Immunosurveillance constitutes the first step of cancer immunoediting in which developing malignant lesions are eliminated by antitumorigenic immune cells. However, the mechanisms by which neoplastic cells induce an immunosuppressive state to evade the immune response are still unclear. The transcription factor STAT3 has been implicated in breast carcinogenesis and tumor immunosuppression in advanced disease, but its involvement in early disease development has not been established. Here, we genetically ablated Stat3 in the tumor epithelia of the inducible PyVmT mammary tumor model and found that Stat3-deficient mice recapitulated the three phases of immunoediting: elimination, equilibrium, and escape. Pathologic analyses revealed that Stat3-deficient mice initially formed hyperplastic and early adenoma-like lesions that later completely regressed, thereby preventing the emergence of mammary tumors in the majority of animals. Furthermore, tumor regression was correlated with massive immune infiltration into the Stat3-deficient lesions, leading to their elimination. In a minority of animals, focal, nonmetastatic Stat3-deficient mammary tumors escaped immune surveillance after a long latency or equilibrium period. Taken together, our findings suggest that tumor epithelial expression of Stat3 plays a critical role in promoting an immunosuppressive tumor microenvironment during breast tumor initiation and progression, and prompt further investigation of Stat3-inhibitory strategies that may reactivate the immunosurveillance program. Cancer Res; 76(6); 1416–28. ©2015 AACR.
Immunoediting refers to a three step process of interactions between the immune system and emerging cancer comprised of elimination, equilibrium, and escape (1, 2). During the elimination phase, antitumorigenic immune cells are involved in the eradication of cancerous lesions. This may be followed by a long period of equilibrium wherein cancerous cells are kept in check by the immune system. Eventually, these cancerous lesions escape through various mechanisms, which, though not well understood, can include cell-intrinsic alterations or the suppression or inactivation of the immune system (1, 2). Modulation of the tumor immune microenvironment is an important factor in tumor progression and disease outcome. Both a high CD4:CD8 T-cell ratio and/or elevated presence of intratumoral macrophages are associated with poor outcome and metastatic disease in human breast cancer (3–5) while higher Th1/Th2 ratios are excellent predictors of positive patient outcome (6). In addition, M2-macrophage polarization and metastasis has been shown to be induced by a Th2 response in murine models of breast cancer (7). However most of these studies focus on the immune microenvironment in invasive disease, leaving the mechanisms that promote antitumor immunity during the ductal carcinoma in situ (DCIS) to invasive disease transition poorly understood.
The transcription factor STAT3 has been observed to be constitutively active in 35% to 60% of human breast cancers (8, 9) as well as many other types of cancer (10, 11). In breast cancer, its expression and activation correlates with tumor grade, stage, presence of metastases and a higher risk of recurrence (8, 9). In addition, loss of Stat3 function has been shown to decrease tumor cell growth and angiogenesis, increase apoptosis in vitro (10) and decrease lung metastasis in a transgenic mouse model (12). Furthermore, Stat3 activation in the immune microenvironment is linked to the tumor-promoting effects of myeloid-derived suppressor cells (MDSC) and M2 tumor-associated macrophages (10, 13, 14). Stat3 activation has also been inversely correlated with immune cell infiltration in vivo (15–17) and its inactivation elicits antitumor immune responses in melanoma, lung cancer, and glioblastomas (15, 16). However, these studies mainly utilize xenograft models to examine Stat3-mediated immune suppression in advanced disease, leaving its interactions at tumor initiation largely unexplored.
In this study, we investigated the role of Stat3 in mammary tumor progression by crossing conditional Stat3 (Stat3flx) mice (18) with a mouse model of mammary tumor progression that expresses the Polyomavirus middle T (PyVmT) and Cre recombinase (Cre) in a doxycycline-inducible fashion (MMTV-MTB/TetO-MIC; ref. 19). This novel model allows for a temporal analysis of tumor initiation and progression. Mammary epithelial-specific disruption of Stat3 resulted in a profound delay in mammary tumor onset and penetrance. In addition, the Stat3-deficient tumors that eventually emerged demonstrated a major metastatic defect. Despite this delay in tumor onset, mammary epithelial ablation of Stat3 did not prevent the initiation of early hyperplastic lesions in these mice. However these lesions were rapidly cleared by a robust immune response, driven by myeloid and T-cell populations. Furthermore, the Stat3-deficient tumors lacked a transcriptional program involved in proinflammatory recruitment of myeloid cells and implicated in metastatic progression. Together, these observations recapitulate all three stages of the immunoediting program and indicate that Stat3 promotes an immunosuppressive tumor microenvironment involved in modulating early tumor outgrowth and later stages of metastasis.
