LIN9 and NEK2 Are Core Regulators of Mitotic Fidelity That Can Be Therapeutically Targeted to Overcome Taxane Resistance

Taxanes such as paclitaxel are a standard-of-care therapy for many malignancies. These drugs function as microtubule-stabilizing agents and while they can be highly effective, resistance is common. Identifying therapeutically targetable mechanisms underlying therapeutic resistance will be necessary to improve long-term survival. At clinically relevant doses, paclitaxel induces multipolar spindles and chromosomal missegregation, leading to chromosomal instability (CIN) and mitotic catastrophe (1, 2). These errors manifest as micro- and multinucleated cells, often with abnormally shaped nuclei (1, 3). Paclitaxel also directly inhibits centrosome separation during mitosis, preventing chromosome detachment from microtubule minus-ends (4). This aberrantly promotes movement of centrosomes away from nucleating chromosomes and disrupts the proper positioning of spindle poles during mitosis (4, 5). As a result of these defects, taxanes induce extensive mitotic abnormalities, resulting in mitosis-associated cell death or mitotic catastrophe.

Mechanisms of taxane resistance are not fully understood. They include β-tubulin overexpression or mutations (6), overexpression of P-glycoprotein and other transport or efflux proteins (7), abnormalities in mitotic and spindle checkpoints (8), modulation of pro- or antiapoptotic proteins (9), and overexpression of Aurora kinase A (AURKA) (10). Thus far, none of these targets have produced effective therapeutic approaches in patients for preventing resistance or restoring sensitivity to taxanes. It is notable that combining atezolizumab, an anti-PD-L1 antibody that acts as a checkpoint inhibitor, with nab-paclitaxel has recently been shown to improve outcomes of patients with triple-negative breast cancer (TNBC; ref. 11), but the basis for this improvement likely extends beyond the specific mechanisms underlying taxane resistance and may be attributable to generation of neoantigens. In this study, we used TNBC as a model for identifying a therapeutically targetable mediator of taxane resistance.

TNBC is associated with poor patient survival. While the mean time to recurrence of all breast cancer subtypes is 5 years, the mean for TNBC is 2.6 years (12). Taxanes are one of the most commonly used cytotoxic chemotherapies for patients with TNBC in neoadjuvant, adjuvant, and metastatic disease settings, alone or in combination with other antineoplastics such as gemcitabine or anthracyclines (13). While initially effective in this disease, taxanes are hampered by dose-limiting toxicities and rapid development of resistance. Of the few treatments available for patients with TNBC who have developed taxane resistance, less than 25% of resistant tumors will respond, with the average response lasting less than 6 months (14). Identifying therapeutically targetable mechanisms of taxane resistance should significantly expand the number of patients that can be effectively treated with these drugs.

As indicated above, taxanes primarily function by disrupting the cell cycle, particularly the mechanical events of mitosis. On a transcriptional level, cell-cycle progression is controlled by the MuvB complex. This complex is comprised of five subunits (LIN9, LIN54, LIN52, LIN37, and RBBP4) that together bind and regulate expression of cell-cycle genes (15). We previously reported that overexpression of LIN9, which encodes the core scaffold of MuvB, is associated with worse outcomes in breast cancer and that LIN9 is overexpressed in approximately 65% of TNBC cases (16). Moreover, suppressing LIN9 induces multinucleation and subsequent apoptosis or senescence of TNBC cells. Herein, we report the discovery of a novel, druggable mechanism underlying taxane resistance in TNBC that involves upregulation of LIN9 and its downstream transcriptional target, NEK2, a centrosomal kinase. Genetically suppressing LIN9 or NEK2 causes profound mitotic defects that synergize with taxanes to induce cell death. Most importantly, therapeutically targeting the LIN9/NEK2 pathway restores taxane sensitivity in resistant cells and xenografted tumors. These data provide a new mechanism-based, two-pronged approach to induce excessive mitotic progression errors in TNBC and ensure taxane response that may be useful for improving patient outcomes.

Additional methodologic details may be found in Supplementary Materials.

All cell lines were acquired from the ATCC and cultured at 37°C with 5% CO₂. Cells were authenticated using short tandem repeat profiling (BDC Molecular Biology Core Facility, University of Colorado, Boulder, CO) or were purchased within 6 months from ATCC. MDA-MB-231, MDA-MB-468, HCC70, HCC38, and HCC1143 cell lines were maintained in RPMI1640 with 10% FBS. Insulin (0.023 IU/mL) was added to this media for the BT-549 cell line. SUM159 cells were cultured in Ham's F12 with 10% FBS, insulin (10 mg/mL), and hydrocortisone (1 mg/mL). SK-BR-3 cells were maintained in McCoy's 5A medium with 10% FBS. MCF7 cells were cultured in DMEM with 10% FBS. All cell lines were tested monthly for Mycoplasma pulmonis and Mycoplasma spp. according to the manufacturer's protocol (Bimake, B39032). Cells never exceeded ten passages after thawing. Paclitaxel (Selleckchem, S1150), docetaxel (LC Laboratories, D-1000), JQ1 (Cayman Chemical, 1268524-70-4), CMP3a (MedKoo, 2225902-88-3), and INH1 (R&D Systems) were dissolved in DMSO. Transient mRNA silencing was conducted using 100 nmol/L nontargeting siRNA (Dharmacon, D-001810-02-20) or siRNA targeting LIN9 (L-018918-01), NEK2 (L-004090-00-0020), and LIN37 (L-013311-02-0005) with Lipofectamine 2000 (Invitrogen, 11668-027) in Opti-MEM media (Invitrogen, 31985088) for 6 hours, after which, they were maintained in complete media for 24 hours. For paclitaxel and docetaxel dose–response curves, cells were treated with the indicated concentration of drug in addition to 250 nmol/L JQ1, 10 μmol/L INH1, or 3 nmol CMP3a for 4 days. Viable cells were counted by Trypan blue exclusion on a Countess II FL (Thermo Fisher Scientific, AMQAF1000).

