An essential role of CBL and CBL-B ubiquitin ligases in mammary stem cell maintenance

Postnatal organ development and maintenance require organ-resident stem cells. While mechanisms of hematopoietic stem cell (HSC) maintenance are well understood, those that maintain epithelial stem cells have only recently become a focus of investigation. This is in part due to the rapid turnover of many epithelia and an emerging interest in the stem cell-like hierarchy of cancers, the clear majority of which originate in epithelial tissues. The mammary gland (MG) provides an excellent model to examine the mechanisms of adult epithelial stem cell maintenance, since a large part of ductal growth and branching takes place during pubertal development (Robinson et al., 2001; Gjorevski and Nelson, 2011).

Recently developed approaches to isolate and propagate mammary stem cells (MaSCs) in vitro, and to examine their self-renewal, differentiation and organ-regenerating abilities, make them excellent models to investigate mechanisms of epithelial stem cell homeostasis (Stingl et al., 2006a,b; Visvader, 2009). A better understanding of these mechanisms is important given the similarities of normal and cancer stem cell hierarchies and gene expression (Beck and Blanpain, 2013; Nguyen et al., 2012), as well as the prognostic significance of MaSC transcriptional signatures in breast cancer patients (Pece et al., 2010; Siegel and Muller, 2010; Soady et al., 2015).

Signaling pathways that regulate MG development also regulate MaSCs. For example, the Wnt signaling pathway regulates mammary development as well as MaSC self-renewal (van Amerongen et al., 2012). Wnt ligands help propagate MaSCs in vitro (Kessenbrock et al., 2013; Wang et al., 2013; Zeng and Nusse, 2010) and partly mediate the hormonal regulation of MaSCs (Cai et al., 2014). The Wnt pathway target gene Lgr5, which marks epithelial stem cells in the intestine, skin and other tissues (Barker et al., 2007, 2013), was recently shown to mark MaSCs (Plaks et al., 2013; Rios et al., 2014). Lgr5⁺ MaSCs can regenerate a MG upon transplantation (Plaks et al., 2013). Recently, Lgr5 was reported to be among the top 5% of 329 genes that are highly expressed in the MaSCs present among basal cells, as compared with other mammary epithelial cell (MEC) populations (Soady et al., 2015).

Receptor tyrosine kinases (RTKs), including members of the ErbB family, c-MET, RON, IGFR1/2, ephrin A2, among others, regulate MG development and MaSC maintenance (Gjorevski and Nelson, 2011; Hynes and Watson, 2010). FGFR2, for example, is crucial for MaSC maintenance (Pond et al., 2013) and RTK ligands are required for in vitro culture of MaSCs (Dontu et al., 2003; Guo et al., 2012). Dysregulation of precise signaling from RTKs and other receptors often leads to oncogenesis (Hynes and Watson, 2010; Korkaya et al., 2008).

Members of the CBL family (CBL, CBL-B and CBL-C in mammals) of ubiquitin ligases serve as negative regulators of protein tyrosine kinases (PTKs), including RTKs and non-receptor PTKs (Mohapatra et al., 2013). In contrast to substantial evidence supporting key physiological roles of CBL proteins (CBL/CBL-B) in hematopoietic and immune systems (An et al., 2015; Duan et al., 2004; Naramura et al., 2010; Thien and Langdon, 2005), their roles in epithelial tissues are essentially unknown. Cbl-c (also known as Cblc) mRNA is primarily expressed in epithelia, but a protein product has not been clearly demonstrated in such epithelia, and germline Cbl-c deletion is without an overt phenotype (Griffiths et al., 2003). A mammary epithelium-intrinsic role of CBL family proteins remains unknown. Transcriptome data show that CBL and CBL-B are expressed in the mammary epithelium, with CBL-B expression enriched in MaSCs (Lim et al., 2010). The embryonic lethality of germline Cbl and Cbl-b (also known as Cblb) double knockout (Cbl/Cbl-b DKO) in mice (Naramura et al., 2002), the exaggeration of immune phenotypes of Cbl-b deficiency by conditional Cbl deletion in immune cells (Kitaura et al., 2007; Naramura et al., 2002), a myeloproliferative disorder (MPD) upon Cbl/Cbl-b DKO in HSCs (An et al., 2015; Naramura et al., 2010), and the apparent lack of mammary epithelial-intrinsic and other epithelial phenotypes in Cbl^−/− or Cbl-b^−/− mice strongly suggest redundant functions of CBL and CBL-B in epithelia.

To investigate the epithelial cell-intrinsic roles of CBL and CBL-B, we used a conditional DKO model in which floxed Cbl was selectively deleted in the mammary epithelium on a germline Cbl-b^−/− background using MMTV-Cre (Wagner et al., 1997). Since concomitant DKO in a small fraction of HSCs in this model leads to a MPD (An et al., 2015; Naramura et al., 2010), we characterized the MG development prior to significant MPD and by using a transplant approach. These analyses revealed a redundant but essential epithelium-intrinsic requirement for CBL and CBL-B in pubertal MG development. Cbl/Cbl-b DKO mammary epithelium exhibited shrinkage of the MaSC-containing basal compartment, which led us to develop a novel MaSC-specific Cbl/Cbl-b DKO model in which floxed Cbl is inducibly deleted only in Lgr5⁺ MaSCs. We also generated a novel mouse model in which floxed Cbl and Cbl-b can be inducibly deleted in isolated basal MECs upon tamoxifen treatment (Goetz et al., 2016). Complementary evidence from these genetic models establishes that CBL and CBL-B are redundantly required to maintain MaSCs, apparently by controlling the level of AKT-mTOR signaling.

