Background

Prolactin (PRL) is a pituitary hormone that is critical for normal mammary gland development by promoting proliferative expansion of the secretory alveolar cell compartment during pregnancy and for terminal differentiation of these milk-producing cells during lactation. Prolactin is also strongly implicated in breast cancer. On one hand, accumulating evidence suggests that PRL promotes breast cancer initiation and growth. In vivo, PRL over-expressing transgenic mice have an increased incidence of mammary tumors [1], while PRL knock-out mice have a reduced incidence of mammary tumors [2]. In women, elevated PRL is associated with increased risk of ER-positive breast cancer [3, 4]. Furthermore, 70-95% of human breast cancers express PRL receptor (PRLR) [5, 6], and many breast cancer cell lines express high levels of PRLR with evidence of proliferative or survival responses to PRL in vitro[7, 8]. PRL was also able to enhance 17β-Estradiol (E2) dependent proliferation of breast cancer cells [9–11]. On the other hand, evidence suggests that PRL acts to preserve cellular differentiation of breast cancer. Stat5 transcription factors, principal mediators of PRL effects [12, 13], are frequently inactivated during metastatic progression in clinical breast cancer specimens, and loss of Stat5 signaling is associated with unfavorable prognosis and increased risk of anti-estrogen therapy failure [14–17]. In experimental breast cancer models, activation of Stat5 increased cell surface E-cadherin expression, induced homotypic cell clustering, and reduced invasion through Matrigel [18, 19]. Restoration of PRL-Stat5 signaling in the mesenchymal-like MDA-MB-231 cells reverted their invasive phenotype, while blocking autocrine PRL signaling in the epithelial T47D cell line was associated with EMT and enhanced invasive properties [20]. In T47D cells, PRL also blocked progestin-induction of a tumor-initiating CK5-positive cell population through a mechanism that involved PRL-suppression of progestin-induced BCL6 [21].

Due to the importance of PRL in breast cancer growth and differentiation, identifying genes regulated by the PRL-Stat5 pathway may provide new insights into the pleiotropic effects of PRL in breast cancer. Several studies have identified genes regulated by PRL in the normal mouse mammary gland [22–26], but only a limited number of studies have been carried out in human breast cancer cells. More importantly, global profiling for PRL-modulated gene expression in human breast cancer in vivo has not been reported. One in vitro study identified genes regulated by PRL, E2, and PRL + E2 in cultured ER-positive T47D cells using genome-wide transcript profiling [9], while a second in vitro study used subtractive hybridization to identify PRL-regulated genes in the ER-negative, Her2-overexpressing SKBR3 cell line [27]. However, neither study confirmed whether the identified transcripts were regulated by Stat5 or remained PRL-modulated in vivo. Work from our laboratory used selective overexpression of either Stat5a or Stat5b in the ER-positive MCF-7 human breast cancer cell line in vitro followed by PRL exposure to explore differences in Stat5a and Stat5b regulated transcripts on the Affymetrix platform, but this study also failed to confirm any of the modulated transcripts by qRT-PCR or validate the data in vivo[17]. An additional very recent study also examined the MCF-7 model and reported PRL regulated genes, as well as genes uniquely modulated through activation of the PRLR transactivation domain [28], but the investigators did not examine whether the identified transcripts were regulated by Stat5 or remained PRL-modulated in vivo.

Based on observed differences in hormone-modulated transcriptional programs in human cancer cells in vitro and in vivo[29], PRL-modified transcripts identified in an in vivo environment are expected to be more clinically relevant than transcripts modulated in cells cultured on plastic. The present study reports a panel of 130 PRL regulated transcripts in the human T47D breast cancer xenotransplant model in estrogenized nude mice. T47D xenografts were established in nude mice and mice were exposed to human PRL or saline for 48 h before RNA isolation from tumor extracts. Of the 130 transcripts, 75 were up-regulated and 55 were down-regulated. Modulated transcripts were identified based on fold change (>1.6) and P-value (<0.05). From this initial transcript set, 18 transcripts were selected based on known breast cancer relevance for validation by qRT-PCR. Many but not all of the in vivo validated transcripts were PRL-modulated in vitro. We also determined the dependence of PRL-modulated transcripts on Stat5, Jak1 or Jak2, and whether individual modulated genes are co-regulated by PRL and E2. One of the PRL-modulated genes identified, parathyroid hormone-related peptide (PTHrP), was found by fluorescence-based quantitative immunohistochemistry to positively correlate with levels of nuclear localized, tyrosine phosphorylated Stat5 (Nuc-pYStat5) at the protein level in clinical human breast cancer specimens. Gene Ontology (GO) analysis of PRL-upregulated genes demonstrated enrichment in differentiation pathways. Finally, a gene signature based on PRL-upregulated genes was associated with prolonged relapse-free and metastasis-free survival in human breast cancer patients. Studies are ongoing to determine how modulation of these genes, including PTHrP, may mediate PRL effects in breast cancer.