Materials and Methods
Onset of mammary tumors was determined by biweekly physical palpation. Animals were sacrificed at predetermined time points postinduction of doxycycline treatment (2 mg/mL in drinking water) or at an ethical maximum tumor burden. Tumor, mammary gland, and lung material was frozen in liquid nitrogen, set in optimal cutting temperature (OCT) medium then frozen or fixed with either 10% neutral buffered formalin or a commercially available zinc fixative (BD Pharmingen: 550523) and embedded in paraffin. The fixed and embedded materials were sectioned at 4 μm and stained by hematoxylin and eosin (H&E) or processed as indicated below. Metastatic lesions were detected and scored by sectioning the lungs at 50 μm intervals and stained for H&E.
Immunoblotting, immunofluorescent, and immunohistochemical analyses
Frozen mammary tumors were lysed using a PLC-γ buffer and run on SDS-PAGE gels. Immunofluorescent and immunohistochemical analyses were performed on paraffin or OCT-embedded sections as described previously (12). Staining was quantified when indicated using slides scanned with a Scanscope XT Digital Slide Scanner (Aperio) and corresponding positive pixel and nuclear immunohistochemical algorithms.
All antibodies were purchased from BD Pharmingen, eBioscience, Invitrogen, BioLegend, the UCSF hybridoma core, or were produced in the Krummel's laboratory. For surface staining, cells were incubated with anti-Fc receptor antibody (clone 2.4G2) and stained with antibodies in PBS + 2% FCS for 30 minutes on ice. Viability was assessed by staining with fixable Live/Dead Zombie (BioLegend) or DAPI. All flow cytometry was performed on a BD Fortessa flow cytometer. Analysis of flow cytometry data was done using Flowjo (Treestar).
qRT-PCR was performed using a Roche LC480 SYBR Green RT-PCR Kit (Roche). Samples were run using a LightCycler (Roche) and each reaction was run in triplicate. The resulting crossing point values were normalized against GAPDH to generate the relative transcript levels using the formula: 2^(average GAPDH crossing point − average target × crossing point).
Microarray results have been deposited in the Gene Expression Omnibus database under the accession number GSE75325.
Additional information can be found in Supplementary Materials and Methods.
Stat3 loss results in a significant delay in mammary tumorigenesis
To evaluate the role of Stat3 in PyVmT-driven tumor progression, we interbred the conditional Stat3 strain (18) with mice carrying the MMTV-rtTA (MTB; ref. 20) and the inducible TetO-PyVmT-IRES-Cre recombinase (MIC; ref. 19) transgenes. This doxycycline-inducible PyVmT model couples the expression of the PyVmT oncogene and Cre, and subsequent deletion of the conditional Stat3 allele(s), specifically in the mammary epithelium upon induction with doxycycline (Fig. 1A). Cohorts of virgin female mice bearing the wild type, one conditional or both conditional Stat3 alleles (Stat3wt/wt/MTB/MIC, Stat3wt/flx/MTB/MIC, and Stat3flx/flx/MTB/MIC) were generated. Whereas the Stat3wt/wt/MTB/MIC and Stat3wt/flx/MTB/MIC mice developed mammary tumors with an average onset of 20 or 28 days, respectively, the Stat3-deficient MIC mice displayed a significant delay in tumor onset to an average of 275 days (Fig. 1B). Furthermore, the Stat3flx/flx/MTB/MIC mice demonstrated a decrease in tumor penetrance, as only 31% developed tumors compared with 80% and 73% of the Stat3wt/wt/MTB/MIC and Stat3wt/flx/MTB/MIC mice, respectively. We confirmed Stat3 ablation and oncogene expression by PCR (Supplementary Fig. S1A), immunoblot and IHC (Fig. 1C and D). Total levels of Stat3 and its activated form (Stat3-pY705) were significantly reduced in the Stat3-deficient tumors. In addition to a similar average onset, the tumors from the Stat3wt/flx/MTB/MIC mice expressed similar levels of Stat3 protein and activation to the parental MIC tumors (data not shown). Residual Stat3 protein observed in both the immunoblot and immunohistochemical staining (Fig. 1C and D) is a result of Stat3 retention in the stroma and tumor-infiltrating cells. The epithelial specificity of the Stat3 ablation was confirmed by immunofluorescence and immunohistochemical staining (Supplementary Fig. S1B and S1C). Histologic analysis of end-stage tumors from all genotypes revealed a similar solid adenocarcinoma phenotype. Thus, the mammary epithelial ablation of Stat3 delays tumor onset and decreases penetrance, suggesting a requirement for additional cooperating events.
