AKT-mediated phosphorylation of Sox9 induces Sox10 transcription in a murine model of HER2-positive breast cancer

Fragments of the Sox10 promoter were amplified by PCR (see Supp. Table 1) from the bacterial artificial chromosome (BAC) RP23-424P8 which contains up to 12 kb of the Sox10 promoter and was purchased from the Centre of Applied Genomics (SickKids, Toronto, Canada). Each fragment of the Sox10 promoter was amplified using the indicated primers in Supplementary Table 1, digested with KpnI and XhoI (which were included in the forward or reverse primer, respectively) and ligated into pGL3P which was linearized with KpnI and XhoI.

pGEX-Sox9 was generated by PCR amplifying Sox9 cDNA from pCMV6-myc-DDK-Sox9 with the indicated primers in Supplementary Table 1, digested with BamHI and XhoI (which were included in the forward or reverse primer, respectively) and ligated into pGEX-4T2 which was linearized with BamHI and XhoI.

pGEX-Sox9^1-223 was generated by linearizing and purifying pGEX-Sox9 with SmaI and NotI. The linear fragment was then blunted with Klenow (New England Biolabs) and purified from a 0.8% agarose gel using the QIAGEN Gel Extraction Kit (28704). The linearized vector was then incubated with T4 DNA ligase (New England Biolabs) to generate the circular plasmid.

Sox9 S181A mutants were generated using either pGEX-Sox9^1-223 or pCMV6-myc-DDK-Sox9 as template. The point mutation was generated using the QuickChange XL Site-directed Mutagenesis Kit according to the manufacturer’s protocol. The primers used to generate the point mutations are listed in Supplementary Table 1.

All mammary tumor cell lines were maintained in DMEM/F12 containing 10% FBS, 1% mammary epithelial growth supplement (Life Technologies), 1% penicillin/streptomycin, and 1% l-glutamine. Mammary tumor cell lines were isolated as described from individual tumors at endpoint [23]. All experiments were performed on three independent isolates and representative data are shown. All mammary tumor cell lines were used within 30 passages of initial isolation. All other cell lines used were maintained in DMEM containing 10% FBS, 1% penicillin/streptomycin, and 1% l-glutamine. Cells were cultured at 37°C in a humidified incubator set at 5% CO₂. Cells were routinely passaged at a one in ten dilution every 2–3 days.

For plasmid and siRNA (Supplementary Table 3) transfections, cells were seeded such that they would reach approximately 75% confluency the day of the transfection. Lipofectamine 3000 was used for all transfections according to the manufacturer’s protocol using 8 μg of plasmid DNA or 200 nM of each siRNAs. Transfections were performed for 48–72 h prior to cell harvest.

The tissues were harvested and fixed in 10% buffered formalin phosphate for 24 h and then transferred to 70% ethanol for storage. The tissues were formalin fixed, paraffin-embedded, and sectioned at a 5-μm thickness. Tissue sections were deparaffinized and subjected to antigen retrieval in 10 mM citrate buffer (pH 6.0) and quenched with 3% hydrogen peroxide. Sections were blocked in 5% donkey serum and incubated overnight with the indicated primary antibody (Supplementary Table 2) at 4°C. Sections were washed followed by incubation with the appropriate HRP-conjugated secondary antibody. For mouse monoclonal antibodies on mouse tissues, mouse-on-mouse blocking solution (Vector Laboratories) was added to the blocking step. All IHC staining were carefully controlled with no primary antibody controls. Staining was developed using DAB substrate (Sigma Aldrich), and sections were counterstained with hematoxylin. Sections were dehydrated in ethanol then xylene and mounted. Sections were imaged using the Aperio Scanscope (Leica Biosystems), and images were processed and/or analyzed using Imagescope or ImageJ. For quantification, stained and scanned sections were opened in ImageScope and the percentage of DAB-positive (brown) pixels for each core was quantified using the built-in Aperio positive pixel count algorithm. Only pixels that fell into the “strong positive” default setting were counted. For the murine hyperplastic lesions, a total of 10 mammary glands were surveyed for each genotype.

