High SEPT9_v1 Expression in Human Breast Cancer Cells Is Associated with Oncogenic Phenotypes

Although breast cancer results from the accumulation of multiple genetic alterations in mammary epithelial cells, the molecular pathways underlying this heterogeneous disease are complex, and many remain to be characterized (1, 2). Toward this effort, we positionally cloned SEPT9 from a discrete region of allelic imbalance on human chromosome 17q25.2 identified in breast and ovarian cancer cells (3, 4). Originally called “MSF” for its recognition as a mixed lineage leukemia (MLL) septin-like fusion partner in acute myelocytic leukemia patients with (11;17)(q23;q25) translocations (5), it was renamed SEPT9 in recognition of its homology to other human septins (6).

SEPT9 belongs to an evolutionarily conserved family of septin proteins containing GTP-binding domains that were first characterized in Saccharomyces cerevisiae by mutations that affected budding morphology and cell cycle progression. Septins can form microfilaments by interacting with each other or with cytoskeletal and filamentous proteins, indicative of their role in cytokinesis during contractile ring formation, cell morphology, and dynamic scaffolds (7–10). In fact, SEPT9 has proven to be critical for normal division of mammalian cells (10). Characterization of mutants also suggests that septins may help regulate cell cycle progression by triggering the mislocalization of inhibitory cyclin-dependent kinases, preventing degradation of inhibitors needed for cell cycle progression through G₂-M (11). Current evidence suggests that septin-associated cytoskeleton organization is tightly coupled to cell cycle progression in yeast cells, creating a septin-associated cytokinesis-related cell cycle checkpoint (11). Homologies between mammalian and yeast septins and cell cycle kinases suggest that this checkpoint described in yeast may be conserved and crucial for proper mitotic cell cycle progression.

Human septins, including SEPT9, have been implicated in cancer. The potential role of SEPT9 in cancer is surmised, in part, by its expression patterns in cancer cells and its homologies. Almost all human septins exhibit extensive alternative splicing and diversity at the 5′ and 3′ ends of the transcript, which suggests spatial and/or temporal regulation of expression (12–17). SEPT9 has at least seven transcripts encoding five distinct isoforms (SEPT9_v1-SEPT9_v5; ref. 16). Altered expression of SEPT9 variants has been implicated in the pathogenesis of several cancers, including ovarian and prostate cancer (12, 16–18). Specifically, Amir et al. (18) found that overexpression of SEPT9_v1 augments hypoxia-inducible factor-1α transcriptional activation, increasing tumorigenicity and angiogenesis in in vitro and in vivo models of prostate cancer. Septins may belong to the class of cancer genes in which alterations in the expression profile (including changes in the spectrum of the transcripts expressed) may underpin their role in neoplasia, as opposed to specific mutational events as pathogenic mutations in SEPT9 have not been found in any of the exons, splice sites, or introns in cancer cells (19). In support of this model, significant amplification and overexpression of Sept9/Sint1, a murine orthologue of SEPT9 that is a recognized proto-oncongene with 96% homology, was found in a mouse mammary tumor virus transgenic mouse model of mammary adenocarcinomas (20). In addition, the same report also noted that mRNA from the SEPT9 locus was overexpressed in a few human BCCs, some of which are also reported here (20).

Here, we describe altered expression of SEPT9 in breast cancer cell lines (BCC) and primary breast cancers and show that altered expression of SEPT9 transcripts, specifically SEPT9_v1, affects pro-oncogenic phenotypes by enhancing cell proliferation, invasiveness, cell motility, foci formation, and aneuploidy. In addition, high expression of SEPT9_v1 changes the morphology of mammary epithelial cells from an epithelial to mesenchymal transition phenotype, a hallmark feature of epithelial cancers. Our data strongly suggest that SEPT9_v1 has important functional roles important to mammary tumorigenesis.

Cell lines. SUM1315, SUM102, SUM190, SUM159, SUM149, SUM52, and the nontumorigenic human papillomavirus (HPV)-immortalized mammary cell lines were developed and provided by S.P. Ethier (Karmanos Cancer Institute, Wayne State University, Detroit, MI). Additional BCCs and immortalized human mammary epithelial cell lines (IHMECs) were obtained from the American Type Culture Collection and grown under recommended conditions.

Genomic DNA and total RNA were isolated from BCCs and IHMECs following the manufacturer's instructions with the DNeasy and RNeasy kits, respectively (Qiagen, Inc.). cDNA was synthesized from 1 μg of total RNA using random primers and the Qiagen Omniscript Reverse Transcription kit.

