Human RASSF7 regulates the microtubule cytoskeleton and is required for spindle formation, Aurora B activation and chromosomal congression during mitosis

RASSF7 (Ras association domain family 7), previously known as HRC1 (HRAS1 cluster 1) [1] and C11orf13, is a member of the N-terminal RASSF family [2], a group of four proteins (RASSF7–RASSF10) which are structurally distinct from the classical RASSF proteins (RASSF1–RASSF6) [3,4]. There is emerging evidence to suggest that at least two members of the N-terminal RASSF family may be tumour suppressors. RASSF8 has reduced expression in lung cancers and knocking down its expression causes an increase in growth [5–7], and RASSF10 expression is epigenetically inactivated in leukaemias and thyroid cancers [8,9]. In contrast, microarray studies have demonstrated that RASSF7 expression is up-regulated in a range of cancers [10–14], but little is known about its function. Our previous work showed that knocking down the expression of Xenopus rassf7 in embryos caused a failure in spindle formation, nuclear fragmentation and apoptosis [15]. This suggests that RASSF7 is required for mitosis, a complex process involving microtubule nucleation from the spindle pole and the stable attachment of microtubules to the kinetochore [16–18]. However, RASSF7 function has not been studied in species other than Xenopus and the role that it plays in mitosis remains unknown. In the present study, we investigated the function of human RASSF7 and established why it is required for mitotic progression. Our data show that knocking down RASSF7 causes a reduction in cell growth and multiple defects in mitosis, including a failure in spindle formation, Aurora B activation and chromosomal congression. In addition, we have shown that RASSF7 localizes to the centrosome and that it plays a crucial role in regulating microtubule growth. These results indicate that RASSF function is crucially important for mitosis because it regulates microtubule growth from the centrosome.

Cell culture medium and reagents were purchased from Sigma. H1792 and HeLa cells were maintained at 37 °C under 5% CO₂ in DMEM (Dulbecco's modified Eagle's medium), with 10% fetal bovine serum, 1% penicillin/streptomycin and 1% glutamine. Mitotic cell enrichment and the microtubule-depolymerization and -regrowth assays were carried out as described in [19,20]. To expose cells to hypoxic conditions, cultures were placed in a humidified 37 °C multigas incubator with 5% CO₂/95% nitrogen for 3 h. Residual O₂ levels were monitored regularly and were found to be consistently less than 0.1%. Control cultures, maintained under nomoxic conditions (21% O₂), were processed in parallel.

The coding sequence of human RASSF7 (IMAGE clone: 40032773) was cloned into the expression vector pcDNA3-DEST53 (Invitrogen) using TOPO® and Gateway® cloning (Invitrogen) to generate a GFP (green fluorescent protein)-fusion protein. This was transfected into HeLa cells using Lipofectamine™ (Invitrogen). H1792 cells were transiently transfected with siRNA (small interfering RNA) oligonucleotides targeting RASSF7, RASSF7 KD1 (Ambion oligonucleotide ID 13242) and RASSF7 KD2 (Ambion oligonucleotide ID 13338) or a control oligonucleotide targeting luciferase (MWG) using Oligofectamine™ (Invitrogen). The soft agar colony-forming assay was conducted as described in [7]. shRNA (short hairpin RNA) silencing of gene expression was achieved using RASSF7 (code KHO9867N) and scrambled control shRNA. pGeneClip/Neo vectors from SuperArray Bioscience. Approx. 2×10⁵ HeLa cells were seeded 1 day before transfection. Transfected cells were selected for 14 days in 1.2 mg/ml G418 (Sigma) and subsequently counted and analysed. The experiment was repeated as described above, except that the seeding of 2×10⁵ HeLa cells was carried out after selection.

Immunoblotting was carried out and quantified using an Optichem detector with associated software (Ultra Violet Products) as described in [21]. After transfer, the membranes (Whatman) were blocked in 5% (w/v) non-fat dried skimmed milk powder in Tris-buffered saline containing Tween 20. Antibodies used were rabbit anti-RASSF7 (Aviva Systems Biology ARP34390_T100), mouse anti-β-tubulin (Sigma T4026), mouse anti-(phospho-histone H3) (Abcam 14955). The anti-RASSF7 antibody produced a band at 34 kDa as predicted.