Tissues and cell lines were homogenized in RIPA lysis buffer containing protease inhibitors (0.05% SDS, 1% Triton X-100, 1% NP-40, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, pH 8.0, 12 mM Na-Deoxycholate, 10 mM NaF, 1 mM DTT, 10 mM β-glycerophosphate, 0.6 mM NaVO₃, 1 mM PMSF, 10 μg/mL leupeptin, 10 μg/mL aprotinin, 10 μg/mL pepstatin, and 100 μM benzamide). Lysate was cleared by centrifugation and protein concentrations were determined using Bradford reagent (Bio-Rad). Equal amounts of protein lysate were denatured and subject to SDS-PAGE. Samples were then transferred to a PVDF membrane and probed with the indicated primary antibody (Supplementary Table 2) overnight at 4°C in 5% BSA. The membranes were washed and incubated with the appropriate HRP-conjugated secondary antibody. Reactive proteins were detected by chemiluminescence (Perkin Elmer) and exposure to X-ray film.

RP bacteria were transformed with the indicated GST or GST-fusion plasmids and glycerol stocks were stored at −80°C. Glycerol stocks were used to inoculate 3 mL of LB broth containing the appropriate antibiotic, and cultures were left to grow overnight at 37°C with shaking at 225 rpm. The 3-mL cultures were then expanded to 30 mL and left to shake at 225 rpm for 1 h at 37°C. GST-fusion protein production was then induced with 1 mM IPTG for 2 h at 37°C with shaking at 225 rpm. The bacterial cultures were then pelleted at 3000 rpm for 15 min at 4°C. Bacterial pellets were lysed in 500 μL of RIPA lysis buffer containing protease inhibitors and incubated on ice for 15 min with occasional vortexing. Lysates were then sonicated with three 15-s pulses and cleared by centrifugation at 14000 rpm for 30 min at 4°C. GST-fusion proteins were then purified with 20 μL of GST-sepharose beads (GE Healthcare) using the immunoprecipitation protocol.

In vitro kinase assays were performed as previously described [11]. Briefly, GST samples were pulled down and washed beads were resuspended in 1x kinase buffer (20-mM Tris-HCl, pH 7.4, 0.25-mM NaVO₃, 1-mM NaF, 10-mM β-glycerophosphate, 1-mM DTT, and 15-mM MgCl₂). The kinase assay was initiated by the addition of 1 μL of [³²P]γATP (5 μCi/μL, Perkin Elmer) and incubated for 30 min at 30°C. Recombinant human GST-AKT fusion protein (200ng; Sigma Aldrich, SRP-5001) was added to the kinase assay master mix. The reactions were terminated with the addition of 4x SDS sample buffer (200-mM Tris-HCl, 400-mM DTT, 8% SDS, 0.4% bromophenol blue, and 40% glycerol) and subject to SDS-PAGE. The incorporation of the radiolabeled phosphate was detected by autoradiography and the efficiency of immunoprecipitation was determined by Western blot.

The total RNA was isolated using Trizol (Invitrogen). cDNA synthesis was performed using the Superscript III Reverse Transcriptase (Life Technologies) according to the manufacturer’s protocol. qRT-PCR was performed using an Applied Biosystems 7500 Real-Time Fast PCR thermocycler. Relative mRNA expression was calculated using the ΔΔC_T method and normalizing to total levels of ribosomal 18S. For microarray analysis, total RNA was hybridized to the Mouse Gene 2.0 ST Array (Affymetrix). Microarray data analysis was carried out in the r statistical programming environment with Bioconductor (GSE128514). Primers used for qRT-PCR are listed in Supplementary Table 1.

Genomic DNA for bisulfite conversion was isolated using the DNeasy Blood and Tissue Kit following the manufacturer’s protocol (QIAGEN). Bisulfite conversion of genomic DNA was performed using the EpiTect Plus DNA Bisulfite Kit (QIAGEN) following the manufacturer’s protocol. Promoter regions of control and bisulfite converted DNA were amplified using the primers indicated in Supplementary Table 1 using Taq polymerase (Invitrogen). Amplicons were subcloned into pGEM-T (Promega) according to the manufacturer’s protocol. Five clones for each conversion reaction were picked, miniprepped, and sequenced to identify methylated cytosine residues.

