A new role of SNAI2 in postlactational involution of the mammary gland links it to luminal breast cancer development.

Breast cancer is a major cause of mortality in women. The transcription factor SNAI2 has been implicated in the pathogenesis of several types of cancer, including breast cancer of basal origin. Here we show that SNAI2 is also important in the development of breast cancer of luminal origin in MMTV-ErbB2 mice. SNAI2 deficiency leads to longer latency and fewer luminal tumors, both of these being characteristics of pretumoral origin. These effects were associated with reduced proliferation and a decreased ability to generate mammospheres in normal mammary glands. However, the capacity to metastasize was not modified. Under conditions of increased ERBB2 oncogenic activity after pregnancy plus SNAI2 deficiency, both pretumoral defects-latency and tumor load-were compensated. However, the incidence of lung metastases was dramatically reduced. Furthermore, SNAI2 was required for proper postlactational involution of the breast. At 3 days post lactational involution, the mammary glands of Snai2-deficient mice exhibited lower levels of pSTAT3 and higher levels of pAKT1, resulting in decreased apoptosis. Abundant noninvoluted ducts were still present at 30 days post lactation, with a greater number of residual ERBB2+ cells. These results suggest that this defect in involution leads to an increase in the number of susceptible target cells for transformation, to the recovery of the capacity to generate mammospheres and to an increase in the number of tumors. Our work demonstrates the participation of SNAI2 in the pathogenesis of luminal breast cancer, and reveals an unexpected connection between the processes of postlactational involution and breast tumorigenesis in Snai2-null mutant mice.

Introduction

Breast cancer is one of the leading causes of death among women worldwide. It is a complex and heterogeneous disease with substantial variations in its molecular and clinical characteristics.[1] ERBB2/NEU/HER2 (henceforth ERBB2)-positive breast cancers constitute 20-30% of all mammary gland tumors, and ERBB2 amplification and overexpression are markers of poor prognosis.[2] Breast cancer evolution is characterized by different stages of progression; the final one is distant metastasis, responsible for the poor survival seen in this disease.

The transcription factor SNAI2/SLUG is involved in the process of epithelial to mesenchymal transition (EMT) by repressing E-cadherin expression.[3, 4] SNAI2 has been associated with tumor invasion and metastasis and the poor prognosis of several tumor types, including breast carcinoma.[3] Furthermore, SNAI2 has additional cellular functions such as the inhibition of apoptosis,[5-7] the regulation of cell movement, adhesion, proliferation[8] and stem-cell properties,[9, 10] all of them important in tumor pathogenesis and evolution.[11] SNAI2 has also been linked to early differentiation and morphogenesis in several tissue types.[4]

The importance of SNAI2 in the determination of the basal phenotype in breast cancer of both sporadic and BRCA1-mutated origins has been demonstrated previously.[12, 13] However, SNAI2 is rarely expressed in luminal breast cancer cells[14] (Supplementary Figure S1a). In the mammary gland, SNAI2 is expressed in normal breast tissue and is localized in the proliferative basal compartment during mammary gland morphogenesis,[9, 15] but it is not known whether SNAI2 might play a role in the development of luminal tumors. To address this point, we studied the participation of SNAI2 in the development of breast cancer in MMTV-ErbB2 mice,[16] which develop breast tumors of the luminal phenotype.[17] Here, we demonstrate that SNAI2 is important for the development of luminal tumors induced by ERBB2. In particular, SNAI2 deficiency protected mice from breast tumorigenesis, led to an extended tumor latency and overall survival with fewer tumors, with only a very weak effect on dissemination. Since tumor latency and yield have their origin in pre-tumoral stages, this demonstrates the participation of SNAI2 in early phases of breast cancer development of luminal origin. Interestingly, in response to an increased expression of the ERBB2 transgene during pregnancy, this protective effect observed in the absence of SNAI2 disappeared, but the defect in the dissemination of metastases was exacerbated. We also report results demonstrating the importance of SNAI2 for proper mammary gland involution, and a connection between this process and breast tumorigenesis in Snai2 knockout (KO)ERBB2+ mice.

Results

SNAI2 deficiency alters the susceptibility and evolution of breast cancer of luminal origin induced by ERBB2

To determine whether SNAI2 was involved in the susceptibility to and evolution of luminal breast cancer, we crossed B6;129S1-Snai2/J[18] with FVB/N-Tg(MMTVneu)202Mul/J transgenic mice, which develop luminal tumors.[16, 17] Thus, we established a Snai2-deficient line that overexpresses the ERBB2 protooncogene (henceforth Snai2 KOERBB2+). To avoid the C57BL/6 resistance effect in breast cancer development,[19-21] F1 transgene-positive/Snai2-deficient males were backcrossed with FVB females to obtain the F4 generation with more than 90% FVB genetic background,[22] as described in the Methods section. To better evaluate the effect of SNAI2 in the different features of breast cancer, we differentiated several pathophenotypes in the disease. We evaluated latency, disease duration and lifespan, and tumor traits of progression, such as the number of tumors, local growth and metastases. We observed that Snai2 KOERBB2+ mice had a longer tumor latency than Snai2 WTERBB2+ animals (median of 54.1 versus 47.7 weeks, respectively), with no significant differences in the duration of the disease, thus leading to longer life-spans (Figures 1a-c and Table 1). In addition, Snai2 KOERBB2+ mice developed significantly fewer tumors than Snai2 WTERBB2+ mice, with no differences in local tumor growth (Figure 1d and Table 1). Because SNAI2 has been implicated in metastasis,[4, 5, 8, 23] we compared the frequency of metastasis in the Snai2 WTERBB2+ and Snai2 KOERBB2+ mice. Unexpectedly, we did not find differences in the median and incidence of metastasis (62.5% of Snai2 WTERBB2+ mice developed metastasis versus 50.0% of Snai2 KOERBB2+ mice). No differences were seen in the multiplicity of metastatic impacts in the lungs either (i.e. the percentage of mice already having metastasis with two or more impacts) (Figures 1e and 1f, and Table 1).

Figure 1

SNAI2 participates in the early stages of luminal breast cancer development

Comparison of different tumor pathophenotypes in nulliparous female mice: (a) Latency. (b) Duration of disease. (c) Survival. (d) Number of tumors. (e) Incidence of metastases. (f) Multiplicity of metastases. (g) Detection of SNAI2 in mammary glands and luminal tumors (western blot). (h) Comparison between tumors from Snai2 WT ERBB2+ and Snai2 KOERBB2+ mice of the percentage of viable Lin- cells corresponding to the luminal lineage (CD24hi CD29lo), basal lineage (CD24lo CD29hi), and the stem-cell-enriched population (CD24lo CD29hi CD49fhi).