Real-time qPCR analyses of FACS-purified luminal and basal cell fractions of the mouse MG confirmed that all three CBL family genes are expressed in epithelial compartments (Fig. S1A). Since an endogenous CBL-C protein remains to be demonstrated (Mohapatra et al., 2013), while strong evidence supports redundant but crucial roles of CBL and CBL-B (Mohapatra et al., 2013; Naramura et al., 2002), we investigated the impact of mammary epithelial-intrinsic Cbl and Cbl-b DKO using Cbl-b null mice with MMTV-Cre-induced mammary epithelial deletion of floxed Cbl and expression of lacZ reporter (Naramura et al., 2010). The Cre⁺ Cbl^f/+; Cbl-b^+/− littermates served as Cre controls. X-gal staining of MG whole-mounts at 5-6 weeks of age indicated efficient Cre-mediated recombination in both control and DKO mice (Fig. S1B). Concurrent nuclear Fast Red and X-gal staining confirmed recombination in both luminal and basal compartments (Fig. S1C). Separately, the expression of a GFP reporter confirmed the MMTV-Cre-mediated gene deletion in the DKO and Cre control mice (Fig. S1D).

Since MMTV-Cre-induced DKO leads to MPD by 10 weeks of age, we analyzed the postnatal MG growth in 5-, 7- and 9-week-old virgin females. Compared with littermate controls, the DKO mice exhibited significantly retarded mammary ductal outgrowths (Fig. 1A), with significant reduction in the number of branch points, ductal length and fat pad filling (Fig. 1B-D). Western blotting (WB) confirmed the CBL deletion and the lack of CBL-B expression in DKO MGs (Fig. 1E), and immunohistochemical (IHC) staining revealed this to be in the mammary epithelium (Fig. 1F). Cbl-b KO mice show a compensatory increase in CBL expression in several tissues, including MG (Fig. 1E).

Hematoxylin and Eosin (H&E) staining of sections revealed sparser ducts in the MGs of 6-week-old DKO mice, but the overall basal and luminal layers were intact (Fig. S2A), a conclusion confirmed by immunofluorescence staining for basal [keratin 5 (K5)] and luminal [keratin 8 (K8)] cytokeratins (Fig. S2B).

The smaller size of DKO MGs could reflect reduced proliferation, increased apoptosis, or both, as their balance is important in mammary morphogenesis (Humphreys et al., 1996; McCaffrey and Macara, 2009). Ki67 staining showed reduced proliferation in DKO mammary ducts but not in terminal end buds (TEBs), compared with controls (Fig. S2C-F). Cleaved caspase-3 staining revealed a comparable level of apoptosis in the TEBs of control and DKO mice (Fig. S2G).

Although the analyses above were performed prior to overt MPD development, we used a transplantation approach (Deome et al., 1959; Stingl et al., 2006a,b) to rule out the possibility that subclinical MPD might have impaired MG development. Contralateral transplantation of control and DKO mammary tissue was performed in NSG hosts. Carmine alum and X-gal staining of MG whole-mounts 10 weeks post-transplant revealed significantly less fat pad filling and reduced branching in DKO transplants (Fig. S3A,B). Compared with the extensive growth and alveolar differentiation of control transplants during pregnancy, DKO transplants were dramatically smaller with less fat pad filling and reduced ductal elongation, although the alveolar differentiation appeared intact (Fig. S3B).

Next, to address whether the DKO phenotype was MEC-autonomous, FACS-sorted lineage-negative (Lin⁻; epithelial and stromal cells) or Lin⁻/EpCAM⁺ (epithelial cells only) MECs from 6-week-old Cbl/Cbl-b DKO versus control donors were transplanted in NSG mice. Whereas all the control Lin⁻ or Lin⁻/EpCAM⁺ MEC transplants led to mammary outgrowths, none of the DKO transplants showed any outgrowths (Fig. S3C,D). Together, our results conclusively demonstrate that deletion of Cbl and Cbl-b in MECs leads to a cell-autonomous impairment of MG growth and branching.

Impaired MG development in Cbl/Cbl-b DKO mice and their inability to regenerate the mammary epithelium upon transplant suggested potential impairment of DKO MaSCs. We therefore analyzed dissociated MECs from 6- or 8-week-old control or DKO mice by FACS, with staining for CD29 (ITGB1), CD24 and EpCAM, which separate MECs into basal (CD29^hi CD24⁺ or EpCAM^low CD29^hi) and luminal (CD29^low CD24⁺ or EpCAM^hi CD29^low) populations (Shehata et al., 2012; Smalley et al., 2012). The former includes MaSCs and mature myoepithelial cells, whereas the latter includes luminal progenitors and mature luminal cells (Shackleton et al., 2006; Stingl et al., 2006a,b). Although we used both CD24/CD29 and EpCAM/CD29 combinations, better separation of subsets was seen in FACS plots with EpCAM/CD29 combination (Fig. S4A,B). This analysis revealed that Cbl/Cbl-b DKO MGs yielded significantly fewer Lin⁻/EpCAM⁺ MECs in total, consistent with reduced MG size in these mice (Fig. S5A,B). Notably, the basal cell percentage was significantly lower, both at 6 and 8 weeks, with a corresponding increase in the proportion of luminal cells (Fig. 2A-D). However, the absolute number of luminal cells was reduced in DKO MGs (Fig. S5B,C). Given the overall reduction in the number of MECs in DKO mice (Fig. S5B,D), the smaller size of their basal MEC compartment suggested a particularly severe reduction in all or some of the cell types that comprise this compartment (Yamaji et al., 2009).

