Murine neuroblastoma cells constitute a good model system to study centrosome declustering

To identify an ideal in vitro model system to study interphase-specific centrosome-declustering events, we evaluated murine neuroblastoma N1E-115 cells. We found that 100% of N1E-115 cells harbor amplified centrosomes (5–20 centrosomes per cell). We also found that the centrosomal cluster in N1E-115 cells is a melange of single, free-standing mother and daughter centrioles and a few canonical centrosomes (Supplementary Figure S1). We thus wondered how these cells haul their centrosomal load through the cell cycle phases to accomplish cell division. In N1E-115 interphase cells, the multiple centrosomes localized as a distinct juxtanuclear cluster (Figures 2a and b). However, in ~10% cells, the multiple centrosomes showed significant scattering, and this feature correlated with chromatin condensation and the absence of a mitotic spindle. Lamin A/C immunostaining showed that cells with loose centrosome clusters still had an intact nuclear membrane. Thus these cells were confirmed to be in prophase (Figures 2a and b). About 30% of mitotic cells were in prometaphase (i.e., they lacked a nuclear membrane) and possessed multipolar spindles with multiple MTOCs at each spindle pole. Metaphase cells, by contrast, clustered supernumerary centrosomes into two polar groups to generate a pseudo-bipolar spindle (Figures 2a and b). The centrosomes at the poles of the spindle were often arranged in ‘ring-like' or ‘V-shaped linear' configurations (Figures 2a). Only ~20% cells displayed equal centrosome counts at the two spindle poles. CREST antibody labeling revealed that, in metaphase cells, the kinetochores did not line up immaculately along the spindle equator perhaps due to widespread merotelic attachments (Figure 2c). Centrosomes remained clustered at the two spindle poles through anaphase, and the occurrence of occasional lagging chromosomes indicated chromosome missegregation (Figure 2d). Telophase was marked by nuclear envelope reassembly around the decondensing chromatin and the inheritance of a juxtanuclear centrosomal cluster by each daughter cell (Figures 2a and b). These observations indicated that murine N1E-115 cells are a great model system to study centrosome clustering and declustering as they show both interphase and mitotic centrosome clustering similar to patient tumor cells.

Open in a separate window

Figure 2

Cell cycle phase characterization of N1E-115 cells. (a) Representative immunofluorescence confocal micrographs depicting centrosome status in all cell cycle phases of N1E-115 cells. N1E-115 cells in interphase possess an enormous number of centrosomes as evident by γ-tubulin immunostaining (green). We acquired images as z-stacks with the slice interval of 0.40μm. Z-stack slices encompassing the entire depth of the cell were then merged, and γ-tubulin-positive spots were counted in interphase cells from randomly selected fields totaling 200 interphase cells. (b) Representative immunofluorescence confocal micrographs showing lamin A/C staining (red) across all cell cycle phases to visualize nuclear membrane in order to distinguish interphase declustering from prophase scattering of centrosome cluster. (c) Confocal micrograph of a metaphase cell stained with CREST antibodies (white), antibodies against α-tubulin (red) and γ-tubulin (green) and 4,6-diamidino-2-phenylindole (DAPI; blue) to detect microtubule-kinetochore attachments and DNA in a pseudobipolar mitotic spindle. (d) Representative micrograph of an anaphase cell immunostained for α-tubulin (red), γ-tubulin (green) and DAPI (blue) showing lagging chromosomes. Scale bar, 5μm

In order to probe the mechanisms facilitating interphase clustering in N1E-115 cells, we examined the involvement of two major microtubule motors (human spleen embryonic tissue and testis (HSET) and dynein) in centrosome clustering in interphasic and mitotic N1E-115 cells. We found that siRNA knockdown of the kinesin-14 protein KifC1/HSET resulted in robust mitotic declustering generating ~65% multipolar mitotic cells, but negligible interphase declustering (15% as compared with 12% in vehicle-treated controls; Supplementary Figure S2). However, dynein inhibition by ciliobrevin treatment demonstrated substantial scattering of interphase centrosomal clusters (~50%) and considerable mitotic declustering (~35%) (Supplementary Figure S2). Based on these data, it seems that, while HSET is crucially involved in mitotic centrosome clustering, dynein has the major role in maintaining the centrosomal cluster during the subsequent interphase.