Table 1

Pathophenotypes studied in nulliparous and parous Snai2 WTERBB2+ and Snai2 KOERBB2+ female mice

Tumor traits were compared using the Mann Whitney U test, and the temporal stages of the disease were evaluated using the Kaplan-Meier estimator and the Log-Rank test. The incidence and multiplicity of tumors and metastases were compared with Fisher's exact test. The numbers of mice used in the experiments are shown in the respective figures along the manuscript. N: number of mice. The N values shown in the local growth rate represent all tumors analyzed. IR: interquartile range; n.s.: non-significant P values.

A. TEMPORAL STAGES			Snai2 WTERBB2+	Snai2 KOERBB2+	P value
Tumor Latency (weeks)
Latency interval (N)	Nuliparous		28.8-90.9 (N=41)	29.7-114 (N=33)
	Parous		21.6-61 (N=24)	13.1-80.8 (N=45)
Median (IR)	Nuliparous		47.7 (13.8)	54.1 (21.04)	0.0127
	Parous		40 (14.82)	40.4 (13.21)	n.s.
		P value	0.0038	0.0004
Duration of Disease (weeks)
Disease interval (N)	Nuliparous		4.28-23.86 (N=36)	4.57-21.14 (N=22)
	Parous		6.7-20 (N=24)	5.14-22.29 (N=45)
Median (IR)	Nuliparous		12.9 (12.75)	10.36 (10.39)	n.s.
	Parous		13.3 (7.46)	11.6 (5.14)	n.s.
		P value	n.s.	n.s.
Lifespan (weeks)
Lifespan interval (N)	Nuliparous		39.4-90.9 (N=41)	38.3-114 (N=32)
	Parous		20-79.1 (N=24)	32.1-93 (N=45)
Median (IR)	Nuliparous		60.42 (14.28)	66.14 (32.93)	0.0145
	Parous		53.38 (17.07)	50 (16.14)	n.s.
B. TUMOR TRAITS		P value	0.0089	0.0009
Tumor Number
Incidence % (rate)	Nuliparous		87.8% (36/41)	66.67% (22/33)	0.0453
	Parous		87,5% (21/24)	76,09% (35/46)	n.s.
		P value	n.s.	n.s.
Multiplicity (2 or more)	Nuliparous		88.89% (32/36)	59.09% (13/22)	0.0204
	Parous		90.48% (19/21)	94.29% (33/35)	n.s.
		P value	n.s.	0.0017
Median (IR)	Nuliparous		3 (2)	1 (3)	0.0012
	Parous		3 (4)	3 (3.25)	n.s.
		P value	n.s.	0.0076
Local Growth Rate
Median (IR) (N)	Nuliparous		0.32 (0.17) (N=59)	0.33 (0.136) (N=26)	n.s.
	Parous		0.24 (0.21) (N=46)	0.32 (0.27) (N=72)	n.s.
		P value	n.s.	n.s.
Metastasis
Incidence % (rate)	Nuliparous		63.64% (21/33)	50% (9/18)	n.s.
	Parous		90.47% (19/21)	61% (20/33)	0.0276
		P value	0.0535	n.s.
Multiplicity (2 or more)	Nuliparous		100% (21/21)	100% (9/9)	n.s.
	Parous		78.9% (15/19)	90% (18/20)	n.s.
		P value	0.0424	n.s.
Median (IR)	Nuliparous		5 (8.5)	1 (5.25)	n.s.
	Parous		8 (18)	3 (10)	0.0349
		P value	n.s.	n.s.

Interestingly, SNAI2 was expressed in the mammary gland but not in the bulk of the tumor (Figure 1g and Supplementary Figure S1b). The breast tumors generated in Snai2 KOERBB2+ mice were ER-negative, PR-negative, and ERBB2-positive. They were positive for the luminal cytokeratins CK8 and CK18 and negative for the basal cytokeratin CK6. In addition, there were no significant differences in the expression levels of several luminal markers, such as Gata3, Sox9, and CK18, between tumors from the Snai2 KOERBB2+ and Snai2 WTERBB2+ mice (Supplementary Figures S1c-g). To gain further insight into the characterization of ERBB2-positive tumors generated in the absence of Snai2, we evaluated the distribution of the epithelial cell hierarchy in the tumors. The bulk of the epithelial population in the ERBB2+ tumors was formed by CD24hiCD29lo luminal cells in agreement with previous studies,[24] and there were no differences in the luminal, basal (CD24loCD29hi) and stem-cell-enriched populations (CD24loCD49fhiCD29hi) between tumors from Snai2 WTERBB2+ and Snai2 KOERBB2+ mice (Figure 1h). These results suggested that Snai2 KOERBB2+ mice develop tumors of luminal origin, as has been reported previously for Snai2 WTERBB2+ mice.[17]

Pregnancy recovers local breast tumorigenesis in Snai2 KOERBB2+ mice but increases the proportion of metastasis-free mice

It has been reported that pregnancy increases the expression of the ErbB2 protooncogene, driven by the MMTV promoter, resulting in a more aggressive disease.[16, 25] We therefore wondered to what extent SNAI2 would be necessary for luminal breast cancer development generated under increasing oncogenic activity of ERBB2 after pregnancy. As expected, we observed an exacerbation of the disease in both Snai2 WTERBB2+ and Snai2 KOERBB2+ parous mice, resulting in a statistically significant decrease in latency and lifespan, but with no modification in the duration of the disease (Supplementary Figures S2a-f, and Table 1).

