The secreted protease Adamts18 links hormone action to activation of the mammary stem cell niche

Estrogens and progesterone control breast development and carcinogenesis via their cognate receptors expressed in a subset of luminal cells in the mammary epithelium. How they control the extracellular matrix, important to breast physiology and tumorigenesis, remains unclear. Here we report that both hormones induce the secreted protease Adamts18 in myoepithelial cells by controlling Wnt4 expression with consequent paracrine canonical Wnt signaling activation. Adamts18 is required for stem cell activation, has multiple binding partners in the basement membrane and interacts genetically with the basal membrane-specific proteoglycan, Col18a1, pointing to the basement membrane as part of the stem cell niche. In vitro, ADAMTS18 cleaves fibronectin; in vivo, Adamts18 deletion causes increased collagen deposition during puberty, which results in impaired Hippo signaling and reduced Fgfr2 expression both of which control stem cell function. Thus, Adamts18 links luminal hormone receptor signaling to basement membrane remodeling and stem cell activation.

T he breast is the only organ to develop mostly after birth. Milk ducts arborize from the nipple and grow into a specialized subcutaneous stroma called the mammary fat pad in mice. The ductal wall comprises a bi-layered epithelium with the inner luminal cells and outer myoepithelial cells. The epithelium is separated from the stroma by specialized extracellular matrix (ECM), the basement membrane (BM). The ovarian hormones, estrogens and progesterone, are key drivers of mammary gland development and also influence breast carcinogenesis 1 . Both estrogen receptor α (ER) and progesterone receptor (PR) are members of the nuclear receptor family and are readily detected by immunohistochemistry (IHC) in a subset of luminal cells 1 . Activation of hormone receptor signaling in cells with high hormone receptor expression, termed sensor cells 2 , triggers the expression of paracrine factors such as amphiregulin and Rankl, which are required for mammary epithelial cell proliferation 3,4 as well as Wnt4 and Cxcl12, which activate stem/progenitor cells 5,6 . Mammary stem and progenitor cells have been identified and characterized based on cell surface markers and functional assays 7 . However, the precise cellular and biochemical components of the stem cell niche and its endocrine regulation remain poorly defined.
Evidence has been provided that mammary ECM can reprogram non-mammary cells to form mammary glands 8,9 , suggesting that it contains critical cues for epithelial development. Hedgehog signaling acts via Gli2 downstream of growth hormone receptor signaling in fibroblasts to trigger changes in paracrine signaling and ECM proteins that affect stem cell function 10 . This suggests that stromal fibroblasts are part of the niche under direct endocrine control by growth hormone. Stromal changes accompany different morphogenic processes induced by epithelial hormone signaling and are a hallmark of breast carcinogenesis. Indeed, high radiographic density, which reflects an increase in fibrillar collagen content in the breast stroma, is the single most important risk factor for breast cancer and correlates with progesterone exposure 11,12 . How ECM and stroma are controlled by the major endocrine drivers of breast development and carcinogenesis, epithelial ER and PR signaling, remains elusive.
ADAMTS18 is an orphan member of the A Disintegrin-like And Metalloproteinase domain with ThromboSpondin type 1 Motifs (ADAMTS) family of secreted Zn-dependent metalloproteinases 13 that comprises 19 members 14,15 . Like other zinc metalloproteinases, ADAMTS catalytic activity depends on zinc ion binding within the active site; unique to ADAMTSs is an ancillary domain containing thrombospondin type 1 repeats 16 . ADAMTS proteases are synthesized as precursors with an N-terminal propeptide, which is excised by pro-protein convertases such as furin 14 . Some ADAMTSs process ECM components such as fibrillar collagens, while others are implicated in turnover of the chondroitin sulfate proteoglycans aggrecan and versican 14 , and ADAMTS13 uniquely cleaves von-Willebrand factor to maturity 17 . We have previously reported that Adamts18 is required for eye, lung and female reproductive tract and kidney development in the mouse 18 . It is highly homologous to Adamts16, which has a role in renal development and fertility 19,20 and can cleave fibronectin 21 .
Here, we show that Adamts18 provides a mechanistic link between epithelial steroid hormone receptor signaling and changes in the ECM, in particular the BM, that regulate mammary epithelial stemness.

Results
Adamts18 expression is driven by the PR/Wnt4 axis. To elucidate the mechanisms, by which PR signaling in luminal mammary epithelial cells may elicit ECM changes, we sought genes induced in vivo by progesterone treatment 22,23 that fulfilled two criteria: (1) They encoded secretory proteins and (2) They showed delayed induction by progesterone as expected of any indirect PR target which is expressed by myoepithelial cells and can hence directly interact with the BM. Adamts18 induction was detected at 16 hours (h) and 78 h but not at 4 h 22 and at 24 h but not 8 h following progesterone stimulation 23 . RT-PCR analysis of fluorescence activated cell sorting (FACS)-sorted cells from adult mammary glands showed a 7-fold enrichment of Adamts18 mRNA in myoepithelial (Lin − CD24 + CD49f + ) over luminal (Lin − CD24 + CD49f − ) cells (Fig. 1a), in line with recent single cell RNA sequencing data 24,25 , confirming expression in myoepithelial cells.
Analysis of Adamts18 transcript levels at different stages of mammary gland development revealed low prepubertal expression that increased 2.7, 7-and 8.6-fold in 4-, 6-and 8-week-old females, respectively; expression rose further during pregnancy with a peak at mid-pregnancy day10.5/12.5 (Fig. 1b). RNAscope in situ hybridization for Adamts18 transcripts combined with immunofluorescence (IF) for the myoepithelial marker α-smooth muscle actin (Sma) confirmed myoepithelium-specific expression of Adamts18 in pubertal and adult mammary ducts (Fig. 1c, d). The increased Adamts18 expression during pregnancy was not attributable to generalized but rather to myoepithelium-specific upregulation of expression (Fig. 1e). Thus, Adamts18 expression in the mammary epithelium is developmentally regulated, and its mRNA is enriched in myoepithelial cells, making it an attractive candidate to mediate ECM changes downstream of epithelial hormone action.
