The oncogene cyclin D1 promotes bipolar spindle integrity under compressive force

The mitotic spindle is the bipolar, microtubule-based structure that segregates chromosomes at each cell division. Aberrant spindles are frequently observed in cancer cells, but how oncogenic transformation affects spindle mechanics and function, particularly in the mechanical context of solid tumors, remains poorly understood. Here, we constitutively overexpress the oncogene cyclin D1 in human MCF10A cells to probe its effects on spindle architecture and response to compressive force. We find that cyclin D1 overexpression increases the incidence of spindles with extra poles, centrioles, and chromosomes. However, it also protects spindle poles from fracturing under compressive force, a deleterious outcome linked to multipolar cell divisions. Our findings suggest that cyclin D1 overexpression may adapt cells to increased compressive stress, possibly contributing to its prevalence in cancers such as breast cancer by allowing continued proliferation in mechanically challenging environments.


Introduction
The spindle is the macromolecular machine that segregates chromosomes at each cell division.In mammalian cells, mitotic spindles are bipolar structures with one centrosome at each spindle pole.Errors in cell division are associated with genomic instability and disease, and aberrant spindles are hallmarks of cancer [1].Extra centrosomes [2][3][4], continuously evolving karyotypes known as chromosomal instability [5,6], and multipolar spindles are elevated in tumors across many tissues of origin and diverse cancer genotypes.Oncogenes can also induce defects in spindle assembly even in the absence of gross spindle abnormalities; for example, MYC overexpression prolongs mitosis and increases chromosome segregation errors [7].Paradoxically, while such multipolar, clustered pseudo-bipolar, or otherwise aberrant spindles are generally adverse for mitotic outcomes [8,9], they can promote tumorigenesis by increasing genetic diversity [1] and potentially other unknown mechanisms.How oncogenic transformation affects spindle assembly remains poorly understood.
Dividing cells in solid tumors are subject to dramatically different mechanical environments than their counterparts in healthy tissue [10][11][12].Spindle poles in dividing cultured cells often fracture under compressive force, leading to mitotic delays, multipolar anaphases, and subsequent cell death [13][14][15][16].Tumors have been shown to be confining microenvironments due to their increased cell density, elevated interstitial fluid pressure [17], and increased extracellular matrix deposition and crosslinking [18], raising the question of how cells continue to divide under this high compressive stress.In breast tumors, compressive stress is high enough to deform and damage interphase nuclei [19], and nearby mitotic cells presumably experience similarly high forces that may interfere with mitotic rounding or spindle assembly.In multicellular tumor spheroid models, compressive stress reduces cell proliferation [20][21][22][23] and has been shown to disrupt bipolar spindle assembly in cells that continue to divide [24].Due to the challenges of making controlled mechanical perturbations at the cellular scale, the mechanisms underlying the spindle's mechanical integrity remain poorly understood [25].Little is known about whether and how the spindles of transformed cells mechanically differ from wild-type spindles as they adapt to the tumor environment.
Cyclin D1, overexpressed in 50-70% of breast cancers [26], is an oncogene with pleiotropic effects in the cell.Acute overexpression of cyclin D1 leads to spindle and karyotypic defects [27], and long-term overexpression is sufficient to drive breast cancer in mice [28].In addition to its canonical role in complex with CDK4/6 in controlling cell cycle progression at the G1/S transition, cyclin D1 may contribute to tumorigenesis through its roles in cytoskeletal remodeling and CDK-independent transcriptional programs [26].Many other oncogenes commonly dysregulated in breast cancer, such as Ras and ErbB2, are upstream of cyclin D1 [29][30][31], making cyclin D1 overexpression a good model to probe changes in spindle mechanics after oncogenic transformation.
Here, we compare control and cyclin D1-overexpressing breast epithelial cells to investigate their spindle architectures and responses to compressive stress.We find that cyclin D1 increases the proportion of spindles containing extra poles, chromosomes, and centrosomes.However, cyclin D1 overexpression also promotes bipolar spindle integrity during cell compression, preventing spindle pole fracture that results in multipolar cell divisions.We propose that cyclin D1 overexpression mechanically adapts cell division to compressive environments, potentially contributing to its prevalence in cancer despite the aberrant spindles it induces.