The purity of FACS-sorted basal or luminal cells was verified by K5 or K8 expression, respectively (Lim et al., 2010) (Fig. S6A,B). qPCR analysis of purified cell populations showed that Cbl and Cbl-b were expressed in control MECs but undetectable (Cbl signal <0.02% and Cbl-b signal <0.003% of wild type) in DKO MECs (Fig. S6C). Cbl-b expression was enriched in the basal compartment of control mice, consistent with previous microarray data (Lim et al., 2010). Recent studies have shown that the basal K5⁺ postnatal mouse MEC population contains bi-potent MaSCs, whereas the luminal K8⁺ population includes the luminal lineage-restricted progenitors (Rios et al., 2014). Quantification of mRNA levels in FACS-sorted populations by qPCR revealed reduced K5 levels in the basal compartment and increased K8 levels in the luminal compartment of DKO versus control mice (Fig. S6D,E).

We confirmed these findings using 6-week-old GFP reporter-bearing control and DKO mice, in which the GFP⁺ MECs represent the progeny of MMTV-Cre-expressing cells. These analyses also demonstrated a reduction in the basal cell population and a relative increase in the luminal fraction in DKO mice (Fig. 2E,F).

CD61 (ITGB3) and CD29 co-staining separates virgin MG MECs into populations that harbor MaSCs (CD29^hi CD61⁻), luminal progenitors (CD29^low CD61⁺), mature luminal cells (CD29^low CD61⁻) and mature myoepithelial cells (CD29^hi CD61⁺) (Carr et al., 2012; Shackleton et al., 2006; Stingl et al., 2006a,b). CD29 and CD61 co-staining and FACS analysis of MECs from 6-week-old control and DKO mice revealed that DKO MECs exhibited a significant increase in the proportion of luminal progenitors and a marked reduction of MaSCs, suggesting that Cbl/Cbl-b DKO promotes a shift of MaSCs towards luminal progenitors (Fig. 2G,H). However, we did not observe a proportionate increase in the mature luminal cell pool, suggesting that the DKO luminal progenitors are defective in their capacity to generate mature cells. The proportion of mature myoepithelial cells was comparable between control and DKO MECs.

Organoid-forming ability in Matrigel reflects the stemness of MECs, with the MaSC-containing basal population yielding solid organoids and the luminal progenitor-containing luminal population yielding acini (Lim et al., 2009; Shackleton et al., 2006). Indeed, Matrigel culture of basal and luminal MEC populations, FACS sorted based on CD29/EpCAM (plus GFP reporter for DKO mice), showed that control MaSC-enriched basal cells form solid organoids with ∼6% efficiency, whereas luminal cells formed acini with hollow lumens at higher efficiencies (than 6%) but solid organoids only rarely (Fig. 3A). The ability of DKO basal MECs to form solid organoids was significantly reduced, whereas acini formation of DKO luminal cells was comparable to that of controls (Fig. 3B). Similar results were obtained after harvesting and replating the first passage organoids or acini for a second passage (Fig. 3C). qPCR at the end of passage 1 confirmed lack of Cbl and Cbl-b mRNA in organoids generated from DKO basal and luminal cells, excluding the possibility of organoids/acini having arisen from precursors in which floxed Cbl was not deleted (Fig. 3D). Altogether, these results further established that loss of CBL and CBL-B in the mammary epithelium impairs the ability of MaSCs to self-renew.

The Wnt target gene Lgr5 was recently found to mark MaSCs within the basal MEC compartment (de Visser et al., 2012; Plaks et al., 2013); such cells were shown to be bi-potent and to regenerate the MG upon transplant (Plaks et al., 2013; Rios et al., 2014). We crossed the Lgr5-EGFP-IRES-CreERT2 mice with Cbl^f/f; Cbl-b^−/− mice to create mice for MaSC-selective and tamoxifen-inducible Cbl/Cbl-b DKO within the mammary epithelium. FACS analysis of MECs from parental Lgr5-GFP-Cre mice confirmed the presence of a small pool of Lgr5⁺ (Lgr5-driven GFP⁺) cells nearly exclusively in the basal population (Plaks et al., 2013) (Fig. S7A,B). Importantly, within 5 days of initiating tamoxifen-induced Cbl deletion in the DKO model, the Lgr5⁺ cells underwent a significant shrinkage compared with control (Cbl^f/+; Cbl-b^+/−) mice (Fig. 4A,B). CD29/EpCAM-based assignment of GFP⁺ cells to basal and luminal compartments indicated that the proportion of Lgr5⁺ cells in the basal compartment decreased upon tamoxifen induction, whereas the proportion in the luminal compartment increased significantly (Fig. 4C), indicating a severe depletion of MaSCs, with a proportion of the DKO MaSCs transitioning into the luminal compartment. The tamoxifen-induced deletion of Cbl (and constitutive deletion of Cbl-b) in the sorted GFP⁺, but not GFP⁻, MEC fraction was confirmed by qPCR (Fig. 4D,E), clearly establishing the MaSC-selective DKO in the Lgr5-Cre-based model.

