Hsp72 and Nek6 Cooperate to Cluster Amplified Centrosomes in Cancer Cells

The mitotic spindle is bipolar in nature as it is organized around two poles, each possessing a single centrosome (1, 2). However, the majority of cancer cells have extra or “amplified” centrosomes that would be expected to form multipolar spindles, leading to mitotic catastrophe and cell death (3, 4). Cancer cells solve this problem by either inactivating the extra centrosomes or clustering them into two poles to form a pseudo-bipolar spindle (5). The added time taken to organize such a spindle and presence of more than two centrosomes in early mitosis favors generation of merotelic microtubule–chromosome attachments that are poorly recognized by the spindle assembly checkpoint (SAC; refs. 6, 7). Hence, amplified centrosomes are not only tolerated but promote chromosome segregation errors, genome instability, and cancer progression (8–10).

Centrosomes act as the dominant microtubule organizing centers (MTOC) in proliferating animal cells because they recruit the γ-tubulin ring complex (γ-TuRC) from which microtubules are nucleated (11, 12). The γ-TuRC is concentrated at centrosomes through binding to components of the pericentriolar material (PCM), a highly ordered matrix of coiled-coil proteins that surrounds and is scaffolded by a pair of centrioles (13). The centrioles are composed of nine triplets of highly stable microtubules organized into a polarized barrel that is duplicated once per cell cycle. Hence, cells in G₁ have two centrioles, while cells in late G₂ have four centrioles (14). These centrioles remain in close proximity during interphase due to proteinaceous ties that link them together (15). Although each centriole is associated with its own PCM, the close juxtaposition of new centrioles with their parental centrioles in S–G₂ mean that only two distinct PCM clouds are visible both before and after duplication. Hence, when scoring centrosome amplification, we use the criteria that cells should have more than two PCM foci and more than four centrioles.

Experimental studies have revealed a number of microtubule-based processes that promote centrosome clustering, including motor protein activity, microtubule dynamics, and microtubule attachments to the centrosome, chromosomes, and cell cortex (16–18). In addition, the extra time taken to organize a pseudo-bipolar spindle in cells with amplified centrosomes means that clustering requires cells to have a robust SAC that can delay anaphase onset until complete chromosome congression has been achieved (16, 19). Targeting these different clustering processes could provide a selective approach to killing cancer cells with amplified centrosomes (20). However, the underlying mechanisms remain far from understood.

Uncoupling of the centrosome duplication cycle from the cell cycle can result from altered activity of tumor suppressor genes and oncogenes and is a major route to centrosome amplification in cancer cells (21–23). This could arise from changes in gene expression of centrosome duplication regulators caused by deregulated signal transduction pathways. Alternatively, it could be due to reduced integrity of checkpoints, such as the SAC or DNA damage checkpoints, which disturb the timing of cell-cycle progression and perturb coordination with the centrosome duplication cycle. Cancer cells also need to tolerate the presence of amplified centrosomes and destabilization of p53 through inactivation of the Hippo pathway may be a common mechanism in this regard (24).

Oncogenesis is associated with proteotoxic stress that leads to induction of heat shock proteins (HSP) required to maintain protein folding and homeostasis (25, 26). As such HSPs represent cancer targets that are not necessarily restricted to tumors with particular driver mutations (27, 28). Using RNAi depletion and chemical inhibitors, we recently found that Hsp72 is essential for mitotic spindle assembly in cancer cells, promoting stabilization of kinetochore-associated microtubules, (K)-fibers, through recruitment of the chTOG–TACC3 complex (29). This function of Hsp72 is dependent upon its phosphorylation by the Nek6 mitotic kinase. As Hsp72 and Nek6 are required for K-fiber stability and robust mitotic spindle assembly, we reasoned that they might also be required for clustering of amplified centrosomes. Here, we set out to test this hypothesis using fixed and time-lapse imaging of MDA-MB-231 (human breast carcinoma) and NIE-115 (mouse neuroblastoma) cell lines that harbor amplified centrosomes and exhibit centrosome clustering activity (16, 30, 31). Our data reveal a new pathway of centrosome clustering and lead us to propose a therapeutic role for Hsp72 and/or Nek6 inhibitors as selective declustering agents in cancer cells with amplified centrosomes.