Results

Global transcript profiling reveals novel PRL-modulated genes in human T47D breast cancer xenotransplants

Global gene expression analysis was performed using the hormone receptor positive T47D xenotransplant model. Tumor-bearing nude mice received either human prolactin (N = 10) or vehicle (N = 10) subcutaneously every 12 h for 48 h. Representative images of T47D xenograft tissues display robust tyrosine phosphorylated Stat5 (pY-Stat5) staining in response to PRL (Figure 1A). For the microarray analyses, RNA isolated from individual tumors was pooled into 3 groups from PRL injected mice and 3 groups from control mice, with each group containing RNA from tumors of 3 or 4 mice. Global profiling on the Affymetrix U133 platform identified 75 upregulated transcripts (Table 1) and 55 down-regulated transcripts (Table 2) based on P-values (<0.05) and fold difference (>1.6). From this panel, 18 transcripts were selected for further analysis based on established breast cancer relevance (Table 3). CISH was induced by PRL and included in subsequent analyses as a positive control, since CISH is an established STAT5 target gene [30]. Based on the microarray data of this panel of 18 transcripts, 15 were upregulated and three were down-regulated 1.6-fold.

Figure 1
figure 1

qRT-PCR validation of 18 candidate prolactin modulated transcripts. (A) Representative images of T47D xenografts treated with or without PRL for 48 h and stained with tyrosine phosphorylated Stat5 (pY-Stat5)(red) or DAPI (blue) using immunofluorescence. (B) RNA isolated from T47D xenografts were tested for 18 breast cancer relevant genes using qRT-PCR analysis. Bars represent an average of fold change from 3 independent xenograft RNA samples. (C) Immunoblot of pY-Stat5 and Stat5 in T47D cells grown in vitro and placed in serum starvation media for 24 h before treatment with vehicle or 10 nM PRL for 24 h. (D) RNA isolated from T47D cells grown in vitro were tested for the same 18 breast cancer relevant genes using qRT-PCR analysis. Bars represent an average of fold change from 3 independent experiments.

Full size image
Table 1 Upregulated genes from microarray with fold change >1.6 and p < 0.05
Full size table
Table 2 Downregulated genes from microarray with fold change > −1.6 and p < 0.05
Full size table
Table 3 18 genes chosen for further study based on breast cancer relevance
Full size table

qRT-PCR validation of 18 candidate prolactin modulated genes

To validate the data from the global microarray analysis, qRT-PCR analysis was carried out on the 18 selected genes using the same xenograft RNA samples that were used for the microarray-based profiling. Of the upregulated transcripts, 12 out of 15 transcripts were confirmed to be upregulated by qRT-PCR >1.6-fold, whereas upregulation of GADD45A, ERAP1, FGFR2 was not confirmed by qRT-PCR. Among the selected three down-regulated transcripts, BCL6 was confirmed down-regulated >1.6 fold by qRT-PCR, whereas FLT4 and SOX4 were down-regulated 1.3-fold and 1.6-fold, respectively, by qRT-PCR analysis (Figure 1B).

We then identified PRL-modulated transcripts in T47D xenotransplants that were also modulated by PRL T47D cells in vitro. Immunoblot analysis of T47D cells cultured in vitro displayed robust pY-Stat5 in response to 10 nM human PRL for 24 h (Figure 1C). qRT-PCR analysis of mRNA extracted from T47D cells treated with or without PRL for 24 h established that 9 of the 12 qRT-PCR validated upregulated transcripts also were upregulated over 1.6-fold in vitro, while three transcripts, CCR6, AMIGO2, and DUSP4 were not modulated by PRL in vitro (Figure 1D). Transcripts for GADD45A, ERAP1, FGFR2, which were not confirmed by qRT-PCR analysis of transcripts in vivo, remained unmodulated in vitro (Figure 1D). Out of the three selected down-regulated transcripts, BCL6 maintained PRL induced down-regulation of >1.6-fold in vitro, whereas FLT4 and SOX4 again were reproducibly down-regulated but only 1.5-fold and 1.4-fold respectively (p < 0.05) (Figure 1D). Since these values were only marginally lower than 1.6, we elected to keep FLT4 and SOX4 in the subsequent analyses. In subsequent experiments we focused on the set of 12 transcripts out of original selected panel of 18 transcripts that were PRL-responsive both in vivo and in vitro.