Constitutive cyclin D1 overexpression promotes aberrant spindle architectures
To determine the effects of cyclin D1 overexpression on spindle architecture, we compared MCF10A breast epithelial cell lines stably overexpressing cyclin D1 or a puromycin resistance gene as a control [32].The parental MCF10A cells are diploid and non-transformed, but are sensitive to transformation by a variety of oncogenes [32][33][34].We confirmed overexpression of cyclin D1 by western blot (Fig 1A ), and used immunofluorescence to quantify spindle pole, centriole, and kinetochore numbers by staining for α-tubulin, centrin, and CREST respectively (Fig 1B ).
While most control spindles (94%) had two centrioles at each of two spindle poles, supernumerary centrioles were more common in the cyclin D1-expressing cells (20% of cells; Fig 1C).These centrioles were either associated with multipolar spindles or clustered into pseudo-bipolar spindles, a known mechanism by which cancer cells adapt to extra centrosomes in order to avoid multipolar divisions [35][36][37].To gain insight into cyclin D1's effect on genomic integrity, we next counted the kinetochores in each spindle.Several mechanisms, including the clustering of extra centrosomes [8,9] and reduced kinetochore-microtubule dynamics [38], have been shown to give rise to aneuploidy and chromosomal instability in cancer cells, while cytokinesis failure leads to larger-scale genomic duplications.Cyclin D1 overexpression was associated with a broader range of chromosome numbers than in controls, with only a small number of cells containing a near-doubling of chromosome number, indicating that it induces aneuploidy (Fig 1D).The average chromosome number increased from 95.4 in control cells to 100.3 in cyclin D1-overexpressing cells, a level of aneuploidy that we do not expect to affect spindle size.In summary, constitutive overexpression of the oncogene cyclin D1 leads to an increased incidence of spindles with extra poles, centrioles, and chromosomes, even when cells are allowed to adapt to elevated cyclin D1 over many passages.

Cyclin D1 overexpression promotes bipolar spindle integrity under compressive stress
Although we observed that cyclin D1 overexpression gave rise to higher rates of spindle defects in cultured cells (Fig 1), cyclin D1 is known to be overexpressed in many tumors such as breast cancers.Because cells in breast tumors have also been reported to be subject to increased compressive stress [12,19,26], we wondered whether cyclin D1 may be adaptive in the context of a compressive solid tumor.Thus, we hypothesized that cyclin D1 overexpression might alter the spindle's biophysical properties to improve mitotic outcomes in cells dividing under compressive stress.We compared the mechanical robustness of control versus cyclin D1-overexpressing spindles by compressing cells in PDMS-based microfluidic devices and performing live imaging (Fig 2A) [39].Cells were pre-treated with the proteasome inhibitor MG132 to prevent anaphase entry, allowing us to focus on the metaphase spindle's response to compressive stress, and gradually compressed to a final height of 5 μm via a computer-controlled vacuum pump over 4 minutes.Compression was then sustained for an additional 70 minutes.This perturbation was reproducible from cell to cell, reducing spindle height from an average of 10.55 ± 1.54 μm to 4.72 ± 0.31 μm (mean ± standard deviation of all cells) (Fig 2B and 2C).Spindles in control and cyclin D1-overexpressing cells had indistinguishable average heights prior to compression and were compressed to a similar final height (Fig 2C).Spindles also widened and elongated as compression was applied, consistent with previous work [40][41][42].Spindle lengths before compression were similar between the control and cyclin D1 cells, as were spindle lengths at 10 minutes post-compression onset, when spindle shape had stabilized (Fig 2D).Spindles were wider in control cells vs. cyclin D1-overexpressing cells, both before and after compression, but the difference was slight (Fig 2E).Thus, our assay probes how cyclin D1 overexpression affects the spindle's intrinsic ability to adapt to a well-defined confined geometry, rather than probing cellular shape change in response to a defined compressive stress.
During compression experiments, we monitored changes in spindle integrity in addition to changes in spindle shape.Control spindle poles fractured into multiple foci during the 74 minutes of compression 47.4% of the time, with kinetochore-fibers detaching and splaying laterally from the original spindle pole (Fig 3A and 3B; S1 Movie).Interestingly, bipolar spindles in the cyclin D1-overexpressing line fractured significantly less often, in just 20.8% of compressions (Fig 3B; S2 Movie).Although these spindles experienced similar compression-induced deformations, most spindles maintained all kinetochore-fibers focused into the two original spindle poles.
To probe the consequences of spindle pole fracture, we imaged control cell compressions without the addition of MG132 to follow spindles into late mitosis.Fractured spindles were still able to progress to anaphase, but they segregated chromosomes into three or more masses, depending on the number of new poles created by fracture ( Finally, we compared the levels of several spindle pole proteins-NuMA, HSET, Aurora A kinase, and TACC3-in the two cell lines, to test whether any were differentially regulated as a consequence of cyclin D1 overexpression.Each of these proteins localizes to spindle poles, and they play roles in assembling and maintaining the spindle pole (NuMA) [43][44][45], clustering supernumerary centrosomes (HSET) [37], phosphorylating other spindle proteins and localizing them to the pole (Aurora A kinase) [46], and regulating microtubule dynamics at centrosomes (TACC3) [47,48].We found that NuMA, but none of the other candidates tested, was upregulated in the cyclin D1 cell line (Fig 4A -4C), suggesting that NuMA could mediate cyclin D1's effect on pole robustness.Our results suggest that compressive force on mitotic MCF10A cells often causes spindle poles to fracture, leading to abnormal chromosome segregation at anaphase, but that overexpression of the oncogene cyclin D1 is protective against spindle fracture.

