Loss of Nuclear Localized Parathyroid Hormone-Related Protein in Primary Breast Cancer Predicts Poor Clinical Outcome and Correlates with Suppressed Stat5 Signaling

Parathyroid hormone-related protein (PTHrP; gene name PTHLH) was discovered as a humoral factor secreted by tumors to cause the paraneoplastic syndrome of hypercalcemia of malignancy (1). Secreted PTHrP activates the G-protein–coupled receptor PTHR1 in paracrine and endocrine manner (2). PTHrP is particularly important for bone and mammary gland development (3, 4). In the embryo, epithelial PTHrP acts in a paracrine manner on the developing mammary mesenchyme, which returns critical specification signals required for normal mammary epithelial development (5). PTHrP is expressed at basal levels in nonlactating mammary epithelial cells and at elevated levels during lactation, leading to extensive maternal bone loss during lactation due to endocrine PTHrP-induced calcium mobilization into milk (6). In addition to paracrine and endocrine roles of PTHrP, intracrine signaling by PTHrP through nuclear translocation has also been documented (7). However, despite the extensive focus on PTHrP biology, the regulation of PTHrP expression and its effects in breast cancer remain poorly defined.

Recent evidence has implicated the PTHrP gene (gene symbol PTHLH) as a heritable risk factor for development of breast cancer (8, 9). A body of literature has firmly established PTHrP protein as a promoter of growth of breast cancer bone metastases, an effect attributed in part to TGFβ-driven paracrine secretion of PTHrP and subsequent osteolysis and hypercalcemia (10). High levels of PTHrP protein in primary tumors were associated with increased risk of bone metastases (11). Consistent with this, several small studies of very limited patient cohorts have reported unfavorable clinical outcome associated with high PTHrP protein expression in primary breast cancer (12–15). However, contradictory findings in a better powered study of a larger patient cohort indicated that primary breast cancers with highest PTHrP protein levels were associated with favorable outcome (16, 17). The prognostic value of PTHrP protein levels in primary breast cancer therefore remains controversial. Furthermore, current dogma holds that PTHrP protein is frequently overexpressed in malignant breast tumors (5, 18, 19), a view supported by numerous reports of frequent expression in primary breast cancers (12–18, 20, 21) and by studies indicating absence of PTHrP immunoreactivity in accompanying normal tissue (18) or in nonlactating breast epithelia (22). However, to our knowledge, there is no conclusive published evidence for elevated PTHrP protein in malignant primary breast tumors relative to normal breast epithelia.

To determine levels of PTHrP protein in breast cancer and their association with clinical outcome, we applied automated quantitative IHC to measure both nuclear localized and cytoplasmic PTHrP in archival breast cancer specimens. We now report marked reduction in nuclear and cytoplasmic PTHrP expression levels in cancer cells of malignant breast tumors compared with normal breast epithelia. Furthermore, loss of nuclear but not cytoplasmic PTHrP protein expression was associated with unfavorable prognosis in human breast cancer patients based on analyses of two large independent breast cancer patient cohorts treated at different institutions. Cancer cell levels of nuclear localized PTHrP in both breast cancer cohorts were highly positively correlated with nuclear levels of Stat5a, Stat5b, and tyrosine phosphorylated Stat5a/b, and the correlations held up across breast cancer subtypes. Stat5a/b activation by prolactin stimulated PTHrP mRNA and protein expression in human breast cancer cell lines in vitro and in vivo. Mechanistically, increased PTHrP expression could be attributed to direct binding of Stat5 within the P2 promoter region of the PTHrP gene. Collectively, we have identified PTHrP as a direct Stat5-target gene and provide novel data associating loss of nuclear PTHrP levels in primary breast cancer with unfavorable prognosis.

A breast cancer progression tissue array was generated by cutting-edge matrix-assembly (23) containing 180 unmatched patient specimens with 40 normal breast tissues, 20 ductal carcinoma in situ (DCIS), 100 invasive breast carcinomas [IBC grade 1 (IBC1; N = 20), IBC2 (N = 40), IBC3 (N = 40)], and 20 lymph node breast cancer metastases as described previously (24). This progression cohort of specimens from Thomas Jefferson University Hospital pathology archives does not have available clinical outcome data.

