Generation of PROCR+/ZEB1+/PDGFRα+ (PZP) cell lines from breast tissues of women of African ancestry
We created nine immortalized PZP cell lines from breast tissues of women of AA (KTB40, KTB42, KTB32, KTB53, KTB57, KTB59, KTB104, KTB106, and KTB109181) by overexpressing human telomerase gene (hTERT) using primary cells isolated and propagated from core breast biopsies. Enrichment of AA informative markers in these donors was confirmed through genotyping (Fig. 1a, and Fig. S1a). To phenotypically characterize and document heterogeneity, these cells were subjected to flow cytometry using various epithelial, stem, mesenchymal, and fibroblast markers. These cell lines were predominantly PROCR+/EpCAM− (Fig. 1b, c, and Fig. S1b–d). To further characterize stem cell-related gene expression, we compared these immortalized variants with the immortalized PROCR±/EpCAM+ luminal/basal cell variants from women of AA and EA we described previously30. PROCR+/EpCAM– cell lines expressed significantly higher levels of ZEB1 compared to PROCR±/EpCAM+ (KTB34 and KTB39) breast epithelial cell lines (Fig. 1d, and Fig. S1e). The gene expression pattern in KTB34 cell line from woman of EA overlaps with the Luminal-A intrinsic subtype of breast cancer, whereas gene expression in KTB39 from woman of AA overlaps with Normal-like intrinsic subtype of breast cancer30.
A recent study described PDGFRα+ stromal cells as adipogenic progenitors of the mammary gland that trans-differentiate into epithelial cells and migrate into the duct when stimulated by Platelet-derived growth factor-C (PDGF-C)25. Interestingly, these cells also express PROCR25. We examined whether human breast PROCR+/ZEB1+ cells express PDGFRα. Indeed, >70% of cells were PDGFRα+ (Fig. 1e, f; Fig. S1f–h), and these cells also expressed PDGFRβ (Fig. S1i). CD49fhigh/EpCAMlow, CD49fhigh/EpCAMmedium, and CD49flow/EpCAMhigh cells are described as breast basal, luminal progenitor, and mature/differentiated epithelial cells, respectively31. None of the PZP cell lines were positive for CD49f and EpCAM (Fig. S2a). Morphologically, PZP cell lines showed features of mesenchymal stem and fibroblast like cells, which are loosely attached to cell culture plates, unlike epithelial cells (Fig. S2b).
Trans-differentiating properties of PZP cells
Both adipogenic progenitors described in the mouse mammary gland25 and fibroadipogenic progenitors (FAPs) described in the skeletal muscle32 are able to undergo trans-differentiation into adipogenic and osteogenic lineages. In order to determine to what extent PZP cells show similarity to these other progenitors, PZP cells were subjected to adipogenic and osteogenic differentiation growth conditions. Indeed, PZP cells differentiated into adipocytes (Fig. 2a) and the lipid content was significantly increased in the differentiated PZP adipocytes (Fig. 2b). The PZP differentiated cells expressed adipocyte differentiation marker PPARγ (Fig. 2c). Adipogenic trans-differentiation capability of PZP cells was further confirmed through RUNX1 expression analysis (Fig. 2d). The PZP cells cultured in osteogenic media for 21 days showed osteogenic differentiation as evident from mineralization of matrix with Ca2+ and positive alizarin red staining (Fig. 2e, f). Thus, PZP cells enriched in the breasts of women of AA could correspond to multipotent cells that can trans-differentiate into different cell types based on environmental cues.
PZP cells express lobular fibroblast markers
In breasts, loose connective tissue is unique to the terminal duct lobular units (TDLUs), which drain into dense connective tissue embedded interlobular ducts33. CD105high TDLU-resident lobular fibroblasts display properties different from interlobular fibroblasts34. While the CD105high lobular fibroblasts resemble mesenchymal stem cells (MSCs) both by phenotype and function, CD26high interlobular cells remain fibroblast restricted34. CD105high/CD26low and CD105low/CD26high lineages are considered to represent lobular and interlobular human breast fibroblastic cells (HBFCs), respectively35. To further characterize PZP cell lines, we examined the CD105 and CD26 staining pattern. PZP cells are enriched for CD105high/CD26low population with inter-individual variability in the ratio between CD105high/CD26‒ and CD105high/CD26low cells (Fig. 3a–c, and Fig. S2c, d), which suggest that PZP cells reside predominantly in the lobules of breasts.
