Prolactin cooperates with loss of p53 to promote claudin low mammary carcinomas

Abstract

TP53 is one of the most commonly mutated genes in cancer. In breast cancer, it is mutated in about 40% of primary clinical tumors and is associated with poor survival. The mammotrophic hormone, prolactin (PRL), and/or its receptor are also expressed in many breast cancers, and accumulating epidemiologic data link PRL to breast cancer development and progression. Like TP53 mutations, evidence for PRL activity is evident across several molecular cancer subtypes, and elevated PRL expression and loss of p53 have been observed in some of the same clinical tumors. In order to examine the interaction of these factors, we employed genetically modified mouse models of mammary-specific p53 loss and local overexpression of PRL. We demonstrated that mammary PRL decreased the latency of tumors in the absence of p53, and increased the proportion of triple negative claudin-low carcinomas, which display similarities to human clinical metaplastic carcinomas. Moreover, PRL/p53^−/− carcinomas displayed higher rates of proliferation and more aggressive behavior. Transcripts associated with cell cycle progression, invasion and stromal reactivity were differentially expressed in carcinomas that developed in the presence of elevated PRL. PRL/p53^−/− carcinomas also exhibited selectively altered expression of Activating Protein-1 components, including higher levels of c-Jun and FosL1, which can drive transcription of many of these genes and the epithelial-mesenchymal transition. The ability of PRL to promote claudin-low carcinomas demonstrates that PRL can influence this subset of triple negative breast cancers, which may have been obscured by the relative infrequency of this cancer subtype. Our findings suggest novel therapeutic approaches, and provide a preclinical model to develop possible agents.

INTRODUCTION

TP53 is mutated in more than 40% of breast cancers of multiple subtypes, and its mutation is associated with poor outcomes.^{1, 2} Mouse models have permitted study of the consequences of loss of this powerful tumor suppressor on mammary pathology.^{3, 4} Transplantation of p53^−/− mammary epithelium into wildtype recipients leads to stochastic development of carcinomas that display a range of histological and molecular subtypes, modeling the human disease and facilitating study of tumorigenesis in this context.^{5, 6} The interaction of mammotropic hormones with pathology induced by dysfunctional p53 has been of particular interest. Although only a minority of the tumors that develop from p53^−/− transplanted cells express ERα,⁶ prolonged postpubertal supplementation with estrogen and/or progesterone reduces tumor latency.⁷ Consistently, treatment with the anti-estrogen, tamoxifen, dramatically reduces tumor incidence.⁸ However, deletion of the progesterone receptor in the transplanted cells only partially reduces tumor number.⁷ The reduction in circulating prolactin (PRL) levels following tamoxifen treatment ⁸ suggests that this hormone may also contribute to tumorigenesis in the absence of p53.

PRL is a critical player in the development and differentiation of the mammary gland.⁹ Epidemiologic evidence also supports a role for PRL in breast cancer. Elevated circulating PRL is linked to the risk of particularly ER+ breast cancer.¹⁰ Moreover, locally synthesized PRL further increases exposure of breast tissue in women.^11–16 Circulating PRL and tumor PRL expression have been associated with metastasis and reduced survival in several studies.^17–20 In addition, PRL receptor (PRLR) transcripts are elevated in many tumors, compared to adjacent normal tissue, and PRLR protein can be stabilized by oncogenic pathways.^21–23 Interestingly, elevated PRL expression and loss of p53 have been observed in some of the same tumors.^{2, 19} However, the mechanisms whereby PRL interacts with other oncogenic factors, including loss of p53, to promote breast cancer are not well understood.