Stat3 is dispensable for the initial outgrowth of PyVmT hyperplasias but is critical for their immune evasion
An advantage of the PyVmT tumor model is the ability to investigate tumor progression through distinct phases from early mammary hyperplasias and adenomas to end-stage adenocarcinomas (19). To evaluate these early stages, we performed histologic analyses on the transformed mammary glands after 2, 4, and 6 weeks of doxycycline induction. These analyses revealed that the Stat3flx/flx/MTB/MIC mice developed mammary epithelial hyperplasias and adenomas with 100% penetrance comparable with their Stat3-proficient counterparts, albeit involving a lower proportion of the mammary epithelial tree (Fig. 2A, top; Supplementary Fig. S2A–S2C). These Stat3-deficient lesions had begun to regress by the fourth week of induction (Fig. 2A, middle). Remarkably, by 6 weeks of induction, the Stat3-deficient mammary glands were completely devoid of hyperplasias or adenomas, whereas Stat3-proficient glands exhibited invasive adenocarcinomas (Fig. 2A, bottom). We confirmed mammary epithelial deletion of Stat3 by both immunohistochemical (Fig. 2B) and immunofluorescence analyses (Supplementary Fig. S2D).
Given that Stat3 has been implicated in the proliferative capacity of mammary tumors and cancer cell lines (10–12), we examined the proliferative or apoptotic profiles of these Stat3-deficient lesions by staining for Ki67 as a proliferative marker and cleaved caspase-3 and terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) as markers of apoptosis. Neither the proliferative nor apoptotic status of the Stat3-deficient lesions was altered between genotypes after 2 weeks of induction (Fig. 2C). Taken together, these findings suggest that Stat3 is not necessary for the initiation of these early lesions but is crucial for their progression to invasive carcinomas.
Another explanation for the regression of the Stat3-deficient lesions is clearance by an activated immune response. It should be noted that although 100% of MIC animals develop mammary epithelial hyperplasias, only 80% develop end-stage tumors. To investigate this we used both IHC and flow cytometry to identify the nature of the immune infiltrate to these lesions. Initially, we examined the extent of overall leukocyte infiltration as well as specific myeloid components. Immunohistochemical analyses revealed that Stat3-deficient lesions were preferentially enriched in both CD45+ leukocyte and F4/80+ macrophage populations that circumscribed the emerging Stat3-deficient lesions (CD45 - Stat3wt/wt 20.96% ± 2.80% vs. Stat3flx/flx 31.33% ± 2.49%, F4/80 - Stat3wt/wt 16.67% ± 1.77% vs. Stat3flx/flx 27.95% ± 1.59%; Fig. 3A). Given that CD45+ population is comprised of a plethora of distinct immune effector cells, we utilized flow cytometric analyses to distinguish between specific cell populations. We initially sorted the CD45+ cells on CD11b (myeloid-specific marker) and Ly6C (monocyte marker) expression. Ly6C-negative cells were gated on MHCII+ expression and further distinguished by F4/80hi, CD24lo expression (macrophages), and CD24hi, F4/80lo expressers (dendritic cells, DC). Macrophages were divided on the basis of differential expression of CD11c and CD11b and are here identified as TAM1 (CD11clo, CD11bhi) and TAM2 (CD11chi, CD11blo; Fig. 3B). These macrophage populations correspond to similarly delineated MHCIIlo (TAM1) and MHCIIhi (TAM2) populations, where CD11c high TAM2 macrophages display classically activated M1 gene signatures (21). DCs were subgated into two populations based on CD11b and CD103 expression as described previously (Fig. 3C; ref. 22). Consistent with the immunohistochemical data, we found that the CD45+ compartment was dramatically enriched in the Stat3-deficient mammary glands after 