Southern blot analysis. Ten micrograms of genomic DNA from each cell line were digested to completion with excess restriction endonuclease Taq1 under standard reaction conditions. Digested DNA was separated on a 1% agarose gel in 1× Tris-borate EDTA and transferred to a HybondN membrane (Amersham) in 20× SSC. A 0.5-kb PCR product from SEPT9 (forward, 5′-GGGTCCAGACTCCCCTACTC-3′; reverse, 5′-CTCTGCAGCTGAGACACAGG-3′) was used to generate an [α-³²P]dCTP radiolabeled probe using the random priming labeling system from Amersham Bioscience. Membranes were fixed in 0.4N NaOH and rinsed in 5× SSC and then preblocked with 250.0 μg human Cot1-DNA at 65°C. Hybridization was done overnight in 0.5 mol/L NaH₂PO₄, 1 mmol/L EDTA, and 7% SDS buffer (21). Blots were washed at 65°C in 2× SSC and 0.1% SDS buffer and exposed to Kodak XAR-5 film. The ratio between the intensity of the 4.4-kb SEPT9 band and the 3.1-kb glyceraldehyde-3-phosphate dehydrogenase (GAPDH) band used as the loading control was determined by densitometry (Alpha Innotech IS-1000 Digital Imaging System, version 2.00).

Gene amplification (duplex PCR) and expression (duplex reverse transcription-PCR). Conditions for semiquantitative duplex PCR were optimized using methods described previously (21). Genomic DNA and cDNA samples from IHMECs and BCCs were amplified with duplexed primers for GAPDH and SEPT9 using the number of cycles corresponding to the exponential range of amplification (before saturation). PCR products were visualized on agarose gels stained with ethidium bromide and quantified by densitometry. The ratios of the band intensity between GAPDH and SEPT9 in BCCs were normalized and compared with the ratios obtained from the IHMECs. The primer sets used for the amplification and expression profiles of SEPT9 are as follows (forward/reverse, 5′-3′): duplex PCR, SEPT9, TCCGTCTCCCCTCTGACTCT/TCTAGGGCTGACTCTGGGTG (197 bp); duplex PCR and reverse transcription-PCR (RT-PCR), GAPDH, GGGAGCCAAAAGGGTCATCA/TTTCTAGACGGCAGGTCAGGT (407 bp); and duplex RT-PCR, SEPT9-X (all variants), CAGTGATTGACACACCAGGG/ATCCGCTGTTTGAAGTGGAC (309 bp).

Primer sets used to analyze expression of the SEPT9 variants are listed in Supplementary Table S1.

Real-time quantitative RT-PCR. cDNA samples from IHMECs and BCCs were amplified in triplicate from the same starting material following the manufacturer's instructions. Samples were amplified using Taqman MGB FAM dye-labeled probes from Applied Biosystems in an ABI7900HT model real-time PCR machine. The following probes were used to amplify SEPT9 and the control GAPDH cDNA: Hs99999905_m1 (GAPDH) and MSFA-80 (custom primers for SEPT9_v1, probe sequence: 5′-AGCGGCGGCCACGG-3′).

Immunohistochemistry. Immunohistochemistry was done with the anti–SEPT9_v1-specific antibody at 1:200 dilution on paraffin-embedded or frozen sections of human breast tissue using standard methods. Primary antibody was detected following protocols described by the manufacturer (DakoCytomation), with diaminobenzidine as a chromogen and with hematoxylin counterstain. Optimization and validation of the immunostaining dilution conditions were done on several paraffin slices made from BCC pellets, which reflected different levels of SEPT9_v1 expression corresponding to the endogenous level of expression described by other methods in this article. Cells were visualized with an Olympus BX-51 microscope with a 60× objective lens. Samples were assessed by two independent scientists (M.E.G. and E.M.P.) and scored from 0 to +4 to indicate intensity, where “0” represented no SEPT9_v1 staining and “+4” represented dark, intense staining.

Western blot. Western blot analysis was done as described previously (21) using 50.0 μg of whole-cell lysates. The following antibodies were used: a custom-made rabbit polyclonal anti-SEPT9_v1 (NH₂-terminal epitope: KKSYSGGTRTSSGRLRR, BioCarta) at a 1:1,000 dilution, a rabbit polyclonal anti-SEPT9 kindly provided by W.S. Trimble (10) at a 1:1,000 dilution, mouse ascites anti-β-actin at a 1:10,000 dilution, mouse anti-vimentin at a 1:1,000 dilution, goat anti-rabbit horseradish peroxidase (HRP) secondary antibody at a 1:10,000 dilution, and goat anti-mouse HRP also at a 1:10,000 dilution (or 1:20,000 when used with anti-β-actin primary antibody). Antibodies were purchased from Sigma, unless otherwise noted, and diluted in 5% milk, 3% bovine serum albumin (BSA), and 0.05% Tween 20 in 1× TBS. The SuperSignal West Pico Chemiluminescent kit (Pierce Biotechnology) was used for detection before exposure to Kodak XAR-5 film. Relative to the loading control, semiquantitative protein expression levels were determined by densitometry (Alpha Innotech IS-1000 Digital Imaging System, version 2.00).