LIN9 (Hs00542748_m1), NEK2 (Hs05021038_g1), ABCB1 (Hs00184500_m1)ABCC2, (Hs00960489_m1) ABCC3, (Hs00978452_m1) ABCG2, (Hs01053790_m1) TUBB1, (Hs00917771_g1), TUBB3 (Hs00801390_s1), and GAPDH (Hs02758991_g1) TaqMan Gene Expression Assays (Thermo Fisher Scientific) were used.

Primary antibodies are LIN9 (Thermo Fisher Scientific, PA5-43640), NEK2 (Bioss, bs-5732R and BD Biosciences, 610594), BcL-XL (Cell Signaling Technology, 2764), β-actin (Sigma, A5316), PARP (Cell Signaling Technology, 9542), and α-actin (Sigma, A1978 clone AC-15).

Cells were grown on coverslips and were fixed with 3.7% formaldehyde for 10 minutes and permeabilized with 0.1% Triton X-100. They were stained with Texas Red-X phalloidin (Invitrogen, T7471) in 1% BSA/PBS for 20 minutes. The slides were blocked for 1 hour in PBS containing 1% BSA, 10% normal goat serum, 0.3 mol/L glycine, and 0.1% Tween. γ-Tubulin primary antibody (Abcam, ab205475) was added at a 1:500 dilution in blocking solution overnight. VectaShield mounting medium with DAPI (Vector Laboratories, H-1500) was used to counterstain the nuclei. Cells were imaged using an inverted Leica fluorescence microscope.

Chromatin immunoprecipitation (ChIP)-PCR was performed as reported previously in MDA-MB-231 cells (17).

Cell-cycle analysis was performed as reported previously (18) with the following modifications: cells were fixed in 70% ethanol and analyzed using the Attune NxT Flow Cytometer (Thermo Fisher Scientific). Gating was performed during the analysis to remove doublets.

MDA-MB-231 cells were transfected with siNS, siLIN9, or siNEK2 (described above) and after 1 day, 1,000 live cells were seeded in 24-well plates. Each transfection was plated in duplicate. Cells were grown for 7 days before being fixed and stained in 0.05% crystal violet, 1% formaldehyde, and 1% methanol for 20 minutes at room temperature. To quantify staining, 10% acetic acid was added for 15 minutes. Sterile water was added 1:4. One-hundred microliters of this solution was read at 590 nm absorbance in duplicate.

Cells were treated for 4 days with either vehicle, 250 nmol/L JQ1, 6 nmol/L paclitaxel, or the combination. Cells were then stained with 10 μmol/L Hoechst (Thermo Fisher Scientific, 62249) for 10 minutes at room temperature. Cells with and without pyknotic nuclei were quantified to calculate the percent of apoptotic cells.

MDA-MB-468 cells were transfected with siNS, siLIN9, or siNEK2 and were imaged using the Incucyte Zoom System (Essen BioScience) starting 42 hours after transfection. Cells were imaged every 20 minutes from 36 to 96 hours after the start of image collection at ×20 magnification. Individual cells were tracked from the beginning to end of mitosis and assessed for the time needed to traverse through mitosis.

Tissue was fixed in 4% paraformaldehyde and paraffin-embedded. Antigen retrieval with citrate buffer pH 6.0 was performed before incubation with the Ki67 primary antibody (Abcam, ab66155) at 1:400 dilution. Slides were incubated in secondary antibody and DAB stained using the EnVision Detection Systems Peroxidase/DAB kit (Agilent, K406511-2), followed by hematoxylin counterstain (Vector Laboratories, H-3401). Scoring was blinded for all slides.

Cells were plated on glass coverslips in 6-well plates and transfected with siNS or siLIN9 as described above. After 72 hours, 100 μmol/L Z-VAD-FMK (550377) or the negative control 100 μmol/L Z-FA-FMK (550411) was added. VectaShield mounting medium with DAPI (Vector Laboratories, H-1500) was used to counterstain nuclei. Cells were quantified using an inverted Leica microscope.

All in vivo experiments were performed with approval from the Institutional Animal Care and Use Committee at Case Western Reserve University (Cleveland, OH), which is certified by the American Association of Accreditation for Laboratory Animal Care. Mice were housed in microisolator units, given standard sterile chow and water ad libitum, and maintained on a 12-hour light/dark cycle. MDA-MB-231 or MDA-MB-468 paclitaxel–sensitive cells, MDA-MB-231 or MDA-MB-468 paclitaxel–resistant cells, or patient-derived xenograft (PDX) TNBC models (TM00091 and TM00098-from The Jackson Laboratory) were xenografted into the two inguinal mammary fat pads of adult female NOD/scid/γ (NSG) mice. At least 10 tumors were evaluated per xenograft/treatment group. Once measurable tumors formed (∼15 mm³), mice were randomized into treatment groups of at least 4 mice per group. For the MDA-MB-231–sensitive model, these groups were: vehicle (1:1 propylene glycol:water), INH1 [100 mg/kg intraperitoneally (i.p.) every other day], paclitaxel (15 mg/kg i.p. biweekly), or the combination, or vehicle (1:1 propylene glycol:water), JQ1 (50 mg/kg i.p. daily), paclitaxel (15 mg/kg i.p. biweekly), or the combination. For all other models, the treatment groups were: vehicle (1:1 propylene glycol:water), CMP3a (10 mg/kg, i.p. 5 days/week), paclitaxel (15 mg/kg i.p. biweekly), or the combination. Tumor size was measured twice per week using calipers. Mouse weight was measured once per week to assess toxicity. After various treatment times (MDA-MB-231–sensitive INH1 model, 16 days; MDA-MB-231–sensitive JQ1 model, 15 days; MDA-MB-231–resistant model, 21 days, MDA-MB-468–sensitive model, 17 days of treatment; MDA-MB-468–resistant model, 28 days; PDX TM00091, 27 days; PDX TM00098, 29 days) that were limited by the extent of tumor growth occurring in vehicle-treated mice, tumors were removed and processed for analysis.

