My Weblog: umraniye elektrikci uskuadar elektrikci usta elektrikci sisli elektrikci

Page 3

Legitimizing Bodily Integrity Identity Disorder: A Discussion of the Ethical Issues Surrounding Elective Amputations

Juan C. Infante ’19


Body Integrity Identity Disorder (BIID) is a collection of symptoms that manifest themselves in a patient’s desire to amputate otherwise healthy limbs. BIID has traditionally been misunderstood and understudied, but recent findings linking it to specific neural abnormalities warrant a re-evaluation of its definition and available treatment options. Analysis of recent neurological studies suggest that BIID must be recognized as a legitimate disorder with a clear physiological basis. The re-classification of BIID leads to the refutation of classical pro-patient-autonomy arguments and instead calls for paternalistic restrictions on patient autonomy that adhere to the Theory of Future Consent. Recent advances in treatments that safely alter brain function raise crucial questions about physicians’ duty to provide treatment and have far-reaching implications on societal opinions and obligations of and toward patients with BIID.



In September of 1997, Scottish surgeon Robert Smith amputated the leg of Kevin Wright, a man who claimed that he simply “did not want [his leg] because it did not feel [as if it were] a part of [him]” (“My left foot”, 2000, para. 1-4).  Mr. Wright claimed that “amputation above the knee was the only feasible option” and that “he simply could not bear his life with his leg intact” (“My left foot”, 2000, para. 14). Reports detail how Mr. Wright was extremely satisfied with the results of his amputation, and how he believed that “his life had been transformed” (Bayne & Levy, 2005, p. 75). The peculiarity of this case lies in the fact that there was nothing medically wrong with Mr. Wright’s leg; he simply did not feel “complete” with his full set of limbs, and underwent an amputation even though there was no apparent physiological need for it (“My left foot”, 2000, para. 3). In essence, Mr. Wright could not contain the desire to get rid of his healthy leg. His apparent cure after the amputation, however desirable, does not obscure the multiple ethical issues surrounding the elective amputation of healthy limbs. Mr. Wright’s discomfort with one of his healthy limbs is, perhaps surprisingly, not an isolated case, but rather a common symptom of individuals suffering from what has been called Body Integrity Identity Disorder (BIID) (Bayne & Levy, 2005, p. 75).

In 2005, Dr. Tim Bayne and Dr. Neil Levy published an analysis of the possible causes for peoples’ desire to become amputees as well as of the ethical considerations of allowing such procedures. While they acknowledged that multiple explanations for this desire could be correct, including other illnesses such as Body Dysmorphic Disorder, they concluded that BIID was “most plausible,” and defended the patients’ autonomous decision to pursue amputations on the grounds that they allegedly met “reasonable standards of rationality” (2005, p. 75). Nevertheless, as this paper will demonstrate, BIID was not completely understood and was largely disputed at the time of Bayne and Levy’s publication. In fact, a portion of their publication focuses on ways to conceptualize BIID, which reflects the lack of understanding present at the time (2005, pp. 76-77). Clearly, BIID was nothing more than a hypothesis. Recent evidence, however, has confirmed the undeniable role of BIID in patients’ desire to undergo amputations and has explained this condition from a perspective that was not previously available. Although BIID continues to be a rare condition, it now raises crucial questions about patients’ right to bodily autonomy, patients’ rationality, and physicians’ duty to provide treatment.

As Bayne and Levy’s writing suggests, at the time of their publication, BIID was primarily considered a “rare psychological disorder” whose neural basis was poorly understood. (2005, p. 75). However, new evidence demonstrates that there are marked neurological differences in specific areas of the brain in patients who have been diagnosed with BIID. This evidence calls for a decisive reassessment of Bayne and Levy’s conclusions in terms of the modern, and now accurate, characterization of BIID. Given that patients with BIID display peculiar neural functions, that were not understood at the time of Bayne and Levy’s work, elective amputations warrant a reevaluation of patients’ rationality and decision-making capacity. In this paper, I will first argue that the now well-understood neurological differences observed in patients with this disorder –which are also explained in the text- require that BIID is approached with treatments that specifically target its known pathology. Then, I will argue that the now understood pathology of BIID calls for placing temporary, paternalistic limits upon the autonomy of patients with BIID; since physicians have a duty to alleviate the suffering of these patients, these restrictions would allow doctors to delay the immediate gratification associated with elective amputations in favor of treatments that adhere to the Theory of Future Consent. The reassessment of this disorder will have far-reaching implications on societal opinions and obligations of and toward patients with BIID.

Understanding the Neurological Origin of BIID

There is widespread agreement that psychological disorders are instantiated in the brain. Thus, the issue with BIID was never whether it was caused by some kind of neural basis, but rather that the particular neural basis was unknown. To understand the neuropathology of BIID it is necessary to consider recent neurological studies that demonstrate a significant link between differential brain function in specific areas of the brain and the “limb ownership” abnormalities that patients with BIID often present (van Dijk et al., 2013, p.1).  Using functional MRI techniques, researchers at the University of Amsterdam and New York University found that “activity in the ventral premotor cortex depended on the feeling of ownership [of the limb] and was reduced during stimulation of the alienated compared to the owned leg” (van Dijk et al., 2013, p.1). In other words, individuals who displayed symptoms of BIID showed reduced brain function in the ventral premotor cortex.  To fully grasp the relevance of these results, it is necessary to consider further remarks by Arzy, Overney, Landis, and Blanke who note, “Damage to the premotor cortex has been associated with the lack of awareness of a limb in a case report” (as cited in van Dijk et al., 2013, p.1).  In addition, a psychological perception study which had subjects watch a rubber hand being stroked by a paintbrush, while their own hand was hidden from view and also stroked, demonstrated that the “subjects experienced an illusion in which they seemed to feel the touch not of the hidden brush but that of the viewed brush, as if the rubber hand had sensed the touch” (Botvinick & Cohen, 1998, p. 756). Remarkably, Ehrsson and colleagues along with Petkova and colleagues note that “neuroimaging studies of this rubber hand illusion in healthy individuals have implicated the ventral premotor cortex in the feelings of body ownership” (as cited in Dijk, 2013, p.1).

These recent scientific findings demonstrate that the brains of patients who do not identify one of their limbs as their own, and who may desire elective amputations, show significant functional differences relative to the “normal brain,” particularly in the ventral premotor cortex. It is therefore evident that the neural basis of BIID is now much better understood. While this new understanding may at first seem insignificant, we will soon discover that it has important implications for determining treatment options and thus that it is relevant for discussing the ethicality of elective amputations.

Bayne and Levy demonstrate that some experts, such as Bruno, believe that “psychotherapy is the appropriate response to [BIID], not surgery. The patient needs to develop insight into the real source of [their] problems before [they] can solve them” (as cited in Bayne & Levy, 2005, pp. 82-83). Statements of these kind highlight the idea that specific, targeted treatments are not considered valid alternatives when the precise pathology of a disease remains unknown. Establishing the clear neuropathology of BIID thus proves to be of vital importance for two main reasons. First, the diagnosis of BIID and its link to the desire for elective amputations becomes undeniable. As such, there can be no doubt that patients who do not identify their limbs as their own require the necessary medical attention and intervention that may have been previously denied due to disputes regarding the nature and legitimacy of BIID. In the past, it was simply not clear whether BIID was actually responsible for the unmanageable desire for amputation. Second, this addresses societal misconceptions and ensures that BIID is recognized as a “real” disorder, since psychological disorders with unclear pathologies may sometimes be given less validity and therefore may not be treated in the same manner as established physiological disorders. The legitimization of BIID as a physiological disorder also shifts the responsibility to treat it from psychologists to medical doctors, thus emphasizing the idea that BIID requires the same attention and treatment as diseases like cancer and diabetes.

Placing Paternalistic Limits on Patient Autonomy

Now that the neural basis of BIID is better understood and BIID has been established as a disorder with clear physiological causes, one may instinctively rush to the conclusion that any treatment (even elective amputation) that alleviates patients’ symptoms of discomfort is justified. In order to fully assess the permissibility of elective amputations as a treatment, it is necessary to examine the functions of the ventral premotor cortex in more detail.  Researchers at the University of Santiago de Compostela have demonstrated that, in monkeys, “the ventral premotor cortex is involved in the use of recent and long-term sensory memory to decide, execute, and evaluate the outcomes of [a] subject’s choices” (Pardo-Vasquez, Padron, Fernandez-Rey, & Acuña, 2011, p.1).   These “choices,” however, are not choices such as deciding to undergo an amputation, but rather motor responses to perceptual stimuli (Pardo-Vasquez et al., 2011, p.1).  In other words, the premotor cortex “takes part in unique aspects of motor planning and [motor] decision making” (Hoshi & Tanji, 2007, p. 234). This additional function of the premotor cortex is best understood with respect to BIID by considering the case study of an anonymous patient called “John.” In an interview with FOX News, “John” recounts the following childhood story, “I remember two buses going in the same direction, and I was standing by the second bus, and I said to myself ‘if I just stick my leg under the rear wheel of the bus, it will run over it and it will have to get cut off’” (Moran & Pouliot, 2009, para. 5). John’s anecdote demonstrates exactly how the decision-making functions of the premotor cortex may act erratically in a patient with BIID. If patients with BIID have uncharacteristic brain functions in the ventral premotor cortex (as has been demonstrated), and the premotor cortex is involved in motor decision-making, then by extension, patients with BIID must also experience uncharacteristic and untraditional patterns of motor decision-making. Such functional abnormalities in the brain may serve as a reasonable explanation for why BIID patients often report that certain limbs are not part of them, as in the case of Mr. Wright. (“My left foot”, 2000, para. 1). After all, why should they feel at ease with their limbs if they feel the impulse to act out motor decisions with them that they cannot rationalize? Clearly, it is reasonable to argue that the compulsion to remove these healthy limbs stems from the fact that these limbs are used to act out the motor decisions that are directed by faulty brain functions, but that patients simply cannot seem to rationalize or understand.

After taking into account the full range of behavioral functions that the premotor cortex regulates, it becomes increasingly difficult to accept Bayne and Levy’s autonomy argument. Bayne and Levi justify their support of patient autonomy by asserting that patients with BIID are rational and can therefore provide fully informed consent for elective amputation procedures (2005, pp. 75, 80). However, I argue that this is far from the case. To understand the ethical implications of providing a less than rational form of consent, it is necessary to consider the ethical principle of Paternalism. According to Dr. Onora O’Neill, a renowned philosopher at Essex University, “patients’ reduced capacities…demand paternalistic treatment” (1984, p. 173). In her discourse on the limited scope of human autonomy, she provides a “framework for working out boundaries of permissible medical paternalism” (1984, p. 177). One of her stipulations involves situations in which patients have a “temporarily impaired capacity for autonomy” (1984, pp. 177-178).  In essence, Dr. O’Neill argues that a patient may be prevented from making medical decisions, such as petitioning the amputation of a healthy limb, in cases where the patient may not meet “reasonable standards of rationality” (Bayne and Levy, 2005, p. 75).

Having qualified BIID as a disorder with a clear pathology that can be explained by irregular brain function in the premotor cortex, it becomes evident that patients’ capacity to make decisions is highly compromised by their neurological irregularity. A patient with BIID could never be thought of as a fully autonomous and rational subject, because their autonomous decision to desire an amputation is ill-informed by their frustration with their seemingly incomprehensible patterns of motor decision-making. As such, instances in which patients desire to undergo elective amputations prove to be instances in which ethics requires placing limits on autonomy, since patients have a neurologically-determined “impaired capacity for autonomy” as outlined by O’Neill (1984, pp. 177-178).

Some may argue that this line of paternalistic reasoning suggests that any neural impairment would lead to an impairment of autonomy. They may argue that this leads to the conclusion that, for instance, an individual with a damaged auditory cortex (who is thus deaf) would have impaired decision-making capacity. I propose two counter-arguments to dismiss such concerns. First, such a line of reasoning may justly be qualified as a fallacious reductio ad absurdum; clearly the symptoms of a condition must be analyzed within the context of each disease, to determine whether the disease warrants placing limits on autonomy. Second, the symptoms of BIID lead patients to desire treatments that would physically disable them. Easing patients’ discomfort by allowing them to undergo an elective amputation becomes increasingly problematic when one considers potential treatments and the Theory of Future Consent.

Physicians’ Duty and Potential Treatments: The Theory of Future Consent

Now that BIID has been established as a disorder with a clear neurological basis and the necessity of imposing limits on the autonomy of patients with BIID has been demonstrated, the question becomes whether physicians can reduce the suffering of patients with BIID without the need to perform these elective amputations. It is crucial to note that one of the conditions for O’Neill’s framework is that the “temporary loss of autonomy [must] offer grounds for paternalistic intervention to restore autonomy” (1984, p. 178). In other words, paternalism is mostly justified when there is a belief that the patient may one day be able to provide autonomous informed consent. What O’Neill suggests by this is that a complete analysis of this idea requires an understanding of the Theory of Future Consent. Tis theory expresses the notion that the application of paternalism, which imposes limits on a person’s autonomy when a person is thought not to be fully rational, is justified by the idea that if the person whose autonomy is being limited could somehow regain their autonomy, then they would have agreed with placing restrictions on their own autonomy in the first place (Kairalla, 2007, pp. 31-32). In other words, placing restrictions on a person’s autonomy, when that person is not fully rational, is justified if that person would agree with the restrictions once they become rational again. This idea leads to the following question: Are there alternative forms of treatment that can help alleviate the suffering of patients with BIID while also changing their brain function in a way that allows them to regain their full autonomy?

When considering potential treatments for patients with BIID, Dr. Christopher Ryan, a professor of psychiatry at The University of Sydney, recognizes the limitations that physicians face. Dr. Ryan observes, “The two approaches often held out as alternatives [to amputation] are psychotherapy and pharmacotherapy. Unfortunately, at this point, neither of these has good evidence of efficacy” (2009, p.26). However, experiments similar to the rubber-hand study, which was discussed earlier, can provide valuable insight into potential treatment options that would be much more desirable than amputation. Giraux and Sirigu conducted a study on patients experiencing phantom limbs that analyzed “the effects of visuomotor training on motor cortex activity” (2003, p.107). When subjects “matched voluntary movements of the phantom limb with prerecorded movements of a virtual hand…subjects showed increased activity in the contralateral primary motor area” (2003, p.107). These findings have led Giraux and Sirigu to hypothesize that “artificial visual feedback on the movements of the phantom limb may thus fool the brain and reestablish the original hand/arm cortical representation” (2003, p.107). Although phantom limb syndrome is not nearly the same as BIID, clearly there are certain methods through which uncharacteristic patterns of brain function may be corrected or at the very least attenuated.

The positive result of treating patients with phantom limbs provides hope for finding a form of treatment for BIID. Caloric vestibular stimulation involves placing water of different temperatures into the ear canal in order to test the functionality of acoustic nerves (“Caloric Stimulation”, n.d.).  As Lenggenhager and colleagues note, “Ramachandran and McGeoch suggested that caloric vestibular stimulation might alleviate the desire for amputation in individuals suffering from BIID” (as cited in Lenggenhager, Hilti, Palla, Macauda, & Brugger, 2014, pp. 1-2). However, researchers studying the effects of caloric vestibular stimulation on the desire to undergo an elective amputation found that there was “no significant impact of the [stimulation] on body ownership” (Lenggenhager et al., 2014, p. 2). Nevertheless, findings by Braam et al. indicate that a combination of Selective Serotonin Reputake Inhibitors (SSRIs) and antidepressants “seemed to lessen the distress [but] not the desire [for an amputation]” (as cited in Ryan, 2008, p.26). Since the desire for amputation remains, abnormal patterns of brain function are not modified by pharmacotherapy and thus patient autonomy is not restored.

As such, the literature suggests that there are currently no permanent alternative forms of treatment that alleviate the suffering of patients with BIID while also restoring their rationality and autonomy. Does this finding mean that the paternalistic limit on patient autonomy is not justified and that physicians should proceed with elective amputations? Given the promising advances that have been made in treating patients with phantom limbs, the answer continues to be a resounding no. Lenggenhager and colleagues point out the fact that caloric vestibular stimulation “alters activity in core regions of the vestibular cortex” (2014, p. 3). However, as noted earlier in this paper and acknowledged by Leggenhager and colleagues, “fMRI data in individuals with [BIID] suggest alterations in sensory mechanisms in the pre-motor cortex rather than in posterior areas” (van Dijk et al. as cited in Leggenhager et al., 2014, pp.2-3).  It is entirely possible that the failure of caloric vestibular stimulation lies in the fact that it did not directly affect the activity of the premotor cortex, the area of the brain thought to be responsible for BIID. If researchers are able to develop a mechanism for stimulating the premotor cortex, there is no reason to believe that the irregular brain activity patterns in patients with BIID will not be normalized. Developing such a mechanism should not be inconceivably difficult given that deep brain stimulation systems are already FDA-approved for Parkinson’s treatment (Brooks, 2017).

Therefore, I argue that the application of paternalism in order to prevent patients from exercising their autonomous right to undergo an elective amputation is justified by the Theory of Future Consent. It is reasonable to argue that, in the very near future, brain stimulation techniques will be able to address the functional brain deficiencies observed in patients with BIID. Once the uncharacteristic patterns of brain activity in these patients are restored, the patients will likely display a radical improvement with regard to their feelings of limb ownership and would utterly reject the idea of voluntary mutilation; the patients’ autonomy would have therefore been restored and they would likely agree with the limits that were placed on their autonomy. In the meantime, through the administration of SSRIs and antidepressants, the suffering of patients with BIID can be minimized while also preventing them from undergoing elective amputations that they would likely consider harmful if they were to regain a fully autonomous status. It is evident that placing limits on the autonomy of patients with BIID protects their bodily integrity, prevents them from making ill-informed medical decisions, and allows physicians to alleviate the suffering of their patients (despite the fact that they must delay their patients’ cure.)


While Bayne and Levy’s argument for permitting elective amputations was more plausible when the neural basis of BIID was poorly understood, recent findings have strongly linked BIID with functional abnormalities in the premotor cortex. The legitimization of BIID as a condition with a biological cause dispels common misconceptions about the validity of psychological illnesses and establishes an imperative for promptly finding an effective cure for BIID. Nevertheless, since patients desiring elective amputations suffer from impaired brain function that leads them to desire treatments that would physically disable them, certain limits must be placed on their autonomy and, as such, not all forms of treatment for BIID are ethically acceptable. As new forms of brain stimulation that could reestablish patients’ feelings of limb ownership loom on the horizon, and pharmacotherapy is available to temporarily alleviate suffering, the imposition of paternalistic limits on patient autonomy proves to be the soundest ethical choice. Paternalism is further supported by the belief that, once these patients regain their autonomy, they would agree that preventing them from undergoing elective amputations was the most morally responsible action. This leads one to further consider the extent to which BIID will be recognized by society as the devastating illness that it is, and how current healthcare delivery models and insurance policies will adjust to encompass this previously overlooked and misunderstood condition.






Arzy, S., Overney, L. S., Landis, T., & Blanke, O. (2006). Neural mechanisms of embodiment: Asomatognosia due to premotor cortex damage. Archives of neurology, 63(7), 1022-1025.

Bayne, T., & Levy, N. (2005). Amputees By Choice: Body Integrity Identity Disorder and the Ethics of Amputation. Journal of Applied Philosophy, 22(1), 75-86.

Botvinick, M., & Cohen, J. (1998). Rubber hands ‘feel’ touch that eyes see. Nature, 391(6669), 756.

Braam, A. W., Visser, S., Cath, D. C., & Hoogendijk, W. J. G. (2006). Investigation of the syndrome of apotemnophilia and course of a cognitive-behavioural therapy. Psychopathology, 39(1), 32-37.

Brooks, M. (2017). FDA Clears Deep-Brain Stimulation System for Parkinson’s. Retrieved March 28, 2018, from

Bruno, R. L. (1997). Devotees, pretenders and wannabes: two cases of factitious disability disorder. Sexuality and Disability, 15(4), 243-260.

Caloric Stimulation. (n.d.). Retrieved May 9, 2016, from

Ehrsson, H. H., Spence, C., & Passingham, R. E. (2004). That’s my hand! Activity in premotor cortex reflects feeling of ownership of a limb. Science, 305(5685), 875-877.

Giraux, P., & Sirigu, A. (2003). Illusory movements of the paralyzed limb restore motor cortex activity. Neuroimage, 20, 107-111.

Hoshi, E., & Tanji, J. (2007). Distinctions between dorsal and ventral premotor areas: anatomical connectivity and functional properties. Current opinion in neurobiology, 17(2), 234-242.

Kairalla, R. (2007). The Ethics Bowl: Adventures in Reasoning. Miami, FL: University of Miami Ethics Society.

Lenggenhager, B., Hilti, L., Palla, A., Macauda, G., & Brugger, P. (2014). Vestibular stimulation does not diminish the desire for amputation. Cortex, 54, 210-212.

Moran, J. M., & Pouliot, K. (2009). Determined to Amputate: One Man’s Struggle with Body Integrity Identity Disorder. FOX News. Retrieved April 23, 2016, from

‘My left foot was not part of me.’ (2000). The Guardian. Retrieved April 23, 2016, from

O’Neill, O. (1984). Paternalism and partial autonomy. Journal of Medical Ethics, 10(4), 173-178.

Pardo-Vázquez, J.L., Padrón I., Fernández-Rey J., y Acuña C. (2011). Decision-making in the ventral premotor cortex harbinger of action. Frontiers in Integrative Neuroscience, 5, 54, 1-14.

Petkova, V. I., Björnsdotter, M., Gentile, G., Jonsson, T., Li, T. Q., & Ehrsson, H. H. (2011). From part-to whole-body ownership in the multisensory brain. Current Biology, 21(13), 1118-1122.

Ramachandran, V. S., & McGeoch, P. (2007). Can vestibular caloric stimulation be used to treat apotemnophilia?. Medical hypotheses, 69(2), 250-252.