CD90‒/CD73high and CD90high/CD73high cells are described as rare endogenous pluripotent somatic stem cells and potential mesenchymal stem cells, respectively36. PZP cells contained CD90‒/CD73high and CD90high/CD73high subpopulations, with notable inter-individual variability in the ratio between CD90‒/CD73high, CD90low/CD73high, CD90med/CD73high and CD90high/CD73high populations (Fig. 3d–g; Fig. S2e–h). CD44 and CD24 are the “original” markers used to characterize cancer stem cells (CSCs) in breast cancer37. Interestingly, PZP cells displayed CD44high/CD24low phenotype (Fig. 3h, i; Fig. S2i, j). However, lack of EpCAM expression in PZP cells suggest that they are not the “normal counterparts” of cancer stem cells. CD10 marker is used to isolate myoepithelial cells, although a recent study showed CD10 positivity of cancer associated fibroblasts38,39. PZP cells showed CD10+ phenotype (Fig. 3j, k; Fig. S2k and l).
CD44+/CD24– and Aldehyde Dehydrogenase+ (ALDH+) cancer cells of the breast have been described as mesenchymal and epithelial-like breast cancer cells, respectively, and gene expression in these cells overlap with basal and luminal epithelial cells of the normal breast40. While PROCR±/EpCAM+ epithelial cell lines (KTB34 and KTB39) contained ALDH+ cells, as measured through ALDEFLUOR assay, none of the PZP cell lines contained significant levels of ALDH+ cells (Fig. S3a, b).
PZP cells are enriched in the normal breasts of women of African compared to European ancestry
In our previous study, we had demonstrated that ZEB1+ cells in the normal breast are located in the stroma with close proximity to ductal epithelial cells8. We generated a tissue microarray (TMA) comprising healthy breast tissues from 49 and 154 women of AA and EA, respectively. The TMA was first created based on self-reported Race/ancestry and subsequently linked to genetic ancestry. Ancestry marker distribution pattern of donors is shown in Fig. S4a, b and details of ancestry analysis are provided in Materials and Methods. We used recently suggested race and ethnicity reporting guidelines to describe study populations41. AA marker levels ranged from 30 to 99% in self-reported Black women. EA marker levels ranged from 20-98% in self-reported non-Hispanic White women. The TMA was analyzed for protein levels of PROCR, ZEB1 and PDGFRα by immunohistochemistry (IHC). Because of our previous observation of elevated ZEB1+ cells in the non-tumor adjacent to tumor tissues (NATs) compared to breast tissues from healthy donors of European ancestry8, the TMA also contained paired breast tumors and NATs of self-reported race (African American and White). However, for simplicity and to avoid confusion, NATs and tumors from self-reported African American and White women are labelled as AA and EA in figures and figure legends. To distinguish tissues of clinically healthy donors from NATs, clinically healthy tissues are labelled as Normal-Healthy. Clinicopathologic features of tumors are described in Table S1. Representative IHC staining patterns of PROCR in Normal-Healthy, NATs, and tumors are shown in Fig. 4a, b and statistical analyses are presented in Fig. 4c–e and in Table S2. PROCR expression was observed in both ductal and stromal cells (Fig. 4a, b) with ~5-fold higher expression in stromal cells compared to epithelial cells (Fig. S5a). PROCR-expressing cells were enriched in the Normal-Healthy breast tissues of women of AA compared to EA (Fig. 4a–c; Table S2). PROCR H-Score remained significantly higher in women of AA compared to EA when the analysis included only those with >75% African and >75% European ancestry markers (H-score: 56.41 in 17 AA vs 21.97 in 70 EA, p value < 0.0001). Furthermore, PROCR H-Score positively correlated with African ancestry proportions (Coefficient 0.347, p = 2.75E−06) and negatively with European ancestry proportions (Coefficient −0.265, p = 4.00E−04). NATs of women of AA and EA contained significantly higher levels of PROCR+ cells compared to Normal-Healthy suggesting the field effect of tumors on PROCR expression and NATs are not “normal” (Fig. 4d). Tumors of women of EA displayed a modest increase in PROCR+ cells compared with those of NATs, while in women of AA, there were no differences between NATs and tumors (Fig. 4e). Thus, PROCR+ cells are intrinsically higher in the normal breasts of women of AA.