To assess the impact of in vivo Lgr5-Cre-based Cbl/Cbl-b DKO in MaSCs on their stemness, GFP⁺ (Lgr5⁺) MECs sorted from Lgr5-EGFP-Cre⁺ (MaSC control) or conditional DKO (MaSC DKO) mice 10 days post-tamoxifen were subjected to organoid-forming assays. Whereas Lgr5⁺ cells from MaSC control MGs formed solid organoids, Lgr5⁺ cells from tamoxifen-treated MaSC DKO mice produced significantly fewer and smaller organoids (Fig. 5A-C).

The number and size of organoids formed after 7 days of Matrigel culture from FACS-sorted GFP⁺ (Lgr5⁺) MECs from uninduced MaSC control or conditional MaSC Cbl/Cbl-b DKO mice were comparable (Fig. 5D,G). Next, 4-hydroxytamoxifen (4-OH-T) was added on day 7 of culture and organoids were harvested after 96 h, replated, and examined after 8 days. Compared with uninduced cultures, 4-OH-T induction of Lgr5⁺ MECs significantly impaired the ability to form organoids upon replating (Fig. 5E,H). Further passaging confirmed these results (Fig. 5F,H). WB of harvested organoids verified CBL and CBL-B deletion (Fig. 5I,J). Organoid assays on Lgr5⁺ cells isolated from Lgr5-GFP-Cre⁺ conditional DKO mice and cultured in the presence of 4-OH-T also showed reduced organoid-forming ability, as compared culture in the absence of 4-OH-T, owing to Cbl deletion (as confirmed by qPCR; Fig. 6A-C).

Consistent with results in the MMTV-Cre-based DKO model, qPCR analysis showed that the levels of K5 and K14 were reduced in 4-OH-T-induced DKO organoids compared with non-induced controls, whereas the expression of K8 increased (Fig. 6D-F). Immunofluorescence analyses showed an increase in the proportion of K8⁺ cells and a decrease in the proportion of K14⁺ cells in MaSC DKO organoids (Fig. 6G,H).

RTK signaling is crucial for MG development and the maintenance of MaSCs (Brisken et al., 2000; van Amerongen et al., 2012). PI3K-AKT-mTOR is a key signaling node downstream of activated RTK signaling (Sun and Bernards, 2014), and studies in cell models have shown that loss of Cbl or Cbl-b increases the levels of phosphorylated (p) AKT (An et al., 2015; Fang and Liu, 2001). Importantly, hyperactivation of AKT-mTOR signaling has been shown to promote differentiation and stem cell exhaustion in hematopoietic, neural and epidermal stem cell systems (Iglesias-Bartolome et al., 2012; Kalaitzidis et al., 2012; Magri et al., 2011). In view of reduced MG development and of MaSC depletion upon Cbl/Cbl-b DKO, we assessed whether AKT-mTOR signaling is hyperactivated in DKO MECs. As the number of cells available from the Lgr5-Cre-driven model was small, we used a fully tamoxifen-inducible Cbl and Cbl-b deletion model (Goetz et al., 2016) in which both Cbl and Cbl-b are floxed, and the ROSA26 locus CreERT and dual fluorescent reporter (mT/mG) alleles allow verifiable 4-OH-T-inducible in vitro Cbl and Cbl-b deletion. The MaSC-containing basal cells from Cbl^f/f; Cbl-b^f/f; CreERT; mT/mG or Cbl^f/f; Cbl-b^f/+; CreERT; mT/mG mice were plated in organoid cultures and induced with 4-OH-T (Fig. 7A). Whereas uninduced organoids were uniformly RFP⁺, 4-OH-T induction led to a switch to GFP⁺ organoids, indicating robust CreERT-dependent gene deletion. WB verified CBL and CBL-B deletion, concurrent with upregulation of the levels of total and pEGFR, pAKT and pmTOR in 4-OH-T-induced DKO organoids as compared with uninduced organoids (Fig. 8A). Replating of uninduced versus induced Cbl^f/f; Cbl-b^f/f; CreERT; mT/mG organoids showed a substantial reduction in organoid growth upon in vitro Cbl and Cbl-b deletion (Fig. 7B,C), consistent with results of the MMTV-Cre and Lgr5-Cre models. PCR analysis of Cbl and Cbl-b expression showed significant reduction upon 4-OH-T induction (Fig. 7D,E). Egfr mRNA levels were not significantly upregulated upon 4-OH-T induction (Fig. 7F). Basal cell-derived organoids from Cbl^f/f; Cbl-b^f/+; CreERT; mT/mG mice showed only a slight reduction in organoid numbers upon induction (Fig. 7C, lower panels). qPCR analysis showed lower levels of K14 and increased levels of K8 in induced versus non-induced organoids, indicating differentiation towards luminal lineages (Fig. 7G,H). Luminal cells from Cbl^f/f; Cbl-b^f/f; CreERT; mT/mG mice yielded comparable numbers of organoids with and without 4-OH-T induction (Fig. S8A), and WB analysis showed no changes in cyclin D1 (proliferation marker) expression (Fig. S8B). WB analyses also verified CBL and CBL-B depletion and an increase in pEGFR but without an increase in total EGFR, but relatively little impact on pAKT or pERK levels (Fig. S8B). Collectively, the results of inducible deletion of both Cbl and Cbl-b further support our conclusion that CBL and CBL-B are required for MaSC maintenance, and show that their deletion hyperactivates the AKT-mTOR signaling axis.