Statistical analyses were performed using two-tailed Student t test for all in vitro data and Mann–Whitney U test for all in vivo data. Significance was concluded if the P value was less than 0.05. Unless noted otherwise, all in vitro data are represented as means with SDs of three independent experiments each completed in triplicate.

We previously reported that LIN9 is overexpressed in the majority of TNBCs and its suppression induces multinucleation (16). Further examination of the expression pattern of LIN9 across breast cancer subtypes revealed that it is most highly expressed in the basal subtype with a median increase of 3.1-fold compared with the normal-like subtype in the The Cancer Genome Atlas (TCGA) dataset (Fig. 1A; ref. 19). While suggesting that cancers with high LIN9 may depend on it for maintaining ploidy, the specific roles of this protein in sustaining the growth and aggressiveness of TNBC are unknown. We used live-cell imaging to quantify both the duration of mitosis and cell fate with and without LIN9 silencing in TNBC cells. Reducing LIN9 expression led to a significant (P < 0.05) increase in the duration of mitosis (Fig. 1B; Supplementary Fig. S1A). LIN9 silencing also resulted in an 8.6-fold increase in the number of cells that enter a prolonged interphase and, after successfully completing one round of mitosis, did not divide again (Fig. 1C). These data indicate that sustained expression of LIN9 is necessary for normal mitotic progression of TNBC cells.

The mitotic defects induced by LIN9 silencing are similar to those observed in paclitaxel-treated breast cancer cells (1–3) and we postulated that LIN9 expression may contribute to paclitaxel response. Supporting this possibility, we found that LIN9 protein expression measured by Western blot analysis is positively correlated (R² = 0.6528) with the intrinsic IC₅₀ for paclitaxel across nine breast cancer cell lines (Fig. 1D). We then determined whether expression of LIN9 is elevated in two TNBC cell lines representing basal (MDA-MB-468) and mesenchymal (MDA-MB-231) subtypes with acquired paclitaxel resistance that were generated by continually increasing paclitaxel exposure over several months. The MDA-MB-231 and MDA-MB-468–sensitive/parental cell lines had IC₅₀ values of 7 nmol/L and 6 nmol/L for paclitaxel, respectively, while the IC₅₀ of the resistant derivatives was increased to 60 nmol/L. Both LIN9 mRNA and protein were increased approximately 2-fold (Fig. 1E–G) in resistant compared with sensitive/parental cells, indicating that resistance is associated with elevated LIN9 expression.

To assess the role of LIN9 in paclitaxel resistance, we used RNAi to determine whether changing its expression can alter paclitaxel potency. Both sensitive/parental and resistant cells were transfected with either siLIN9 or a nonsilencing control (siNS) and treated with increasing concentrations of paclitaxel. LIN9 silencing shifted the dose–response curve of sensitive/parental cells to the left, reducing the IC₅₀ by 50% (from 7 to 4 nmol/L) in MDA-MB-231 and 67% (from 6 to 2 nmol/L) in MDA-MB-468 cells (Fig. 1H–L). More strikingly, silencing LIN9 in the resistant derivatives restored paclitaxel sensitivity to a level that was similar to parental/sensitive cells, with a 3- to 10-fold shift in the IC₅₀ from 48.2 to 4.06 nmol/L in MDA-MB-231 cells and from 10.72 to 3.59 nmol/L in MDA-MB-468 cells. LIN9 silencing also increased sensitivity to paclitaxel in an intrinsically resistant TNBC cell line (BT-549, Fig. 1M and N). To determine whether the impact of LIN9 was generalizable to taxanes as a class, we treated MDA-MB-231–sensitive/parental and resistant cells with docetaxel and found that suppressing LIN9 expression similarly restores sensitivity to this drug (Supplementary Fig. S1B). The effect of LIN9 on taxane sensitivity was not a general response to modulating MuvB because genetically silencing LIN37, another component of the complex, did not alter paclitaxel response (Supplementary Fig. S1C and S1D). These data indicate that LIN9, but not all members of MuvB, controls paclitaxel sensitivity in TNBC cells. It also suggests that resistant cells may depend on LIN9 expression to minimize catastrophic mitoses induced by taxanes. To determine whether elevated LIN9 is sufficient to induce taxane resistance, we used standard and inducible overexpression paradigms in MDA-MB-231 and MDA-MB-468 cell lines; however, all induced rapid cell death. Because LIN9 expression in resistant cells is just 2-fold greater than sensitive/parental cells, and approximately 3-fold higher in TNBC compared with normal-like tumors, these data suggest that a finely tuned level of LIN9 is necessary to ensure TNBC cell viability.

Several mechanisms of taxane resistance have been identified previously (6–10). We assessed whether these may promote resistance in the models used herein. However, we found that none of the four ABC transporters (ABCB1, ABCC2, ABCC3, or ABCG2) nor TUBB1 or TUBB3 were differentially expressed in resistant versus sensitive/parental cells (Supplementary Fig. S1E–S1H). Furthermore, none of the previously described mutations in TUBB1 were detected. These data indicate that the role of LIN9 in resistance is independent of these previously characterized factors.