Next, we tested whether endocrine factors contribute to developmental Adamts18 expression. First, we mimicked pubertal estrogen stimulation by injecting ovariectomized 21-day-old mice with 17-β-estradiol. Within 18 h of injection, Adamts18 transcript levels in extracts from total mammary glands increased 1.76-fold (Fig. 1f). Second, we asked whether changes in progesterone levels as they occur during estrous cycles affect Adamts18 transcript levels and obtained mammary gland extracts from mice in estrus and diestrus. Progesterone plasma levels determined by liquid chromatography-mass spectrometry were on average 2.8-fold higher in diestrus than in estrus (Fig. 1g); Adamts18 transcript levels in the mammary glands were 1.6-fold higher in diestrous over estrous (Fig. 1h). Thus, physiological Adamts18 expression correlates with plasma progesterone levels, suggesting that it is progesterone-responsive. The subtle increases in transcript levels are consistent with myoepithelial cells representing a minor fraction of the mammary cell types and hence of the total RNA in the whole tissue extracts we analyzed.
To determine whether epithelium-intrinsic PR signaling is required for Adamts18 mRNA expression, mammary epithelia from WT.EGFP + and PR −/− .EGFP + mice were grafted to contralateral fat pads of WT recipients surgically cleared of the endogenous epithelium and allowed to grow out for six weeks. At sacrifice, reconstitution was validated by fluorescence stereomicroscopy of the engrafted glands. Adamts18 transcript levels in the mammary glands successfully reconstituted with PR −/− epithelium were on average 27% of those in the contralateral controls (Fig. 1i). Thus, epithelial PR expression is required for Adamts18 mRNA expression.
Wnt4 is a plausible candidate to induce Adamts18 expression in myoepithelial cells because it is a PR target 26 transcribed exclusively in PR+ luminal cells 6 and activates canonical Wnt signaling in the myoepithelial cells 6 , which express Adamts18. We analyzed expression of various Wnt signaling components expressed in the mammary epithelium by RT-PCR in contralateral glands engrafted with WT.EGFP + and PR −/− .EGFP + mammary epithelia. Among the Wnt genes, only Wnt4 transcript levels were significantly lower in the mutant grafts, furthermore the transcript levels of the stem cell marker Lgr5 and the Wnt coreceptor Lpr6 were decreased (Fig. 1j). Consistent with canonical Wnt signaling activation downstream of PR/Wnt4 controlling Adamts18 expression, TCF4 binding sites were reported in the Adamts18 promoter by ChIP-seq analysis 27 . To assess whether canonical Wnt signaling controls Adamts18 expression in vivo, we analyzed Adamts18 expression in mammary glands with hyperactive canonical Wnt signaling in the myoepithelium 6 due to the presence of an MMTV-Wnt1 transgene 28

. Ectopic
Wnt1 expression was readily detected in transgenic glands and expression of the canonical Wnt signaling target, Axin2, was increased 5-fold over the non-transgenic control while Adamts18 mRNA levels were increased 7-fold (Fig. 1k). RNAscope for Adamts18 transcripts combined with IF for Sma showed the increased expression specifically in myoepithelial cells (Fig. 1l).
To test whether Wnt4 was furthermore required for Adamts18 expression, we engrafted contralateral cleared fat pads with WT.  EGFP + and Wnt4 −/− .EGFP + mammary epithelia and harvested the transplanted glands on day 8.5 of pregnancy when Wnt4dependent canonical Wnt signaling activity peaks 6 . Levels of Wnt4 expression in the mutant grafts were 1% of WT levels and Adamts18 expression was reduced to 35% of WT levels (Fig. 1m). Thus, increased canonical Wnt signaling induces Adamts18 expression and both PR and Wnt4 are required for Adamts18 mRNA expression. This indicates that myoepithelial Adamts18 expression is downstream of the luminal PR/Wnt4 axis. Potentially, our conclusion could be confounded by lineage differentiation and cell specification defects resulting from PR and Wnt4 deletions. In light of the finding that both PR −/− and Wnt4 −/− epithelial cells can differentiate into milk secreting alveolar cells 29,30 , major cell specification defects are improbable. Nevertheless, we examined the possibility of a lineage differentiation defect by determining the ratio of luminal and myoepithelial cells in the two mutants. FACS analysis of lineage-depleted WT and PR −/− mammary cells showed no significant difference in the two cell lineages ( Supplementary Fig. 1a). As the Wnt4 −/− mice die on embryonic day 13, we resorted to transplanting WT.EGFP + and Wnt4 −/− .EGFP + mammary epithelia derived for embryonic mammary buds 6 to contralateral fat pads and quantified the percentage of Sma+ epithelial cells by IF. The percentage of myoepithelial cells was decreased from 34% in the WT to 26% in the Wnt4 −/− epithelium ( Supplementary Fig. 1b). To gain more insights into the lineage deregulation, we went on to compare FACS-sorted GFP+ luminal and myoepithelial cells from conditionally Wnt4-deleted (MMTV::Cre + .Wnt4 fl/fl .mT/ mG) and control (MMTV::Cre + .Wnt4 wt/wt .mT/mG) epithelia by Affymetrix microarray analysis. The number of genes differentially expressed between the two genotypes was almost twice as high in the myoepithelial than in the luminal cell populations ( Supplementary Fig. 1c-e). Hence, despite a lineage defect, there are major gene expression changes in the myoepithelium. Gene set enrichment analysis (GSEA) of the differentially expressed genes revealed that signatures reflecting the activity of the canonical Wnt signaling target, Myc, and the expression of its target genes were decreased in the Wnt4 −/− myoepithelial but not luminal cells ( Supplementary Fig. 1f). Together these findings are consistent with the model that Wnt4 secreted by luminal cells activates canonical Wnt signaling in the myoepithelial cells 6 . Wnt4 was the most significantly down-modulated gene in the luminal compartment ( Supplementary Fig. 1d). While expression of Cytokeratin 5 (Krt5) a gene typically enriched in myoepithelial cells, was increased in the Wnt4 −/− luminal cells no cell typerelated gene signatures were identified ( Supplementary Fig. 1d). In the myoepithelial cell population, the secreted Wnt signaling inhibitor, Wif1, was the most significantly down-modulated gene suggesting the existence of a negative feedback loop in intraepithelial homeostasis ( Supplementary Fig. 1e). The stem and progenitor cell markers, Sox9 and Lgr5, were decreased ( Supplementary Fig. 1e). Adamts18 was also among the down modulated genes but failed to reach statistical significance ( Supplementary Fig. 1e). GSEA revealed furthermore a decreased stem cell signature and an increase in Tgf-β targets in the Wnt4 −/− myoepithelial cells ( Supplementary Fig. 1g). Reactome pathway analysis revealed a protein interactome centered around cell-cell junction and cell junction organization as well as cell-cell communication ( Supplementary Fig. 1h). Taken together, while the deletion of Wnt4 results in a stem cell defect with some consequent cell lineage defect, the gene is expressed in the luminal compartment and its deletion affects transcription mostly in the myoepithelial compartment where Adamts18 is expressed.