In the MMTV-ErbB2 transgenic mice, these effects of SNAI2 deficiency on luminal tumor development were compensated after pregnancy. Parous Snai2 WTERBB2+ mice displayed a significantly shorter latency than their nulliparous counterparts, as reported,[16, 26] but no significant modifications in the number of tumors were observed (Supplementary Figures S2a and S2g). By contrast, parous Snai2 KOERBB2+ mice showed both a shorter latency and a significant increase in the number of tumors with respect to their nulliparous analogues (Supplementary Figures S2b and S2h). In fact, the effect on both the latency and tumor numbers in parous Snai2 KOERBB2+ mice was more evident than in their parous wild-type equivalents (Figure 2a). This meant that the differences in latency and tumor number previously observed between nulliparous Snai2 WTERBB2+ and nulliparous Snai2 KOERBB2+ mice had disappeared in their parous counterparts (Figures 2b, 2d and 2e, and Table 1). However, the induction of ERBB2 by pregnancy did not modify the duration of the disease (Figure 2c and Supplementary Figures S2e and S2f, and Table 1). This is in agreement with the observation that, because metastases are well tolerated in this model, local tumor growth is the main pathophenotype that determines the duration of the disease, and latency is the one that mainly determines lifespan (Table 1).[27]

Figure 2

Effect of pregnancy in breast cancer development under Snai2 deficiency

(a) Tumor latency in nulliparous and parous mice. Note that Snai2 KOERBB2+ mice showed the greatest shortening of tumor latency after pregnancy. (b-h) Comparison of different tumor pathophenotypes in parous female mice: (b) latency; (c) duration of disease; (d) lifespan; (e) number of tumors; (f) incidence of metastases; (g) multiplicity of metastases. (h) Representative dotplots of viable Lin- tumor cells from parous Snai2 WTERBB2+ (upper panel) and parous Snai2 KOERBB2+ (lower panel) mice, showing the gating strategy followed to quantify luminal, basal, and stem-cell-enriched populations. (i) Comparison between tumors from parous Snai2 WTERBB2+ and Snai2 KOERBB2+ mice of the percentage of viable Lin- cells corresponding to the luminal lineage (CD24hi CD29lo), basal lineage (CD24lo CD29hi), and the stem-cell-enriched population (CD24lo CD29hi CD49fhi).

Regarding tumor dissemination, parity significantly increased the incidence and multiplicity of metastases in Snai2 WTERBB2 mice with respect to their nulliparous counterparts (Table 1). The absence of SNAI2 increased the percentage of metastasis-free mice under increasing ERBB2 oncogenic activity. Thus, the incidence of metastasis in parous Snai2 KOERBB2+ mice was significantly lower than in parous Snai2 WTERBB2+ animals (61% versus 90.47%, respectively; P = 0.0276) without modification of multiplicity (Figures 2f and 2g, and Table 1).

Moreover, there were no significant differences in ERBB2 expression between tumors from nulliparous and parous mice of both the WT and KO Snai2 genotypes (Supplementary Figures S1c and S2i-k). This is in agreement with the fact that there were no significant differences in the local growth speed, proliferation and apoptosis between tumors from nulliparous and parous Snai2 WTERBB2+ and Snai2 KOERBB2+ mice (Supplementary Figures S2l and S2m). Furthermore, we analyzed the tumor epithelial cell hierarchy from parous mice. Again, the bulk of the population was formed by epithelial cells without significant differences between tumors from parous Snai2 WTERBB2+ and parous Snai2 KOERBB2+ mice. Surprisingly, we saw a slight but significant difference in the basal (CD24loCD29hi) and the stem-cell-enriched (CD24loCD49fhiCD29hi) subpopulations in favor of the tumors from the Snai2 KOERBB2+ mice after pregnancy conditions (Figures 2h and 2i).

Mammary glands from nulliparous Snai2 KOERBB2+ mice show defects in cellular turnover

Since the Snai2 deficiency modified tumor characteristics originating in pretumoral stages, such as latency and the number of tumors, we evaluated the morphology of mammary glands from adult nulliparous mice at two months of age. Normally, the extension of fat pad filled by ducts is used as a criterion for assessing the ability of mammary epithelial cells to undergo ductal growth.[28] The fat pads of the Snai2 KOERBB2+ mice were completely filled by ducts, similar to wild-type mice (Figure 3a), but the ducts in the glands of the Snai2 KOERBB2+ female mice were characterized by a reduced presence of side-branching as compared with their Snai2 WTERBB2+ counterparts (Figures 3b), as recently reported in non-transgenic Snai2 KOWT mice.[9] However, we did not find differences in the number of terminal end-buds (Supplementary Figures S3a and S3b). Previously, an impairment in ductal growth has been reported in Snai2 WTERBB2+ transgenic mice with respect to their control Snai2 WTWT non-transgenic counterparts,[29] as reflected by a shorter average ductal length (defined as the distance from the center of the lymph node to the leading edge of the gland); this shortening was also confirmed here (P = 0.018) (Figure 3c). In addition, mice that overexpressed ERBB2 in the context of Snai2 deficiency showed a similar shortening of average ductal length with respect Snai2 WTWT non-transgenic mice, but there was only a statistical trend (P = 0.083) (Figure 3c). Furthermore, there were no significant differences in the expression levels of several markers of luminal lineage described previously,[9] such as the progesterone receptor (PR), Gata3 and Ck18, between mammary glands from Snai2 KOERBB2+ and Snai2 WTERBB2+ mice (Supplementary Figure S3c). In addition, transgenic expression of ErbB2 in the mammary gland did not modify the SNAI2 localization already described in the basal compartment[9] (Supplementary Figure S3d).

Figure 3

Mammary glands from nulliparous Snai2 KOERBB2+ mice show defects in cellular turnover and stem-cell populations

(a) Fat pad in mammary glands normally filled by ducts (whole-mount). (b) Defect in side branching in Snai2 KO mice. (c) Average length of primary ducts. Distances were estimated based on whole-mount preparations. (d) Ki67 detection in mammary glands from female mice (immunohistochemistry) (left). Percentage of Ki67-positive cells per gland determined in five different mice (right). (e) Detection of CYCLIN D1 in mammary glands by western blot. Each number represents an epithelial mammary gland from an individual mouse. (f) Cleaved CASPASE-3, in nulliparous mice (immunohistochemistry) (left). Percentage of cleaved CASPASE-3-stained cells per gland of five different mice (right). (g) Ability to generate mammospheres ex vivo. Mammospheres were obtained from the mammary glands of 6-week-old mice. (h) Propagating ability of mammospheres in vitro. (i) SNAI2 protein expression in epithelial mammary glands versus mammospheres obtained from the same preparation by western blot.

Latency is the time required for the initiated target cells to acquire the fully transformed phenotype, which is related to the baseline proliferation of a tissue. Thus, the higher the proliferation itself, or in the presence of promoting agents, the easier it becomes for the cells to acquire new secondary oncogenic events and the shorter the latency time.[30] In light of this, we analyzed cellular turnover in terms of the rate of cell proliferation and apoptosis in histologic sections of mammary glands from Snai2 WTERBB2+ and Snai2 KOERBB2+ mice. Interestingly, according to Ki67 staining (Figure 3d) mice deficient in Snai2 showed a significantly decreased proliferation in the mammary epithelia of nulliparous mice and lower levels of CYCLIN D1 (Figure 3e). Surprisingly, nulliparous Snai2 KOERBB2+ mice showed discretely but significantly lower levels of baseline apoptosis (active CASPASE-3 staining) than their wild-type counterparts (Figure 3f).