Sustained expression of LIN9 is necessary for mitotic progression and taxane resistance. Because taxanes induce tumor cell death by causing chromosomal missegregation (20), we postulated that LIN9 may promote proliferation, but suppress mitotic progression errors and chromosomal instability even in the absence of taxane exposure. Following transient transfection with siRNA to LIN9 or a nonsilencing control, we quantified four characteristics of mitotic dysfunction and chromosomal instability: supernumerary (3+) centrosomes, micronuclei, and multiple and dysmorphic nuclei. LIN9 silencing in sensitive/parental MDA-MB-231 and MDA-MB-468 cells causes a significant increase in all four characteristics of chromosomal instability (Fig. 2A–E; Supplementary Fig. S2A–S2E). In some cases, we also observed fragmented nuclei following LIN9 silencing. We classified these as dysmorphic and not apoptotic because treatment with a caspase inhibitor, Z-VAD-FMK, in the presence of LIN9 silencing did not prevent the acquisition or number of fragmented nuclei (Supplementary Fig. S2F and S2G). Notably, we also found that taxane-resistant cells have a basal increase in mitotic errors compared with sensitive cells, even when cultured in the absence of paclitaxel and without LIN9 silencing (Fig. 2, Res. siNS compared with Sens. siNS), yet these cells are viable. This suggested that resistant cells may acquire an ability to suppress apoptosis as a means to tolerate moderately increased genomic defects. Indeed, BCL-xL, an antiapoptotic protein, was increased approximately 3-fold in resistant compared with sensitive cells (Supplementary Fig. S2H–S2J). BCL-xL was also modestly increased in sensitive cells when LIN9 was silenced, suggesting that BCL-xL upregulation is an intrinsic response to increased genomic instability.

Finally, we asked whether the increase in LIN9 expression that occurs in resistant cells may be an adaptive response that allows a low level of genomic defects but suppresses accumulation of rampant chromosomal instability. Indeed, we found that reducing LIN9 expression in taxane-resistant cells further increased the extent of mitotic progression errors (Fig. 2A–E). Together, these data indicate that LIN9 is necessary to sustain cancer cell fitness and suppress genomic defects, both in taxane-sensitive and -resistant cells.

The ability of LIN9 silencing to restore taxane sensitivity in resistant cells suggests that suppressing its activity may provide a means for ensuring taxane response in patients. However, due to its function as a scaffolding protein that controls transcription (15), LIN9 will likely be difficult to therapeutically target. We previously reported that JQ1, a Bromodomain and ExtraTerminal protein inhibitor (BETi), repressed LIN9 (16) expression and confirmed this finding in MDA-MB-231 cells (Fig. 3A and B). To determine whether BETi suppression of LIN9 levels can potentiate taxane response, sensitive/parental and resistant TNBC cells were treated with paclitaxel in the presence and absence of JQ1. Using the Chou–Talalay approach, we treated cells with escalating doses of combined JQ1 and paclitaxel. This revealed that JQ1 only minimally synergizes with paclitaxel to inhibit growth of taxane-sensitive cells (CI values of 0.880 and 0.955, respectively). In contrast, these two drugs were robustly synergistic in the taxane-resistant lines (CI values of 0.198 and 0.366, respectively; Fig. 3C). To determine whether BETi would mimic the effects of LIN9 silencing on the paclitaxel dose–response relationship (Fig. 1H–J; Supplementary Fig. S1E), taxane-sensitive and -resistant TNBC cells were treated with a minimally effective concentration of JQ1 (250 nmol/L) and increasing doses of paclitaxel. JQ1 recapitulated the effects of LIN9 silencing by shifting the paclitaxel dose–response curve to the left for both sensitive/parental and resistant cells. However, adding JQ1 was much more impactful when examining paclitaxel-resistant cells (Fig. 3D; Supplementary Fig. S3A). In addition, treating resistant cells with the combination of paclitaxel (6 nmol/L, IC₅₀ of paclitaxel-sensitive cells) and JQ1 (250 nmol/L) resulted in a dramatic increase in multi- and micronucleated cells compared with either drug alone (Fig. 3E and F; Supplementary Fig. S3B). This result was similar to that observed with LIN9 silencing in the presence of paclitaxel (Fig. 2B and C; Supplementary Fig. S2B and S2C), and further supports the role of LIN9 in repressing excessive genomic instability and ensuring cell fitness and viability. Indeed, treating resistant cells with combined paclitaxel and JQ1 caused a substantial increase in the percent of sub-G₀ cells compared with either drug alone (Fig. 3G; Supplementary Fig. S3C). Hoechst staining further revealed that combining JQ1 with paclitaxel induced greater apopotic cell death than either drug alone (Fig. 3H; Supplementary Fig. S3D).

Having observed the ability of BETi to restore paclitaxel sensitivity in resistant TNBC cells, we attempted to assess the efficacy of combining JQ1 with paclitaxel in vivo. However, the drug combination caused substantial mouse weight loss with multiple dosing strategies, indicating toxicity (Supplementary Fig. S3E). This is likely due to adding a global transcriptional inhibitor (BETi/JQ1) with a significant toxicity profile of its own, to a cytotoxic drug (paclitaxel). Thus, we turned to identifying a downstream transcriptional target and mediator of LIN9, with the goal of discovering a more precise and less toxic approach for modulating its impact on paclitaxel response.

