Inhibitors Targeting Mitosis: Tales of How Great Drugs against a Promising Target Were Brought Down by a Flawed Rationale

On the premise that tumors harbor a (much) larger fraction of actively dividing cells than do normal tissues and should therefore be more vulnerable to agents aimed at cell division, researchers have developed drugs to target numerous components of this intricate process as chemotherapeutics. Chief among these are agents that target the processes of mitosis and the accompanying cytokinesis, with inhibitors of mitotic kinases being among the most recent entries (1). Although the clinical efficacy of agents that target the process of cell division [e.g., microtubule-targeting agents (MTA)] provided support for this rationale, accumulating clinical and preclinical evidence is encouraging a reassessment of how MTAs act (2). Our increasing understanding of the working of cells in general and cancer cells in particular, together with greater knowledge about the diversity of targets of our chemotherapeutic agents, suggests that a more complex approach than only targeting cell division may be warranted. Among the components of cell division, mitosis, the process whereby a eukaryotic cell separates its replicated chromosomes into identical sets in 2 offspring, has long been viewed as an attractive target for chemotherapeutic agents. Together, mitosis and the accompanying cytokinesis that divides the nucleus, cytoplasmic organelles, and cell membrane into 2 cells define the mitotic or M-phase of the cell cycle. In most cells, this accounts for less (and often substantially less) than 10% of the cell cycle. Emerging knowledge is revealing the complexity of the mitotic process and allowing the identification of a diverse group of proteins whose activity is precisely orchestrated during mitosis. Mitosis requires many other players in addition to tubulin and microtubules (MT). Many of these other proteins are primarily functional only during mitosis and are responsible for controlling different steps in the assembly and function of the complex machinery of the mitotic spindle. Among these proteins, the Aurora kinases (AK), Polo-like kinases (PLK), and kinesin spindle protein (KSP) have emerged as targets for cancer therapeutics (3, 4).

As shown in Fig. 1, ordered arrays of MTs play essential roles in both interphase (G₁-, S-, and G₂-phases) and mitosis. In animal cells, centrosomes act as MT-organizing centers (i.e., the site of MT nucleation), the structures from which MTs emerge (5). Centrosomal nucleation of polarized arrays of MTs is essential for mitosis and for cellular organization during the large portion of the cell cycle that is not mitosis. In interphase, the MT-organizing center organizes the array of MTs that provides polarity to the cytoplasm. This array is an essential structure for the cellular trafficking of a myriad of proteins, including many important oncoproteins. Polarized arrays of MTs grow outward from near the nucleus in growing epithelial cells, and from apical to basal in polarized cells (6). Their polarity is recognized by motor proteins of the dynein and kinesin superfamilies, allowing for directed movement of cargo on MTs (7). Dyneins move toward the minus (nuclear) end of MTs, whereas kinesins mostly move to the plus (peripheral) end of MTs. In cells undergoing cell division, a major function of the centrosome is to organize the 2 opposing arrays that form the mitotic spindle apparatus, which is required for separation of chromosomes during cell division.

MTs provide the structure and machinery for chromosome segregation during mitosis, a process that also requires an array of other proteins, including essential serine/threonine kinases. Figure 2 depicts the principal localization of the mitosis-associated proteins that have been targets of drugs in clinical trials (8). Figure 3 shows the proteins that modulate the activity of the mitotic kinases and those through which the effects of these mitosis-associated proteins are mediated. Although the mitotic kinases often act in concert, they also perform unique functions without overlap. During S-phase, Aurora kinase A (AKA) localizes to the centrosome. It then translocates to the mitotic poles and the adjacent spindle MTs as the cell progresses through prophase, metaphase, and anaphase, eventually locating in the midbody during telophase and cytokinesis (9). During translocation from the centrosome, the level of AKA increases. This indicates that expression is largely restricted to mitosis and that agents designed to target AKA would be inactive in cells in other phases of the cell cycle (this point is discussed further below). Aurora kinase B (AKB) localizes to bundles of specialized MTs termed K-fibers and helps connect the kinetochore, a protein structure that assembles on the centromere of chromosomes, to spindle fibers during prometaphase and metaphase. It then relocalizes to the midzone or central spindle at the metaphase–anaphase transition, influencing chromosome separation and cytokinesis. Like AKB, Aurora kinase C (AKC) is a chromosome passenger protein that localizes first to the inner centromere, then to the central spindle, and finally to the midbody of mitotic cells as the cell cycle progresses (10). As with AKA, the mRNA and protein expressions of AKB and AKC are maximally elevated during the G₂/M-phase of the cell cycle and decrease rapidly as the cells enter G₁ (10).

PLK 1–4 localize to the centrosome during interphase and prophase, move to the spindle poles after prophase, and relocate to the midbody for telophase and cytokinesis (11).