Stat5 regulates novel PRL-modulated transcripts

PRL activates multiple signaling pathways in breast cancer cells [31], with Stat5 constituting a principal mediator of PRL actions during development and differentiation of the mammary gland [32]. To determine whether the observed PRL-modulated transcripts were regulated through the Stat5 pathway, we overexpressed Stat5a, Stat5b, or a dominant-negative Stat5a/b (DNStat5) in T47D cells using adenoviral gene delivery, and treated cells with or without PRL for 24 h. Immunoblot analysis of total cell lysates verified over-expression of the correct Stat5 variants and robust PRL-induction of pY-Stat5 in cells following adenoviral gene delivery (Figure 2A). The overexpression was sufficiently high that basal levels of Stat5 were not detectable without further exposure of the blots (not shown). qRT-PCR analysis revealed that DNStat5 blocked PRL induction of each of the 9 upregulated transcripts. PRL induction of most of these 9 transcripts were further enhanced by overexpression of Stat5a or Stat5b, either basal or PRL-induced (Figure 2B). Although levels of PRL-induced CISH and WT1 transcripts did not rise further by Stat5 overexpression, PRL-induction of both transcripts was effectively abrogated by DNStat5, thus supporting Stat5-mediated regulation of these genes. Stat5 overexpression also enhanced down-regulation of both SOX4 and FLT4. The ERAP1 transcript was included as a negative control since it was not regulated by PRL as shown in Figure 1. Stat5a, Stat5b, or DN-Stat5 overexpression had no effect on ERAP1 transcript levels, indicating that enhancement of PRL-induction detected for the other transcripts is specific (Figure 2). BCL6 was omitted from this analysis since we have published this separately [33].

Figure 2
figure 2

Stat5 regulates novel prolactin modulated transcripts. (A) Immunoblot of pY-Stat5 and Stat5 in cells treated with adenovirus containing LacZ (Control), Stat5a, Stat5b, or Dominant-Negative Stat5 (DNStat5) for 24 h, then serum starved and treated with vehicle or 10 nM PRL for 24 h. (B) RNA isolated from T47D cells treated with adenovirus was tested for 12 PRL modulated transcripts using qRT-PCR analysis.

Full size image

Jak2 and Jak1 are critical for PRL gene regulation in T47D breast cancer cells

Conditional gene knock-out in mice demonstrated that Jak2 is the key Stat5 tyrosine kinase in breast epithelial cells during and outside of pregnancy and lactation [34]. However, we have reported that in human breast cancer cell lines Jak1 is also recruited in a Jak2-dependent manner for maximal PRL-activation of Stat5 and other signaling mediators [31]. To determine whether PRL-recruitment of Jak1 was required for maximal modulation of PRL-regulated transcripts, T47D cells were infected with lentivirus carrying shRNAs targeting either Jak1 or Jak2 followed by treatment with or without PRL for 24 h. Jak1 shRNA was effective and selective, as judged by marked down-regulation of Jak1 mRNA but not Jak2 mRNA (Figure 3). Conversely, Jak2 shRNA effectively suppressed Jak2 mRNA but not Jak1 mRNA (Figure 3). All nine PRL-upregulated transcripts showed complete dependence on Jak2 (Figure 3). Importantly, all nine PRL-induced transcripts were also partially suppressed by Jak1 knockdown, consistent with a significant role for Jak1 recruitment by PRL in breast cancer cells to maximize downstream signals. Among the three down-regulated transcripts, BCL6 down-regulation by PRL was dependent on Jak2 but not on Jak1, whereas down-regulation of FLT4 and SOX4 by PRL was not significant under these experimental conditions, possibly due to cell stress during lentiviral infection (Figure 3). Collectively, these data are consistent with a general model in which Jak2 is critical for PRL receptor signaling and Jak1 recruitment is needed for maximal signal.

Figure 3
figure 3

Jak2 and Jak1 are critical for PRL transcript regulation in T47D breast cancer cells. T47D cells were treated with lentivirus containing NTC, Jak1 shRNA, or Jak2 shRNA overnight. Then, cells were placed in serum starvation media for 24 h and then treated with vehicle or 10 nM PRL for another 24 h before RNA isolation and qRT-PCR analysis of the 12 PRL regulated transcripts.