For luciferase assays, 1.5 × 10⁵ cells were seeded in triplicate wells of a 6-well plate. The following day, the cells were transfected using lipofectamine 3000 with 1.25 μg of the appropriate pGL3P reporter construct and 1.25 μg of pRL-CMV (encoding Renilla luciferase) for 48 h. Luciferase assays were performed using the Dual Luciferase Assay Reporter System (Promega). Cells were collected in 1.5 mL Eppendorf tubes and lysed in 100 μL of 1X passive lysis buffer with rotation at room temperature for 30 min. Lysates were cleared by centrifugation at 13000 rpm for 10 min at room temperature. 20 μL of each sample was transferred in triplicate into a black-sided 96-well plate with a clear bottom. 100 μL of resuspended Luciferase Assay Substrate was added to each well and read using a luminometer with a read time of 10 s and a 2-s delay per well. 100 μL of Stop and Glo solution was then added to each well and read using a luminometer with a read time of 10 s and a 2-s delay per well. Luciferase counts were normalized to the pRL-CMV counts to account for differences in transfection efficiency between wells.

For chromatin immunoprecipitation, 5×10⁶ cells from each condition were seeded in two 15-cm plates and cultured for 48 h. Cells were then cross-linked in 1% formaldehyde for 10 min at room temperature. The formaldehyde was then quenched with the addition of glycine to a final concentration of 125 mM. Nuclear extraction, chromatin digestion, and immunoprecipitation and DNA elution were all performed using the SimpleChip Kit (Cell Signaling Technologies) according to the manufacturer’s protocol with the addition of a pre-clearing step with beads alone prior to the immunoprecipitation. 10 μg of chromatin was used per immunoprecipitation with 1 μg of the indicated antibody. Following chromatin IP and DNA cleanup, 6.5 μL of chromatin was mixed with 35 μL of iTaq Universal SYBR Green Supermix (BioRad), 6.5 μL of the indicated primers (5μM), and 21 μL of nuclease-free water. The PCR reaction was started with an initial hold and denaturation at 50°C for 5 min and 95°C for 10 min, respectively, followed by 40 cycles of denaturation at 95°C for 15 s and Annealing and Extension at 60°C for 60 s. The percent input was then calculated as follows: percent input = 2% × 2^{(C[T] 2% input sample – C[T] IP sample)}. The antibodies used for chromatin immunoprecipitation can be found in Supplementary Table 2. The primers used for ChIP-qPCR can be found in Supplementary Table 1.

We have previously shown that deletion of the Ste20-like kinase SLK induces Sox10 expression, enhancing tumorigenesis in vivo [23]. This is also accompanied by AKT upregulation and increased tumor stem/progenitor cell activity [23]. To gain further insights into the molecular mechanisms regulating Sox10 expression, we first assessed any potential epigenetic changes at the SOX10 locus. The SOX10 promoter has previously been shown to be epigenetically silenced in human gastric cancers and metastatic melanoma [24‐26], the basal level of Sox10 expression is low in vitro, we hypothesized that Sox10 gene induction could be due to gene demethylation in SLK^-/- NDL cells. We identified two putative CpG islands immediately upstream of the Sox10 transcription start site using the MethPrimer CpG island prediction software (Supp. Figure 1A) [26]. To assess whether the Sox10 gene was methylated, we treated SLK^fl/fl and SLK^-/- NDL cells with 5-aza-2’-deoxycytidine for 5 days to effectively eliminate any methylation marks within the genome. Following treatment, only a modest increase in Sox10 mRNA was observed in the SLK^-/- NDL cells with no significant differences observed in SLK expressing controls (Supp. Figure 1B). The lack of a robust Sox10 induction following 5-aza-2’-deoxycytidine treatment suggests that these putative CpG islands are not heavily methylated in either cell line. To validate this, we isolated genomic DNA and performed bisulfite sequencing on both putative CpG islands and compared the methylation pattern in SLK^fl/fl and SLK^-/- NDL cells. Using this approach, we found that only approximately 20% of cytosines within these CpG islands are methylated in either cell line with no significant differences observed following SLK deletion (Suppl. Figure 1C). Together, these results indicate that demethylation of the Sox10 promoter is not the main mechanism of Sox10 induction following SLK deletion.