Consistent with our previous report regarding ZEB1 expression8, Normal-Healthy breast tissues of women of AA displayed higher levels of ZEB1 H-score compared to women of EA. (Fig. 5a–c; Table S2). ZEB1 H-Score negatively correlated with EA proportions (Coefficient −0.208, p = 6.45E−03). ZEB1 H-Score was not significantly different between AA and EA women when the analysis included only those with >75% African and >75% EA ancestry markers, possibly due to small sample size (H-score: 1.66 in 13 AA vs 1.37 in 83 EA, p value 0.9658). NATs of women of both AA and EA contained significantly higher levels of ZEB1+ cells compared to corresponding Normal-Healthy (Fig. 5d). Tumors of women of EA displayed higher levels of ZEB1+ cells compared to NATs (Fig. 5e).
We examined the ZEB1 expression with respect to income level as a surrogate marker of socioeconomic stress (<20 K, 20–50 K, 50–100 K and >100 K in $), but did not observe any significant differences based on income levels (Fig. S5b, c). While this is an incomplete measure of socioeconomic stress, it supports the findings that the normal breasts of women of AA exhibit intrinsically higher levels of ZEB1+ cells but ZEB1+ cell numbers increase in the breast with cancer in women of EA, consistent with our previous report8.
PDGFRα expression in the Normal-Healthy breast tissues was low and mostly in stromal cells (Fig. 6a and Fig. S5d). However, tumor epithelium showed some degree of positivity. PDGFRα-expressing cells were enriched specifically in the Normal-Healthy breast tissues of women of AA compared to women of EA (Fig. 6a–c; Table S2). PDGFRα H-Score remained significantly higher in women of AA compared to women of EA when the analysis included only those with >75% African and >75% European ancestry markers (H-score: 38.62 in 16 AA vs 13.41 in 87 EA, p value 0.0027). The PDGFRα H-Score positively correlated with AA proportions (Coefficient 0.484, p = 1.62E−10) and negatively with EA proportions (Coefficient −0.404, p = 1.74E−07). Normal-Healthy breast tissues contained significantly higher levels of PDGFRα+ cells compared to NATs only in women of AA, while no difference between Normal-Healthy breasts and NATs of women of EA was observed (Fig. 6d). Furthermore, we did not observe a significant difference in expression between NATs and tumors of women of AA (Fig. 6e), suggesting that PDGFRα+ cells are intrinsically higher in the normal breasts of women of AA. Taken together, our data suggest that PZP cells are elevated in the normal breasts of women of AA compared to women of EA, although expression of PROCR and PDGFRα in epithelial cells makes data interpretation a bit difficult.
We next examined the same TMA for luminal epithelial markers to rule out possible bias in our observations. Because FOXA1 along with another pioneer factor GATA3 and ERα form a lineage-restricted hormone-responsive signaling network in the normal breast42, we examined the expression levels of these markers. Representative staining patterns of ERα, GATA3, and FOXA1 are shown in Fig. S6, Fig. S7, and Fig. S8, respectively. We did not observe genetic ancestry-dependent differences in ERα levels in Normal-Healthy tissues (Fig. S6a–c). NATs of women of AA contained modestly higher levels of ERα+ cells compared with those of Normal-Healthy, while this difference was not observed in NATs of women of EA (Fig. S6d). Tumors of women of EA contained significantly higher levels of ERα+ cells compared with those of NATs, while this difference was not observed in women of AA (Fig. S6e). This is consistent with pathologic features of tumors in women of EA as the majority of them were ER+ tumors (Table S1, 75% ER+ in EA vs 60% ER+ in AA; 8% TNBCs in EA vs 16% TNBCs in AA). Similar to ERα, GATA3 expression in the normal breasts did not show genetic ancestry-dependent differences (Fig. S7a–c). NATs of women of AA contained significantly higher levels of GATA3+ cells compared with those of Normal-Healthy, while this difference was not observed in women of EA (Fig. S7d). Tumors of women of EA contained significantly higher levels of GATA3+ cells compared with those of NATs, which is consistent with the clinical features of tumors, while this difference was not observed in women of AA (Fig. S7e). FOXA1 also did not show any genetic ancestry-dependent differences in Normal-Healthy tissues (Fig. S8a–c). NATs of women of AA and EA contained significantly higher levels of FOXA1+ cells compared with those of Normal-Healthy (Fig. S8d). As expected, tumors of women of EA contained significantly higher levels of FOXA1+ cells compared with those of NATs, while this difference was not observed in women of AA (Fig. S8e). These results suggest that the stromal but not the epithelial compartment of the normal breast shows genetic ancestry-dependent variability in cell composition, at least with the markers examined.