Full size image

Synergistic enhancement by E2 of PRL-induced proliferation and select PRL-modulated genes

Previous studies have indicated that PRL may enhance E2-induced proliferation as well as additively or synergistically increase transcription of certain PRL or E2 target genes [9–11]. To verify that the PRL-E2 interaction occurred in T47D cells under our culture conditions, we treated T47D cells with varying doses of PRL (0, 1, 10, 20, 37, and 100 nM) while keeping E2 constant (1 nM), and determined cell number after 72 h. PRL concentrations as low as 10 nM in the presence of E2 were associated with an increase in cell number compared to E2 alone, and this effect was maintained at higher PRL concentrations (Figure 4A). Next, we maintained constant PRL concentration (20 nM) while varying the E2 dose (0.001, 0.01, 0.1, 1, 10 nM), and counted cell numbers after 72 h. At every concentration of E2, we observed a PRL-induced increase in cell number (Figure 4B). We also determined the effect of PRL on E2-driven growth in soft agar and measured colony size after 2 weeks. While PRL alone had limited effect on colony size, E2 increased colony size dramatically, and PRL further increased E2-driven colony size (Figure 4C). Representative images of the colonies in soft agar are displayed in Figure 4D. These data established under our culture conditions that PRL positively interacts with E2 to induce proliferation of luminal T47D breast cancer cells.

Figure 4
figure 4

Enhancement by E2 of PRL-induced proliferation and select PRL modulated transcripts. (A) T47D cells were treated with varying doses of PRL (0, 1, 10, 20, 37, and 100 nM) in the presence of constant E2 (1 nM) for 72 h. Bars represent triplicates of each condition that were counted and averaged. (B) T47D cells were treated with varying doses of E2 (0.001, 0.01, 0.1, 1, 10 nM) in the presence of constant PRL (20 nM) for 72 h. Bars represent triplicates of each condition that were counted and averaged. (C) T47D cells were treated with vehicle, 20 nM PRL, 1 nM E2, or 20 nM PRL + 1 nM E2 for 2 weeks. Images from four independent wells were analyzed for colony size through ImageJ. (D) Representative images of T47D cells grown on soft agar and treated with vehicle, PRL, E2, or PRL + E2 for 2 weeks. (E) T47D cells were treated with PRL, E2, or PRL + E2 for 24 h. RNA was isolated and qRT-PCR analysis was performed on the 12 PRL modulated transcripts.

Full size image

To determine whether the PRL-modulated transcript panel was affected by co-treatment with E2, we treated T47D cells with vehicle, PRL, E2, or PRL + E2 for 24 h and performed qRT-PCR analysis of the 12 transcripts. Seven out of the nine upregulated transcripts displayed further induction with E2 present (NTN4 and TMC5 had no further induction) (Figure 4E). None of the three PRL-downregulated genes displayed further downregulation in the presence of E2 (Figure 4E). This data indicates that E2 is selectively modulating some but not all PRL-Stat5 regulated transcripts, and these specific transcripts may be playing a role in PRL’s ability to increase E2-driven breast cancer cell proliferation.

PTHrP protein levels correlate with levels of pYStat5 in human breast cancer tissues

To begin to assess the clinical relevance of the newly identified PRL-modulated transcripts, we selected the gene product most strongly upregulated by PRL in the in vivo xenotransplant tumors, PTHrP, for protein expression analysis in clinical breast cancer specimens. We hypothesized that cellular PTHrP protein levels would be positively correlated with levels of nuclear localized and tyrosine phosphorylated Stat5 (Nuc-pYStat5). Nuc-pYStat5 is an indirect measure of PRL receptor activation in breast epithelia, and we documented evidence of Stat5-dependence of PRL-upregulation of PTHrP in T47D cells. Levels of cellular PTHrP and Nuc-pYStat5 were analyzed using fluorescence-based quantitative immunohistochemistry on a breast cancer progression array containing 40 normal and 140 malignant breast tissues. Representative images of PTHrP and pY-Stat5 staining are shown in Figure 5A, where Case 1 has high PTHrP and Nuc-pYStat5 levels, while Case 2 has low PTHrP and Nuc-pYStat5 levels. Evaluable levels of cellular PTHrP and Nuc-pYStat5 were obtained in 92 breast cancer specimens and in support of our hypothesis, a statistically significant positive correlation was detected (r = 0.51, P < 0.001) (Figure 5B).