2 weeks of induction (Stat3wt/wt 49.43% ± 6.64% vs. Stat3flx/flx 71.90% ± 1.33%; Fig. 3D). Interestingly, we did not see such a dramatic difference in F4/80+ macrophages by FACS (Stat3wt/wt 10.65% ± 2.88% vs. Stat3flx/flx 11.89% ± 0.84%; Fig. 3B and D). However, there was a shift towards TAM2 macrophages in the Stat3-deficient glands. One marker of the protumorigenic M2 class of macrophages is CD206 (the mannose receptor; ref. 23). Consistent with the protumorigenic role of M2 macrophages, the Stat3-deficient lesions exhibited a marked decrease in CD206 expression in the TAM1 macrophage population compared with the macrophages found in their Stat3-proficient counterparts (Fig. 3B and E). DCs are another key myeloid population that modulates T-cell responses in cancer (22, 24, 25). Consistent with the antitumor function of the CD103 subclass of DCs, the Stat3-deficient lesions exhibited a significant increase in the ratio of CD103+ DCs to CD11b+ DCs (CD103:CD11b ratio of 2.10 ± 0.17) compared with Stat3-proficient lesions (CD103:CD11b ratio of 0.69 ± 0.11; Fig. 3D). Taken together, these observations argue that the loss of Stat3 in the tumor epithelia results in an increase of antitumor myeloid populations in these lesions.
T-cell populations work in concert with these myeloid cells to elicit an antitumor immune response. To identify the nature of the T-cell infiltrates, we again employed both IHC and flow cytometry, utilizing various T-cell–specific markers. Using CD3ϵ as a pan T-cell marker, we observed a dramatic increase in the number of CD3+ T cells to the Stat3-deficient lesions relative to wild-type counterparts (Stat3flx/flx 10.41% ± 1.88% vs. Stat3wt/wt 2.14% ± 0.37%; Fig. 4A). The observed increase in the CD3 population was comprised of an increase in both the CD4+ [regulatory T (Tregs) and effector T cells (Teff)] and CD8+ (CTLs) T-cell subsets in the Stat3-deficient lesions (CD4+ Stat3wt/wt 6.79% ± 0.80% vs. Stat3flx/flx 19.66% ± 1.59%, CD8+ Stat3wt/wt 1.34% ± 0.25% vs. Stat3flx/flx 6.28% ± 0.97%; Fig. 4A). Furthermore, we utilized a progressive gating strategy to evaluate the activation status of the different T-cell populations. The CD45+ compartment was isolated, then gated on the basis of CD90.1+ (T-cell marker) expression. T-cell populations were further divided on the basis of CD8 (CTLs) and CD4 expression, which was further divided into CD4+ Foxp3+ (Tregs) and CD4+ Foxp3− (Teff) populations. The activation status of each of these T-cell populations was further assessed using CD44 and PD1 as markers of activation (Fig. 4B and D). These flow cytometric analyses revealed that the Stat3-deficient mammary lesions have elevated levels of CD8+ CTLs (Stat3flx/flx 8.96% ± 1.43 vs. Stat3wt/wt 3.03% ± 0.96; Fig. 4B and C). While both genotypes showed similar levels of activated effector T cells, there was a trend towards a decrease in activated Tregs in the Stat3-deficient mammary lesions and more significantly a substantial increase in activated CTLs (Stat3wt/wt 0.45% ± 0.02 vs. Stat3flx/flx 3.26% ± 0.67; Fig. 4D). Given that a high CD8:CD4 ratio is indicative of an antitumor response (6), we examined this relationship in the Stat3-deficient lesions and found they exhibited a significant increase in CD8:CD4 ratios (CD8:CD4 ratio Stat3flx/flx 1.35 ± 0.21 vs. Stat3wt/wt 0.53 ± 0.07; Fig. 4C). When activation status of these T-cell populations was taken into account, the activated CD8:activated CD4 ratio was even more strikingly elevated in the Stat3-deficient lesions (actCD8:actCD4 ratio Stat3flx/flx 4.37 ± 0.78 vs. Stat3wt/wt 0.32 ± 0.03; Fig. 4D). Collectively, these observations indicate that tumor epithelial expression of Stat3 plays a critical role in establishing immunosuppressive tumor microenvironment early in tumor development.