Ryan, C. J. (2009). Out on a limb: the ethical management of body integrity identity disorder. Neuroethics, 2(1), 21-33.

Van Dijk, M. T., Van Wingen, G. A., Van Lammeren, A., Blom, R. M., De Kwaasteniet, B. P., Scholte, S. H., & Denys, D. (2013). Neural basis of limb ownership in individuals with body integrity identity disorder. PLOS One, 8(8), 1-6. doi:10.1371

From a Whisper to a Shout: Whisper Networks and #MeToo as Forms of Resistance

Amelia Goldberg ’19


Contrary to romantic notions about the sudden explosion of women’s anger after decades of sexual harassment, the #MeToo movement is strongly grounded in a history of resistance. This paper traces the roots of the #MeToo movement in whisper networks that women use to pass along information about which men are harassers and should be avoided. Drawing on a variety of published accounts as well as my own experience as an advocate and a member of a whisper network, I analyze the characteristics of whisper networks. I argue that these whisper networks constitute a moral community of women opposed to sexual harassment. Furthermore, whisper networks enable women’s covert resistance by giving meaning to their small, isolated acts of defiance. This argument provides support for the relevance of peasant resistance studies to understanding contemporary subaltern groups. At the same time, whisper networks mark the tacit acceptance of sexual harassment as a culturally intimate aspect of most American institutions. While this intimacy gains public expression in modern populist politics, the #MeToo movement realizes a countertendency by publicizing and attempting to institutionalize whisper networks.


From Whisper Networks to #MeToo[1]

A recent string of high-profile sexual harassment allegations, dubbed the #MeToo movement, is bringing much-needed attention to the far-reaching nature and implications of sexual harassment. As has become clear, sexual harassment is a widespread abuse of power tacitly condoned by men within most American institutions. Yet beyond bringing this to light, the #MeToo movement also represents the beginning of a new political contest over gender hierarchies in the workplace. Without a doubt, #MeToo has the potential to transform some aspects of American society. Nevertheless, decontextualizing this movement from a long history of women’s resistance to institutionally tolerated sexual harassment risks romanticizing resistance as the eventual explosion of a human tendency for justice. In reality, the #MeToo movement has its roots in whisper networks that women use to pass along information about which men are harassers and should be avoided. In this paper, I argue that these whisper networks constitute a moral community of women opposed to sexual harassment, which gives meaning to individual practices of everyday resistance, and generates the moral outrage expressed in #MeToo. At the same time, whisper networks mark the tacit acceptance of sexual harassment as a culturally intimate aspect of most American institutions. As this intimacy gains public expression in modern populist politics, the #MeToo movement realizes a countertendency by publicizing and attempting to institutionalize the heretofore covert resistance of whisper networks.

Whisper networks are the chains of communication between women in institutions that spread warnings about who has a history of sexual harassment (Creswell and Hsu 2017; McKinney 2017; Meza 2017; Tolentino 2017).[2] These networks have long existed, but became the subject of public scrutiny when it became clear that most women professionally connected to the men implicated in the recent string of sexual harassment accusations knew of their behaviors. In the highly publicized case of film magnate Harvey Weinstein, Hollywood women warned each other about his style of operating and shared advice, such as to “dress frumpy” when meeting with him (Farrow 2017). In another case, women in small-town Alabama knew that Judge Roy Moore, who is alleged to have assaulted a number of young girls, was a frequent visitor to a local mall, and advised young girls to stay away (Meza 2017). In these and many other instances, the #MeToo movement brought to light long-standing whisper networks (Creswell and Hsu 2017). Importantly, women form whisper networks within workplaces rather than around individual abusers. Institutions and industries where women and men are professionally affiliated, including but not limited to universities, large companies, Hollywood, New York journalism, Silicon Valley, and Wall Street all contain their own whisper networks. For this paper, I will examine the whisper networks in Hollywood and in New York media by drawing on the many firsthand and journalistic accounts of both shared in the months following the Weinstein allegations. I will also draw on my personal experience as an anti-sexual violence advocate and a member of the whisper network at Harvard, where I am an undergraduate.

While whisper networks sometimes contain widely shared open secrets, they are exclusive to women and, to a limited degree, gay men. In the conversations that I have had in preparing this paper, most women that I spoke with were not familiar with the term “whisper network,” but immediately understood what I meant when I described the practice of asking for and passing along information about sexual harassers. Most men, however, were not aware of this practice. For instance, when a male friend overheard myself and two other friends discussing his buddy’s sexist behavior, he was appalled to hear me passing along such damning criticisms. I struggled to explain to him that women share such information in order to protect themselves, because it is often hard to tell which “nice guys” are actually harassers or misogynists, and not just to attack unfortunate men. My friend’s confusion demonstrated that whisper networks are exclusive to women, who are particularly vulnerable to sexual harassment. Gay men are sometimes included, but often not. As Jesse Dorris has written in response to allegations of sexual harassment within the gay community, “gay men need a whisper network, too” because existing networks do not serve them (Dorris 2017). Some whisper networks also exclude on the base of race or class, and few accommodate the particular sexual harassment experienced by LGBT people. The impact of this exclusion is magnified because those excluded from whisper networks are in many cases already more vulnerable to sexual harassment – for instance, LGBT people, people of color, and people new to an industry (Shafrir 2017). The discriminatory nature of whisper networks is likely the most significant limitation of their usefulness, as many voices in the #MeToo movement have emphasized. However, whisper networks generally extend to all women within a given institution. Critically, this includes women who have not personally experienced harassment but still receive and pass along information.

In contrast to whisper networks that are local to a particular institution, the #MeToo movement is publicly exposing, on an unprecedented scale, high-profile sexual harassers. The concept of “me too” originated with activist Tarana Burke in 2006, who at the time was working with black and brown girls and wanted to validate their experiences of sexual assault. It therefore began as an inward-facing project of “survivors helping survivors” (Lopez and Snyder 2017). More recently, the hashtag #MeToo went viral on social media in the wake of revelations that Harvey Weinstein had for decades used his power and influence to assault young actresses (Dibdin 2017). Actress Alyssa Milano suggested on Twitter that “If all the women who have been sexually harassed or assaulted wrote ‘Me too’ as a status, we might give people a sense of the magnitude of the problem” (Fance 2017). This viral hashtag shifted #MeToo from a focus on support and healing to and outward-facing focus on revealing the extent of sexual assault for a broader public, including men. However, while most social media posts of #MeToo left the perpetrators anonymous, they triggered a third phase of #MeToo, which I make the main focus of this paper: a series of public accusations against men in powerful positions that they have used their power to commit sexual assault or harassment. Most of these accusations have gained considerable press coverage, and have been corroborated by photographic evidence or by the agreement of multiple accusers. As of this writing, over 200 public figures have been accused (North 2017). While public accusations of sexual assault and harassment have a long history, the #MeToo movement marks the first time that so many accusations have occurred in such a short span of time, and the first time that such accusations could constitute a movement. By comparison, just last year, a series of assault accusation against Bill Cosby resulted in a hung jury, but nothing like the outpouring of the #MeToo movement (Zeitchik 2016). As proponent Kelsey McKinney wrote, “the dam broke” (McKinney 2017). While its impact remains indeterminate, the #MeToo movement signifies an important change in how sexual assault claims are received.

As survivors of sexual harassment, anti-sexual violence activists, and defenders of the accused have argued, the #MeToo movement bears certain imperfections. For instance, many survivors have expressed frustration about the pressure that the movement exerts on them to publicly share their intimate and often traumatic experiences. Additionally, the #MeToo movement does not effectively extend to the many people, typically working-class women, whose harassers are not high-profile enough to generate media interest but still use their positions of power to abuse female workers (Alianza Nacional de Campesinas 2017). Another concern is that the media exposition of alleged harassers often forces them out of positions without a fair trial or a chance to defend themselves against the charges. Anti-sexual violence advocates argue that the system of trial by press is not a solution, but merely reflects that the systems intended to provide justice, from human resources departments within institutions to the justice system within the United States, make it nearly impossible for survivors to successfully make claims against perpetrators. Of course, two wrongs do not make a right, and there is a small possibility that false accusers would exploit #MeToo movement for their own ends.[3] However, the focus of this paper will not be on defending the #MeToo movement or re-litigating the accusations against its growing list of abusers. Instead, I explain the social foundations for the movement, and argue that these account for many of the tensions between it and existing judicial structures.


Whispering Together: Gossip as a Form of Resistance

Whisper networks are a form of gossip, but they are not mere gossip. Rather, as anthropologists Gluckman and F.G. Bailey argue, gossip is often critical to establishing and maintaining a moral community. Developing his study of European peasant societies in Gifts and Poison: The Politics of Reputation, Bailey writes that a moral community is a “sphere of action in which moral claims are made” (Bailey 1971, 191). In other words, moral communities are composed of people who are prepared to issue moral judgments about each other. In doing so they draw on a set of shared values, although the exact application of those values is a site of social contest. Thus, the moral community is both contested and maintained through the exchange of moral judgment – gossip. Gossip regulates membership within a moral community by establishing each member’s reputation relative to the values of the culture within which they are judged. As a result, gossiping reifies those values (Bailey 1971, 7). Gossip itself is a mode of communication for the moral community. Gluckman’s description of the Makah, a small American Indian tribe, illustrates this point nicely: “values of the group are clearly asserted in gossip and scandal…[gossip can] control disputation by allowing each individual or clique to fight fellow members of the larger group with an acceptable, socially instituted customary weapon” (Gluckman 1963, 313). Gossip is a particular language for members of a group to talk about how well each other member performs as a member of that group and therefore, indirectly, to talk about the group itself. Thus, when gossip says that Max is not a true Makah, at the same it asserts that the gossiper is a true Makah, and knows what it means to be Makah. Gossiping about shared information also reminds any outsider listening that they do not belong. In this way, gossip highlights the exclusion of those who and are not members of the group by drawing on the shared knowledge and history of the moral community (Gluckman 1963, 313). Moreover, as Bailey argues, gossip shapes the reputation of the gossiper as well as the object of their gossip. To improve their reputation, the gossiper must convey true information with the correct moral valence to their audience (Bailey 1971, 7). If the gossiper is concerned that the story will turn out to be untrue, or is concerned about the reception of their moral interpretation of the story, they often sign it with a phrase like “I have just heard from Frederick that…” (Bailey 1971, 7). This signature passes along the responsibility to another person and limits the ability of any individual to deploy gossip for character assassination.

Constituted by gossip, a whisper network upholds the values and reputations of the moral community of women within an institution. Despite being deeply embedded in a broader institution, such a moral community maintains its distinctness through particular language, codes, shared histories and values, and reputations. In a particularly extreme example, many women in journalism communicated which senior editors were harassers through a code: he is either “on the Island” or not (Friedman 2012). More common idioms include “watch his hands” and “make sure you’re sober if you hang out with him”. Beyond this particular language, whisper networks also carry a memory of harassment that can help communicate new instances. For instance, if the whisper network has parsed the stories about Steve, a serial harasser, a member might say that “the new professor is a total Steve” to warn those in the know. Most importantly, women use whisper networks to transmit judgment about harassers based on the moral valence of certain acts. Such judgments maintain a set of shared (and contested) values about what constitutes harassment and how wrong it is. Finally, consistent with Bailey’s observations, gossip shapes the reputation of each member of a whisper network. Jia Tolentino, herself a member of multiple whisper networks over the course of her career, writes that “women ask for and examine sourcing” and differentiate between firsthand and second- or third-hand claims (Tolentino 2017). These source-examinations reflect the tags that gossipers use to protect their reputations while passing along valuable information. As Tolentino writes, “if I give you false information, then my credibility and relationships will suffer,” because the group of people engaged in the whisper network typically “ask around, monitor social situations, [and] shut down the rare false rumor” (Tolentino 2017). Through such checks, whisper networks protect against their abuse by women seeking to tarnish a man’s reputation. In all these ways, whisper networks help constitute and maintain the moral community of women within an institution with its own history, reputations, and common values.

Beyond maintaining this moral community, whisper networks mediate between women in an institution and the men who have power over them. Gluckman and Bailey, by focusing on gossip within relatively isolated communities, do not examine the importance of gossip about non-members. Gluckman’s teleological approach assumes that gossip’s function must accord with the community in question, disregarding outsiders and often excluding the anthropologist as well. Meanwhile, Bailey argues that the reputations of those outside the community are not judged by the same moral standards, and instead are “destroyed or employed in whatever fashion serves our interests” (Bailey 1971, 7). However, a moral community often cannot afford to simply treat outsiders in this instrumental fashion if it is, like a whisper network, deeply embedded in another community. In particular, for Bailey and Gluckman, gossip is fundamentally public, because it distributes the shared knowledge of an all-encompassing moral community. Yet even Bailey’s own ethnography does not support this position. In a telling example, he describes the small village of 400 called Valloire, in the French Alps, where it is perfectly acceptable for men to sit in public and gossip, but women go to great length to avoid talking to each other in the open and appearing to gossip (Bailey 1971, 7). This is because men’s gossip is assumed to be bavarder, or idle chatter and news-passing; while women’s gossip is socially condemned and assumed to be mauvaise langue, or scandal, the passing on of defamatory information (Bailey 1971, 7). Bailey fails to explain why women are considered especially prone to mauvaise langue, and therefore must resort to covert gossip. It is not clear why women’s gossip was considered so dangerous in Vallorie, but in the case of whisper networks, it is obvious: women use gossip to share their frustration with the men who have power over them. Indeed, the particular role of gossip for a subordinated group is explored in James Scott’s theory of peasant resistance.

In Weapons of the Weak, Scott argues that gossip is one of the tools of everyday resistance that are widely employed in peasant societies. Scott, although a political scientist by training, bases his theory on fieldwork in Sedeka, a village in rural Malaysia with stark inequality between a class of peasants and one of landlords. Although the peasants live in a situation of near-total subjugation, Scott argues, they nevertheless engage in consistent, everyday resistance. Some common acts of everyday resistance in Sedeka include pilfering grain, informal boycotts, and squatting on public land, all of which contest power structures without endorsing “public and symbolic goals” (Scott 1985, 23). Unlike peasant rebellions, this resistance is covert, but it is just as significant in shaping society and how people live in it (Scott 1985, 23). Also unlike rebellions, everyday resistance requires barely any coordination, uses implicit understandings shared within informal networks, looks like individual self-help, and avoids direct confrontation with authority, especially on a symbolic level (Scott 1985, 23). In fact, Scott writes, “the success of de facto resistance is often directly proportional to the symbolic conformity with which it is masked” (Scott 1985, 23). Given the poverty of the peasants and their dependence on the landlords, open insubordination would have severe consequences, but covert resistance is more likely to allow its protagonists to survive. Therefore, such resistance conforms to formal hierarchies and symbolic power (Scott 1985, 23). At the same time, however, everyday resistance relies on gossip to coordinate and justify acts against power.

Gossip is a means for the dominated group to contest symbolic hierarchies of power and maintain a distinct set of shared values by which they judge themselves and others. In Sedeka, the values of the peasantry include the expectation that wealthy landlords should be generous in providing employment and assistance to the peasants of the village, and justify resistance against those who do not (Scott 1985, 23). As a result, while everyday resistance is not institutionalized, it is not fully uncoordinated (Scott 1985, 23). Individual acts are linked by peasant subcultures that contest the symbolic power of the landowners, contained in folk traditions, tales, and critically, gossip and jokes about the landowners (Scott 1985, 23). Such stories provide practical survival tips, but they also carry an implicit rejection of the status quo that makes such adaptations necessary (Scott 1985, 23). Scott terms these “transcripts” of resistance, but “scripts” might be a better term to describe this phenomenon: there are certain accepted and popular ways of speaking about the landlords and the peasant situation that contain an implicit criticism of the symbolic power structure itself. For instance, the wealthiest landlords have denigrating nicknames, Haji Broom and Kadir Ceti, that peasants use widely behind their backs to refer to their ungenerous and unacceptable behaviors (Scott 1985, 23). Unlike the landlords, whose full transcripts are generally public, “the exercise of power nearly always drives a portion of the full transcript underground” (Scott 1985, 23). Nevertheless, the transcript coordinates individual actions. Gossip justifies and encourages resistance against those landlords who fall short of the moral demands of the peasantry, transforming individual and self-serving acts into a pattern of resistance (Scott 1985, 23). For instance, everyone might pilfer grain individually, and if caught use the appropriate term of address to indicate their submission to the village hierarchy of wealth and power. However, the wealthier and less generous the landlord, the more they discover that their grain is pilfered. By contesting these power structures, peasants also demonstrate their understanding of the symbolic hierarchies of power they inhabit (Scott 1985, 23). Therefore, everyday resistance occupies a middle ground between overt resistance and passive submission, where peasants do not actively pursue a radically different society but do actively imagine alternate social forms. By establishing the shared values of a moral community and constantly applying them to particular cases, gossip is critical in constructing this alternative.

Similar to peasant folk traditions, whisper networks coordinate and justify women’s everyday resistance within patriarchal institutions. Like the peasants Scott describes, women in patriarchal institutions are embedded in systems of power. In fact, Scott himself notes the parallel between his theory of resistance and the feminist literature on the myth of male dominance in peasant society, which argues that women can exert power in male-dominated societies only so far as they do not openly challenge the “formal myth of male dominance” (Scott 1985, 23). Instead, by paying lip service to gender hierarchies, just as Scott’s peasants do to the hierarchies of income and power, women are able to conceal their actual resistance. This structural analogy extends to women in whisper networks. Like the peasant resistance in Sedeka, the immediate aim of using a whisper network is survival, not revolution. As McKinney describes, “the network carries the worst nights of people’s lives…but it also carries warnings” (McKinney 2017). Such warnings include strategies to stymie sexual harassment, such as not showing up to meetings alone, inviting someone else to a lunch, and never staying late or getting drinks (Petersen 2017). While the continuous presence of whisper networks acknowledges harassment as a fact of life, it does not therefore accept such treatment as just or necessary. To the contrary, the role of whisper networks in resisting patriarchal power is particularly evident in gossip that transitions easily from discussions of harassers, to strategies of avoidance, to complaining about how difficult it is to do anything about men in such positions of consolidated power. Thus, as with the peasants Scott studied, whisper networks are made up of people who know the causes of their own oppression in that they do not merely seek protection against individual assailants, but also recognize the inadequacies of bureaucratic and institutional structures that are supposedly meant to protect them from harassment. In voicing a moral opposition to these structures and the behaviors they permit, whisper networks contest sexual harassment on a symbolic level and thereby coordinate actual resistance to its occurrence.

However, like the peasant resistance, the strategies of the whisper network are masked by symbolic conformity to patriarchal power. Such conformity is necessary because the barriers to overt resistance are quite significant. To be sure, many women are currently making claims against figures such as Weinstein without facing terrible opposition. However, this is a very new phenomenon. Previous to this year, most women believed that “normal routes of protection — HR complaints, direct confrontation, the police — simply won’t work…and the price of becoming an accuser is so steep” (Petersen 2017). As I will argue, the price of accusing a powerful man of sexual harassment results from institutional cultures that protect and condone such acts. Whisper networks often were the only mechanisms women had to protect themselves and to oppose the acts they considered immoral. Yet because it is covert, a whisper network does not significantly destabilize structures of power. In fact, many of the activists in the #MeToo movement have criticized whisper networks for placing the burden on women to protect themselves from harassment, rather than on men in power to stop harassing women in their employ. Some activists allege that such networks accept “the status quo, in which women work around abusers rather than forcing them out of our workplaces” (Press 2017). For instance, a recent publication on “indecent advances” in academia includes a set of “prevention tips” including to “find out if there are rumors of sexual harassers in your field,” and, in the case that a researcher does experience harassment, to “confide in a trusted colleague or friend and discuss the pros and cons of filing a report” (Gewin 2015). In other words, this report recommends that young researchers turn to the whisper network before the institutional structures supposedly meant to protect them. However, by prioritizing the safety of individual women, the network leaves a culture of sexual harassment unchallenged. Therefore, like peasant resistance, it does not directly challenge gender hierarchies.

Nonetheless, as Matthew Gutmann argues in Rituals of Resistance, covert and overt resistance are not mutually exclusive responses to oppression. Gutmann claims that Scott’s description of peasant resistance is unsatisfying because the peasants ultimately accept society as it is, and that Scott’s argument comes at the cost of acknowledging the importance of overt resistance. For instance, Gutmann attributes to Scott the claim that peasants “[accept] society as it is – its inevitability if not its justice” (Gutmann 1993, 75). This is an unfair criticism, because Scott does state that “the notion of inevitability itself can be, and is, negated by the historical practice of subordinate classes,” not to mention the ways that resistance challenge the justice of the status quo (Scott 1985, 23). The peasants Scott describes, indeed, might happily participate in a revolution were the possibilities more amiable. Rather, Scott’s argument rests on the empirical claim that in many cases, everyday resistance “has been the only option” (Scott 1985, 23). Yet while Gutmann’s criticism of Scott is ultimately off-target, his more important and helpful claim is that covert and overt resistance are two responses to conditions of oppression and may occur simultaneously: “it is not a question of overt of covert in isolation; rather…these forms occur together, alternate, and transform themselves into each other” (Gutmann 1993, 75). Gutmann does not offer a clear framework for understanding how covert and overt resistance are transformed one into the other, but the emergence of the #MeToo movement is an excellent example of how covert resistance can generate overt opposition.

Whisper networks laid the groundwork for resistance by maintaining a moral community with a set of values in opposition to sexual harassment, but the #MeToo movement is transforming this moral orientation into overt resistance. As one journalist and #MeToo activist wrote, “this is the year the whisper network went viral” (Meza 2017). With the #MeToo movement, information that had long circulated in whisper networks is becoming public knowledge shared far beyond small moral communities. Concurrently, the moral judgments and values implicit in whisper networks are also being explicitly voiced. Complaints about how easily men in power can abuse women have generated public allegations and calls for those men to relinquish their power. The tensions between the movement and the status quo reflect the revolutionary tendencies of the whisper networks themselves. For instance, in order to justify their inattention to sexual harassment, institutions must assert that sexual harassment is highly uncommon. In direct contradiction, however, the very underlying principle of the whisper network is that sexual harassment so frequent, it is a fact of life that drives women to formulate specific coping strategies. #MeToo has sought to expose that by revealing publicly how many people within these industries have experienced sexual harassment and how many people in positions of power are harassers.