As epigenetic silencing is not responsible for the differential expression of Sox10 in SLK^fl/fl and SLK^-/- NDL cells, we sought to identify potential signaling systems regulating Sox10 expression. As we have shown that AKT activity is required for maintenance of Sox10 expression in SLK knockout mammary tumor cells [23], we first assessed the levels and nuclear localization of known transcription factors that are AKT-responsive. Direct AKT targets implicated in transcription control include CREB and Forkhead box factors (FoxA1 and FoxO family [15]). Cellular fractionation and Western blot analysis (Fig. 1a) showed no difference in the levels and localization of those factors, suggesting that Sox10 is likely activated through other AKT-dependent mechanisms. To bootstrap our way back to transcriptional regulators controlling Sox10 expression in SLK-null cells, we analyzed potential regulatory regions around the Sox10 gene. Comparative sequence analyses have previously identified multiple-species conserved sequences (MCS) that control Sox10 expression [27]. As two of these MCS (4 and 7) have been shown to regulate Sox10 expression in mammary epithelial cells (Fig. 1b for schematic), we tested whether these enhancer elements were differentially regulated in SLK^fl/fl and SLK^-/- NDL cells. Both enhancer elements drove luciferase activity above the levels of the control vector; however, no significant differences between cell lines was observed (Fig. 1c), suggesting that these enhancer sequences do not mediate the differential Sox10 expression observed in the SLK^-/- NDL cells but perhaps mediate basal expression of Sox10.

Since little is known about the mechanisms regulating Sox10 transcription other than the MCS sequences, we generated luciferase constructs driven by five independent Sox10 promoter elements within −7 kb of the transcription start site. Interestingly, the −3484/−2495 fragment of the promoter contains a repressive element which decreased luciferase activity by approximately 75% in both SLK^fl/fl and SLK^-/- NDL cell lines and may account for the lower basal expression of Sox10 in cultured mammary epithelial tumor cell lines (Fig. 1d). Differential regulation of luciferase activity was observed in the two most distal promoter fragments between SLK^fl/fl and SLK^-/- NDL cells (Fig. 1d). The −6904/−5995 fragment showed a three-fold increase in luciferase activity in the SLK knockout cells, whereas that same element was inactive in the control cell line (Fig. 1d), suggesting that it harbors a differentially regulated enhancer region within the Sox10 promoter.

Interestingly, sequence analysis of the −6904/−5995 enhancer element revealed three consensus SoxE (Sox8, 9, and 10) binding sites [27] within this 909 base pair (bp) region (see Fig. 1c). However, Q-PCR analysis revealed no detectable expression of Sox8 and no differences in Sox9 expression in either cell lines grown in culture (Fig. 2a). As Sox10 is highly expressed in the SLK-deficient cells, we investigated the possibility that Sox10 could control its own expression by assessing the levels of endogenous Sox10 in stable pBABE-Sox10 overexpressing SLK^fl/fl NDL cells. Using primers located in the 3′UTR, we found that exogenous Sox10 was unable to induce endogenous gene expression (Fig. 2b).

One possibility is that the locus is in an inactive topology or, alternatively, the Sox10 target sites are bound by another SoxE transcription factor. Interestingly, the Sox9 transcription factor has been shown to induce Sox10 gene expression through SoxE binding sites [16, 28‐32], As Sox8 is not expressed in NDL cells, we performed chromatin immunoprecipitation (ChIP) for Sox9 at the consensus SoxE binding sites within the active −6904/−5995 element in SLK-deficient NDL cells (Fig. 2c). Although we observed a significant enrichment for Sox9 above IgG pull down in both control and SLK knockout cells, no significant differences between the two cell lines were observed (Fig. 2c). Similar results were observed when using an anti-K27 acetylated histone H3 antibody, suggesting an open chromatin conformation around those enhancers (Suppl. Figure 2A). As a negative control, we also performed anti-Sox9 ChIP on a putative SoxE binding site within exon 1 that was unresponsive in our luciferase assay (Fig. 1d). This element was not bound by Sox9 in either cell line (Fig. 2c), suggesting that the enrichment observed within the −6904/−5995 element is not due to non-specific chromatin pull down by the Sox9 antibody. Importantly, these data show that the SoxE sites within the −6904/−5995 Sox10 genomic region are bound by Sox9, suggesting that it plays a role in Sox10 gene expression.