To assess whether AKT-mTOR hyperactivation might mediate the loss of MaSCs upon Cbl and Cbl-b deletion, we tested if specific AKT or mTOR inhibitors could rescue MaSCs from Cbl/Cbl-b DKO-induced depletion. The allosteric AKT inhibitor MK-2206 (Hirai et al., 2010; Yap et al., 2011) was added to preformed organoids generated from a Cbl^f/f; Cbl-b^f/f; CreERT; mT/mG basal cell population with or without concurrent 4-OH-T induction at 50 nM or 100 nM, based on IC₅₀ values reported in other systems (Sangai et al., 2012). MK-2206 treatment markedly reduced AKT-pS473 levels, and partially reduced the levels of pmTOR (Fig. 8B). Direct mTOR inhibition with rapamycin, at 1 nM and 10 nM, led to reduction of pmTOR as well as AKT-pS473 (Fig. 8B), the latter an expected effect of prolonged rapamycin treatment causing inhibition of mTORC1 and mTORC2 complex activity (Ballou and Lin, 2008; Sarbassov et al., 2006; Weichhart et al., 2015). Replating of untreated versus inhibitor-treated organoids showed an inhibitor dose-dependent partial (AKT inhibitor) or complete (mTOR inhibitor) rescue of secondary organoid formation by Cbl/Cbl-b DKO cells as compared with the untreated control (Fig. 8C,D). These results support the conclusion that hyperactivation of mTOR signaling downstream of RTKs, in part due to hyperactivity of AKT, promotes the depletion of MaSCs in Cbl/Cbl-b DKO mammary epithelium.

CBL family ubiquitin ligases target active pools of PTKs for ubiquitylation and degradation, thus serving as activation-dependent negative-feedback regulators of cellular signaling. Although the role of CBL-C is unclear, mouse knockouts have revealed key physiological roles of CBL or CBL-B, individually as well as in combination, especially as negative regulators of immune responses and hematopoiesis (Duan et al., 2004; Mohapatra et al., 2013; Thien and Langdon, 2005). Relatively little is known about the role(s) of CBL proteins in mammalian epithelia despite their broad epithelial expression and apparent redundancy. In this study we used three complementary conditional mouse gene deletion models to reveal a novel and crucial requirement for CBL and CBL-B in the maintenance of MaSCs. Multiple lines of evidence support our conclusion.

Mammary epithelium-selective Cbl/Cbl-b DKO via floxed Cbl deletion with MMTV-Cre (Wagner et al., 1997) on a whole-body Cbl-b null background led to a significant block in MG development (Fig. 1A). Transplantation of MMTV-Cre DKO MECs demonstrated that the MG developmental block was not due to preclinical MPD driven by MMTV-Cre-induced DKO HSCs (An et al., 2015; Naramura et al., 2010) (Fig. S3B). No defect in MG development was seen in Cbl null or Cbl-b null mice, indicating that CBL and CBL-B play a redundant, cell-autonomous regulatory role in the mammary epithelium. Redundancy between CBL and CBL-B and a compensatory increase in the expression of family members, as seen in Cbl-b null MGs (Fig. 1E), is likely to account for the lack of MG phenotypes in MGs of single-KO mice. Increased mammary branching as previously reported in Cbl null mice (Murphy et al., 1998) was later shown to be due to a non-cell-autonomous increase in growth factor secretion by the stroma (Crowley et al., 2005).

Surface marker-based analyses revealed a significant reduction in the overall numbers of MECs and their basal and luminal subsets in DKO mice, with a more substantial reduction in the MaSC-containing basal population, leading to a relative increase in the proportion of luminal cells (Fig. 2A-D). Interestingly, CD61/CD29-based MEC subset analysis indicated an increase in luminal progenitors but not the mature luminal cell pool. Together with a smaller basal population, this finding is consistent with a more rapid transition of DKO MaSCs into luminal progenitors. We suspect that such cells are functionally impaired since there was not a proportionate increase in mature luminal cells (Fig. 2G,H). Importantly, in vitro and in vivo assays on basal MECs derived from the MMTV-Cre-driven DKO mice demonstrated impaired MaSC self-renewal and ability to regenerate a MG.

The requirement of CBL and CBL-B for the maintenance of MaSCs was independently confirmed by selectively inducing Cbl/Cbl-b DKO in Lgr5⁺ MECs, which have recently been established as bi-potent MaSCs with MG-regenerating ability (Barker et al., 2007, 2013). Tamoxifen-induced, MaSC-specific Cbl/Cbl-b DKO led to a rapid depletion of Lgr5⁺ cells (Fig. 4B), with the redistribution of a significant proportion of the remaining Lgr5⁺ MECs into luminal compartment. The reduction in MaSC numbers and their apparent transition into luminal cells appear to represent an abnormally accelerated differentiation of MaSCs at the expense of their self-renewal, as demonstrated by organoid studies (Fig. 5A,B,E,F,H, Fig. 6A,B). In vitro deletion of Cbl and Cbl-b in basal MECs from CreERT; Cbl^f/f; Cbl-b^f/f mice also led to reduced organoid-forming ability, complementing results from the other models (Fig. 7B,C). Although our immunofluorescence and qPCR analyses in multiple models support the idea of differentiation of MaSCs towards luminal lineages, mammary lineage tracing in Cbl^flox/Cbl-b^flox mice bearing the dual reporter, possibly using K14-Cre (Van Keymeulen and Blanpain, 2012; Visvader and Stingl, 2014), will be needed to definitively test this idea.