Primary tumor tissues from a patient outcome cohort 1, composed of 619 unselected invasive breast carcinomas, were represented in a tissue microarray YTMA-49 generated from Yale University pathology archives containing formalin-fixed, paraffin-embedded (FFPE) tissues (25). These patients were diagnosed between 1962 and 1982 with a median disease-free survival (DFS) follow-up of 8.8 years. Evaluable IHC PTHrP, Stat5a, Stat5b, and pY-Stat5a/b data were obtained from 410 of the 619 cases. From outcome cohort 2 cases, represented in a FFPE tissue microarray generated from Thomas Jefferson University Hospital pathology archives, corresponding IHC evaluable data were obtained from 387 to 540 unselected cases of invasive breast cancer. These patients were diagnosed between the years 1988 and 2000. Median recurrence-free survival (RFS) was 7.5 years. All histological breast cancer tissues were archival, de-identified specimens approved for use under waiver of consent by MCW IRB protocol PRO00028590. Supplementary Table S1 provides clinical characteristics of the outcome cohorts 1 and 2 (Supplementary Table S1). Publicly available PTHrP transcript data from breast cancer specimens with clinical outcome data of cohort 3 were from the gene expression-based Outcome for Breast cancer Online (GOBO) database (N = 737; ref. 26) and for cohort 4 from the KM-plotter database (n = 3,951; ref. 27).

IHC and AQUA analyses were performed on sections containing archival specimens in tissue array format or FFPE xenotransplant tumor tissues. IHC and AQUA analysis (HistoRx/Genoptix) for Stat5a (Advantex BioReagents, 1:8,000), Stat5b (Advantex BioReagents, 1:4,000), pY-Stat5 (Epitomics, 1:200), and PTHrP (Santa Cruz, H137, 1:200) were performed as previously described (28–30) on an autostainer (Dako). Stained slides were scanned on the PM2000 automated microscope (HistoRx/Genoptix) and fluorescent images were captured in three channels (Cy5, FITC/Alexa Fluor 488, and DAPI). AQUA scores were calculated for Stat5a, Stat5b, pY-Stat5, and PTHrP as mean signal intensity (MSI) within the epithelial cell compartment as defined by pan-cytokeratin-positive mapping and DAPI staining of cell nuclei. Signal intensities from three fluorescence channels are available at the pixel level, and MSI of each marker across cancer cell region at the whole cell level, cell cytoplasm level, or cell nucleus level are computed as previously described (30).

Analysis of association of quantitative PTHrP protein expression levels with DFS (cohort 1) and RFS (cohort 2) was conducted using recursive partitioning with 10 cross-validations in R package “rpart” (http://CRAN.R-project.org/package=rpart) to establish optimal cutpoint for dichotomization (high vs. low) of PTHrP protein levels in training cohort 1. This objective cutpoint was then validated in test cohort 2. The differences between patients with high versus low PTHrP levels were evaluated using the Kaplan–Meier estimator of the survival curves and log-rank test. Analyses of PTHrP transcript levels association with clinical outcomes in cohort 3 and cohort 4 data were likewise performed using the Kaplan–Meier estimator and log-rank test, as well as the multivariable analysis tool based on Cox regression analysis, available for cohort 3 in the online GOBO web portal (26). PTHrP transcript expression related to breast cancer molecular subtype was studied using the GOBO database. For correlation analyses, due to the presence of potential outliers, nuclear pY-Stat5, Stat5a, or Stat5b and nuclear PTHrP were evaluated using the Robust minimum covariance determinant (MCD) estimator (31). The 95% confidence intervals (CI) for these correlation coefficients were computed using 10,000 bootstrap samples. The regression lines for association between PTHrP and Stat5 markers were fitted using robust MM-estimators (32). The PTHrP protein levels in a progression array of normal, DCIS, primary invasive breast cancer (grades 1–3), and lymph node metastases were analyzed using the robust one-way analysis of variance model (33). Statistical analyses were performed in R.

T-47D and SK-BR-3 cells (ATCCs) were cultured in RPMI medium containing 10% FBS and 1 mmol/L sodium pyruvate. MDA-MB-231 cells (ATCC) and HEK-293 cells (Invitrogen) were cultured in DMEM containing 10% FBS and 1 mmol/L sodium pyruvate. The cell lines obtained from the ATCC were authenticated by short tandem repeat profiling. All cell lines were cultured for <6 months and confirmed to be mycoplasma-negative using the Universal Mycoplasma Detection Kit (ATCC). Recombinant human prolactin (AFP795) was provided by Dr. A.F. Parlow (National Hormone and Pituitary Program). For pharmacologic inhibition of MEK–Erk1/2 or Akt pathways, confluent serum-starved T-47D cells were incubated with vehicle, 10 μmol/L U0126 (Signagen), or 10 μmol/L LY294002 (Signagen) for 1 hour prior to prolactin stimulation.

qRT-PCR assays were performed with RNA isolated from cell lines using RNeasy kit (Qiagen). cDNA was generated using iScript (Bio-Rad). Both cDNA and ChIP DNA were subjected to quantitative PCR using corresponding primers (Supplementary Table S2; ref. 34).