Figure 5
figure 5

PTHrP protein levels correlate with levels of nuc-pYStat5 in human breast cancer tissues. (A) Representative immunofluroescent images of two cases of breast cancer stained for pY-Stat5 (Red), PTHrP (Red), cytokeratin (Green), and DNA (Blue) using immunohistochemistry. (B) Scatter plot and Pearson correlation analysis between nuc-pYStat5 and cellular PTHrP protein levels on 92 breast cancer specimens.

Full size image

Gene ontology (GO) terms are enriched in differentiation markers

Analysis of GO biological process terms using the 75 PRL-upregulated transcripts identified in this study revealed 24 GO terms that had a false discovery rate (FDR) below 25% (Table 4). Many of the pathways that were identified were homeostasis-related, correlating with the ability of the PRL-Stat5 pathway to maintain cellular differentiation, consistent with established pro-differentiation effects of PRL on normal and malignant luminal breast epithelial cells. In addition, proliferation-regulation and negative regulation of apoptosis were also identified, which is consistent with the reported role of PRL contributing to breast cancer initiation and growth.

Table 4 Gene Ontology (GO) analysis of PRL upregulated genes with false discovery rate (FDR) < 25
Full size table

Prolactin-upregulated gene signature is associated with prolonged time to relapse and metastasis

We then determined whether the PRL-upregulated gene signature was associated with clinical outcome, using an available 49 of 75 PRL-upregulated genes (Table 5) in a cohort of 936 primary invasive breast cancer patients. We divided the patients into three groups based on their degree of expression of the PRL-gene signature (upper quartile, interquartile range, and lower quartile). The patients in the upper quartile had significantly prolonged time to metastasis compared to patients associated with the lower quartile (Figure 6A). In addition, patients that were in the upper quartile also had significantly prolonged disease-free survival than patients in the lower quartile (Figure 6B).

Table 5 PRL-upregulated genes used to generate PRL gene signature
Full size table
Figure 6
figure 6

PRL upregulated gene signature is associated with metastasis-free and relapse-free survival. (A) Patients with high levels of a genes signature of PRL-upregulated had reduced risk of developing metastasis compared to patients with low levels of the PRL-upregulated gene signature. (B) Patients with high levels of the PRL-induced gene signature had prolonged relapse-free survival compared to patients with low levels of the PRL gene signature.

Full size image

Discussion

The present study reports a novel panel of PRL-modulated transcripts based on analysis of human breast cancer xenograft tumors in vivo. We identified 75 up-regulated and 55 down-regulated transcripts that were significantly modulated based on at least a 1.6-fold change with a P-value less than 0.05. From this panel of 130 PRL-modulated transcripts, a subset of 18 transcripts with established breast cancer-relevance was selected for further analysis and validation. Validation by qRT-PCR documented significant modulation of 12 of 18 transcripts in vitro. We further documented that the majority of in vitro-modulated transcripts were Stat5- and Jak2-dependent, and showed that Jak1 was required for maximal PRL-modulation. Consistent with PRL-enhancement of E2-driven proliferation of breast cancer cells, select PRL-modulated transcripts displayed positive co-regulation by E2, including the growth factors EREG and AREG. These molecular results are summarized in Table 6. Furthermore, quantitative immunofluorescence analyses of clinical breast cancer specimens from a cohort of 92 patients documented a significant positive correlation between tumor levels of PTHrP protein, one of the PRL-stimulated transcripts, and nuclear localized and tyrosine phosphorylated Stat5, a marker of PRL signaling. Gene ontology analysis revealed that prolactin-upregulated genes were associated most frequently with terms involved in homeostasis and differentiation. Finally, a gene signature generated with PRL-upregulated genes was associated with prolonged relapse-free survival as well as metastasis-free survival in a cohort of breast cancer patients. Collectively, the validation studies support the value of the transcript data and are expected to facilitate better understanding of PRL action in breast cancer.