Nulliparous Snai2 KOERBB2+ mice have defects in stem-cell populations in the mammary gland

Mammary stem cells have been implicated in the origin of breast cancer,[31] and the number of stem cells is reflected in the number of mammospheres generated in vitro.[32] Moreover, SNAI2 participates in the stem-cell biology of the mammary gland.[9, 10, 33] Because latency and tumor numbers are both traits of breast cancer determined in pretumoral stages, we evaluated the generation of mammospheres in the mammary glands of nulliparous Snai2 KOERBB2+ mice. Snai2 KOERBB2+ mice had a diminished capacity to generate mammospheres, but these maintained their ability to expand in vitro (Figures 3g and h). Interestingly, the expression of SNAI2 in mammospheres was enriched in comparison with its levels of expression in mammary gland epithelial cells (Figure 3i), in agreement with the role of SNAI2 in mammary stem-cell biology.[9, 10, 34] Nevertheless, we did not observe differences in mammosphere formation in Snai2 KOWT non-transgenic mice (Supplementary Figure S3e).

Since normal mammary glands from Snai2 KOERBB2+ mice generated fewer mammospheres than nulliparous Snai2 WTERBB2+ mice, suggesting that the nulliparous Snai2 KOERBB2+ animals would have fewer target mammary stem cells susceptible to transformation than Snai2 WTERBB2+ mice, we used flow cytometry to evaluate the number of stem cells and other subpopulations in the mammary glands of the nulliparous mice. We did not observe clear differences in the percentage of the CD24+/CD49fhi/CD29hi subpopulations identified as mammary stem cells in the literature[31, 35] (Supplementary Table S1). Thus, in the Snai2 KOERBB2+ mice we evaluated other cell subpopulations described in the mammary gland from Snai2 WTERBB2+ transgenic mice.[24, 36-39] The nulliparous Snai2 KOERBB2+ mice had no differences in the population of Sca1pos cells, reported to be enriched for functional stem/progenitor cells with increased regenerative potential,[36, 37] and we also failed to find differences in the CD24+Sca1+ subpopulation, specifically described to have self-renewal capability in Snai2 WTERBB2+ mice.[24] Moreover, neither did we find differences in the CD49f+CD61hi population described as being enriched in tumor-initiating cells in Snai2 WTERBB2+ transgenic mice.[38] Finally, we did not find differences in the number of luminal progenitors CD24hiCD29med with self-renewal capability.[39] EpCAM has recently been described as a better marker that permits greater resolution of the epithelial mammary gland subpopulations in the mouse.[40, 41] Thus, we re-analyzed all these populations with the anti-EpCAM antibody, but failed to find any difference either (Supplementary Table S1). We were also unable to demonstrate differences in the mammary glands from nulliparous Snai2 KOERBB2+ mice in luminal cells in spite of the function of SNAI2 described in the luminal and stem-cell compartments[10, 12, 33] (Supplementary Table S1).

All the data presented above suggested that there seem to be no-differences in the number of putative targets of tumors in MMTV-ErbB2 mice in the absence of Snai2. Nevertheless, the fact that we observed a diminished capability in mammosphere production (Figure 3g) suggests that there is a functional defect in self-renewal capability in MSCs from Snai2 KOERBB2+ mice. Moreover, because we observed that the propagating capability of mammospheres was normal (Figure 3h), this suggests the defect in MSCs would be non-cell-autonomous due to other compartments, probably in the niche.

The defect in mammosphere yield in nulliparous Snai2 KOERBB2+ mice is compensated after pregnancy

Since the defect in latency and number of tumors disappeared after pregnancy, we assessed the cellular turnover and the ability to generate mammospheres in parous mice. To evaluate the behavior of the mammary gland after the most active phases of involution, we carried out this study at thirty days after the end of lactation. Under these conditions, and despite a higher epithelial cellularity (addressed in the next section), there was also a reduced proliferation in the mammary glands of parous Snai2 KOERBB2+ mice (Figure 4a), associated with lower levels of CYCLIN D1 (Figure 4b). Parous Snai2 KOERBB2+ mice also exhibited less apoptosis in the mammary glands, but this was not statistically significant (Figure 4c).

Figure 4

Snai2 KOERBB2+ mice compensate the defect in the number of mammospheres after pregnancy

All analyses from this figure were performed using parous mice at 30 days post-lactation. (a) Immunohistochemical staining for Ki67 (left). Percentage of Ki67-positive cells per gland (right). (b) Detection of CYCLIN D1 by western blot. Each number represents an epithelial mammary gland from an individual mouse. (c) Tissue staining against cleaved-CASPASE-3 (left). Percentage of CASPASE-3-positive cells per gland (right). (d) Mammosphere yield obtained from mammary glands.

In addition, the generation of mammospheres by parous Snai2 WTERBB2+ and parous Snai2 KOERBB2+ mice was equivalent, indicating similar stem-cell activity in the mammary glands (Figure 4d and Supplementary Figure S3f). This is consistent with the fact that no statistical differences were observed in the number of tumors generated between parous Snai2 WTERBB2+ and parous Snai2 KOERBB2+ mice (Figure 2e). Neither were there any differences between the number of mammospheres generated by nulliparous Snai2 WTERBB2+ mice versus parous Snai2 WTERBB2+ mice (Supplementary Figure S3g), and again this was correlated with the absence of differences in the number of tumors generated (Supplementary Figure S2g). Increased mammosphere production was only seen in the parous Snai2 KOERBB2+ mice (Supplementary Figure S3h), which, as indicated, could explain the increase in tumor susceptibility observed in these rodents as compared to nulliparous Snai2 KOERBB2+ mice (Supplementary Figure S2h).

Parous Snai2 KOERBB2+ mice display a delay in mammary gland involution

We next evaluated the histology of the mammary glands of parous mice. The observation of the mammary glands of Snai2 KOERBB2+ mice at thirty days post-lactation revealed the presence of more residual ducts with proteinaceous material inside than in the mammary glands from the wild-type mice, suggesting a delay in involution (Figures 5a and 5b). Mammary gland involution is a complex process in which the lactating gland returns to a previous morphologically pre-pregnant state. This process is characterized by substantial apoptosis in epithelial cells, the regeneration of adipose mammary tissue, and global tissue remodeling.[42] Although remains of the lactating mammary gland can be found as vestiges in normal involuted breast tissue,[43] the mammary glands of Snai2 KOERBB2+ mice showed vestiges much more frequently than the Snai2 WTERBB2+ glands (Figures 5a and 5b).