The shift from a local whisper network has also raised new challenges, as #MeToo activists seek to realize their values in new public institutions. Significantly, claims within the whisper network are typically assessed and passed along based on reputational politics within a local community. No evidence is required apart from a narrative that is believable given the purported assailant’s character. The aim of passing along such information, moreover, is self-protective rather than a direct attempt at character assassination. These ends are maintained in the #MeToo movement, but come into significant tension with a media establishment that airs claims on a broad scale, and a justice system that typically operates on the basis of “innocent until proven guilty”. To some extent, the typical function of a whisper network is replicated when whole groups of women (and in some cases, men) come forward to affirm each other’s accusations. However, most of the perpetrators named have been forced out or stepped down without due process, raising red flags in some quarters about the possibility of false accusations and the miscarriage of justice (Yoffe 2017). A particularly telling example was one woman’s attempt to institutionalize and democratize the whisper network in the New York media industry by sharing a spreadsheet online that anyone could anonymously edit to add the names and deeds of “shitty media men”. The document was removed within a few days, but not before accumulating 74 names (Chapin 2017). Even #MeToo advocates, such as journalist Alex Press, have expressed concerns about the possibility of false claims gaining steam in such a format: “while false reporting is far from common, the ability to input an allegation anonymously and online runs the risk of declaring men guilty without verification” (Press 2017). Press invokes the possible reputational costs of gossip when it occurs in a typical whisper network, which prevents people from making false accusations, and keeps unbelievable ones from further propagating. In the spreadsheet, unlike in the whisper network, a gossiper did not have to stake their reputation on the received validity of their complaint. As a result, most journalists writing about the list noted one or two claims which they were certain would not be substantiated (Tolentino 2017). These tensions reflect the differences between covert and overt resistance. They also demonstrate the schism between the primary moral orientation of whisper networks – that sexual harassment is wrong – and the powerful institutions that do not fully accept this value. Thus, #MeToo activists draw on whisper networks to rethink the polity, even if they have not yet presented an explicit and coherent alternative.

In fact, while many institutional structures overtly agree that sexual harassment is wrong, whisper networks at work reveal this to be a mask. As Lila Abu-Lughod argues, everyday resistance can serve as a diagnostic for often covert or complex functions of power. Scott’s work, she argues, manifests a widely-shared tendency to romanticize resistance, and to “read all resistance as signs of the ineffectiveness of systems of power and of the resilience of the human spirit in is refusal to be dominated” (Abu‐Lughod 1990, 42). In other words, we tend to look at resistance as evidence that apparently hegemonic power structures can never truly dominate us, and that humans have a certain indomitable agency even in deeply oppressive conditions. However, this romantic notion inaccurately disconnects resistance from power, when in fact it typically operates from an opposing field of power. Therefore, she argues, “we should use resistance as a diagnostic of power” (Abu‐Lughod 1990, 42). Among the Bedouin, for instance, Abu-Lughod documents that women’s resistance to patriarchal power takes many of the forms Scott describes: they tell sexually irreverent jokes, songs, and folktales; they keep secrets for each other; and they tell elaborate stories of those who have resisted pressure to marry. In other words, they gossip. In each of these instances, however, women’s resistance to patriarchal power is actually grounded in those forms native to the society itself. For example, the folktales and songs express Bedouin values of resistance to power; and the strict segregation of women’s and men’s spaces make it easy for women to gossip and keep secrets together. Therefore, Abu-Lughod writes, “women take advantage of [the] contradictions in their society to assert themselves and to resist…through locally given traditional forms, a fact which suggests that in some sense at least, these forms have been produced by power relations and cannot be seen as independent of them” (Abu‐Lughod 1990, 42). On some level, then, women’s resistance is also a product of gender hierarchy itself. It both opposes dominating power structures and uses them for its own ends. Abu-Lughod takes her argument too far when she argues that this means that young Bedouin “unwittingly [enmesh] themselves in an extraordinarily complex set of new power relations” (Abu‐Lughod 1990, 42). She does not present compelling evidence for this false-consciousness argument. It is not clear that this choice is unwitting; and in fact, the very act of resistance suggests that Bedouin women are aware of the power structures in which they are enmeshed. Nevertheless, Abu-Lughod’s argument that resistance both opposes and is rooted in power reflects the relation between whisper networks and gender hierarchy.


The Open Secret of Sexual Harassment

Sexual harassment is a culturally intimate practice that combines tacit acceptance with official disavowal, a contradiction that provides some power to whisper networks. As defined by Michael Herzfeld in his seminal text, Cultural Intimacy, “cultural intimacy is the recognition of those aspects of an officially shared identity that are considered a source of external embarrassment but that nevertheless provide insiders with their assurance of common sociality” (Herzfeld 2016, 7). The culturally intimate is notoriously difficult for outsiders to identify, as it contains those aspects of a culture that everyone within the community knows about, but conceals from outsiders. Such concealment is necessary when the culturally intimate practice contradicts official political cultures. As a result, representatives of the community disavow such acts and claim that they never happen anymore. At the same time, however, a wide network of collusion, extending to bureaucracies and political structures that either participate or look the other way, maintains culturally intimate practices. As Herzfeld explains, culturally intimate acts constitute “the tolerance of democratic state systems for various minor offenses against formal law and morality” (Herzfeld 2016, 7). Such acts violate both formal and moral laws, but are knowingly upheld by members of a community. They generate a sense of “embarrassment, rueful self-recognition….not solely personal feelings, but…the collective representation of intimacy” (Herzfeld 2016, 11). In other words, members of a group represent their shared secrets by performing culturally intimate behaviors when outsiders are absent or unable to recognize what is happening. A culturally intimate “open secret” differs from the covert resistance because it is a source of embarrassment and pride, rather than moral sanction. While the prime examples of cultural intimacy occur on the level of the nation-state, others occur on the level of an institution; as Herzfeld writes in Cultural Intimacy, “all institutional structures are capable of generating their own peculiar intimacies” (Herzfeld 2016, 54). Within most American institutions, and the country more broadly, one of the most culturally intimate acts is sexual harassment. This covert power marks a critical departure from Scott’s thesis, which expects power to express itself freely (Scott 1985, 23). To the contrary, covert resistance may oppose covert power. At the same time, as with the Bedouin, whisper networks ground the key principle of their resistance – that sexual harassment is, indeed, bad – in the politically correct culture that many harassers publicly endorse but secretly undermine.

As with most culturally intimate elements, sexual harassment was once widely accepted. As Herzfeld writes, cultural intimacy is “highly labile. It shifts with the ideological winds of history” (Herzfeld 2016, 62). For example, many countries that once proudly embraced patriarchal models now find those models to be embarrassing in the context of a gender-egalitarian political consensus. Yet patriarchal behaviors, including harassment, remains a source of identification and rueful pride for “at least the male segment of these countries’ populations” (Herzfeld 2016, 62). Although women have sometimes been implicated in covering up for harassers, the vast majority of cases involve men harassing women, meaning that the culture of sexual harassment is primarily intimate to the men of a community (Pina, Gannon, and Saunders 2009, 128). Indeed, it was not long ago that relationships between men in power and their female employees or students were widely accepted. In Hollywood, the long myth (and reality) of the “Casting Couch,” where women would be asked to exchange sexual favors for roles, reflects the widespread acceptance of sexual harassment. As a 1956 exposé in Picturegoer described, “the terrible thing is that these facts seem to be taken for granted by people in show business” (Dessem 2017).  In the context of academia, sexual relationships between professors and students were also widely accepted until recently. For example, a 1993 panel on whether such relationships should be prohibited unanimously agreed against the prohibition, with English Professor William Kerrigan from University of Massachusetts at Amherst reasoning that he sometimes met “a female student who, for one reason or another has unnaturally prolonged her virginity… There have been times when this virginity has been presented to me as something that I… can handle” (Mcmillen 2017). Sexual harassment was not so much an open secret, as not a secret at all. Such widely practiced behaviors were not seen as morally wrong and likely would not have been termed harassment. Over time and for reasons beyond the scope of this paper, the political culture has shifted to condemn sexual harassment, making it a secret but one widely shared by most men in the institution.

As a result, the cultural intimacy surrounding sexual harassment both accounts for the extent of this “open secret” and provides the basis for women’s resistance in the whisper network. The abuses of most of the men named in the #MeToo movement, including Weinstein, were “open secrets” facilitated by a wide network of collusion. Many of Weinstein’s staff and company reported knowing about his behavior and sometimes being enlisted in facilitating his encounters with young actresses (Farrow 2017). Seth MacFarlane even made a joke about the sexual price of entry to the film industry at the Oscars in 2013, telling the five nominees for best supporting actress, “congratulations! You five ladies no longer need to pretend to be attracted to Harvey Weinstein” (Mason 2017). It would be incorrect to interpret these colluders as either bad eggs or pathetic assistants sucked into the dirty schemes of their bad-egg bosses. Within Hollywood, as within many other institutions, sexual harassment is widely and implicitly accepted even while it is publicly disavowed (Stephens 2017). Moreover, bureaucratic structures technically intended to prevent sexual harassment, from the Screen Actor’s Guild (the union for actors), to university Title IX offices responsible for implementing sexual and gender-based harassment policies, to large companies’ human resources departments, often simply look away from such acts. Most are set up to prevent sexual harassment allegations from being heard outside the institution. For instance, the Screen Actor’s Guild responds to sexual harassment allegations by asking the production house or studio to conduct an internal investigation, creating a rich space for bureaucratic collusion – especially as harassers often run or own the houses (Kirshner 2017). Human resources departments typically silence allegations in order to protect companies (Smith 2017). At universities, policies that make all employees mandatory reporters, who are legally responsible for passing along claims of sexual harassment, actively undermine whisper networks and make it much less likely that students will report their experiences (Flaherty 2015). There are limits to such collusion, but as the saying goes, “everyone knows” what goes on behind all the declarations of adherence to official political morality. As a result, while representatives of institutions claim that sexual harassment hardly happens, statistics suggest otherwise. For instance, recent surveys indicate that nearly two-thirds of women working in academic research at remote field sites experience sexual harassment, while one-fifth experience assault (Gewin 2015). A 2009 survey of the United States found that nearly half of women experienced sexual harassment at work, with higher levels among women of color, suggesting an intersection between the culturally intimate acts of racism and sexism (Rospenda, Richman, and Shannon 2009). In Hollywood, 94% of women reported experiences of harassment (Puente and Kelly 2018). Such widespread harassment reveals a culturally intimate practice that is rarely challenged by the structures of institutions, or by their members.

Currently, the cultural intimacy of sexual harassment appears to be confronted from two directions: the #MeToo movement, that confronts it, and a new populism that embraces it. As Herzfeld argues, “cultural intimacy, though associated with secrecy and embarrassment, may erupt into public life and collective self-representation” (Herzfeld 2016, 7). Populist politicians are particularly skilled at channeling and legitimizing the content of cultural intimacy in order to build a political movement by transforming “potentially offensive speech, mannerisms, and attitudes” into “legitimate alternatives to establishment values and practices” (Herzfeld, manuscript under review (2018), p. 1, cited with permission). Such performances “draw on a repertoire of culturally intimate secrets,” including prejudices and vulgarity, which had previously been met with embarrassment but are mimicked and brought into the open by populist leaders (Herzfeld, 2018, p. 2). Populism is attractive because it allows people to escape embarrassment for the attitudes and practices they already espouse. Populist leaders may not openly espouse these embarrassing attitudes, themselves, but could also engage in specific performative acts that demonstrate their membership in the intimate space, particularly working-class styles of speech an action; they may also insist “I’m not sexist, but…” or “accidentally” fail to respond to particular incidents. From the release of the Access Hollywood tape and perhaps even before then, it was clear that President Trump participates in the culturally intimacy that tacitly approves of sexual harassment. However, he has never explicitly condoned sexual harassment and would insist that he did not mean his comments and has not ever been an abuser. Rather, he performs his populism through culturally intimate styles of speech, such as assigning women numbers on the basis of their attractiveness. His cynical deployment of a culture of sexual harassment is particularly evident in his choices to condemn politicians of the opposite party such as Bill Clinton who have been accused of such acts, but stand firmly behind similarly-situated members of his own party including Alabama senate candidate Roy Moore, who has been accused of pedophilic acts. As Herzfeld discusses in the context of racism, populist leaders like Trump do not make give people these attitudes freshly made, but rather have “rendered them acceptable” (Herzfeld, 2018, p. 7). Trump has merely made acceptable certain behaviors and attitudes towards women that were already widespread in the American male. With dismissive descriptions of sexual harassment as “locker-room talk” and “boy talk,” Trump has transformed the shameful into the quotidian (Fitzgerald 2017, 485). He has turned a covert operation of power into one that is overt.


Conclusion: A New Confrontation

Whisper networks, long-standing forms of women’s resistance to regular sexual harassment within institutions, have recently generated a new overt resistance in the form of the #MeToo movement. Whisper networks had functioned in the space provided by the contradictions of male power, which simultaneously condemned and condoned sexual harassment. Yet in the same historical moment, a new populist leader has validated sexual harassment, and women have begun to vocally argue their opposition to such treatment. More research is needed to understand the relationship between these occurrences. Nevertheless, it is clear that #MeToo results from larger trends than just the Weinstein case. Weinstein’s much publicized crimes, while repellent, are not unique and therefore no more than the proximate catalyst. On the other hand, the rise of populist acceptance of sexual harassment under President Trump is a new phenomenon that could have triggered #MeToo. Whisper networks constituted a covert opposition to a covert function of power. When the covert power transformed into overt and public affirmations of sexism and sexual harassment, however, it began to seem that the key space for women’s resistance – the political acceptance that sexual harassment meets a particular moral valence – was crumbling. This forced a covert whisper network to generate an overt form of resistance, as well. In other words, the battle over sexual harassment has come somewhat into the open. As a result, previously inchoate opposition to sexual harassment is becoming increasingly legible as whisper networks become public. #MeToo activists are developing a more tangible vision of the polity, with specific understandings about how sexual harassment allegations are assessed, what exactly constitutes sexual harassment, what sanctions are appropriate, and how gender hierarchies should operate more broadly. Thus, two alternate visions of the polity – Trump’s brand of populism and the #MeToo movement – are transitioning from hidden into open conflict. What will happen next is an empirical question that will certainly provide rich insights into the relationship of power and resistance, and the possibilities for social transformation.


Works Cited


Abu-Lughod, Lila. 1990. “The Romance of Resistance: Tracing Transformations of Power through Bedouin Women.” American Ethnologist 17 (1): 41–55.

Alianza Nacional de Campesinas. 2017. “700,000 Female Farmworkers Say They Stand With Hollywood Actors Against Sexual Assault.” Time Ideas, November 10, 2017.

Bailey, F. G. 1971. Gifts and Poison: The Politics of Reputation. Pavilion Series. New York: Schocken Books.

Chapin, Angelina. 2017. “Naming and Shaming the Weinsteins of the Workplace.” The Cut, October 13, 2017.

Creswell, Julie, and Tiffany Hsu. 2017. “Women’s Whisper Network Raises Its Voice.” The New York Times, November 4, 2017, sec. Business Day.

Dessem, Matthew. 2017. “In 1956, a Fan Magazine Published a Four-Part Casting Couch Exposé. It Didn’t Go Well.” Slate, October 13, 2017.

Dibdin, Emma. 2017. “A Full List of Harvey Weinstein’s Accusers and Their Allegations.” Elle, December 13, 2017.

Dorris, Jesse. 2017. “After Kevin Spacey, Gay Men Need a Whisper Network, Too.” The New Yorker, November 8, 2017.

Fance, Lisa Respers. 2017. “#MeToo: Social Media Flooded with Personal Stories of Assault.” CNN, October 16, 2017.

Farrow, Ronan. 2017. “From Aggressive Overtures to Sexual Assault: Harvey Weinstein’s Accusers Tell Their Stories.” The New Yorker, October 10, 2017.

Fitzgerald, Louise F. 2017. “Still the Last Great Open Secret: Sexual Harassment as Systemic Trauma.” Journal of Trauma & Dissociation, April, 1–7.

Flaherty, Colleen. 2015. “Faculty Members Object to New Policies Making All Professors Mandatory Reporters of Sexual Assault.” Inside Higher Ed, February 4, 2015.

Friedman, Ann. 2012. “‘Is He on The Island?’ On the (Lack of) Professional Consequences for Sexual Harassers.” Ann Friedman (blog). August 10, 2012.

Gewin, Virginia. 2015. “Social Behaviour: Indecent Advances.” Nature 519 (7542): 251–53.

Gluckman, Max. 1963. “Papers in Honor of Melville J. Herskovits: Gossip and Scandal.” Current Anthropology 4 (3): 307–16.

Gutmann, Matthew C. 1993. “Rituals of Resistance: A Critique of the Theory of Everyday Forms of Resistance.” Latin American Perspectives 20 (2): 74–92.

Herzfeld, Michael. 2018 (forthcoming). How Populism Works.’In Populism, Essentialism, and the (new) World Order (eds) Bruce Kapferer and Dimitrios Theodossopoulos. Oxford: Berghahn.

———. 2016. Cultural Intimacy. Taylor and Francis.

Kelly, Liz, Jo Lovett, and Linda Regan. 2005. A Gap or a Chasm? Attrition in Reported Rape Cases. Home Office Research Study 293. London: Home Office Research, Development and Statistics Directorate.

Kirshner, Mia. 2017. “I Was Not Protected from Harvey Weinstein. It’s Time for Institutional Change.” The Globe and Mail, October 13, 2017.

Lisak, David, Lori Gardinier, Sarah C. Nicksa, and Ashley M. Cote. 2010. “False Allegations of Sexual Assault: An Analysis of Ten Years of Reported Cases.” Violence Against Women 16 (12): 1318–1334.

Lopez, Linette, and Chris Snyder. 2017. “Tarana Burke on Why She Created the #MeToo Movement — and Where It’s Headed.” Business Insider, December 13, 2017.

Mason, Amelia. 2017. “How Weinstein’s Sexual Misdeeds Remained An Open Secret For So Long.” WBUR.

McKinney, Kelsey. 2017. “Almost Every Single Woman I Know Has Been the Victim of Sexual Assault.” Village Voice, October 12, 2017.

Mcmillen, Sheila. 2017. “Dirty Old Men on the Faculty.” The Chronicle of Higher Education, December 6, 2017.

Meza, Summer. 2017. “What Is a Whisper Network? How Women Are Taking Down Bad Men in the #MeToo Age.” Newsweek, November 22, 2017.

North, Anna. 2017. “202 Celebrities, Politicians, CEOs, and Others Who Have Been Accused of Sexual Misconduct since April 2017.” Vox, December 22, 2017.

Petersen, Anne Helen. 2017. “Here’s Why So Many Women Knew The Rumors About Harvey Weinstein.” BuzzFeed, October 8, 2017.

Pina, Afroditi, Theresa Gannon, and Benjamin Saunders. 2009. “An Overview of the Literature on Sexual Harassment: Perpetrator, Theory, and Treatment Issues.” Aggression and Violent Behavior 14 (March): 126–38.

Press, Alex. 2017. “It’s Time to Weaponize the ‘Whisper Network.’” Vox, October 16, 2017.

Puente, Maria, and Cara Kelly. 2018. “How Common Is Sexual Misconduct in Hollywood?” USA Today, 20 2018.

Rospenda, Kathleen M., Judith A. Richman, and Candice A. Shannon. 2009. “Prevalence and Mental Health Correlates of Harassment and Discrimination in the Workplace: Results From a National Study.” Journal of Interpersonal Violence 24 (5): 819–43.

Scott, James C. 1985. Weapons of the Weak: Everyday Forms of Peasant Resistance. New Haven: Yale University Press.

Shafrir, Doree. 2017. “What To Do With ‘Shitty Media Men’?” BuzzFeed, October 12, 2017.

Smith, Tovi. 2017. “When It Comes To Sexual Harassment Claims, Whose Side Is Human Resources On?” All Things Considered. National Public Radio.

Stephens, Bret. 2017. “Weinstein and Our Culture of Enablers.” The New York Times, October 11, 2017, sec. Opinion.

Tolentino, Jia. 2017. “The Whisper Network After Harvey Weinstein and ‘Shitty Media Men.’” The New Yorker, October 14, 2017.

Yoffe, Emily. 2017. “Why the #MeToo Movement Should Be Ready for a Backlash.” Politico, December 10, 2017.

Zeitchik, Steven. 2016. “Prosecutor Says Bill Cosby’s Lawyers Are Trying to Intimidate 13 Women Who Want to Testify against Him.” Los Angeles Times, November 2, 2016.



[1] Many thanks to friends and fearless organizers Kay Xia, Bella Roussanov, Niharika Singh, and Sejal Singh for discussing these issues with me, to Professor Michael Herzfeld for supporting the development of this paper, and to my mom for driving me home while I finished writing it.

[2] In this paper I use the term “sexual harassment” as the broad category under which most acts circulated by whisper networks are described; however, a number of other acts extending to sexual assault and rape also fall into this category.

[3] While false accusations do occur, and raise a significant concern, they are much less frequent than critics of the #MeToo movement imply. For instance, a recent study at a large Northeastern university found that the false accusation rate was between 2 and 10% (Lisak et al. 2010). Moreover, few false accusations have consequences, as most are withdrawn; for instance, in a study by the British Home Office of rapes reported to the police, only 6 of 216 false allegations led to an arrest, and only 2 led to charges (Kelly, Lovett, and Regan 2005).