All media were from Invitrogen and supplemented with 10% heat-inactivated fetal bovine serum (FBS), 100 IU/mL penicillin, and 100 mg/mL streptomycin. MDA-MB-231, HeLa, and HBL-100 cells were obtained from Cell Lines Service, while NIE-115 and RPE1-hTERT cells were obtained from ATCC. MDA-MB-231, NIE-115, HeLa, and HBL-100 cells were cultured in DMEM, while RPE1-hTERT cells were grown in DMEM-F12 with 0.25% sodium bicarbonate. Acute lymphoblastic leukemic (ALL) cells were obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (http://www.celllines.de) and grown in RPMI 1640 medium. Peripheral B-lymphocytes (PBL) are Epstein-Barr virus–transformed B-cells obtained from the Coriell Institute for Medical Research Biorepository (https://catalog.coriell.org). All cells were obtained within the last 10 years, stored in liquid nitrogen, and maintained in culture at 37°C in a 5% CO₂ atmosphere for a maximum of 2 months. We relied on the provenance of the original collections for authenticity, while mycoplasma infection was avoided through bimonthly tests using an in-house PCR-based assay. RNAi-resistant wild-type and kinase-dead (KD, K75M) Flag-Nek6 constructs were generated using the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies) and transfections performed with Lipofectamine 2000 reagent (Life Technologies) according to manufacturer's instructions. Unless otherwise indicated, the following inhibitors were added to cells for 4 hours: VER-155008 (Hsp70i, 10 μmol/L; Tocris Bioscience); Griseofulvin (10 μmol/L; Sigma); MLN8054 (Aurora-Ai; 500 nmol/L; MedChem Express); BI-2536 (Plk1i; 100 nmol/L; MedChem Express). Control cells were treated with the same volume of DMSO. For microtubule regrowth, cells were incubated for 16 hours with 500 ng/mL nocodazole (Sigma) followed by treatment with the 10 μmol/L VER-155008 for 2 hours. M-phase arrested cells were prepared by mitotic shake-off after 16 hours of nocodazole treatment.

Adherent cells were grown on acid-etched glass coverslips while suspension cells were seeded onto Superfrost Plus glass slides (Thermo). Cells were fixed with ice-cold methanol before being processed for immunofluorescence microscopy as previously described (29, 32, 33). For Hsp72 staining, cells were incubated for 30 seconds in pre-extraction buffer (60 mmol/L Pipes, 25 mmol/L Hepes, pH7.4, 10 mmol/L EGTA, 2 mmol/L MgCl₂, and 1% Triton X-100) prior to fixation. Primary antibodies were against α-tubulin (0.3 μg/mL; Sigma), γ-tubulin (0.5 μg/mL; Sigma), centrin-2 (N-17; 0.4 μg/mL; Santa Cruz Biotechnology), CEP135 (2 μg/mL; OriGene), pericentrin (2 μg/mL; Abcam), CenpA (2 μg/mL; Abcam), Hsp72 (0.8 μg/mL; C92F3A-5; Enzo Life Sciences), BubR1 (10 μg/mL; Abcam) and phospho-Histone H3 (2 μg/mL; Abcam). DNA was stained with 0.8 μg/mL Hoechst 33258. Secondary antibodies were Alexa Fluor-488 and -594 goat anti-rabbit and goat anti-mouse IgGs (1 μg/mL; Invitrogen). Imaging was performed on a Leica TCS SP5 confocal microscope equipped with a Leica DMI 6000B inverted microscope using a 63x oil objective (numerical aperture, 1.4). Z-stacks comprising of 30–50 × 0.3-μm sections were acquired. Time-lapse imaging was performed on a Leica TCS SP5 confocal microscope equipped with a Leica DMI 6000B inverted microscope using a 63× oil objective (numerical aperture, 1.4). Cells were cultured in glass-bottomed culture dishes (MatTek Corp.) and maintained at 37°C in an atmosphere supplemented with CO₂ using a microscope stage temperature control system (Life Imaging Services). SiR-tubulin (50 nmol/L; Spirochrome; SC002) was added 7 hours before imaging when required. Z-stacks comprising 20 × 0.5-μm sections were acquired every 5 minutes for a minimum of 16 hours. Stacks were processed as maximum intensity projections using LAS-AF software (Leica) and movies prepared using ImageJ (v.1.51). To quantify spindle association of Hsp72, the mean pixel intensity over the whole spindle was measured using Volocity software and normalized to spindle microtubule intensity, while BubR1 association with kinetochores was quantified by measuring total pixel intensity over the chromatin using ImageJ software.