T-47D and SK-BR-3 cell lysates were subjected to SDS-PAGE and immunoblotted with mouse pY-Stat5 (AX1, 1:10,000; Advantex BioReagents), rabbit PTHrP (H137; 1:500; Santa Cruz), rabbit Stat5a (1:3,000; Advantex BioReagents), or rabbit Stat5b (1:1,500; Advantex BioReagents) antibodies followed by secondary HRP-conjugated anti-mouse or anti-rabbit antibodies. In some instances, pY-Stat5 protein was immunoprecipitated prior to immunoblot assays using either Stat5a or Stat5b antibody (35).

Stat5a, Stat5b, dominant-negative Stat5a-Δ713 (DN-Stat5), and constitutively active Stat5a-710F (CA-Stat5a) adenovirus preparations were produced using double cesium chloride centrifugation for gene delivery into T-47D cells (4 × 10⁶/T25; MOI = 5). After 24 hours, cells were incubated with or without prolactin (10 nmol/L) in the absence of FBS for another 24 hours and subsequently harvested for qRT-PCR analysis.

Twenty nude female mice age 6 to 8 weeks were implanted with T-47D (4 × 10⁶ cells) orthotopically into the fourth paired mammary glands. Once tumors reached an average size of 500 mm³, mice were matched based on tumor size into two groups and injected with either vehicle or human prolactin (5 μg/kg) every 12 hours for 48 hours. At that point, mice were euthanized and tumors were harvested and split into two, and either snap frozen or formalin fixed for paraffin embedding.

Confluent T-47D cells serum-starved for 16 hours were treated with or without prolactin (10 nmol/L) for 1 hour and exposed to 1% formaldehyde for 5 minutes. Reactions were terminated with 0.125 M glycine. Cells were lysed in lysis buffer (50 mmol/L Tris-HCl, pH 8.0, 10 mmol/L EDTA, 1% SDS) for 1 hour and subsequently sonicated 10 times on ice. Lysates were incubated with binding buffer (0.1% SDS, 1.1% Triton-X 100, 167 mmol/L NaCl, 16.7 mmol/L Tris-HCl, pH 8.1) with a pan-Stat5 antibody (N20; Santa Cruz) overnight at 4°C, followed by capture with protein A Sepharose (GE Healthcare Life Sciences) for 1 hour. Samples were washed with binding buffer and resuspended in 100 μL of TE prior to immunoblot and qRT-PCR analyses.

PTHrP-P2 (P2) and PTHrP-P3 (P3) promoter gene constructs (pGL4-PTHrP-P2 and pGL4-PTHrP-P3) were generated by PCR using specific primers for each region (Supplementary Table S2) to amplify the P2 and P3 promoters of the PTHrP gene, digested with XhoI and EcoRV, and cloned into pGL4 vector. For PTHrP reporter assays, MDA-MB-231 and HEK-293 cells (1.5 × 10⁵) were transiently cotransfected with either β-casein-(35), P2-reporter, or P3-reporter constructs along with pcDNA3-hPRLR (36) and pXM-Stat5a. After 12 hours, cells were fed with DMEM media containing 10% horse serum for 16 hours and subsequently incubated with or without human prolactin (10 nmol/L) for 24 hours. Luciferase assays were performed 24 hours post-prolactin stimulation (BMG PolarStar Optima luminescence reader; BMG Technologies).