To identify direct transcriptional targets of LIN9 that may modulate breast cancer progression and be druggable, we used an in silico approach with the following criteria: (i) candidate target mRNAs must correlate with LIN9 mRNA in human breast cancer (positive Pearson correlation greater than 0.5 with LIN9 in TCGA and METABRIC; refs. 19, 21, 22), (ii) candidate genes must contain a LIN9-binding site (using the only publicly-available ChIP-seq dataset for LIN9; ref. 23), and (iii) high expression of candidate genes must associate with reduced breast cancer survival. This approach yielded nine candidates (NEK2, DTL, CENPF, CKAP2L, RBBP5, NUSAP1, EXO1, DKC1, and ASPM). Because LIN9 silencing causes abnormalities in centrosome composition (Fig. 2A; Supplementary Fig. S2A) similar to those observed with taxane treatment (2, 4), we further refined this list to four candidates that are established regulators of centrosome function (NEK2, DTL, CENPF, and CKAP2L). Of these, only NEK2 (NIMA-related Kinase 2) encodes an enzyme for which inhibitors currently exist, making it readily druggable. NEK2 is a serine-threonine centrosomal kinase that controls mitotic progression by ensuring proper centrosome separation and bipolar spindle formation (24, 25). Dysregulated NEK2 expression leads to a variety of mitotic failures including abnormal spindle formation, supernumerary centrosomes, micro- and multinucleation, aneuploidy, and mitotic catastrophe (26, 27). Several of these phenotypes are similar to those occurring with LIN9 silencing, suggesting that NEK2 may be a major mediator of the effects of LIN9 on mitosis.

Like LIN9, NEK2 is more highly expressed in the basal subtype of breast cancer compared with normal-like tumors (TCGA data; Fig. 4A; refs. 28, 29). Meta-analyses of a collection of breast cancer gene expression datasets revealed that higher expression of NEK2 correlates with a lower probability of disease-specific and relapse-free survival across the diverse cohort of patients with breast cancer (Fig. 4B; Supplementary Fig. S4A). Expression of LIN9 and NEK2 are also highly correlated in breast cancers (Fig. 4C; Supplementary Fig. S4B), suggesting that LIN9 may regulate NEK2 gene expression. Indeed, transiently silencing LIN9 decreased NEK2 mRNA and protein in parental MDA-MB-231 and MDA-MB-468 cells (Fig. 4D–F). Analysis of a LIN9 ChIP-seq dataset indicated that LIN9 binds to the NEK2 gene in HeLa cells (Supplementary Fig. S4C) and ChIP/PCR confirmed LIN9 binding to the NEK2 gene in MDA-MB-231 cells similar to its interaction with a positive control, AURKA (Fig. 4G). Finally, we treated cells with JQ1 to reduce LIN9 levels, and this also decreased NEK2 protein expression (Fig. 4H and I). To determine whether NEK2 is downregulated by JQ1 in a LIN9-dependent manner, we silenced LIN9 and treated with either DMSO or JQ1. We found that JQ1 treatment does not decrease NEK2 to a greater extent than LIN9 silencing alone, indicating it does not independently regulate NEK2 (Supplementary Fig. S4D). These data indicate that NEK2 is a transcriptional target of LIN9 that is associated with worse breast cancer patient outcomes.

Because it is a downstream target of LIN9, it was not surprising that NEK2 mRNA and protein levels were also elevated with the acquisition of taxane resistance in both MDA-MB-231 and MDA-MB-468 TNBC cell lines (Fig. 4J–L). Live-cell imaging also revealed that similar to LIN9, reducing NEK2 expression increased the duration of mitosis (Fig. 4M–O) as well as the percentage of cells that entered prolonged interphase or that died following mitotic exit (Fig. 4P). NEK2 silencing did cause a higher rate of death following mitotic exit compared with LIN9 silencing, while LIN9 silencing resulted in more cells that entered prolonged interphase. This difference is likely due to the ability of LIN9 to regulate the expression of many target genes other than NEK2, some of which may suppress cell death such as BCL-xL (Supplementary Fig. S2J). Together, these data indicate that mitotic fidelity is dependent upon the LIN9-NEK2 transcriptional pathway and that its upregulation may convey a level of genomic instability that promotes the tumorigenic potential of cells and taxane resistance while also ensuring cell fitness and viability.

To determine whether transient loss of LIN9 or NEK2 has a sustained impact on cell viability and growth, we transfected sensitive/parental cells with siLIN9, siNEK2, or a nonsilencing control and assessed colony formation after 8 days. Reducing either LIN9 or NEK2 caused a profound decrease in colonies (Fig. 4Q; Supplementary Fig. S4E and S4F), underscoring the ability of these two proteins to phenocopy one another. In addition, these data indicate that a transient loss of LIN9 or NEK2 activity can have a sustained impact on tumor cell growth, further supporting their potential utility as therapeutic targets.

To determine whether loss of NEK2 generates similar mitotic dysfunction and chromosomal instability phenotypes as LIN9 silencing in the absence of taxane exposure, paclitaxel-sensitive/parental and resistant MDA-MB-231 cells were transiently transfected with siRNA targeting NEK2 or a nonsilencing control. Reducing NEK2 (Fig. 5A) caused an increase in supernumerary centrosomes as well as micro-, multi-, and dysmorphic nuclei in sensitive and resistant cells, with the effects being more pronounced in resistant cells (Fig. 5B–E). Notably, the extents of these changes were identical to those observed with LIN9 silencing (data included from Fig. 2A–D to facilitate direct comparison), revealing that NEK2 silencing phenocopies the mitotic defects observed with the loss of LIN9. Comparable results were observed in MDA-MB-468 cells (Supplementary Fig. S5A–S5E). Notably, these defects occurred in the absence of taxane exposure, again indicating that the reprogramming that occurs during the acquisition of resistance makes cells particularly sensitive to disrupting the LIN9-NEK2 cascade.