In order to experimentally examine the actions of PRL in mammary pathogenesis, we have developed a transgenic mouse model that expresses PRL in mammary epithelium (NRL-PRL), mimicking the local expression of PRL observed in women.²⁴ In this model, local PRL exposure is elevated without altering circulating PRL or estrous cycling, permitting study of PRL actions without secondary effects on ovarian hormones. Nulliparous NRL-PRL females develop diverse aggressive carcinomas, including many adenocarcinomas with molecular characteristics of the luminal tumor subtype of clinical breast cancer, as well as a small subset of spindle cell carcinomas.²⁵ While some of these tumors display phosphorylation of the classical PRL-induced signaling mediator, STAT5, others exhibit high levels of AP-1 proteins, which are induced by MAP kinases.²⁵ This murine model has elucidated the ability of PRL to cooperate with other important players in breast cancer, including growth factors and estrogen.^{26, 27} Here we have examined the interplay of PRL and the absence of p53 in mammary pathology, by crossing NRL-PRL mice with p53 null mice in the FVB/N genetic background, and transplanting the genetically modified mammary cells into syngeneic wildtype recipients. Our studies demonstrate that PRL potently promotes the development of aggressive claudin-low carcinomas, associated with selectively altered expression of AP-1 proteins, including higher levels of c-Jun and FosL1.

RESULTS AND DISCUSSION

PRL accelerates tumorigenesis in the context of p53^−/− and shifts the spectrum of tumor histotypes toward triple negative claudin low carcinomas

Nullizygous p53 mice on the FVB/N genetic background developed multiple tumors (chiefly hemangiosarcomas and lymphomas) as young adults and often did not live beyond 3 months of age (data not shown).²⁸ Therefore, to examine interactions between p53 status and PRL in mammary tumorigenesis in vivo, we transplanted genetically modified mammary tissue from young adult females to epithelium-free fat pads of wildtype 3 week old syngeneic females, as reported to study mammary p53 mutations in BALB/c mice.⁵ Donor glands of p53^−/− females exhibited normal morphology (Supplementary Figure 1a,e; c,g), as reported in BALB/c mice,⁵ and glands expressing the PRL transgene (NRL-PRL) displayed ductal dilation, alveolar development and focal hyperplasias (Supplementary Figure 1b,e), as previously described.²⁴ However, glands which contained cells with elevated PRL combined with p53 loss (NRL-PRL/p53^−/−; PRL/p53^−/−) showed markedly irregular ductal epithelium in addition to these other abnormalities, even at this young age (Supplementary Figure 1d,h).

Nulliparous nontransgenic recipients bearing transplanted PRL/p53^−/− mammary cells developed carcinomas more rapidly than p53^−/− cells without PRL (p<0.05) (Figure 1a and Table 1). In contrast, transplanted cells from wild-type and NRL-PRL donors did not develop tumors by one year of age, consistent with the long latency of mammary carcinomas in NRL-PRL females.^{24, 25} In addition, transgenic PRL significantly shifted the spectrum of tumor histotypes that developed from p53^−/− transplants. FVB/N p53^−/− donor cells developed carcinomas with diverse histotypes (Table 1, Figure 1b, 2a–c), as reported for the BALB/c strain.⁶ However, nearly all of the tumors that developed from PRL/p53^−/− donor cells consisted of anaplastic spindle-shaped cells (Figure 1b, Figure 2c,d). The proportion of tumors exhibiting this histotype was significantly higher in PRL/p53^−/− transplanted glands (p=0.024), and the proportion of adenocarcinomas tended to be lower (p=0.06) than transplanted p53^−/− cells without the PRL transgene (Figure 1b, Table 1). Many cells within these spindle cell tumors expressed high molecular weight cytokeratins (Supplementary Figure 2A) and cytokeratin 8 (K8, Figure 2e,f). Many of these carcinomas contained large giant cells (Figure 2d), some but not all of which were K8+ (Figure 2e,f). The PRL/p53^−/− carcinomas did not express immunologically detectable ERα, progesterone receptor, or ErbB2 (HER2) (Supplementary Fig. 2B), establishing their resemblance to clinical triple negative metaplastic carcinomas. In light of the ability of PRL to induce changes in collagen alignment in 3-dimensional cultures of breast cancer cells,²⁹ we examined collagen after staining with picrosirius red. As shown in Figure 2g,h, many of the carcinomas that developed from PRL/p53^−/− cells contained straight, directionally oriented collagen fibers, particularly near the tumor margins, which indicate a more aggressive tumor,³⁰ compared to the wavy fibers observed in the spindle cell carcinomas that developed from p53^−/− cells.

