Stat3 is a member of the signal transducer and activator of transcription (Stat) family of transcription factors and exerts different and sometimes contrasting functions in normal and transformed cells (1–3). In the breast, Stat3 regulates mammary gland development, is activated during the proliferative phase of pregnancy, and plays roles during lactation and involution (3, 4). Stat3 activation is a transient and tightly regulated process in normal tissues, but occurs constitutively in many human tumors, including breast cancers, most notably in triple-negative breast cancer (TNBC). Stat3 signaling contributes to tumor cell survival, proliferation, migration, invasion, and chemoresistance, as well as influencing angiogenesis, immunosuppression, and cancer stem cell (CSC) self-renewal/differentiation (1, 5, 6). Whilst Stat3 regulates many genes and processes, its regulation of normal and cancer stem cells has been linked particularly with direct transcriptional activation of ΔNp63, itself a regulator of normal and CSCs and associated specifically with basal-like TNBC (1, 7–10).

Despite the observations for oncogenic roles, immunohistochemical studies of Stat3 give conflicting results for prognosis (11–15). Importantly, there is increasing preclinical and clinical evidence that Stat3 inhibitors reduce breast cancer growth and metastasis (6,16), indicating the potential clinical value of measuring Stat3 activity. Stat3 activity is controlled by phosphorylation of tyrosine 705 (pTyr705-Stat3), which leads to dimerization, translocation to the nucleus, and transcriptional activation of target genes (17). Stat3 is also phosphorylated on serine 727 (pSer727-Stat3), and phosphorylation at both sites provides maximal activity (18, 19). Thus, it is generally thought that the two phosphorylation events act in concert. However, Ser727 phosphorylation can reduce Tyr705-Stat3 phosphorylation (20) and Stat3 can be activated through Ser727 phosphorylation in the absence of pTyr705 (21–23). Moreover, pSer727-Stat3 has been reported to localize to mitochondria and the endoplasmic reticulum to mediate their functions, including mitochondrial apoptosis pathways (24). Stat3 phosphorylation also has roles in defining cell identity and differentiation, where pTyr705-Stat3 induces an epithelial phenotype and pSer727-Stat3 a mesenchymal phenotype in pancreatic and lung cancers (25), and pTyr705-Stat3 maintains pluripotency whilst pSer727-Stat3 induces differentiation in embryonic stem cells (26).

These findings suggest that Tyr705- and Ser727-Stat3 phosphorylation may be independent regulators of cancer cell phenotypes. Thus, we studied pTyr705- and pSer727-Stat3 in a set of well-characterized TNBCs and correlated the relative levels of each Stat3 phosphorylation with each other and with clinicopathological and phenotypic features. We show that pTyr705-Stat3 and pSer727-Stat3 are often independent of each other and influence the basal/luminal phenotype of TNBC cells in opposite directions. We also show that high levels of pStat3 are associated with good prognostic indicators but not with improved survival suggesting it may be a useful target for therapeutic intervention in these patients.

Immunostaining data were obtained for pTyr705- and/or pSer727-Stat3 in 173 TNBCs. Data were available for both in 151 tumors. Uninterpretable spots were due to loss of tissue on the array or insufficient numbers of tumor cells.

pTyr705-Stat3 localized to the nuclei of tumor cells and stromal cells including lymphocytes and fibroblasts, which served as convenient internal controls for staining. pTyr705-Stat3 was not seen in the cytoplasm. In contrast, pSer727-Stat3 was observed in the nucleus and the cytoplasm of tumor cells and stromal cells, compatible with previous data on the location of these two phosphorylated forms of Stat3 (25). To investigate the effects of phosphorylation events specifically on transcriptional activity within tumor cells, only nuclear labeling of tumor cells was used for quantification of the selected areas in QuPath, resulting in an H-score for each phosphorylation event for each tumor. H-scores for pTyr705-Stat3 ranged from 1.59 to 213.20 (mean 43.59 ± 36.02 SD) and from 8.10 to 242.75 for pSer727-Stat3 (mean 120.50 ± 53.20 SD). For comparison with clinicopathological variables, nuclear H-scores were used to divide tumors into groups above and below the mean value for each phosphorylation.