The T-cell compartment is involved in early stages of immune surveillance of the Stat3-deficient lesions
Although we have shown that loss of Stat3 in emerging hyperplasias results in the rapid recruitment of both myeloid and T-cell populations, the relative contribution of each these immune infiltrates to the immune clearance is unclear. To directly investigate the contribution of the T-cell compartment, we introduced both the Stat3-proficient and deficient MIC lines onto the SCID genetic background, which lacks both B and T cells (26). After a 6-week doxycycline induction, we examined the glands by histology. This showed that, in contrast to the invasive adenocarcinomas exhibited by the Stat3-proficient tumors, the Stat3-deficient lesions were primarily comprised of hyperplasia-like lesions in the immunodeficient SCID background (Fig. 5A, bottom). Notably, at this stage of induction, on immunocompetent FVB genetic background, the Stat3-deficient PyVmT lesions have become completely cleared (Figs. 5A, top and 2A, bottom). Despite the difference in histologic grades between the Stat3-deficient and proficient SCID backgrounds, both sets of lesions exhibited a similar proliferative and apoptotic status (Fig. 5B). While loss of the T- and B-cell compartment rescued the hyperplastic phase of tumor progression, infiltration of non-T- and B-cell immune cell types was still significantly increased to the Stat3-deficient lesions as assessed immunohistochemically with leukocyte and macrophage markers (CD45 and F4/80, respectively), relative to their wild-type counterparts (CD45, Stat3flx/flx 10.69% ± 1.21% vs. Stat3wt/wt 3.33% ± 0.70%; F4/80, Stat3flx/flx 4.59% ± 1.04% vs. Stat3wt/wt 1.10% ± 0.19%; Fig. 5C). Collectively, these observations indicate that full clearance of Stat3-deficient lesions requires an active adaptive immune response. However, given that Stat3-deficient PyVmT lesions did not fully evolve into invasive adenocarcinomas, the Stat3-deficient tumor cells may still be susceptible to the remaining immune populations to limit malignant progression.
A Stat3-dependent inflammatory network is required for the metastatic phase of PyVmT tumor progression
Previous studies with a MMTV-driven ErbB2 model of mammary tumorigenesis have demonstrated that, while mammary epithelial ablation of Stat3 does not impact the induction of ErbB2 mammary tumors, metastasis of these tumors to the lungs is severely impaired (12). To evaluate whether activation of Stat3 plays a similar role in the inducible PyVmT model, we took the lungs of tumor-bearing mice from both the Stat3-proficient and deficient strains at the same tumor burden and scored them for metastatic lesions. Strikingly, none of the Stat3-deficient tumor bearing animals developed lung metastases compared with 100% of their Stat3-proficient counterparts (Fig. 6A and C). Interestingly, although all the tumor-bearing PyVmT mice heterozygous for the Stat3 conditional allele did develop lung metastases, the lungs had, on average, fewer metastatic lesions compared with the wild-type mice (Stat3wt/wt 65.17 ± 16.61 vs. Stat3wt/flx 8.75 ± 3.80; Fig. 6B). To assess whether this metastatic defect was due to an inability to extravasate or a difficulty colonizing a secondary site, we injected freshly dissociated tumors directly into the vasculature of NCr mice. In the absence of any immunosurveillance, the Stat3-deficient cells were still unable to colonize the lungs to the same extent as their wild-type counterparts (Supplementary Fig. S3A–S3C). This suggests an inability of the Stat3-deficient tumors to develop a metastatic niche, survive, and grow in the lungs, independent of an active immune response.