Figure 1.

PRL decreases the latency of mammary tumors that develop in the absence of p53 and increases the proportion of claudin low spindle cell carcinomas. NRL-PRL mice (line 1647-13, TgN(Nrl-Prl)23EPS), were generated in the FVB/N strain as described.²⁴ C57BL/6 mice carrying a deletion for p53⁷⁵ were backcrossed more than ten generations into the FVB/N genetic background.⁷⁶ Mammary fragments from 10–13 week old donors of the genotypes of interest were transplanted to mammary fat pads of 3 week old wildtype FVB/N mice which had been cleared of endogenous epithelium as described⁷⁷ (5–9 recipients per donor). Nulliparous recipients were examined weekly for tumors until one year of age. (a) Latencies of the tumors that developed from p53^−/− (N=12) and PRL/p53^−/− (N=15) donor cells were compared by Kaplan-Meier analysis. PRL/p53^−/− tumors had a significant shorter latency than p53^−/− tumors (p<0.05). All (100%) of the recipients of wild-type and NRL-PRL cells remained tumor free at 1 year of age. (b) Distribution of carcinoma histotypes that developed from p53^{-/ -} and PRL/p53^−/− donor cells. The proportion of spindle cell carcinomas was significantly higher in the tumors that developed from PRL/p53^−/− donors (*, p=0.024, Fishers exact test), and the proportion of adenocarcinomas tended to be lower (†, p=0.06, Fishers exact test). (c,d) RNA from mammary epithelial cells (MECs) pooled from 12 week old wildtype FVB/N females, individual adenocarcinomas that developed in NRL-PRL females²⁵ (N=4), and spindle cell carcinomas that developed from PRL/p53^−/− and p53^−/− donor cells in the current study was examined for the transcripts indicated by quantitative RT-PCR (qRT-PCR). Results were analyzed via the delta-delta C(t) method normalized to 18S ribosomal RNA, as described.²⁵ Primers used for qRT-PCR are listed in Supplementary Table 1. (N=4–6; mean +/− S.E.M.) ***p<0.0001 compared to NRL-PRL adenocarcinomas by the Kruskal-Wallis test for non-parametric data, followed by the Mann-Whitney post test. (e) Heat map of relative levels of the indicated transcripts, normalized to FVB/N MECs. Left, single adenocarcinoma from p53^−/− donor cells; middle, spindle cell carcinoma from p53^−/− donor cells; right, spindle cell carcinoma from PRL/p53^−/− donor cells. Levels of these transcripts in tumors that developed in these genotypes were not significantly different.

Table 1.

Effect of PRL in the presence of p53^−/− on mammary carcinomas

Genotype	p53−/−	NRL-PRL/p53−/−
Tumor latency (mean +/− s.d., days)	272 +/− 70	214 +/− 74*
Tumors/# recipients	12/22	15/23
Tumor histotype (# / total tumors)	Spindle cell carcinoma (6/12) Adenocarcinoma (5/12) Squamous cell carcinoma (1/12)	Spindle cell carcinoma (14/15) † Adenocarcinoma (1/15) ‡ Squamous cell carcinoma (0/15)
Tumors that invadedbeyond mammary fat pad (# / total tumors)	0/12	5/15

^*significantly reduced latency (p<0.05, Student’s t test).

^†significantly increased frequency (p=0.024, Fishers exact test).

^‡trend toward reduced frequency (p=0.06, Fishers exact test).

Figure 2.

Both NRL-PRL/p53^−/− and p53^−/− donor cells give rise to mammary carcinomas of multiple histotypes. (a-d) Hematoxylin/ eosin stained micrographs. (a) Squamous cell carcinoma that developed from p53^−/− donor cells. (b) Adenocarcinoma that developed from p53^−/− donor cells. (c) Spindle cell carcinoma that developed from PRL/p53^−/− donor cells. (d) Giant cells (arrows) in a tumor that developed from p53^−/− donor cells. (e,f) Cytokeratin 8 staining of spindle cell carcinomas that developed from p53^−/− donor cells (e) and PRL/p53^−/− donor cells (f). Note the cytokeratin 8+ giant cells (arrows). (g,h) Picrosirius red-stained spindle cell carcinomas that developed from p53^−/− donor cells (g) and PRL/p53^−/− donor cells (h), visualized with a polarizing microscope (Spot-advanced Olympus BX51 camera, Olympus, Germany).⁷⁸ (a-f), Original magnification x100. Bars= 100µ. Insets in (g,h), 200x.