Table 6 Summary of transcript regulation
Full size table

PRL activates both Stat5a and Stat5b, which have 92% amino acid similarity [35], but are encoded by different genes and may mediate overlapping and distinct effects in breast cancer cells [17, 33, 36–38]. In the present study, which focused on PRL-modulated transcripts in the T47D breast cancer model, experimental overexpression of Stat5a or Stat5b enhanced to a comparable extent PRL-modulation of most transcripts tested. However, TMC5, NTN4, and AREG displayed greater degree of enhanced PRL-modulation when Stat5a was overexpressed rather than Stat5b, supporting the notion that certain genes are more responsive to Stat5a than Stat5b. PRL-modulation of all examined transcripts with the exception of PRL-induced down-regulation of FLT4 mRNA was disrupted by overexpression of a C-terminally truncated Stat5 variant that acts as a dominant-negative molecule for both Stat5a and Stat5b. This exception is consistent with previous reports that repression of certain Stat5 target genes is unaffected by alterations in the Stat5 transactivation domain [33, 39]. We have recently reported that Stat5a but not Stat5b expression was lost during progression of human breast cancer, and in cultured MCF-7 cells there was only a limited overlap in transcripts modulated by the two PRL-activated transcription factors [17]. It is possible that in T47D cells the higher basal levels of Stat5a and especially Stat5b make it more difficult to ascertain differences between the two transcription factors in overexpression studies. Future transcript analyses will focus on further characterizing the differences in gene regulation between Stat5a and Stat5b in human breast cancer.

Jak1 was reported to be activated by PRL signaling in human breast cancer lines and cooperate with Jak2 to enhance signaling pathways downstream of PRL receptors, including Stat5, Stat3 and Erk activation [31]. The present study validates the notion that Jak1-coactivation enhances PRL-Jak2 signaling by demonstrating that maximal PRL-modulation of target transcripts required PRL-induced co-activation of Jak1. Future studies to identify the mechanism of activation of Jak1 by PRL receptors in breast cancer and the effect of Jak1-activation on PRL-modulated biology of breast cancer are now warranted.

Furthermore, whereas PRL alone exerted limited proliferative effect on T47D breast cancer cells in vitro, PRL enhanced E2-driven cell proliferation both on plastic and soft agar. PRL positively interacted with E2 to further elevate several transcripts encoding growth and progression factors for breast cancer, including AREG, EREG, PTHrP and WT1. Considering the established role for AREG as a paracrine mediator of E2-induced proliferation of luminal breast epithelial cells during pubertal growth [40], AREG may be directly involved in PRL stimulation of E2-driven growth of human breast cancer. A recent study has implicated PRL receptors in the maintenance of ER expression and responsiveness of breast cancer cells to estrogen [28], which is consistent with our findings which demonstrate significant crosstalk between the two pathways. Since Stat5 has been associated with response to anti-estrogen therapy in clinical breast cancer specimens [16, 17], this synergistic stimulation of proliferation with estrogen may be a mechanism behind increased responsiveness to anti-estrogen treatments. In addition, PRL biological action may vary depending on the hormonal environment, especially given our recent observation that PRL effectively counteracts progestin-induction of the tumor-initiating CK5-positive cell population [21].

We identified PRL-suppression of BCL6 transcript and protein based on this global transcript analysis of T47D xenotransplants tumors, and we reported a negative correlation between levels of BCL6 protein and Nuc-pYStat5 in clinical breast cancer specimens [33]. The fact that the protein products of two of the PRL-modulated genes identified in this global transcript profiling, PTHrP and BCL6, both correlated with Nuc-pYStat5 in human clinical breast cancer specimens indicates that transcripts identified in the present study are relevant in clinical specimens and may become useful human breast tumor markers of PRL activation.

Supporting the validity of the identified panel of PRL-modulated transcripts in T47D cells in vivo, 12 out of the 57 transcripts identified as PRL-modulated transcripts in a recent in vitro transcript profiling study of T47D cells overlapped with our panel (AREG, WT1, PTHrP, IER3, TMC5, CISH, BCL6, DUSP4, TNIK, EGLN3, FBXO32, and AKAP9) [9]. Only WT1 and IER3 transcripts were tested and confirmed by qRT-PCR in the previous report. In addition, select transcripts identified in the previous study such as WT1 and IER3 demonstrated co-regulation by E2 [9], consistent with the findings of the present study. In vitro transcript profiling for PRL-modulated transcripts in another ER-positive human breast cancer cell line, MCF7, using Stat5 overexpression to enhance PRL effects, yielded 300 candidate PRL-modulated transcripts among which 12 overlapped with the panel identified in the present study of T47D cells (CISH, EGLN3, KCNK6, PTHrP, FN1, CHSY1, BCL6, DUSP4, IGFBP5, TNIK, ABCC4, and MYC) [17]. The limited overlap between our current study of T47D cells and two recent studies of MCF-7 cells in vitro may be due to lower expression levels of Stat5 in MCF-7 cells, making it necessary to overexpress Stat5 [17] or PRL receptor [28]. While previous studies were performed in vitro and did not broadly validate identified candidate PRL-modulated transcripts by qRT-PCR, the present study also provides novel data by demonstrating PRL-regulation of transcripts in vivo and through Jak-Stat5 dependent mechanisms.