Cellular transfection and retroviral transduction. SEPT9 cDNAs were cloned into the retroviral pLNCX2 and pLPCX vectors (BD Biosciences). Constructs were transfected using FuGENE6 transfection reagent according to the manufacturer's instructions (Roche Diagnostics) into the PT67 packaging cell line. Retrovirus was then collected, filtered, and added to the target cell lines for 24 h. Stable polyclonal transductants in the MCF10A, HPV4-12, and Hs578T cell lines were selected for 1 week in complete medium supplemented with either geneticin for the pLNCX2 vector (1 mg/mL for HPV4-12 and Hs578T cell lines or 50 μg/mL for MCF10A cells) or puromycin for the pLPCX vector (0.5 μg/mL for all cell lines).

Cell proliferation assays. Two methods were conducted to examine cell proliferation. For one, cells were plated in triplicate at the same density and counted at different time points over the course of several days. For the other, cells were plated at the same density and cultured for 24 h (MCF10A and transductants) or 72 h (HPV4-12 and transductants) in a 96-well microplate. WST-1 reagent was added and absorbance at 450 nm was measured after 3 h of incubation following the manufacturer's instructions (Roche Molecular Systems).

Apoptosis assay. Subconfluent cultures of HPV4-12, MCF10A, and transductants were treated with 10 μmol/L camptothecin for 24 h to induce apoptosis; untreated cells were used as controls. Annexin V antibody conjugated to Alexa Fluor 488 was hybridized to the cells, which were then counted by flow cytometry, following the manufacturer's instructions (Vybrant Apoptosis Assay 2, Molecular Probes, Invitrogen Corp.). Propidium iodide was used as a counterstain to distinguish between apoptotic cells and dead/necrotic cells.

Invasion analysis. Transwell membranes coated with Matrigel were used to assay invasion in vitro. A suspension of HPV4-12 and MCF10A cells stably transduced with SEPT9_v1, SEPT9_v3, or empty vector and of MDA-MB-231 and BT-549, mock control, and SEPT9_v1-depleted cells was added to 24-well BD BioCoat Matrigel invasion chambers at 5 × 10⁴ cells/mL (BD Bioscience Discovery Labware). The invading cells were fixed and stained using the Protocol Hema stain set (Fisher Diagnostics). Three independent 20× fields for each well with chemoattractant (5% fetal bovine serum for HPV4-12, MDA-MB-231, and BT-549 and 5% horse serum + 20 ng/mL epidermal growth factor for MCF10A) and without (“no chemo,” only HPV4-12 and MCF10A) were counted for quantification.

Scrape motility assay. Untransduced cells, SEPT9-overexpressing stable transductants, and SEPT9-ablated cells were grown to confluency on two-chambered glass slides. The cell monolayer was mechanically scarred using a glass Pasteur pipette (22) and visualized for movement into the wound 22 h later with a Leica inverted microscope (phase-contrast optics, 20× objective). The open wound areas at the initial and end point for each motility assay were calculated using the Applied Biosystems ImageQuant TL version 2005 program. The ratio between time 0 and 22 h later (0 h/22 h) represented the motility for each cell line through the wound.

Foci formation assay. Hs578T transductants and parental cells were plated in 100-mm dishes at 80% confluence. Medium was changed every 3 to 4 days over 30 days. Cells were stained with methylene blue, and colonies were photographed.

Immunofluorescence. For immunofluorescence analysis, stable transductants and parental cell lines were grown on two-chambered glass slides and fixed with 4% paraformaldehyde for 40 min at room temperature. Slides were then washed thrice in 1× PBS for 10 min, blocked for 1 h in blocking solution (5% dry milk, 1% BSA, and 0.025% Triton X-100 in 1× PBS), and incubated overnight at 4°C in anti-SEPT9_v1, anti-SEPT9_X, monoclonal mouse anti-α-tubulin, or anti-vimentin (Sigma-Aldrich Corp.) antibody at 1:30 dilution in blocking solution. Phalloidin conjugated to Alexa Fluor 568 was used to identify filamentous actin (F-actin). Alexa Fluor 488 and Alexa Fluor 633 were used as secondary antibodies (Molecular Probes, Invitrogen) at a 1:500 dilution in blocking solution for 1 h at room temperature. Cells were visualized using an Olympus FV-500 confocal microscope (100× objective).

Mitotic index. HPV4-12, MCF10A, empty vector, and SEPT9 transductants were prepared as described previously (23) and stained with 0.54 mg/mL Giemsa solution and then destained in deionized water. Cells were visualized using a Leica DAS model microscope, and 1,000 cells were counted per sample in triplicate. The mitotic index was calculated as the percentage of cells with condensed chromosomes and lacking a nuclear membrane out of 1,000 cells.