Modeling the effects of PZP cells on breast tumorigenesis
To obtain potential insight into signaling pathway alterations in epithelial and PZP cells as a consequence of their crosstalk, we performed cytokine/chemokine profiling of factors secreted by an immortalized luminal epithelial cell line, a PZP cell line and both co-cultured together (50% of each cell line) for ~24 h. If the expression under the co-culture condition was much higher than expression in either cell type alone, we interpreted those results as showing a cooperative effect on gene expression. If the expression under co-culture conditions was the same or lower than expression in either cell type alone, we interpreted those results as showing no effect of cell-cell interaction. We selected the Luminal-A epithelial cell line KTB34 (EA donor) for this experiment because this cell line upon transformation generates adenocarcinoma, similar to human breast cancer30,43. While the luminal epithelial cell line expressed several ligands such as PDGF-AA and osteopontin, which can affect trans-differentiation of PZP cells, PZP cells expressed factors such as EGF, HGF and SDF-1α, which can signal in luminal cells (Fig. 7a; Table S3). Interestingly, IL-6 was produced only under co-culture conditions (Fig. 7a; Table S3). We further confirmed IL-6 production under co-culture conditions at the mRNA level by qRT-PCR (Fig. 7b). We suspect PZP cells produce IL-6 in response to interaction with luminal cells as the basal expression of IL-6 was much higher in PZP cells compared to the luminal cell line (Fig. S9a), and luminal cells secreted IL-1α, which we have previously shown to induce IL-6 in stromal cells44. There appears to be specificity in cytokine production under co-culture conditions as we did not observe an elevated production of IL-8 under co-culture conditions of PZP and epithelial cells (Fig. S9b).
PZP-epithelial cell interaction alters the expression of specific genes
We performed a literature search to identify potential genes whose expression could be altered due to stromal-epithelial cell interaction. For example, transgelin (TAGLN) is known to be a specific marker of smooth muscle differentiation and is highly expressed in the myoepithelial cells and fibroblastic cells of benign breast tissue with limited expression in luminal cells45,46. A population of subepithelial cells that lines the entire villus-crypt axis of the intestine expresses high levels of PDGFRα, Delta Like Canonical Notch Ligand 1 (DLL1), F3, and EGF-family ligand Neuregulin 1 (NRG1)47. In addition, NRG1 is also expressed in mesenchymal cells adjacent to the proliferative crypts47. Expression of these genes were examined in individual cell types and under co-culture. Three PZP cell lines (KTB32, KTB40, KTB42) and two epithelial cell lines (KTB34, KTB39), as well as co-culture of PZP and epithelial cell lines (50% of each cell line) were used. We observed abundant TAGLN expression in PZP cells, while epithelial cells expressed it at low levels. The expression of TAGLN was further increased under co-culture condition (Fig. 7c). We found a low level of DLL1, F3, and NRG1 expression in PZP cell lines, except for a high level of NRG1 in KTB42. DLL1, F3 and NRG1 are expressed mostly in epithelial cell lines. In co-cultured cells, expression of DLL1 and F3 was additive depending on the cell type (Fig. S9c–e). Taken together, these results indicate that PZP cells correspond to stromal cells that interact with epithelial cells to alter gene expression in a reciprocal manner. However, these cells are unlikely to function similar to subepithelial mesenchymal cells described in the intestine47.