To further characterize these tumors, we examined levels of transcripts for markers that distinguish breast cancer subtypes.⁶ All of the spindle cell carcinomas examined contained very low levels of Cldn 3 and Cldn7 mRNA, in marked contrast to wildtype FVB/N mammary epithelial cells and adenocarcinomas induced by the PRL transgene in the context of wildtype p53 (Figure 1c,d, p<0.0001). Furthermore, these tumors exhibited the transcript signature of the epithelial-to-mesenchymal transition (EMT).³¹ Together with the cytokeratin staining, these data confirm that the carcinomas of this histotype that developed from both p53^−/− and PRL/p53^−/− transplanted cells were “claudin-low” (Figure 1e).^{6, 32, 33}

The triple negative claudin-low subtype of breast cancer comprises about 7–14% of primary clinical breast cancers, and has a worse prognosis than luminal A tumors.^{2, 33} This subtype exhibits some resistance to anthracycline/taxane-based chemotherapy,³³ and residual tumors after neoadjuvant chemotherapy or anti-estrogen treatment are enriched in its transcript signature.^{33, 34} Clinical claudin-low tumors express high levels of markers for mammary stem cells,^{33, 35} and examination of experimental p53^−/− tumors in BALB/c mice demonstrated enrichment in tumor initiating cells.⁶ However, the factors that lead to the development and progression of these tumors are not known. Of particular interest in light of the dramatic increase in the proportion of these tumors that developed with PRL exposure in the current study, PRL was elevated in 37% of claudin-low tumors in a clinical study, although the sample size was small.²

Transgenic PRL increases tumor growth and aggression

Although the claudin low carcinomas that developed from both p53^−/− and PRL/p53^−/− cells were locally invasive, the PRL/p53^−/− tumors were more aggressive. While all of the tumors from p53^−/− transplants remained confined to the mammary fat pad, 32% of the PRL/p53^−/− carcinomas invaded into the peritoneal cavity (Table 1). Consistently, carcinomas that developed in the context of locally elevated PRL were significantly larger at end stage, despite the decreased latency (Figure 3a, p<0.05). Moreover, the rate of proliferation of these tumors was significantly higher (p<0.01, Figure 3b). In order to evaluate potential mechanisms of the increased proliferation, we examined levels of mRNA for multiple cell cycle regulators in claudin low carcinomas that developed in both genotypes. Interestingly, we identified no differences in transcript levels of cyclins (listed in Supplementary Table 1), but those for two regulators, p16^Ink4a and p19^Arf, were significantly reduced in PRL/p53^−/− compared to p53^−/− carcinomas (Figure 3c).

Figure 3.

Carcinomas that developed from PRL/p53^−/− donor cells were significantly larger despite the reduced latency and exhibited higher indices of proliferation, and contained different levels of mRNAs for cell cycle inhibitors, proteases associated with invasion, and stromal changes. (a) Tumor wet weight at end stage. (N=9–14; mean +/− S.E.M.) *p<0.05. (b) Proliferation indices were determined by counting PCNA-labeled cells in 1000 total cells from at least 5 randomly chosen microscopic fields in divergent regions of each carcinoma. (N=6–7; mean +/− S.E.M.) **p<0.01, Kruskal-Wallis test for non-parametric data, followed by the Mann-Whitney post test. (c-e) RNA from individual p53^−/− and PRL/p53^−/− spindle cell carcinomas was examined for the transcripts indicated by qRT-PCR as described for Figure 1. (N=4–6; means +/− S.E.M.) (c) Transcripts for p16^Ink4a and p19^Arf . *p<0.05, 2-tailed Student’s t test. (d) Transcripts associated with invasion. *p<0.05 by 2-tailed Student’s t test; ‡p=0.03 by 1-tailed Student’s t test. (e), Transcripts associated with stromal changes. †p=0.08, 1-tailed Student’s t test; §p=0.086, 1-tailed Student’s t test.