Aneuploidy analysis. HPV4-12, MCF10A, empty vectors, and SEPT9 transductants were grown to 70% confluence, trypsinized, and collected by centrifugation at 900 rpm for 5 min. Cells were then washed with 1× PBS and fixed by dropwise addition of a 3:1 mixture of methanol and glacial acetic acid and incubated overnight at 4°C. Fresh fixative was then added, and the resuspended cell pellets were dropped onto clean microscope slides. Slides were air dried, stained in 0.54 mg/mL Giemsa solution, and destained in deionized water. After air drying, chromosomes were visualized and counted using a Leica DAS model microscope. Twenty-five metaphases were counted for aneuploidy, in triplicate, for each sample.

Knockdown of SEPT9_v1 expression in BCCs. Transient small interfering RNA (siRNA) knockdown was used in MDA-MB-231 cells, a cell line primarily expressing the SEPT9_v1 isoform, following the manufacturer's protocol. Cells were plated in triplicate at a density of 4.0 × 10⁵ per well in a six-well plate and subsequently transfected with DharmaFECT2 and the MSF siRNA pool. This pool of four siRNAs targets all SEPT9 isoforms (Dharmacon Research, Inc.). Target sequences for the four duplexes in the siRNA-SEPT9 (MSF) pool correspond to the following loci of the SEPT9_v1 transcript (accession number AF189713): 5′-GGAGAUCACCAUCGUCAAA (546–564 bp); 5′-GGUAAAUCCACCUUAAUCA (1,024–1,042 bp); 5′-GAGAAAGGCGUCCGGAUGA (1,147–1,165 bp); and 5′-CCAACGGCAUCGACGUGUA (1,496–1,514 bp). Mock-transfected and siCONTROL RISC-Free siRNA supplied by the manufacturer were used as negative controls. Functional studies were done 72 h after transfection, which was the time point with the greatest decrease in expression.

Stable knockdown in BT-549 cells was achieved using the pRNA-H1.1/Retro vector with either a SEPT9_v1-specific short hairpin RNA (shRNA) construct or a scrambled sequence shRNA cassette provided by the manufacturer as a negative control (GenScript Corp.). The SEPT9_v1 shRNA construct targeted the unique 5′-untranslated region (UTR), spanning 15 to 33 bp (sense strand target sequence: 5′-GGCCCAGGATTAGCGCCCT-3′; Genbank accession number AF189713). Following sequence verification of the constructs, 5 μg of purified plasmid DNA were transfected into PT67 cells to produce viruses, which were then used to retrovirally transduce the BT-549 target cell line. Polyclonal stable cell lines were selected in hygromycin (40 μg/mL) for 1 week.

Statistics. Data are presented as the mean of triplicate experiments, and error bars represent the SE. The ANOVA test, Student's t test, and the χ² test were used to determine statistical significance as indicated.

SEPT9 is amplified in BCCs. Because activation of proto-oncogenes can occur through gene amplification, we examined the genomic copy number of SEPT9 in 22 breast cell lines using semiquantitative duplex PCR and Southern blotting. Both methods showed amplification in 12 of 18 (67%) BCCs compared with no amplification in IHMECs (P < 0.025; Fig. 1A). A representative comparison between a subset of BCCs and IHMECs using the two methods is shown in Fig. 1A (right). Of note, SEPT9 copy number was low in CAL51 cells, a line containing a confirmed 17q25 deletion (24), supporting the sensitivity of our assays to detect genome copy number differences.

SEPT9_v1 is an overexpressed transcript in BCCs. We did not find any mutations in the coding sequence or intron-exon boundaries for the SEPT9 locus (data not shown), so we next assessed expression of SEPT9 in BCCs. We initially used semiquantitative duplex RT-PCR to investigate the expression of SEPT9 in our series of 18 BCCs and 4 IHMEC lines (Supplementary Fig. S1A). Using primers that recognized all variants (SEPT9-X), we detected high expression in 72% of BCCs when compared with the mean of the expression level of all four IHMECs (Supplementary Fig. S1A). Only one of four IHMEC lines, MCF10A, showed significant expression. Nine of the 12 BCCs with amplification at the DNA level (Supplementary Fig. S1A, asterisk) had high levels of gene expression compared with the IHMECs, suggesting that gene amplification might be one mechanism for increased mRNA expression from the SEPT9 locus in some BCCs.

