Background
Approximately, 12% of all new cancers yearly are reported as breast carcinomas. However, extensive heterogeneity and the complex biology and etiology for the various breast cancer subtypes make treatment difficult. Breast cancer subtypes are generally classified based on the expression of the estrogen receptor (ESR1), progesterone receptor (PGR), or the human epidermal growth factor receptor 2 (HER2) [
1,
2]. The HER2-positive subtype, characterized by overexpression and amplification of HER2, accounts for 20–30% of all breast cancers and is associated with poor prognosis and aggressive cancers [
3,
4]. The activation of the HER2 receptor through tyrosine phosphorylation [
5,
6] results in the upregulation of proliferative and survival pathways [
7,
8]. Although HER2 amplification is a critical event in the etiology of HER2+ breast cancers, the molecular mechanisms regulating its progression are not fully understood.
The Ste20-like kinase (SLK) has been shown to regulate multiple biological responses [
9]. In addition to cell migration [
10,
11], SLK has been observed to play an important role in the breakdown of E-cadherin and ZO-1-positive junctions following TGFβ stimulation [
12]. The loss of SLK delayed EMT, suggesting that it regulates the cytoskeletal changes associated with this process [
12,
13]. We have previously shown that SLK is activated downstream of Neu (rat homolog of HER2) and requires PI3K or PLCγ activity for maximal activation [
14]. Supporting a role for SLK in HER2-driven signaling, expression of a dominant negative SLK K63R reduced HER2-dependent chemotaxis in human breast cancer cell lines.
The Akt/PKB literature is constantly growing since the cloning of v-Akt in 1987 (extensively reviewed in [
15]). It is now well established that Akt activation occurs downstream of phosphoinositide-3-kinase (PI3K), a lipid kinase implicated in tumorigenesis and the insulin response. The PI3K-dependent generation of PtdIns-3,4-P3 (PIP3) recruits and activates Akt at the membrane in concert with PDK1 and mTORC2. There are well over 100 reported Akt substrates implicated cell survival, proliferation, metabolism, neuronal functions, and angiogenesis. To date, the only transcription factors reported to be Akt targets are the Forkhead Box O proteins (FoxO1, 3, 4, and 6), regulating multiple transcription programs. Their phosphorylation induces their cytosolic retention through interactions with 14-3-3 proteins, blocking their access to target genes. The importance of turning off the PI3K-Akt pathway is underscored by the sheer number of negative feedback and cross-talk pathways. The most critical signal terminator is the phosphatase tumor suppressor PTEN, capable of converting the activator PIP3 to PIP2. Obviously, perturbations in AKT signaling lead to numerous pathological conditions such as overgrowth syndromes, autoimmune diseases, and cancer.
The Sry-HMG-box (Sox) family of transcription (reviewed in [
16‐
18]) plays critical roles in many developmental processes. The SoxE group proteins, including Sox8, 9, and 10, have been extensively studied in the context of reproductive system development, neural crest cell-derived tissues, and cell types such as melanocytes. Although Sox9 and Sox10 have been widely studied during development, little is known about the mechanisms that regulate their activities other than their nucleocytoplasmic shuttling. The role of the Sox proteins in cancer progression remains elusive and controversial (reviewed in [
18]). Different members have been shown to play a role both as tumor promoters and tumor suppressors in various types of cancers through the regulation of oncogenic pathways. Supporting this, Sox2 and Sox9 have been shown to be critical for the persistence of quiescent stem-like cancer cells through immune evasion [
19]. We and others have recently identified Sox10 as a marker of triple-negative breast cancers (TNBC) [
20] and secretory carcinomas, often triple negative and basal-like. Recently, Sox10 was found to be specifically expressed in mammary progenitor cells, including fetal and adult mammary cells in vivo [
21] and be critical to maintain the stem cell state and reprogramming in breast cancer [
22]. Strikingly, Sox10 deletion impairs mammary gland reconstitution whereas its overexpression increases it [
21].