Motivated by the lack of published evidence to support the prevailing view of frequent elevation of PTHrP in malignant breast epithelia compared with normal breast epithelia, we quantified levels of PTHrP in a breast cancer progression tissue array containing 180 unmatched specimens. Included in the cohort were cases of normal breast tissue, ductal carcinoma in situ (DCIS), invasive breast carcinoma (IBC) grade 1 (IBC1), IBC grade 2 (IBC2), IBC grade 3 (IBC3), and lymph node metastases. PTHrP expression across this breast cancer progression series showed significant and gradual decrease from highest levels in normal epithelia to IBC1, IBC2, and IBC3 and lymph node metastasis, both nuclear localized PTHrP (P < 0.001; Fig. 1A) and cytoplasmic PTHrP (Supplementary Fig. S1). Nuclear PTHrP levels in DCIS were intermediate and were also significantly lower than those in normal breast epithelia. Generally, the intensity of PTHrP immunostaining was higher in cell nuclei than in cytoplasm of both healthy and malignant breast epithelia (Fig. 1A and B; Supplementary Fig. S1; panels i–iv). The progressive reduction in PTHrP levels across the breast cancer progression series paralleled the previously reported (30) progressive reduction in activated, nuclear localized signal transducer and activator of transcription-5 (pY-Stat5; Fig. 1A), and showed a marked positive correlation (R = 0.536, N = 92, P < 0.001; Supplementary Fig. S2) with highest levels in normal breast epithelia.

To reconcile the conflicting reports on the prognostic value of PTHrP in primary human breast cancer, we next measured cancer cell levels of PTHrP by immunofluorescence-based IHC staining and quantitative image analysis of two independent patient cohorts totaling 797 patients. Patient and tumor characteristics of clinical outcome cohorts 1 and 2 are provided in Supplementary Table S1. Cohort 1 of 410 unselected breast cancer patients with clinical outcome was used as a training set. Survival analysis using an objective, data-driven cutpoint revealed that lower tumor levels of PTHrP were associated with unfavorable clinical outcome (Fig. 1C; HR 1.37; 95% CI, 1.04–1.80; P = 0.028). To validate this observation, we applied the percentile-based cutpoint for PTHrP-positivity identified in cohort 1 (54% lowest PTHrP expressers) to an independent cohort 2 of 387 unselected breast cancer patients and confirmed based on this cutpoint an unfavorable clinical outcome in patients with the lower PTHrP-expressing tumors (Fig. 1D; HR 1.86; 95% CI, 1.16–3.00; P = 0.011). In contrast, the generally lower cytoplasmic PTHrP levels did not correlate with clinical outcome (not shown). Transcript analyses of PTHrP from the publicly available cohort 3 (N = 937; GOBO web portal; ref. 26) further supported the notion that loss of PTHrP in breast tumors is associated with unfavorable outcome (Fig. 1E). Importantly, the association between low PTHrP transcript levels and unfavorable outcome held up after multivariable adjustment for standard clinicopathological parameters (Fig. 1F). In subgroup analyses the outcome association remained significant in both ER-positive and node-negative breast cancer cases (Supplementary Fig. S3). Independent transcript analyses of PTHrP transcripts from a second publicly available cohort 4 (N = 3,951; KM-Plotter web portal; ref. 27) provided further evidence for the notion that loss of PTHrP in breast tumors is associated with unfavorable outcome (Supplementary Fig. S4).

Furthermore, PTHrP mRNA (not shown) and protein levels were both significantly higher in well-differentiated, low-grade tumors (Supplementary Fig. S5A) and were also significantly higher in ER-positive than in ER-negative breast cancer (Supplementary Fig. S5B). Collectively, quantitative IHC revealed a general reduction in PTHrP levels in primary breast cancer relative to normal breast epithelia, and tumors retaining most PTHrP were associated with favorable prognosis.