Like the mitotic kinases, KSP is mitosis-specific and thus has attracted attention as a target for anticancer agents targeting mitosis (12). KSP is a homotetrameric kinesin motor protein and a member of the kinesin superfamily of proteins that binds and hydrolyzes ATP, coupling the energy released from hydrolysis with force production that allows for unidirectional movement along MTs. Unlike transport kinesins, which function in cytosolic movement and the localization of organelles and vesicles, mitotic kinesins such as KSP participate in the assembly, maintenance, and elongation of the mitotic spindle, chromosome alignment and segregation, and MT depolymerization. KSP's antiparallel MT-sliding function is required for separation of the 2 centrosomes in early mitosis, an essential event for bipolar spindle formation. Consequently, inhibition of KSP results in mitotic arrest because the failure of centrosomal separation results in an MT array with only one pole.

Numerous studies have shown that AK, PLK, and KSP are expressed primarily in the M-phase of the cell cycle and, to a lesser extent, in G₂ (8). During G₁-, G₀-, and S-phases, expression either does not occur or occurs only at very low levels. Consequently, drugs aimed at these proteins will only find their target and have an effect on cells that are rapidly cycling (dividing) and likely to be in or passing through the G₂- or M-phase while the drug is present. This highly restricted, cycle-specific expression likely explains why these agents have failed to affect tumor growth in patients.

Inhibitors of mitotic proteins all cause disruption of normal mitotic function, as do MTAs, which additionally disrupt interphase functions. The mechanism whereby mitosis is disrupted differs among these agents. MTAs and inhibitors of KSP and PLK interfere with proper assembly of the mitotic spindle and lead to arrest in mid-mitosis (4, 13). Inhibitors of AKA cause transient arrest followed by mitotic exit with misaligned chromosomes, and inhibitors of AKB cause premature mitotic exit with major defects in chromosome attachment (13). Although the mechanistic details differ, all of these compounds, including MTAs, disrupt the normal process of mitosis, but of course, they only do so when mitosis occurs. None of these agents produces or promotes the initiation of mitosis, and hence will only cause damage to cells when mitosis happens to occur. If mitosis occurs rarely, agents whose action is restricted to mitosis will rarely cause damage. However, MTs are present throughout the cell cycle, and MT arrays are most vulnerable during mitosis due to the increased rate of MT turnover. Hence, MTAs potently target mitosis when it occurs, but they can also target interphase cells.

Studies involving a number of cell types and a variety of methods, including the use of mitosis inhibitors, showed that mitotic arrest can only be sustained for ∼1–2 days in human cells with constant drug exposure, and less than that in rodent cells (14). Although 1–2 days is a significant duration, given that the normal duration of mitosis is only 1–2 hours in many different cell types and tissues, it is by no means indefinite (15). Mitotic arrest causes many different types of stress. For example, condensed chromosomes cannot be transcribed, which makes it impossible to replenish needed transcripts—a state that cannot be sustained. Consequently, this period is limited by mitotic slippage, especially in normal cells, or cell death, especially in cancer cells (14). A further complication is that cells display significant intra- and interline variability in the duration of mitotic arrest, and great variation among different drugs and the mode of exit from arrest (16). The latter presumably happens when the slow degradation of cyclin B1, which continues during arrest, is reduced below the level that is required to maintain the mitotic state. At this point some cells die; some exit mitosis (termed mitotic slippage), remain tetraploid, and possibly die later; and some show other fates (reviewed in ref. 17). Whatever the details, the data show that mitotic arrest is not maintained for any extended period. Furthermore, it is unclear how any of these results would change with pulsed or varying drug concentrations. Very few data are available regarding mitotic duration in patients. In one study, mitotic counts that were obtained before and after treatment with an MTA (paclitaxel) showed great variability from patient to patient in the extent and duration of mitotic arrest (18).

Cancer is often mistakenly thought of as a mass of abnormal cells growing rapidly in an uncontrolled manner. This is misleading for at least 2 reasons. First, the data show that cell division is a very precisely regulated process that follows similar pathways in both normal and cancer cells and involves highly specialized proteins. Second, one might think that frequent cell division is a hallmark of tumor cells. This misconception is encouraged by the rapid rate of cell division of cancer cells observed in vitro and in xenograft models. Although it can be challenging to measure tumor doubling times in humans, studies have shown that tumors do not double as rapidly as one might think. As summarized in Table 1, data obtained by a variety of radiologic imaging modalities show median doubling times for many human tumors of >100 days, which is much longer than the doubling times observed in preclinical animal models (2). This means that at any one time, only a very small percentage of tumor cells are undergoing mitosis. Indeed, the mean mitotic index in a variety of tumor types has been shown to be <1% (19, 20). Additionally, the mean labeling index measured by radiolabeled imaging [e.g., tritiated thymidine ([3H]-TdR)] was only a few percent of the tumor (21). Together, these data show that in contrast to in vitro and xenograft models, human tumors have very long doubling times. This makes such tumors indifferent to drugs that target cell proliferation or proteins whose expression is highly restricted to one phase of the cell cycle.