To evaluate the potential clinical significance of NEK2 in taxane response, we assessed NEK2 levels in a gene expression dataset of breast cancers prospectively collected prior to neoadjuvant treatment with a taxane/anthracycline regimen (30). For TNBC tumors that underwent a pCR, NEK2 levels were unable to distinguish long-term clinical outcomes (Fig. 5F). In contrast, for patients with TNBC that had residual disease following neoadjuvant treatment, basal elevation of NEK2 expression was highly prognostic for shorter relapse-free survival (Fig. 5G). Similar results were observed when examining the broader group of patients with breast cancer regardless of subtype (Supplementary Fig. S5F and S5G). These data suggest that NEK2 may be a predictive biomarker for taxane/anthracycline response. They also support the possibility that NEK2 may be a primary effector of LIN9 that controls taxane efficacy in breast cancer.

To determine whether NEK2 could also phenocopy the impact of LIN9 on taxane sensitivity, we examined the effects of NEK2 silencing on the paclitaxel dose–response relationship in MDA-MB-231 and MDA-MB-468 cells. As with LIN9 suppression, NEK2 silencing modestly shifts the dose–response curve of sensitive/parental cells to the left (Fig. 5H; Supplementary Fig. S5H). In contrast, when examining resistant cells, NEK2 silencing has a substantial impact on the IC₅₀, fully restoring taxane responsiveness of resistant cells to an indistinguishable level from the sensitive/parental cells. Together, these data support the hypothesis that NEK2 is a transcriptional target of LIN9 that potentiates its effects on mitotic progression and taxane sensitivity.

LIN9 or NEK2 suppression causes profound mitotic progression errors in TNBC cells that are either sensitive or resistant to taxanes (Figs. 2 and 4). To determine whether these defects would be even more severe in the presence of paclitaxel, sensitive/parental and resistant cells were transfected with siRNAs to LIN9 or NEK2 and then treated with 6 nmol/L paclitaxel. As expected, LIN9 or NEK2 silencing in sensitive/parental cells caused significant increases in supernumerary centrosomes as well as micro-, multi-, and dysmorphic nuclei (Fig. 6A–D). These effects were profoundly increased in paclitaxel-resistant cells, with the vast majority of cells (>90%) undergoing mitotic defects following LIN9 or NEK2 silencing in combination with paclitaxel (Fig. 6E–H). These data indicate that the acquisition of taxane resistance causes cells to become wholly dependent on LIN9 and NEK2 to sustain chromosomal stability in the presence of paclitaxel. These data also suggest that pharmacologically targeting these proteins should have a robust ability to ensure taxane responsiveness.

Targeting LIN9 remains challenging. The only inhibitors currently available are global regulators of transcription (BETi) that suppress the expression of many genes beyond LIN9. Moreover, combining the BETi, JQ1, with paclitaxel displayed significant toxicity in vivo. To more precisely target the mitotic impact of LIN9, we used selective inhibitors of NEK2 to determine whether these may provide a viable approach for improving taxane responsiveness. INH1 causes proteasome-mediated decline of NEK2 protein by disrupting its interaction with HEC1/NDC80 (31). Treatment with 10 μmol/L INH1 fully resensitized resistant cells to paclitaxel while having no impact on its own (Fig. 7A; Supplementary Fig. S6A). We confirmed the ability of pharmacologically inhibiting NEK2 to restore taxane responsiveness using a second inhibitor with a distinct mechanism of action. CMP3a is an inhibitor of NEK2 catalytic activity that represses glioma growth in mice (32), but whose ability to potentiate taxane response has not been assessed. Treating paclitaxel sensitive or resistant MDA-MD-231 cells with a low dose (3 nmol/L) of CMP3a alone had no impact on growth (Supplementary Fig. S6B). In contrast, this dose of CMP3a acted similarly as INH1 to modestly sensitize parental cells and profoundly improve the response of resistant cells to paclitaxel (Fig. 7B). These effects were nearly identical to that observed with NEK2 silencing (Fig. 5H). Notably, NEK2 inhibition was more efficacious at potentiating paclitaxel response than JQ1. This is likely due to differences in the specificity of the targets. While JQ1 regulates the transcription of many genes, including induction of p21, a suppressor of cell-cycle entry and inhibitor of taxane efficacy, NEK2 inhibitors are much more selective for processes associated with centrosomes and microtubules.