Mammary gland development in Adamts18 −/− mice. To assess the functional importance of Adamts18 in mammary gland development, we generated mice homozygous for an allele lacking exons 8 and 9, which encode the Zn-binding catalytic site 31 and analyzed their inguinal mammary glands at critical developmental stages by whole mount stereomicroscopy. In prepubertal, 14-dayold WT and Adamts18 −/− littermates, the ductal system was rudimentary and of similar size in both genotypes (Fig. 2a). Consistently, extent of fat pad filling (Fig. 2b) and the number of branching points were comparable in prepubertal, 14-day-old, WT and Adamts18 −/− littermates (Fig. 2c). In pubertal, 4-6-week-old, WT females, milk ducts grew by characteristic dichotomous branching, extended beyond the subiliac lymph node, and had enlarged tips, terminal end buds (TEBs) characteristic of this stage (Fig. 2d). In the Adamts18 −/− littermates, ducts barely reached the lymph node (Fig. 2d). The extent of fat pad filling was reduced by 50% (Fig. 2e), the number of branching points by 60% (Fig. 2f) and the number of TEBs by 40% compared to the WT counterparts (Fig. 2g). In adult, 14-week-old, females, the milk ducts reached the edges of fat pads in both genotypes. In WT females, ductal complexity was increased through side branching whereas ducts of the Adamts18 −/− littermates were simple (Fig. 2h) and the number of branching points was 58% of WT (Fig. 2i). Thus, Adamts18 is required for ductal development both during puberty and adulthood.
Histological examination of mammary glands from 6-week-old mice revealed structurally normal ducts with intact luminal and myoepithelial layers in both genotypes (Fig. 2j). To address whether the observed delay in ductal elongation was due to increased cell death and/or decreased cell proliferation, we stained sections from pubertal glands for cleaved-caspase 3 and phosphorylated histone H3 (pHH3). The proportion of cleaved caspase3+ cells did not differ significantly ( Fig. 2k) but the pHH3-index in Adamts18 −/− mammary epithelia was reduced to 64% of WT levels (Fig. 2l, m). Thus, the delayed ductal elongation is due to decreased cell proliferation.
Adamts18 function in the mammary epithelium. Adamts18 −/− pups show a transient growth delay 18 , which may indirectly affect mammary gland development. In addition, subfertility associated with abnormalities in the female reproductive tract, such as dorsoventral vagina or imperforate vagina of Adamts18 −/− females 18 precluded analysis of mammary gland development during pregnancy. To discern the epithelial-intrinsic role of Adamts18 in ductal growth at later developmental stages, we grafted mammary epithelium from WT.EGFP + and Adamts18 −/− .EGFP + females to contralateral inguinal glands of 3-week-old WT female mice surgically divested of their endogenous epithelium. To unequivocally distinguish the engrafted epithelium from host epithelium that could have been inadvertently left behind during surgery, the donor cells constitutively expressed an enhanced green fluorescent protein (EGFP) under control of a chicken β-actin promoter 32 . Six weeks after engraftment, outgrowths derived from WT donors filled the host fat pads whereas the contralateral Adamts18 −/− epithelia failed to do so (Fig. 3a) and the branching points were decreased by 33% (Fig. 3b). Twelve weeks after engraftment, both WT.EGFP + and Adamts18 −/− .EGFP + outgrowths filled the host fat pads but side branching was decreased in Adamts18 −/− .EGFP + epithelial grafts (Fig. 3c). Flow cytometry of dissociated glands showed a 30% reduction in EGFP + cells (Fig. 3d) consistent with decreased cell proliferation resulting in lower epithelial cell numbers and delayed branching. Thus, the mammary branching phenotype in Adamts18 −/− females is intrinsic to the mammary epithelium.
At 14.5 days of pregnancy, epithelia of both genotypes showed widespread alveoli both by fluorescence stereomicroscopy and histology ( Fig. 3e). At day 1 of lactation, alveoli were fully distended ( Fig. 3f) suggesting normal lactogenic function. However, at both time points, spaces between EGFP+ epithelial structures were larger in Adamts18 −/− .EGFP + grafts than in the WT counterparts consistent with reduced side branching at earlier stages ( Fig. 3e-g). In line with the morphologic analysis and the decreased number of MECs, transcript levels of lactogenic differentiation markers such as Lalba, Wap, and CsnA were lower in mutant glands compared to WT controls but failed to reach statistical significance when normalized to the epithelial marker Krt18 (Fig. 3h). Thus, while epithelial cell numbers are decreased in the absence of Adamts18, the protease is not required for cytodifferentiation.
Adamts18 expression has been reported in the stromal compartment and was confirmed by semi quantitative RT-PCR analysis of WT fat pads engrafted with Adamts18 −/− epithelium showing 25% of the Adamts18 transcript levels detected in WT recombinants ( Supplementary Fig. 2a). To determine the functional importance of this stromal expression, WT.EGFP +   Fig. 2d). Thus, stromal Adamts18 expression is not required for ductal branching.