We have recently demonstrated that conditional SLK deletion in a MMTV-Neu background activates the PDK1-Akt system and accelerates breast tumor onset [
23]. Although they are HER2/Neu+, those tumors display a basal-like phenotype. Strikingly, early lesions and tumor-derived cell lines express high levels of Sox10, a marker of TNBC [
20]. This is accompanied by increased tumor stem cell activity in vitro and enhanced tumor growth in xenograft models. Interestingly, this phenotype is dependent on AKT activity and Sox10 expression. To gain insights into the molecular mechanisms regulating Sox10 expression, we have further assessed the role of the PI3K-PDK1-AKT pathway on Sox10 regulation. Our data show that AKT-driven Sox10 expression is dependent on Sox9 activation. Furthermore, this activation requires direct Sox9 phosphorylation by AKT. Our studies have uncovered Sox9 as a novel substrate for AKT, providing a novel link between AKT and tumor progression.
Discussion
Here, we have shown that the observed induction of Sox10 in SLK-/- NDL tumor cells is dependent on a novel enhancer located at about −7kb upstream of the putative Sox10 start site. Our data show that Sox10 expression is also directly correlated with Sox9 phosphorylation and activity in our SLK-/- NDL mouse model and human tissue microarrays. We demonstrate that AKT can directly phosphorylate DNA-bound Sox9 at serine 181, increasing its transcriptional activity on the −7kb Sox10 enhancer. These data also extends the growing list of AKT substrates.
Our data demonstrate that an AKT-dependent pathway regulates Sox10 transcription specifically through an enhancer fragment located between −6904 and −5995 from the putative start site. Scanning of this enhancer revealed three SoxE sites. Although we have observed some variability, no significant differences were found in Sox9 binding to those SoxE sites, suggesting that Sox9 is bound equally to these elements in both SLKfl/fl and SLK-/- NDL cells. This suggests that Sox9 phosphorylation at S181 activates transcription without altering its DNA binding activity. Given the large increase in Sox10 expression in the SLK-/- cells, it is likely that, in addition to Sox9 binding to SoxE sites, other transcriptional mechanisms are activated. One possibility is that phosphorylation at this site is required to recruit transcriptional cofactors which may be required for maximal activity.
Using amino acid alignment and kinase assays, we have shown that AKT can directly phosphorylate Sox9 in vitro and that enhancer activity is lost following AKT inhibition (Fig.
4). This suggests that the inability for AKT to phosphorylate Sox9 prevents its activation and Sox10 induction. Although we predict that AKT directly phosphorylates Sox9 at serine 181, this remains to be demonstrated in vivo. However, this may be more challenging as a number of other kinases have been shown to phosphorylate that site in other contexts [
33‐
36]. Alternatively, it is possible that AKT directly phosphorylates and inhibits the activity of an unknown phosphatase that targets pSox9 S181. It is also possible that this phosphorylation may be mediated by another kinase in the AKT pathway such as S6 kinase (S6K). S6 Kinase is activated by mTOR downstream of AKT and has a consensus phosphorylation motif matching that of Sox9 S181 [
39]. Combined with in vitro kinase assays, rapamycin treatment to inhibit mTOR would address the potential role of S6K in Sox9 activation and
Sox10 expression.
To validate the role of Sox9 in regulating the transcription of
Sox10, we performed luciferase assays using the AKT-responsive element from the Sox10 promoter. Surprisingly, myc-Sox9 or the myc-Sox9 S181A mutant is both capable of activating luciferase expression from the −7kb enhancer under basal conditions (Fig.
5). This was also observed for the collagen II promoter, where the induction by both Sox9 constructs is relatively high above background [
34]. In the context of the collagen II promoter, wildtype Sox9 was sufficient to boost luciferase activity following co-transfection with the catalytic subunit of PKA, whereas the Sox9 S181A mutant failed to do so [
34]. Similarly, treatment of the myc-Sox9 transfected cells with heregulin increased luciferase activity two-fold above the untreated sample (Fig.