Next, we used transcript-specific primers for semiquantitative duplex RT-PCR analysis to determine which of the SEPT9 transcripts influenced the overall expression profile (Supplementary Table S1; Supplementary Fig. S1A). Ten of 18 (55%) showed slightly higher levels of SEPT9_v4* expression and 6 of 18 (33%) BCCs overexpressed the distinct SEPT9_v1-5 transcript compared with only one of four (25%) IHMECs. However, SEPT9_v4, SEPT9_v2, and SEPT9_v5 were expressed in few BCCs and IHMECs and SEPT9_v3 expression showed no significant difference between BCCs and IHMECs. Of significant, isoform SEPT9_v1 was highly expressed in 61% of BCCs compared with just one (25%) IHMEC line (Fig. 1B; Supplementary Fig. S1A). Of note, all BCCs with high endogenous expression of SEPT9_v1 are in the group that showed high expression of SEPT9-X, indicating that this transcript is responsible, at least in part, for the overexpression of the SEPT9 gene. These results were confirmed by quantitative real-time RT-PCR using a SEPT9_v1-specific probe normalized to GAPDH expression (Fig. 1B and C).

SEPT9_v1 protein is overexpressed in primary breast cancers. We tested 17 matched pairs of normal and primary breast cancer tissues by immunohistochemistry for SEPT9_v1 expression using an isoform-specific antibody. Strikingly, 12 of the 17 paired samples (70%) were negative or very weak (0 or +1) for SEPT9_v1 staining in primary normal mammary epithelial tissues but the patient-matched tumor tissue stained very intensely (+3 or +4) for SEPT9_v1 (P < 0.02, χ² test; Fig. 1D; color images in Supplementary Fig. S1C).

SEPT9_v1 protein is overexpressed in BCCs. We did Western blotting analysis using our polyclonal SEPT9_v1 antibody and a polyclonal SEPT9-X antibody, which recognized all isoforms but SEPT9_v5, to characterize protein expression of the variants (Fig. 2A). When using the SEPT9-X antibody for Western blotting, the smaller size of the SEPT9_v4*/SEPT9_v4 isoforms allowed us to analyze their protein expression. However, as the other isoforms were very similar in sizes, individual quantification was not possible. It was noted that SEPT9_v4 was found overexpressed in 38% of BCCs compared with IHMECs.

Using the isoform-specific antibody, we found high SEPT9_v1 protein expression in 55% of BCCs compared with IHMECs (P < 0.025). This was validated by immunohistochemistry on cell pellets from BCCs (Supplementary Fig. S1B). Seven of the BCCs had high SEPT9_v1 mRNA and protein levels (Fig. 2A , asterisk). Differences between the mRNA and protein levels in the remaining BCCs suggested post-translational modification to regulate expression. These results supported our RT-PCR observations and further suggested that preferentially increased expression of the SEPT9_v1 isoform may be relevant to breast cancer. Interestingly, the four BCCs with no SEPT9_v1 protein expression have an increased expression of alternative isoforms, the relevance of which remains to be elucidated.

Overexpression of SEPT9_v1 leads to the acquisition of tumorigenic phenotypes. To examine the role of increased expression of SEPT9 on cellular characteristics, we retrovirally transduced several cell lines that showed no or low endogenous SEPT9_v1 expression, HPV4-12, MCF10A, and Hs578T, with cDNA constructs of two SEPT9 isoforms, SEPT9_v1 and SEPT9_v3 (Fig. 2B). Interestingly, overexpression of isoform SEPT9_v3 dramatically lowered the growth rate of the cells, suggesting a potential tumor-suppressive or dominant-negative function, which will be characterized further in future studies (Fig. 2C,, left). However, overexpression of SEPT9_v1 dramatically increased the proliferation rate of all the cell lines, both by manual counting of cells and by the WST-1 proliferation assay, when compared with controls (Figs. 2C and 3A, respectively).

To determine if impaired apoptosis contributed to the increased growth rate, we assessed the percentage of apoptotic cells following camptothecin treatment to induce apoptosis. HPV4-12 cells were insensitive to camptothecin-induced apoptosis, likely due to impaired p53 function as a result of the HPV E6/E7 immortalization (25). There was no significant difference in apoptotic response between the controls and SEPT9_v1-overexpressing cells for all cell lines without camptothecin treatment. However, overexpression of SEPT9_v1 significantly decreased the percentage of apoptotic cells in the MCF10A line following exposure to camptothecin (Fig. 3B).

We also noted striking morphology changes for cells overexpressing SEPT9_v1, and for some cells overexpressing SEPT9_v3, compared with control cells in the Hs578T and MCF10A lines (Fig. 4A). The controls displayed the usual cuboidal appearance of epithelial cells. The cells ectopically expressing SEPT9_v1 became elongated, almost fibroblastic or mesenchymal in appearance, which is typical for cancers of epithelial origins.

As oncogenes often will promote cellular invasion when overexpressed, we used the Matrigel model of a basement membrane to test our overexpressing transductants. We found that ectopic expression of SEPT9_v1, but not SEPT9_v3, in MCF10A, HPV4-12, and Hs578T cells significantly increased invasiveness compared with controls (Fig. 4B). Additionally, multiple polyclonal lines of MCF10A cells overexpressing SEPT9_v1 showed increased invasion, indicating that the results were not due to a predominant clone (Supplementary Fig. S2A). Using the scrape motility assay, we found that SEPT9_v1 overexpression increased motility compared with the other variants and the controls used in this assay (Fig. 4C; Supplementary Fig. S3A). The ability of cancer cells to metastasize indicates that both invasion and motility are relevant to tumorigenesis.