Cells were seeded at 30% to 40% confluency in Opti-MEM Reduced Serum Medium and transfected using Oligofectamine (Life Technologies) according to manufacturer's instructions. ON-TARGETplus siRNA duplexes (100 nmol/L) were used against human Hsp72 (J-005168-06, J-005168-07), Nek6 (J-004166-06, J-004166-09), or Nek7 (J-003795-12, J-003795-14), or mouse Hsp72 (J-054644-05, J-054644-07) obtained from Thermo Fisher Scientific, or GAPDH (4390850), obtained from Life Technologies. siRNA duplexes were transfected. Cells were prepared for microscopy or Western blot analysis 72 hours after transfection of siRNA duplexes.

Preparation of cell lysates, SDS-PAGE, and Western blotting were performed as previously described (34). For Western blotting, primary antibodies were against Nek6 (1 μg/mL, Abcam), Nek7 (0.5 μg/mL, Abcam), Hsp72 (0.5 μg/mL, Enzo Life Sciences), Hsc70 (1 μg/mL, Santa Cruz Biotechnology), and GAPDH (1:1,000, Cell Signaling Technology). Secondary antibodies were goat or donkey anti-rabbit and anti-mouse horseradish peroxidase (HRP)-labeled IgGs (1:1,000).

All quantitative data represent means and SD of at least three independent experiments. Statistical analyses were performed using a one-way ANOVA analysis; ****, P < 0.0001; ***, P < 0.001; **, P < 0.01; *, P < 0.05.

To examine the role of Hsp72 in centrosome clustering, we first assessed the frequency of centrosome amplification in a set of cancer and noncancer-derived cell lines using antibodies against γ-tubulin to count PCM foci and centrin-2 to score centrioles (Fig. 1A). Consistent with previous studies (16, 30, 31), 33.5% MDA-MB-231 human breast cancer, and 96% NIE-115 mouse neuroblastoma cells were found to have >2 centrosomes in interphase, whereas only 8.3% HeLa human cervical cancer and 1% human RPE1 (retinal pigment epithelial cells immortalized with hTERT) had >2 centrosomes (Fig. 1B). To assess spindle polarity, cells were stained with antibodies against α-tubulin to detect microtubules and centrin-2 to score centrosomes. Scoring only mitotic cells with >2 centrosomes (i.e., >4 centrin-2 stained centrioles), 82% MDA-MB-231, and 76% NIE-115 cells formed pseudo-bipolar spindles with the amplified centrosomes clustered at two poles (Fig. 1C and D).

Spindle organization was then assessed in MDA-MB-231 and NIE-115 cells treated with increasing doses (5–20 μmol/L) VER-155008, a chemical inhibitor of Hsp70 (35). This revealed a dose-dependent increase in the frequency of cells with amplified centrosomes that had multipolar spindles following treatment with the Hsp70i (Fig. 1E–G). As VER-155008 inhibits all members of the Hsp70 family, we confirmed the specific role of the inducible Hsp72 isoform in centrosome clustering using two independent siRNA oligonucleotides to deplete the protein from MDA-MB-231 and NIE-115 cells. Both siRNAs caused a significant increase in multipolar spindles in cells with amplified centrosomes as compared with mock-depleted cells, while Western blotting confirmed that siRNAs depleted Hsp72 but not the constitutive Hsc70 in each cell line (Fig. 1H and I, and Supplementary Fig. S1A and S1B). Hence, we conclude that Hsp72 is essential for centrosome clustering in cancer cells with amplified centrosomes.