Figure 5

SNAI2 is necessary for normal mammary gland involution

(A) Mammary glands from parous mice at 30 days post-lactation (hematoxilin-eosin). (b) Quantification of intact mammary ducts: 10 fields were counted per mammary gland from five individual mice of each genotype. (c) Mammary glands from parous mice at three days post-lactation (hematoxilin-eosin). Arrowheads in Snai2 WT mice indicate an apoptotic body. (d) SNAI2 expression at 3 days post-lactational involution (X-Gal staining and hematoxilin-eosin). (e) Cleaved CASPASE-3 in mammary glands from parous female mice at 3 days post-lactation involution (immunohistochemistry). (f) Quantification of CASPASE-3 activity per gland from parous female mice by ELISA (fluorometric assay). (g) pSTAT3(Y705) in mammary glands from parous mice at 3 days post-lactation (immunohistochemistry). Arrowheads indicate positive cells. (h) Quantification of pSTAT3(Y705) positive cells in parous mice at three days post-lactation (10 fields were counted per gland in three different mice). (i) Detection of pAKT1(S473) and pSTAT3(Y705) in organoids from mammary glands at 3 days post-lactation. Each number corresponds to an individual mammary gland from four individual mice.

In light of the foregoing, we evaluated the delay in the involution of the mammary gland in greater detail. We first assessed the global morphology of the mammary gland in the Snai2 KOERBB2+ mice immediately after lactation, at three days of involution. In whole-mount evaluations, we observed smaller hypoplastic glands at necropsy (Supplementary Figure S4a), with a less dense branching pattern. This was particularly evident at the periphery of the gland, indicating a less hyperplastic mammary gland. This permitted the lymph node to be visualized in 6 out of 11 mammary glands, whereas the node was visible only in 1 out of 10 mammary glands from independent parous Snai2 WTERBB2+ female mice (Supplementary Figures S4b and S4c). This result is in agreement with a recent report that suggested that Snai2 deficiency affects mammary gland development during pregnancy.[34] Despite this, as indicated above, in the mammary glands of Snai2 KOERBB2+ mice we observed more residual ducts than in the mammary glands from the wild-type mice due to a delay in involution (Figures 5a and 5b).

In addition, the mammary glands of the Snai2 KOERBB2+ mice also showed histopathological features of delayed involution at three days post-lactation, exhibiting fewer apoptotic cells inside, as well as a lower overall staining of active-CASPASE-3 as compared with their wild-type counterparts (Figures 5c-f). Importantly, the delay in the involution process was already present in non-transgenic parous Snai2 KOWT mice (Supplementary Figures S5a-c), indicating that SNAI2 was necessary for normal mammary gland involution. This is in agreement with the broad expression of SNAI2 at three days of post-lactational involution in the mammary gland observed here (Figure 5d).

SNAI2 is required for STAT3 activation during post-lactational involution of the mammary gland

To study the molecular mechanism that led to a defect in the involution of the mammary gland under Snai2 deficiency, we studied the levels of some targets primarily involved in the involution process. Activation of STAT3 is essential for triggering post-lactational involution; pSTAT3 in turn interferes with the PI3K pathway, leading to lower levels of pAKT1 and apoptosis in mammary epithelial cells.[44-46] Here we show that mammary glands from parous Snai2 KOERBB2+ mice had significantly lower levels of pSTAT3(Y705), together with higher levels of pAKT1(S473) than their wild-type counterparts at three days of post-lactational involution. This could explain the delay in mammary gland involution in the Snai2 KOERBB2+ mice (Figures 5g-i). The delay in the involution process already present in non-transgenic parous Snai2 KOWT mice also showed the same defect in STAT3 activation, indicating that this was specifically due to the SNAI2 deficiency (Supplementary Figures S5d and S5e). We did not observe differences in the baseline levels of pAKT1 or pSTAT3 in the mammary glands of nulliparous mice (Supplementary Figures S6a and S6b).

We observed that the defect in mammary gland involution in parous Snai2 KOERBB2+ mice was still detectable at thirty days post-lactation (Figure 5a). Accordingly, we analyzed the ERBB2 status and found that there was more ERBB2 expression, and more ERBB2+ cells at 30 days post-involution in parous Snai2 KOERBB2 mice, probably as consequence of a defect in elimination during the involution process (Figures 6a and 6b). These results could explain the later increased latency and tumor numbers. Furthermore, the highest levels of pAKT1(S473) were also detected at 30 days post-involution, together with abnormally high levels of pSTAT3 in the mammary glands of parous Snai2 KO mice (Figure 6c). The activation of both AKT1 and STAT3 at 30 days is probably a consequence of the presence of more ERBB2+ cells. Indeed, STAT3 expression is correlated with ERBB2 expression in human tumors.[47, 48]

Figure 6

Downregulation of Snai2 in HC11 cells leads to an increase in phosphorylation of AKT1 and a decrease in pSTAT3

(a) Detection of ERBB2 from parous mammary glands at 30 days post-lactation by western blot. Each number corresponds to an individual mammary gland from an individual mouse. (b) ERBB2 expression in mammary glands from parous mice at 30 days post-lactation (immunohistochemistry). Arrowheads indicate complete membrane stain-positive cells. (c) Detection of AKT and pSTAT3 from parous mammary glands at 30 days post-lactation by western blot. Each number corresponds to an epithelial mammary gland from an individual mouse (N = 4). pAKT and pSTAT3 bands were quantified, and normalized intensities were calculated as averages from 4 individual mice. (d) HC11 cells transfected with Snai2 shRNA were treated with prolactin for three days; GFP-positive cells were sorted by flow cytometry, and AKT and pSTAT3 protein levels were assessed. (e) Levels of total and pAKT1 (S473) from tumor samples (ELISA). (f) Detection of pSTAT3 from nulliparous and parous tumors by western blot.

To further study the influence of SNAI2 levels in AKT1 and STAT3 activation, we used HC11 cells, a clonal epithelial cell line derived from a mid-pregnant mouse mammary gland capable of lactogenic differentiation under hormonal stimuli.[49] In this cell line, both prolactin and progesterone were able to induce the phosphorylation of AKT and increase the levels of SNAI2 (Figure 6d and Supplementary Figure S6d). Interestingly, after down-regulation of Snai2 by shRNA, the activation of AKT and the decrease in pSTAT3 were more evident (Figure 6d and Supplementary Figures S6c and S6d). Furthermore, this effect was also seen even in the absence of prolactin (Supplementary Figure S6c), suggesting that the downregulation of Snai2 would be sufficient to raise pAKT1(S473) and reduce pSTAT3(Y705) levels. In conclusion, SNAI2 participates in mammary gland involution, helping to activate STAT3 and to inhibit pAKT1. However, the mechanism by which SNAI2 exerts these effects remains unknown.

