p63 suppresses the ability of pregnancy-identified mammary epithelial cells (PIMECs) to drive HER2-positive breast cancer

While pregnancy is known to reduce a woman’s life-long risk of breast cancer, clinical data suggest that it can specifically promote HER2 (human EGF receptor 2)-positive breast cancer subtype (HER2+ BC). HER2+ BC, characterized by amplification of HER2, comprises about 20% of all sporadic breast cancers and is more aggressive than hormone receptor-positive breast cancer (the majority of cases). Consistently with human data, pregnancy strongly promotes HER2+ BC in genetic mouse models. One proposed mechanism of this is post-pregnancy accumulation of PIMECs (pregnancy-identified mammary epithelial cells), tumor-initiating cells for HER2+ BC in mice. We previously showed that p63, a homologue of the tumor suppressor p53, is required to maintain the post-pregnancy number of PIMECs and thereby promotes HER2+ BC. Here we set to test whether p63 also affects the intrinsic tumorigenic properties of PIMECs. To this end, we FACS-sorted YFP-labeled PIMECs from p63+/−;ErbB2 and control p63+/+;ErbB2 females and injected their equal amounts into immunodeficient recipients. To our surprise, p63+/− PIMECs showed increased, rather than decreased, tumorigenic capacity in vivo, i.e., significantly accelerated tumor onset and tumor growth, as well as increased self-renewal in mammosphere assays and proliferation in vitro and in vivo. The underlying mechanism of these phenotypes seems to be a specific reduction of the tumor suppressor TAp63 isoform in p63+/− luminal cells, including PIMECs, with concomitant aberrant upregulation of the oncogenic ΔNp63 isoform, as determined by qRT-PCR and scRNA-seq analyses. In addition, scRNA-seq revealed upregulation of several cancer-associated (Il-4/Il-13, Hsf1/HSP), oncogenic (TGFβ, NGF, FGF, MAPK) and self-renewal (Wnt, Notch) pathways in p63+/−;ErbB2 luminal cells and PIMECs per se. Altogether, these data reveal a complex role of p63 in PIMECs and pregnancy-associated HER2+ BC: maintaining the amount of PIMECs while suppressing their intrinsic tumorigenic capacity.


Introduction
HER2 (human epidermal growth factor receptor 2)positive breast cancer (HER2+ BC) comprises 15-20% of all sporadic breast cancers and is an aggressive subtype. It is characterized by gene amplification/protein upregulation of HER2 receptor tyrosine kinase and presents with reduced survival and high rate of relapse after chemotherapy, due to enhanced cell proliferation, angiogenesis, metastasis, and reduced apoptosis 1 . HER2+ BC is frequently diagnosed at widely metastatic stage III/IV and in younger patients 2 . Although HER2-targeted therapies (Trastuzumab, Lapatinib) have greatly improved management of this malignancy, there is a significant rate of primary and acquired resistance 3,4 , urging to identify additional factors that contribute to HER2+ BC pathogenesis and survivorship.
In contrast to other subtypes, HER2+ BC seems to be associated with pregnancy. Thus, the incidence of HER2+ BC among so called pregnancy-associated breast cancers is increased to 28-58%, compared to 16-22% of age-matched control patients or 19% in the general population of reproductive age [5][6][7][8][9][10] . Moreover, parity can increase the life-long risk of HER2+ BC. Indeed, two large studies (2710 and 28,095 patients, respectively) found a significant association of parity with the risk of HER2+ BC, whereas ≥3 full-term pregnancies had an even greater association 11,12 . In agreement, parity accelerates tumor onset and mortality in two genetic mouse models of HER2+ BC, the ErbB2 and Neu mice (with constitutively active and amplified HER2, respectively) [13][14][15] . Compelling explanation for this came from mouse studies showing that a major cancer stem cell/ tumor-initiating cell population for HER2+ BC are PIMECs (pregnancy-identified mammary epithelial cells) 13,16 . PIMECs are multipotent alveolar progenitors that comprise 0.8-4% of mammary epithelial cells (MECs) in virgins, but undergo enormous expansion in late pregnancy to give rise to essentially all milk-producing alveoli [17][18][19] . Importantly, PIMECs-unlike the rest of MECs-are largely resistant to apoptosis during post-lactation gland involution and now comprise 20-30% luminal cells and greatly contribute to gland expansion in subsequent pregnancies 17,[19][20][21] . This significant increase in PIMECs content in parous females is likely the basis for their increased susceptibility to HER2 tumorigenesis. Of note, even in virgin mice, all HER2driven tumors arise from lineage-traced PIMECs, highlighting their bone fide role as tumor-initiating cells for HER2+ BC, at least in mice 13,16 .
We previously showed that p63, an epithelial master regulator and a homologue of the tumor suppressor p53, is a critical novel regulator of PIMECs and pregnancyassociated HER2+ BC 15 . Specifically, mammary glands from heterozygous p63+/− females (homozygous p63−/− animals die perinatally 22,23 ) exhibit enhanced apoptosis in post-lactation gland involution, mediated by Oncostatin M/ Stat3 and reduced Neuregulin/Stat5 signaling 15 . Moreover, the post-involution p63+/− mammary glands contain on average 40% fewer PIMECs than p63+/+ glands 15 , suggesting that p63 is required to maintain the PIMECs pool. Consistently, p63+/−;ErbB2 females are partially protected from HER2+ BC (which is not observed in virgins), pointing to the reduced PIMEC content as the likely underlying mechanism 15 . Since p63 is a known regulator of normal epithelial stem cells and cancer stem cells [24][25][26][27][28] , here we set to test whether, besides the PIMECs content, p63 also regulates their intrinsic tumorigenic properties. Surprisingly, we found that p63 does play a role, but as a tumor suppressor rather than an oncogene. This seems to be due to an interplay between p63 isoforms: tumor suppressor TAp63 and oncogenic ΔNp63.
photographed at ×20 magnification, and the percent of Ki67-positive cells was calculated. For the in vivo proliferation assay, mammary gland #4 from p63+/+ and p63 +/− sisters were flattened on filter paper, fixed in formalin, embedded in paraffin, and sectioned (5 μm), followed by immunofluorescent staining as described below. Ten random non-overlapping fields were photographed at ×20 magnification, and the percent of Ki67-positive cells within YFP-positive cells was calculated.