Cells were grown on coverslips, fixed and permeabilized at −20 °C in methanol for 10 min and incubated in blocking solution (PBS, 1% BSA and 3% normal goat serum). Primary antibodies were: rabbit anti-pericentrin (Covance PRB-4325), mouse anti-γ-tubulin (Sigma T6557) and mouse anti-RASSF1A (eBioscience 14-6888). The rabbit anti-RASSF7 antibody required antigen retrieval to unmask the epitope, which involved EDTA treatment (1 mM, pH 8.0) for 1.5 h at 37 °C before blocking. For mouse anti-α-tubulin (Sigma T9026), mouse anti-(Aurora A) (BD Biosciences 610938), mouse anti-(Aurora B) (BD Biosciences 611082), rabbit anti-phospho-Aurora (Cell Signaling Technology 2914), rabbit anti-phospho-PLK1 (Polo-like kinase 1) (Biolegend 618601), mouse anti-INCENP (inner centromere protein) (Upstate Biotechnology 05-940), rabbit anti-phospho-CENP-A (centromere protein A) (Millipore 04-792) and rabbit anti-(active caspase 3) (Abcam ab13847) cells were fixed in 4% (w/v) paraformaldehyde and permeabilized in 0.1% Triton X-100 at room temperature (22 °C). Incubation with relevant secondary antibodies and DAPI (4′,6-diamidino-2-phenylindole) (1 μg/ml; Sigma) followed. Cells were then mounted in Mowiol (Sigma), visualized using a Zeiss LSM 510 confocal microscope, and images were captured and processed using LSM image browser software.

COBRA (combined bisulfite restriction analysis) [8] was used to determine the methylation status of the 5′ CpG island associated with the RASSF7 gene. Primers used for methylation analysis were 5′-TTTAGGAGYGGGGTTAGATATTTATTT-3′ and 5′-TTAACTACCTCTATCACRCCCCTCCC-3′.

In situ hybridization was carried out as described previously [7]. The mouse Rassf7 probes were generated using a pCMV.sport6 vector (IMAGE clone 4206694) and an SP6 (sense) or T7 polymerase (antisense) following digestion with StuI or AhdI enzymes respectively. H a E (haematoxylin and eosin) staining was performed on paraffin sections using standard methods.

Means and S.D. values were calculated and plotted using Microsoft Excel. Each experiment was repeated at least three times. P<0.05 was considered significant. Statistical analysis was carried out using unpaired Student's t tests.

We began by establishing which tissues and cell types express RASSF7. In situ hybridization was carried out on mouse embryos and adult organs, and Rassf7 was found to be expressed in all tissues examined; however, the level of expression varied, with particularly high levels seen in the lung and the brain (Figures 1A–1C). Immunoblotting of lysates from mammalian and human cell lines showed a major band which migrated at the predicted size of RASSF7 (34 kDa) (Figure 1D) and this band is specific for RASSF7 because it was reduced by two different knockdown approaches (Figures 2A and 2B). The immunoblotting showed that RASSF7 was expressed in all cell lines tested (Figure 1D). Consistent with the fact that RASSF7 is expressed in a number of different cancerous cell lines, and in contrast with other RASSF proteins [3,4], we found no evidence of RASSF7 gene silencing by promoter methylation in 57 cancer cell lines (see Supplementary Figure S1A at http://www.BiochemJ.org/bj/430/bj4300207add.htm, and results not shown). RASSF7 mRNA is up-regulated by hypoxia [22,23], which, given the hypoxic nature of solid tumours, might explain the increase in RASSF7 expression seen in tumour cells [10–14]. Western blotting demonstrated that RASSF7 protein levels are also increased by hypoxic insult (Figure 1E), showing that RASSF7 protein is likely to be increased by the hypoxic environment found in solid tumours.

Several members of the RASSF family are believed to be tumour suppressors and act to restrain growth [3,4]. Consistent with this idea, knockdown of RASSF6 or RASSF8 causes an increase in the anchorage-independent growth of the human lung adenocarcinoma cell line H1792 [7,24]. To establish whether RASSF7 functions in a similar way, we analysed the effect of RASSF7 knockdown on anchorage-independent growth. RASSF7 expression was knocked down in cells transfected with RASSF7 KD2 siRNA (Figure 2A). In contrast, RASSF7 KD1 did not efficiently knock down RASSF7 and this siRNA was used as an additional negative control along with the standard control siRNA (Figure 2A). H1792 cells showed significantly lower colony-forming activity in soft agar after RASSF7 KD2 siRNA treatment as both the size and the number of colonies was reduced compared with the controls (Figure 2A). These findings show that, unlike RASSF6 and RASSF8, which act to restrain growth, RASSF7 is essential for anchorage-independent growth. The fact that RASSF7 is required for growth may explain why its expression is not silenced in the cancer cell lines analysed above.