Potential role of PZP cells in immune cell modulation in the microenvironment
Several other factors aside from IL-6 have been shown to alter the immune microenvironment. For example, WNT1-inducible-signaling pathway protein 1 (WISP1) expression affects the clinical prognosis by promoting macrophage M2 polarization and immune cell infiltration in pan-cancer and helps to maintain CSC properties in glioblastoma48. PZP cell lines displayed higher expression of WISP1 compared to luminal cell lines, while additive expression was observed under co-culture conditions (Fig. 7d). Tenascin-C (TNC) promotes an inflammatory response by inducing the expression of multiple proinflammatory factors in innate immune cells such as microglia and macrophages. TNC drives macrophage differentiation and polarization predominantly toward an M1-like phenotype49. TNC is expressed mostly by epithelial cells and this expression was unaffected under co-culture conditions (Fig. 7e). Colony stimulating factor 1 (CSF1) controls both the differentiation and immune regulatory function of macrophages50. PZP cell lines displayed higher expression of CSF1 compared to epithelial cell lines and the expression was unaffected under co-culture conditions (Fig. 7f). Osteopontin (SPP1 or OPN), secreted by myofibroblasts, promotes M2 macrophage polarization through the STAT3/PPARγ pathway51. PZP cell lines displayed high expression of SPP1 compared to epithelial cell lines and the expression remained additive under co-culture conditions (Fig. 7g). IL-33 is known to be upregulated in metastases-associated fibroblasts, and the upregulation of IL-33 activates type 2 inflammation in the metastatic microenvironment and facilitates eosinophil, neutrophil, and inflammatory monocyte recruitment to lung metastases52. Co-culturing of PZP and epithelial cells did not affect IL-33 expression (Fig. 7h). CMTM6 maintains the expression of PD-L1 in tumor cells to regulate anti-tumor immunity53. Epithelial cells but not PZP cells expressed higher levels of CMTM6, which was unaffected by co-culture conditions (Fig. S9f). Macrophage migration inhibitory factor (MIF) is an essential cytokine that is involved in the regulation of macrophage function in host defense through the suppression of anti-inflammatory effects of glucocorticoids54. Both PZP and epithelial cell lines expressed MIF, but co-culture conditions did not further affect expression (Fig. S9g). Secretion of MFGE8 can reprogram macrophages from an M1 (proinflammatory) to an M2 (anti-inflammatory but pro-tumorigenic) phenotype55. MFGE8 also induces the production of basic fibroblast growth factor, which is responsible for fibroblast migration and proliferation56. PZP cell lines displayed high expression of MFGE8 which was unaffected by co-culture conditions (Fig. S9h). Periostin (POSTN) is predominantly secreted by stromal fibroblasts to promote the proliferation of tumor cells. POSTN is also an essential factor for macrophage recruitment in the tumor microenvironment and is involved in the interactions between macrophages and cancer cells57. PZP cell lines displayed high expression of POSTN, and the expression was unaffected by co-culture conditions (Fig. S9i). We noted variation in the expression of select genes amongst PZP cell lines (WISP1, CSF1, NRG1, and POSTN, for example), suggesting inter-individual variability in gene expression in stromal cells, similar to epithelial cell types we described previously30 and consistent with phenotypic heterogeneity displayed by PZP cell lines (Fig. 3d, for example). Collectively, these results suggest that PZP-epithelial cell interaction in the breast could impact the levels of select chemokines/cytokines in the breast environment (IL-6 and TAGLN, for example) with consequential effects on the normal and/or tumor immune environment.
The effects of breast epithelial: PZP cell interaction on trans-differentiation of epithelial luminal progenitor cells
In order to further investigate the intercellular communication between PZP and epithelial cells, we generated stable tdTomato-labeled KTB40 and KTB42 cell lines using pCDH-EF1-Luc2-P2A-tdTomato lentivirus (Fig. S10a). To determine whether co-culturing alters the phenotype of epithelial cells, tdTomato-labeled PZP (KTB40, KTB42), epithelial (KTB34, KTB39), and co-cultured PZP and epithelial cells were analyzed by flow cytometry using CD49f and EpCAM, which can differentiate luminal mature (CD49f−/EpCAM+), luminal progenitor (CD49F+/EpCAM+) and basal (CD49f+/EpCAM−) cells. Co-cultured epithelial cells displayed an increase in CD49f+/EpCAM−/low subpopulation of cells. Reduced CD49f-/low/EpCAM+/high cell subpopulations were observed in co-culture conditions (Fig. 8a–c). Isotype controls are shown in Fig. S10b. The tdTomato-Red+ PZP cells on their own were CD49f−/EpCAM−; a small fraction of these cells acquired a CD49f+/EpCAM+ phenotype under co-culture conditions (Fig. 8d, e). These results suggest that similar to mouse mammary fibroadipogenic cells25, PZP cells can potentially acquire epithelial characteristics under specific conditions. We could not successfully characterize flow sorted CD49f+/EpCAM+ trans-differentiated PZP cells because of dominant growth of a few contaminating PZP cells in the sorted population of cells.