The established breast cancer relevance for AREG, WT1, and IER3 is discussed in a previous transcript profiling study [9]. NTN4 is a transmembrane protein whose expression levels positively correlates with better prognosis in breast cancer [41]. TMC5 is a transmembrane channel that was overexpressed in PIK3CA-mutated breast cancer [42]. FN1 is a protein present in the extracellular matrix that is a candidate serum biomarker for detecting breast cancer [43], and disrupting the interaction between FN1 and integrins in breast cancer cells led to increased apoptosis and response to radiation [44]. EREG is a ligand for EGFR and Her4, and was reported to be part of a set of four genes that promote breast cancer intravasation and metastasis to the lung [45]. PTHrP is a secreted protein critical for mammary gland development [46], and extensive research has been performed on its role in mediating breast cancer metastasis to bone [47–49]. SOX4 is a transcription factor regulated by progesterone in breast cancer cells [50], and was identified as an oncogene in prostate cancer [51]. FLT4 is a member of the VEGF receptor family, and expression in vessels surrounding breast tumors was correlated to lymph node positivity and poor clinical outcome [52]. These genes clearly demonstrate the complexity of PRL effects, since PRL has the ability to suppress oncogenes (SOX4, FLT4, BCL6) and upregulate favorable prognostic markers (NTN4), while also upregulating genes involved in breast cancer growth and progression (AREG, WT1, IER3, EREG, TMC5, FN1, PTHrP). These observations are consistent with the many reported conflicting and likely context-dependent effects of PRL in breast cancer.

Furthermore, gene ontology (GO) analysis based on PRL-upregulated genes demonstrated a concentration in homeostasis pathways, consistent with the known pro-differentiation role of PRL in breast cancer. However, other terms such as cell proliferation and anti-apoptosis were also enriched, most likely reflecting the duality of PRL action. Consistent with the enrichment of differentiation terms in the GO analysis, a PRL gene signature generated on PRL-upregulated genes was associated with prolonged time to relapse and metastasis-free survival. These associations are consistent with the multiple reports that Stat5 is associated with favorable prognosis in breast cancer patients [14–16]. In addition, we have recently reported that prolactin can suppress a therapy-resistant, tumor-initiating CK5-positive population induced by progestin [21]. Our results are consistent with a role of PRL in reducing the tumor-initiating CK5-positive cell population, which is implicated in breast cancer metastasis and relapse.

Conclusions

The present study is the first to report a panel of PRL-modulated transcripts based on global transcript profiling of human breast cancer xenotransplant tumors in vivo. Some but not all transcripts were also modulated by PRL in vitro. PRL-enhancement of E2-driven proliferation of T47D cells in vitro may be mediated by observed co-regulation by PRL and E2 of growth-promoting genes including AREG, EREG, WT1 and PTHrP. PRL-modulated transcripts reported in this study are expected to facilitate deciphering of the mechanisms underlying the pleotropic effects of PRL on breast cancer. PRL-upregulated genes were frequently associated with differentiation pathways. Finally, select transcripts or their protein complement identified in this study also may be useful as breast cancer tumor marker signatures of PRL activation, which is highly relevant considering our report that PRL gene signature is associated with relapse-free survival and metastasis-free survival, and the already documented association of Stat5 with breast cancer prognosis and hormone therapy responsiveness.

Materials and methods

Tissue culture

T47D cells were cultured in RPMI (Cellgro) containing 10% fetal bovine serum (FBS) and 1 mM sodium pyruvate (Cellgro). For PRL induction, confluent T47D cells were put in serum starvation media (RPMI without FBS) for 24 h, and then stimulated with either PBS vehicle or 10 nM of recombinant human prolactin (AFP795, provided by Dr. A.F. Parlow at National Hormone and Pituitary Program) for 24 h. 24 h prior to experiments involving β-Estradiol (Sigma), media was changed to RPMI containing 5% Charcoal Stripped Serum (Thermo Scientific) and 1 mM sodium pyruvate.