To analyze the role of Hsp72 in centrosome clustering, we took advantage of a far-red fluorescent probe, SiR-tubulin, that can stain microtubules in live cells (33, 36). Addition of SiR-tubulin revealed the initial assembly of multipolar spindles in NIE-115 cells following mitotic entry (Fig. 2A and B). However, following a mean time of 78 minutes in a multipolar state, these spindles subsequently resolved into a bipolar state before cells progressed into anaphase and completed division (Fig. 2C). In the presence of the Hsp70i, there was little difference from untreated cells in the length of prophase, measured from when a robust microtubule array formed to when nuclear envelope breakdown occurred. However, after assembling a multipolar spindle, this state was maintained for extended time in the presence of the Hsp70i with only approximately 50% cells resolving the spindle into a pseudo-bipolar state within the time-frame of the experiment (700 minutes); for those cells in which the spindle did resolve into a pseudo-bipolar state they spent a mean of 480 minutes in metaphase (Fig. 2A–C). Analysis of cells depleted of Hsp72 using the same approach gave similar results confirming the specific role of Hsp72 in this process (Fig. 2C). Likewise, inhibition of Hsp70 or depletion of Hsp72 in MDA-MB-231 cells led to multipolarity with a mean time of >500 minutes in metaphase (Fig. 2D). These data reveal that in these cell types the multiple MTOCs generated by the presence of amplified centrosomes are resolved to a pseudo-bipolar state as a result of centrosome clustering rather than centrosome inactivation. Moreover, they confirm that loss of Hsp72 function prevents this from occurring, thereby stopping cells from satisfying the SAC and proceeding into late mitosis.

Our previous studies revealed that Hsp72 promotes spindle assembly in mitotic cells downstream of the protein kinase Nek6 (29). To determine whether a similar pathway is involved in centrosome clustering, we depleted MDA-MB-231 cells of either Nek6 or the closely related Nek7 kinase. Although Nek6 and Nek7 have almost identical catalytic sites, they have distinct substrate specificities that result from their distinct N-termini (37), and Nek6 but not Nek7 interacts with Hsp72 (29). Each kinase was independently depleted with two different oligonucleotides and depletion confirmed by Western blot (Supplementary Fig. S1C). Whereas removal of Nek7 had no effect on centrosome clustering, depletion of Nek6 led to a substantial increase in the frequency of cells with amplified centrosomes that had multipolar spindles indicative of a failure in centrosome clustering (Fig. 3A and B). Strikingly, wild-type but not kinase-inactive Nek6 was capable of rescuing centrosome clustering in cells from which Nek6 had been depleted demonstrating the requirement for Nek6 kinase activity in this process (Fig. 3B).

In MDA-MB-231 cells that had been detergent-extracted prior to fixation to remove soluble protein, Hsp72 was observed at spindle poles consistent with previous observations in other cell types (29, 38–40). However, this localization was lost upon Nek6, but not Nek7, depletion despite no change in overall expression of Hsp72 (Fig. 3C and D; Supplementary Fig. S1D). Further evidence that Nek6 and Hsp72 cooperate in centrosome clustering came from demonstrating that chemical inhibition of either Aurora-A or Plk1, both of which act upstream of Nek6 and induce monopolar spindle formation (41, 42), prevent localization of Hsp72 to spindle poles without disturbing its expression (Fig. 3C and D and Supplementary Fig. S1E and S1F). Similar loss of Hsp72 localization upon Aurora-A or Plk1 inhibition was observed in NIE-115 cells (Fig. 3E). Thus, we propose that a pathway acting through the Aurora-A, Plk1, and Nek6 kinases targets Hsp72 to the spindle apparatus.

As a potential strategy for targeting cancer cells with amplified centrosomes, we compared the phenotypes generated upon loss of Hsp72 or Nek6 function, with the action of a well characterized centrosome declustering agent, griseofulvin (43). As expected, griseofulvin induced generation of multipolar spindles in MDA-MB-231 cells with amplified centrosomes (Fig. 4A and B). However, it led to an even higher frequency of multipolar spindles in HeLa cells, which do not exhibit substantial centrosome amplification (Fig. 4A and B). Analysis of these cells with centriole markers revealed that griseofulvin induces formation of microtubule asters in mitotic cells that lack centrioles, so called “acentrosomal” poles (Fig. 4C and D). In contrast, MDA-MB-231 cells treated with the Hsp70i, or depleted of Hsp72 or Nek6, did not form acentrosomal poles, with all poles in the multipolar spindles containing centrioles (Fig. 4C and D).

