Cell Death and Cancer Therapy: Don't Forget to Kill the Cancer Cell!

The selective targeting of cancer-specific molecular vulnerabilities by targeted therapies has demonstrable clinical benefit. Although the modern discovery of targeted therapies is largely driven by the vast clinical oncology needs that are not met by conventional chemotherapy, it should be acknowledged that to date, this broad category of therapy has been by far the most successful category in the medical oncologist's armamentarium. “Conventional chemotherapy” here refers to those drugs, used for decades, that primarily target microtubules (like vinca alkaloids and taxanes) or DNA, whether by damaging it or affecting DNA metabolism (like anthracyclines, platinums, alkylating agents, topoisomerase inhibitors, and antimetabolites).

In this day of exciting targeted therapies, why should we care about conventional chemotherapy? Conventional chemotherapy has cured millions of people of otherwise fatal diseases, and provided clinical benefit in millions of others (1). “Cured” here means that patients presented to their doctor with a fatal cancer, were given a finite course of therapy, the cancer went into a complete remission (CR), and never returned. One way to gauge the magnitude of benefit of chemotherapy is to compare the number of new cases and deaths in leukemias and lymphomas, for which conventional chemotherapy is the primary modality. In 2015 in the United States alone, there were 135,000 diagnoses, but only 45,000 deaths (2). Clearly, this type of exceptional, life-saving benefit has been realized primarily in the hematologic malignancies, including aggressive lymphomas, Hodgkin disease, and acute leukemias. However, cures are also frequently obtained in testicular cancer, anal cancer, and osteosarcoma, often in combination with radiotherapy. Moreover, in breast cancer there have been millions of women whose undetectable residual breast cancers were rendered clinically silent, if not cured, by adjuvant therapy, or whose cure rates were enhanced by neoadjuvant chemotherapy (3). Although in these cases it is not possible to identify specific individuals who benefitted, it is clear that roughly one third of those so treated would otherwise have relapsed with incurable disease if they had not received chemotherapy. Finally, there are those patients who have received meaningful clinical responses, even CRs, to conventional chemotherapy, but who eventually succumbed to their disease. These patients, again numbering in the millions, were not cured, but received benefit in terms of increased length and quality of life.

In describing the impressive clinical benefits of conventional chemotherapy, significant shortcomings in terms of both clinical benefit and toxicities should not be overlooked. Conventional chemotherapy provides all-too-limited a benefit in the majority of advanced solid tumors and many hematologic tumors. Although it performs impressively in subsets described above, the vast majority of patients with advanced cancers are not curable by conventional chemotherapy. Moreover, conventional chemotherapy regimens often have significant side effects that run the gamut from alopecia and fatigue to nausea and inanition to life-threatening febrile neutropenia. There is a desperate need for improving medical therapy for cancer, and modern targeted agents will doubtless play an important, perhaps dominant, role in the future. However, talk of eliminating conventional chemotherapy to give way to targeted therapy is probably premature until we can identify targeted agents, or combinations thereof, that can provide similar curative potential. Indeed, it seems likely that ideal regimens might contain elements of both conventional and targeted therapies.

Given the impressive, albeit restricted, successes of conventional chemotherapeutics, what is the source for their therapeutic index? Indeed, given the aggregate benefit of conventional chemotherapy, there is shockingly little investigation into the source of the therapeutic index. Their targets, largely microtubules and DNA, alone do not seem to provide the answer as they are present in all cells, malignant and nonmalignant, in the patient. Textbook explanations favor a relatively vague appeal to differential proliferation rates, and the tendency of conventional agents to selectively target rapidly dividing cells (4). Although some support exists for this model, surprisingly few clinical studies have simply compared tumor cell proliferation rate and response with chemotherapy (5–10). Moreover, many cancers generally do not proliferate at rates greatly in excess of many normal tissues (11, 12). In addition, many tumors, such as indolent lymphomas, that divide more slowly than gut epithelium, are sensitive to conventional chemotherapy agents that do not cause significant gut toxicity (13, 14). It therefore appears clear that there must be other important determinants of chemosensitivity.

In this regard, it is important to recognize that sensitivity to one type of conventional chemotherapy is commonly associated with sensitivity to other types. For instance, advanced non–small cell lung cancer provides a clinical conundrum not only because it is relatively insensitive to cisplatin, but also because it is relatively insensitive to anthracyclines, vinca alkaloids, alkylating agents, and everything else. On the other hand, pediatric acute lymphoblastic leukemia is curable not merely because it is sensitive to vinca alkaloids, but rather because it is also sensitive to alkylating agents, anthracyclines, and antimetabolites. Although there are likely agent-specific mechanisms, such as selective drug efflux or selective metabolism, that can affect agent-specific cancer sensitivity, it appears that there must also be mechanisms common to how a wide variety of agents kill that explain the therapeutic index. Because nearly all conventional chemotherapeutic agents have the ability to induce cell death by the mitochondrial pathway of apoptosis, my laboratory has tested differences in this pathway as a potential explanation for the therapeutic index.