Xenotransplants

Nude mice (N = 20) were implanted with 17β-estradiol pellets (0.72 mg; Innovative Research of America) and injected subcutaneously with T47D cells (5 × 106) suspended in Matrigel into two dorsolateral sites. Tumors were allowed to grow for 6 weeks and subsequently were injected subcutaneously with either saline (N = 10) or 5 μg/g body mass of human prolactin (N = 10) every 12 h for 48 h total. Tumors were harvested and processed for immunohistochemistry and qRT-PCR. All research involving mice were approved by Thomas Jefferson IACUC (protocol 789C to H.R.) in accordance with international guidelines for ethical treatment of animals.

Microarray

RNA was pooled into 3 groups from PRL injected mice and 3 groups from saline injected mice, with each group containing RNA from 3–4 mice. Microarray analysis was performed for each group using the Affymetrix HG-U133 GeneChip Set (Expression Analysis). Two group comparison analyses were conducted on normalized expression values that were individually transformed using base 2 logarithms. On the log-transformed scale, the mean is calculated for every gene within each group and a two-sample, two-sided t-test is conducted to test the equality of those means. The P-value indicates the significance level of this test.

Quantitative reverse transcription polymerase chain reaction

Quantitative RT-PCR assays were performed with RNA samples isolated from T47D cells using RNeasy kit (Qiagen). cDNA was generated using Iscript (Bio-Rad) according to the manufacturer’s protocol. cDNAs were subjected to qPCR using corresponding primers (Table 7).

Table 7 Primer sequences
Full size table

Immunohistochemistry

Immunohistochemistry and AQUA analyses were performed on a tissue array generated by cutting-edge matrix-assembly containing 100 deidentified primary invasive breast carcinoma specimens in a cohort described previously [33]. Immunohistochemistry was performed as described previously [33] using pY-Stat5 (Epitomics, 1:200), PTHrP (Santa Cruz, 1:200), and cytokeratin (DAKO, 1:100). AQUA analysis was performed using AQUA/PM2000 (HistoRx) as described previously [33].

Adenoviral and lentiviral production and infection

Lentiviral particle production was performed as described previously [33]. shRNA lentiviral vectors (Open Biosystems, Lafayette, CO, USA) were obtained for scrambled control (SC002), and Jak1 (TRCN0000003102), and Jak2 (TRCN0000003180). The cells were infected with lentivirus overnight and allowed to grow for 48 h before hormone induction for an additional 24 h. LacZ, Stat5a, Stat5b, and Dominant-negative-Stat5 (DN-Stat5) adenoviruses were prepared using double cesium chloride centrifugation [53] and used for gene delivery into T47D cells (1.5×106/well in 6 well dish; multiplicity of infection = 5). 24 h after infection, cells were incubated with hormones for another 24 h and subsequently harvested for immunoblot analysis.

Cell proliferation and soft agar assays

T47D cells were treated with vehicle, PRL, or 17β-Estradiol (E2) (Sigma) for 72 h. Triplicates of each condition were plated and counted using the Countess Automated Cell Counter (Invitrogen). For the soft agar assay, T47D cells suspended in 0.3% agarose were plated on top of 0.6% agarose. T47D cells were treated with media containing either vehicle, PRL (20 nM), E2 (10 nM), or PRL + E2 for 2 weeks, with fresh media and hormones added every 3 days. Each condition was done in quadruplicate. Images were taken from each well at 2 weeks and were analyzed for colony size using ImageJ.

Immunoblotting

T47D cells were lysed as described previously [54]. Proteins were resolved by SDS-PAGE and immunoblotted with antibodies to phospho-Stat5 (AX1, Advantex), and total Stat5 (BD #610192), followed by secondary antibodies Alexa Fluor 680-conjugated goat anti-mouse IgG (Invitrogen) or IRDye 800 CW-conjugated goat anti-rabbit IgG (Licor, Lincoln, NE, USA) depending on primary antibodies. Immunoblots were scanned using the Odyssey Infrared Imaging System (Licor).

Gene ontology analysis and survival analyses

The PRL-regulated genes identified using mRNA expression microarrays were analyzed for enrichment of Gene Ontology Biological Process terms using the Database for Annotation, Visualization and Integrated Discovery (DAVID ) v6.7 [55, 56]. A previously described collection of mRNA microarray datasets compiled from public repositories [57] was used to evaluate the set of PRL-induced transcripts in the context of clinical outcome. Transcription levels for PRL-upregulated genes were averaged into an expression signature and used to divide samples into the lower quartile, interquartile range, and upper quartile. Relapse-free and metastasis-free survival differences in these groups were evaluated for significance using the log rank test.