Acentrosomal poles also arise in mitotic cells following nocodazole-induced microtubule depolymerization and subsequent nocodazole washout. However, these quickly resolve to form a bipolar spindle. To observe, how cells with acentrosomal poles respond in the absence of Hsp70 activity, NIE-115, MDA-MB-231, and HeLa cells were treated for 16 hours with nocodazole and then released into nocodazole-free media in the presence or absence of the Hsp70i (Supplementary Fig. S2A). In the absence of Hsp70i, NIE-115 cells initially had multiple microtubule asters but these largely resolved within 60 minutes to two asters (Fig. 4E and Supplementary Fig. S2B). However, in the presence of the Hsp70i, these multiple asters did not resolve and the percentage of cells with >2 asters remained relatively constant. With the MDA-MB-231 cells, there was a partial resolution in the presence of the Hsp70i consistent with the lower frequency (∼35%) of cells with amplified centrosomes (Fig. 4F). In contrast, in HeLa cells that do not have amplified centrosomes, the multiple asters resolved to form a bipolar spindle at a similar rate in the presence or absence of Hsp70i (Fig. 4G and Supplementary Fig. S2C). Hence, Hsp70 is required to cluster microtubule asters organized by bona fide centrosomes, but not to cluster microtubule asters generated by acentrosomal poles.

To address the role of inhibiting Hsp70 in a particular cancer setting, we opted to assess a panel of ALL cell lines, on the basis that childhood ALL is a cancer typically treated with the microtubule poison, vincristine (44). This treatment is effective but associated with significant toxicity and risks of secondary cancers due to the lack of selectivity for cancer cells. For this analysis, ten ALL cell lines were analyzed as well as two noncancer-derived PBL cell lines. Despite being nonadherent cell lines with high nuclear-to-cytoplasmic ratios that make them suboptimal for imaging, we optimized techniques to assess centrosome number using antibodies against Cep135 to stain centrioles and either γ-tubulin or pericentrin to stain the PCM (33). Although PBLs had a very low frequency (<5%) of cells with amplified centrosomes, the majority of ALL cell lines exhibited centrosome amplification within a range of 10% to 30% cells (Fig. 5A and B). Treatment with the Hsp70i revealed a dose-dependent increase in the frequency of multipolar spindles that correlated strongly with extent of centrosome amplification (Fig. 5C and D). In contrast, griseofulvin induced multipolar spindle formation to the same extent in all cells, including the PBLs, due to generation of acentrosomal poles (Fig. 5E and F). Western blotting revealed increased expression of Hsp72 in most, but not all ALL cell lines compared with PBLs, while Nek6 was present at similar levels in ALL and PBL cell lines (Supplementary Fig. S1F). Hence, inhibiting the Nek6–Hsp72 pathway represents a potential route to selective targeting of ALL cells with amplified centrosomes.