Understanding Immigration: A Closer Look at the Fiscal Impact of Immigrant Charactersitics

Jean Thirouin


This paper attempts to further research on the fiscal impact of immigration. My findings are illustrated by computing the net fiscal impact, in present value terms, of admitting one additional immigrant, conditional on education, gender, and age at the time of immigration. I demonstrate that the average immigrant arriving past age 34 has a lifetime negative fiscal impact. Additionally, a college educated immigrant arriving past age 52 will have a lifetime negative fiscal impact while a non-college educated immigrant will roughly have a lifetime negative fiscal impact, regardless of age at arrival. Further, I confirm that age at arrival matters, and determine that arrival prior to working age influences educational attainment. Finally, I provide a household life-cycle model that sheds light on the fiscal contribution of immigrating families.


With an aging population and a social security system set to run out of funds in the foreseeable future, is immigration the key to slowing the decline in the working-age share of the population while helping bolster a strained fiscal deficit? This question is at the heart of the public policy debate on immigration. The key to answering this question lies in understanding the age structure of immigrants and their life-cycle fiscal impact.

Immigration has greatly increased in the past decades. From 1940-1950 there were about 1mil immigrants, 1950-1960 had 2.5mil immigrants, 1960-1970 had 3.3mil immigrants, 1970-1980 had 4.5mil immigrants, 1980-1990 had 5.7mil immigrants, 1990-2000 had 11.3mil immigrants, and 2000-2010 had 8.8mil immigrants (Gibson 1999). Along with the increase in immigration has been an increase in public opinion on the best public policy pathway. Some have viewed immigrants as a “fiscal drain”, while others have believed in their ability to provide a solution to the aging population problem. Skilled workers that would immigrate early and immediately contribute through taxes would likely lead to large positive net fiscal effects, even accounting for the subsequent costs of retirement (Kjetil 2000).

A particular interest of immigration is their impact on the host country’s welfare system. The United States’ pay-as-you-go system relies on the working age population to support retirees, which lends itself to the argument in favor of increased immigration. From 1980 to 2010, though, the share of inflows of immigrants aged 50 to 74 increased from 8.9% to 14.5% (Yi 2014). As can be seen in Figure A, the mean age of immigrants has been steadily rising since the 1970s, with women, on average, older than men. If immigrants are entering the country at a later age, they may be worsening the aging situation. On the other hand, Figure B and C paint a different picture as they show an upward trend in the educational attainment of immigrants. Between 1950 and 2007, the foreign-born share of employees in the U.S with a masters, professional, or doctorate degree rose from 5.9% to 18.1% (Peri 2010). They may be coming in later, but if they are more skilled, their impact on the welfare system may also be larger.

Although immigration is a hot policy topic, there have only been three major changes in U.S immigration policy in recent history: 1) the 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) or Welfare Reform, 2) the 1920s national-origin quotas, and 3) the 1965 elimination of such quotas (Klopfenstein 1998). The 1996 Welfare Reform served to restrict immigrant access to Federal welfare benefits such as Medicare and food stamps for their first five years in the country, however, states could grant aid out of their own funds. Additionally, in 2004 the annual cap for new H-1B visas was lowered from 195,000 to 65,000 – essentially an attempt to reduce competition between similarly-educated immigrants and natives (Peri 2010).

What is the best course of action for the government to solve a growing fiscal deficit tied to an aging population problem? Should the government focus more on changing the level or mix of immigrants (Auberbach Oreopoulos 2000)? The goal of this paper is to shed light on the quantitative aspect of this debate.

The remainder of the paper proceeds as follows. I provide a brief literature review in Section I. Section II describes my data, Sections III, IV, V provide my method along with my results, and Section VI constitutes a household life-cycle model. Section VII discusses limitations, government policy implications, and concludes.


I. Literature Review

 Immigration has a growing body of literature with two mains sides: 1) the labor market outcomes, and 2) the fiscal impact (Borjas 2013). Much of the focus has been on the demand side of the labor market impact caused by immigrants. More specifically, attention has been paid to the effects that immigrants have had on native wages. While research has shown immigrants to lower native wages, especially those with less than a college level education, others have argued for the complementary nature of immigrants (Ottaviano Peri 2011). My paper focuses on the fiscal impact of immigrants therefore I will mainly reference the literature on that front.

Despite the strong implications of immigration for public finance, there is a limited amount of literature addressing the cost-benefit life-cycle aspect of immigration (Friedburg Hunt 1995). However, the related literature is important in formulating hypotheses and making assumptions in my later models. One main focus has been the importance of immigrating at an early age (OECD)(Shaafsma Sweetman 2001)(Myers et al. 2009)(Seeborg Sandford 2003). Comparably, human capital accumulation and language proficiency have been determined to be two of the most important characteristics of immigrants at arrival, closely tying in to age of arrival effects (Lagakos 2016). Intuitively, an immigrant arriving as a young child has a much higher likelihood of assimilating into the culture and surpassing the language barrier than an immigrant in his late thirties (Hoyt Chin 2016). Additionally, immigrants from poor countries will tend to accumulate much less human capital in their birth countries before migrating (Lagakos 2016).

No clear consensus has been reached about the use of welfare by immigrants, however. While some studies have shown that longer time spent in the host country (Sweden) have led to decreased rates of welfare use, others have shown the exact opposite (Hansen Lofstrom 2010). Another issue concerning welfare literature has been the lack of separation between welfare usage and welfare eligibility (Pekkala Kerr 2011). This is especially important to consider due to changes in eligibility over time.

There are two main techniques for evaluating the fiscal impact of immigrants. The first estimates the growth to GDP due to growth in labor supply (Borjas 1995)(Freeman 2006)(Drinkwater et al. 2007). The second relies mostly on accounting methods and estimating the total cost and benefits that immigrants will have on the economy, which varies greatly by stage of life (Pekkala Kerr 2011). The accounting method is the type my paper focuses on. More specifically, my paper sheds light on the varying net impact at all stages in an immigrant’s life. As well, I expand on my different life-cycle models to estimate household impacts of migrating families for alternate family structures.

II. Data

My data comes from the March Annual Social and Economic Supplement (ASEC) supplemental survey of the Current Population Survey (CPS). It is all downloaded from the Integrated Public Use Microdata Series (IPUMS) website which allows the identical coding of variables across time for easier cross-time comparison. Additionally, IPUMS-CPS provides all estimations with the person weights in the census data files (Flood King 2015). My sample consists of pooled microdata, information on individual persons and households, for the years 1994 to 2016.[1] Data from previous years was not used since the age classification of immigrants could only be constructed from a variable in my time frame.

The IPUMS-CPS data source provides variables crucial to my research such as age, year of immigration, gender, education, taxes, benefits received, among many others. I classify immigrants by marking immigration population (impop) = 1 if they answered the question about year of arrival. Unlike most of the variables in the CPS, the major tax variables such as FEDTAXSTATETAX, and FICA are not the result of direct questioning of respondents. Instead, the values for the variables come from the Census Bureau’s tax model which was updated in the fall of 2004 to produce more accurate tax estimates. The model simulates tax returns for each individual to produce the estimates needed by incorporating information from non-CPS sources such as the Internal Revenue Service’s Statistics of Income series, the American Housing Survey, and the State Tax Handbook. I adjust all monetary values for inflation to the 1999 CPI indicator using a predefined IPUMS-CPS variable which multiplies a factor for each non-1999 year to a dollar amount in that year.

III. Net Yearly Contribution Model

 A. Net Yearly Contribution – Average

The most important variable I constructed for each observation in my data set is an individual’s net contribution (netcontri) to the government. This is defined as their total taxes paid (taxtot) minus the total benefits received from the government (incgov). Since my data set did not contain predefined variables for total taxes nor total benefits received, using IPUMS’ recommendations, I constructed them by summing across individually reported values of its components. The compositions are defined below:Since netcontri is central to the remainder of my analysis, it is crucial that the variables it is comprised of are as closely reflective to the actual values as possible. I perform a sensitivity analysis in order to determine their viability. Taxtot has relatively little room for error since it already accounts for the three major sources of tax revenue – federal tax, state tax, and FICA. However, an individual’s income received from the government is more complex since it can come from many different sources. I experiment with alternate compositions such as the inclusion of disability income, survivor’s benefits, and earned income tax credit. Plotting the average difference between these compositions and the one mentioned above by age, I notice a negligible impact when looking at immigrants only. I run the same analysis for natives and notice a slightly greater impact past the age of 50 yet it remains insignificant. I plot both immigrants and natives separately because I am (correctly) predicting that the composition of variables such as survivor’s benefit differs for immigrants and natives.

After constructing my net contribution variables for each observation in my data, the first step in my analysis is to find the average contribution at each age. Plotting the results gives me a first pass look at the impact an immigrant has on the government budget at each age in their life. As can be seen in Figure 1, an immigrant’s impact over their life follows an intuitive understanding of one’s life-cycle earnings. Earnings begin around 15 years of age and sees its greatest growth between the early twenties to the mid-thirties, where it stands around $5,300/year. It begins roughly leveling off to around $6,200/year, the peak earnings stage during the fifties. From there, the decline begins, which becomes sharp around the early sixties, the age of retirement, and roughly levels off around -$6,000/year for the remainder of an immigrant’s life.

To compare immigrants’ average earnings to natives, I run the same analysis restricting the data to natives only. Shown in Figure 1, a native’s average impact at each age follows a similar pattern with slight differences. Earnings begins at age 15 and rises faster than immigrants until the mid-thirties where it reaches around $6,500/year. Instead of leveling off, though, earnings continue to increase, albeit at a slower rate, until the fifties where they peak at $8,200/year. A similar sharp decline can be seen around the early sixties before roughly leveling off at -$7,500/year for the remainder of their life.

Figure 1 shows the aggregation of all immigrants and all natives, whether male, female, college educated, living in the US for a short/long time, etc. In other words, the results discussed above can be generalized to “an average immigrant” and “an average native”. However, when facing the decision to accept a foreigner into the country, other differentiating characteristics mentioned above are observed. Below I provide a contribution analysis based on the main characteristics observed when a foreigner attempts to immigrate to the United States, limited to the ones observed in my data set.

Net Yearly Contribution – Education

I continue my analysis of an immigrant’s yearly impact on the government budget by taking into account their level of education. I construct a binary variable college that is defined as 1 if an individual has completed at least a year of college all the way up to doctorate degrees, and 0 for any educational attainment below one year of college. I then find the average contribution per year at each age for both immigrants and natives, with and without a college education.

The results for college and non-college educated immigrants and natives are displayed in Figure 2. We can see that both college and non-college educated immigrants follow a very similar life-cycle pattern mentioned earlier in our “average immigrant” case. The stark difference is revealed in their earnings growth and peak earnings. An immigrant that has at least one year of college education will, on average, peak at around $10,000/year net contribution in his mid-fifties. A less educated immigrant, on average, will peak and stagnate closer to $2,000/year. Said in another way, a college educated immigrant, on average, will have a five times greater positive impact on the government budget than a non-educated immigrant.

Looking at the impact that education can have on an immigrant’s earnings clearly reveals its importance, yet even more can be learned when compared to the impact that education has on a native’s earnings. A college educated native mirrors the earnings of a college educated immigrant, with peak earnings slightly surpassing $11,000/year. Interestingly, non-college educated natives perform much better than non-college educated immigrants. A non-educated native reaches peak contribution closer to $3,500/year, or almost twice that of an identical immigrant.

Figure 2 reveals another comparison between immigrants and natives. We saw that natives are reaching higher peak earnings during their working life and therefore have a higher net contribution. Once retirement age in the early sixties is attained, though, we now see natives have a much higher negative contribution than immigrants. Are contributions during one’s working life offsetting the income received from the government post-retirement? If not, at what age is the cutoff? This phenomenon motivates the life-cycle model I build later in the paper that can shed light on immigrant and native impacts on the government budget throughout the entirety of their life.

It is impossible to tell strictly from the graphs what factors may be causing the discrepancy in earnings between immigrants and natives[1]. Non-educated immigrants most likely suffer from the language barrier, strongly decreasing their already limited employment options. A non-educated native may be able to benefit from relatively higher earnings based on an ability to be employed in communication-based jobs. At higher levels of education, however, immigrants don’t suffer as much from the language barrier as they can go into high earnings fields that are focused more on quantitative skills rather than communicative ones.

C. Net Yearly Contribution – Gender

Should the government have the right to discriminate against incoming immigrants depending on their education? If so, should they be able to do the same regarding an immigrant’s gender? Whether or not they should, a lot can be learned by analyzing the differential earnings of men and women. The trends observed earlier in both “average immigrants” and educated immigrants can further be explained by separating the data according to an immigrant’s gender. Additionally, these trends prove useful later in understanding the dynamics of the household life-cycle model.

Figure 3 illustrates the net contribution gap between male and female immigrants. It is evident that men have much higher earnings growth throughout their working life. Both men and women seem to reach peak contributions around their fifties, but men on average contribute close to $8,500/year while women peak closer to $4,000/year. Once retirement is reached, both men and women follow similar net contribution trajectories into the negatives.

Figure 3 also allows for the comparison of the effect of gender on the net contributions of both natives and immigrants. Native men and women, similarly to immigrants, have differential impacts on the government budget at each age before retirement. However, the men’s highest contributions are about $10,500/year with the women’s around $5,500/year, both almost $2,000 more than immigrants for each gender. Unlike immigrants, native women have a slightly smaller negative impact than native men into their seventies and beyond.

IV. General Lifecycle Model

The net contribution model reveals the general trends that are typically seen in an immigrant’s life-cycle impact on the government budget at each age. Throughout their working life, fifteen to early sixties, they have a net positive effect on the budget each year by paying more in taxes than they are receiving in welfare. Additionally, the model provides insight into the effect of differing characteristics on the magnitude of immigrants’ impacts. As shown, education and gender play an important role in predicting future earnings.

However, the model fails on two fronts: 1) it doesn’t reveal whether an immigrant (or native) has a net positive impact on the government budget over the entirety of their life, and 2) it does not factor in a potentially crucial factor, the amount of time spent in the United States. I build two different models to address each problem. The first sums an immigrant’s contribution over their lifetime, showing at what age an incoming immigrant no longer benefits the country’s government budget. The second expands on this model by restricting the sample for each age of arrival, essentially allowing me to elicit the effect that duration in the country has on earnings.

A. Lifecycle Model 1 – Average

My first pass life-cycle model is generalized to the “average immigrant” and builds off the previous yearly net contribution model. I use the net contribution at each age and sort them negatively, from oldest to youngest[3]. I then sum up every yearly contribution. Each resulting data point, plotted in Figure 4, indicates an immigrant’s future impact on the government budget at that age. This allows me to answer the question, “What is an immigrant’s lifetime impact if they arrive at age x?”.

Figure 4 also indicates the point at which positive contributions during the working life offset negative contributions after retirement. An immigrant arriving at age 39 will break even in terms of lifetime contributions to the government budget. From a pure government budget standpoint, this indicates that any immigrant over the age of 39 should not be allowed to enter the United States. In Figure 4, we can also observe the lifetime contributions of natives. It is quite interesting to see that natives have the exact same cutoff point, 39 years of age – even though they contribute more during their working life, they receive more benefits post-retirement. Lastly, we can see that, on average, an immigrant and native who start off at the same point, age 15, will differ in their lifetime impact by $19,000, with natives contributing $91,000 and immigrants contributing $72,000.

B. Lifecycle Model 1 – Education

Analogous to the yearly contribution education model, I use the lifetime contribution method developed in the “average immigrant” case to demonstrate the effects of education on an immigrant’s total impact to the government budget. Using the previously constructed college variable, I separate the data into college educated and less than one year of college education. Sorting the ages from oldest to youngest, I sum up all observations and plot the resulting data points shown in Figure 5.  I include the equivalent analysis for natives to use as a comparison “control” group.

The results seen in Figure 5 illustrate the drastic impact of education that was not as evidently clear from the yearly contribution model. No matter at what age they arrive to the United States, an immigrant with a high school diploma and less will never have a net positive impact on the government budget. In other words, they will always take out more from the United States than they will be able to give back over their entire lifetime. In stark contrast, an immigrant with at least one year of college education and above will be profitable over their entire lifetime so long as they arrive before age 53. In addition, a college educated immigrant who starts working at age 15 will, on average, contribute $250,000 to the government budget.

Figure 5 reveals another interesting trend that was unapparent in the yearly contribution model when comparing natives and immigrants. Although college educated immigrants seemed to reach lower peak contributions than natives during their working life, over the course of their entire life they outperform them in net positive contribution. It doesn’t make intuitive sense in this context to compare the break-even point since natives don’t just “arrive” in the country at a later age. However, it is still possible to draw conclusions about efficiency in impact over a lifetime by noting that it takes an immigrant arriving at age 53 or later to become a net loss while a native must start working at age 49 or before to positively contribute. Non-college educated natives and immigrants follow very similar impact trends over their lifetime, with natives’ post-retirement benefits outweighing their higher earnings during their working life, leading to very similar total impacts of -$66,000 and -$75,000, respectively.

C. Lifecycle Model 1 – Gender

Lastly, I expand upon the trends seen in the yearly contribution gender model by looking at the impact gender has on lifetime government budget impact. As mentioned earlier, discriminating due to gender may not be the right policy decision, but understanding the life-cycle paths of both men and women can be useful in forecasting household impacts. This is especially relevant for a country with an immigration policy like the United States, which places a large emphasis on allowing family members to re-unite through immigration (Borjas 1996).

Figure 6 follows the same method as the education and average lifecycle models, this time separating yearly net contributions based on gender, and summing up each age into a lifetime impact. Figure 6 demonstrates the lifetime disparity in contributions between immigrant men and women. Assuming a working life starting at 15, men, on average, will contribute $141,000 while women, on average, will basically break even at $2,500. It is interesting to note the parabolic shape of the lifetime impact curve, which indicates that the worse age at which an immigrant can arrive is 64 for women and 66 for men.

Figure 6 also compares the gender impact between immigrants and natives which allows us to conclude that both native men and women, on average, have a greater lifetime impact, assuming they begin working at age 15. Having seen that, on average, immigrants perform worse than natives, it makes sense that both immigrant men and women perform worse than natives. Since educated immigrants, on average, perform better than natives, however, then it must mean that one of two things (or both) could be happening. Either there are less educated immigrants than natives in my sample, pulling the immigrant averages down, or there is an omitted effect, such as the duration of stay.

My sample data reveals that 38% of immigrants have at least one year of college education and 37.7% of natives have the same educational attainment. Therefore, the ratio of college educated to non-college educated is irrelative in this context. Earlier, we explored the age at immigration trends which revealed that, on average, immigrants have started to come in their late twenties, early thirties. I believe that the “average” immigrant’s life-cycle impact is picking up this effect of later arrival in one’s life.

V. Age at Immigration Lifecycle Model

In my second version of the life-cycle model, I test the hypothesis that earlier immigration and longer duration in the United States positively impacts earnings and therefore an immigrant’s net contribution. My belief is that a 35 year old immigrant who arrived at age 1 has a different impact on the government budget than a 35 year old immigrant that arrived at age 35. The topic of many studies, it is continually shown that the earlier an immigrant arrives to the United States, the higher the likelihood that their earnings will surpass those that arrive later in their life. This phenomenon has been attributed to different factors, with higher cultural assimilation and education attainment two of the most important (Sandford Seeborg 2003)(Hoyt Chin 2010)(Schaafsma Sweetman 2011). In the same vein, the earlier an immigrant arrives, the longer he/she can positively contribute through lengthened working life taxation.

A. Lifecycle Model 2 – Average

The basis for my second life-cycle model is very similar to the previous one. The main difference is that I restrict the sample for each age of arrival. To do this, my first step is constructing the variable ageimmig. For each observation, I take the year at which they immigrated and subtract it from the year in which they responded to the survey. I then build a loop that runs through each age, starting at 0 and ending at 90. For each iteration of the loop, I restrict the sample to that loop number’s age.

Similarly to previous models, my next step is calculating the average yearly contribution at each age. This time, having a restricted sample means that I am averaging the yearly contributions for only those that immigrated at age x. I then sort my data from oldest to youngest, and sum each age to arrive at a lifetime impact. The lifetime contribution of arriving at age x is the only value that I am interested in since I am calculating that same impact for each different age of arrival. With this in mind, for each iteration of the loop, I only save that one observation. Running another loop, I append the lifetime contribution for each age of immigration to build my final model. Each observation in this model shows the lifetime impact of immigrating at age x, and unlike previous models, it accounts for the time spent in the United States. The results are plotted in Figure 7.

There are a few different results to focus on from this graph. The most important one, this time more robust than in previous models, is the age of immigration where working life contributions will offset retirement benefits received. Figure 7 suggests that any immigrant arriving after the age of 34 will, on average, have a total negative impact on the government budget if they remain in the United States throughout the remainder of their life.

The second result, which confirms my hypothesis regarding the positive effects of arriving early and having a longer duration of stay, can be observed by looking at the trends of immigrants arriving between age 0 to 15. Since all immigrants arriving in that age range start working at age 15, in theory, if there is no effect of duration of stay or earlier arrival, they should all have the same lifetime impact. Clearly, Figure 7 illustrates the opposite – a child immigrating at age 1 will contribute on average roughly $287,000 while a child immigrating at age 10 will contribute on average only $179,000. These findings are in line with (Schaafsma Sweetman 2011) and (Hoyt Chin 2010).

Thirdly, there is one interesting “spike” around age 20-27 (there are a few others around age 7 and age 45 but those qualify more as noise than structural trends). Why is it that immigrants arriving between the ages of 20 and 27 incrementally begin earning more while the overall trend shows that no matter what age you arrive, the older you are, the lower your total impact? We can’t draw any conclusions about this phenomenon from Figure 7 but subsequent variations to the model will attempt to answer this question. My theory is that those arriving between 20-27 have a higher likelihood of having completed college, are looking for work abroad, and therefore are, on average, more educated than those coming in around 18 years old.