Snai2 deficiency modifies AKT1 levels in tumors from Snai2 KOERBB2+ mice through a non-cell autonomous mechanism

Since the Snai2 deficiency modified the levels of activated STAT3 and AKT1 in the mammary gland, we wondered whether the status of these molecules in the tumors was affected. Interestingly, tumors from parous Snai2 KOERBB2 mice showed significantly higher total and phosphorylated levels of AKT1 than tumors from parous Snai2 WTERBB2+ animals, which could contribute to increasing their relative local aggressiveness and the lower capacity to disseminate.[50, 51] However, we did not see differences in the levels of activated STAT3 (Figures 6e and 6f). Since SNAI2 is not expressed in tumors from Snai2 WTERBB2+ mice (Figure 1g and Supplementary Figure S1b), this effect on AKT1 levels would be exerted by a non-cell autonomous mechanism. In this sense, a recent report has described a non-cell autonomous effect of SNAI2 in skin cancer progression by promoting the recruitment of myeloid cells.[52] Accordingly, we evaluated myeloid infiltration in breast tumors and we did not see differences between the wild-type and KO mice or between nulliparous and parous mice (Supplementary Figure S6e). Thus, other studies will be needed to clarify the non-cell autonomous effect of SNAI2 deficiency in breast cancer development.

Discussion

In this work we found that Snai2 KOERBB2+ mice were partially protected from breast tumorigenesis, exhibited an extended latency and therefore lifespan, and had fewer tumors than their Snai2 WTERBB2+ counterparts. Since both disease pathophenotypes have their origin in pretumoral stages, and because SNAI2 is expressed in the mammary gland but not in the bulk of the tumor (Figure 1g and Supplementary Figure 1b), these results demonstrate the participation of SNAI2 in the early stages of luminal breast cancer development. Surprisingly, in this model of luminal breast cancer, Snai2 deficiency did not affect the metastatic capacity in these mice.

The increased oncogenic activity of ERBB2 after pregnancy compensated the lower local tumor susceptibility previously demonstrated in nulliparous Snai2 KOERBB2+ mice regarding tumor latency and the number of tumors; by contrast, it exacerbated the defect in metastatic capacity. These results suggest that a complete inhibition of SNAI2 (e.g. a pharmacological inhibition) could significantly raise the incidence of metastasis-free individuals in luminal breast cancer. This effect of SNAI2 deficiency on luminal tumor dissemination is in agreement with the fact that a minority of cells inside luminal tumors express basal epithelial genes, and proves that targeting the basal invasive program in luminal tumors could limit metastatic progression.[53] However, we were unable to detect SNAI2 expression in the invasion front either (Supplementary Figure S1b).

Mice deficient in Snai2 showed a significantly decreased proliferation in the mammary epithelium of nulliparous mice. It has been reported that SNAI2 could be located downstream of ERBB2 in breast cancer cells and could contribute to ERBB2 tumorigenesis.[54] Thus, the absence of SNAI2 might contribute directly to reducing ERBB2 signaling. However, this low proliferation has also been reported in wound healing in Snai2 KO mice without the ERBB2 transgene,[55] indicating that this low proliferation could affect other tissues without the influence of ERBB2, at least in vivo. This effect of Snai2 deficiency on a lower proliferation rate could be explained in terms of the lower levels of CYCLIN D1 observed here (Figure 3e), in agreement with a previous report that SNAI2 increases CYCLIN D1 levels by inhibiting its ubiquitination.[56] Moreover, CYCLIN D1 is necessary for tumor development in MMTV-ERBB2 transgenic mice.[57, 58] It is possible that the low baseline proliferation in mammary glands under Snai2-deficient conditions would hinder the acquisition of secondary events for the complete transformation of the target cells already initiated by the first oncogenic event (i.e. the ErbB2 transgene), leading to longer tumor latency.[30, 59] Despite the function of SNAI2 as an inhibitor of apoptosis,[5, 60] nulliparous Snai2 KOERBB2+ mice showed lower levels of baseline apoptosis than their wild-type counterparts (Figure 3f), in agreement with the effect of Snai2 deficiency on the skin observed in a recent report.[55]

Interestingly, the nulliparous Snai2 KOERBB2+ mice showed a lower capacity to generate mammospheres, but these maintained their ability to expand in vitro, suggesting a smaller number of functional stem cells in vivo but a normal propagation capacity ex vivo (Figures 3g and 3h). Thus, we cannot exclude the possibility of the existence of an extrinsic effect of Snai2 deficiency through other compartments, such as the local niche where stem cells reside. We did not observe differences in mammosphere formation in Snai2 KOWT non-transgenic mice (Supplementary Figure S3e), which is partially in agreement with a recent report[9] describing some differences in what the authors referred to as microspheres. Thus, the differences in the mammosphere yield observed here in Snai2 KO mice were only present in the context of ERBB2 overexpression. In this sense, ERBB2 has also been related to mammosphere formation and stem-cell expansion in mice.[61-64] In addition, CYCLIN D1 is required for the self-renewal of mammary stem cells, which are targets of MMTV-ErbB2 tumorigenesis.[65] It is possible that the observed defect in CYCLIN D1 (Figure 3e) could contribute to the lower mammosphere yield in nulliparous Snai2 KOERBB2+ mice. Finally, lower numbers of mammospheres and cancer-initiating cells have been described in MMTV-ErbB2 mice deficient in IKK-alpha, these animals also showing a reduction in mammary tumor incidence and multiplicity.[64] In conclusion, we found a positive association between the number of mammospheres, and therefore functional mammary stem cells, and the number of tumors originating in the ErbB2 transgenic mice. Furthermore, in response to pregnancy the Snai2 KOERBB2+ mice compensated the defect in the number of mammospheres observed in their nulliparous counterparts. It is possible that the defect in CYCLIN D1 (Figures 3e and 4b) could contribute to the lower mammosphere yield in nulliparous Snai2 KOERBB2+ mice but that it could somehow be compensated in parous Snai2 KO ERBB2+ mice.