Figure 1 displays examples of pTyr705- and pSer727-Stat3 immunohistochemical staining and corresponding examples of QuPath selections. We initially examined whether the two phosphorylation events were concordant or were independent of each other. A common trend of staining was seen in 100 cases (66%), in which low pTyr705- and low pSer727-Stat3 were present in 60 tumors, and high pTyr705- and high pSer727-Stat3 in 40 samples. An opposite trend was observed in 51 patients (34%), with low pTyr705- and high pSer727-Stat3 in 32, and high pTyr705- and low pSer727-Stat3 staining intensity in 19 tumors (Figure 2A). Pearson correlation coefficient showed a positive correlation between pTyr705- and pSer727-Stat3 normalized histoscores within individual tumors (r = 0.509; Figure 2B). Whilst the correlation is statistically significant, it is also evident that many individual tumors (over one-third of the cohort) showed discordant values for phosphorylation at the two sites.

The role of Stat3 in cancer is controversial, with several studies demonstrating an oncogenic role in the development of various cancers, whereas others have indicated that Stat3 behaves as a tumor suppressor (1, 25). These contrasting data imply dependency on tumor type, stage of the disease, and local microenvironment; Stat3 activation or inactivation can each support disease progression depending on such factors (33, 34). In breast cancer, Stat3 is more often overactive in TNBCs than other types (5) and has been implicated in inducing CSC properties (35).

Stat3 activation involves phosphorylation to allow dimerization, nuclear transport and binding to response elements in the promoters of target genes. In general, it is thought that Tyr705 phosphorylation is the initial event for Stat3 activation, followed by Ser727 phosphorylation to provide higher activity. Thus, most immunochemical studies evaluate total Stat3, or pTyr705-Stat3 as a general indicator of activity and only exceptionally pSer727-Stat3. However, data suggest that these two events have distinct functional effects during embryonic stem cell differentiation and in dictating epithelial cell identity in pancreatic and lung cancers (25, 26).

In our study, we found that pTyr705- and pSer727-Stat3 are related events that are often associated with each other. On the other hand, individual tumors show discrepancies, where one phosphorylation event occurs in the absence of the other, suggesting that different signaling events are operating in these tumors. Most importantly, each phosphorylation is associated with a distinct tumor cell phenotype, indicating that pTyr705- and pSer727-Stat3 regulate distinct pathways within the tumor. In particular, high levels of pTyr705- but not pSer727-Stat3 were associated with ERβ alone and with ERβ and AR combined. Our results further showed that patients with high levels of pTyr705-Stat3 were more likely to be MUC1-positive. High-level AR is the defining feature of luminal-AR TNBCs (36) and Burstein et al. (37) further characterized this phenotype by showing these tumors also over-express MUC1. Thus, pTyr705-Stat3 associates with an epithelial cell phenotype in TNBC, analogous to its specific effects in lung and pancreatic cancers (25). It is also important to note that transmembrane C-terminal MUC1 (MUC1-c) is necessary for Tyr705-Stat3 phosphorylation in breast cancer cells and promotes Stat3-mediated transcription in an auto-inductive regulatory loop (38). Thus, whether pTyr705-Stat3 is a cause or effect of this specific phenotype will require further study. We also found an association between high levels of pTyr705-Stat3 and BRCA2 mutation, which is more common in ER-positive luminal B tumors (39). Moreover, consistent with our findings, pTyr705-Stat3 has been reported recently to disrupt CHK1 phosphorylation, thereby impairing BRCA2-mediated RAD51 recruitment for homologous recombination repair of double-strand DNA breaks (40).