To investigate this metastatic defect, we compared the gene expression profiles of the Stat3-proficient and deficient tumors. The results revealed that the Stat3-deficient tumors exhibited a significant decrease in many proinflammatory genes that are both Stat3 transcriptional targets and are functionally involved in the metastatic cascade (Fig. 6D), many of which we have previously identified in a MMTV-driven ErbB2 model (12). For example C/EBPδ and oncostatin M receptor (OSMR), which are known to potentiate the acute phase response (APR; refs. 27, 28), are significantly downregulated in Stat3-deficient tumors. Interestingly, other downregulated genes in the Stat3-deficient tumors include the serum amyloid A genes, Saa1, Saa2, and Saa3, which are known transcriptional targets of both C/EBPδ and Stat3 through its binding interaction with NFκB transcription factor (29–31). Additional genes involved in inflammation and metastasis were significantly downregulated in Stat3-deficient tumors including Gata2 (5.8-fold), Bcl3 (2.2-fold), Spp1 (3.7-fold), and Rb1 (2.5-fold; Fig. 6D).
To confirm the gene expression profiling data, we performed quantitative real-time PCR (qRT-PCR) on total RNA extracted from the Stat3-proficient and deficient tumors on these key Stat3 target genes. After establishing significantly reduced stat3 transcript levels in the Stat3-deficient tumors compared with the wild-type tumors (19 fold decrease), we established that the expression levels of Cebpd and Osmr were also downregulated (5.7- and 5.2-fold, respectively) in the Stat3-deficient tumors (Fig. 6D). Consistent with the array data, the Stat3-deficient tumors exhibited a dramatic downregulation of the transcript levels of Saa1, Saa2, Saa3, and S100a8. (Fig. 6D).
Given that many of the downregulated Stat3 target genes are involved in promoting a proinflammatory tumor microenvironment, we investigated whether the Stat3-deficient tumors exhibited a deficit in immune infiltrates. To accomplish this, we stained both wild-type and Stat3-deficient tumors with myeloid cell markers CD68 and F4/80. In contrast to the robust immune recruitment, we saw in early Stat3-deficient lesions (Figs. 3 and 4) or end-stage wild-type tumors (Fig. 6E), the Stat3-deficient tumors had a significant reduction in infiltrating CD68+ and F4/80+ populations (Fig. 6E). Furthermore we show that these Stat3-deficient tumors have managed to effectively suppress the immune response as exhibited by low overall leukocyte and T-cell infiltration compared to the wild-type tumors (Supplementary Fig. S3D). Taken together, these results argue that a Stat3-dependent transcription network involved in promoting an inflammatory tumor microenvironment that plays a critical role these tumors' metastatic capacity. Thus, we have demonstrated the importance of tumor epithelial expression of Stat3 in promoting immune suppression during early stages of tumor progression and a prometastatic inflammatory tumor microenvironment in late-stage tumors (Fig. 7). In addition, we propose an interesting model to investigate the mechanisms of immune-mediated tumor elimination.
Several studies have highlighted altered tumor immune microenvironments as an important factor in tumor progression and malignancy. Indeed, several studies have shown that an elevated Th1/Th2 ratio is the best predictor of good outcome in patients with invasive breast cancer (6), while others have shown that increased presence of intratumoral macrophages are associated with metastasis and poor outcome (3–5). However, most of these studies have focused on invasive metastatic disease leaving the molecular mechanisms that promote antitumor immunity to prevent the DCIS to invasive ductal carcinoma (IDC) transition poorly defined.
Stat3 has been well characterized as an important player in inflammation and its activation is often dysregulated in human breast cancers (8, 9, 11). Several in vitro and a few in vivo models have highlighted the importance of Stat3 in tumor growth, invasiveness, and metastasis (11, 12). The recent development of a PyVmT-IRES-Cre mouse model of breast cancer that can be induced in temporal fashion has enabled us to investigate the role of Stat3 throughout all stages of mammary tumor progression. As mammary epithelial–specific expression of PyVmT is dependent on doxycycline, unlike the conventional germline MMTV/PyVmT model, the PyVmT oncogene product is not immunologically tolerized and behaves as a tumor antigen. This unique model has also allowed us to evaluate how tumor epithelial ablation of Stat3 can impact on the tumor immune microenvironment at tumor initiation and progression in a transgenic, immunocompetent context. Furthermore, our model, in the absence of Stat3, nicely recapitulate the 3 phases of immunoediting; elimination, equilibrium, and escape. As we have shown, early Stat3-deficient lesions form to a comparable level to the parental MIC line. However, in the absence of Stat3 these lesions are eliminated by the immune system (Figs. 2–4) followed by a long period of latency, which may be indicative of the equilibrium phase, wherein dormant PyVmT cells continue to reside within these Stat3-deficient mammary glands and are kept in check by the adaptive immune system. Eventually a proportion of these latent PyVmT tumor cells are able to circumvent this equilibrium phase and escape to form detectable tumors (Fig. 1), though they are unable to metastasize to the lung, an observation that correlated with the loss of a proinflammatory tumor microenvironment (Fig. 6).