Our recent analysis of a small cohort of 92 human breast cancers indicated that total cellular PTHrP protein levels were positively correlated with levels of nuclear localized phosphorylated Stat5a/b (pY-Stat5a/b; ref. 29). However, in that preliminary study we did not examine nuclear localized PTHrP. Because breast cancer cell PTHrP is predominantly localized within the cell nuclei, we evaluated correlations between in situ protein expression of nuclear localized PTHrP and pY-Stat5a/b, as well as Stat5a and Stat5b, in the two large and independent breast cancer cohorts 1 and 2 using robust correlation analysis (32). Nuclear translocation of Stat5a/b is required for its canonical transcriptional regulatory activity (37). The analyses identified strong positive correlations between nuclear PTHrP and nuclear pY-Stat5a/b in cohort 1 [Fig. 2A, left; minimum covariance determinant (MCD) R = 0.66 (0.56–0.73), N = 359] and cohort 2 [Fig. 2B, left; MCD R = 0.56 (0.36–0.66), N = 250]. Representative images illustrate IBC cases with concordantly high (Case 1) or low (Case 2) nuclear PTHrP and nuclear pY-Stat5a/b levels (Fig. 2C). Likewise, PTHrP demonstrated significant positive correlations with both nuclear Stat5a and nuclear Stat5b in cohort 1 [Stat5a: Fig. 2A, middle, MCD R = 0.61 (0.53–0.70), N = 359; Stat5b: Fig. 2A, right, MCD R = 0.48 (0.22–0.64), N = 359] and cohort 2 [Stat5a: Fig. 2B, middle, MCD R = 0.80 (0.67–0.86), N = 250; Stat5b: Fig. 2B, right, MCD R = 0.67 (0.50–0.76), N = 250). Interestingly, nuclear PTHrP and nuclear pY-Stat5 remained positively correlated independent of breast cancer subtype, with robust correlation coefficients for Luminal A, Luminal B, Her2-positive, and triple negative breast cancer ranging from 0.52–0.67 (MCD R) or 0.49–0.71 (Spearman ρ; Supplementary Fig. S6).

Our global prolactin target gene study identified PTHrP as a candidate Stat5-target gene in the T-47D breast cancer cell line (29). This observation was unexpected based on previous data indicating that PTHrP, at least in healthy mammary glands, is being stimulated by serotonin in a prolactin-independent manner (38, 39), and a role for prolactin in PTHrP regulation has been considered permissive rather than direct (40, 41). Furthermore, PTHrP promoters are complex and a direct transcriptional role for Stat5 on PTHrP gene expression has not been established. Time course analyses determined that prolactin can effectively induce PTHrP mRNA expression in vitro in the two breast cancer cell lines, T-47D and SK-BR-3. Cells were treated with or without prolactin for 12, 24, and 48 hours and qRT-PCR analysis revealed that both T-47D and SK-BR-3 cells responded to prolactin treatment with a rapid and sustained increase in PTHrP mRNA (Fig. 3A–C). Transcripts of the established prolactin-Stat5 target gene CISH were also effectively induced by prolactin, serving as a positive control (Fig. 3A–C). Immunoblot analysis of lysates from prolactin-treated T-47D cells further revealed that prolactin-induced tyrosine-phosphorylation of Stat5a/b paralleled expression of PTHrP protein levels at both 24 and 48 hours (Fig. 3D). These initial analyses demonstrated that prolactin effectively induces both PTHrP mRNA and protein levels in human breast cancer cell lines.

Prolactin receptors signal through multiple pathways in breast cancer, including Stat5a/b, MEK–Erk1/2, and PI3K–Akt (42). We first tested whether Stat5a/b transcription factors mediated prolactin-induced PTHrP expression in breast cancer. We used adenoviral gene delivery to elevate Stat5 constructs into cultured breast cancer cells, including Stat5a, Stat5b, dominant-negative DN-Stat5, and constitutively-active CA-Stat5a (Supplementary Fig. S7). The DN-Stat5a/b protein lacks the transcriptional activation domain and, therefore, selectively binds Stat5a/b-response elements and blocks both Stat5a- and Stat5b-mediated transcription (43). In contrast, the CA-Stat5a protein contains the S710F mutation, which causes the protein to become hyperactive, even in the absence of induction by prolactin (24). T-47D cells were infected with the individual Stat5 adenoviruses, incubated with or without prolactin (10 nmol/L) for 12 hours, and tested for PTHrP protein expression by immunoblotting of cell lysates. T-47D cells contain endogenous Stat5a/b, and prolactin-induction of PTHrP protein was further elevated by adenoviral overexpression of wild-type Stat5a or Stat5b (Fig. 4A). Introduction of DN-Stat5 abolished prolactin-induced PTHrP protein, whereas introduction of CA-Stat5a induced PTHrP protein in the absence of prolactin and showed further increase in response to prolactin (Fig. 4A). Parallel transcript analyses by qRT-PCR were generally consistent with the protein data, documenting that DN-Stat5 overexpression blocked prolactin-induction of PTHrP mRNA, whereas overexpression of wild-type Stat5a, Stat5b, and CA-Stat5a enhanced prolactin-induction of PTHrP mRNA (Fig. 4B). Furthermore, the relationship between PTHrP and pY-Stat5 was confirmed by their nuclear colocalization in normal human lactating mammary gland where prolactin-induced Stat5 activation has been well documented (Fig. 4C) and in breast cancer (Fig. 4D). Colocalization provides further evidence for direct upregulation of PTHrP by Stat5, and also indicates a role for prolactin in maintaining nuclear PTHrP in lactating mammary glands, a condition under which large amounts of PTHrP are also secreted into milk and bloodstream.