As epithelial ER and PR signaling drive pubertal dichotomous branching and estrous cycle-induced side branching, respectively 29,33 , we asked whether receptor expression was affected by   Adamts18 inactivation. IHC of sections from contralateral glands engrafted with WT.EGFP + and Adamts18 −/− .EGFP + epithelia revealed comparable proportions of ER+ (Fig. 3i) and PR+ cells (Fig. 3j) indicating that Adamts18 is not required for ER or PR protein expression.
The role of Adamts18 in mammary epithelial self-renewal. Delayed pubertal ductal outgrowth and reduced side branching together with normal alveologenesis and cytodifferentiation were previously observed in Wnt4 −/− epithelia 26 and shown to result from a stem cell defect 6 . To test whether Adamts18 deletion also affects mammary stem cells (MaSCs), we analyzed cells from dissociated WT and Adamts18 −/− mammary glands by FACS 34 using CD24 and CD49f detection after depletion for lineage positive cells (Fig. 4a). The number of lineage-depleted cells obtained from mammary glands of 14-week-old females was one third less in Adamts18 −/− compared to WT (Fig. 4b). The percentage of both luminal (Lin − CD24 + CD49f − ) and myoepithelial (Lin − CD24 low CD49f low ) cells was not significantly altered in Adamts18 −/− glands but the stromal cell fraction (Lin − CD24 − CD49f − ) increased by 19% in the mutant glands (Fig. 4c). Mammary progenitors, which give rise to colonies and are called colony forming cells (CFCs) represented <1% of the lineage negative cells in both genotypes whereas the number of MaSCs (Lin − CD24 med CD49f high ) also defined as mammary repopulating units (MRUs) was decreased by 43% in Adamts18 −/− glands (Fig. 4a, d).
To functionally evaluate stem cell frequency in WT and Adamts18 −/− mammary epithelia, we injected serially diluted single cells from WT.EGFP + and Adamts18 −/− .EGFP + mammary glands to contralateral cleared fat pads of 3-week-old WT.EGFP − female mice. After 8 weeks, we determined frequency and extent of outgrowth by fluorescence stereomicroscopy combined with image analysis. The repopulating cell frequency 35 of Adamts18 −/− cells was 10% of the WT cells with 1/20,000 vs. 1/2000 (Fig. 4e).
The single cell-based in vivo reconstitution assay can be confounded by cell adhesion and/or cell migration defects as well as by increased susceptibility to apoptosis. All these factors impact on any cell's ability, whether stem cell or not, to establish itself after injection in the fat pad, which is, of course, a prerequisite for the generation of any progeny. A complex assay that overcomes these limitations is the serial transplantation of pieces of intact epithelium. WT epithelium that is serially grafted can fill cleared fat pads for up to 7 generations 36 . Indeed, WT. EGFP + epithelium filled host fat pads efficiently over 5 transplant generations, however, the reconstitution ability of the contralaterally grafted Adamts18 −/− .EGFP + epithelium decreased progressively to cease completely upon the 5th transplant (Fig. 4f, h,  i). Histological analysis of the 4th generation transplants by H&E revealed no obvious difference (Fig. 4g). Thus, Adamts18 is required for the regeneration capacity of the mammary epithelium, albeit to a lesser extent than PR and Wnt4, whose deletion blocks reconstitution at the 4th and 3rd generation, respectively, by the same assay 6 .
The basement membrane is part of the stem cell niche. To address the mechanisms by which Adamts18 affects stem cell activity, we searched for its binding partners. In light of the myoepithelial cell-specific expression of the protease, we chose the human breast epithelial cell line, MCF10A, which has myoepithelial/basal characteristics 37 , as a model. We ectopically expressed V5-tagged ADAMTS18 in these cells, immune precipitated it from the conditioned medium, and analyzed coimmunoprecipitated proteins by mass spectrometry. We discovered 238 proteins cumulatively in 3 independent experiments (Supplementary Data 1), of which 31 were identified in ≥2 experiments (Fig. 5a). Transforming Growth Factor Beta-Induced (TGFBI), a secreted molecule that contains RGD domains similar to fibronectin and laminin and inhibits cellular adhesion to the ECM, was among the 12 proteins identified in all 3 experiments 38 . Bioinformatic analysis with MetaCore showed that top enriched MetaCore processes related to ECM organization and hemidesmosome assembly (Fig. 5b, Supplementary Table 1). The top localizations of the putative ADAMTS18 interactors were ECM, laminin-5 complex, and BM (Fig. 5c, Supplementary  Table 2). Together, these findings support the hypothesis that Adamts18 function relates to the ECM and, more specifically, to the connection between epithelium and BM. This implies that the BM may be part of the stem cell niche.
To seek in vivo evidence for a role of the BM as part of the stem cell niche we turned to mice deficient for Col18a1 because this heparin-sulfate proteoglycan is specifically localized to BMs 39 . Whole mount stereo-microscopy and morphometric analysis showed that Col18a1 −/− females like their Adamts18 −/− counterparts had delayed ductal elongation and fewer TEBs compared to their WT littermates (Fig. 5d). Adamts18 and Col18a1 doubledeficient (DKO) mice showed a further decrease in TEB numbers, fat pad filling, and branching points at 6 weeks compared to single knockouts (Fig. 5e) indicating that Adamts18 and Col18a1 have additive roles in ductal elongation. To assess whether this genetic interaction affects stem cell function, we serially transplanted the DKO epithelium. While the contralateral WT epithelium reconstituted glands over 5 transplant cycles, the DKO epithelium failed to reconstitute by the 3rd generation ( Fig. 5f-h). Thus, Adamts18 and Col18a1 cooperate in mammary stem cell control, providing in vivo evidence for a role of the BM in stem cell function, likely as part of the stem cell niche.