Discussion
Many oncogenes induce aberrant spindle architectures, yet they also promote uncontrolled cell proliferation in tumorigenesis.One explanation for this apparent paradox is that the elevated rate of chromosome mis-segregation in these spindles accelerates genome evolution and gives some cells a selective advantage [49].Here, we describe another mechanism by which an oncogene could act as a double-edged sword, with detrimental consequences for the cell in some contexts but conferring a proliferative advantage in others.Overexpression of cyclin D1 increases the prevalence of mitotic cells containing extra poles, centrioles, and chromosomes  3).Our assay was conducted in two-dimensional culture and with compressive force that may differ in magnitude and direction from that experienced by cells in vivo.Indeed, a recent study using HeLa cells found that while confinement-induced cell flattening led to increased pole fracturing, confining cells into elongated, narrow channels was protective against pole fracturing [16].However, an increase in spindle multipolarity has also been observed in confined HCT116 colorectal cancer cell spheroids [24], suggesting that our assay mimics compressive forces that exist in a crowded three-dimensional environment.
Although the fractured spindles we followed into anaphase segregated chromosomes into more than two masses (Fig 3C ), many of these mitoses presumably resolved into two daughter cells due to the lack of an anaphase central spindle competent to recruit the cytokinetic machinery between the newly separated poles.However, rapid nuclear envelope reformation at mitotic exit may prevent these multiple DNA masses from merging and lead to genomic instability or cell cycle arrest.Because cyclin D1 overexpression has a protective effect on bipolar spindle integrity under compressive force, we propose that it helps to prevent multipolar anaphases and speculate that it may allow cells to continue proliferating under compressive stress in the tumor context.Intriguingly, the cyclin D1 interactors pRb, p27, and p21 have been shown to mediate a G1 arrest in cells subjected to compressive stress [22,23,50], suggesting that cyclin D1 levels may affect the likelihood both that cells will continue to divide under compression and that they will complete these divisions successfully.
This work poses the question of the mechanisms by which cyclin D1 overexpression protects against spindle fracture.We find that cyclin D1-overexpressing cells upregulate NuMA (Fig 4 ).A role for NuMA in mediating the pole-protective effect downstream of cyclin D1 would be consistent with NuMA's function of clustering microtubule minus ends at spindle poles [44,45], and with previous observations that it promotes pole integrity under compressive stress [42].Cyclin D1 could directly or indirectly regulate additional factors involved in the cell's or the spindle's response to compression through its kinase-dependent or transcriptional roles [26].These factors could be regulated by cyclin D1 during interphase, and we note that cyclin D1 itself need not be expressed at high levels during mitosis.Pharmacologically inhibiting CDK4/6, the partner kinases of cyclin D1, and testing whether spindles in cyclin D1-overexpressing cells are sensitized to compressive stress could help determine whether cyclin D1's pole protective effect is kinase dependent.Any effects of CDK4/6 inhibitors on spindle mechanics may also be therapeutically relevant, because these inhibitors are widely used to treat metastatic estrogen receptor-positive breast cancer [51].
Indirect consequences of cyclin D1 overexpression could also underlie the spindle pole protection we observe.Incomplete mitotic rounding has been shown to lead to pole fracturing [14], and oncogenic h-Ras G12V has been shown to prevent pole fracture in MCF10A cells by enhancing mitotic rounding under stiff gels [15].We propose that different mechanisms are at play in the protective effect we observe here, because spindles in cyclin D1-overexpressing cells underwent fewer fractures despite being compressed to the same flattened height as spindles in control cells (Fig 2C).Supernumerary centrioles could contribute to the protective effect of cyclin D1 by increasing the density of microtubules and/or pericentriolar material at poles [2,52,53].Indeed, the proportion of bipolar spindles containing extra centrioles was increased from 1.3% of controls to 13.1% in the cyclin D1 cell line (Fig 1C ), and whether the pole-protective effect of cyclin D1 occurs specifically in cells with centriole amplification is an important question.Finally, other proteins that are differentially regulated during oncogenic transformation (but not specifically downstream of cyclin D1) could affect pole integrity.Future work dissecting the mechanism(s) by which cyclin D1 promotes bipolar spindle integrity under compression will be important to predict how generalizable this phenomenon is likely to be among tumors with diverse driver oncogenes.More broadly, achieving this goal will require a better understanding of the physical and molecular basis of spindle mechanical integrity [54][55][56].
The biochemical hallmarks of cancer, including anti-apoptotic signaling, metabolic reprogramming, and cell cycle dysregulation, are well-established [57].By contrast, our knowledge of the biophysical hallmarks of cancer lags behind, and addressing this gap could reveal new insights into disease progression.Our application of controlled, cellular-scale force suggests that cyclin D1 overexpression may adapt dividing cells to the mechanical burdens of the tumor environment.Better understanding the biophysical adaptations of cancer cells could lead to new ways to selectively target these cells for therapeutic gain.