Since negative regulation of PTKs is the primary biochemical function of CBL family proteins (Mohapatra et al., 2013), we surmised that dysregulated signaling downstream of PTK-coupled receptors in MaSCs was responsible for the rapid MaSC depletion upon Cbl and Cbl-b deletion. As multiple potential CBL/CBL-B target RTKs, including ErbB receptors, c-MET, IGF1/2 receptors and FGF receptors, are known to regulate MG development (Gjorevski and Nelson, 2011; Hynes and Watson, 2010; Sternlicht, 2006), we focused on AKT-mTOR signaling because it is a pivotal signaling axis downstream of all RTKs and non-receptor PTKs (Sun and Bernards, 2014), and its hyperactivity is associated with stem cell exhaustion in hematopoietic, neural and epidermal stem cells (Iglesias-Bartolome et al., 2012; Kalaitzidis et al., 2012; Magri et al., 2011). Indeed, inducible CreERT-mediated Cbl/Cbl-b DKO demonstrated an upregulation of pAKT and pmTOR in addition to the expected increase in pEGFR and EGFR levels, as organoids are cultured in EGF (Fig. 8A). CBL/CBL-B proteins are well-established negative regulators of RTKs. We chose EGFR as an index RTK since the organoids are grown in the presence of EGF and FGF2. As the use of an EGFR inhibitor (or lack of addition of EGF in the culture medium) would make the organoid assay unfeasible, we focused on downstream signaling pathways that are shared between RTKs activated by the added ligands as well as those that might be produced by MaSCs themselves. Our results raised the possibility that hyperactive mTOR signaling might commit Cbl/Cbl-b DKO MaSCs to premature differentiation, thereby reducing their self-renewing pool. A chemical inhibitor strategy supported such a role of hyperactive AKT-mTOR signaling. However, whether hyperactive mTOR is sufficient for MaSC depletion needs to be directly tested; our attempts to test this by expressing constitutively active or dominant-negative RHEB (Nie et al., 2010) were unsuccessful due to technical difficulties in lentiviral infection of primary organoids. The partial rescue of MaSC self-renewal by AKT inhibition versus an essentially complete rescue by mTOR inhibition (Fig. 8B-D) supports a model in which the hyperactivity of AKT and other upstream signaling pathways increases mTOR signaling to deplete MaSCs by promoting their rapid differentiation. Among the upstream pathways that might add to increased mTOR activity, ERK is a likely candidate as it is hyperphosphorylated when Cbl and Cbl-b are genetically deleted in HSCs (An et al., 2015) or depleted by shRNA in human MECs (Duan et al., 2011). Culture of human mammary organoids with amphiregulin has been found to maintain MECs in a less differentiated state, whereas growth in EGF led to myoepithelial differentiation; this dichotomy was shown to be related to sustained ERK/RSK signaling by EGF (Pasic et al., 2011).

Although our results support the likelihood that hyperactive mTOR results in rapid differentiation and loss of MaSCs, it is possible that the increased differentiation is a consequence rather than a cause of MaSC depletion. In this context, CBL is unequally distributed in daughter cells of asymmetrically dividing neural stem cells, with higher CBL correlating with lower EGFR levels in daughter cells that retain stemness (Ferron et al., 2010). Future studies to investigate a potential role of CBL and CBL-B in promoting asymmetric division in MaSCs will therefore be of interest.

Our studies provide the first evidence for a cell-intrinsic and redundant requirement of CBL and CBL-B in epithelial stem cell self-renewal and maintenance, provide a mechanistic basis for this requirement and suggest that fine-tuning of mTOR signaling is crucial for MaSC maintenance. These insights, together with models generated here, should help uncover an unanticipated positive role of CBL family proteins in epithelial stem cell maintenance, with implications for epithelial stem cell biology and cancers.

The following antibodies were used for WB: mouse monoclonal antibodies to CBL (clone 17/c-Cbl, BD Biosciences, 610442; 1:1000), HSC70 (clone B6, Santa Cruz, SC-7298; 1:10,000), cytokeratin 8 (Troma-1, Developmental Studies Hybridoma Bank; 1:200) and Ki67 (Alexa 488 conjugated, BD Biosciences, B56 clone, 558616; 1:200); rabbit monoclonal antibodies to CBL-B [clone D3C12, Cell Signaling Technology (CST), 9498; 1:1000], phospho-AKT (Ser473, CST, 4060; 1:500), pan-AKT (CST, 4691; 1:1000), phospho-mTOR (Ser2448, CST, 5536; 1:1000), mTOR (7C10, CST, 2983; 1:1000) and cyclin D1 (SP4, Abcam, ab16663; 1:200) and rabbit polyclonal antibodies to cytokeratin 5 (Covance, PRB-106P; 1:500), cytokeratin 14 (Covance, PRB-155P; 1:200), cleaved caspase 3 (CST, 9662; 1:200), EGFR (sc-03, Santa Cruz, 1005; 1:500) and phospho-EGF receptor (Tyr1068, CST, 2234; 1:1000). Anti-CD24-PE (clone M1/69, BD Biosciences, 553262; 1:50), anti-EpCAM-PE (eBioscience, 12-5791-82; 1:50), anti-CD29-APC (clone eBioHMb1-1, eBiosciences, 17-0291-82; 1:50) and anti-CD61-BV421 (clone 2C9.G2, BD Horizon, 562917; 1:50) were used for flow cytometry. Biotinylated anti-CD45 (clone 30-F11, BD Biosciences, 553077; 1:50), anti-Ter119 (clone ter119, BD Biosciences, 553672, 1:50) and anti-CD31 (clone MEC 13.3, BD Biosciences, 553371; 1:50) antibodies were used for AUTOMACS depletion of Lin⁺ cells.