To address why RASSF7 is required for growth, we carried out shRNA knockdowns in HeLa cells. The alternative knockdown strategy ensures the specificity of the phenotype, and HeLa cells were used as they are easier to analyse than H1792 cells. Consistent with the data in H1792 cells, shRNA knockdown of RASSF7 in HeLa cells caused a reduction in cell number (Figure 2B). This could be caused by increased apoptosis; however, the percentage of active caspase 3-positive RASSF7 KD cells was not significantly different from controls (see Supplementary Figure S2 at http://www.BiochemJ.org/bj/430/bj4300207add.htm). To establish whether the reduction in cell number was caused by a defect in mitosis, analogous to that seen in Xenopus, knockdown cells were examined and mitotic aberrations in metaphase RASSF7-knockdown cells were clearly evident. Loss of RASSF7 resulted in a failure in chromosomal congression, demonstrated by an increase in the number of mitotic cells which had failed to align their chromosomes (63% compared with 18% in controls, n=300, P<0.01) (Figure 2C), an increase in metaphase cells with lagging chromosomes (21.5% compared with 6% in controls, n=300, P<0.01) (Figure 2C, white arrows) and a decrease in metaphase cells with correctly aligned DNA (15.7% compared with 75.7% in controls, n=300, P<0.001). Spindle defects were also found in RASSF7-deficient cells, with spindles showing less pronounced polarization towards the DNA and a more radial organization of microtubules (Figure 2C, red arrows). Finally, there was a small increase in the rate of multi-polar spindles (32.6% compared with 23.7% in controls, n=300, P<0.05) (Figure 2C, arrowheads).

In order to understand the molecular basis for these mitotic defects, we examined the effect of RASSF7 knockdown on established regulators of mitosis. Analysis of the activation of PLK1 [25], using an antibody which recognizes the active form, showed similar staining in RASSF7-knockdown and control cells (Figure 3A). The localization of RASSF1A, which is required for progression through mitosis [26–28], to the centrosome appeared to be the same in knockdown and control cells (Figure 3B). The localization and activation of Aurora A and Aurora B [29] was also examined (Figures 3C and 3D). An antibody against Aurora A and B and a phospho-specific antibody which recognizes the active form of Aurora A and B enzymes were used. Aurora A was present at the centrosome in RASSF7 (Figure 3C, arrows) and Aurora B was present at the kinetochores (Figure 3D, arrowheads). The active Aurora staining was present at the centrosomes (Figures 3C and 3D, arrows), but strongly reduced at the kinetochore (Figures 3C and 3D, arrowheads) in the RASSF7 knockdowns, with only 9.3% of cells showing strong Aurora B staining compared with 83.8% in the controls (n=240, P<0.001), suggesting a failure in Aurora B activation. This loss appeared to be restricted to metaphase as the activation of Aurora B during cytokinesis appeared normal (see Supplementary Figure S3A at http://www.BiochemJ.org/bj/430/bj4300207add.htm). To confirm the loss of Aurora B activity, the phosphorylation of an Aurora B target, CENP-A [30], was analysed. Phospho-CENP-A levels were reduced (Figure 3E), with only 22% of cells showing strong staining compared with 93.4% of control cells (n=210, P<0.001), providing further evidence that Aurora B is not activated normally in the absence of RASSF7. The loss of Aurora B activity is likely to contribute to the mitotic defects seen in RASSF7 knockdowns and, consistent with this hypothesis, inhibiting Aurora B causes chromosomal congression defects similar to those in RASSF7-knockdown cells [31,32].

To begin to understand the mechanistic basis for these mitotic defects, we analysed the subcellular localization of RASSF7. Xenopus rassf7 fusion proteins localize to the centrosome [15] and a GFP-tagged human RASSF7 protein also localized to the centrosome in HeLa cells (see Supplementary Figure S4A at http://www.BiochemJ.org/bj/430/bj4300207add.htm). To establish the localization of endogenous RASSF7, HeLa cells were stained with an anti-RASSF7 antibody. This showed strong staining which co-localized with γ-tubulin, a component of the pericentriolar material (Figure 4A) and this staining was specific as the signal was not present in RASSF7-knockdown cells (see Supplementary Figure S4B). RASSF7 staining was found at the centrosome in interphase cells and throughout all stages of mitosis (Figure 4A), showing that RASSF7 is not recruited to the centrosome during mitosis, but localizes there in dividing and non-dividing cells. The amount of RASSF7 protein was also not increased in mitotic cells (see Supplementary Figure S1B), so, despite RASSF7 having a role in mitosis, RASSF7 protein levels do not appear to be regulated during the cell cycle. The localization of RASSF7 at the centrosome was not affected by microtubule depolymerization with nocodazole (Figure 4B), indicating that human RASSF7 is an integral centrosomal component whose localization is microtubule-independent. Our previous data with Xenopus rassf7 suggested that its localization was microtubule-dependent [15]; this discrepancy could well be due to the use of a fusion protein in the Xenopus experiments.