We later observed that SNAI2 was necessary for normal mammary gland involution. Here again, the lower apoptosis observed during the involution process in Snai2 KO mice is paradoxical, considering the function described for SNAI2 as an inhibitor of apoptosis in other systems.[5, 11, 60] The relative expansion of mammospheres, and hence mammary stem cells in parous Snai2 KOERBB2 mice, could thus be explained in terms of the defect in post-lactational involution. There are many examples of mouse models where delayed involution has been associated with greater subsequent tumor susceptibility, due to a defect in the elimination of cell populations susceptible to transformation.[66] Remarkably, in the transgenic mouse model used in this work,[16] a specific defect in elimination during the involution of tumor cells that form in situ carcinomas, but normal elimination of mammary cells surrounding tumoral ones have been reported. [67] This indicates that the already fully-transformed cells may not be eliminated easily after post-lactational involution. Although there were more ERBB2-positive cells in the mammary glands of parous Snai2 KO mice than in the Snai2 WT animals at 30 days post-lactational involution, both showed the same capacity to generate mammospheres (Figure 4d). This would suggest the same number of target stem cells, in agreement with the fact that there were no differences in terms of latency and tumor numbers.

It has been reported that overexpression of activated AKT1 leads to short tumor latency in MMTV-ErbB2 transgenic mice.[50] Here, we report that the mammary glands from parous Snai2 KOERBB2+ mice exhibited higher levels of pAKT1 than parous Snai2 WTERBB2+ mice, which could explain the tumor latency compensation seen after pregnancy (Figure 2a). Moreover, this increase in pAKT1 levels could contribute to the observed delay in mammary gland involution,[45, 46] which could lead to more target cells becoming available. pAKT1 levels would be increased by pSTAT3 through the regulation of PI3K, modifying the levels of some regulatory subunits;[46] as reported in a recent paper, this could also be performed directly by SNAI2 in colon cancer cells.[68]

In conclusion, we have demonstrated that SNAI2, although not expressed in the tumors, is important for the pathogenesis of breast cancer of luminal origin. This demonstrates the usefulness of SNAI2 inhibition for therapy or prevention in luminal breast cancer and for preventing metastasis. We also found that tumors from parous Snai2 KOERBB2+ mice had higher levels of total and phosphorylated AKT1, suggesting that this alteration in AKT1 levels would be exerted by a non-cell autonomous mechanism. It remains to be clarified how SNAI2 deficiency in other local or long-distant compartments might also contribute to the pathogenesis of luminal breast cancer. Finally, the connection observed here, as in other models, between a defect in mammary remodeling and increased susceptibility to breast cancer highlights the possibility of modifying this physiological process to develop preventive strategies against breast cancer.[66]

Materials and Methods

Animals

All mice were housed at the Animal Research Facility of the University of Salamanca. All practices were accepted by the Institutional Animal Care and Bioethical Committee of the University of Salamanca. FVB/N-Tg(MMTVneu)202Mul/J, which are FVB mice carrying the ErbB2 protooncogene under the control of the Mouse Mammary Tumor Virus (MMTV) 3′promoter,[16] were obtained from the Jackson Laboratory (Bar Harbor, ME, USA), and wild-type (WT) FVB/N mice were purchased from Charles River (Wilmington, MA, USA). B6;129S1-Snai2/J Snai2-deficient mice were generated by Dr. Gridley (Scarborough, ME, USA).[18] Mice were weaned at 3-4 weeks of age and analyzed for the presence of the ErbB2 transgene. The initial F1-Snai2-deficient mice were generated by mating C57BL/6 Snai2 KO males with MMTV-ErbB2 transgenic females in a FVB genetic background. To avoid the C57BL/6-resistance effect in breast cancer development,[19-21] F1 transgene-positive/Snai2-deficient males were backcrossed with FVB females to obtain the F4 generation with a 90% FVB genetic background;[72] mice were intercrossed to obtain Snai2 WTERBB2+ and Snai2 KOERBB2+ Because Snai2 KO mice are born at lower than the expected frequency and show perinatal mortality, only those KO mice that surpassed five weeks of life were included in the experiments. To generate the parous cohort, Snai2 WT ERBB2+ and Snai2 KOERBB2+ female mice were bred twice, and after the second lactation the cohort was followed for mammary tumor formation. All mice were maintained in ventilated filter cages under SPF conditions and were fed ad libitum. We estimated the number of mice based on Statgraphics Centurion software (Statpoint Technologies, Warrenton, VA, USA); thus, a minimum of total sample size of 23 animals per group was necessary to provide a 90% chance of detecting difference ≥ 1.5 standard deviations between the two populations means with 95% confidence. Taking this into account, we adapted the sample size based on the results of our previous study.[27] The usual mammary gland studies were performed in mice after a single pregnancy and lactation. All analyses were performed by certified pathologists (M.C. from the CNIO and an additional pathologist from the Comparative Pathology Core Facility) blinded to sample identity.

Protein analyses

The epithelial mammary gland component (organoids) and tumors were collected at necropsy, snap-frozen in liquid nitrogen and kept at -80 °C. Proteins were extracted from frozen tissues. Ceramic beads (Precellys, Bertin Technologies, #03961-1-009, Montigny le Bretonneux, France) were added to tissues (10-50 mg) and these were homogenized for 10 seconds, 5.5 m/s (twice), using the FastPrep Homogenizer (Thermo Savant, Thermo Fisher Scientific Inc., Waltham, MA, USA) in RIPA buffer (150 mM NaCl, 1% (v/v) NP40, 50 mM Tris-HCl at pH 8.0, 0.1% (v/v) SDS, 1 mM EDTA, 0.5% (w/v) Deoxycholate) containing cocktails of protease and phosphatase inhibitors (Roche, Basel, Switzerland) for tumors and organoids. Samples were incubated for 20 minutes on ice and centrifuged for 10 min at 13000 r.p.m. and 4°C. Supernatants were collected and quantified using the BCA Protein Assay Kit (Thermo Fisher Scientific Inc.,catalog number 23228, Waltham, MA, USA). Equivalent amounts of proteins were resolved by SDS-PAGE, and transferred to polyvinylidene difluoride membranes (Immobilon-P, Millipore, Darmstadt, Germany). Immunoblotting was performed using the following primary antibodies: anti-phospho-AKT(Ser473) (D9E, #3787), anti-AKT (11E7, #4685), anti-AKT1 (2H10, #2967), anti-SNAI2 (#9585), anti-pSTAT3(Y705) (#9145) from Cell Signaling (Danvers, MA, USA); anti-ERBB2 (ab2428), anti-phospho-ERBB2(Tyr1248) (ab47755) from Abcam (Cambridge, United Kingdom); anti-TUBULIN (DM1A, #T6199, Sigma, St Louis,MO, USA), anti-ACTIN (C4, sc-47778, Santa Cruz, Dallas, TX, USA), anti-GAPDH (sc-25778, Santa Cruz, Dallas, TX, USA), and anti-CYCLIN D1 (C-20, sc-717, Santa Cruz, Dallas, TX, USA), anti-Estrogen Receptor alpha (ab39642, Abcam, Cambridge, United Kingdom) and subsequently with horseradish peroxidase-conjugated anti-mouse, anti-rabbit or anti-goat secondary antibodies (1:10000 dilution) (BIO-RAD, Berkeley, CA, USA), and visualized by enhanced chemiluminescence (Thermo Scientific). A total amount of protein of 100 μg from the organoids was assayed for active CASPASE-3 using the fluorometric CASPASE 3 Assay Kit (ab39383, Abcam) according to the manufacturer's instructions. For ELISA assays, the levels of phosphorylated and total AKT1 were measured using Sandwich ELISA Kits (#7142 and #7143, respectively, PathScan, Cell Signaling).