All reagents and media used for primary mammary epithelial isolation and in vitro organoid and mammosphere cultures were from Stem Cell Technologies, unless otherwise specified. ROCK inhibitor was from Tocris Bioscience. Tamoxifen, 4-OH-T, recombinant EGF, FGF and rapamycin (R0395) were from Sigma. MK-2206 2HCl was from Selleckchem (S1078). Reduced growth factor Matrigel was from BD Biosciences.

Mice carrying MMTV-Cre [stock Tg(MMTV-Cre)4 Mam/J] for conditional floxed Cbl deletion (Cbl^tm2Hua/Cbl^f/f) (Naramura et al., 2010) on a Cbl-b null (Cbl-b^tm1Hua) background (Cbl/Cbl-b DKO) (all genotypes are listed in Table S1) as well as Cbl KO (Cbl^tm1Hua) (Naramura et al., 1998) and Cbl-b KO (Cbl-b^tm1Hua) (Chiang et al., 2000) mice have been described previously. ROSA26 locus lacZ reporter [B6.129S4-Gt(ROSA)26Sor^tm1Sor/J] and EGFP reporter (CAG-CAT-EGFP) (Naramura et al., 2010) were incorporated to track cells with Cre-dependent target gene deletion. To generate a tamoxifen-inducible MaSC Cbl/Cbl-b DKO model, the Lgr5-GFP-IRES-CreERT allele (The Jackson Laboratory) was incorporated into Cbl^f/f; Cbl-b^−/−; R26R-lacZ mice. Control mice were littermates obtained during breeding. We engineered a Cbl-b^f/f mouse model and used it to generate a Cbl^f/f; Cbl-b^f/f; CreERT; mT/mG strain that was used for in vitro Cbl and Cbl-b deletion in MaSCs in certain experiments (Goetz et al., 2016). Three-week-old female NSG mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ) (The Jackson Laboratory) were used as recipients for mammary transplantation. All mouse strains were maintained on a C57BL/6J background under specific pathogen-free conditions and genotyped using tail DNA PCR (sequences are provided in Table S2). All experimental analyses were carried out on female mice. The Institutional Animal Care and Use Committee of University of Nebraska Medical Center approved all mouse experiments.

Fourth (inguinal) MGs were excised and processed according to a standard protocol (LaMarca and Rosen, 2007). Whole-mounts were photographed and various parameters (branch points, ductal length and fat pad filling) were quantified using Nikon Elements software. Whole MG X-gal staining, to visualize lacZ-encoded β-galactosidase activity as a reporter of gene deletion, was performed as described (LaMarca and Rosen, 2007). At least four animals were analyzed per genotype and time point.

Staining of paraffin-embedded MG sections and mammary organoids with fluorescently labeled antibodies was carried out using standard protocols as previously described (Pasic et al., 2011). WB of freshly isolated MG and organoids was performed as described in the supplementary Materials and Methods.

Formalin-fixed and paraffin-embedded tissue sections were stained with anti-CBL antibody using the Mouse on Mouse (MOM) and ABC Amplification IHC Kit (Vector Laboratories) as per the manufacturer's instructions. The slides were counterstained with Hematoxylin. In all cases, a species-matched (and isotype-matched for mouse monoclonal antibodies) control IgG was used as a negative control.

The mammary epithelial tissue between the nipple and lymph node of the fourth MG of 3-week-old immune-compromised NSG female mice was removed to yield epithelium-cleared fat pads (Deome et al., 1959). Complete removal of the mammary epithelium was verified by whole-mount staining. Recipient mice received a control mammary transplant on one side and a Cbl/Cbl-b DKO transplant on the contralateral side. The transplants were harvested from virgin hosts 8 weeks post-transplantation. Alternatively, the recipients were bred with fertile males and the glands harvested at various points of pregnancy. The whole-mounts of transplants were stained with carmine alum or X-gal.

To test MG repopulating activity, 50,000 Lin⁻ single MECs isolated from 6-week-old DKO or control MGs were suspended in 100 μl 1:1 mix of Matrigel and Complete Mouse Epicult-B medium (STEMCELL Technologies, 05610) and then injected into epithelium-cleared mammary fat pads of 3-week-old NSG mice. In other experiments, 5000 Lin⁻/EpCAM⁺ cells were directly sorted into Epicult-B medium, centrifuged (400 g for 5 min at 4°C) and resuspended in 50 μl 1:1 Epicult-B medium and Matrigel, and injected into epithelium-cleared fat pads. Mammary fat pads were harvested 8 weeks after transplant and stained with carmine alum to assess the extent of mammary outgrowths.