Adamts18 modulates the ECM. To probe for structural alterations in the ECM related to Adamts18 deletion, we used picrosirius red to stain Adamts18 −/− and WT pubertal mammary glands. Fibrillar collagen was increased around the ducts and TEBs in Adamts18 −/− relative to WT (Fig. 6a). Immunoblotting of protein lysates from pubertal WT and Adamts18 −/− glands and quantification showed that levels of the important BM components, laminin and collagen IV increased 1.7-and 3.9-fold, respectively, in Adamts18 −/− glands (Fig. 6b, c). Levels of the major fibrillar collagen, collagen I, were increased 6.2-fold (Fig. 6b, c). Assembly of nascent collagen I, laminin and collagen IV matrices rely on initial assembly of fibrils composed of the primordial ECM glycoprotein fibronectin, the first ECM protein to be expressed during tissue development and wound healing 40,41 . Fibronectin levels were 3.2fold higher in the mutants than in WT (Fig. 6b, c). IF showed increased staining intensity for all these proteins around ducts and TEBs in Adamts18 −/− relative to WT pubertal glands (Fig. 6d). The staining was restricted to the BM for laminin and collagen IV but extended to the interstitial ECM for collagen I and fibronectin. Thus, in the absence of Adamts18, major ECM/BM components accumulate in the pubertal mammary gland in line with an important role for Adamts18 in ECM/BM remodeling.
Interestingly, analysis of mammary glands from 14-week-old WT and Adamts18 −/− littermates showed that protein levels of laminin, collagens I and IV as well as fibronectin did not differ significantly between the two genotypes (Fig. 6e, f). This shows that Adamts18 is critical for ECM/BM modulation during pubertal ductal elongation and suggest that this specific developmental window determines mammary stem cell function.
Adamts18 cleaves fibronectin. In contrast with the increased fibronectin protein levels, its mRNA levels were unaltered in the pubertal Adamts18 −/− mammary glands (Fig. 6g) suggesting that the observed increased staining could result from translational or posttranslational changes attributable to lack of Adamts18. As fibronectin is the prime component of nascent ECM fibers and a substrate of the Adamts18 homolog Adamts16, we tested whether it is equally an Adamts18 substrate. We purified the secreted active form of ADAMTS18 from HEK-293T cells and incubated it with N-terminal 70 kDa fibronectin. The exogenous fibronectin fragment migrated slightly faster when co-incubated with EDTA and was undetectable in the presence of ADAMTS18 after 24 h. When the digest was supplemented with EDTA, which chelates the bivalent metal ions required for ADAMTS activity, no change in fibronectin abundance was seen (Fig. 6h). Additionally, HEK-293T cells expressing ADAMTS18 or a control vector were incubated without or with the 70 kDa recombinant fibronectin. By western blot, the medium of cells expressing ADAMTS18, but not the control vector, showed a readily detectable 30 kDa fibronectin fragment (Fig. 6i) similar to that detected after cleavage by Adamts16 21 . The amount of cleaved fibronectin increased 16-fold in the presence of ADAMTS18 (Fig. 6j). Thus, the presence of ADAMTS18 leads to fibronectin proteolysis, which may influence abundance of other ECM proteins and indirectly regulate growth factor availability and signaling.
Stem cell signaling in Adamts18 −/− glands. Our findings pointed to the observed stem cell defect being secondary to changes in the ECM/BM. To elucidate the mechanisms by which altered ECM affected stem cell signaling, we transcriptionally profiled 3 pairs of contralateral glands engrafted with either WT. EGFP + or Adamts18 −/− .EGFP + epithelia using RNA-seq. PCA analysis was used to identify and visualize possible batch effects due to sources of variation in the mice used ( Supplementary  Fig. 3a). After removing these effects by applying 2-way ANOVA correction, samples clustered by biological subgroups (Supplementary Fig. 3b). Expression of Adamts18, Fgfr2, and Ctgf was tested and found reduced in all 3 Adamts18 −/− samples after read   Rel. TEB numbers * * **   count normalization (Supplementary Fig. 3c). Overall, in Adamts18 −/− transplanted glands, expression of 313 genes decreased (FC < 0.8, p < 0.05) and that of 273 genes increased (FC > 1.25, p < 0.05) (Fig. 7a). Analysis of the differentially expressed genes by pathway enrichment analysis using both ReactomePA 42 and ClusterProfiler 43 showed that cell junctions and ECM were affected, in particular various collagens and laminins (Supplementary Fig. 3d-g). More specifically, out of 40 significant GO terms, 11 were related to the ECM and 10 to Fgfr signaling, a pathway critical for stem cells 44,45 (Supplementary Table 3). Two of the 40 terms related to Hippo-Yap/Taz signaling another pathway critical for stem cell differentiation, which is upstream of Fgfr2 46 . When we specifically interrogated the genes whose expression decreased, Reactome pathway analysis revealed Yap/ Taz-mediated gene expression (Fig. 7b) and a protein interactome centered around cell-cell communication and cell-cell junctions as well as ECM, laminin and collagen complexes and assembly (Fig. 7c) that partly overlap with the Wnt4 specific interactome (Supplementary Fig. 1h). In light of the increased ECM deposition, the differential expression of various ECM-related genes as well as the involvement of the Yap/Taz signaling pathway, we evaluated integrin expression in the Adamts18 −/− glands. We extracted 27 Integrin genes, α and β Integrin subunits, from the RNAseq analysis and generated a heatmap (Supplementary Fig. 3h). No integrin-related gene was significantly altered by adjusted p-value, but Itga3, Itgb4, and Itgb7 were significantly altered by p-value. Analysis of their expression levels by qRT-PCR at puberty in mammary glands from WT and Adamts18 −/− mice showed Itga3 and Itgb4, two integrins previously implicated in mammary stem cell function 47,48 and part of laminin 5 receptors, to be significantly down modulated in the mutants (Fig. 7d).