Although Stat5 was required for PTHrP expression, prolactin-activated MEK–Erk1/2 or P13K–Akt signaling pathways were not required for prolactin-induced PTHrP protein expression. T-47D cells were treated with prolactin for 12 hours in the presence or absence of the MEK-inhibitor U0126 or the PI3K-inhibitor LY294002. Cells were then subjected to immunoblot analyses. Neither U0126 nor LY294002 blocked prolactin-induction of PTHrP protein expression, whereas the inhibitors effectively blocked prolactin-induced phosphorylation of Erk1/2 and Akt (Fig. 4E). Collectively, these data indicate that prolactin-induction of PTHrP expression is mediated by Stat5 and is not dependent on either Erk or Akt signaling pathways.

A T-47D xenograft model was used to validate the relationship between prolactin and PTHrP in breast cancer in vivo. Twenty nude mice were implanted with T-47D cells, and tumors were established. Ten mice were injected with vehicle, and 10 mice were injected with prolactin every 12 hours for 48 hours. qRT-PCR on tumor extracts from the 20 mice demonstrated that tumors treated with prolactin exhibited elevated levels of PTHrP transcripts (Fig. 5A). Importantly, tumors of prolactin-treated mice displayed markedly elevated levels of both nuclear localized pY-Stat5 and nuclear PTHrP protein over that in controls (Fig. 5B), thus, verifying the ability of prolactin to stimulate PTHrP mRNA and nuclear PTHrP expression in vivo.

Three alternative PTHrP transcripts AT5.2, AT6, and AT7 encode the distinct PTHrP proteins PTHrP 1-139, PTHrP 1-173, and PTHrP 1-141, respectively (Fig. 5C). We first determined which of the three transcripts are induced by prolactin in T-47D cells using transcript-specific qRT-PCR analysis. Prolactin treatment of T-47D cells selectively induced the AT6 transcript encoding PTHrP 1-173 and not AT5.2 or AT7 transcripts (Fig. 5D). We therefore explored whether the preferentially prolactin-stimulated PTHrP transcript AT6 encoding PTHrP-1-173 correlated favorably with clinical outcome in breast cancer using the publicly available data in the TCGA SpliceSeq database (44). Indeed, high levels of transcript AT6 encoding PTHrP-1-173 in primary breast cancer correlated with favorable clinical outcome (HR 1.61; 95% CI, 1.17–2.21; P = 0.004; N = 1,079), whereas transcripts AT5.2 and AT7, encoding PTHrP-1-139 and PTHrP-1-141, respectively, did not show statistically significant correlation with outcome.

We next determined whether prolactin-activated Stat5 directly regulates PTHrP promoters using chromatin immunoprecipitation (ChIP) and promoter-luciferase reporter assays. The PTHrP gene is controlled by three alternative promoters termed P1, P2, and P3 (Fig. 5C). Although the three promoters are relatively well characterized in chondrocytes (3, 4), much remains unknown as to how the individual promoters are regulated in breast cancer. For ChIP, Stat5 interaction with the PTHrP promoters was detected using promoter-specific qPCR primers following immunoprecipitation by Stat5 antibody of sheared chromatin from untreated and prolactin-stimulated T-47D cells. Stat5-ChIP analyses of PTHrP gene promoters revealed that Stat5 bound to the P2 and P3 promoter regions, but not to the P1 promoter region (Fig. 6A). Consistent with these observations, there are no consensus Stat5 binding sites in the P1 promoter, three in the P2 promoter, and one in the P3 promoter (Fig. 6B). To determine whether Stat5 activates transcription from the P2 or P3 promoters in response to prolactin, P2- or P3-luciferase reporter constructs were cotransfected with Stat5 and prolactin receptor (PRLR) expression vectors into MDA-MB-231 and HEK-293 cells, which were then treated for 48 hours with prolactin or vehicle control. The Stat5-responsive β-casein-reporter was used as a positive control. As expected, the β-casein-reporter was activated following Stat5 and PRLR coexpression plus prolactin stimulation, confirming the induction of Stat5 target genes by prolactin (Fig. 6C and D). In comparison, only the P2-luciferase reporter construct was significantly activated by Stat5 following prolactin stimulation, whereas the P3-luciferase promoter was unresponsive (Fig. 6C and D). Collectively, these data provide evidence that Stat5 directly binds to and activates the P2 promoter of the PTHrP locus in response to prolactin.