In contrast to the association of pTyr705-Stat3 with an AR⁺/ERβ⁺/MUC1⁺ epithelial phenotype, pSer727-Stat3 is associated with a basal cell phenotype, again analogous to the mesenchymal phenotype induced by pSer727-Stat3 in other cancers (25). In particular, tumors with high levels of pSer727-Stat3 were more often CK5/6-positive, the most useful marker for the basal subgroup of TNBC (41). On the other hand, basal cancers, especially the BL2 subtype, are characterized by high levels of EGFR (42), whereas high levels of pSer727-Stat3 were associated with low EGFR in our samples. These findings are in agreement with the increasing recognition of intra-tumor heterogeneity that can be caused by a variety of cell-intrinsic and extrinsic traits that, together with inter-tumor heterogeneity, significantly affect a patient’s prognosis, therapy response and clinical outcome (43–45). In this respect, the use of a tissue core in a tissue microarray may not provide an overall picture of the entire tumor. In our study, this potential problem of intra-tumor heterogeneity was mitigated by using two independent cores taken from non-adjacent regions of the tumor for all patients, and this approach has been shown to accurately represent the overall tumor phenotype (46, 47).

Stat3 is an important regulator of CSC phenotypes (1, 48). It is becoming evident that various CSC subtypes exist in breast and other cancers, have different effects on prognosis, and are another source of tumor heterogeneity (43, 49–51). For example, ΔNp63 marks a specific subtype of basal/mesenchymal CSCs in breast cancer, as opposed to the luminal/epithelial CSC subtype in these tumors (27, 52, 53) and there is a similar population of ΔNp63-positive basal/mesenchymal CSCs in prostate adenocarcinomas (54). ΔNp63 and Stat3 are involved in several common pathways regulating CSC properties (1, 7), Stat3 is a direct regulator of ΔNp63 transcription (9, 10) and they cooperate in regulating epithelial cell identity in KRas-driven lung and pancreatic cancers (25), whilst ΔNp63 overexpression itself alters pTyr705- and Ser727-Stat3 levels in TNBC cells (55). However, our results show that pStat3 is not associated with ΔNp63 (or TAp63) in TNBC. Thus, although either pSer727-Stat3 or ΔNp63 may associate with basal cell phenotypes in TNBC they are independent of each other, suggesting at least two transcriptional activation routes to this phenotype and providing further evidence for heterogeneity of CSC phenotypes in breast cancer.

High pTyr705-Stat3 is associated with less aggressive tumor characteristics such as smaller tumor size, lower clinical stage, and absence of lymph node metastasis, similar to some previous studies (11, 56–58), but different from others (59). Levels of pSer727-Stat3 were associated with tumor size, and, together with high levels of pTyr705-Stat3 also with clinical stage. This is inconsistent with data from Yeh et al. (30), where high pSer727-Stat3 was correlated with larger tumors and higher clinical stages. Contributors to these differences include method sensitivity and the scoring systems and cut-offs used (58). In our study, we employed an objective rather than subjective approach to delineate pStat3 levels based on the mean values obtained for each antibody.

Our data show that although pStat3 phosphorylated at either site is associated with less aggressive tumor characteristics that are independently associated with better survival, such as tumor size or clinical stage, pStat3 itself is not associated with survival. Thus, these data imply that high pStat3 is an unfavorable indicator in tumors with a good prognosis according to their clinicopathological factors. This observation may relate to the ability of Stat3 to induce drug resistance (60, 61). Furthermore, recent data indicate that cancer cells dynamically exploit Stat3 activity, whereby both Stat3 activation and inactivation support cancer progression in a time- and space-dependent manner (34).

In conclusion, we show that pTyr705- and pSer727-Stat3 are associated with more favorable clinicopathological features but not with better survival, inferring its negative role in TNBC prognosis. Our results further indicate that Stat3 phosphorylation at Tyr705 and Ser727 have different effects on the phenotype of TNBCs, analogous to their distinct roles in embryonic stem cell self-renewal/lineage commitment (26) and cellular phenotype in pancreatic and lung cancers (25). In turn, these data indicate complex regulation and roles for Stat3 in TNBC, helping to explain the discordant results of studies that report the impacts of total Stat3 levels or single phosphorylation site analysis. The site-specificity of Stat3 phosphorylation should be taken into account in clinical trials that aim to disrupt Stat3 signaling in breast and other cancers (6, 48).