Next, we tested if SEPT9_v1 ectopic expression could promote a phenotype indicative of cellular transformation by doing a focus formation assay. Foci formation was not observed in the immortalized cell lines HPV4-12 and MCF10A (data not shown), indicating that ectopic expression of SEPT9_v1 alone is not sufficient to transform normal cells. However, the BCC Hs578T transduced with SEPT9_v1 showed enhanced foci formation when compared with SEPT9_v3 transductants or empty vector and parental cells (Fig. 4D).

SEPT9_v1 interacts with vimentin in BCCs. The BCC Hs578T is classified in the mesenchymal-like group and expresses high levels of vimentin, a cytoskeletal protein marker for mesenchymal cells that has been shown to increase cellular motility and invasiveness in MCF7 cells (26). We hypothesized that this feature of Hs578T cells, in addition to the overexpression of SEPT9_v1, could contribute to the increased invasion and foci formation of the SEPT9_v1 transductants when compared with the IHMEC lines, which do not express high levels of vimentin. Therefore, we studied the interaction between vimentin and SEPT9_v1 by immunoprecipitation. We found coimmunoprecipitation between vimentin and SEPT9_v1 only in Hs578T cells overexpressing SEPT9_v1 but not in HPV4-12 cells overexpressing SEPT9_v1 or either parental cell line, indicating that up-regulation of both vimentin and SEPT9_v1 was needed for their interaction (Supplementary Fig. S4A). In support of this, we also noted strong colocalization of SEPT9_v1 and vimentin by immunofluorescence when SEPT9_v1 was overexpressed (Supplementary Fig. S4B, j and t). In addition, overexpression of SEPT9_v1 in MCF10A and Hs578T cells increased vimentin expression as shown by immunofluorescence (Supplementary Fig. S4B, i and s versus d and n). These data further support the epithelial to mesenchymal morphology change described above and are suggestive of coordination between vimentin and SEPT9_v1 for cellular transformation.

SEPT9_v1 overexpression interferes with tubulin filaments. To determine which properties of overexpressed SEPT9_v1 versus SEPT9_v3 (which differ only by their first 25 amino acids) could promote cellular transformation, we stained HPV4-12 cells overexpressing each variant with either anti-SEPT9_v1 or anti-SEPT9-X (to detect SEPT9_v3) antibodies for immunofluorescence. Endogenous SEPT9_v1 localized to the cytoplasm in interphase cells with minimal nuclear staining (Fig. 5A,, b; Supplementary Fig. S5A, b). However, ectopically expressed SEPT9_v1 showed this isoform was more abundant in both the nucleus and the cytoplasm, with an altered morphology of the cells (Fig. 5A,, f; Supplementary Fig. S5A, n). In contrast, ectopic SEPT9_v3 showed a distinctive localization to the periphery of the cells, around the cellular membrane, not seen in parental cells (Supplementary Fig. S5A, h). SEPT9_v3 colocalized strongly with F-actin and only minimally with microtubule filaments with no observable effect on cytoskeletal architecture (Supplementary Fig. S5A, k and l, respectively). SEPT9_v1 colocalized with both F-actin and α-tubulin in interphase cells (Fig. 5A,, d and h; Supplementary Fig. S5A, e, f, q, and r). Interestingly, when SEPT9_v1 was overexpressed, there was a dramatic disorganization of the tubulin filaments in 68% of interphase cells (Fig. 5A , g; Supplementary Fig. S5A, p) compared with 0% in cells overexpressing SEPT9_v3 (Supplementary Fig. S5A, j) and 11% of cells in the parental HPV4-12 line (P < 0.001, ANOVA). This was in strong agreement with the differences in the subcellular localization of the two variants.

SEPT9_v1 overexpression causes localization to the nucleus and interferes with mitosis. To assess whether overexpression of SEPT9_v1 can affect the mitotic machinery, we stained SEPT9_v1-HPV4-12 transductants with either SEPT9_v1 or SEPT9-X antibodies for immunofluorescence. Only 1% of cells overexpressing SEPT9_v3 showed localization to the nucleus and no other effects were found under the conditions assayed here (data not shown). In contrast, SEPT9_v1 overexpression had a greater effect on cell division: 10% of cells were binucleated, 10% localized SEPT9_v1 to the nucleus, and ∼2% showed a combination of these (Fig. 5B,, b–d, respectively, including graph; Supplementary Fig. S5B). However, the nuclear localization could not be directly associated with the cytokinesis defect. In addition, SEPT9_v1 overexpression in HPV4-12 and MCF10A cells increased the mitotic index versus the parental and empty vector controls (Fig. 5C). These results suggested that overexpression of the SEPT9_v1 transcript is causing the deregulation of mitosis by promoting mitotic entry earlier than the parental and empty vector counterparts.