Fourth, as we saw with previous lifetime contribution models, Figure 7 also highlights a parabolic shape to the age of arrival curve. We can see that (ignoring the noise spike around age 44), the curve levels off around age 60 before rising again. This result makes intuitive sense since 60 is right around the age of retirement. If an immigrant arrives around that age and never contributes, only receiving benefits from the government, the later the immigrant arrives past that point, the smaller his negative impact will be on the government since his life expectancy will decrease.

Lastly, I want to point out that we are dealing with age of arrival observations across many different years so taking averages mitigates the effects of economic downturns or booms in certain periods. It is also important to note that Figure 7 has more noise than previous models. This is in part due to the sampling size I am using and having further restricted each data point to observations for that age of immigration only.

The age at immigration trends discussed earlier revealed that immigrants have, over the past five years, been coming in to the United States around age 28. From an immigration policy standpoint, it is interesting to see the lifetime impact that the average 28-year-old immigrant has on the government budget amounts to roughly $99,500. Although the graph has not been included, it is also interesting to see that a 27-year-old man has a $234,000 lifetime impact and a 29-year-old woman has a $25,600 lifetime impact, on average.

B. Lifecycle Model 2 – Education

 I continue to build on variations to my second life-cycle model by evaluating the impact trends due to differing educational attainment. Comparable to my hypothesis regarding the effects of earlier arrival, I also believe that the effect of education will shed more light on the effect of age at arrival. Simultaneously, age at arrival has an impact on educational attainment, as a younger immigrant will have an easier time assimilating to the culture, therefore giving him/her a better chance at higher human capital accumulation (Schaafsma Sweetman 2011)(Myers et al. 2009).  Once again, I run the same loops as discussed in the previous section, this time separating my sample into those with at least one year of college education and those with less.

In Figure 8 we can see that earlier arrival evidently has an impact on lifetime contribution. Restricting once again our analysis to those arriving between ages 0 and 15, however, we don’t observe the same major differences in lifetime contributions for those college educated. While age 0 seems to be an outlier, ages 1 to 14 all oscillate in the earnings ranges of $345,000 to $400,000. What could be the reason for not seeing a clear impact of earlier arrival? This is most likely because they all end up with a college education. The impact of earlier arrival may be increased educational attainment, manifested in the oscillation observed in Figure 8. The homogeneity in my sample would simultaneously suggest that earlier arrival, age 1 versus age 10, affects future lifetime contribution by affecting the chances of achieving a college education. Therefore, when looking at a sample of only college educated immigrants, the effect of earlier arrival (before working age) is non-existent.

Around age range 40-47, we observe a significant spike most likely due to highly-educated immigrants, those with PhDs, arriving at those ages. Another interesting feature of Figure 8 is the lifetime impact cutoff point of $0, which takes place at age 52; this is much older than the 34 years of age suggested for the “average” immigrant. This lifetime graph is also very particular in that it doesn’t have the same parabolic shape as all the previous ones. Rather, past age 52 it oscillates around $0. I do not have any intuitive theory as to why this takes place, however, my best guess is that older college educated immigrants coming to the United States are most likely wealthier. Instead of coming for work or medical care, they come to settle and retire – creating minimal impact on the government budget. Additionally, I addressed this data concern in the previous section, and it may have an even stronger impact since my sample has greater restrictions, but my results may be in part due to a lack of enough observational data. This may also explain the increased noise in the college-educated data relative to the non-college educated.

For non-college educated immigrants, the earlier arrival effect does take place. Within that sample of immigrants, arriving at age 1 makes a significant difference from arriving at age 10. Since their educational attainment remains low throughout their life, a potential theory might be the effect that the language barrier has on future earnings – the later they arrive, the stronger the effect. The most puzzling aspect of Figure 8 for non-college educated immigrants is that the lifetime impact $0 cutoff takes place at age 12. Arriving at age 12 versus arriving at age 13, unlike arriving at age 24 instead of age 25, will not increase your net contribution by an extra year since you may only start working at age 15. This provides even stronger evidence that there must be some significance in arriving at an earlier age, even prior to working age eligibility. Unfortunately, I am unable to explain why the parabolic shaped curve’s inflection point takes place around age 45. The only theory I could have for a 55-year-old making less of a negative impact on the government budget than a 45-year-old is if they both come in, receive benefits right away, and don’t find work. The plausibility of a working age foreigner receiving benefits upon immigration is relatively low, however.

VI. Household Lifecycle Model

Immigrants who come to the United States are often not alone – they come with families. I test my hypothesis that the impact of a family on the government budget differs from the impact of a sole immigrant through the construction of a household life-cycle model. This model helps me better approximate the impact of children on a household while varying the different types of household possibilities.

I construct four different types of immigrant households: 1) a parent and child, 2) a father, son, and daughter, 3) a mother, son, and daughter, and 4) a father, mother, son, and daughter. I assume that the parents have children at age 25 and that once the parents die the children keep contributing until their own death. For each household, I attempt to answer the question, “What is the contribution of a household when the parent(s) immigrate(s) at age x with child(ren) aged x-25?”. More importantly, the answer to this question helps shed light on the optimal immigration policy when considering multi-person households. I plot my results in Figure 9.

To project the 2-person household’s net contribution, I sum the contribution of a parent immigrant arriving at age x, with their child’s contribution being an immigrant arriving at age x–25. Doing this for every x>24 provides a full view of all age possibilities at parent arrival. I construct the 3-person households by summing the contribution of either a male or female parent arriving at age x>24 with their son and daughter arriving at age x-25. Similarly, to construct a 4-person household I sum the contribution of a male immigrant and female immigrant arriving at age x>24 with a male immigrant and female immigrant arriving at age x-25.

My results show that a 2-person household will positively impact the government budget over their lifetime if the parent arrives before age 57. Similarly, a 3 and 4-person household will also positively impact the government budget as long as the parents arrive prior to age 57. The shape of the household lifecycle curves reveal that a 3-person male parent household, on average, will have a greater positive contribution on the government budget if the parents arrive prior to age 57, however, the 4-person household will have a greater negative contribution if they arrive past that age.

The 4-person household follows a very similar pattern to the 3-person male parent household, revealing that a female spouse does not seem to have much of an impact on the government budget. In fact, around parent arrival age of 30-57 the 3-person male parent household shows a stronger positive impact than the 4-person household. The 3-person female parent household and 2-person household both have less positive and negative overall impacts on the government budget. A 3-person household led by a male shows a stronger positive impact prior to arriving at age 57 than one led by a female, however, past age 57 the magnitudes of their impact are very similar. It is also important to note that around age 47, the 3-person households both begin to have a more positive impact than the 4-person household. This is likely due to the parents hitting retirement age sooner and with 2 parents rather than 1 in the household, the negative impact is amplified.

Interestingly, all curves follow a slight parabolic shape whereby a parent arriving around age 72 will have the worse impact on the government budget than if they came at an earlier or later age. This may be explained by the amount of time that both the parent and child will spend receiving aid from the government past retirement age. A parent coming in at age 72 can expect to receive welfare until death, while their child at age 47 will contribute less over their working life than when they hit retirement. Past that point, the older households will not receive welfare for as many years and therefore will have less of a negative impact.

VII. Conclusion

A. Limitations

One of the main limitations of my paper is that it ignores the impact of immigrants on native wages and employment displacement effects, and instead relies purely on estimating their fiscal impact. The argument is usually made that immigrants are displacing native workers when they gain employment, which would mean that the taxes paid for the job displaced does not change whether an immigrant or native holds the position. Under my model it would seem as if the immigrant is benefitting the government by the total amount of taxes paid. Additionally, while the immigrant may not receive welfare, the now displaced worker may begin receiving welfare from the government, creating a net loss.

Another potential limitation in my paper is the lack of a death discount factor, which would better predict lifetime impacts by discounting it based on life expectancy. According to the Social Security Administration’s Actuarial Life Table, male life expectancy ranges roughly between 75-80 years old while female life expectancy ranges between 80-85. These ranges are very similar to where my data cuts off, which mitigates the effects of a large impact on my results.

Lastly, this paper does not account for the public cost of education. My results show that the younger an immigrant arrives, the higher the lifetime contribution. However, while an immigrant arriving prior to schooling age is shown to have a greater positive fiscal impact than one that may arrive at age 25, for example, the cost of education could impact the net impact. If both immigrants end up having the same level of schooling, one’s native country will have paid for it while the other will have it paid for by the United States. It makes it extremely difficult to compare such a scenario since other factors such as cultural integration can have a large impact on future success, regardless of whether two immigrants have the same level of education.

B. Policy Implications

 To better understand policy implications of the results displayed in this paper, it’s important to understand the current U.S. Immigration Policy Goals. These revolve around 4 axes: 1) Economic: increase labor supply, especially where skill deficits exist, 2) Humanitarian: reunite families, 3) Cultural: ethnic and racial diversity, 4) Political: allowing or refusing certain political refugees. This contrasts to the Canadian immigration “points system”, which allocates visas to prospective immigrants by awarding points based on language proficiency (28), education (25), experience (15), age (12), employment contract (10), and adaptability (10).

My results clearly indicate that the United States should actively solicit young, highly-skilled immigrants. Similar to Canada, my results for the average immigrant point to the optimal working ages of 18-35, my cutoff point showing age 35 as the beginning of a net loss. The only difference being that they award no points for immigrants below the age of 18, while my results all indicate that the younger the immigrant arrives, the higher the lifetime contribution.

Based on the two main metrics observed, age at arrival and education, my results suggest that immigration can be a solution to the aging population and fiscal deficit. If the U.S. goal of immigration aims to solve this growing problem, the government has a clear incentive to implement a similar “points” system to Canada. Other potential government interventions could include ways to reform social benefits provided to immigrants through optimal time-dependent structures. These would ensure that they are making a net positive impact on the system over their lifetime, while also considering their impact on the native population.

C. Final Remarks

This paper demonstrates that immigrants have strong quantitative implications for fiscal policy in the United States. In particular, this paper investigates the optimal lifetime contributions based on age at arrival, education, and gender. The findings throughout the paper are illustrated by computing the net fiscal impact, in present value terms, of admitting one additional immigrant to the United States, conditional on education, gender, and age at time of arrival. The lifetime contributions vary considerably across these three characteristics, with large and positive values for college-educated immigrants arriving in the earlier part of their life.

Using a yearly net contribution model, two life-cycle models, and a household contribution model, I demonstrate that the average immigrant arriving past age 34 has a lifetime negative fiscal impact. Additionally, a college educated immigrant arriving prior to age 52 will have a lifetime positive fiscal impact while a non-college educated immigrant will roughly have a lifetime negative fiscal impact, regardless of age at arrival. Further, I confirm that age at arrival matters, and determine that arrival prior to working age influences educational attainment. Finally, I provide a household life-cycle model that sheds light on the fiscal contribution of immigrating families.

My research has a few avenues for add-on research. First, I do not distinguish between the level of welfare and taxes for each individual’s contribution. Understanding that dynamic might better motivate welfare or tax reform. Second, I focus on the mix of immigrant characteristics, but not on the level of immigration[4]. Third, while I have data on education, age, and gender, a further analysis could look at the effects of language proficiency, skill, and native country of birth (Lagakos 2016). Lastly, it is important to note that while these results are based on U.S. immigration policy, there is an external validity concern, as each country has its own unique welfare system policies that also differ in magnitude.



[1] 2017 NAS report on fiscal impact of immigration justifies pooling by finding great similarity between patterns in 1995-97 and 2011-13

[2] Literature review addresses this point.

[3] Neat trick I learned from Professor Hendren!

[4] I do this because Kerr shows only minor displacement effects, even after large immigration inflows.



Auerbach, Alan J., and Philip Oreopoulos. “The Fiscal Effect Of U.S. Immigration: A Generational-Accounting Perspective”. Tax Policy and the Economy 14 (2000): 123-156.

Bleakley, Hoyt, and Aimee Chin. “Age At Arrival, English Proficiency, And Social Assimilation Among US Immigrants”. American Economic Journal: Applied Economics 2.1 (2010): 165-192.

Borjas, G. J., and L. Hilton. “Immigration And The Welfare State: Immigrant Participation In Means-Tested Entitlement Programs”. The Quarterly Journal of Economics 111.2 (1996): 575-604.

Borjas, George. “Immigration And The American Worker: A Review Of The Academic Literature”. Center for Immigration Studies (2013).

Flood, Sarah, Miriam King, Steven Ruggles, and J. Robert Warren. Integrated Public Use Microdata Series, Current Population Survey: Version 4.0. [dataset]. Minneapolis: University of Minnesota, 2015.

Friedberg, Rachel M, and Jennifer Hunt. “The Impact Of Immigrants On Host Country Wages, Employment And Growth”. Journal of Economic Perspectives 9.2 (1995): 23-44. Web.

Gibson, Campbell, and Emily Lennon. “Historical Census Statistics On The Foreign-Born Population Of The United States: 1850 To 1990”. Working Paper no.29, U.S. Census Bureau (1999).

Hansen, Jorgen, and Magnus Lofstrom. “Immigrant Assimilation And Welfare Participation: Do Immigrants Assimilate Into Or Out Of Welfare?”. The Journal of Human Resources 38.1 (2003): 74.

Klopfenstein, Kristin. “A Reconsideration Of Immigrant Assimilation And Welfare Participation”. (1998).

Lagakos, David. “Life-Cycle Human Capital Accumulation Across Countries: Lessons From U.S Immigrants”. National Bureau of Economic Research 21914 (2016): n. pag. Print.

Myers, Dowell, Xin Gao, and Amon Emeka. “The Gradient Of Immigrant Age-At-Arrival Effects On Socioeconomic Outcomes In The U.S”. The International Migration Review 43 (2009): 204-229.

Ottaviano, Gianmarco I. P., and Giovanni Peri. “Rethinking The Effect Of Immigration On Wages”. Journal of the European Economic Association 10.1 (2011): 152-197. Web.

Pekkala Kerr, Sari, and William R. Kerr. “Economic Impacts Of Immigration: A Survey”. SSRN Electronic Journal (2011): n. pag. Web.

Peri, Giovanni. “Highly-Educated Immigrants And Native Occupational Choice”. (2010).

Sandford, Jeremy, and Michael C. Seeborg. “The Effects Of Ethnic Capital And Age Of Arrival On The Standard Of Living Of Young Immigrants”. Journal of Economics 29 (2003): 27-48.

Schaafsma, Joseph, and Arthur Sweetman. “Immigrant Earnings: Age At Immigration Matters”. Canadian journal of Economics (2011): 1066-1099. Print.

Storesletten, Kjetil. “Sustaining Fiscal Policy Through Immigration”. Journal of Political Economy 108.2 (2000): 300-323. Web.

“The Fiscal Impact Of Immigration In OECD Countries”. International Migration Outlook (2013).

Zhang, Yi. “A Reconsideration Of Labor Supply Of Immigrants And Social Welfare Programs”. State Univeristy of New York at Stony Brook (2014).

Evaluating Brazil’s Federal Affirmative Action Policy

Ike Okonkwo ’18


Expanding access to Brazil’s institutions of higher education, particularly for historically underrepresented racial minorities and low-income students, has long been a topic of interest in Brazilian affairs. Although racial mixing in Brazil certainly produced a spectrum of races and phenotypes that span economic classes, racism and colorism are still present in society. It can be argued that the racial democracy in Brazil asserts the absence of racism as a byproduct of the country’s history with racial mixing1. However, evidence of racial inequality specifically in the Brazilian education sector can easily be found. In Chapter 1 of Maria Susana Arrosa Soares’, Higher Education in Brazil, Soares portrays the foundation of Brazilian colleges and their backwardness. When Brazil became independent in 1822, education was restricted to the descendants of Portuguese families. In 1931, President Getúlio Vargas created the Ministry of Education opening courses that perpetuated this exclusion in the education sector.2 Even in 1934, with the foundation of the Federal University of São Paulo (USP, widely regarded as the highest quality university in Brazil), the ‘Paulista’ elite continued having privileged access to professional courses in medicine and law3. Clearly, this complex colonial history influenced a gap in who had political power, financial stability, and job opportunities. This raises the questions: How, if at all, can we repair these inequities? More, can this be accomplished through Affirmative Action (AA) policy?

In order to make sense of Brazil’s AA policy, it is first necessary to understand Brazil’s various existing models of higher education. In this text we will focus on Brazil’s public universities, which must adhere to Brazilian federal AA law. By examining the work of Brazilian scholars like Maria Soares, we can conceptualize the various models of financing for these institutions and the implications that they pose for under-represented students. These institutions include both state and federal colleges, referred to as public universities. For state colleges, the state is the principal source of funding for the institution, while the federal colleges are maintained by the federal government and depend on its tax revenue as the primary source of income4. Since education in these public institutions is free to its students, the application process to attend them remains supremely competitive. This is not to say, however, that Brazilian private colleges are necessarily uncompetitive. In fact, Soares highlights how private universities are also growing uncontrollably each year, citing the 1,004 private institutions within Brazil serving 1.8 million undergraduates (roughly 2/3 of the country’s student population in higher education)5. Still, given that Brazilian private institutions do not offer free tuition, they are less preferred or financially inaccessible to many students, especially those from low-income households.

Current trends in Brazilian education demonstrate an increase in university applicants accompanied by increasing debate over quota policies. It is worth recognizing the increased discourse surrounding AA policy as it has turned into a highly controversial topic in Brazil. Studying the student population and implementation of AA at federal institutions helps us betterunderstand the competitive nature of Brazil’s school system as it relates to socio-economically disadvantaged students. Brazil’s Census of Higher  Education confirms a recent growth in the enrollments to its colleges stating: “the percentage of people attending university is almost 30% of the Brazilian population in the age group of 18 to 24 years, and 15% is at the age theoretically adequate to attend this level of education.”6 This information (perhaps) reflects a general distrust in the labor market such that more people feel that higher education is needed to successfully secure employment. Even distance learning courses have been increasing in annual enrollment7. The creation of more vacancies on college campuses generates expenditures that are not sustainable in the long-term, exacerbating the need to create effective access to education and to rethink plans to increase the quality of all the available institutions (both private and public). To that end, I propose to approach the topic of affirmative action in Brazil taking into account its racial and economic quotas and the ways in which they operate.


It is important to review the implementation of Brazil’s affirmative action policies in order to understand the demographic makeup of Brazil’s public colleges. In 2012, Brazil’s Federal Affirmative Action Law (Lei de Cotas) established a reserve of 50% of vacancies in federal institutions for students who had completed public high schools. Of the half of the reserved slots, 50% were earmarked for students with per capita family income of up to a minimum wage and a half.8 That is to say, a total of 25% of the university spaces were reserved for students from low-income families. In addition, the law also implemented a racial distribution across the reserved spots among Black, Mixed-Race (pardos), and Native-Indian students with respect to the proportion of ethnic groups in each state. The adoption of racial sub- quotas was the most polemical of the policy as it established an unprecedented legal process for Brazilians to identify themselves racially.9 This polarizing policy gave rise to two factions: the supporters of racial quotas and its opponents.







This infographic illuminates the distribution of spaces in Public Universities

By the same token, we must stop to address the consequences of applying to or not applying to the universities’ quota pools. For students who have entered university through the quota system, they risk being identified as ‘quota holders’ (cotistas). In recent years, quota holders have faced social challenges as negative attitudes about their status have manifested in forms of bullying including cases of physical violence used against them.11 That is to say, being a quota recipient (likely) equates to having greater chances of being beaten, isolated, or having one’s academic merits and/or ethnic identity questioned. These factors present a negative cost for minority students, discouraging their use of the quota policy. Conversely, for students who choose not to apply to the quota system, they run the risk of being less likely to receive a seat in a public university. All of these challenges to (AA) policy are significant because they limit the access to higher education, a potentially powerful agent in combating cycles of intergenerational poverty and increasing an individual’s social mobility.

In short, when doing an analysis of the quota system in Brazil it is crucial to take all these details into account alongside the students’ experiences. Therefore, we investigate the effects of the 2012 AA law. Specifically we ask: how has federal AA policy in Brazil been received by its students and how has it influenced students on a day-to-day basis? We then address which measures can be applied to improve its implementation. The data collected from this study suggests an internalized discourse regarding race and social status that reflects the need to change federal quota policies and potentially the funding of higher education in Brazil as a whole.


1. The Methods

This project used both quantitative and qualitative methods to explore the relationships between participants socioeconomic profile and their daily experiences as students of public universities. For the quantitative portion, we used SPSS to breakdown correlations between student attributes like race, family income, quota status, college, and rating of the quota system. We utilize regression analysis on these student attributes and their personal ratings on a scale of 1-5 of the quota policy (from least to most favorable), attempting to explain the variance in the students ratings. It is important to note that our quantitative methods report on a small number of participants (50 students), which could potentially invalidate the nature of such methods. However, we still find some statistically significant data that could at least justify the continuation of further studies of a larger magnitude. For the qualitative portion of our study, students wrote open-ended responses about their most present challenges (specifically on college campuses) and also had the option to participate in oral-interviews.

2. The Participants

This study was conducted with the help of 50 participants who were randomly selected from various Brazilian universities including the Federal University of Amazonas, the State University of Amazonas, the University of São Paulo, the Federal University of São Paulo, and the Federal University of Pernambuco. Participants were recruited using fliers that were posted on their respective campuses. Of the 50 participants who responded to these postings, 41 completed the questionnaire and agreed to have their answers published. Figures (1-4) depict demographic details of these participants (see appendix).

3. Instruments

The data for this study were collected through two research instruments: audio recordings of in-person interviews and an online questionnaire developed to access the students’ opinions of the quota policy.