Centrosome-declustering agents disperse interphase clusters and set the stage for a catastrophic mitosis

Given the limited mitotic populations in human cancers, centrosome declustering during mitosis alone would fail to achieve sufficient elimination of cancer cells. On the other hand, interphase declustering may not only prime the cell for catastrophic mitosis but also ensure disruption of interphase-specific cellular processes that undergird migration. Thus we investigated how declustering agents affect centrosome clustering during interphase. We tested three declustering drugs (RedBr-Nos, Griseofulvin and PJ-34)^{13, 14, 15, 16, 17} and compared them with Paclitaxel, a tubulin-polymerizing drug. RedBr-Nos, Griseofulvin and Paclitaxel are known to bind tubulin^{18, 19, 20} but PJ-34 is a poly-ADP-ribose polymerase inhibitor with no known tubulin-binding property. However, they share common phenotypes, such as mitotic arrest and multipolar mitoses.^{13, 14, 15, 16, 17, 21} We found N1E-115 cells to be more sensitive to these drugs compared with other cancer cell lines (for instance, MDA-MB-231, HeLa) with IC₅₀ values ranging between 0.05μM for Paclitaxel and 25μM for Griseofulvin (data not shown). To evaluate their effect on interphase clustering, we treated N1E-115 cells with drugs at their respective IC₅₀ concentrations for 0, 3, 6 and 9h and co-immunostained for γ-tubulin and α-tubulin to evaluate centrosomal spread and microtubule nucleation status, respectively (Figure 3a). RedBr-Nos and Griseofulvin inflicted more severe interphase declustering compared with PJ-34 and Paclitaxel. We also verified the cell cycle phases via lamin A/C immunostaining to distinguish interphase declustering events from prophase centrosomal spread. To quantitate the spread of the interphase centrosomal cluster, we generated a 3-D reconstruction of z-stack images of 25 randomly-selected interphase cells from the 6-h treatment group of each drug. By defining an ROI (region of interest) around the interphase centrosomal cluster, we calculated volume of the cluster spread using the Volocity software as shown in Figure 3bi. Likewise, defining an ROI using the cell periphery provided the cell volume. We defined the interphase declustering index (DI) for each drug as the ratio of the average volume of clusters to the average volume of the corresponding cell. Quantitative evaluation of DI revealed RedBr-Nos as the strongest declustering agent (DI=0.36), followed by Griseofulvin (DI=0.28) and PJ-34 (0.14). Paclitaxel showed the least declustering effect with a DI of 0.08 as compared with 0.02 in control cells (Figure 3bii). We also found that dispersal of the interphase centrosome cluster precipitated multipolar mitoses in the treated cells (Figure 3c). Again, the proportion of multipolar cells was higher in RedBr-Nos- and Griseofulvin-treated cells as compared with cells treated with PJ-34 and Paclitaxel, which mirrored the trend in interphase declustering (Figure 3c). These observations suggest that interphase declustering of centrosomes compels cells into catastrophic multipolar mitoses.

Open in a separate window

Figure 3

Interphase declustering induced by centrosome-declustering agents. (a) Confocal micrographs showing interphase declustering induced by 6-h treatment with RedBr-Nos (10μM), Griseofulvin (50μM), PJ-34 (25μM) and Paclitaxel (0.1μM). The percentages indicate proportion of interphase cells with declustered centrosomes. (bi) 3-D representation and quantitative volume analysis of control and drug-treated interphase cells using the Volocity 6.3 software. Cells were co-immunostained for lamin A/C (red) and γ-tubulin (green), and z-stacks were acquired with a 0.35μm z-step. Z-stack slices were then used to construct a 3-D image, and ROIs were defined to generate DI measurements. (bii) Quantitative bar graph representing the DI of the four drugs. P<0.05. (c) Confocal micrographs showing spindle multipolarity induced by 18h treatment with declustering agents RedBr-Nos (10μM), Griseofulvin (50μM), PJ-34 (25μM) and Paclitaxel (0.1μM). Cells were co-immunostained for α-tubulin (red) and γ-tubulin (green). The percentages indicate proportion of mitotic cells with declustered centrosomes. (d) Confocal micrographs showing Golgi dispersal concomitant with interphase declustering upon 6-h treatment with all the four declustering agents at the stated concentrations. Cells were co-immunostained for GM130 (red) and γ-tubulin (green). DNA was DAPI (4,6-diamidino-2-phenylindole) stained. Scale bar, 5μm

Centrosome declustering in interphase disrupts Golgi coalescence and inhibits migration

The Golgi, which is primarily responsible for posttranslational modification and protein sorting, also functions as an MTOC.²² It has been hypothesized that supernumerary centrosomes may better organize the Golgi to enhance directional cell migration.²³ We therefore investigated what happens to the Golgi upon declustering drug-induced dispersal of the interphase centrosomal cluster. We co-immunostained drug-treated cells for GM130 (a cis-Golgi matrix protein crucial for maintaining its structure) and γ-tubulin. Following treatment with declustering drugs, the interphase Golgi complex fragmented, and the distribution of Golgi fragments closely mimicked scattering of the centrosomal cluster, with the most robust effect seen with RedBr-Nos and Griseofulvin (Figure 3d).