Single-cell RNA sequencing (scRNA-seq)
Cell suspensions of p63+/+ and p63+/− PIMECs obtained as described above from four pooled p63+/+ or three pooled p63+/− mice, respectively, were loaded on a 10x Genomics Chromium instrument to generate singlecell gel beads in emulsion (GEMs). Approximately 45,000 cells of each genotype were loaded per channel. ScRNAseq libraries were prepared using the following kits: Chromium Next GEM Single Cell 3′ GEM, Library & Gel Bead Kit v3.1, PN-1000121; Chromium Next GEM Chip G Single Cell Kit, PN-1000120 and Single Index Kit T Set A PN-1000213 (10x Genomics) as described 31 and following the User Guide (manual part #CG000204 Rev D). Libraries were run on an Illumina NovaSeq 6000 pairedend reads, read 1 is 28 cycles, i7 index is 8 cycles, and read 2 is 91 cycles, one lane per sample, for approximately >57% and >48% sequencing saturation (p63+/+ and p63 +/−, respectively). The Cell Ranger Single Cell Software Suite (v1.3) was used to perform UMI processing and single-cell 3′ gene counting. Cell Ranger (v5.0.1), Loupe Browser (v5.0.0), and R package Seurat (v4.0.0) were used to visualize gene expression and find significantly altered genes. For best cell type identification and separation, we chose K-mean clustering (k = 8) and UMAP for dimensionality reduction. Gene expression analysis of luminal cells (Figs. 3b-e, 4, 5a) was performed using Loupe browser and the Significant Feature Comparison/Locally Distinguishing function. Gene expression analysis of PIMEC cells (Fig. 5b-e) was performed using Seurat, where exogenous YFP gene was added to identify PIMECs. The control and experimental samples were processed with "cellranger count", followed by reaggregation into a combined dataset using "cellranger aggr". Again, UMAP was used (resolution = 0.015), and differentially expressed genes between p63+/− and p63 +/+ PIMECs were found by comparing YFP/Krt8 doublepositive cells (expression > 0), using FindMarkers function. Wilcoxon rank-sum test and Bonferroni corrections were used to calculate adjusted p-values.