We further confirmed PZP cell mediated transition from luminal to basal characteristics of epithelial cells by ALDEFLUOR assay. Epithelial cells co-cultured with PZP cells lost ALDH-positivity (Fig. S10c). Interestingly, consistent with PZP cells acquiring epithelial characteristics upon co-culture with epithelial cells (based on CD49f+/EpCAM+ positivity noted above), a few of the PZP cells became ALDH+ under co-culture conditions (Fig. S10d). These results suggest that PZP-epithelial cell interactions lead to trans-differentiation of both cell types.
IL-6 plays a partial role in PZP cell-induced trans-differentiation of epithelial cells
Because co-culturing of PZP cells with epithelial cells caused increased expression of IL-6 in PZP cells, we next examined whether IL-6 plays any role in trans-differentiation of epithelial cells when in contact with PZP cells. The ability of the luminal-like KTB34 cell line to trans-differentiate into CD49f+/EpCAM– basal like cells when co-cultured with PZP cells was significantly reduced when co-culture conditions contained neutralizing antibody against IL-6R (Fig. S11a, b). The effect of IL-6R antibody in reducing trans-differentiation of KTB39 was limited. Unlike KTB34 cells, which have luminal features, KTB39 from a woman of AA displays normal-like intrinsic subtype features30 and is possibly susceptible to trans-differentiation by other cytokines/chemokines.
We evaluated the effect of IL-6 inhibition on CD44, EpCAM, CD24 and CD10 cell surface marker profiles of KTB34 and KTB39 cells with and without co-culturing with KTB40 or KTB42 PZP cell lines. While co-culturing increased the percentage of CD44+/EpCAMlow KTB34 and KTB39 cells, IL-6R neutralizing antibody had minimum effect on this trans-differentiation (Fig. S11c, d). Similarly, IL-6R antibody treatment did not influence the effects of PZP cells in epithelial cells losing CD10 expression (gain of CD24+/CD10low phenotype) (Fig. S12a, b).
We next evaluated the effect of IL-6 inhibition on the STAT3 signaling pathway and measured the phosphorylation levels of STAT3 (Y705 and S727) in KTB34 cells treated with conditioned media (CM) from KTB34, KTB40, and co-cultured KTB34 and KTB40 cells. These two phosphorylation were selected because they differentially regulate epithelial-mesenchymal-transition (EMT) to mesenchymal-epithelial-transition (MET) switch during metastasis58. CM from KTB34 and KTB40 cell lines induced phosphorylation of STAT3 (Y705) at variable levels but the induction was highest when CM from co-cultured cells was used (Fig. 8f). IL-6R antibody drastically reduced KTB40 and co-culture CM-mediated pSTAT3 (S705) phosphorylation. Because of very high levels of basal phosphorylation at S727 residue due to culture conditions, the effect of PZP cell CM on S727 phosphorylation was modest. These results suggest that enhanced IL-6 production by PZP cells as a consequence of their interaction with epithelial cell leads to STAT3 activation in epithelial cells.
Normal breast and/or breast tumors of women of African ancestry contain higher levels of phospho-STAT3
Since PZP-epithelial cell interaction resulted in elevated IL-6 expression in vitro and PZP cells are present at higher levels in the normal breast of women of AA, it is expected that signals downstream of IL-6 should be higher in the breast tissues of women of AA compared to women of EA. We used phospho-STAT3 as a surrogate marker to determine IL-6 activity in the normal breast tissues. Antibody against S727 but not S705 was sensitive in immunohistochemistry. Although cells in the Normal-Healthy breast tissues contained lower levels of phospho-STAT3 as we could estimate only positivity, positivity was still higher with healthy breast tissues of women of AA compared to women of EA (Fig. 9a–c). Modestly higher levels of phospho-STAT3 were also observed in tumors of women of AA compared to women of EA (Fig. 9c). Consistent with elevated levels of PZP cells in NATs compared to Normal-Healthy in women of EA, phospho-STAT3 levels were higher in NATs compared to Normal-Healthy of women of EA (Fig. 9d). We analyzed several publicly available proteomics and transcriptomics datasets59,60 using the UALCAN database61 to determine whether STAT3 activity levels, as measured through phosphoprotein levels or STAT3 gene signatures62, are higher in breast tumors of African American women compared to non-Hispanic White women (note information in these cases is limited to self-reported race and ethnicity). Significantly elevated STAT3 phosphorylation at specific residues of STAT3 as well as STAT3 gene signatures were noted in tumors from African American women compared to tumors from Asian women but not between tumors from African American and non-Hispanic White donors. Overall, our results and these publicly available data support the impact of genetic ancestry on phosphorylation/activity status of STAT3.