While Stat3 deletion had little impact on the initial development of mammary epithelial hyperplasias, which retained comparable proliferative and apoptotic potential (Fig. 2A–C), these Stat3-deficient lesions demonstrated a dramatic infiltration of myeloid populations and T-cell infiltrates compared with the parental Stat3-proficient lesions, which eventually led to clearance of these lesions (Fig. 2). Furthermore, we demonstrated that this tumor immune clearance is mediated through a shift away from CD206+ M2-polarized macrophages and an increase in CD103+ DCs and activated CTLs in the absence of epithelial Stat3 (Figs. 3 and 4).
M2-polarized macrophages are thought to play critical role in facilitating breast cancer progression (24, 25). Consistent with these results, epithelial-derived Stat3 has previously been shown to play an important role in the macrophage polarization event, skewing the polarization towards an M2, immunosuppressive state (32). Interesting future directions may include investigating the functional changes in these myeloid cells. By isolating the myeloid cells recruited to these lesions, we could examine differences in their cytokine production, expression of costimulatory molecules and T-cell activation capacity, thus providing additional insight into the mechanism by which they clear the malignant lesions in the absence of Stat3.
Dendritic cells have also come to light as important modulators of tumor immune response (24, 25). The ratio of CD103:CD11b expressing DCs plays a key role in the antitumor activation of CD8+ CTLs and further correlates with better prognosis in human lung and breast cancer (22). Consistent with this, epithelial ablation of Stat3 at early tumor initiation stages results in a higher CD103:CD11b DC ratio (Fig. 3) and an increase in highly activated CD8+ CTLs (Fig. 4). In addition, the Stat3-deficient lesions exhibit a marked increase in their CD8:CD4 ratio, an elevated ratio that correlates with better patient outcome (6).
Further investigation into the molecular mechanism by which Stat3 promotes this immunosuppressive tumor microenvironment is essential. Production of such immunosuppressive factors as IL10, IL23, and TGFβ by Stat3 in a number of hematopoietic lineages has been shown to cause an overall dampening of the immune response (13, 14, 16). Similarly, in the context of the mammary epithelium, Stat3 has been shown to attenuate of the APR, resulting in a decrease in proinflammatory cytokine signalling during mammary gland involution (32) and a Stat3-dependent wound healing–like immune program involving IL10 has been identified in the post-partum involuting mammary gland (33). Investigation into a Stat3-dependent cytokine network responsible for establishing this immunosuppressive tumor microenvironment during DCIS to invasive disease transition may have important therapeutic implications.
The importance of T cells in establishing immune surveillance is further supported by the observation that Stat3-deficient hyperplasias and adenomas can be partially rescued by introducing Stat3-deficient PyVmT lesions into a SCID genetic background (Fig. 5). However, even on this SCID genetic background, these hyperplastic lesions were not able to progress to fully invasive adenocarcinomas, indicating that while the T cells play an important role in the immune clearance of these nascent lesions, other components of the innate immune system are involved in suppressing later stages of tumor progression. Natural killer (NK) cells, which remain intact in the SCID mouse, may play an important role in this. A role for NK cells in antitumor immunity has been well documented in various mouse models and their presence has been correlated with better prognosis in some cancers (34). In fact, the presence of molecular signatures of NK cells were shown to be positive predictors of relapse-free survival in a retrospective study of breast cancer patients (35). Therefore, while we have shown that T cells are major effectors of the immune surveillance observed in our model, the roles of the remaining innate immune cells in suppressing tumor growth remain to be elucidated.