Research suggests that Golgi-derived microtubules are not sufficient to preserve cell polarization; instead, they need to act in concert with the centrosome to establish and maintain cell polarization.²⁴ In cancer cells harboring a supercentrosomal cluster, we predict that disrupting the cytoskeletal and organellar framework organized by a strongly polarizing supercentrosomal cluster will present a setback to the mechanical thrust that such a cluster can empower a migrating cell with; this in turn, we predict, will lead to impaired directional migration. As a surrogate for the polarization that underlies directional migration, we decided to examine neuritogenesis, a process in nerve cells involving the extension of polarized, elongated neurites. N1E-115 cells usually extend only one major neurite per cell, which can vary in length from 5 to 500μm. The growth cones of the neurites serve as primary focal points of motility. We evaluated the effect of declustering agents on cell motility by assessing the length and frequency of neurites formed in a serum-free medium on a laminin-coated surface. Neurite growth under these conditions is linear for up to 24h, reaching a maximum around 36–48h after plating. Phase-contrast imaging showed the presence of several elongated (10–200μm long) neurites upon 48h of serum starvation (Figure 4bi). We observed 70–80% inhibition of neurite extension when treated with RedBr-Nos and Griseofulvin and moderate inhibition with PJ-34 and Paclitaxel treatment (Figures 4a and bii). Confocal imaging confirmed that inhibition of neurite formation was accompanied by dispersal of the interphase centrosome cluster, which is normally situated near the base of the tubulin-rich neurite shaft (Figure 4c).

Open in a separate window

Figure 4

Inhibition of neuritogenesis by centrosome-declustering agents. (a) Phase-contrast images of N1E-115 cells in SS medium showing neurite formation after 48h of SS or with RedBr-Nos (5μM), Griseofulvin (10μM), PJ-34 (10μM) and Paclitaxel (0.05μM) treatment. Scale bar, 10μm. (bi and bii) Quantitative bar graphs representing the average length of neurites and the percentage of population of cells showing neurite length >10μm, respectively. Hundred cells were counted in each case. P<0.05. (c) Confocal micrographs showing neurite outgrowth after 48h without SS, with SS or SS along with drug treatment, respectively. Cells were co-immunostained for α-tubulin (red) and γ-tubulin (green). DNA was 4,6-diamidino-2-phenylindole (DAPI) stained. (di) Immunoblot showing the CLASP1 expression levels in control and CLASP1 siRNA-transfected N1E-115 cells. (dii) Confocal micrographs showing Golgi network immunostained for GM130 (red) and centrosome cluster immunostained for γ-tubulin (green) in control and CLASP1 siRNA-transfected N1E-115 cells. (diii) Confocal micrographs showing neurite outgrowth in control and CLASP1 siRNA-transfected N1E-115 cells. (e) Confocal micrographs showing vinculin localization during neurite outgrowth after 48h without SS, with SS or SS along with drug treatment, respectively. Cells were stained for F-actin using rhodamine-phalloidin, immunostained for vinculin (green) and DNA was DAPI stained. Scale bar, 5μm

In order to establish whether Golgi-dependent vesicular trafficking lies downstream of interphase centrosome clustering during cell polarization and neuritogenesis in N1E-115 cells, we studied the effect of centrosome-declustering-independent Golgi scattering on neuritogenesis. To accomplish this, we used CLASP1 siRNA to disrupt the Golgi-nucleated microtubules (Figure 4di), thus disarraying the directionality of post-Golgi vesicular trafficking but leaving the centrosome cluster intact, and evaluated whether these cells can generate neurites. We observed ~50% Golgi scattering upon CLASP1 knockdown (Figure 4dii). We observed that cells with CLASP1 siRNA formed significantly fewer neurites compared with control cells (Figure 4diii,Supplementary Figure S3C). This observation suggests that (i) disruption of Golgi network impedes Golgi polarization-dependent neuritogenesis, and (ii) Golgi complex integrity and polarized post-Golgi trafficking lie downstream of interphase centrosome clustering.