Due to the presence of binucleated giant cells described earlier, we assayed the cells overexpressing SEPT9_v1 for aneuploidy, a characteristic of genomic instability in many cancer cells. We found a 10-fold increase in aneuploid cells when SEPT9_v1 was overexpressed compared with the parental and the vector controls for the HPV4-12 and MCF10A cell lines (Fig. 5D). Similar results were observed for multiple polyclonal populations of MCF10A cells overexpressing SEPT9_v1 (Supplementary Fig. S2B).

Depletion of SEPT9_v1 in BCCs reverses tumorigenic phenotypes. To test further our hypothesis that the SEPT9_v1 isoform contributes to tumorigenic phenotypes, we questioned whether lowering its expression would reverse these phenotypes in BCCs that normally showed high endogenous expression. We used two different approaches in two BCCs. First, we transiently transfected MDA-MB-231 cells, which predominantly overexpressed SEPT9_v1 protein compared with other isoforms, with a pool of four siRNAs that targeted the SEPT9 locus. Seventy-two hours after transfection, there was an 80% decrease of SEPT9_v1 mRNA and a 60% decrease in SEPT9_v1 protein levels compared with the mock-transfected and nontargeting (siCONTROL) negative controls (Fig. 6A,, left). Other SEPT9 variants were not affected (Fig. 6A,, left; data not shown). In addition, we used a stable shRNA construct to permanently and specifically decrease SEPT9_v1 expression in BT-549 BCCs by targeting the isoform-specific 5′-UTR. This resulted in an 80% decrease in endogenous SEPT9_v1 mRNA expression and decreased protein expression by >70% compared with the parental and the scrambled shRNA controls (Fig. 6A , right).

For both lines, SEPT9_v1-depleted cells showed inhibited cellular proliferation (Fig. 6B), decreased invasion in the presence of chemoattractant (Fig. 6C), and impaired cellular motility (Fig. 6D; Supplementary Fig. S3B) when compared with mock-transfected and nontargeting/scrambled construct controls. In MDA-MB-231 cells in which SEPT9_v1 expression is transiently decreased by SEPT9 locus siRNA, the growth curves indicated a significant difference between the SEPT9 knockdown cells and the control cells at day 4 (P = 0.03, Student's t test), whereas there is not a significant difference between the parental and the negative controls. However, at day 8, the difference in growth rates becomes less significant, likely due to a return of SEPT9_v1 expression because loss of expression was transient and increasing by day 6 as indicated by the Western blot. As expected, the results obtained with the stable knockdown were much more striking than the transiently depleted cells, likely due to a more efficient and isoform-specific knockdown of SEPT9_v1 in these cells. These two strategies support the initial hypothesis implicating SEPT9_v1 in cell transformation in cell culture studies.

To begin testing whether altered SEPT9 expression was relevant to the pathogenesis of breast cancer, we did mutation analysis, genomic amplification studies, and expression studies. Although mutations in the open reading frame were not found, we detected increased SEPT9 genomic copy number and a corresponding increase in mRNA expression in the majority of BCCs. Analysis of SEPT9 variant transcript expression in BCCs by multiple RT-PCR methods using transcript-specific primers revealed that SEPT9_v1 was the SEPT9 variant most often overexpressed when compared with IHMECs followed next by SEPT9_v4*. Similar results were found when protein expression for the different isoforms was examined. It is interesting that the SEPT9 transcripts SEPT9_v1 and SEPT9_4* have also been shown to be up-regulated in ovarian tumors (12, 17). This suggested that the regulation of alternative splicing of the SEPT9 transcript is altered in multiple types of cancers. Importantly, we also found that, in 70% of patient-matched normal and primary breast cancer tissue samples we tested, SEPT9_v1 expression is absent in normal tissue but highly expressed in the cancer tissue. Future studies should determine if there is a significant correlation between SEPT9_v1 isoform expression and clinical and pathologic variables relevant to tumorigenesis, cancer progression, and treatment.