Transformed PZP cells generate metaplastic carcinoma
We recently reported that cell-of-origin but not oncogenic mutations determine the histotypes of breast tumors using breast epithelial cell lines derived from multiple donors and a defined set of oncogenes43. We used the same strategy to determine whether PZP cells are cell-of-origin of specific malignancy of the breast. Mutant Ras is one of the potent oncogenes used to transform breast epithelial cells in vitro63. Although initially considered not a relevant oncogene in breast cancer and breast cancer-related studies that utilized Ras oncogenes were often dismissed upon by reviewers, recent studies have clearly shown the role of Ras oncogene in endocrine resistance and metastasis of luminal breast cancer64. Furthermore, the majority of breast cancer cell lines used as model systems have either mutation in Ras genes themselves or in downstream effectors or negative regulators of Ras pathway. Ras pathway is the major signaling pathway activated in TNBCs and Claudin-low breast cancer subtype65. We transformed PZP cell lines with HRasG12V, SV40-T/t antigen and the combination of both HRasG12V and SV40-T/t antigens using lentivirus, since this combination was the most effective in breast epithelial cell transformation43. SV40-T/t antigens impair tumor suppressor functions of retinoblastoma and p53 proteins, inhibit the activity of protein phosphatase 2 A and generate gene expression signatures found in aggressive breast cancers66. Western blotting was used to detect the overexpression of mutant HRasG12V (Fig. S13a), SV40-T/t antigens (Fig. S13b), and combination of both HRasG12V and SV40-T/t antigens (Fig. S13c). Phase contrast images of PZP transformed cell lines are shown in Fig. S13d. Transformation of PZP cells with activated HRasG12V increased the fraction of cells that have acquired epithelial phenotype and express EpCAM, particularly in KTB42 (Fig. 10a, b; Fig. S13d). Transformed PZP cells expressed the basal cell marker CD49f with inter-individual variability (Fig. 10a; Fig. S14a for isotype control). Transformation also altered the cell surface profiles of the mesenchymal stem cell marker CD90 with a few of PZP lines demonstrating loss of CD90 expression (Fig. 10c; Fig. S14a). Transformed PZP cells were PROCR+ and CD44+ (Fig. 10b Fig. S14b). Thus, PZP cells further undergo trans-differentiation upon transformation by acquiring CD49f expression.
We next determined whether cells expressing oncogenes are tumorigenic in NSG mice. Indeed, five million transformed cells in 50% matrigel implanted into the mammary gland of 6–7-week-old female NSG (NOD/SCID/IL2Rgnull) mice (n = 5) progressed into tumors. All animals injected with HRasG12V + SV40-T/t and four out of five animals injected with HRasG12V transformed cells generated tumors. Parental immortalized cell lines (n = 5) did not generate tumors. Tumor was resected and analyzed by H&E staining and expression of luminal markers estrogen receptor alpha (ERα), GATA3, and FOXA1 and cytokeratins CK5/6, CK8, CK14 and CK19 was examined. The luminal cells express cytokeratin 19 (CK19), while basal cell types express cytokeratin 5/6 (CK5/6) and cytokeratin 14 (CK14), and cells expressing both CK14 and CK19 show luminal progenitor phenotype or correspond to luminal/basal hybrid cells of the breast67. KTB42-HRasG12V cell-derived tumor was ERα–/GATA3–/FOXA1+ (Fig. 10d). KTB42-HRasG12V-derived tumor was also CK5/6−/CK8−/CK14−/CK19− (Fig. S14c). KTB42 cell line transformed with both mutant HRasG12V and SV40-T/t antigen also developed tumors in NSG mice. KTB42- HRasG12V + SV40-T/t cell-derived tumor was ERα−/GATA3−/FOXA1− (Fig. 10e), and CK8−/CK14−/CK19− (Fig. S14d). Unlike luminal breast epithelial cell-derived tumors obtained after transformation with the same set of oncogenes43, which metastasized to lungs, these tumors did not show extensive lung metastasis (Fig. 10f, g). Histologically, these tumors were metaplastic carcinomas, which comprise 0.08–0.2% of all breast neoplasms (Fig. 10g)29.