Commitment of a cell to programmed cell death via the mitochondrial pathway of apoptosis is governed by the BCL-2 family of proteins at the mitochondrion (15). This family comprises both anti- and proapoptotic proteins, and to a rough approximation, the balance of these two classes determines whether a cell chooses survival or commitment to death. The point of commitment to apoptosis is reached when the outer membrane of the mitochondrion is permeabilized by the combined action of proapoptotic proteins. Permeabilization releases factors that cause proteins and DNA to be cleaved and the plasma membrane to be marked with “eat me” signals that promote rapid phagocytois. Permeabilization is dependent on the so-called BH3 domains of the proapoptotic proteins, amphipathic alpha-helical segments of approximately 20 amino acids. These domains by themselves can replicate much of the function of some proapoptotic proteins (16). Indeed, by measuring mitochondrial permeabilization in response to titrated doses of synthetic BH3 peptides, one can create a useful index of mitochondrial sensitivity to proapoptotic signaling, an index we call “priming” (17, 18). Highly primed cells, containing mitochondria that permeabilize to relatively low doses of BH3 peptides, are those that are very close to the threshold of apoptosis. One can thus perform experiments to rank different cell samples based on their mitochondrial priming in a technique we call “BH3 profiling” (Fig. 1).

Armed with this tool, we then tested the hypothesis: “Is baseline apoptotic priming a key determinant of the therapeutic index in cancer?” We obtained pretreatment patient tumor samples and measured priming with BH3 profiling and then compared with clinical response to standard chemotherapy regimens. We performed similar BH3 profiling on normal human and mouse tissues. We found that chemosensitive tumor types consistently had high baseline priming, whereas normal somatic tissues generally had dramatically lower baseline priming (Fig. 2; ref. 19). Important exceptions were hematopoietic cells, which were the most primed normal cells studied. It is probably not coincidental that hematopoietic cell toxicity is the most common dose-limiting toxicity for conventional chemotherapy regimens.

Chemoresistant tumor types exhibited baseline priming similar to that of normal tissues. These results supported our hypothesis, therefore. We also found that baseline apoptotic priming could discriminate the even more subtle distinction of chemosensitive and chemoresistant tumors within the same histology. We have thus far replicated these findings in acute myelogenous leukemia (AML), acute lymphoblastic leukemia, chronic lymphocytic leukemia, multiple myeloma, and ovarian cancer (19–21). In Fig. 3, an example in AML is provided. It can be seen that BH3 profiling performed on myeloblasts from newly diagnosed patients not only provides predictive information about the ability of chemotherapy to induce a CR, but also to consolidate a CR. The lesson is that if the myleoblasts are more primed than the hematopoietic stem cell (HSC), then there is a great likelihood of cure. If myeloblasts are less primed than HSC, then cure with chemotherapy is very unlikely. Moreover, we have found that when tumors have a CR and later relapse, the relapsed sample tends to select for lower priming. This suggests that priming is not simply a marker, but a determinant of chemosensitivity, and indeed that selection for reduced mitochondrial apoptotic priming may be an important cause of the pan-resistant phenotype of many relapsed tumors. More recently, it has even become apparent that response to targeted therapies, including kinase inhibitors, can be predicted with a modification of BH3 profiling called dynamic BH3 profiling (22). In this version, patient cells are first briefly exposed to agents of interest, and then the priming change induced by the agents is measured. The drug-induced priming changes correlate well with in vivo cytotoxicity.

Others have used different methods for measuring the pretreatment apoptosis (23, 24). The Prehn laboratory measures absolute protein levels of key BCL-2 family members in pretreatment samples. These values are then put into a systems model predictor called DR_MOMP to predict the impact of apoptosis-sensing drugs. This tool also can discriminate between clinically chemosensitive and chemoresistant colon and rectal cancer cases.

The results of these studies have several implications. First, they suggest that measurement of pretreatment conditions of the mitochondrial apoptotic pathway are sufficient to predict response to therapy, so that perhaps DR_MOMP or BH3 profiling might have utility as a clinical decision-making tool. Such studies are underway. Second, and more fundamentally, they demonstrate that to a first approximation, differences in dynamic response to chemotherapy are not what differentiate sensitive and insensitive cells, but rather the preexisting sensitivity to death signaling. Third, because the measurement of apoptotic parameters alone affords such predictive power, they suggest that at least in the clinical contexts studied, apoptosis via the mitochondrial pathway is the predominant form of chemotherapy-induced cell death in vivo. Finally, it is worth noting that many cite the famous Hanahan and Weinberg hallmark of “resisting cell death” as support for the notion that cancer cells are more resistant to apoptosis than are normal cells. In fact, the opposite may more commonly be true—cancer cells are often more sensitive to apoptotic signaling than normal cells are, or else chemotherapy would be far less effective.

Conventional chemotherapy takes advantage of the therapeutic index afforded by the enhanced apoptotic priming of cancer cells, but it does so as a blunt tool. Consequently, there are many accompanying toxicities, including carcinogenic genotoxicity. It is therefore desirable to consider if the apoptotic pathway can be more precisely targeted. The first clinical agents to directly target the BCL-2 family are the so-called BH3-mimetic drugs (25). These are large (nearly 1 kDa) compounds that mimic the amphipathic BH3 domain alpha-helices. As such, they compete for the BH3-binding sites on antiapoptotic proteins, freeing proapoptotic proteins from sequestration. These freed proapoptotic proteins can then participate in forming the outer mitochondrial membrane pores that commit the cell to programmed cell death.