As Sox9 is bound to the Sox10 promoter in both wildtype and SLK knockout cell lines, we reasoned that the activity of Sox9 may be enhanced in the absence of SLK, through post-translational modifications, without affecting its DNA binding capacity. Interestingly, Sox9 has been shown to be phosphorylated at S64, S181, and S211 to regulate its nuclear import, stability, and DNA-binding/transcriptional activity [33‐36]. Using phospho-serine 181 (pSox9 S181) as an indicator of Sox9 activity, we assessed the levels of pSox9 S181 in both cell lines. Western blot analysis showed a marked increase in the levels of Sox9 S181 phosphorylation in SLK knockout tumor cell lines that was correlated with elevated Sox10 levels in these cells (Fig. 3a), suggesting increased Sox9 activity in SLK-null tumor cells compared to the control. This was also correlated with higher levels of active AKT (pAKT S473; Fig. 3a). We have previously reported a two-fold increase in the number of Sox10+ nuclei from SLK-null mammary hyperplasia at 16 weeks of age [23]. Supporting this, immunohistochemical analysis for pSox9 S181 on hyperplastic lesions from SLK expressing and knockout hyperplastic lesions shows a similar increase (18 ± 4% vs 35 ± 7%) in the number of pSox9 S181-positive nuclei following SLK deletion in vivo (Fig. 3b). We did not observe any differences in the proportion of Sox10+ or pSox9 S181+ nuclei at endpoint, suggesting that SLK deletion affects tumor initiation as previously described [23]. Our previous analyses showed that SLK deletion results in a basal-like phenotype in MMTV-Neu mice [23]. Furthermore, as high as 13% of HER2+ patients fall within the SLK-low/SOX10-high subtype [23] and that Sox10 is a biomarker of the TNBC subtype [20]. Corroborating recent findings [37], analysis of those TCGA datasets reveals a strong correlation between high levels of Sox9 and Sox10 in TNBC samples (Fig. 3c, blue dots). To establish a potential link with active Sox9 in Sox10+ human samples, we assessed the levels of pSox9 S181 and Sox10 by immunohistochemistry across HER2-positive, luminal, and TNBC tumor cores. Consistent with our TCGA analysis (also see [23, 37]) and murine data, we observed that 43% of TNBC cores were Sox10^hi and that all Sox10^hi nuclei in those cores also showed a pSox9 S181^hi signal (Fig. 3d), suggesting that the activation of Sox9 and Sox10 induction is also observed in human breast cancer samples. Together, these studies suggest that the loss of SLK results in the activation of Sox9 that can directly induce Sox10 expression in murine tumors as well as human breast cancers.

We have previously shown that Slk deletion results in enhanced mammary tumorigenesis with the activation of PDK1 and AKT [23]. Therefore, we tested the possibility that Sox9 could be activated downstream of that pathway. Interestingly, amino acid sequence alignment and analysis of Sox9 revealed a putative AKT consensus phosphorylation site at serine 181, resembling several bona fide AKT targets (Fig. 4a). To assess whether Sox9 was a direct target of AKT, we performed in vitro kinase assays using recombinant proteins and monitored direct phosphorylation. As the full-length GST-Sox9 fusion protein is unstable and readily broken down in bacteria (not shown), we tested whether Sox9 was directly phosphorylated at serine 181 by AKT using a GST-Sox9^1-223 truncation. Recombinant AKT was able to efficiently phosphorylate the GST-Sox9^1-223 fusion protein, which contained the serine 181 residue (Fig. 4b, c). Mutation of serine 181 to alanine (S181A) in the GST-Sox9^1-223 truncation completely abolished AKT-dependent phosphorylation of Sox9 (Fig. 4c), suggesting that Sox9 is a novel substrate for AKT at the serine 181 consensus site.