Table 2 and Fig. 4 show preclinical in vivo data obtained for mitotic kinase and KSP inhibitors, and they provide insight into why this class of agents has not been successful clinically. Also shown in Table 2 are the doubling times of the preclinical models used in various studies (see also Table 1). As noted above, this rate of doubling is markedly faster than that found in any solid tumor in humans, and this difference may be significant for target genes/proteins whose expression is highly restricted during the cell cycle, such as occurs with mitosis-specific proteins. Furthermore, in most models the agents were administered frequently (often daily and even twice daily). Frequent administration is essential for agents that target proteins whose expression is restricted to a very short part of the cell cycle (22). The importance of frequent drug administration is underscored by the results depicted in Fig. 4, which show a decline in activity when drug administration was changed from twice per week to once per week. Finally, we note, as summarized in Table 2, that complete responses were achieved in only 2 of the 53 models evaluated, underscoring the fact that despite a frequent schedule of administration to rapidly growing tumors, complete tumor disappearance could not be achieved. From these data, it is clear that in the models with rapid doubling times, cells that are not in G₂/M-phase are refractory to therapy and can repopulate the rapidly dividing tumor mass unless frequent administration is used (23).

Few (if any) agents have shown as much activity as paclitaxel did during its development, in terms of both breadth and magnitude. A review in 1995 of early evaluations of the drug's activity showed single-agent response rates of 17% to 62%, 20% to 48%, and 21% to 41% in breast, ovarian, and lung cancers, respectively (24). Activity was subsequently reported in other cancers, including Kaposi sarcoma and esophageal, urothelial, and head and neck cancers. In our own tabulation of 29 studies of single-agent paclitaxel in these seven cancers, we recorded response rates ranging from 7% to 71%, with a median of 37% and an overall response rate of 28%, in a group of 2,271 patients (data not shown; see references in ref. 25). With this as background, we summarize in Table 3 and Supplementary Tables S1A and S1B the clinical results of 46 studies conducted with 20 different agents targeting the AKs, PLK, and KSP. The range and median values for all patient groups are presented at the bottom of each table. The overall response rate of 1.6% for all studies shows unequivocally that these agents lack activity against solid tumors [787 (AKs, Table 3) + 345 (PLK, Table S1A) + 267 (KSP, Table S1B) = 1,399 patients total; objective response rate: 22/1399 = 1.6%]. Such a response rate might even be recorded with a placebo (26, 27). Some may wonder whether low bioavailability can explain why these drugs were inactive in patients. We think this unlikely given that poor results were seen even in cases where adequate biomarkers for mitotic arrest (mitotic index or histone H3 phosphorylation most often in skin biopsies) showed target engagement (28, 29). Moreover, we would argue that the neutropenia observed in many of the treated patients represents a biomarker or surrogate that clearly proves the drug hit the target and inhibited mitosis (even if it was not in the tumor). In addition, we emphasize our belief that a stable disease rate of ∼20% can only be interpreted as a measure of the inherent biology of the tumors being treated. As reviewed above, mitotic arrest can only be sustained for ∼1–2 days, so any assumption that this class of agents (or any other) can arrest cells in mitosis for any period that might be scored clinically as stable disease is not scientifically defensible. Similarly, given the restricted expression of the targets in mitosis, only a small percentage of cells would prove vulnerable to inhibitors targeting mitosis. This would argue against a situation in which the fraction of killed cells equals the quantity that has been newly added to the tumor by division.

Not surprisingly, we note that in the subset of patients with hematologic malignancies, the response rate was higher, but still low, at 8.2%. Given that hematologic cancers often have faster doubling times than solid tumors, one could envision a potential role for agents that target mitosis in these malignancies. However, the neutropenia that will invariably occur might limit treatment such that both the amount and frequency of drug administration would be inadequate.

Finally, although infrequent responses have been observed clinically, occasionally an off-target effect inhibiting a kinase other than a mitotic kinase has resulted in activity. For example, activity has been reported with an AK inhibitor, tozasertib (MK-0457/VX680), in patients with imatinib-resistant chronic myeloid leukemia (CML) harboring the T315I mutation (30). However, these effects are independent of the cell cycle and do not contradict the reasons given herein to explain why these drugs have failed to benefit the overwhelming majority of patients with solid tumors.