Quantitative Results and Discussion

The following tables (1-2) depict the most relevant data on the relationship between students’ course selection and his/her attributes such as: their rating of the quota policy, level of family income, and status as a quota holder. What we can see from this information is that there is a negative and significant correlation between the courses of business administration and the ratings of the quotas. In other words, as the level of satisfaction with the quota policy increases, the number of administration students drops. Most of the administration students in this sample come from the University of São Paulo (USP) and are also responsible for the positive correlation between high-family income and enrollment in the administration course. That is to say, that with an increase in the number of USP administration students, there is an accompanying increase in family income and lower satisfaction with quotas. It is also interesting to note that a parallel trend exists with the law course, though it is not statistically significant (perhaps) due to the sample’s fewer number of law students. Our sample of quota holders relative to other students is relatively small, rightfully so, as the percentage of total quota holders overall is much smaller in many courses. In short, these correlational data may reflect a negative attitude among students in the most competitive courses (administration and law) in relation to quotas. Conversely, we find that the less competitive course, social service, possesses the opposite relationship regarding student enrollment and family income; social service also possesses the largest number of quota holders in total.


The continuation of these correlations can be seen in table two. However, table 2 depicts the relationship between university and the students’ attributes (ratings of the quota system, family income, and status as shareholder). The most relevant data here show a negative and statistically significant correlation between quota satisfaction and enrollment at USP. Again, this factor is largely due to to the administration students in the sample. More, we see a positive correlation between students attending this university and higher family income. On the other hand, we find a negative correlation between the enrollment at the University of the State of Amazonas and wealth. All of this is to say that our data confirms a relationship between the level of resources students have, what the student chooses to study, and where they choose to study.


Finally, we analyze the multi-variable regression that uses students’ ratings of the quota system as a dependent variable and their attributes (course, college, family income, etc.) as the independent variables. In other words, this regression seeks to predict how students would rate the quotas conditional on their socioeconomic background. What we found is that 57.7% of the variation found in the student responses can be explained by their socio-economic profile. However, this data needs to be seen as speculative because its ‘F-value’ remains affected by the sample size which was not significant. (see the ANOVA in the appendix – Figure # 5). Still, a similar discourse has been observed by (Bourdieu, 2007) as he summarizes his quota research stating – “these [quota] students do not choose, they are those chosen to occupy careers and courses of lesser prestige.”


a. Predictors: (Constant)

  • School, Course, Race
  • Gender, Income, Age
  • Quota Student, Entry Year

b. Dependent Variable: Quota Rating

Qualitative Results and Discussion

To answer the following short-response question: describe your college experience?, we use data collected on a scale of 1 to 5 whereby students would rank the most present challenges in their experiences. Of the 41 participants, 24 answered the first question and 29 answered the follow-up question. What the responses reveal is simply a division between quota-holders and non-quota holders. For the quota holders, the most pressing challenges they detail (in descending order) were: cost of materials/food/rent/ transportation, balancing working hours with class (many of them take night classes and work during the day), and social exclusion. In other words, most of their problems posed literal consequences for their attendance in the university. On the other hand, the challenges described by non-quota holders were not related to their permanence in the institution (e.g. securing an internship, participating in extracurricular, accessing to teachers, transferring what they learn in the classroom into tangible skills). It is interesting to note that the challenges reported by the students in the state of Amazonas were outliers. For these students, both quota holders and non-quota holders alike described facing a lack of infrastructure on their campus. This may suggest that the students’ problems are not centralized, but rather, that the challenges are nuanced and move depending on the region.

In addition, several quota students shared in the interviews a feeling of “happiness to simply attend a public university”, as they view it as a life-changing opportunity to improve their socioeconomic status. For many of these students, studying at the public university is the most constructive option in their lives. However, this attitude may come with a conundrum of encouraging these students to apply for a place in the public colleges regardless of their course of study. This may explain the tendency for quota students to look for the least competitive courses in order to better their chances of acceptance into a free institution. If these assumptions are valid, they would contribute to an internalized discourse on race and socioeconomic status that dictates that underrepresented students attend whichever courses are offered at night, whichever courses require the least materials (to minimize expenditures), and whichever course they have the best chance of gaining acceptance. In this case, if quotas are unsuccessful in changing these attitudes, then they are not effectively solving the problem of unequal access to educational resources.


The proposal of this text was not to validate or invalidate the presence of quotas but rather to examine using a microscopic-lens how students interact with the quotas, to potentially justify alternatives or alterations to existing policy. Consequently, this study was designed to access the beliefs and attitudes of Brazilian students about their experience in public as it relates to AA policy. Based on the collected qualitative data of the student narratives in interviews and questionnaires, we confirm evidence of an internalized discourse on race and socioeconomic status that dictates how these factors interact with students’ access to higher education. This discourse associates the most competitive courses of greater prestige (such as law or business administration) with higher family income. Indeed, our qualitative data implicates the presence of this narrative positively disproportionately correlating underrepresented students with less prestigious institutions/course selection. It may be that underrepresented students are subject to a selection bias whereby they are selected by courses that are less competitive and have a lower cost of materials.

Furthermore, from this small, random, sample size we can glean that the general perception of Brazilian AA policy is that it is ineffective. We have seen nuances in student accounts of the challenges they face on their college campuses and these challenges move depending on the region, this perhaps means that the problems of Brazil’s education system operate on various geographic and governmental levels. More, what we can deduce from this point is that simply reserving seats for underrepresented students without other measures is not enough to reduce inequalities in the education system. Preparing underrepresented students to receive higher level education is another important part of closing the education gap. One step in particular to do this would involve investing in the lower levels of the public school system (e.g. high schools) that most underrepresented applicants come from.

Solving inequality in the education system is not only a matter of distributing seats, but also, financial management is a factor that directly affects underrepresented students. For this reason, to address the challenges mentioned in the Brazilian education system, I propose a three- point plan that requires greater transparency in the educational budget, followed by a reassessment of the distribution of educational funds by the Ministry of Education and the implementation of a new tuition-system for Public Universities, in which students pay tuition on a sliding percentage scale based on their family income or attend free of cost if they are below a certain income level. With the remaining savings from the new tuition based school system, I would recommend reinvestment into the country’s primary and secondary education system. I would also submit that it is important to start campaigning to change the narrative on college campuses surrounding AA policy so that these policies can be better understood as inclusive efforts rather than alienating. Everyone stands to benefit from attending a more diverse campus.12 However, when quotas are casted as a socio-economic/racialized competition, underrepresented students face large social costs which may deter them from applying to particular courses. When viewed as a cut-and-dry competition, AA policy appears to polarize campuses over issues of race, class, and who does or does not deserve to attend. For this reason, the status-quo on many Brazilian campuses is tangibly hostile towards quota holding students. In the future, I propose studying the trajectory of these students, investigating their attrition rates in their programs of study to see if the cause is due to lack of scholarships/aid, lack of night school offerings, etc. I would additionally obtain a larger sample to gain better statistical analysis and further confirm or refute the notion that quota students actually face more “bullying” for their status.



Figure #1

Figure #2

Figure #3

Figure #4

Figure #5




  1. Beting, Joelmir. “Canal Livre Discute Cotas Raciais.” Canal Livre, Rede Bandeirantes, São Paulo, São Paulo, 10 Apr. 2009.
  2. Soares, Maria Susana Arrosa., and Arabela Campos. Oliven. “Educação Superior No Brasil.” Educação Superior
    No Brasil, CAPES, 2002, pp. 24-37.
  3. Soares, Maria Susana Arrosa., and Arabela Campos. Oliven., pp. 24-37.
  4. Soares, Maria Susana Arrosa., and Arabela Campos. Oliven. “Educação Superior No Brasil.” Educação Superior No Brasil, CAPES, 2002, pp. 194–218.
  5. Soares, Maria Susana Arrosa., and Arabela Campos. Oliven. “Educação Superior No Brasil.” Educação Superior No Brasil, CAPES, 2002, pp. 113–130.
  6. Instituto Nacional de Estudos e Pesquisas (INEP) Educacionais Anísio Teixeira. Censo Da Educação Superior 2013, Ministério Da Educação, 2013,
  7. INEP. Censo Da Educação Superior 2013.
  8. Beting, Joelmir. “Canal Livre Discute Cotas Raciais.”
  9. Educarargentina, director. Raça Humana. Raça Humana, YouTube, 29 Oct. 2010,
  10. Beting, Joelmir. “Canal Livre Discute Cotas Raciais.”
  11. Vermelho, Portal. “Indígenas Da UFSC Repudiam Agressão a Cotista Kaingang No Sul.” Portal Vermelho, Portal Vermelho, 29 Mar. 2016,
  12. Gurin, Patricia, Biren Ratnesh A. Nagda, and Gretchen E. Lopez. “The benefits of diversity in education for democratic citizenship.” Journal of social issues 60.1 (2004): 17-34.

Pursuing Synergy: Combining a Pan-Class PI3K Inhibitor with Novel Small Molecule Inhibitors of MYC

Scientist pipetting growth medium for human cells

Muhammed Ors ’17


One of the most common aberrations in human cancers is the overexpression of MYC, a master regulator transcription factor which functions in a heterodimer with the MYC-associated protein X (MAX) (Arvanitis & Felsher, 2006). Previously, it was shown that transduction of MYC caused resistance to PI3K inhibitor GDC-0941 (Muellner et al., 2011), and that the drug JQ1 inhibited MYC expression-induced PI3K rescue through JQ1’s inhibitory effects on the protein BRD4, which functions as a MYC enhancer (Stratikopoulos et al., 2015).

Here we use novel small molecule modulators of MYC transcriptional activity, KI-MS1- 001 and KI-MS2-008, in a comparative study across three human cell lines against the aforementioned PI3K inhibitor GDC-0941 and BET inhibitor JQ1. Co-immunoprecipitation experiments between purified Streptavidin Binding Protein (SBP) tagged MYC, the SBP tag being used to purify the MYC from whole cell lysate, and pure MAX protein in the presence and absence of the compounds imply that the novel MAX modulator KI-MS2-008 does not disrupt the MYC/MAX heterodimer binding, but data also seemed inconclusive and more work needed to be done for a definitive answer. Cell viability assays in the breast lines HCC1599 and MDA-MB-468 confirms that KI-MS1-001 and KI-MS2-008 (in addition to JQ1 and GDC-0941) reduces the viability of these cell lines, more so than these modulators have in previous studies done by the Koehler lab in other lines, but the brain line U87MG showed non convergent IC50 values for all compounds. Experiments are currently being done to determine if the compounds lower MYC and MAX protein levels in a dose dependent fashion in these cell lines. While this work does not rigorously support the hypothesis that the novel modulators would perform better than the other compounds, it contributes preliminary data for any future work the Koehler Lab will do that could more thoroughly test this hypothesis and optimize the modulators.



MYC and Justification for its Therapeutic Potential

 A transcription factor is a protein that can bind to DNA to help control the rate of transcription. Dysregulated transcription factors can lead to unregulated cell growth and gene expression and therefore is a key therapeutic target in cancer research (Darnell, 2002).

Transcription factors as a class of therapeutic targets that would be both unique enough that drugs could be designed specifically to target the transcription factor of interest, and broad enough that treatment techniques would be more than just a single-disease solution.

MYC is a master regulator transcription factor due to its pervasive necessity in transcriptional activity; MYC is part of the MYC/MAX/MAD family of basic helix-loop-helix leucine zipper (bHLHZip) domain transcription factors (Adhikary & Eilers, 2005). It functions in a heterodimer with the protein MAX, the two intrinsically disordered individual proteins undergoing a coupled folding and binding to DNA and a number of additional cofactors in order to then either activate or repress transcription (Adhikary & Eilers, 2005). As demonstrated both in Burkitt’s lymphoma and in a study of the transcriptional network of Drosophila genome, the

MYC/MAX heterodimer occupies approximately 15% of gene promoters in cells, healthy and cancerous (Li et al., 2003; Orian et al., 2003). These occupied genes are incredibly diverse: MYC regulates cellular processes such as angiogenesis, apoptosis, differentiation, growth, metabolism, proliferation, protein synthesis, ribosomal biogenesis, and self-renewal (Adhikary & Eilers, 2005; Dang, 1999; van Riggelen, Yetil, & Felsher, 2010). In any one of these fields, MYC would be taken seriously as an important protein. As a result, the overexpression of MYC is one of the most common aberrations in human cancers, having been linked to upwards of 70% of cancers (Dang, 2012). Naturally, these aberrations range greatly from uncontrolled cell proliferation and immortalization to escape from immune surveillance and growth factor independence (Vita & Henriksson, 2006). Current estimates directly attribute 100,000 annual cancer deaths in the U.S. to the deregulation of MYC activity (Boxer, 2001).

One important concern when considering to target MYC as a cancer therapy is that for the same reason that inhibiting dysregulated MYC would be widely applicable and desirable, it could also cause dangerous toxicity to the healthy cells that rely on MYC activity for normal processes (Soucek et al., 2008). Fortunately, research has shown inactivation of MYC leads to tumor regression and apoptosis, and systemic effects of MYC inhibition are well tolerated and are completely reversible (Arvanitis & Felsher, 2006; Soucek et al., 2008). Furthermore, in osteogenic sarcoma even brief inactivation of MYC was sufficient, and reactivation of MYC did not restore the cancer (Jain et al., 2002). An important consideration for why systemic MYC inhibition may not be as harmful as once feared is because at any given time the normal cells in the body are quiescent and do not express much MYC anyway; side effects may not be any worse than side effects occurring in hematopoietic and gastrointestinal tissues – cells that are known to replicate often – due to current medicines (Prochownik & Vogt, 2010).

While effects of MYC inhibition appear to be tolerable for normal cells, this is decisively not the case for cancer cells where dysregulation of MYC is a big driver in their tumorigenesis. In these cases, cancers ranging from lymphoma, leukemia, osteosarcoma, and many more, inactivation of the MYC pathway alone can lead to sustained tumor regression through proliferative arrest, apoptosis, senescence, etc. (Felsher, 2010). This phenomenon is referred to as oncogene addiction: when a tumor heavily dependent on a single oncogenic pathway dies due to inhibition of that pathway. Therefore, it is clear that identifying a small molecule that disrupts the MYC/MAX heterodimer or otherwise modulates MYC function in vivo can have immense implications in drug design, cancer treatment, and survival rates. Unfortunately, MYC historically being considered “undruggable” due to lack of typical binding pockets has stunted research on it (Darnell, 2002). However, this moniker may be more outdated than accurate: research on a variant ETS transcription factor (ETV1) that undergoes undergoes chromosomal translocation in prostate cancers and Ewing sarcomas, among other conditions, using small-molecule microarray screens found a drug-like compound named BRD32048 that binds ETV1 directly and modulates its transcriptional activity (Pop et al., 2014). Simply considering oncogenic transcription factors “undruggable” is no longer good enough.

MYC Modulators Past and Present

With the goal of finding inhibitors of the MYC/MAX heterodimer, Yin, Giap, Lazo and Prochownik screened a chemical library of 10,000 low molecular-weight compounds using a yeast two-hybrid system in which disruption of the MYC/MAX heterodimer led to a decrease in beta-galactosidase activity (Yin, Giap, Lazo, & Prochownik, 2003). From this, seven compounds were found to be specific inhibitors of the MYC/MAX heterodimer in a control yeast strain (Yin et al., 2003). Compounds from this initial set have since been used as positive controls in inhibition experiments, although synthetic chemistry work has been done to try and increase their potency; one compound in particular, 10058-F4 (IC50 = 49 µM, in HL60 cells), can inhibit growth of fibroblasts, but improved synthetic chemistry efforts from the same lab still only led an increases in potency to approximately IC50 = 20 µM (Wang et al., 2007).

The Koehler Lab has been working to discover and develop new probes with better potency or selectivity over these existing MYC/MAX modulators. Previously, Bradner, McPherson, and Koehler developed small-molecule microarrays (SMMs), glass microscope slides onto which solid substrates have been arrayed with the purpose of identifying potential small molecule modulators for transcription factors (Bradner, McPherson, & Koehler, 2006). These SMMs were developed using a novel isocyanate-based covalent attachment technique which allows for the inclusion of compounds like bioactive small molecules, FDA-approved drugs, synthetic drug-like compounds, and natural products into screens—compounds which had not been intentionally synthesized to be arrayed onto a solid substrate, but because they had certain functional groups on them, like thiols, primary alcohols, and amines, the researchers were able to adhere them to the slides (Bradner et al., 2006). Following the attachment of compounds, SMMs are incubated with pure protein of interest or cellular lysate, with protein-small molecule interactions being detected through fluorescence (Bradner et al., 2006). Using this novel SMM development method, Clemons et al. performed a screen of more than 20,000 compounds against 100 transcription factor proteins, including MYC and MAX, and identified potential small molecule binders to MYC and MAX (Clemons et al., 2010). Koehler then performed another screen against pure MYC and pure MAX of more than 40,000 compounds, including ones previously identified in the Clemons et al. screen. It is from here that the prospective MYC/MAX modulators KI-MS1 and KI-MS2 emerged, and this is currently being written into another paper.

The goal of this unbiased-binding approach was to find novel direct probes of MYC independent of it being bound to MAX, for this could provide new information to the field as all known direct probes of MYC have been found through binding against the MYC/MAX heterodimer. The 313 positive hits discovered through the SMMs were subsequently evaluated in secondary reporter gene assays in HEK293T cells; 32 of these compounds specifically inhibited MYC-dependent transcription. Currently, the Koehler Lab is characterizing the mechanisms of action of the most promising of these hit compounds using cell-free and cell- based approaches, and further chemically optimize their potency and selectivity. An ultimate goal is to use these putative MYC modulators in translational studies involving actual cancers where dysregulated MYC has a large impact and thus break the ground around exploring therapeutic modulation of an “undruggable” oncogenic transcription factors like MYC.

MYC’s Role in the PI3K/AKT pathway

 While inhibition of proteins and enzymes with roles in important kinase pathways provide attractive approaches to cancer treatment, sustaining this inhibition has proven to be a major challenge. One such pathway that develops inhibitor resistance is the PI3K/AKT/mTOR pathway, which is centered on phosphoinositide-3 kinase (PI3K) and is important in regulating processes involved in the cell cycle, proliferation, and longevity (Figure 1). Relevantly, MYC has been proven to be key player in it: during a study on the PI3K pathway using PIK3CAH1047R– initiated mammary tumors Liu et al. found that the tumors able to escape oncogene addiction had elevated levels of MYC, with knockdown of MYC reducing the incidence of recurrent tumors (Liu et al., 2011). Muellner et al. similarly observed that MYC mRNA translation and MYC transcriptional activity were amplified more often among human cell lines resistant to PI3K- mTOR inhibitors, and subsequently demonstrated that resistance to PI3K inhibitors was dependent on MYC induction (Muellner et al., 2011). Furthermore, RNA interference (RNAi) of MYC mRNA reversed the observed inhibitor resistance, showing its necessity and sufficiency for PI3K-mTOR resistance (Muellner et al., 2011).

Inspired by these results, Stratikopoulus et al. pursued a strategy to overcome PI3K inhibitor resistance using a metastatic breast cancer mouse model driven primarily by proteins PI3K and MYC. To make this model they crossed mice strains that expressed MYC mutations to strains that had either PIK3CA overexpression (H1047R mutation) or loss of PTEN expression (Ptenflox/flox) to create strains that overexpressed both MYC and PI3K, referred to as PI3K;MYC mice (Stratikopoulos et al., 2015). This model resisted a pan-class 1 PI3K inhibitor that hit both p110α and p110β subunits (GDC-0941) through feedback activation of tyrosine kinase receptors (RTKs), AKT, mTOR, and MYC. Another experiment showed their mouse model was also resistant to inhibitors of bromodomain and extra terminal domain (BET) proteins, which recognize acetylated-lysine residues in nucleosomal histones and facilitate recruitment of transcription proteins to chromatin. Stratikopoulos et al. chose to use the BET inhibitor JQ1 in order to target a member of the BET protein family BRD4, due to the fact that JQ1 had been previously shown to suppress MYC activity by downregulating MYC transcription and thereby causing potent anti-proliferative effects associated with cell cycle arrest and cellular senescence (Delmore et al., 2011; Stratikopoulos et al., 2015). Stratikopoulus et al. found that BET and PI3K inhibitors combined led to a sustained decrease of signaling of PI3K and proteins downstream of it in cell lines from the PI3K;MYC mice, causing cancer death and tumor regression both in vitro and in vivo (Stratikopoulos et al., 2015). They believe these results provide an appropriate step forward in pursuit of preventing the rescue of PI3K inhibition (Stratikopoulos et al., 2015).

Project and Hypothesis

This study seeks to capitalize on the preliminary data and approach communicated by Muellner et al. and Stratikopoulos et al. by performing an in vitro synergistic study of PI3K and BET inhibitors—but with the additional dimension of directly modulating MYC activity using the novel compounds of the Koehler Lab. The cell lines used include commercially available human cell lines corresponding to two breast cancers, HCC1599 and MDA-MB-468, and one brain cancer, U87MG. The latter two were chosen due to the loss of PTEN expression and/or mutation in PIK3CA which leads to overexpression of the PI3K pathway, and because both were present in the Stratikopoulos et al. study. HCC1559 will be a control line which does not have either loss of PTEN or a PIK3CA mutation. The compounds used are JQ1 and GDC-0941 from the previous study, and the Koehler Lab’s KI-MS1-001 and KI-MS2-008. The first aim is to determine if the MYC modulators disrupt the MYC/MAX heterodimer. The second aim is to characterize the independent IC50s of the four inhibitor compounds using cell viability assays, followed by western blotting to look for any dose-dependent decrease in MYC and p-AKT protein following incubation. The third aim, which was not completed during the thesis project period, involves determining the IC50s of the synergistic combinations of the PI3K/BET and PI3K/c- MYC inhibitors. The hypothesis is that the synergistic effects of the PI3K/MYC inhibitors will produce a lower IC50 than those of the PI3K/BET inhibitors.