Both p16^Ink4a and p19^Arf are encoded by the Cdkn2a locus. As tumor suppressors, they are disabled in many cancers.³⁶ Despite the overlap in their genomic organization, these genes have distinct promoters and play different roles to cooperatively regulate the cell cycle and integrate cellular stresses and outcomes. p16^Ink4a represses the activity of CDK4/6 cyclin complexes, and p19^Arf acts in part by blocking degradation of p53 mediated by MDM2, inducing cell cycle arrest or apoptosis. The cooperation of p16^Ink4a and p53 deficiencies in tumorigenesis is well recognized.³⁷ Interestingly, the BALB/c strain, which has been used extensively to model p53 in mammary tumorigenesis, has a mutated allele of Cdkn2a that impairs the activity of p16^Ink4a.³⁸ This genetic variation may have obscured a role for PRL on p53-initiated mammary tumorigenesis in this strain.^{7, 8} In combination with the ability of PRL to post-transcriptionally stabilize other cell cycle regulators,^{39, 40} these findings further illuminate the mechanisms whereby PRL increases mammary cell proliferation.^{24, 41–43}

To determine effects on enzyme expression that might impact the observed invasiveness, we examined mRNA for the matrix metalloproteinases (MMP) -2, and -9, which have been associated with tumor invasion and metastases.^{44, 45} As shown in Figure 3d, Mmp9, but not Mmp2, mRNA was significantly elevated in PRL/p53^−/− carcinomas. Moreover, transcripts for cathepsin L, another protease important in matrix degradation and invasion in breast cancer,⁴⁶ were also elevated in tumors of this genotype.

In light of the ability of PRL to increase transcripts for Type I collagen⁴⁷ (Rugowski, unpublished observations) and the marked differences in the appearance of collagen fibers in PRL/p53^−/− , compared to p53^−/− spindle cell carcinomas (Figure 2g,h), we examined Type I collagen mRNAs in these tumors. As shown in Figure 3e, levels of Col1a1, but not Col1a2, mRNA tended to be higher in PRL/p53^−/− carcinomas than p53^−/− carcinomas (p=0.08, 1-tailed Student’s t test). Transcripts for the pro-angiogenic ligand, Vegfa, also tended to be higher (p=0.086, 1-tailed Student’s t test).

Together, these data indicate that PRL modulates expression of genes which are likely to contribute to the more aggressive phenotype of the claudin low carcinomas that develop from PRL/p53^−/− , compared to p53^−/− cells.

PRL/p53^−/− carcinomas express higher levels of AP-1 components associated with aggression

PRL/p53^−/− carcinomas exhibited uniformly very low levels of transcripts for two well-characterized mediators of PRL’s physiologic actions in the mammary gland, Elf5 and Stat5^{48, 49} (data not shown), in contrast to their variable levels in PRL-induced adenocarcinomas.²⁵ Indeed, these mediators are not associated with the aggressive phenotype of claudin-low tumors, such as those observed in our studies: clinically, activated Stat5 is associated with better differentiated tumors and therapeutic responses,^{50, 51} and experimentally, Elf5 inhibits the EMT.⁵²

Instead, the PRL/p53^−/− carcinomas in the current study displayed altered levels of Activating Protein -1 (AP-1) components, compared to p53^−/− carcinomas. AP-1 complexes, composed of homo- or hetero-dimers of AP-1 proteins, transcriptionally regulate many key genes controlling proliferation, the EMT and invasion, and are often deregulated in cancers (for reviews,^53–58). As shown in Figure 4a, PRL/p53^−/− carcinomas contained significantly higher levels of mRNA for c-Jun and FosL than p53^−/− alone (p<0.05). In contrast, transcripts for c-Fos trended lower (p=0.16, 2-tailed Student’s t test), while those for JunB, as well as JunD (not shown) were similar in tumors of both genotypes. Analysis of protein levels of these components in the carcinomas demonstrated similarly significant patterns (Figure 4b).