Pharmacologic inhibition of NEK2 also potentiated taxane response in several mouse models of TNBC, in vivo. Mice with orthotopic xenografts of sensitive/parental MDA-MB-231 cells were treated with vehicle, paclitaxel, INH1, or the combination of both drugs for 16 days. While each drug alone had modest effects on final tumor size, the combination substantially inhibited growth (Fig. 7C; Supplementary Fig. S6C) without causing significant weight changes, suggesting minimal toxicity (Supplementary Fig. S6D). Likewise, combining CMP3a with paclitaxel caused tumor regression of orthotopic xenografts of sensitive/parental MDA-MB-468 cells while either drug alone had minimal effects on final tumor size (Fig. 7D; Supplementary Fig. S6E). None of the mouse studies combining CMP3a with paclitaxel induced significant weight loss, suggesting this combination is also nontoxic (Supplementary Fig. S6F–S6I). We then assessed the efficacy of combining paclitaxel and a NEK2 inhibitor in orthotopic xenografts of taxane-resistant MDA-MB-231 cells. CMP3a was used because it appeared that it may be more efficacious than INH1 when combined with paclitaxel in the sensitive/parental xenograft models (Fig. 7C vs. 7D). In the resistant xenografts, neither paclitaxel nor CMP3a significantly inhibited tumor growth compared with vehicle. However, combining paclitaxel with CMP3a significantly reduced tumor growth (Fig. 7E; Supplementary Fig. S6J). Histologically, no differences were observed in the morphology of viable residual tumor or the percentage of tumor cell necrosis across treatment groups. However, proliferation as measured by Ki67 staining was significantly repressed in tumors treated with combined CMP3a and paclitaxel compared with vehicle or either drug alone (Fig. 7F). The combination of paclitaxel and CMP3a was also significantly more efficacious in reducing tumor volume in resistant MDA-MB-468 xenografts than either drug alone (Fig. 7G; Supplementary Fig. S6K). To provide a more clinically relevant assessment of the CMP3a/paclitaxel combination, we evaluated its efficacy in two PDX models. The TM00098 model has 2.5-fold higher LIN9 and NEK2 expression than TM00091 (Fig. 7H) providing an approach to determine whether higher LIN9/NEK2 may distinguish response to combining a NEK2 inhibitor with a taxane. We found that the model with lower LIN9 and NEK2 (TM00091) responded well to the combination (Fig. 7I; Supplementary Fig. S6L), but that response was not significantly greater than paclitaxel alone, likely due to variability in tumor response. In contrast, the model with elevated LIN9 and NEK2 (TM00098) was highly responsive to the drug combination versus either drug alone, with near-complete suppression of tumor growth in the combination group (Fig. 7J; Supplementary Fig. S6M). As another indicator of toxicity, we measured spleen weight and found no differences between groups (Supplementary Fig. S6N). Finally, we also assessed the ability of the combination to induce apoptosis by measuring PARP cleavage. As expected, combining the NEK2 inhibitor with a taxane greatly increased PARP cleavage compared with either drug alone in the more responsive (high LIN9/NEK2 expressing, TM00098) model, but was unable to induce such cleavage in the model (TM00091) with low LIN9/NEK2 (Supplementary Fig. S6O–S6Q). Taken together, these data suggest that elevated LIN9/NEK2 provides a targetable vulnerability that can be exploited in the presence of taxanes and support the future assessment of NEK2 inhibitors for their potential to improve taxane response in patients with TNBC.

Taxanes remain one of the most commonly used cytotoxic therapies for breast cancer. However, the high rate of resistance to these drugs necessitates that mechanism-based, multifaceted combination treatments be developed to ensure therapeutic response. Although combinations with anthracyclines, such as doxorubicin, or nitrogen mustards, such as cyclophosphamide, have been used, resistance is still a critical issue (33, 34). Here, we report that loss of LIN9 expression thwarts proper mitosis and increases taxane sensitivity in TNBC models. Taxane-resistant cells require LIN9 to maintain a functional level of mitotic errors, while also preventing excessive genomic instability that would lead to death upon taxane treatment. We also identified NEK2 as a key transcriptional target of LIN9 that phenocopies the effects of LIN9 on mitotic progression, chromosomal stability, and taxane responsiveness that is more readily druggable than LIN9. Elevated expression of NEK2 is also associated with recurrence in patients that have residual disease following neoadjuvant treatment with combined taxane and anthracycline therapy. These data suggest that NEK2 may be a predictive biomarker of patient outcomes postneoadjuvant taxane+anthracycline treatment. More importantly, they also provide support for combining NEK2 inhibitors with taxane therapy in the adjuvant setting. Indeed, we found that two different NEK2 inhibitors can potentiate the efficacy of paclitaxel in xenograft mouse models of TNBC.

LIN9 has been reported to be a tumor suppressor in colon carcinoma cell lines and in mouse and human fibroblasts (35–37). In contrast, we have identified protumorigenic properties of LIN9 in breast cancer. These include the amplification and overexpression of the LIN9 gene in the majority of TNBCs and the association of high LIN9 levels with worse breast cancer outcomes that we described previously (16). In addition, we now report that sustained LIN9 expression is necessary for normal mitotic progression and that reducing LIN9 sensitizes breast cancer cells to taxanes. The ability of LIN9 to have both tumor-suppressive and -promoting properties is likely due to its role in regulating mitosis, which is exquisitely dependent upon the precise temporal and quantitative expression of specific proteins. Indeed, other proteins that control mitosis and chromosomal stability have also been reported to be tumor suppressive as well as oncogenic. These include Aurora kinase A and Polo-like Kinase (38–40), with therapeutic inhibitors to each being clinically evaluated.