In this CCR Focus, Gibson and Davids (26) review the development and impressive clinical activity of BH3 mimetics, with particular emphasis on the BCL-2–selective venetoclax (ABT-199; ref. 27). Several pharmaceutical programs are in the process of developing selective inhibitors of other antiapoptotic proteins like BCL-XL and MCL-1. Indeed, it is not out of the question that a pan-inhibitory compound that inhibits all of the antiapoptotic proteins would also have clinical utility, taking advantage of priming differences like conventional chemotherapy, but without the myriad side effects that accompany genotoxicity.

It is still early days in the field of clinical BH3 inhibition, but given the activity and low toxicity of venetoclax it seems reasonable to foresee this class of drug potentially being combined with nearly any other agent(s) that also induce apoptosis. Initial studies of BH3 mimetics have focused on the hematologic malignancies. Indeed, given that hematologic malignancies are generally more apoptotically primed than solid tumors, it is likely that single-agent activity of BH3 mimetics will be greater in hematologic malignancies. Nonetheless, the modalities of conventional chemotherapy and kinase inhibition both kill via apoptosis (22, 28–31), and both are used to treat solid tumors. It seems likely, therefore, that there are potential combinations in solid tumors to be exploited. Developing the tools to identify these combinations will be an important task.

Second mitochondrial activator of caspases (SMAC) is released from the mitochondrion following its permeabilization (32). It has been found to bind and inhibit, via a tetrapeptide sequence, members of the inhibitor of apoptosis (IAP) family. On the basis of the results from invertebrate model organisms, members of the IAP family were initially thought to act primarily as direct inhibitors of caspases, serine proteases activated during apoptosis that are responsible for many of the phenotypes of apoptotic cells. However, the reality in mammals has turned out to be much more complicated. Although certain IAP proteins, such as XIAP, can directly inhibit caspases, others, such as CIAP1 and CIAP2, are weak caspase inhibitors, and instead more prominently function as E3 ubiquitin ligases (33, 34).

SMAC mimetics are small molecules that inhibit the activity of IAP proteins by mimicking the key tetrapeptide sequence that facilitates binding to IAP proteins (35). These compounds have biologic activity, but their success in the clinic will likely depend greatly on identifying the right combinations in which to include them, as reviewed by Fulda in this CCR Focus (36). Intriguingly, these agents have the potential to activate caspases independently of the mitochondrion, and, indeed to even provoke nonapoptotic cell death. Therefore, they provide an intriguing possible alternative for tumors that are resistant to apoptosis via the mitochondrial pathway.

Autophagy is the program of sequestering intracellular organelles and proteins into autophagosomes and then lysosomes for degradation. It was initially noted that autophagy was active in the context of many stresses that induced cell death, and was hence considered to be a mechanism of cell death. However, subsequent work has shown that it is more often a mechanism of cell survival in the face of stress (37). The role of autophagy in the cellular response to extreme stress is more analogous to that of firemen at a fire: firemen are invariably present at fires, but are there to put them out, not to start them.

Autophagy is necessary for recycling metabolic intermediates under many types of cell stress. It may be that in certain cases, oncogenesis carries with it certain inherent metabolic stresses, so that inhibition of autophagy would be selectively harmful to the cancer (38–40). In general, however, whether autophagy promotes or inhibits cancer is a complicated and incompletely answered question (41). Indeed, the answer may well rest upon the particular phase of carcinogenesis being considered, and the specific cancer in question. In this CCR Focus, White and colleagues (42) review preclinical and clinical data to address the issue of where autophagy inhibition might be beneficial in cancer treatment.

There is perhaps no mode of cancer therapy that has generated more excitement in recent years than that of harnessing the immune system to eradicate cancer cells. Both adoptive cell therapies and immune checkpoint blockade inhibitors have scored recent major clinical successes (43–47). However, there continues to be great variation in the quality of response both among different tumor types and within tumor types. The depth and duration of clinical responses in patients with otherwise refractory cancers have been tremendously exciting, however, adding great impetus to the efforts to better understand the mechanism of good response.

Most of the attention in trying to understand differential response has been paid to aspects of tumor recognition by the adaptive immune system. In this CCR Focus, Martínez-Lostao and colleagues (48) address the less frequently considered question of how the adaptive immune system actually kills target tumor cells. In addition, they consider how different mechanisms of tumor cell death can themselves affect recognition by the immune system.

These articles study a wide variety of therapies, but unite in a common theme: A successful therapy must not merely engage, but must kill cancer cells. Understanding how therapies kill is, therefore, a critical part of understanding how to make them perform better clinically.

A. Letai reports receiving commercial research grants from and is a consultant/advisory board member for AbbVie, AstraZeneca, and Tetralogic, and is listed as a co-inventor on pending patent applications, which are owned by Dana-Farber Cancer Institute, related to BH3 profiling. No other potential conflicts of interest were disclosed.