Targeting the Nek6–Hsp72 pathway for cancer therapy would only improve selectivity if these proteins are not required for proliferation of normal cells. In fact, we have previously shown that this pathway is required for robust spindle assembly in HeLa cells arguing that their role extends beyond centrosome clustering in cancer cells (29). Importantly, we found that while Hsp70 inhibition blocked chromosome congression in both HeLa and MDA-MB-231 cells, treatment with the Hsp70i did not lead to chromosome congression defects in the noncancer-derived RPE1 and HBL-100 cell lines (Fig. 6A and B). Encouragingly, we also found that depletion of Hsp72 or Nek6 led to a similar level of chromosome congression defects in HeLa and MDA-MB-231 cells, but did not interfere with chromosome congression in RPE1 and HBL-100 cells (Fig. 6C). Measurement of mitotic index by analysis of condensed chromatin and staining for the phospho-histone H3 (pHH3) marker confirmed that Hsp70 inhibition caused a significant delay in mitotic progression in HeLa and MDA-MB-231 cells, but not in RPE1 and HBL-100 cells (Fig. 6D, and Supplementary Fig. S3A). Time-lapse imaging of RPE1 cells incubated with the SiR-tubulin probe also revealed no change in mitotic duration or spindle shape upon treatment with the Hsp70i or depletion of Hsp72 (Fig. 6E and Supplementary Fig. S3B and S3C). As expected, Western blotting revealed that the noncancer-derived cell lines, RPE1 and HBL-100, exhibited lower expression levels of the inducible Hsp72 protein than the cancer lines, MDA-MB-231 and HeLa (Supplementary Fig. S3D and S3E). Finally, we showed that while Hsp70 inhibition can perturb K-fiber assembly, it did not prevent SAC activation as determined by recruitment of BubR1 in nocodazole-treated MDA-MB-231 cells (Fig. 6F and G). Thus, we propose that the Nek6–Hsp72 pathway is required for spindle organization and centrosome clustering, but not SAC activity, in cancer cells but not essential for mitotic progression in normal cells. This pathway, therefore, represents an attractive target for development of novel anti-cancer therapies.

Centrosome amplification is frequently observed in cancer cells and associated with poor prognosis (3). To allow cell survival, the extra centrosomes are clustered to enable formation of a pseudo-bipolar spindle that can support chromosome segregation. Centrosome clustering has, therefore, emerged as a potentially exciting avenue for cancer drug discovery based on the principle of selective disruption of mitosis in cells with amplified centrosomes (18, 45).

Here, we demonstrate that the Nek6–Hsp72 axis, together with the upstream Aurora-A, Plk1, and presumably Nek9 kinases, represents a novel pathway for regulating centrosome clustering (Fig. 6H). Nek6-phosphorylated Hsp72 regulates the spindle association of the chTOG–TACC3 complex (29). This complex was previously implicated in centrosome clustering through its regulation by integrin-linked kinase (ILK), inhibition of which prevents centrosome clustering (30). Spindle association of chTOG–TACC3 also requires clathrin, a known binding partner of Hsp70 chaperones (46). Our working hypothesis is that phosphorylation of Hsp72 by Nek6 modulates the interaction of clathrin with TACC3 and/or microtubules to favor spindle association of the complex, thereby promoting K-fiber stability (47). However, there could equally be other Nek6 and/or Hsp72 targets associated with centrosome and/or kinetochore attachments (including motors, such as Eg5, HSET, or dynein) that promote tension in K-fibers. These mechanisms are under investigation.

A major challenge is to identify mechanisms that play an essential role in centrosome clustering but that can be safely inhibited in normal cells. Multipolar spindles can result from perturbation of many different spindle associated proteins, but there are few examples so far of proteins that show selective effects in cancer cells with amplified centrosomes and that are amenable to inhibition with a drug-like molecule. One example is HSET/KIFC1, a minus end-directed microtubule motor of the kinesin-14 family (48, 49). Multipolar spindle formation is induced upon treatment of cancer cells harboring amplified centrosomes with either of two distinct HSET inhibitors: AZ82, which blocks the ATP binding pocket, or CW069, which binds to an allosteric site. However, cells with a normal complement of centrosomes are still affected by these inhibitors and, whilst only transient, multipolarity can occur in these cells that may either trigger cell death or induce genetic instability.

One concern with certain compounds developed as centrosome declustering agents, such as griseofulvin and its derivatives (43), is that they induce formation of acentrosomal spindle poles, that is, poles that do not contain or arise from bona fide centrosomes. Similar acentrosomal poles are formed upon treatment of cells with Taxol-based drugs, and this could be a common feature of declustering agents that work by disturbing microtubule dynamics or motors that crosslink microtubules. The fact that we never saw acentrosomal poles upon loss of Hsp72 or Nek6 function argues that this pathway acts through some other mechanism. The fluorescent tubulin probe allowed us to directly monitor microtubule organization in live mitotic cells and revealed that it is resolution of a multipolar to bipolar state that is inhibited in cells with amplified centrosomes. This was supported by nocodazole washout experiments that showed acentrosomal poles formed in HeLa cells with normal centrosome numbers clustered efficiently in the presence of the Hsp70 inhibitor. These data rather argue for a specific role for this pathway at points of microtubule attachment to centrosomes, kinetochores, or the cell cortex and is reassuring from a therapeutic perspective as it demonstrates selectivity toward cells with amplified centrosomes.