To further establish relationship between PZP cells and metaplastic carcinomas, we compared characteristics of PZP cells with published metaplastic carcinoma protein/gene expression datasets68,69. We had previously reported RNA-seq data of immortalized PZP cell lines KTB40 and KTB42 and six immortalized breast epithelial cell lines (KTB21, 22, 26, 34, 37, and 39)30. Similar to metaplastic carcinomas that show EMT characteristics with enhanced mesenchymal gene expression and claudin low phenotype, PZP cells expressed higher levels of EMT-associated genes ZEB1, ZEB2, SNAI1 and TWIST1 but lower levels of E-Cadherin and Claudin-7 (CLDN7) compared to epithelial cells. PZP cells did not express CLDN3 and CLDN4. Gene Set Enrichment Analysis (GSEA) of PZP cells compared to epithelial cells revealed enrichment of EMT, IFNγ and apical junction signatures, and reduced OXPHOS signatures (Fig. S14E), similar to metaplastic carcinomas68. Since cell-of-origin gene signatures instead of mutation driven signatures predominate in 33 cancer types that have been examined70, these results further support the possibility of PZP cells being one of the cells-of-origin of metaplastic carcinomas.
To ensure that tumors originate from PROCR+/EpCAM− cells instead of a minor fraction of contaminating epithelial cells, we sorted PROCR+/EpCAM− cells from the KTB42 cell line (Fig. S15a). Gate for the sort was placed in a such a way that there is limited possibility of EpCAM+ cell contamination. Sorted PROCR+ KTB42 cells transformed with mutant HRasG12V and SV40-T/t antigen (five million) in 50% matrigel implanted into the mammary gland of 6–7-week-old female NSG mice (n = 6) progressed into tumors. We characterized the cell line established from the tumor to determine the expression of PROCR and other phenotypic markers. Human specific antibody against Na+/K+ ATPase CD298 (ATP1B3) cell surface marker was used to sort CD298-enriched human tumor cell populations from mouse stromal cells (Fig. S15a)43. A cell line established from tumor of KTB42- HRasG12V + SV40-T/t transformed cells showed PROCR+ cells similar to transformed cells in vitro (Fig. S15a). Tumor-derived cell line expressed the basal cell marker CD49f similar to transformed cells. Tumor-derived cell line displayed an increase in CD90high/CD73+ cell population compared to transformed cells in vitro (Compare Fig. S15a with Fig. 10c). Tumor-derived cells were CD44+ and CD10+ similar to transformed cells (Fig. S15a). H&E, PROCR, and EpCAM staining of resected tumors revealed PROCR+ spindle-like tumor cells with pockets of PROCR+/EpCAM+ tumor cells (Fig. S15b, c). No lung metastasis was observed.
To further clarify that tumors are derived from PROCR+/EpCAM− cells and few of the tumor cells trans-differentiate into EpCAM+ cells, we generated a single cell-derived clone (D4) from PROCR+/EpCAM− cell population of KTB42-HRasG12V + SV40-T/t transformed cells by sorting PROCR+/EpCAM− cells using flow cytometry (Fig. S15d). Single cell-derived clone D4 was PROCR+, but a small fraction of cells trans-differentiated into EpCAM+ cells (Fig. S15d). Five million single cell derived PROCR+ KTB42-HRasG12V + SV40-T/t transformed clone D4 cells in 50% matrigel were implanted into the mammary gland of 6–7-week-old female NSG mice (n = 5). All animals injected with the single cell derived PROCR+ KTB42-HRasG12V + SV40-T/t transformed clone D4 generated tumors (Fig. S15e, f). Similar to tumors derived from bulk PROCR+/EpCAM− transformed cells, tumors derived from the single cell derived clone were also PROCR+ and spindle shaped with pockets of EpCAM+ tumor cells. Collectively these results suggest that PZP cells can generate tumors similar to metaplastic breast carcinomas with pockets of malignant epithelial cells71. A summary of major findings of this study is presented schematically in Fig. 10h (All the elements described in figure are generated through this study).