Figure 4.

PRL/p53^−/− carcinomas express higher levels of c-Jun and Fosl1 compared to p53^−/− tumors. (a) RNA from FVB/N MECs and mammary carcinomas that developed from transplanted p53^−/− and PRL/p53^−/− cells was examined for levels of the transcripts indicated by qRT-PCR as described for Figure 1. (N=6, means +/− S.E.M.) *p<0.05 by Kruskal-Wallis test for non-parametric data, followed by the Mann-Whitney post test. (b) Equal amounts of protein from mammary carcinomas that developed from transplanted p53^−/− and PRL/p53^−/− cells were examined for the indicated proteins by immunoblotting, and signals were quantified by densitometry. (means +/− S.D., N=5; *p<0.05, ***p<0.001 by 2 tailed Student’s t test; †p=0.068 by 1-tailed Student’s t test).

The AP-1 family is dysregulated in 10–15% of breast cancers.^{2, 59} The activity of AP-1 enhancers is controlled by not only the levels of these complexes, but also the composition of the AP-1 dimers as well as promoter context.^{54, 56} c-Jun and FosL1, both elevated in PRL/p53^−/− compared to p53^−/− carcinomas, are associated with poor outcomes in breast cancer in clinical studies.^60–64 In contrast, other AP-1 proteins, such as c-Fos, which was reduced in our studies in the presence of PRL, are associated with better outcomes.

PRL initiates activation of many MAP kinases which increase the synthesis and phosphorylation of AP-1 components,⁶⁵ and AP-1 proteins are elevated in a subset of PRL-induced mammary carcinomas that develop in NRL-PRL females.²⁵ The ability of PRL to activate these pathways in vitro is inversely related to its ability to activate Stat5,⁶⁷ which also is reflected in PRL-induced carcinomas in mouse models.²⁵ Although a role for AP-1 signals in hormone-driven postnatal mammary development has been established,⁶⁶ a link to physiological actions of PRL action is unclear. Among the genes regulated by AP-1 complexes are those encoding p16^Ink4a and p19^Arf,^{68, 69} MMP-9 and Cathepsin L,³⁹ Vegfa,⁷⁰ and type I collagen,⁷¹ as well as those regulating the EMT.⁵⁴

Although the transcript profiles of p53^−/− claudin-low carcinomas resemble mammary stem cells,^{33, 35} the cells that give rise to these tumors are poorly understood. PRL may be expanding an epithelial subpopulation that is the origin of these tumors, promoting progression of early lesions, and/or initiating paracrine signals which promote this cancer subtype. The secreted transgenic PRL in this model raises local concentrations within the mammary compartment, where it is available to act on any target cells, modeling that in women. Some evidence suggests that PRL can act on a luminal epithelial subpopulation, increasing RANKL, which acts on cells with surface markers enriched for mammary stem cells.⁷² However, the mammary target cells of PRL are poorly understood. In addition to epithelial cells, stromal cells also contain PRLR in rodent models.^{73, 74} Ongoing studies to identify the target cells and consequences of PRL action will illuminate the direct and indirect mechanisms by which PRL contributes to this disease.

Our studies demonstrate that PRL can drive the development of the aggressive claudin low carcinomas that develop in the genetically unstable environment permitted by the absence of p53. The ability of PRL to promote this breast cancer subtype demonstrates that PRL can influence a subset of nonluminal breast cancers, which may have been obscured by the relative infrequency of these cancers. Our results provide additional mechanistic insight into the effects of PRL on epithelial proliferation, and demonstrate the potential for PRL to directly contribute to other processes in tumor progression via AP-1 regulated genes, including angiogenesis, matrix remodeling, and the EMT. The evidence for PRL-augmented AP-1 activity suggests potential therapeutic and/or preventative strategies for this cancer subtype, which may be targeted alone or in combination with another agent. The preclinical model described here will assist in the development of these approaches.