Together these findings suggest that Adamts18 is required for activation of the Hippo pathway, which in turn induces Fgfr2 expression, activation of which is critical for stem cell function. Consistent with this scenario, the 3 Hippo target genes, Ctgf, Fgfr2, and Gata3 46,49 were reduced to 73%, 68% or 78% of WT levels, respectively, in additional transplants in the absence of Adamts18 (Fig. 7e). Double-IF for Yap and the myoepithelial marker α-smooth muscle actin (Sma) showed expected nuclear localization of Yap in WT myoepithelial cells (Fig. 7f) 50 . In the contralateral Adamts18 −/− .EGFP + epithelia the signal intensity of Yap was decreased in myoepithelial cells (Fig. 7f). Quantitative image analysis revealed that the mean nuclear intensity of the Yap staining in the mutant epithelium was 58% of the contralateral WT.EGFP + transplanted glands (Fig. 7g).
To further support our claim that BM modulation by Adamts18 involves the Yap/Taz signaling pathway, we assessed the expression levels of downstream targets, Cited-1, Ctgf, Fgfr2, Gata3 in pubertal WT, Col18a1 −/− , Adamts18 −/− , and DKO mice. Additionally, we assessed the expression levels of Itga3 and Itgb4 altered in Adamts18 −/− mice. In line with our previous findings (Fig. 7h), we found the Yap/Taz targets to be significantly down modulated in pubertal Adamts18 −/− and the DKO. Col18a1 −/− glands displayed downmodulation in Adamts18, Cited-1, and Ctgf. This suggests that modulation of the BM composition by Adamts18 leads to activation of Yap/Taz signaling with increased Fgfr2 expression and signaling which results in stem cell activation.
ADAMTS18 in the human breast. Our data indicate that Adamts18 translates the hormonal stimuli received by luminal cells into activation of stem cells via changes to the BM in the mouse mammary gland. To assess whether this signaling axis may also operate in the human breast, we generated a polyclonal antibody to ADAMTS18 and validated it on MCF10A overexpressing V5 tagged human ADAMTS18 with or without a short hairpin RNA (shRNA) to knock down overexpressed ADAMTS18 (Supplementary Fig. 4). IHC of reduction mammoplasty sections showed ADAMTS18 expression was not detected in the CK7+ luminal compartment, but in myoepithelial cells identified by p63 immunostaining (Fig. 8a) as observed for the transcripts in the mouse.
To test whether expression of ADAMTS18 transcripts in human breast epithelial cells is similarly controlled by PR signaling, we humanized mouse mammary glands 51 . Human breast epithelial cells isolated from 4 different reduction mammoplasty specimens were infected with lentiviruses expressing luciferase-GFP and injected into the milk ducts of immune-compromised NOD scid gamma females 51 (Fig. 8b). Once photon flux reached 10 7 per gland, the mice received subcutaneous pellets containing either vehicle, 20, or 50 mg progesterone (Fig. 8c, d). The hormonecontaining pellets resulted in 7.2-and 19.7-fold increased plasma progesterone levels, respectively (Fig. 8c); Adamts18 transcript levels were 1.8-and 2.3-fold higher than in noninjected mammary glands from the progesterone-treated mice, respectively, indicating that prolonged progesterone exposure results in increased Adamts18 transcript levels in the mouse mammary glands (Fig. 8d). Next, we dissociated the xenografted glands to single cells and enriched for the human cells by depleting mouse cells with immunomagnetic beads. The xenografted cells from 4 different patients exposed to progesterone showed increased expression of ADAMTS18 compared to control cells with an average 3-fold increase (Fig. 8e). Thus, the progesterone/ ADAMTS18 axis is conserved between mice and humans.

Discussion
Here, we have addressed the longstanding puzzle of how epithelial ER and PR signaling connect to ECM changes that accompany both normal breast development and breast carcinogenesis. We show that the gene encoding Adamts18 is expressed in the myoepithelium downstream of Wnt4 secretion induced by ER/PR signaling luminal sensor cells (Fig. 9). The myoepithelial cells respond by canonical Wnt signaling activation and link luminal hormone receptor signaling to stromal changes with functional consequences. Our finding that altered BM composition affects MaSCs shows that the BM is a central part of the stem cell niche and a critical determinant of stem cell function.
The precise nature of the BM and interstitial ECM changes that alter signaling remain to be determined. Numerous factors, such as tissue stiffness and growth factor availability, directly or indirectly controlled by Adamts18 may be critical. The observed changes in the abundance of collagen I, collagen IV, laminin, fibronectin, and glycoproteins, like collagen XVIII, may be secondary to the reduced fibronectin clearance but Adamts18 may also be directly involved in their processing; other family members have glycoprotein substrates 14 .
Increased laminin expression was also observed in Adamts18 −/− adipose tissue 52 and embryonic brains 53 with effects on early adipocyte differentiation and spine and synapse formation. A detailed analysis of kidney and lung development in WT and Adamts18 −/− mice revealed that expression of the enzyme by branching tips is important for branching and organ size 18 .