Mammary cell preparation and culture

Mammary glands were dissected out from female mice at the ages specified along the manuscript. After mechanical dissociation with two scalpels (Swann-Morton, #0511, Sheffield, United Kingdom), the tissue was placed in culture medium (DMEM/F12, 10% FBS, 1% penicillin/streptomycin, 1% Glutamine) containing 600 units/ml collagenase (Sigma, C2674, St Louis, MO, USA) and 200 units/ml hyaluronidase (Sigma, H3506, St Louis, MO, USA) and digested for 3 hours at 37° C for the mammosphere protocol, and 6 hours for FACS analysis. Organoid suspensions were obtained by differential centrifugation.[69] Mammosphere formation was achieved as previously described[70] with the following modifications: 10000 cells were seeded per well in triplicate on ultralow attachment plates (Corning, #3471, Corning, NY, USA), and after 7 days of culture only mammospheres with a diameter greater than 50 μm were counted. The mammosphere medium was prepared as previously described.[71]

X-Gal staining

For whole mount X-gal staining, Snai2 KOERBB2+ mice, which carry the LacZ gene in the Snai2 locus, [18] and Snai2 WTERBB2+ mice were sacrificed and the fourth mammary gland or tumor samples were harvested and processed with the β-Galactosidase Staining kit according to the manufacturer's instructions (#9866 Cell Signaling Technology).

Mammary gland whole-mounts

The fourth (abdominal) mammary glands were removed surgically, stretched onto a glass slide and fixed with Carnoy's fixative (ethanol / chloroform / glacial acetic acid, at proportions 6/3/1, respectively) for 24 hours at room temperature. Then, they were rehydrated with a graded series of ethanol solutions, washed for 5 minutes in distilled water and stained overnight in alum carmine solution (Sigma, C1022). The following day, the mammary glands were dehydrated with a graded series of ethanol solutions. Mammary glands were cleared in xylene until the fat had been sufficiently removed from the glands. Images were analyzed using ImageJ software (National Institutes of Health, Bethesda, MD, USA).

Immunostaining

Mammary glands or tumors were fixed in 4% paraformaldehyde overnight at room temperature and washed in 70% EtOH before paraffin inclusion. Sections of 3 μm were deparaffinized and processed for antigen retrieval by incubation in a microwave with citrate buffer (pH 6) for cleaved CASPASE-3 and ERBB2, but at pH 8 for pSTAT3. The primary antibodies were incubated using the Ventana Discovery automated immunohistochemistry research slide-staining system (Tucson, AZ85755) with the following antibody dilutions: 1/50 Ki-67 (MAD020310Q, Master Diagnostica, Granada, Spain), 1/50 cleaved CASPASE-3(Asp175) (#9662, Cell Signaling), 1/75 ERBB2 (ab2438, Abcam), 1/50 pSTAT3(Y705) (D3A7, #9145, Cell Signaling), 1/50 CK6 (SPM269, ab75703, Abcam), 1/100 CK8 (ab59400, Abcam) and 1/100 Progesterone Receptor (Y85, MAD210302Q, Master Diagnostica). Ki-67-, CASPASE-3- and ERBB2-positive cells from tumors were quantified with the ARIOL software at the Centro Nacional de Investigaciones Oncológicas (CNIO, Madrid, Spain). To accomplish this, tumors from Snai2 WTERBB2 and Snai2 KOERBB2 nulliparous and parous mice (10 tumors from each group) were stained in triplicate on tissue-array sections.

Cell culture and retroviral infection

The HC11 cell line was obtained from Dr. Montoliú (Centro Nacional de Biotecnología, Spain) in February 2014. Experiments were performed during the ensuing 3 months. The cells showed the correct phenotype, with the expected morphology and growth curves; and were previously tested for mycoplasma contamination. HC11 cells were cultured with RPMI-1640 medium with 10% heat-inactivated FCS, 5 μg/ml insulin, 20 ng/ml epidermal growth factor, and 1% penicillin-streptomycin. HC11 cells were infected with retrovirus carrying Snai2 shRNA or a control.[72] Recombinant retroviruses were produced from Phoenix-A cells, which were transiently transfected with plasmid DNA (shRNA Snai2) using LT-1 reagents (Mirus, Madison, WI, USA). HC11 cells were infected in the presence of polybrene (30 μg/ml). Two days later, the medium was replaced with fresh virus-containing medium and the cells were allowed to grow to confluence for three days in the presence or absence of prolactin at 5 μg/ml or progesterone at 10-8 M (Sigma).

Statistical analyses

Mammary tumor latency, disease duration and lifespans were compared in the Snai2 WTERBB2+ and Snai2 KOERBB2+ female mice using the Kaplan-Meier estimator and the Log-Rank test. The number of tumors, the number of lung metastases and other determinations were evaluated with the Mann-Whitney U test for two groups, and with the Kruskal-Wallis test more than two groups, followed by Dunn's multiple comparison post-test. Incidence and multiplicity were evaluated using Fisher's exact tests. For all analyses, P values ≤ 0.05 were considered statistically significant. The analyses were performed using the SPSS (IBM, Armonk, NY, USA) and JMP/SAS (SAS Institute, Cary, NC, USA) statistical packages.

Supplementary Table S1, Supplementary Figure Legends and Supplementary Materials and Methods

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Figure S5

Supplementary Figure S6