We next hypothesized that increased expression of SEPT9 might contribute to malignant progression in mammary epithelial cells. This hypothesis was supported by reports of increased expression of SEPT9 in human and murine cancers (14, 16, 19, 20). In addition, SEPT9 as well as other human septins (i.e., SEPT5, SEPT6, and SEPT11) are well-described MLL-fusion partners noted in acute leukemia cells. This evidence suggested roles for one or more SEPT9 variants in tumorigenesis, but no compelling data supporting a role for SEPT9 in breast cancer had been reported to date. To test our hypothesis, we retrovirally transduced cDNA constructs of SEPT9 variants into IHMECs (HPV4-12 and MCF10A), which promoted the development of cellular phenotypes characteristic of malignant cells. Specifically, ectopic expression of SEPT9_v1 induced increased growth kinetics, accelerated cell proliferation, enhanced invasiveness, an epithelial to mesenchymal morphology change with increased vimentin expression, and increased cell motility when compared with controls. Similar results were seen, as well as enhanced foci formation, when SEPT9_v1 was ectopically expressed in BCC with low SEPT9-X and no SEPT9_v1 endogenous expression, Hs578T. Knockdown of SEPT9, and specifically SEPT9_v1, in two cancer cell lines with high endogenous levels of SEPT9_v1 (MDA-MB-231 and BT-549) using RNA interference techniques substantiated our conclusions by rescuing tumorigenic phenotypes. Together, our results provide compelling support that SEPT9_v1 may function as an oncoprotein in mammary epithelial cells and drive malignant progression. The association with invasiveness, motility, and a mesenchymal phenotype suggests that up-regulation of SEPT9_v1 may be functionally important in progression from invasive to metastatic phenotypes.

The other abundant SEPT9 variant detected in our assays, SEPT9_v4*, was previously noted to be up-regulated in tumor cells and showed only mild evidences of cell transformation, suggesting that it might be a manifestation of tumorigenesis rather than a significant functional contributor to malignant progression (19). This work, combined with the data presented here, indicates that the SEPT9_v1 isoform contributes to oncogenesis when the SEPT9 locus is amplified or deregulated. However, future work to determine if the other SEPT9 isoforms are also involved in tumorigenesis, either as oncogenes or as tumor suppressors, will be necessary to clarify the role of this locus in breast cancer development.

We also were interested in the potential mechanisms by which SEPT9_v1 may contribute to oncogenesis. Other septin heterotrimers (i.e., SEPT2/6/7) are recognized regulators of microtubule stability, and septin depletion has resulted in marked stabilization of microtubules and mitotic defects in vivo (7). Most relevantly, SEPT9 depletion reportedly caused incomplete cell division and the accumulation of binucleated cells. Its localization along the microtubules of the mitotic spindle in human mammary epithelial cells suggested some function in mitotic processes (10). These findings in the literature suggested a plausible mechanism by which deregulation of SEPT9 expression may affect cell division by altering the microtubules.

Using immunofluorescence, we found different cellular localization patterns for SEPT9_v1 compared with SEPT9_v3, where SEPT9_v1 strongly colocalized with the cytoplasmic tubulin filaments and impaired their organization, and SEPT9_v3 was more strongly aligned with the actin filaments in interphase cells. These data, combined with our observation that the SEPT9_v3 isoform dramatically decreased growth rates in MCF10A cells, support the idea that these isoforms of the same gene could be dynamically rearranged under the guidance of different signaling pathways to accomplish different intracellular functions. In particular, increasing SEPT9_v1 expression in cells resulted in a morphology change, increased intensity of nuclear staining, and destabilization of the tubulin filaments in interphase cells. This indicated a potentially unique cellular role for SEPT9_v1 as a regulator of the microtubule components of the cytoskeleton. In addition, our findings of increased giant binuclear cells, higher mitotic index, and increased aneuploidy in IHMECs overexpressing SEPT9_v1 suggested that this isoform may alter the microtubules of the mitotic spindle and/or impair cytokinesis. This would be highly relevant to malignant progression and could have therapeutic implications. These findings are further supported by reports of cytokinesis defects associated with SEPT9_v1 gene silencing (8). Taken together, this indicates that the amount of SEPT9_v1 present in the cell, or its stoichiometry with other septins, is highly important in properly regulating tubulin filament organization, mitosis, and cytokinesis.

To our knowledge, the effects of SEPT9_v1 on cell proliferation, motility, and invasion in the context of cellular transformation have not been previously examined. Our studies provide compelling evidence supporting a role for SEPT9_v1 in mammary tumorigenesis. Future efforts focused on mechanistic studies, to elucidate how specific SEPT9 isoforms coordinate different cellular functions and to determine what regulates expression of one variant over another, should provide additional novel insights to the functional roles of SEPT9 in health and disease. In addition, further characterization of the SEPT9 variants in in vivo models should enhance our understanding of how cellular division and proliferation normally proceeds and how aberrations can lead to deleterious states, such as breast cancer.

Grant support: NIH National Cancer Institute grant R01CA072877 (E.M. Petty), Department of Defense Breast Cancer Research Predoctoral Fellowship BC050310 (L.M. Privette), NIH National Research Service Award 5-T32-GM07544 from the National Institute of General Medicine Sciences (L.M. Privette and E.A. Peterson), and Department of Defense USAMRMC BC 980697 (L.M. Kalikin).

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.

We thank Hernan Roca for helpful data analysis and discussion and Donita L. Sanders for assistance with immunohistochemical protocols.