Cell compression
Cell compressions were performed using a 1-well dynamic cell confiner with 5 μm PDMS micropillars, or 4 μm micropillars for the example shown in Fig 3C (4DCell).The device was attached to an AF1 Dual vacuum/pressure controller (Elveflow) and negative pressure was controlled using the Elveflow ESI software.Prior to imaging, a seal was established between the compression device and the dish of cells by applying a negative pressure of -10 mbar.At the start of imaging, a linear pressure ramp was applied from -10 to -150 mbar over a period of 4 minutes to lower the pillared coverslip onto the cells.Once the PDMS pillars contacted the dish, compression was maintained for 70 minutes.Z-stacks were acquired before each compression and after each timelapse acquisition to determine spindle height before and after each compression and quantitatively compare compression outcomes.

Data and statistical analysis
Immunofluorescence images show maximum intensity projections (Fig 1B) and time strip images show single spinning disk confocal Z-slices (Fig 3).All images were formatted for publication using FIJI [58].The brightness/contrast for each channel was scaled identically between different example cells for immunofluorescence images.The brightness/contrast for videos and time strips were scaled individually to account for variations in tubulin labeling efficiency.Kinetochores were counted using the multi-point tool in FIJI.For compression experiments, spindle heights were measured from XZ views generated from z-stacks (see Fig 2B) in a vertical direction perpendicular to the coverslip.A fracture was defined as the development of a clear gap in tubulin intensity between a kinetochore-fiber minus-end and the main spindle pole within 74 minutes of compression onset (Fig 3).

Fig 1 .
Fig 1. Cyclin D1 overexpression promotes aberrant spindle architectures.(A) Western blot of α-tubulin and cyclin D1 levels in MCF10A Puro (control) and MCF10A Cyclin D1 cell lines.All images are from the same blot, with intervening lanes removed.(B) Representative confocal immunofluorescence images (maximum intensity projections) of spindles stained for α-tubulin (green), CREST (yellow), centrin (magenta), and Hoechst (blue), with spindle phenotypes cartooned (right).Magnifications of the centrioles at each spindle pole are shown at right.Scale bars = 3 μm.(C) Frequency of the three observed metaphase spindle phenotypes in each MCF10A cell line.The distribution of phenotypes differs between cyclin D1 and control cells (*p = 0.010, Fisher's exact test), with the cyclin D1 line enriched in cells with supernumerary centrioles.(D) Number of kinetochores per spindle.Metaphase spindles in the cyclin D1 cell line had significantly more kinetochores (representing the number of chromatids) than the control line (****p = 2.32x10 -14 , Mann-Whitney U test).Lines indicate mean ± standard deviation.Inset shows a smaller range of kinetochore numbers.For C and D, n = 80 control spindles and 91 cyclin D1 spindles, each pooled from 3 independent experiments.https://doi.org/10.1371/journal.pone.0296779.g001 Fig 3C; S3 Movie; note the example shown was compressed using 4 μm micropillars).Interestingly, poles that separated from each other as a result of fracture were directly connected by few or no microtubules (Fig 3C, white arrow), and cytokinetic furrowing between these fractured poles was delayed and possibly incomplete (Fig 3C, yellow arrow).