5c, lanes 6 and 7). However, expression of the phospho-deficient mutant abrogated this effect (Fig.
5c, lanes 10 and 11). One possibility is that Sox9 activity through phosphorylation at S181 and/or S211 can be further increased by specific signals while bound to DNA as a homodimer. This would explain the slight reduction in enhancer activity when a myc-Sox9 S181A is expressed in SLK
-/-NDL cells. The S181A mutant may have a dominant negative effect on a Sox9/Sox9 S181A dimer.
We have previously shown that Sox10 is biomarker for the TNBC subtype and is induced following the mammary gland-specific deletion of SLK in a MMTV-Neu mouse model, inducing a basal-like phenotype in a HER2+ model [
23]. The induction of Sox10 is accompanied by increased stemness and accelerated tumorigenesis [
23]. Recently, Sox10 has been shown to be expressed in mammary progenitor cells in vivo [
21] and be critical to maintain stemness in breast cancer [
22]. Interestingly, Sox2 and Sox9 have been shown to be critical for the persistence of quiescent stem-like breast cancer cells [
19]. It is not clear whether those Sox9+ stem-like cells also express Sox10.
Supporting our findings in murine mammary tumors, we have also observed a correlation between pSox9 S181 levels and Sox10 expression in a proportion of HER2+ human tumor samples [
23]. In addition, Fig.
3c shows that a high proportion of Sox10
hi TNBC samples also show a Sox9
hi phenotype (blue dots in the SOX9/SOX10 high quadrant of the data set), suggesting that high Sox9 activity could occur without increased Akt activity. Similarly, Sox9 is highly phosphorylated in human triple-negative breast cancers expressing high levels of Sox10 (see Fig.
3 and [
37]). Therefore, it is possible that a similar mechanism exists in breast cancers that display high level of PI3K/AKT pathway activation. In fact, oncogenic activation of the PI3K/AKT signaling pathway is frequent in TNBC and most commonly occurs following
PIK3CA gain-of-function mutations or
P53 inactivation [
40,
41]. Treatment of triple-negative breast cancer cell lines with the allosteric AKT inhibitor, MK-2206, inhibits tumor growth and increases sensitivity to other chemotherapeutic agents [
42‐
44]. Clinical trials for TNBC have shown that AKT inhibitors, including MK-2206, have a synergistic effect with paclitaxel and significantly improve progression-free and overall survival [
42,
43]. In addition to regulating cell survival and promoting tumor growth, we have shown that AKT controls the expression of
Sox10. Therefore, we believe that the therapeutic targeting of AKT could decrease mammary tumor stem/progenitor activity by downregulating
SOX10 expression in TNBC.
A recent murine model for TNBC has identified a high frequency of both
Egfr and
Fgfr2 amplifications [
21,
30]. FGF-signaling has previously been shown to induce the expression of both
Sox9 and
Sox10 [
21,
30]. In light of our observations,
Fgfr2 amplifications in TNBC [
41] may be sufficient to upregulate
Sox9 expression. Combined with gain-of-function mutations in
PIK3CA, this would be sufficient to increase Sox9 activity in an AKT-dependent manner. Therefore, the combinatorial treatment of triple-negative breast cancers with FGFR and AKT inhibitors may target two distinct signaling pathways that drive
SOX10 expression by blocking both the induction and activation of Sox9.
One of the largest barriers to the effective treatment of HER2-positive breast cancers is the rapid acquisition of Herceptin-resistance [
45,
46] often accompanied by significantly elevated levels of phosphorylated active AKT [
47]. As both chronic Herceptin treatment and SLK deletion result in the hyperactivation of AKT, we speculate that these Herceptin-resistant tumors may have also acquired
SOX10 expression. Therefore, it is possible that Herceptin-resistant tumors with high levels of AKT activity may be dependent on the oncogenic and stem/progenitor activities of Sox10.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.