Statistical analysis
Mouse tumor-free survival was analyzed by Kaplan-Meier analysis and log rank statistics, and the p value and the hazard rate were determined using online software (https://www.evanmiller.org/ab-testing/survivalcurves.html). Tumor size and the mammosphere formation efficiency were analyzed by unpaired two-tailed Student's t-test, p < 0.05 was considered statistically significant. No animal randomization was used. No blinding was used. No statistical method was used to predetermine sample size. Normal distribution of data and data variation was not assessed.
Increased self-renewal and proliferation capacity of p63 +/− PIMECs Cancer stem cells are characterized by the ability to give rise to the bulk of the tumor and-similarly to normal stem/progenitor cells-to self-renew and give rise to floating spheres in suspension conditions 20,26 . We hypothesized that the more aggressive nature of p63+/− PIMECs in transplantation assays may be due to their increased self-renewal capacity. To test this, we used the 3D floating mammosphere forming assay, which is widely used to test stem cell activity in tissues, tumors, and cell lines 34 . Spheroids originate from rare cells with stem cell features able to grow in suspension and behaving as tumorigenic in mice 34 . To this end, we plated freshly isolated single-cell suspensions of p63+/+ and p63+/− PIMECs onto ultra-low-adherent 12-well plates (6500-17,500 cells per well, see Supplementary Table 1). This induced formation of mammospheres within 7-12 days, as previously reported 21 . As expected, the mammosphere formation efficiency (MFE) was significantly higher for p63+/− than p63+/+ PIMECs: 2.30 ± 0.62 vs. 1.44 ± 0.55, respectively (Fig. 2a, b). The average mammosphere size was not different between the genotypes (data not shown).
As another possible mechanism of the increased tumorigenic capacity of p63+/− PIMECs, we then assessed their proliferation potential in vitro and in vivo. To this end, we plated freshly isolated p63+/+ and p63 +/− PIMECs onto adherent 24-well plates and, 6-12 days later, stained them for the cell proliferation marker Ki67, frequently used in the clinic to assess tumor aggressiveness. As expected, p63+/− PIMECs had on average 1.8folds higher proliferation index than p63+/+ PIMECs (Fig. 2c, d). We also attempted to analyze Ki67 in the PIMECs of unperturbed mammary glands, i.e., in vivo, but found it problematic, likely due to high-fat content, as previously reported 35 . Nevertheless, a single p63+/+ and p63+/− sister pair again revealed a 1.8-fold higher PIMECs proliferation rate in the p63+/− compared to p63+/+ gland (Fig. 2e). In sum, these data uncovered a novel tumor suppressor-rather than oncogenic-role of p63 in PIMECs, which is associated with altered selfrenewal and proliferation ability.

Downregulation of TAp63 and upregulation of ΔNp63 in p63+/− luminal cells and PIMECs
The uncovered role of p63 as a tumor suppressor in isolated PIMECs is in sharp contrast with our previous report that implicated p63 as an oncogene in ErbB2overexpressing mammary glands 15 . The simplest explanation to unify these observations is that different assays reveal the roles of different p63 isoforms known to play opposite roles in cancer. Two major p63 isoforms are the full-length TAp63, a bona fide tumor suppressor similar to p53, and the N-terminally truncated ΔNp63, a bona fide oncogene [36][37][38] . We speculate that in intact p63+/− mice, ΔNp63-expressed exclusively in the basal cells 39-41maintains the number of post-pregnancy PIMECs, which are luminal cells 16,17 , in a non-cell-autonomous manner and thus, promotes HER2+ BC 15 (see "Discussion"). On the other hand, TAp63-expressed exclusively in the luminal cells [39][40][41] and therefore in PIMECs-cell-autonomously represses their tumorigenic properties, thus suppressing HER2+ BC (this study). To directly test this idea, first, we assessed the levels of TAp63 and ΔNp63 mRNAs by qRT-PCR in the luminal (CD24 high ) mammary cells in lieu of PIMECs, since very low PIMECs yields precluded their direct assessment by qPCR even upon combining mammary glands from several females (data not shown). We found that indeed, TAp63 levels were somewhat reduced in p63+/− luminal cells (Fig. 3a, left). Surprisingly, we also found upregulation of ΔNp63 in the luminal cells that normally is not expressed there (Fig. 3a, right). In order to gain deeper insight into the molecular underpinnings of p63+/− mammary gland, we then performed single-cell RNA sequencing (scRNA-seq) analysis on pooled CD24 pos (i.e., epithelial) mammary cells from p63+/+ (n = 4) and p63+/− (n = 3) mammary glands (Figs. 3b-e, 4, 5). Focusing on the PIMECs-enriched luminal cluster (Fig. 3b, orange), identified by the expression of Krt8, Krt18, Epcam, Gata3 etc. (Fig. 3c,  Supplementary Table 2), we assessed the expression of known TAp63 and ΔNp63 target genes. We found that among TAp63 targets, Casp1 was significantly downregulated, while Puma, Noxa, p21, Mdm2, Dicer etc 37,42-45 were not changed (Fig. 3d, e and data not shown). On the other hand, all significantly changed ΔNp63 targets were upregulated in p63+/− luminal cells (Fig. 3d, e). Subsequently, using more advanced software Seurat, we also found a significant downregulation of Casp1 in p63+/− PIMECs per se (Fig. 5e, left). On a side note, our conventional scRNA-seq (from the 3′ end) did not directly detect TAp63 and ΔNp63 isoforms, likely because they are N-terminal. Altogether, these data are consistent with the idea that the increased tumorigenic capacity of p63+/− PIMECs is due to reduced TAp63 and possibly, induction of oncogenic ΔNp63 in the luminal cells.
Using more advanced scRNA-seq analysis software, Seurat, we then zoomed in on the PIMEC cells (Fig. 5b-e). We found that the number of PIMECs was greatly reduced in p63+/−;ErbB2 compared to p63+/+; ErbB2 luminal cells (more than 10 folds, Fig. 5b-d), even more dramatically than we previously reported for non-ErbB2 glands (by 40%) 15 . Moreover, similarly to the overall luminal cells, p63+/− PIMECs were enriched in the cancer-associated and oncogenic Il-4/Il-13, HSP, TGFβ, FGF, NGF, and MAPK pathways 46,47,51,54 , the selfrenewal Notch and Wnt pathways 28,48,49 , overexpressed a stemness marker Aldh1a3, and had significantly decreased TAp63 target gene, Casp1 (Fig. 5e). The smaller number of significantly altered genes in each pathway is likely due to the smaller number of the PIMEC cells compared to the total luminal population, so that many genes did not reach statistical significance. Altogether, the  Fig. 3b, c) revealed by the Reactome pathway analysis. See Supplementary Fig. 1 for the full list of enriched pathways. b-e Significantly altered genes in the individual pathways (b, d) and UMAP plots of selected genes (c, e). Dark gray bars, genes whose UMAP plots are shown; pink, stemness and self-renewal pathways and genes. *p < 0.05, **p < 0.01, ***p < 0.001.
scRNA-seq analysis revealed broad pro-oncogenic changes in p63+/− luminal cells in general and PIMECs per se, thus providing a mechanistic explanation for their increased tumorigenic capacity.