Supporting our in vitro kinase assays, treatment of SLK knockout cells with the AKT inhibitor MK-2206 significantly reduced the expression of Sox10 and the phosphorylation of AKT and Sox9 (Fig. 4d). Furthermore, MK-2206 treatment was sufficient to abolish luciferase activity driven from the −6904/−5995 enhancer element of the Sox10 promoter in SLK knockout cells (Fig. 4e). Knockdown of AKT1, 2 or 3 showed that a ~50% knockdown of AKT1 reduced Sox10 levels by about 50%. However, knockdown of AKT2 by 50–70% reduced Sox10 levels to about 20% of controls. Knockdown of AKT3 did not show a marked downregulation of Sox10. Together, these data suggest that AKT2 preferentially regulates Sox10 expression and that direct AKT-mediated activation of Sox9 regulates Sox10 transcription from the −6904/−5995 enhancer element.

To assess the effect of Sox9 phosphorylation on Sox10 transcriptional regulation, we first tested whether phosphorylation of Sox9 was required for direct binding to the Sox10 promoter. SLK^-/- NDL cells were transfected with an empty vector, myc-Sox9 or myc-Sox9 S181A and subjected to anti-myc ChIP on the Sox10 promoter (Fig. 5a). No differences were observed in DNA binding activity for the Sox9 S181A mutant (Fig. 5a), suggesting that S181 phosphorylation is not required for enhancer binding. Next, we investigated whether Sox9 S181A expression could impair luciferase expression driven from the Sox10 promoter. Transfection of myc-Sox9 into SLK^-/- NDL cells with the Sox10 −6904/−5995 enhancer element resulted in a 2.5-fold increase in luciferase activity above a Myc control vector and myc-Sox9^1-223, lacking the C-terminal transactivation domain (Fig. 5b). Albeit significantly lower than wildtype Sox9, overexpression of the Sox9 S181A mutant also resulted in an increase in luciferase activity (Fig. 5b). As Sox9 can homodimerize when bound to DNA [38], one possibility is that the Sox9 S181A mutant can form a complex with endogenous Sox9, resulting in reduced transcriptional activity imparted by the mutant Sox9.

Since Sox9 is present on the Sox10 promoter of both cell lines, we hypothesized that the activation of AKT results in DNA-bound Sox9 phosphorylation and increased transcriptional activity. To test this, we transiently overexpressed myc, myc-Sox9, or myc-Sox9 S181A in SLK^-/- NDL cells and performed luciferase assays using the −6904/−5995 fragment of the Sox10 promoter. Consistent with our previous data, we observed a two- to three-fold increase in luciferase activity from the Sox10 promoter fragment with the Myc vector control (Fig. 5c) lanes 1 and 2). Additionally, we observed an AKT-dependent increase in luciferase activity following heregulin-stimulation in these cells (Fig. 5c, lanes 3 and 4). These controls further support the notion that Sox10 induction from the −6904/−5995 promoter element is dependent on Neu signaling through AKT. To validate that these signals are dependent on Sox9 transcriptional activity, we assessed luciferase activity in the presence of myc-Sox9 or a myc-Sox9 S181A mutant with reduced transcriptional activity [34]. As observed in Fig. 5b, in unstimulated cells, both wildtype and mutant Sox9 were sufficient to increase luciferase activity four-fold above background (Fig. 5c, lanes 5 and 6 compared to lanes 9 and 10). However, following stimulation with heregulin, a two-fold increase in luciferase activity was observed in myc-Sox9 expressing cells, whereas no significant change was observed in cells expressing the myc-Sox9 S181A mutant (Fig. 5c, lanes 6 and 7 compared to lanes 10 and 11). As for myc control-transfected cells, the increased luciferase activity following heregulin-stimulation in myc-Sox9 transfected cells was completely dependent on AKT activity (Fig. 5c, lanes 7 and 8). Together, these data show that the −6904/−5995 region of the Sox10 promoter is an AKT-responsive element which requires the transcriptional activity of Sox9. Furthermore, our results suggest that Sox9 S181 phosphorylation does not affect DNA binding but is required for maximal transcriptional activity on this Sox10 enhancer.