The localization of RASSF7 at the centrosome and the spindle defects seen in the knockdowns suggest that RASSF7 might have a role in regulating microtubules. To test whether RASSF7 is required for microtubule polymerization, we investigated whether microtubule regrowth was affected by the loss of RASSF7. Cells were treated with nocodazole and subsequent microtubule regrowth was examined. The initial nucleation of microtubules in knockdown cells appeared normal (5 min), but after 15 min, 53% of RASSF7-knockdown cells showed a delay in microtubule regrowth compared with 14% of controls (n=100, P<0.001) (Figure 4C, arrows). The microtubules in RASSF7-knockdown cells also appeared more bent and looped than controls (Figure 4C, arrows). A similar delay in microtubule regrowth was seen at 30 min. After 60 min, the number of microtubules in the RASSF7-knockdown cells resembled the controls (Figure 4C), showing that RASSF7-knockdown cells eventually produce similar numbers of microtubules to control cells. However, the microtubules still appeared less straight than the control microtubules (Figure 4C, arrows). This assay demonstrates that RASSF7 is a key regulator of microtubule growth and provides an explanation for the spindle defects seen in mitosis. The broad expression of RASSF7 and its localization to the centrosomes of interphase cells suggests that RASSF7 may also regulate microtubules in non-dividing cells.

A role in regulating microtubules can explain the defects in spindle formation seen in the knockdowns, but why is RASSF7, which localizes at the centrosome, required for the activation of Aurora B at the kinetochores? Aurora B activation requires the recruitment of INCENP [33], but RASSF7 knockdown did not compromise the localization of this protein (see Supplementary Figure S3B). Interestingly, Aurora B activation also requires contact with microtubules [34,35]. We propose that the defect in spindle formation seen in the RASSF7-knockdown cells might stop microtubules from correctly interacting with Aurora B and cause the failure in Aurora B activation. This argues that the primary role of RASSF7 is in regulating microtubule growth and that mis-regulation of this process is responsible for the spindle defects and the loss of Aurora B activation. The aberrant spindle and loss of Aurora B would then both contribute to the dramatic failure in chromosomal congression seen in RASSF7-knockdown cells.

In summary, the present study provides the first functional analysis of human RASSF7 and shows that down-regulation of RASSF7 impairs cell growth and causes defects in mitosis, including abnormal spindle assembly, loss of Aurora B activation and a failure in chromosomal congression. We propose that these mitotic defects are caused because RASSF7 localizes to the centrosome and regulates microtubule dynamics.

CENP-A

centromere protein A

GFP

green fluorescent protein

H a E

haematoxylin and eosin

INCENP

inner centromere protein

PLK1

Polo-like kinase 1

RASSF

Ras association domain family

shRNA

short hairpin RNA

siRNA

small interfering RNA

Asha Recino co-ordinated the study, designed experiments, collected and analysed the majority of the data and wrote the paper; Victoria Sherwood designed experiments, and collected and analysed data for preliminary experiments; Amy Flaxman assisted in data collection and analysis; Wendy Cooper designed experiments, and collected and analysed data; Farida Latif designed experiments and analysed data; Andrew Ward designed experiments and analysed data; Andrew Chalmers oversaw the project, designed experiments, analysed data and wrote the paper.

We thank Adrian Rogers (University of Bath Bioimaging Suite), Iryna Withington and Joanne Stewart-Cox for technical support and advice, David Tosh, Will Wood and Cheryll Tickle for comments on the manuscript, Momna Hejmadi (Department of Biology and Biochemistry, University of Bath) for providing the hypoxic chambers and Geoff Holman (Department of Biology and Biochemistry, University of Bath) for providing the Optichem detector for analysing Western blots.

A.R. was supported by a Marie Curie Ph.D. studentship [grant number MCEST-CT-2005-019822]. F.L. was supported by the Breast Cancer Campaign. A.D.C. was supported by a Medical Research Council Career Development Fellowship [grant number G120/844].