Purification of SBP Tagged MYC 

We grew bacteria expressing high amounts of SBP-MYC by using a frozen glycerol stock of E. coli bacteria which had previously been transformed by Dr. Caballero with a MYC vector pET28-6xHis-SBP-TEV (Addgene). After separating the SBP-MYC soluble lysate from the insoluble pellet, my next goal was to purify the SBP-MYC from the rest of the proteins in the cell lysate, and also the MYC protein from the SBP tag. I performed a Coomassie stain and a western blot to check the efficacy of the purification approach. The purified MYC would be found in the sample named FT-TEV on the blot, the supernatant separated from the streptavidin agarose beads after incubation with the TEV Reaction Mix (Figure 2a). A sample of the flow through after the overnight incubation was saved and named FT-Beads as a way to provide a comparison point to the purity of the isolated SBP-MYC (Figure 2a). The FT-Beads sample would be used to demonstrate just how much of the SBP-MYC is left unbound to the Streptavidin agarose beads after incubation, though admittedly this is perhaps not a positive or negative control exactly—I reasoned that it would provide an internal control of the efficacy of incubation conditions. Furthermore, a second internal control was devised to look at the efficacy of the protease reaction by simply adding Laemmli running buffer to the Streptavidin agarose beads following the TEV reaction, creating a sample that would prospectively show what was left attached to the beads and was unable to be isolated through the purification (Figure 2a).

The western blot showed MYC bands around 57 kDa for all three lanes, FT-Beads, FT- TEV, and Beads, which corresponds to the molecular weight of MYC (57-60 kDa) according to anti-MYC antibody manufacturer’s information, though the Koehler Lab has many times seen MYC run closer to 50 kDa. The lightest band is seen from the FT-TEV lane, while the darkest band was from the Beads lane (Figure 2a). From these results, the greatest concentration of MYC remained attached to the streptavidin agarose beads instead of being properly cleaved by the AcTEV Protease reaction. In order to produce cleaner and better quality gels to look at I performed these experiments twice, and the second run provided the images used for Figure 2a. However, the results did not change between these two runs and there was not enough time to try again. Considering that cleaving the SBP tag was not absolutely necessary to use the protein, I decided after talking to Dr. Caballero to change the original plan for the rest of the experiments and simply perform them with the SBP-MYC.

Novel Compounds Do Not Disrupt MYC/MAX heterodimer

 The purpose of these experiments was to ask whether the modulators disrupt the MYC/MAX heterodimer binding. To that end, I chose to look at this question in vitro through co- immunoprecipitation assays between MYC and MAX in the presence and absence of modulator compounds, using the SBP-MYC lysate produced previously as my source of MYC. This first set of assays use the positive MYC/MAX heterodimer inhibitors 10074-G5 or KJ-Pyr-9 as the compounds modulating the interaction between SBP-MYC and MAX; the purpose of this set of assays was to behave as a large control for all of the co-immunoprecipitation assays that were to follow. Within this assay protocol, the negative control was a sample prepared without SBP- MYC, to determine if MAX would be pulled down by the streptavidin agarose beads in the absence of SBP-MYC. The positive control tubes were prepared with increasing concentrations of the MYC/MAX heterodimer inhibitors 10074-G5 (25 µM, 50 µM, and 100 µM) or KJ-Pyr-9 (10 µM, 15 µM, 20 µM). Additionally, all samples were kept at a final concentration of 1% DMSO to be consistent with conditions, as this is what the drug compound stock solutions had been diluted in.

Blots were visualized through chemiluminescence using the ChemiDoc MP Imaging System machine (Bio-Rad) and they were further analyzed using the Image Lab program (Bio- Rad) in order to ascertain the relative intensities of the visible bands (Figure 3). Bands for MYC were seen in all lanes but the lane where there was no SBP-MYC lysate added, as expected (Figure 2b). However, amongst the bands for MAX while there seemed to be a decrease in MAX protein present as 10074-G5 increases in concentration, I was surprised to see a band for MAX coming from the MYC-absent sample and no band from the sample incubated without the inhibitor compound (Figure 2b). The Image Lab program analysis of band intensity confirmed that while the MAX protein concentration was decreasing as 10074-G5 concentration was increasing from 25 µM to 100 µM, as evidenced by the decreasing ratio of MAX/MYC present in each lane, this was not the case when the concentration of 10074-G5 was 0 µM (Figure 3b).

Considering the unexpected developments encountered here, I performed technical replicates with the same lysate and biological replicates with another round of purified SBP-MYC lysate. However, I was unable to get data that would either replicate what was seen in Figure 2b or contradict it—all of the other gels were very blotchy with undecipherable bands. Therefore, I have chosen to include the clearest looking image in this thesis (Figure 2b).

Following co-immunoprecipitations done with incubation with the positive control compounds, the next step would be to use the small molecule modulators developed in lab, currently designated as KI-MS1-001 and KI-MS2-008. This would have been done to test the hypothesis that these too would inhibit the MYC/MAX heterodimer. Unfortunately, due to time constraints and concerns about moving onto other experiments for the thesis, I could not personally test this hypothesis. Luckily, my mentor Andrew Chen was already working on this question independently and allowed me to include his results here to serve as a comparison to the previous step (Figure 2c). Within this assay protocol there were two negative controls in a sense: a sample prepared with only DMSO, without either anti-MAX (C-17) antibody or KI-MS2-008, and secondly, for each condition tested three washes were also run as an internal control, as there should have been no proteins of interest visible from these wash steps. The incubation condition of anti-MAX (C-17) antibody without KI-MS2-008 would be the positive control, as the antibody should bind to MAX which would in turn pull down its heterodimeric partner MYC.

The blot was visualized through chemiluminescence using the ChemiDoc MP Imaging System machine (Bio-Rad). No bands were visible for the flow through (1% loaded) or wash (5% loaded) lanes, but there was a faint band for MYC immunoprecipitated from the control sample incubated with normal IgG rabbit antibody—likely due to non-specific binding (Figure 2c). Most importantly, KI-MS2-008 did not seem to significantly impact amount of MAX pulled down by MYC, suggesting it does not interfere with in vitro MYC-MAX binding (Figure 2c).

KI-MS1-001 and KI-MS2-008 Reduce Cancer Cell Viability

The human cancer cell lines used to perform the cell viability assays were two breast cancer cell lines, HCC1599 (suspension) and MDA-MB-468 (adherent), and one brain cell line, U87MG (adherent). U87MG and MDA-MB-468 both have loss of PTEN expression and/or mutations in PIK3CA which leads to overexpression of the PI3K pathway, and additionally both were present in the Stratikopoulos et al. study in a limited manner. HCC1599 has neither loss of PTEN nor a PIK3CA mutation and is meant to be a control line. These cell lines were chosen because they were all readily available from the MIT Koch Institute’s high throughput screening facility. The purpose in doing cell viability assays was to characterize the independent IC50s of the pan-class I PI3K inhibitor GDC-0941, the BET inhibitor JQ1, and the two MYC and MAX modulators KI-MS1-001 and KI-MS2-008.

In the first biological replicate all four drug compounds were serially diluted two-fold in DMSO from 40 µM to 0.156 µM (results not included in this manuscript). The second and third replicates were further serially diluted from 40 µM to 0.00122 µM due to a desire to better visualize a sinusoidal curve to calculate IC50s from. Cell lines incubated in compounds or DMSO and for both 3 days and 5 days, after which that plate would be visualized. These time points parallel those tested in the Stratikopoulos et al. paper. CellTiter-Glo was used to measure the amount of ATP present in the wells of the plates through luminescence, as these values are considered to be a standard indicator of metabolically active cells, and therefore a reasonable method to draw conclusions on cell viability (Riss, Moravec, & Niles, 2011).

Luminescence values were graphed, and IC50 values calculated, using Prism software (Graphpad). I performed three biological replicates and three technical replicates for each of the conditions tested. The technical replicates are accounted for through the error bars appearing on the graphs, and I have chosen to include both the second and third replicates as they each give interesting and partially conflicting information about the cell lines.

In the HCC1599 cell line all of the drug compounds seemed to be capable of reducing cell viability, with JQ1 (Figures 4a and S4a) and GDC-0941 (Figures 4b and S4b) consistently having IC50s less than 1 µM, while KI-MS1-001 (Figures 4c and S4c) and KI-MS2-008 (Figures 4d and S4d) had higher IC50s, between 4 µM and 14 µM on Day 5. All drug compounds displayed lower IC50 values on Day 5 in comparison to Day 3, which would be expected; JQ1 seemed to be the most potent out of all four, going as low as 0.002 µM (Figure S4a). In the U87MG cell line, none of the drug compounds were particularly effective, with almost all of the IC50 values for the drugs being greater than 30 µM (Figures 5 and S5), and JQ1 even having an IC50 value that actually went up from 24 µM on Day 3 to 32 µM Day 5 (Figure 5a). Interestingly, in the third biological replicate while the Day 5 IC50 values for JQ1, GDC-0941, and KI-MS2- 008 (Figures S5a, S5b, S5d) could not be determined because the graphs did not converge, KI- MS1-001 alone had an IC50 value of 3.7 µM at Day 5 (Figure S5c). The data from the MDA- MB-468 cell line were especially interesting as KI-MS1-001 and KI-MS2-008 showed very different IC50s between the second and third biological replicates. All drug compounds had IC50s around 1 µM in the second replicate (Figure 6) but in the third replicate KI-MS1-001 went up to 5.5 µM (Figure S6c) and KI-MS2-008 jumped up to 30.5 µM (Figure S6d).

Following testing of cell viability under compound and DMSO conditions, HCC1599 and MDA-MB-468 were also prepared for western blotting against MYC, p-AKT, and GAPDH in order to test for potential dose-dependent differences in protein expression. I chose to incubate the cells for 24 hours because that was the time frame the Stratikopoulus et al. paper used and I wanted to be as comparable as possible. I chose to incubate with two-fold serial dilutions from 10 µM to 0.156 µM because the measured IC50s seemed to consistently fall within that range.

Total protein concentration was determined using a BCA assay in order to be able to prepare samples for the gel to have a consistent 30 µg concentration of protein, otherwise the comparison of bands would not be particularly meaningful. However, by the time the project ended I had been unable to produce any western blot data that showing the MYC protein present in the lysate.



MYC is indisputably a genetic heavy weight. Studies in both Burkitt’s lymphoma and the Drosophila genome have shown that the MYC/MAX heterodimer occupies approximately 15% of gene promoters in cells (Li et al., 2003; Orian et al., 2003), affecting processes including angiogenesis, apoptosis, differentiation, protein synthesis, and many others (Adhikary & Eilers, 2005; Dang, 1999; van Riggelen et al., 2010). Furthermore, overexpression of MYC has been linked to upwards of 70% of cancers (Dang, 2012) with estimated 100,000 annual U.S. cancer deaths directly caused due to deregulation of MYC activity (Boxer, 2001). Most importantly, we know that inactivation of MYC can lead to tumor regression and apoptosis without adversely affecting normal tissue (Arvanitis & Felsher, 2006; Soucek et al., 2008). It seems abundantly clear that identifying a small molecule disrupting the MYC/MAX heterodimer or modulating c- MYC function in vivo should be very high on the list of anyone who wants to “cure” cancer.

The fact that MYC has been considered “undruggable” for so long due to its lack of typical binding pockets is a golden opportunity—whoever creates a viable treatment targeting the protein will surely achieve greatness. It is through the Koehler Lab, and the novel small molecule MYC and MAX modulators being developed in-house, that I have had the opportunity to contribute to this cause. When considering my own thesis I was inspired by the papers written by Muellner et al. and Stratikopoulos et al. examining inhibitor resistance in the PI3K/AKT/mTOR pathway. Muellner et al. had demonstrated that resistance to PI3K inhibitors was dependent on c- MYC induction, which was reversible through using RNA interference (RNAi) of MYC mRNA (Muellner et al., 2011). Stratikopoulus et al. examined PI3K inhibitor resistance using a metastatic breast cancer mouse model driven primarily by either PIK3CA overexpression or loss of PTEN, finding it was resistant to both a pan-class 1 PI3K inhibitor GDC-0941 and BET inhibitor JQ1, but not both together (Stratikopoulos et al., 2015). Looking at these results I had to ask: what about a combinatorial study using the PI3K inhibitor GDC-0941 in conjunction with the Koehler Lab’s MYC modulators instead of the BET inhibitor JQ1? KI-MS1-001 and KI- MS2-008 would affect MYC/MAX’s transcriptional ability while JQ1 would affect the ability for MYC to even be transcribed, so I thought there might perhaps be interesting things to be learned from comparing these two approaches.

As the first step in pursuit of my thesis, I wanted to examine whether the novel small molecule modulators behaved as heterodimer inhibitors. This was an ongoing question in the Koehler Lab at the time, so that we better understood the mechanism of action of these modulators to put their effects in the cell lines in context. I began this step by growing E. coli bacteria overexpressing SBP-tagged MYC protein and working to purify the protein from the rest of the lysate, performing a western blot to confirm MYC was being produced and was purified properly. However, as mentioned earlier, the lightest band is seen from the FT-TEV lane while the darkest band was from the Beads lane, the opposite of what I would have wanted (Figure 2a). Although I was able to purify MYC from the rest of the lysate and the SBP-tag as determined from the presence of the FT-TEV band, an ideal purification would have had a much stronger band for MYC in that lane, and weaker bands in both of the other lanes. Especially the Beads lane made from boiling the leftover Streptavidin agarose beads in Laemmli running buffer, as the only thing remaining on these beads should have been the SBP-TEV tag and not MYC.

One explanation why the western blots were against expectations could be because the AcTEV Protease was older than believed and degrading, and thus wouldn’t be able to catalyze the cleavage reaction as well explaining the little MYC in the FT-TEV. Another reason could be because I changed the protocol after the first attempt to purify came out very poorly from the optimal incubation conditions of 3 hours in a 37° C water bath to an overnight incubation at 4° C. Furthermore, I didn’t think to wash the beads after removing the FT-TEV and perhaps there was a lot of cleaved MYC that was just stuck in the beads considering MYC is a very sticky protein. Repeating these purification experiments with new AcTEV Protease would be appropriate, but I decided at Dr. Caballero’s recommendation to simply continue to use the SBP- MYC in further experiments as ultimately the SBP tag is small and would not interfere with the MYC/MAX binding interaction. In addition, it is clear that there was overflow in the lanes of the western blot due to my error in loading them, and a repeated effort should load less sample. Any future attempt to purify SBP-MYC would take these issues into consideration.

To examine whether KI-MS1-001 or KI-MS2-008 inhibited the MYC/MAX heterodimer I first performed co-immunoprecipitation experiments between purified SBP-MYC and pure MAX protein in the presence and absence of the drug compound 10074-G5 to establish data for a positive control (Figure 2b). I expected that increasing concentration of the positive control would lead to decreasing concentration of MAX binding to the SBP-MYC, but this turned out to be harder to prove than anyone had reasonably considered. Generally, an increasing concentration of 10074-G5 caused a progressively lesser concentration of MAX to bind to the SBP-MYC (Figure 3). Unfortunately, there was an extremely small amount of MAX binding to the SBP-MYC in the control sample to which no 10074-G5 was added (Figure 3). Considering that the control within this blot clearly didn’t work, I performed three more co-Immunoprecipitations with 10074-G5 and two with PY9, another positive control inhibitor—and yet, all of these blots were very blotchy so I was still unable to either replicate what was seen in Figure 2b or contradict it.

There are a couple of reasons that could explain why these western blots were unsuccessful. One reason could be that 100 µL of streptavidin agarose beads was insufficient to incubate with, although this amount had been standard in the lab. Another problem could have been that incubating the five samples for 1.5 hours at room temperature, prior to adding MAX and after, was not long enough for meaningful binding. A third reason for these difficulties could be due to the binding buffer solution used in incubation steps being a buffer solution from an EMSA kit, potentially causing issues in a co-immunoprecipitation; the choice to use this buffer solution had been made because the original plan was to later perform an EMSA testing whether the positive control inhibitor, KI-MS1-001, or KI-MS2-008 affected MYC/MAX binding to DNA, and I had been advised to prepare these Western blots using similar conditions to what would come later. This preparation proved completely unnecessary as I did not end up having time to perform an EMSA. Either way, I feel that these positive control data are inconclusive as I was ultimately unable to properly replicate results in readable gels.

Following the co-immunoprecipitation with the positive control inhibitors, the plan was to test whether the novel small molecule modulators inhibit the MYC/MAX heterodimer by incubating a solution of purified MAX and SBP-MYC with increasing concentrations of both KI- MS1-001 and KI-MS2-008. Unfortunately, I was also unable to perform this due to time constraints. Therefore, I have instead included in this study co-immunoprecipitation experiments that my second mentor Andrew Chen performed with KI-MS2-008 to use as a comparison to my experiment (Figure 2c). As expected, no bands were seen at all for the flow through or wash lanes, and there was only a light band in the control lane (Figure 2c). I was a bit surprised to see bands for MYC and MAX coming from the control sample incubated with normal IgG rabbit antibody, but this is most likely due to non-specific binding as MYC is known to be a sticky protein (Figure 2c). Most importantly, KI-MS2-008 did not seem to significantly impact the amount of MAX pulled down by MYC which suggests that it in fact does not interfere with in vitro MYC/MAX binding (Figure 2c). In the years following these experiments the Koehler Lab has also compiled more evidence to support this case.

I recognize that a couple key conditions are different between the two protocol designs and should be addressed, but I do not think any of these distinctions would disqualify the data from being a useful comparison: firstly, my experiment was performed using SBP-MYC purified from E. coli lysate while Andrew Chen’s was performed using normal MYC expressed in the Burkitt’s lymphoma cell line P493-6; secondly, my experiment used streptavidin agarose beads while Andrew Chen’s used magnetic Dynabeads; and thirdly, I incubated using EMSA kit binding buffer and Andrew Chen incubated using the Dynabead kit binding buffer. Given time, I would have liked to perform the ideal co-immunoprecipitation and EMSA experiments with both KI-MS1-001 and KI-MS2-008 under similar conditions to the positive control experiment.

Following this first goal, the second goal of my thesis was to characterize the independent IC50s of the four drug compounds in cell viability assays. Given limited time, I chose cell lines different along multiple axes to amplify information gained i.e. the cell lines that overexpressed the PI3K pathway were in different cancer types, breast (MDA-MB-468) and brain (U87MG), and the third cell line was a control with no overexpression of the PI3K pathway (HCC1599). I reasoned that without hard independent data in these three cell lines with these four drugs (pan-class I PI3K inhibitor GDC-0941, BET inhibitor JQ1, and the two MYC and MAX modulators KI-MS1-001 and KI-MS2-008), it would be unclear later on whether data produced through a characterization of a synergistic drug combination was actually causing synergistic effects or merely additive ones. Due to a desire to gain more data points from which to better visualize a sinusoidal curve and calculate IC50s of the drug compounds, I chose to perform two-fold serial dilutions from 40 µM to .156 µM and then further to 0.00122 µM.

Interestingly, all of the drug compounds seemed to be capable of reducing cell viability in the HCC15599 cell line: JQ1 (Figures 4a and S4a) and GDC-0941 (Figures 4b and S4b) consistently had IC50s less than 1 µM, while KI-MS1-001 (Figures 4c and S4c) and KI-MS2-008 (Figures 4d and S4d) had higher IC50s, between 4 µM and 14 µM. This was against expectations, as the HCC1599 cell line was expressly chosen because it has neither loss of PTEN nor a PIK3CA mutation. Instead, the cell line is known to be negative for expression of Her2- Neu, p53, estrogen receptor (ER), progesterone receptor (PR). Perhaps the results here, that drugs prospectively meant for the PI3K pathway and MYC/MAX modulators affect the viability of triple negative breast cancer, are the most unexpected gains from my thesis and could be further pursued by the Koehler Lab outside of this study. In the U87MG cell line, it did not seem like any of the drug compounds were particularly effective at all, which combined with the HCC1599 data possibly suggests that there is more of an importance placed on the tissue than the specific mutations than I expected (Figures 5 and S5). Finally, the data from the MDA-MB-468 cell line showed internal contradictions, with all drug compounds having IC50s around 1 µM in the second replicate (Figures 6), but KI-MS1-001 and KI-MS2-008 jumping up in the third replicate (Figures S6c and S6d). One commonly used MYC/MAX heterodimer inhibitor compound that was discovered in the same study as the 10074-G5 used in this thesis is 10058-F4 (Yin et al., 2003). This compound initially was recorded by Yin et al. as having an IC50 of 49 µM in HL60 cells, but improved synthetic chemistry efforts from the same lab still only led an increases in potency to approximately IC50 = 20 µM (Wang et al., 2007; Yin et al., 2003). I would want to repeat the cell viability assays a few more times in MDA-MB-468 to look for more consistency in IC50 values due to the internal contradictions present between the replicates, but considering the data from the Wang et al. paper I think KI-MS1-001 and KI-MS2-008 show promise.

During this stage I took care to do both three biological and technical replicates in order to be most thorough, and naturally there were plenty of opportunities for error when collecting data. The potential for human error whenever I was plating drugs, or media, or cells into the hundreds and hundreds of wells over the course of the months is a particularly large problem. A lot of the variation between technical replicates could probably be explained to a deficiency in technique, especially when I was first performing the assays. Performing these experiments with a robotic system would be a good way to avoid this source of error. Additionally, the western blotting to see if there were dose-dependent decreasing concentrations of the MYC and p-AKT proteins as a result of increasing concentrations of the drug compounds also did not work well.

This is perhaps due to the HCC1599 and MDA-MB-468 cell lines not producing enough MYC considering that they are not known to overexpress that protein to begin with. In the future it would be prudent to first determine exactly how concentrated the total protein in the lysate from these cell lines must be to get a detectable band in western blotting.

The third aim had been to bring the thesis study to a close by characterizing the IC50s of the synergistic combinations of the PI3K/BET inhibitors and PI3K/MYC inhibitors, finally testing the hypothesis that the synergistic effects of the PI3K and MYC inhibitors would have a stronger impact on the cell lines and produce a lower IC50 value than the effects of the PI3K and BET inhibitors. As should be clear at this point, I unfortunately wasn’t actually able to perform either a combinatorial or synergistic study of the PI3K inhibitor and MYC/MAX modulators.