There are currently no therapies that directly target LIN9 with or without chemotherapy. However, the AURKA inhibitor, alisertib, is efficacious in patients with breast cancer following cytotoxic therapy (41). AURKA is required for mitotic spindle formation, chromosome alignment, centrosome separation, and cytokinesis (42). Loss of AURKA activity leads to many mitotic defects that we report also occur with LIN9 suppression. Notably, AURKA is a downstream target of LIN9 (16, 43), hence it is likely that inhibition of LIN9 could elicit similar effects in patients as AURKA inhibitors. We previously reported that BETi suppress the expression of both AURKA and LIN9 in TNBC, leading to mitotic catastrophe (16, 44). Thus, a rational approach to inhibiting LIN9/AURKA in TNBC could be the use of BETi. While our studies demonstrated that the BETi, JQ1, does potentiate paclitaxel efficacy in vitro, this led to significant toxicity in vivo. Hence, use of such a broad-scale inhibitor that suppresses the expression of hundreds to thousands of genes is unlikely to improve patient outcomes from taxane therapy. More selective targeting of LIN9-regulated genes/proteins may be more appropriate, and an ongoing phase II clinical trial (NCT02187991) is currently investigating the potential efficacy of combining alisertib with paclitaxel in patients with breast cancer. We report here that selective inhibition of NEK2, another downstream target of LIN9, did not cause obvious toxicity when combined with paclitaxel but did improve paclitaxel response in vitro and in vivo. This is likely due to the more specific impact of NEK2 inhibition on mitosis. Supporting the potential for selectively targeting NEK2 as a means to improve taxane response in patients with breast cancer, NEK2 gene expression correlates with worse patient outcomes overall, is upregulated with the acquisition of taxane resistance, and is associated with disease recurrence following taxane therapy. Notably, combining NEK2 inhibitors with paclitaxel appeared to be well-tolerated wherein no changes in mouse body mass or spleen weight were observed. However, it is possible that other systemic effects of this combination could occur that either contribute to suppressing tumor growth or cause undetected toxicity. For example, we have not determined whether the combination may have an effect on tumor-associated stroma or the immune system and preclinical studies examining such possibilities would help inform future therapeutic trials.

While selective NEK2 inhibition has not previously been reported to potentiate taxane responsiveness, overexpression of its partnering protein, HEC1/NDC80, was found to be associated with poor prognosis in a small cohort of patients with ovarian cancer and to modulate the efficacy of paclitaxel in ovarian cancer cell lines (45). Huang, and colleagues have also reported that a HEC1 inhibitor causes cell death in a variety of breast cancer cell lines (46). These data complement the results reported herein demonstrating that selectively targeting the NEK2 kinase can improve taxane response, particularly in resistant TNBC cells, by inducing extensive mitotic defects and pervasive chromosomal instability.

It is well established that TNBCs exhibit a higher level of aneuploidy and chromosomal instability than other breast cancer subtypes, and such instability is associated with an increased risk of metastasis and worse outcomes (47). Our studies revealed that TNBC cells that have acquired paclitaxel resistance have a basal increase in mitotic abnormalities compared with parental cells. Modest genomic changes associated with low levels of instability such as those that we observed in taxane-resistant cells may lead to increased tumor cell fitness that promotes growth and metastasis (47–49). However, extreme mitotic instability induces cell death stemming from mitotic catastrophe (47, 49). We found that targeting LIN9 or NEK2 in resistant cells causes greater mitotic defects than in sensitive/parental cells. In the presence of paclitaxel, nearly all resistant cells develop numerous defects associated with extreme chromosomal instability. Thus, combining two insults, that is, the increased instability associated with taxane resistance partnered with suppression of LIN9 or NEK2, causes greater instability and mitotic catastrophe than observed in sensitive/parental cells that have only one insult (LIN9 or NEK2 silencing). We postulate that the increased chromosomal instability associated with the development of resistance provides a specific vulnerability to inhibition of the LIN9/NEK2 pathway. Hence, use of a NEK2 inhibitor should prevent the acquisition of taxane resistance because any cells with a resistance phenotype should be particularly sensitive to such inhibitors.

In summary, we found that increased LIN9 expression is correlated with paclitaxel resistance in TNBC cells and that its silencing restores taxane sensitivity though the induction of extensive mitotic abnormalities. We identified NEK2 as a downstream mediator of LIN9 whose suppression also causes chromosomal instability and promotes taxane sensitivity. Importantly, inhibiting LIN9 or NEK2 either pharmacologically (JQ1 or CMP3a/INH1, respectively) or genetically sensitizes cells to paclitaxel and reverses paclitaxel resistance. The ability of NEK2 inhibitors to increase paclitaxel responsiveness was also observed in vivo. These studies provide a rational therapeutic approach for combating and overcoming the major therapeutic challenge of taxane resistance in patients with TNBC through the utilization of NEK2 inhibitors.

V. Varadan is a consultant/advisory board member and reports receiving a commercial research grant from Curis. R.A. Keri is a endocrinology associate editor and is a consultant/advisory board member for Endocrine Society Board of Directors. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.S. Roberts, J.M. Sahni, R.A. Keri

Development of methodology: M.S. Roberts, J.M. Sahni, M.K. Summers, R.A. Keri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.S. Roberts, M.S. Schrock, K.M. Piemonte, K.L. Weber-Bonk, D.D. Seachrist, L.J. Anstine, S.T. Sizemore

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.S. Roberts, J.M. Sahni, K.M. Piemonte, K.L. Weber-Bonk, S. Avril, S. Singh, S.T. Sizemore, V. Varadan, R.A. Keri

Writing, review, and/or revision of the manuscript: M.S. Roberts, J.M. Sahni, K.M. Piemonte, D.D. Seachrist, S. Avril, L.J. Anstine, S. Singh, S.T. Sizemore, V. Varadan, M.K. Summers, R.A. Keri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.S. Roberts, R.A. Keri

Study supervision: M.K. Summers, R.A. Keri

Other (performed IncuCyte live cell imaging experiments): M.S. Schrock

We thank Drs. Kerry Burnstein and Maria Julia Martinez for thoughtful discussions during the execution of this work. This work was supported by R01CA206505 and Velosano Bike to Cure (to R.A. Keri), F99/K00CA212460 (to J.M. Sahni), R01GM108743 (to M.K. Summers), P20CA233216 (to V. Varadan), American Brain Tumor Association Basic Research Fellowship supported by an anonymous corporate partner (to M.S. Schrock), and the Cytometry and Microscopy Shared Resource, and the Athymic Animal and Preclinical Therapeutics Shared Resource of the Case Comprehensive Cancer Center (P30CA043703 and S10OD021559).

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.