Mitotic kinase inhibitors were developed as nonneurotoxic alternatives to MTAs; however, the rationale behind the development of these agents, i.e., the thesis that MTAs kill cells in human tumors only by inhibiting mitosis, meant that if unrecognized mechanisms proved to be important, such agents would fail when they reached the clinic. In an in vitro or a xenograft model with a doubling time of a few days, a large fraction of cells will prove vulnerable to a therapy that targets a protein that is crucial for mitosis. Therefore, it is not surprising that in a patient with a tumor doubling time of ≥30–60 days and an S-phase fraction of at most a few percent, only a small, insignificant fraction of cells will be vulnerable to a drug aimed at a target that is expressed only transiently in most tumor (and normal) cells. Inhibitors that target mitosis could prove effective in rapidly growing (i.e., rapidly dividing) leukemias and lymphomas (i.e., Burkitt lymphoma), and their value in such malignancies should be further explored. However, their lack of activity against a majority of cancers highlights several aspects of cancer drug development in the 21st century and informs our understanding of the mechanisms of drug action, as summarized below.

First, the dose-limiting toxicity of mitotic kinase inhibitors—reversible neutropenia—stands as a testament to the prowess of pharmaceutical drug developers, especially the chemists who synthesized these agents, and the biologists who identified and characterized the targets and developed the models used in their validation. Although the majority of human tumors do not divide rapidly enough to be susceptible to these mitotic poisons, the same cannot be said of vulnerable marrow elements. The doubling time of granulocyte precursors is very short [17 hours for myeloblasts, 63 hours for promyelocytes, and 55 hours for myelocytes (31)]. Thus, reversible neutropenia would be expected of an agent targeting a mitotic kinase or KSP, because at any one time ≥25% of marrow neutrophils are undergoing mitosis. Indeed, we would argue that the neutropenia observed nearly uniformly in clinical trials with agents that inhibit mitosis can be viewed as a biomarker or surrogate of activity.

Second, although this pharmaceutical prowess has positive attributes, it led to the overdevelopment of drugs aimed at targets that have not yet been validated (e.g., AKs, PLKs, and KSP). To date, clinical data have been reported for at least 20 different mitosis-specific poisons, and there are likely more yet to be reported. This has involved a considerable financial outlay and patient recruitment effort without a return on the investment (the response rate to an inhibitor targeting mitosis in clinical trials involving 1,399 patients with solid tumors was 1.6%; see Fig. 4 and Supplementary Tables S1A and S1B). Although a certain amount of redundancy represents attempts to design a better agent, the clinical evidence and the nearly uniform myelosuppression suggest that these drugs were remarkably similar and potent.

Third, the results indicate yet again the deficiencies of the preclinical models used in drug development. Although these preclinical models are far from ideal (but at some level valuable), it is clear that the rapid doubling times in such models compared with human tumors allow for an accelerated drug-development timeline. However, this precludes their use when the doubling time is a critical factor in a drug's activity. The rapid doubling times of the preclinical models explains why agents targeting mitosis proved active in these models but were ineffective against patient tumors.

Fourth, the results highlight the increasing use/abuse of “stable disease” as a measure of drug efficacy. As discussed above, scoring stable disease as a measure of activity cannot be defended scientifically for this class of compounds, arguing against its use as a measure of drug activity. As noted above, studies have shown that inhibitors targeting mitosis cannot be expected to stabilize tumor growth by arresting cells in mitosis for prolonged periods of time. Indeed, one can persuasively argue that these studies have collectively defined ∼20% as the stable disease rate due to the inherent tumor biology of patients harboring advanced solid tumors who enroll in similar phase I/II trials, because in effect, these patients received a drug that had no disease-stabilizing activity. We note that this estimate of 20% as the stable disease rate due to inherent tumor biology is not excessive, because stable disease rates of 55% to 67% have been obtained with placebos in renal cell and hepatocellular carcinomas (26, 27). Finally, the disappointing clinical results ratify the paradigm that MTAs do not kill cancer cells in humans primarily by inhibiting mitosis, even though both MTAs and inhibitors that target mitosis are lethal to bone marrow elements in this way. We previously proposed the inhibition of trafficking on MTs as the principal mechanism of action of MTAs (2). One such example for which clear evidence now exists is the androgen receptor, which plays a crucial role in the growth of prostate cancer (32, 33). The evidence that MT disruption affects androgen-receptor trafficking helps explain the activity of both docetaxel and cabazitaxel in a disease that is often so indolent that questions about the need for therapy remain unresolved. Therefore, when studying cells in which an MTA or a combination of MTAs is active, one should seek to determine which of the critical proteins that traffic on MTs have been affected by the therapy in question.

No potential conflicts of interest were disclosed.