Discussion
Here we uncovered a previously unknown role of p63 (specifically, TAp63) as a cell-autonomous tumor suppressor in PIMECs that curbs their intrinsic tumorigenic properties in parous ErbB2 females. Together with our previous report of an oncogenic role of p63 towards the PIMECs content 15 (likely due to oncogenic ΔNp63, which was not directly tested), this paints a more complete picture of a complex role of p63 in pregnancy-associated HER2+ BC (Fig. 6).
Besides the N-terminal p63 isoforms, several alternatively spliced C-terminal isoforms (α, β, γ etc.) are known and affect p63's activity 77 . However, we could not detect them by scRNA-seq, likely due to low expression and/or shared 3′ UTRs (data not shown). Moreover, it is unclear at present which C-terminal p63 isoforms are the most expressed and active in different MEC populations. Mechanistically, we found downregulation of TAp63 in the luminal cells (Fig. 3a) and downregulation of its target gene Casp1 in the luminal cells and PIMECs (Figs. 3d, e, and 5e, left), but no change in other known TAp63 targets. This is likely because different TAp63 target genes mediate its effects in different settings. Thus, TAp63 activates Puma and Noxa in the oocytes 78 . p21, usually activated upon DNA damage (which is not expected in the normal mammary gland), is mostly activated by TAp63β 79 , which is weakly expressed in the mammary gland 15 . And it is unclear whether Mdm2 is activated by TAp63, ΔNp63 or both 80 . Nevertheless, our data reveal a novel role of p63, likely TAp63, as an essential tumor suppressor in pregnancy-associated HER2+ BC and its tumor-initiating cells PIMECs.
Author contributions C.E.E. contributed to the study design, performed data acquisition, analysis and interpretation, critically read the manuscript, and approved its final version. J.L. contributed to the study design, performed data analysis, critically read the manuscript, and approved its final version. E.M.A. designed the study, performed data acquisition, analysis, and interpretation, wrote the manuscript, and approved its final version.

Funding
This work was supported by NIH/NCI grant K22CA190653 and Stony Brook Cancer Center startup to E.M.A. Fig. 6 A model integrating the roles of p63 isoforms in PIMECs and pregnancy-associated HER2-positive breast cancer. a In the post-parous whole mammary gland, ΔNp63 (expressed in the basal cells) non-cell-autonomously maintains the post-pregnancy PIMEC pool, whereas TAp63 (expressed in select luminal cells) has minimal or no effect. This results in a decreased PIMEC number and suppressed HER2 tumorigenesis in p63+/−;ErbB2 females 15 . b In isolated wildtype PIMECs, the tumor suppressor TAp63 is the main isoform and suppresses their intrinsic tumorigenic properties, as well as keeps at bay oncogenic ΔNp63. In p63+/−;ErbB2 PIMECs, TAp63 is reduced, resulting in increased tumorigenic properties, due to ΔNp63 upregulation and activation of oncogenic and self-renewal pathways. Yellow, PIMECs.

Ethics statement
No human tissues or samples were used. All mouse work was approved by the Stony Brook University Institutional Animal Care and Use Committee (IACUC), protocol 924666.