Therefore, the future directions are clear. Stratikopoulos et al. had demonstrated the BET inhibitors blocked PI3K inhibition resistance caused by exposure to GDC-0941, and PI3K/BET inhibition blocked AKT reactivation (Stratikopoulos et al., 2015). They showed this through in vitro proliferation assays performed while holding one of the compounds at a constant 1 µM and adding to it a serial dilution from 8 µM to .125 µM of the other compound. The Koehler Lab plans on finishing my line of questioning by performing cell viability assays using these cancer cell lines; however instead of holding one compound at a constant 1 µM the plan is to vary the concentrations of both compounds along a two-fold serial dilution from 40 µM to .156 µM, in order to more precisely define where the best synergy is.

Through these future studies, we aim to present the novel MYC and MAX modulator compounds KI-MS1-001 and KI-MS2-008 as viable options to put into the cancer researcher’s toolkit. As dysregulated transcription factors become increasingly attractive targets for cancer research, developing a small molecule modulator of MYC or MAX function in vivo will only increase in importance as well (Darnell, 2002). In particular, evidence demonstrating elevated levels of MYC increases tumor recurrence in PIK3CAH1047R-driven mammary cancers show that MYC is an important oncoprotein that cannot be overlooked even when attempting to study a seemingly unrelated cancer driver (Liu et al., 2011). The clinical significance of MYC cannot be underestimated, especially in consideration of the work done by Stratikopoulus et al. and Liu et al. in showing that a single PI3K inhibitor isn’t enough to use as a treatment for a cancer of the PI3K/AKT/mTOR pathway. Even if the planned synergistic study here does not show that the modulators in their current chemical forms produce better results than the combination of JQ1 and GDC-0941, this wouldn’t be a reason to give up on this line of reasoning. These compounds are continually being improved and optimized by the Koehler Lab with the belief that the failures of today will inform the successes of tomorrow. The main goal is not an immediately translatable treatment, but a solid foundation from which further research and treatments can be built upon.






Bacterial Expression of SBP tagged MYC

SBP-MYC was purified from E. coli bacteria previously transformed by Dr. Caballero with a MYC vector pET28-6xHis-SBP-TEV (Addgene). A frozen glycerol stock of this bacteria was thawed and incubated overnight at 37° C using 25 mg/mL Kanamycin (ThermoFisher) and 25 mg/mL Chloramphenicol (ThermoFisher) in Lysogeny Broth media (Boston Bio Products). The next day a sample of this was incubated in a 500 mL culture of media at the same temperature and same antibiotic concentration. When the OD600 of the bacteria was measured to be between 0.5-1, and thus in the exponential growth stage, bacterial growth was slowed on ice and expression of MYC with a streptavidin binding peptide (SBP) and Tobacco Etch Virus (TEV) tag attached was induced using 0.5 mM IPTG. After the bacteria culture was incubated overnight at room temperature the culture was centrifuged for 30 minutes at 3000 rpm at 4° C. The pelleted bacteria was resuspended in 20 mL of lysis buffer, a solution composed of 20 mL CelLytic B (Sigma-Aldrich), 0.8 mL Lysozyme (Sigma-Aldrich), 2.6 µL benzonase (Sigma- Aldrich), and 2 tablets of protease inhibitor cocktail (Roche). The lysate was incubated for 30 minutes at 37° C before being centrifuged for 15 minutes at 12,000 g at 4° C, discarding the pellet and storing the supernatant at -80° C for use in both purification and co- immunoprecipitation experiments.


Purification of SBP tagged MYC

300 µL of streptavidin agarose beads (ThermoScientific) were washed twice with cold PBS, centrifuging for 4 minutes at 200 g at room temperature. The streptavidin agarose beads were incubated overnight at 4° C with 10 mL of the previously collected lysate. After this incubation the flow through was stored, henceforth referred to as FT-Beads. The beads remaining in the tube were then washed twice again with cold PBS. To separate the MYC from the SBP-TEV tag and have pure MYC, AcTEV Protease (Invitrogem) was used to bind the TEV part and then cleave MYC from the SBP-TEV tag. The 300 µL of streptavidin agarose beads were incubated overnight at 4° C with a Reaction Mix solution composed of 30 µL of 20X TEV buffer, 6 µL of 0.1 M DTT, 4 µL of AcTEV Protease, and 600 µL of pure water. After incubation with the TEV Reaction Mix, the streptavidin agarose beads were centrifuged for 5 minutes at 200 g at 4° C, and the supernatant was then separately stored, henceforth referred to as FT-TEV. Finally, 300 µL of 2x Laemmli sample running buffer (Bio-Rad) was added to the beads and the tubes were boiled for ten minutes at 90° C, and centrifuged for 5 minutes at 12000 g at room temperature.

20 µL of the Laemmli buffer was added to 20 µL samples of the FT-Beads and FT-TEV and these were boiled for ten minutes at 90° C, centrifuged for 5 minutes at 12000 g at room temperature. These samples, in combination with a 40 µL sample of the boiled beads mentioned above, were run on 4-15% polyacrylamide gels (Bio-Rad) in 1X Tris/glycine/SDS buffer at 200 V for about 45 minutes. One gel was analyzed by Coomassie stain, and the other gel was transferred to a PVDF membrane (Bio-Rad) and analyzed by western blot using a 1:500 dilution of anti-MYC (9E10) mouse monoclonal antibody (Santa Cruz Biotech) and a 1:2000 dilution of anti-mouse monoclonal antibody (Santa Cruz Biotech). Blots were developed using a 1:1 solution of the two West Pico Stablizer/Enhancer solutions (ThermoScientific) and visualized through chemiluminescence using the ChemiDoc MP Imaging System machine (Bio-Rad).


Co-immunoprecipitation of SBP-MYC and MAX with 10074-G5

Co-immunoprecipitation assays were performed to look at the ability of SBP-MYC to bind to and pull down MAX in vitro while under the condition of being incubated with modulators of this interaction. This first set of assays use the positive MYC/MAX heterodimer inhibitors 10074- G5 (Selleck Chem) or PY9 (Sigma-Aldrich) as the compounds modulating the interaction between SBP-MYC and MAX. A binding buffer solution was produced that was composed of pure water, 10x Binding buffer (EMSA kit, ThermoScientific) diluted to a 1x final concentration, 50% glycerol diluted to a 2.5% final concentration, 100 mM MgCl2 diluted to a 5 mM final concentration, 1% NP-40 diluted to a 0.05% final concentration. This solution is henceforth referred to as simply binding buffer. 1 mL of Streptavidin agarose beads (ThermoScientific) were washed twice with cold PBS, centrifuging for 5 minutes at 200 g at 4° C. 100 µL of these washed beads were aliquoted into 4 tubes to be incubated with 100 µL of the SBP-MYC lysate previously frozen and 800 µL of the binding buffer. Another tube was prepared to be the negative control, containing only the beads and the binding buffer, no SBP-MYC. All 5 of these were incubated for 1.5 hours at room temperature and then centrifuged for 3 minutes at 300 g at 4° C, removing the supernatant. To all of these tubes 1 µL of pure His-tagged MAX protein (Abcam) and 889 µL of the binding buffer was added. The positive control tubes additionally had increasing amount of MYC/MAX heterodimer inhibitors 10074-G5 (25 µM, 50 µM, and 100 µM) or PY9 (10 µM, 15 µM, 20 µM) added to the solution, whereas the negative control and test tubes just had a final concentration of 1% DMSO, without additional drug. Again all 5 of these were incubated for 1.5 hours at room temperature and then centrifuged for 3 minutes at 300 g at 4° C, removing the supernatant. For the last time, the streptavidin agarose beads were washed twice with cold PBS.

100 µL of the Laemmli buffer was added to 100 µL samples of the beads and boiled for ten minutes at 90° C, centrifuged for 5 minutes at 12000 g at room temperature. 30 µL of each sample were run on 4-15% polyacrylamide gels (Bio-Rad) in 1X Tris/glycine/SDS buffer at 200 V for about 45 minutes. The gel was transferred to a PVDF membrane (Bio-Rad) and analyzed by western blot using a 1:500 dilution of anti-MYC (9E10) mouse monoclonal antibody (Santa Cruz Biotech), a 1:500 dilution of anti-MAX (C-124) rabbit monoclonal antibody (Santa Cruz Biotech), and 1:2000 dilutions of anti-mouse and anti-rabbit monoclonal antibodies (Santa Cruz Biotech). Blots were developed using a 1:1 solution of the two West Pico Stabilizer/Enhancer solutions (ThermoScientific) and visualized through chemiluminescence using the ChemiDoc MP Imaging System machine (Bio-Rad).

Co-immunoprecipitation of MYC and MAX from P493-6 cell lysates with KI-MS2-008

The protocols and experiments described in this section were all performed by my mentor, graduate student Andrew Chen. P493-6, a Burkitt’s lymphoma cell line which is engineered to express high levels of MYC in the absence of doxycycline, was cultured. 400 million P493-6 cells, as counted by the Cellometer Auto T4 Bright Field Cell Counter (Nexcelom Bioscience), were pooled and washed twice in cold PBS. Cells were resuspended in 10 mL of modified RIPA lysis buffer with 0.1% sodium deoxycholate, 1% NP-40, 0.004% buffer benzonase, 200 mM NaCl, 50 mM Tris, 1X protease inhibitor (Sigma-Aldrich), and 1X phosphatase inhibitor (Roche). After incubating on ice for 15 minutes, the cells were centrifuged for 10 minutes at 14000 g at 4° C.

Following this a Bradford assay was performed to determine the total protein concentration of the cell lysate (the supernatant from the previous step), and 4 mg of total protein was incubated for thirty minutes with either 2 µL DMSO or 20 µM KI-MS2-008 and topped off to a total volume of 1 mL with Ab Binding and Washing Buffer from the magnetic Dynabeads kit (ThermoScientific). Then the protein/compound mixtures were incubated overnight at 4° C with either 20 µL MAX (C-17) rabbit antibody (Santa Cruz Biotech) or 10 µL of normal rabbit IgG (Santa Cruz Biotech). 50 µL of the magnetic Dynabeads per tube were placed on a magnetic stand for a minute to settle the beads, and washed in 200 µL of the Ab Binding and Washing Buffer before transferring the protein/compound/antibody mixture prepared the night prior into the tubes. These new tubes were incubated for one hour at 4° C and then placed on the magnetic stand again to settle the beads, saving the flow through to run on the gel. This was washed three times using 200 µL of the Washing Buffer, saving the washes for the gel as well.

30 µL of 2X SDS buffer was added to Dynabead tubes and boiled for five minutes at 90° C, using the magnetic stand to settle the beads and separate the elution. 20 µL of each sample were run on 4-15% polyacrylamide gels (Bio-Rad) in 1X Tris/glycine/SDS buffer at 100 V for 10 minutes, followed by 200 V for about 35 minutes. The gel was transferred to a PVDF membrane (Bio-Rad) and analyzed by western blot using a 1:500 dilution of anti-MYC (9E10) mouse monoclonal antibody (Santa Cruz Biotech), a 1:500 dilution of anti-MAX (H-2) mouse monoclonal antibody (Santa Cruz Biotech), and 1:2000 dilutions of anti-mouse monoclonal antibodies (Santa Cruz Biotech). Blots were developed using a 1:1 solution of the two West Pico Stabilizer/Enhancer solutions (ThermoScientific) and visualized through chemiluminescence using the ChemiDoc MP Imaging System machine (Bio-Rad).


Cell Viability Assays

All cell lines were obtained from the MIT Koch Institute’s High Throughput Screening facility. The suspension human breast cancer cell line HCC1599 was cultured in Roswell Park Memorial Institute (RPMI) 1640 Medium (ThermoFisher), the adherent human breast cancer cell line MDA-MB-468 was cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (ThermoFisher), and the adherent human brain cancer cell line U87MG was cultured in Eagle’s Minimum Essential Medium (EMEM) (ATCC). All culture media was prepared with 10% Fetal Bovine Serum (FBS) (ATCC) and 1% Penicillin/Streptomycin (P/S) (Corning Incorporated).

Cell cultures were grown using an incubator kept at 37° C at 5% CO2 and cultures were split every 3 to 5 days, as recommended according to the manufacturer’s instructions. The compounds used are BET inhibitor JQ1 (Sigma-Aldrich) and PI3K inhibitor GDC-0941 (Selleck Chem), and KI-MS1-001 and KI-MS2-008 which are being produced internally at the Koehler Lab. All four were serially diluted two-fold in DMSO, from 40 µM to 0.156 µM and later from 40 µM to 0.00122 µM, while being kept at a final concentration of 0.4% DMSO.

50 µL of cells were plated in sterile 96 well plates (Corning Incorporated) at a density of 5,000 cells per well, two plates per condition. Suspension cells were directly treated with 50 µL of either JQ1, GDC-0941, KI-MS1-001, KI-MS2-008 or DMSO and placed into the incubator, while adherent cells were first incubated overnight before replacing the culture media and then treating them with compound or DMSO. Each cell line was incubated in compounds or DMSO for both 3 days and 5 days, time points after which a plate would be visualized. 100 µL of CellTiter-Glo reagent (Promega) was added to each well, and plates were incubated for 10 minutes at room temperature on a platform shaker. Luminescence was recorded using a Tecan Infinite 200 Pro spectrophotometer using an integration time of 0.5 second/well. Luminescence values were graphed, and IC50 values calculated, using Prism software (Graphpad).

Following testing of cell viability under compound and DMSO conditions, cells were  also prepared for western blotting against MYC, p-AKT, and GAPDH. Using sterile 12 well plates (Greiner Bio-One) for the HCC1599 suspension cells and sterile 6 well plates (VWR) for the MDA-MB-468 adherent cells, 3 million cells from each cell line were incubated for 24 hours with two-fold serial dilutions, from 10 µM to 0.156 µM, of the four compounds at 0.4% DMSO. Suspension cells were directly treated with the compounds while adherent cells were first incubated overnight in culture media before being treated with compounds. Cells were pelleted and washed with cold PBS, centrifuging for 5 minutes at 200 g at 4° C. Then cells were resuspended in 60 µL of modified RIPA lysis buffer with 0.1% sodium deoxycholate, 1% NP-40, 0.004% buffer benzonase, 200 mM NaCl, 50 mM Tris, 1X protease inhibitor (Sigma-Aldrich), and 1X phosphatase inhibitor (Roche). Total protein concentration was determined using a BCA assay, and samples meant to run on the gel were either prepared to have a consistent 30 µg concentration of protein or 50 µg, depending on the total protein concentration from the cell line.

Enough 2X SDS buffer was added to samples for a 1:1 ratio of sample and running buffer, and samples were boiled for five minutes at 90° C, centrifuged for 5 minutes at 12000 g at room temperature. Samples were run on 4-15% polyacrylamide gels (Bio-Rad) in 1X Tris/glycine/SDS buffer at 200 V for about 45 minutes. The gel was transferred to a PVDF membrane (Bio-Rad) and analyzed by western blot using a 1:500 dilution of anti-MYC (9E10) mouse monoclonal antibody (Santa Cruz Biotech), a 1:500 dilution of anti-MAX (C-124) rabbit monoclonal antibody (Santa Cruz Biotech), a 1:1000 dilution of anti-p-AKT (T308) rabbit monoclonal antibody (Cell Signaling), a 1:1000 dilution of anti-GAPDH (14C10) rabbit monoclonal antibody (Cell Signaling), and 1:2000 dilutions of anti-mouse and anti-rabbit monoclonal antibodies (Santa Cruz Biotech). Blots were developed using a 1:1 solution of the two West Pico Stabilizer/Enhancer solutions (ThermoScientific) and visualized through chemiluminescence using the ChemiDoc MP Imaging System machine (Bio-Rad).




Adhikary, S., & Eilers­, M. (2005). Transcriptional regulation and transformation by MYC proteins. Nat Rev Mol Cell Biol, 6(8), 635-645.

Arvanitis, C., & Felsher, D. W. (2006). Conditional transgenic models define how MYC initiates and maintains tumorigenesis. Semin Cancer Biol, 16(4), 313-317. doi:10.1016/j.semcancer.2006.07.012

Boxer, L. M. D., Chi. V. (2001). Translocations involving MYC and MYC function. Oncogene, 20(40), 5595-5610. doi:10.1038/sj.onc.1204595

Bradner, J. E., McPherson, O. M., & Koehler, A. N. (2006). A method for the covalent capture and screening of diverse small molecules in a microarray format. Nat Protoc, 1(5), 2344- 2352. doi:10.1038/nprot.2006.282

Clemons, P. A., Bodycombe, N. E., Carrinski, H. A., Wilson, J. A., Shamji, A. F., Wagner, B. K., . . . Schreiber, S. L. (2010). Small molecules of different origins have distinct distributions of structural complexity that correlate with protein-binding profiles. Proc Natl Acad Sci U S A, 107(44), 18787-18792. doi:10.1073/pnas.1012741107

Dang, C. V. (1999). MYC Target Genes Involved in Cell Growth, Apoptosis, and Metabolism. Molecular and Cellular Biology, 19(1), 1-11.

Dang, C. V. (2012). MYC on the Path to Cancer. Cell, 149(1), 22-35. doi:10.1016/j.cell.2012.03.003

Darnell, J. E., Jr. (2002). Transcription factors as targets for cancer therapy. Nat Rev Cancer, 2(10), 740-749. doi:10.1038/nrc906

Delmore, J. E., Issa, G. C., Lemieux, M. E., Rahl, P. B., Shi, J., Jacobs, H. M., . . . Mitsiades, C. (2011). BET bromodomain inhibition as a therapeutic strategy to target MYC. Cell, 146(6), 904-917. doi:10.1016/j.cell.2011.08.017

Felsher, D. W. (2010). MYC Inactivation Elicits Oncogene Addiction through Both Tumor Cell– Intrinsic and Host-Dependent Mechanisms. Genes & Cancer, 1(6), 597-604. doi:10.1177/1947601910377798

Jain, M., Arvanitis, C., Chu, K., Dewey, W., Leonhardt, E., Trinh, M., . . . Felsher, D. W. (2002).

Sustained Loss of a Neoplastic Phenotype by Brief Inactivation of MYC. Science, 297(5578), 102-104. doi:10.1126/science.1071489

Li, Z., Van Calcar, S., Qu, C., Cavenee, W. K., Zhang, M. Q., & Ren, B. (2003). A global transcriptional regulatory role for MYC in Burkitt’s lymphoma cells. Proc Natl Acad Sci U S A, 100(14), 8164-8169. doi:10.1073/pnas.1332764100

Liu, P., Cheng, H., Santiago, S., Raeder, M., Zhang, F., Isabella, A., . . . Zhao, J. J. (2011). Oncogenic PIK3CA-driven mammary tumors frequently recur via PI3K pathway-dependent and -independent mechanisms. Nature medicine, 17(9), 1116-1120. doi:10.1038/nm.2402

Muellner, M. K., Uras, I. Z., Gapp, B. V., Kerzendorfer, C., Smida, M., Lechtermann, H., . . . Nijman, S. M. (2011). A chemical-genetic screen reveals a mechanism of resistance to PI3K inhibitors in cancer. Nat Chem Biol, 7(11), 787-793. doi:10.1038/nchembio.695

Orian, A., van Steensel, B., Delrow, J., Bussemaker, H. J., Li, L., Sawado, T., . . . Eisenman, R. (2003). Genomic binding by the Drosophila MYC, MAX, Mad/Mnt transcription factor network. Genes & Development, 17(9), 1101-1114. doi:10.1101/gad.1066903

Pop, M. S., Stransky, N., Garvie, C. W., Theurillat, J.-P., Lewis, T. A., Zhong, C., . . . Garraway, A. (2014). A small molecule that binds and inhibits the ETV1 transcription factor oncoprotein. Molecular cancer therapeutics, 13(6), 1492-1502. doi:10.1158/1535- 7163.MCT-13-0689

Prochownik, E. V., & Vogt, P. K. (2010). Therapeutic Targeting of MYC. Genes Cancer, 1(6), 650-659. doi:10.1177/1947601910377494

Riss, T. L., Moravec, R. A., & Niles, A. L. (2011). Cytotoxicity Testing: Measuring Viable Cells, Dead Cells, and Detecting Mechanism of Cell Death. In M. J. Stoddart (Ed.), Mammalian Cell Viability: Methods and Protocols (pp. 103-114). Totowa, NJ: Humana Press.

Soucek, L., Whitfield, J., Martins, C. P., Finch, A. J., Murphy, D. J., Sodir, N. M., . . . Evan, G. I. (2008). Modelling MYC inhibition as a cancer therapy. Nature, 455(7213), 679-683. doi:10.1038/nature07260

Stratikopoulos, E. E., Dendy, M., Szabolcs, M., Khaykin, A. J., Lefebvre, C., Zhou, M. M., & Parsons, R. (2015). Kinase and BET Inhibitors Together Clamp Inhibition of PI3K Signaling and Overcome Resistance to Therapy. Cancer Cell, 27(6), 837-851. doi:10.1016/j.ccell.2015.05.006

Van Riggelen, J., Yetil, A., & Felsher, D. W. (2010). MYC as a regulator of ribosome biogenesis and protein synthesis. Nat Rev Cancer, 10(4), 301-309.

Vita, M., & Henriksson, M. (2006). The MYC oncoprotein as a therapeutic target for human cancer. Seminars in Cancer Biology, 16(4), 318-330. doi:

Wang, H., Hammoudeh, D. I., Follis, A. V., Reese, B. E., Lazo, J. S., Metallo, S. J., & Prochownik, E. V. (2007). Improved low molecular weight MYC-MAX inhibitors. Molecular cancer therapeutics, 6(9), 2399.

Yin, X., Giap, C., Lazo, J. S., & Prochownik, E. V. (2003). Low molecular weight inhibitors of MYC-MAX interaction and function. Oncogene, 22(40), 6151-6159.