The results of this study challenge the commonly held notion that PTHrP levels are elevated in malignant breast tumors. In fact, expression levels of PTHrP protein in both invasive primary breast cancers and in lymph node metastases were markedly lower than in healthy breast epithelial cells, thus debunking a long-standing misconception that PTHrP is upregulated in malignant breast epithelium compared with the normal breast. Moreover, based on four different breast cancer patient cohorts, we found that loss of PTHrP expression was associated with poorly differentiated breast tumors and unfavorable clinical outcome. We also detected a strong positive correlation between levels of nuclear PTHrP and nuclear pY-Stat5 which may explain, at least in part, the association of low nuclear PTHrP to unfavorable clinical outcome, because loss of Stat5 expression and activation in breast cancer has consistently been associated with poor prognosis (28, 30, 45). Although PTHrP is a complex gene regulated by multiple promoters and numerous transcriptional regulators, including factors downstream of TGFβ (46, 47) and serotonin (38, 39), the new data establish a direct prolactin–Stat5–PTHrP axis in breast epithelia that frequently becomes inactivated during malignant progression. Based on our current findings, we propose that the prolactin–Stat5–PTHrP pathway is part of a broader luminal breast epithelial pro-differentiation and survival program that is frequently inactivated in de-differentiated, invasive primary breast cancer, and yet may later be reactivated in metastases to the bone (11).

In addition to the established role for breast cancer cell PTHrP as a promoter of metastasis growth in bone (3, 4, 10, 11), this study corroborates the controversial but increasingly appreciated notion that patients with breast cancer with higher cancer cell levels of PTHrP have a more favorable prognosis (16, 17). We found in two patient cohorts totaling nearly 800 patients treated at two different institutions that PTHrP protein levels were inversely associated with poorly differentiated and more aggressive breast tumors, which was further replicated at the mRNA level in cohorts of 737 patients within the GOBO dataset (26) and 3,951 patients within the KM-Plotter datasets (27). To our knowledge, this is the most comprehensive and quantitative analysis of PTHrP expression in breast cancer to date. Importantly, we also investigated for the first time the relationship of both nuclear- and cytoplasmic-localized PTHrP, of which nuclear-localized PTHrP levels were higher than cytoplasmic PTHrP and were associated with clinical outcome. However, both nuclear and cytoplasmic levels of PTHrP underwent parallel reduction during breast cancer progression. The PTHrP epitope detected by IHC is shared by all three PTHrP protein variants resulting from alternate splicing. Moreover, each of the three translated PTHrP proteins contains the tetrabasic KKKK-nuclear localization signal. Therefore, more selective antibodies will be needed to determine which PTHrP proteins and possibly protein fragments translocate to the nucleus in breast epithelia. These findings highlight a potentially distinct intracrine role of PTHrP signaling that when more fully explored may help to explain the context-dependent role(s) of PTHrP.

We recently detected PTHrP mRNA among transcripts upregulated in prolactin-stimulated breast cancer cell lines (28, 29). However, it was unknown whether PTHrP promoters were direct targets of prolactin-activated transcriptional regulators. In fact, previous reports indicated that PTHrP, at least in healthy mammary glands, is upregulated by serotonin in a prolactin-independent manner (38, 39) and that prolactin may only indirectly regulate PTHrP (40, 41). Prolactin signaling activates Stat5 in normal and malignant breast epithelia (37, 42) and PTHrP expression was strongly correlated with pY-Stat5 in our breast cancer cohorts, suggesting that Stat5 might mediate prolactin-dependent PTHrP expression via direct activation of PTHrP promoter(s). Indeed, we found in this study that activation of Stat5 by prolactin directly upregulated PTHrP protein expression via transcriptional regulation of one of the three PTHrP promoters in human cancer cell lines in vitro and in vivo. Prolactin-activated Stat5 physically bound to both the P2 and the P3 promoters of the PTHrP gene, but not to the P1 promoter. Unexpectedly, however, only the P2 promoter was activated by prolactin-mediated Stat5 signaling, whereas the P3 promoter was unaffected. The preferential activation of the P2 promoter of PTHrP is consistent with the presence of three consensus Stat5 binding sites, whereas the P3 contains only one and may require transcriptional co-activators, such as those downstream of serotonin (38, 39) or TGFβ (46, 47). Future work will explore the possibility that P3 promoter is also engaged by prolactin under certain conditions, for example, during maximal Stat5 activation at the time of lactation. The possible interaction between prolactin-Stat5 and serotonin signaling as drivers of PTHrP expression will also be subject of future studies.