We identified enhanced Yap/Taz nuclear localization and increased Fgfr2 signaling as potential mechanisms underlying Fig. 6 Biochemical changes and Fibronectin cleavage elicited by Adamts18. a Representative picrosirius red staining for fibrillar collagen (red) on 4th mammary gland sections from 5-week-old, pubertal WT and Adamts18 −/− littermates; n = 5. Scale bar, 100 μm. b Representative western blot analysis on 3rd mammary glands of 5-week-old, pubertal WT and Adamts18 −/− littermates; n = 4. β-actin loading control, MW marker in red. c Dot plots showing relative protein levels of laminin, collagen I, collagen IV, and fibronectin normalized to actin in 4 pubertal WT and Adamts18 −/− littermates. Paired Student t-test, two-tailed; **p < 0.01. d, Fluorescent micrographs showing IF on 4th mammary gland sections from 5-week-old, pubertal WT and Adamts18 −/− littermates for laminin, collagens I and IV as well as fibronectin (green) and DAPI nuclear stain (blue), n = 3. Arrows point to ECM density around TEBs or ducts; scale bar, 100 μm. e Representative western blot analysis on 3rd mammary glands of 14-week-old WT and Adamts18 −/− littermates; n = 3. β-actin loading control, MW marker in red. f Dot plots showing relative protein levels of laminin, collagen I, collagen IV, and fibronectin normalized to actin in 3 adult WT and Adamts18 −/− littermates. Paired Student t-test, two-tailed; n.s. not significant. g Dot plot showing relative transcript levels of Fn1 normalized to Hprt in 3rd mammary glands from 6 pairs of 5-week-old WT and Adamts18 −/− littermates. Paired Student t-test, two-tailed, n.s. not significant. h Representative Western blot analysis of 3 independent experiments in which fibronectin (FN)−70K was incubated with purified active Adamts18 in the presence or absence of EDTA and/or protease inhibitor (PI). Anti-FN antibody specific to the N-terminal heparin-binding domain. i Western blot analysis of FN1-70K incubated with ADAMTS18 overexpressing HEK-293T cells in the presence or absence of EDTA. j Bar graph showing levels of cleaved FN in supernatants from control transfected and Adamts18 overexpressing HEK-293T cells in 2 independent experiments. stem cell activation downstream of Adamts18 activity (Fig. 9). Whether Yap/Taz activation is central to increased Fgfr2 signaling and/or whether biochemical changes in the BM result in increased ligand availability was not addressed in our study. Yap/Taz signaling is typically activated by extracellular cues such as increased stiffness. Our gene expression analysis did not provide direct indications for this; whether the increased expression of musclerelated genes may also impinge on Yap/Taz or whether another stiffness independent mechanism 54 is important, remains to be explored. We speculate that Adamts18-induced modifications of the ECM affect integrin-mediated, F-actin dependent cell-ECM adhesion and contraction, which promote cellular mechanical tension and Yap/Taz activation 55 . As such, the progesterone/ Wnt4/Adamts18 axis provides an entry point for further studies of epithelial-BM interactions.
The regulatory axis we identified genetically in the mouse mammary gland likely operates in the human breast with implication for breast cancer prevention and treatment. Exposure to progesterone as it occurs recurrently during menstrual cycles has been shown to induce WNT4 expression 56 ADAMTS18 expression, as we show here. The resulting BM/ECM remodeling may contribute to the increased breast cancer risk associated with recurrent menstrual cycles. Furthermore, the increased risk of postmenopausal women exposed to combined hormone replacement therapy with ethinyl estradiol and progestins may, at least in part, be attributable to increased stem cell divisions and stromal alterations 58,59 elicited by ADAMTS18.
Premenopausal patients with in situ carcinoma or early stage invasive disease, as well as women with high risk for breast cancer, may benefit from a preventive treatment that interferes with PR signaling or its downstream effectors. Blocking progesterone action, while possibly protective for the breast, will have many side effects as its actions are complex and affect many organs. Similarly, targeting downstream Wnt signaling has potential side effects because this signaling pathway is physiologically important for stem cells in many tissues. Based on the mouse model, ADAMTS18 is important for development of specific organs but it does not appear to have an essential function in adult mice 31 . Furthermore, in its extracellular location ADAMTS18 makes it an excellent target for antibody-mediated therapy. As such, targeting ADAMTS18 appears as a feasible strategy for primary and secondary prevention unlikely to elicit major side effects.
Patient sample processing. The cantonal ethics committee approved the study (183/10). Breast tissue was obtained from women undergoing reduction mammoplasties with no previous history of breast cancer. All human subjects provided Fig. 7 Adamts18 impinges on transcription and regulates cell signaling. a Volcano plot showing genes, which are differentially expressed between contralateral glands transplanted with Adamts18 −/− and WT epithelia; n = 3, Kolmogorov-Smirnov test, all highlighted genes have p-values < 0.05. Genes with log2(FC) >0.5 in red and log2FC <0.5 in blue. Names of selected genes are indicated. b Enrichment map plot of Reactome pathway analysis (ReactomePA) on genes downregulated in 3 pairs of contralateral glands engrafted with WT and Adamts18 −/− epithelia in 3 independent experiments with 3 different donors. c CNE plot of ReactomePA of genes down regulated in contralateral glands transplanted with WT and Adamts18 −/− epithelia. d Bar graphs showing relative transcript levels of Adamts18, Itga3, Itgb4, and Itgbt, normalized to Hprt in 5 pubertal host mice bearing contralateral transplants of WT and Adamts18 −/− epithelia. Data represent mean ± SD. Unpaired Student t-test, two-tailed. e Bar graphs showing relative transcript levels of Fgfr2, Ctgf, and Gata3 normalized to Hprt in contralateral glands transplanted with WT and Adamts18 −/− epithelia, n = 6. f Representative IF for Sma (green) and YAP (red) counterstained with DAPI (blue) of 4th mammary gland sections from 5-week-old WT and Adamts18 −/− littermates; n = 3. Arrows indicate YAP positive nuclei of myoepithelial cells. g Dot plot showing quantification of relative mean intensity of nuclear YAP detected in myoepithelial cells of 5-weekold WT and Adamts18 −/− littermates; n = 3. Each point represents an individual TEB. h Bar graphs showing relative transcript levels of Adamts18, Col18a1, Cited-1, Ctgf, Fgfr2, Gata-3, Itga3, and Itgb4, normalized to Hprt in pubertal WT, Col18a1 −/− , Adamts18 −/− , and DKO; n = 9, 8, 4, and 4, respectively. Data represent mean ± SD, one-way ANOVA. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001, n.s. not significant.  Experimental scheme: dissociated human breast epithelial cells from reduction mammoplasties were injected via the teat into the milk duct system of NSG female mice and establish themselves there. c LC/MS measured serum progesterone levels in mice 60 days after implantation with pellets containing vehicle, 20 or 50 mg progesterone. Data represent mean ± SD from n = 10 (vehicle), n = 7 (20 mg), and n = 4 (50 mg); one-way ANOVA. d Dot plot showing Adamts18 transcript levels as measured by semi qRT-PCR normalized to the geometric mean of Hprt and Gapdh in mammary glands from mice that were subcutaneously engrafted with pellets containing either vehicle (0) or 20 or 50 mg progesterone for 60 days. Data represent mean ± SD from n = 10 (vehicle), n = 7 (20 mg), and n = 4 (50 mg); one-way ANOVA. e Dot plot showing relative ADAMTS18 transcript levels normalized to GAPDH in glands xenografted with human breast epithelial cells from 4 mammoplasty specimens. Recipient mice were either implanted with vehicle-or 20 mg progesterone-containing pellets. Paired t-test, two-tailed. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.
informed consent for use of tissue samples in research. Samples were examined by the pathologist to be free of malignancy.