There is further potential patient benefit in targeting the Nek6–Hsp72 pathway as both components are overexpressed in cancer versus normal cells (50, 51). Moreover, normal cells may have a reduced requirement for this pathway compared with cancer cells with no evidence of mitotic delays or segregation errors in either immortalized RPE1 or HBL100 cells. From a practical perspective though there are limitations in directly targeting Hsp72. First, different Hsp70 family members are highly similar in active site structure and it will be difficult to develop catalytic inhibitors that do not also target other family members with essential roles in normal cell homeostasis. For this reason, simply measuring cell survival in response to Hsp70 inhibition may well not reveal a direct correlation with centrosome number. Second, Hsp70 is proving to be a challenging drug target due to its high affinity for nucleotide. In contrast, having demonstrated a requirement for Nek6 activity for centrosome clustering, this kinase may well make a more attractive target than Hsp72. Indeed, while there is significant sequence similarity within the catalytic sites with other NEKs, particularly Nek7, we are developing approaches to generate selective inhibitors that target not only the catalytic site but also the more divergent allosteric sites required for activation (52).

At the same time as developing Nek6 inhibitors, we need to identify groups of patients who might benefit from centrosome declustering agents. Although inhibiting this pathway causes problems in mitotic progression in cancer cells even without centrosome amplification, we argue that extra centrosomes would sensitize cancer cells to mitotic catastrophe in a physiologic setting and are, therefore, a valuable and relatively easy to measure biomarker, especially in hematologic malignancies (4). Genetic changes known to cause centrosome amplification, such as Aurora-A overexpression or BRCA1 loss, may also represent useful biomarkers for selecting clinical trial cohorts (8). Here, we show that a subset of ALL cell lines harbor amplified centrosomes and most of these have the capacity to cluster the extra centrosomes to avoid generating multipolar spindles. Furthermore, treatment with an Hsp70 inhibitor led to loss of clustering with the frequency of multipolar spindles closely reflecting the extent of centrosome amplification. ALL cells represent a good model for a disease that has traditionally been treated with anti-mitotic chemotherapies and, although we haven't explored cell survival for the reasons given above, it could be that centrosome number acts as a biomarker for stratification. High levels of centrosome amplification have previously been observed in other hematologic malignancies, including DLBCL, MCL, and CML (4). Future work will be important to identify genetic signatures that dictate the response of blood cancers to centrosome declustering strategies and enable the clinical development of new targeted agents.

No potential conflicts of interest were disclosed.

Conception and design: J. Sampson, R. Bayliss, A.M. Fry

Development of methodology: J. Sampson, L. O'Regan, A.M. Fry

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Sampson, A.M. Fry

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Sampson, M.J.S. Dyer, A.M. Fry

Writing, review, and/or revision of the manuscript: J. Sampson, L. O'Regan, M.J.S. Dyer, R. Bayliss, A.M. Fry

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. O'Regan, A.M. Fry

Study supervision: L. O'Regan, M.J.S. Dyer, A.M. Fry

We thank Dr. Sandrine Jayne and Dr. Ildiko Gyory (Leicester) for assistance with ALL cell culture. We acknowledge support of the Advanced Imaging Facility of the University of Leicester Core Biotechnology Services.

This work was funded by grants from Hope Against Cancer (PHD/SELLICKS/BLN/2013), the Medical Research Council (MR/J003972/1), Worldwide Cancer Research (16-0119), The Wellcome Trust (097828/Z/11/Z) and Cancer Research UK (C1362/A18081) to A.M. Fry, and grants from Cancer Research UK (C24461/A23303) and BBSRC (BB/L023113/1) to R. Bayliss.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.