This study has identified a novel prolactin–Stat5–PTHrP signaling axis, which we postulate provides critical pro-differentiation cues that are required for maintenance of normal mammary epithelial function and physiology. Our analyses are the first to systematically compare normal and malignant breast epithelia, revealing an unexpected gradual loss of epithelial PTHrP expression during progression from normal to in situ carcinoma to invasive primary breast cancer, which correlated with downregulation of Stat5 activation and with poorly differentiated and more aggressive primary tumors. Our data on PTHrP and clinical outcome are correlative, and future work will be required to determine whether loss of nuclear PTHrP causes more aggressive tumor behavior. Nonetheless, our findings agree with substantial evidence in several reports that PTHrP suppresses growth, invasion, and metastasis of established breast cancer (16, 17), which when taken at face value appears to contradict alternative evidence that PTHrP promotes cancer cell proliferation, tumorigenesis, and metastatic growth (2, 4, 10). These apparent discrepancies may be reconciled by differing context-dependent roles of PTHrP signaling. PTHrP is, on one hand, critical for specification of well-differentiated mammary epithelial cells (5). On the other hand, in poorly differentiated malignant breast epithelial cells that have metastasized to bone, reactivation of PTHrP expression, for instance by local TGFβ and possibly circulating prolactin, will be advantageous for expansion of osteolytic metastases. Collectively, we have provided evidence for PTHrP as a novel Stat5 target gene, which will facilitate understanding of the role of prolactin–Stat5 signaling in breast biology and breast cancer. Our new observations further overturn the prevailing notion that PTHrP is upregulated in primary breast cancer and identifies a prolactin–Stat5–PTHrP signaling axis that is progressively lost in more aggressive tumors, thus reconciling several controversial concepts related to PTHrP and breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: T.H. Tran, F.E. Utama, A.R. Peck, H. Rui

Development of methodology: T.H. Tran, F.E. Utama, A.R. Peck, Y. Sun, C. Liu, M.A. Girondo, J.P. Palazzo, H. Rui

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.H. Tran, F.E. Utama, T. Sato, A.R. Peck, Y. Sun, C. Liu, M.A. Girondo, C.D. Shriver, H. Hu, J.P. Palazzo, E.P. Mitchell, H. Rui

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.H. Tran, F.E. Utama, T. Sato, A.R. Peck, J.F. Langenheim, S.S. Udhane, Y. Sun, C. Liu, H. Hu, J.P. Palazzo, P.W. Auer, M.J. Flister, T. Hyslop, E.P. Mitchell, I. Chervoneva, H. Rui

Writing, review, and/or revision of the manuscript: T.H. Tran, F.E. Utama, A.R. Peck, J.F. Langenheim, S.S. Udhane, Y. Sun, A.J. Kovatich, J.A. Hooke, C.D. Shriver, H. Hu, J.P. Palazzo, M. Bibbo, P.W. Auer, M.J. Flister, T. Hyslop, E.P. Mitchell, I. Chervoneva, H. Rui

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.H. Tran, F.E. Utama, C.D. Shriver, E.P. Mitchell,

Study supervision: T.H. Tran, H. Rui

We thank Dr. David L. Rimm at the Department of Pathology, Yale University School of Medicine, New Haven, CT, for generously providing the cohort 2 tissue microarray and associated data for this study. The work was supported by Susan G. Komen for the Cure Promise grant KG091116, US National Institutes of Health grants CA188575 and CA185918, Worldwide Cancer Research grant WCR 14-0272, a grant from the Dr. Nancy Laning-Sobczak Fund for Breast Cancer awarded by the Medical College of Wisconsin Cancer Center, Institutional Research Grant #14-247-29 from the American Cancer Society and the MCW Cancer Center (to Y. Sun), and a postdoctoral fellowship grant P2BEP3_168705 from the Swiss National Science Foundation (to S.S. Udhane). The views expressed in this article are those of the authors and do not reflect the official policy of the Department of the Army/Navy/Air Force, Department of Defense (DOD), or US Government.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.