The spatio-temporal arrangement of Golgi apparatus serves as a geometrical regulator of cell migration as well as neurite extension. Thus we wanted to determine whether Golgi disruption upon CLASP1 knockdown affects cell shape and cell adhesion, modulation of which are crucial for cell migration as a precursory step for neurite extension in N1E-115 cells. We observed significant shift in the morphology of cells from majorly mesenchymal-like cell shape in cells transfected with control vector to largely amoeboid-like and more ‘rounded' cell shape in CLASP1 knockdown cells. This shift in cell morphology indicates changes in cell-substrate adhesion properties as a result of Golgi dispersal, which was confirmed by the reduction in vinculin localization at distinct adhesion focal points in CLASP1 siRNA cells (Supplementary Figure S3).

Vinculin stabilizes cell-substrate contacts in neuronal cells undergoing neuritogenesis,²⁵ and activation by actin-binding proteins mobilizes vinculin to focal adhesions.^{26, 27} We therefore determined the localization of vinculin in the neurite extensions and the effect of declustering agents on its localization. To this end, we immunostained cells for vinculin and stained F-actin using rhodamine-phalloidin. Cells in serum-supplemented medium showed vinculin localization at focal adhesions with very little internalized vinculin. Upon serum starvation for 48h, most of the vinculin was localized to the neurite growth cones. However, upon treatment of serum-starved (SS) cells with RedBr-Nos and Griseofulvin, we observed complete internalization of vinculin and complete loss of focal adhesion points. The observed effect was less severe with PJ-34 and Paclitaxel (Figure 4e). Centrosome-declustering drugs thus impair cell polarization and neurite formation and the localization of vinculin, a key player in the establishment of cell-substrate contacts. In order to support our rationale that the dispersion of centrosomal clusters in interphase is directly responsible for anti-migratory effects of these drugs and are not merely side effects of the drugs, we showed that the declustering agents were comparatively less affective in disrupting neuritogenesis of mouse neuroblastoma cells, Neuro-2a (harboring much lesser degree of centrosome amplification) (data shown in Figure S4).

These observations underscore the immense clinical potential of centrosome declustering as a selective therapy for cancer cells harboring excess centrosomes, without affecting cells with normal centrosome content.

Cellular protein preparation, western blotting, immunofluorescence and antibodies and other reagents

Cells were cultured to ~70% confluence, and protein lysates were collected following drug treatment, transfection or otherwise for western blotting following methods described in previous publications.¹³ For immunofluorescence staining, cells grown on glass coverslips were fixed with cold (−20°C) methanol or 4% paraformaldehyde (room temperature) for 10min and blocked by incubating with 2% bovine serum albumin/PBS/0.05% Triton X-100 at 37°C for 1h. Specific primary antibodies were incubated with coverslips for 1h at 37°C at the recommended dilution followed by 1:2000 dilution of Alexa 488- or 555-conjugated secondary antibodies. Antibodies against γ-tubulin, α-tubulin and β-actin were from Sigma (St Louis, MO, USA); cleaved caspase-3 was from Cell Signaling (Danvers, MA, USA). Alexa 488- or 555-conjugated secondary antibodies were from Invitrogen (Carlsbad, CA, USA). Anti-Ki67 antibody was from Abcam (Cambridge, MA, USA). Anti-GM130 antibody was from BD Biosciences (San Jose, CA, USA). Anti-vinculin antibody was purchased from Millipore (Billerica, MA, USA). Anti-lamin A/C, anti-centrin-2 and anti-CLASP1 antibodies and horseradish peroxidase-conjugated secondary antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). CLASP1 siRNA was ordered from Origene (Rockville, MD, USA). SMARTpool: ON-TARGETplus KIFC1 siRNA (Dharmacon, Pittsburgh, PA, USA) was used to knockdown HSET in N1E-115 cells.

Electron microscopy

N1E-115 cells were grown on coverslips made of Aclar film (Electron Microscopy Sciences, Hatfield, PA, USA) and were processed for electron microscopy as described in previous publication.³²

Neurite extension assay

N1E-115 cells were grown on glass coverslips coated with laminin in 35-mm tissue culture dishes and treated with the respective drugs diluted in serum-free medium; the cells were then fixed, labeled and visualized by immunofluorescence microscopy. Cells bearing neurite-like structures with a length of at least one cell diameter were identified by immunofluorescence microscopy using an α-tubulin antibody. At least 200 cells were counted for each condition, and the experiments were repeated three times. The images were taken using a Ziess LSM 700 confocal microscope (Oberkochen, Germany).

This study was supported by grants to RA from the National Cancer Institute at the National Institutes of Health (R00 CA131489, R01 CA169127).