Histology. Inguinal mammary glands were fixed in 4% PFA in phosphate-buffered saline (PBS, pH 7.2) overnight at 4°C, embedded in paraffin and cut into 4 μm sections. Hematoxylin and eosin or sirius red staining were performed according to standard protocols. For immunostaining, sections were de-waxed, rehydrated and subjected to antigen retrieval with 10 mM citrate buffer, pH 6.0 for 20 min at 95°C. Sections were counterstained with Mayer's hematoxylin. For fluorescence microscopy, nuclei were counterstained with DAPI (Sigma). IF images were acquired on Leica DM 4000 B LED with Leica DFC 7000T camera and on Zeiss LSM700 confocal microscope for colocalizations. Primary antibodies A rabbit anti-ADAMTS18 antibody was raised against the peptide GQYKYPDKLPGQIYDA corresponding to ADAMTS18 sequence 502-516 aa (Eurogentec) an epitope that is conserved between human and mouse proteins, absent from other proteins and selected for high antigenic potential. The percentage of ER+ and PR+ cells were quantified using ImageJ, the percentage of SMA+ cells with QuPath software. Antibody list can be found in Supplementary Table 4.
RNA in situ Hybridization. Adamts18 ISH was performed using RNAScope (Advanced Cell Diagnostics, Newark, CA) following the manufacturer's protocol. Briefly, 4 µm sections were deparaffinized and hybridized to a mouse Adamts18 probe set (452251; Advanced Cell Diagnostics) using a HybEZ oven (Advanced Cell Diagnostics) and the RNAScope 2.5 HD Detection Reagent Kit (322360; Advanced Cell Diagnostics) and stained with anti-SMA after the RNAScope procedure.
Transplantation. Fat pads were transplanted onto the abdominal muscle wall of adult WT females 29 . Single cell suspensions of mammary epithelial cells in 20% matrigel were injected and 1 mm 3 of epithelial fragments were transplanted to cleared fat pads. Intraductal injection of human breast epithelial cells was performed via cleaved teat.
Mammary gland wholemounts. Mammary gland whole-mounts were performed as described 65 , and stereomicrographs were acquired with a LEICA MZ FLIII stereomicroscope and Leica MC170 HD. Fluorescence stereomicrographs were acquired on a LEICA M205FA with a Leica DFC 340FX camera. Fat pad filling and branching points were determined using ImageJ software.
Hormone measurements. Progesterone hormone levels in the plasma were measured using LC-MS (Q-Exactive, ThermoFisher Scientific) 66 .
Fluorescence activated cell sorting. Single cell suspensions of mammary glands from 15-to 25-week-old virgin females were processed as described 34 and sorted on a FACSAria (Becton Dickinson).
Hormone treatments. Low consistency silicon elastomer (MED-4011) two parts (part A, MP3745/E81949 and part B, MP3744/E81950) were mixed with hormone powder, incubated at 37°C overnight as described 67 , and implanted subcutaneously. Three-week-old mice were ovariectomized and injected subcutaneously 10 days later with 17-β-estradiol 5 ng/g of body weight (Sigma-Aldrich, St. Louis, MO) using 5 mg/ml in 100% ethanol stock or vehicle. Mammary glands were harvested 18 h after injection.  Fig. 9 Working model of Adamts18 as a modulator of mammary gland development. A schematic representation of the mammary acinar wall shows the spatial relationship between luminal cells, myoepithelial cells, BM and the surrounding interstitial ECM. Estrogen and progesterone induce Adamts18 production in myoepithelial cells via Wnt4-stimulated canonical Wnt signaling. Adamts18 remodels the BM and/or interstitial ECM, as part of the stem cell niche to ensure optimal stem cell regenerative capacity. Loss of Adamts18 alters the stem cell niche and decreases mammary epithelial regenerative potential as its essential ECM modulatory function is abrogated.
RT-PCR. Mammary glands were homogenized with TRIzol reagent (Invitrogen), total RNA was isolated with miRNeasy Mini Kit (Qiagen), cDNA was synthesized with random p(dN) 6 primers (Roche) and MMLV reverse transcriptase (Invitrogen). Real-time PCR analysis in triplicates was performed with SYBR Green Fas-tMix (Quanta) reaction mix. Primers used for RT-PCR, see Supplementary Table 5.
AP-MS analysis for ADAMTS18 binding proteins. MCF-10A cells were spininfected with an ADAMTS18 lentivirus containing a V5 tag or LacZ control virus. Cells were cultured to confluence in 10 cm dishes. Proteins were extracted with RIPA lysis buffer supplemented with protease inhibitors and protein concentration measured with a BCA kit (Pierce). ADAMTS18 was immunoprecipitated from 1 mg of protein using anti-V5 antibody conjugated agarose beads (Sigma A7345). The immune precipitates were subjected to SDS-PAGE, the gel was stained with colloidal Coomasie blue (Biorad), bands were excised and subjected to reduction/ alkylation followed by tryptic digestion and LC-MS/MS proteomic analysis. Detected peptides were mapped against the human protein database, label-free protein quantification was performed and affinity lists were constructed in Scaffold 4 Proteomics Software using a minimum of 2 peptides to identify the proteins with a peptide false discovery rate (FDR) of 0.1% and protein FDR of 0.3%.
Cloning. ΔCT-Adamts18-867aa cDNAs were amplified from cDNA library prepared from eyes and fused to FLAG-tag and His 6 -tag at N-terminus by PCR and cloned into NheI and HindIII restriction sites of pcDNA3.1/Hygromycin expression vector (Invitrogen). Plasmids were purified with HighPure midiprep kit (Invitrogen).