Although these studies provide compelling evidence that tumor epithelial expression of Stat3 is involved in suppressing tumor immune surveillance, 30% of the Stat3-deficient MIC animals develop focal mammary tumors that evaded immune surveillance as evidence by the profound loss of immune infiltrate (Fig. 6) and reached the escape phase of the immunoediting program. The precise molecular mechanism by which this occurs remains to be elucidated. Possible mechanisms involved in this escape from immune surveillance involve major changes at either the tumor cell level or that of the tumor microenvironment, although, in general, these mechanisms are not fully understood (2). On the basis of the long latency of these Stat3-deficient tumors, there is opportunity for stochastic genetic or epigenetic alterations to occur, leading to the eventual outgrowth of the mammary tumors. This may result in reduced immunogenicity of the tumors due to downregulation of immunogenic molecules such as the MHC class I and II molecules or resistance to the cytotoxic functions of the immune cells (2). Indeed other genetic alterations could result in the activation of several other transcription factors have been implicated in supporting an immunosuppressive environment including NF-κB, AP-1, and SMAD (36). Alternatively, this tumor escape could be the result of peripheral tolerance against the PyVmT antigen from CD8+ T-cell anergy or deletion as shown by Willimsky and colleagues using a sporadically expressed dormant viral Tag oncogene (37). Thus, while there is evidence of possible mechanisms to explain this escape, the precise mechanisms in play in our model have yet to be determined.
Despite the capacity to eventually escape immune surveillance, the Stat3-deficient tumors were incapable of metastasizing to the lungs. Consistent with the importance of Stat3 in the metastatic phase of tumor progression, we have previously demonstrated that Stat3 ablation in an ErbB2 model of breast cancer resulted in a similar metastatic blockade (12). Interestingly, in both models, gene expression analyses revealed the downregulation of a Stat3-dependent inflammatory transcriptional network involving Saa1, Saa2, Saa3, S100a8, Cepbδ, and Osmr and implicates it in the metastatic program (12). Indeed, elevated expression of OSMR has been correlated with poor prognosis in several cancers (38). In addition, SAA3 and S100A8 expression at both the primary tumor site and premetastatic niche facilitates myeloid cell recruitment and metastasis (39–41). Consistent with this concept, the Stat3-deficient PyVmT tumors exhibit a marked deficit in TAMs (Fig. 6). Given the importance of macrophage recruitment to primary tumor site in metastasis (7, 42) these data indicate that Stat3 plays a critical role in promoting a metastatic tumor microenvironment.
These observations have several important implications in the therapeutic management of metastatic breast cancers. The observation that Stat3 plays a critical role in suppressing immune surveillance raises the intriguing possibility that small-molecule inhibitors targeting Stat3 may be effective agents to reactivate the immune surveillance program. Interesting combinatory studies looking at Stat3 inhibition along with new immunostimulatory treatments such as anti-PDL1 and anti-CTL4 antibody therapies may also prove efficacious in mobilizing antitumor immune response. Further studies into the mechanism of this Stat3-mediated immune evasion may greatly benefit the current research into various immunotherapies or support the role for direct Stat3 inhibitors in both early and metastatic breast cancers.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: L.M. Jones, J.J. Ranger, W.J. Muller
Development of methodology: L.M. Jones, J.J. Ranger, J. Ursini-Siegel, M. Krummel
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.M. Jones, M.L. Broz, J.J. Ranger, R. Ahn
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.M. Jones, M.L. Broz, J.J. Ranger, J. Ozcelik, J. Ursini-Siegel, M.T. Hallett, M. Krummel, W.J. Muller
Writing, review, and/or revision of the manuscript: L.M. Jones, M.L. Broz, J.J. Ranger, M. Krummel, W.J. Muller
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.M. Jones, R. Ahn, D. Zuo
Study supervision: J.J. Ranger, W.J. Muller
The authors thank Vasilios Papavasiliou and Cynthia Lavoie for technical support and Dr. Connie Krawczyk for helpful discussion.
This work was supported by the Terry Fox Frontier Program Project Team Grant (020002), McGill Integrated Cancer Research Training Program Fellowship and McGill University Health Center Studentship (L.M. Jones), McGill University Canadian Research Chair in Molecular Oncology (W.J. Muller), Genentech Predoctoral Research Fellowship, the Margaret A. Cunningham Immune Mechanisms in Cancer Research Fellowship Award, and the Achievement Reward for College Scientists Scholarship (M. Broz), NIH grants U54 CA163123 and R21CA191428 (M. Krummel), and Canadian Cancer Society (grant 702060; J. Ursini-Siegel).