Single cells were prepared from mammary tissue as previously described (Stingl et al., 2006a,b). Single-cell suspensions were stained with a combination of biotinylated anti-CD45, anti-Ter119 and anti-CD31 antibodies followed by anti-biotin microbeads (Miltenyi Biotech) to magnetically deplete the Lin⁺ cells using AUTOMACS (Miltenyi Biotech). Lin⁻ cells were co-stained with the indicated combinations of anti-CD24-PE, anti-EpCAM-PE, anti-CD29-APC and anti-CD61-BV421. FACS analysis and cell sorting were performed on a FACSAria II (BD Biosciences) and the results analyzed using FlowJo software.

To induce Lgr5-CreERT-mediated floxed Cbl deletion in vivo, 4-week-old mice received three consecutive daily intraperitoneal injections of tamoxifen (1 mg/20 g body weight) in corn oil (10 mg/ml stock). Mice were analyzed at day 5 or 10 as indicated.

In vitro mammary epithelial organoid culture was performed as described (Guo et al., 2012) with minor modifications. FACS sorting of single MECs was used to prepare the GFP⁺ fraction from control or DKO mice carrying the Lgr5-GFP-CreERT allele, or the RFP⁺ basal cell fraction from Cbl^f/f; Cbl-b^f/f; CreERT; mT/mG or Cbl^f/f; Cbl-b^f/+; CreERT; mT/mG (control) mice. Cells were seeded at 2000-5000 per well in 100% Matrigel in 8-well Lab-Tek plates (Nunc) precoated with Matrigel and fed with complete Epicult-B medium and 5 μM Y-27632 (ROCK inhibitor; TOCRIS, 1254). The medium was replaced every 2 days. Organoids (>100 μm in diameter) were imaged and counted 7-14 days after seeding. For immunofluorescence analyses, organoid cultures were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100 in PBS and processed as described previously (Ewald et al., 2008).

In vitro mammosphere assay of MaSC renewal was performed as described (Guo et al., 2012). GFP⁺ basal (CD29^hi EpCAM^low) and luminal (CD29^low EpCAM^hi) cell populations from DKO and control mice carrying the MMTV-Cre-GFP allele or the GFP⁺ fraction of single MECs from control or DKO mice carrying the Lgr5-GFP-CreERT allele, or the RFP⁺ basal cells sorted from Cbl^f/f; Cbl-b^f/f; CreERT; mT/mG and Cbl^f/f; Cbl-b^f/+; CreERT; mT/mG (control) were prepared by FACS sorting and seeded in complete Epicult-B medium containing 5% Matrigel, as above, at 1000-2000 per well in 96-well ultra low-attachment plates (Corning). Mammospheres were imaged and counted (>100 μm in diameter) 7-10 days after seeding. For serial passaging, mammospheres were harvested using 5 mg/ml dispase I (Stem Cell Technologies) and dissociated in 0.25% trypsin-EDTA (Gibco). Single cells were seeded (2000 per well) for 7-10 days followed by imaging, counting, and further passaging of mammospheres.

Sorted GFP⁺ MECs from Cbl^fl/+; Cbl-b^+/−; Lgr5-CreERT (control) or Cbl^fl/fl; Cbl-b^−/−; Lgr5-CreERT (DKO) mice were plated in 100% Matrigel in Matrigel-coated 8-well chambers and cultured in Epicult-B medium with ROCK inhibitor for 7 days (Guo et al., 2012). The medium was replaced every 2 days. For in vitro Cre activation, 400 nM 4-OH-T in Epicult-B medium was added for 96 h beginning at day 3, replaced with regular medium for 1 day, and organoids harvested. 2000 cells were then replated per well in 96-well ultra low-attachment plates to assess mammosphere-forming efficiency.

RFP⁺ basal and luminal cells were isolated from Cbl^fl/f; Cbl-b^f/f; CreERT; mT/mG and Cbl^fl/f; Cbl-b^f/+; CreERT; mT/mG mice and plated in 96-well ultra low-attachment plates for 7 days as described above. 400 nM 4-OH-T was added in Epicult-B medium for 96 h. GFP expression in organoids indicates induction of Cre and deletion of floxed genes. WBs were performed to test the signaling events and CBL and CBL-B deletion. Basal organoids were treated with MK-2206 (50 nM, 100 nM) or rapamycin (1 nM, 10 nM) with and without 4-OH-T induction for 4 days. Single cells were replated, and organoids imaged and quantified. Non-induced cultures were treated with inhibitors alone to control for any non-specific effects. DMSO was used as a vehicle control.

FACS-sorted cells were directly collected in Trizol reagent (Ambion, Life Technologies) for RNA extraction. Detailed procedures are provided in the supplementary Materials and Methods and qPCR primers are listed in Table S3.

Quantified results of whole-mounts, qPCR, IHC and flow cytometry were compared between groups using Student's t-test and are presented as mean±s.e.m., with P≤0.05 deemed significant. Statistical analysis and graphical representation of data were performed using Prism 4.0c (GraphPad).

We thank the H.B. laboratory members for helpful discussions, Kay-Uwe Wagner and Aleata Triplett for advice on mammary transplants, the staff of UNMC Cores for assistance and Dr Andrew Ewald (Johns Hopkins School of Medicine) for sharing his protocol for mammary organoid staining.