Fig 2 .
Fig 2. The cell compression assay is quantitatively reproducible.(A) Schematic diagram of cell compression assay using a microfluidic device.Cells were compressed to a height of 5 μm using computer-controlled negative pressure over 4 min, and compression was sustained for 70 additional minutes.Cells were live-imaged throughout to monitor changes in spindle architecture.(B) Side (XZ) views of a control spindle, labeled with SiR-tubulin, before and after (at 74 min) compression.X and Z scale bars = 3 μm.(C) Between the control and cyclin D1 cell lines, spindle heights did not significantly differ before compression, and spindles were compressed to a similar final height (measured at 74 min).ns, not significant.(D) Spindle lengths before and 10 minutes after compression onset (ns, not significant).(E) Spindle widths before and 10 minutes after compression onset (*p = 0.028, **p = 0.00066).For C-E, two-sample t-tests were performed with n = 57 control and 53 cyclin D1 spindles (C) or n = 48 control and 52 cyclin D1 spindles (D and E), pooled from 13 and 12 independent days, respectively.Spindles were excluded from length and width analysis if both poles were not in focus in the same z-plane.https://doi.org/10.1371/journal.pone.0296779.g002

Fig 3 .
Fig 3. Cyclin D1 overexpression protects against spindle pole fracture during compression.(A) Confocal timelapse images of control and cyclin D1-overexpressing cells undergoing compression, where the control spindle fractures around 60 min after compression onset (cartooned at right).The fractured poles of the control spindle are indicated in the final frame by white arrows.Tubulin is labeled with SiR-tubulin.Scale bars = 5 μm; time stamps are in minutes.(B) Spindles in cyclin D1-overexpressing cells fractured less often than control spindles during the 74 minutes of compression (**p = 0.0048, Fisher's exact test).n = 57 control and 53 cyclin D1 spindles pooled from 13 and 12 independent days, respectively.(C) Confocal time-lapse images of a control spindle undergoing compression to a height of 4 μm, without the addition of MG132.After the spindle fractures between the 10 and 50 min time points, the cell enters anaphase and segregates chromosomes into 3 masses (cartooned below).Interpolar microtubule bundles connect the original pole to each of the fractured poles (white arrowheads), while no interpolar bundles connect the two poles resulting from the fracture (white arrow).The cytokinetic furrow is disrupted between the two fractured poles by the final 80 min timepoint (yellow arrow).Scale bars = 5 μm; time stamps are in minutes.https://doi.org/10.1371/journal.pone.0296779.g003

Fig 4 .
Fig 4. NuMA, but not HSET, Aurora A kinase, or TACC3, is upregulated in cyclin D1-overexpressing cells.(A) Western blot of NuMA, TACC3, HSET, Aurora A kinase, and cyclin D1 levels in MCF10A Puro (control) and MCF10A Cyclin D1 cell lines.The overexpressed cyclin D1 is HA-tagged.Vinculin is shown as a loading control.(B-C) Quantification of NuMA, cyclin D1 (B), HSET, Aurora A kinase, and TACC3 levels (C) normalized to vinculin levels.Lines show mean ± standard deviation of three independent replicates.https://doi.org/10.1371/journal.pone.0296779.g004 Fisher's exact test was used to compare categorical datasets (Figs 1C and 3B); two-sample ttests were used to compare the numerical datasets in Fig 2C-2E based on the assumption that spindle heights, lengths, and widths are approximately normally distributed; and a Mann-Whitney U test was used to compare the numerical dataset in Fig 1D due to kinetochore number distributions that deviated from a normal distribution.Statistical tests were performed using the ttest2, ranksum, and fishertest functions in MATLAB R2022b, and the fisher.testfunction in R for the 2x3 comparison in Fig 1C.All statistical tests were two-sided.P-values are given in the figure legends.Supporting information S1 Movie.A control cell before and after compression.Time stamps are in min:sec, scale bars are 5 μm, and video playback is 30 frames per second.See also Fig 3A.(MP4) S2 Movie.A cyclin D1-overexpressing cell before and after compression.Time stamps are in min:sec, scale bars are 5 μm, and video playback is 30 frames per second.See also Fig 3A.(MP4) S3 Movie.A control cell before and after compression to a height of 4 μm, without the addition of MG132.Time stamps are in min:sec, scale bars are 5 μm, and video playback is 30 frames per second.See also Fig 3C.(MP4) S1 Raw images.Raw images of all western blots.(PDF) S1 File.Source data for all graphs.(XLSX)