Distinct FAK Activities Determine Progenitor and Mammary Stem Cell Characteristics

The mammary epithelium, mainly composed of an inner layer of luminal mammary epithelial cells (MaEC) and an outer layer of basal MaECs, is organized in a hierarchical manner (1–5). A single multipotent mammary stem cell (MaSC) in the basal layer can reconstitute a functional mammary gland by generating lineage-restricted progenitor cells, as shown in transplantation studies (2, 3, 6). In contrast, recent lineage-tracing experiments have alternatively proposed that distinct unipotent MaSC populations, located in the luminal and basal compartments, contribute to mammary gland development and maintenance under physiologic conditions (7). Currently, the signaling mechanisms regulating these MaSC/progenitor populations remain to be characterized.

Breast cancer is a heterogeneous disease with 6 distinct subtypes based on gene expression profiling (8–11), suggesting possible origins from different subsets of MaECs in the mammary epithelial hierarchy. Indeed, genome-wide transcriptome analyses of different subtypes of breast cancers, as well as MaEC subpopulations in human BRCA1 mutation carriers, suggest that basal-like breast tumor may originate from aberrant luminal progenitors, whereas claudin-low subtype is closely associated with the signature of basal MaSC-enriched subsets (5, 12). However, direct experiments involving the selective depletion of potential tumor-initiating cell populations have not been reported.

Focal adhesion kinase (FAK), which mediates signaling pathways initiated by integrins and other receptors to regulate diverse cellular functions via kinase-dependent and -independent mechanisms (13–15), has been implicated in the development and progression of breast and other cancers (16–22). Furthermore, we found that loss of FAK decreased the content of mammary cancer stem cells (MaCSC) and compromised their self-renewal and tumorigenicity (18), suggesting that FAK may serve as a potential target in MaCSCs. However, it is unknown whether and how distinct activities of FAK contribute to different breast cancer subtypes possibly from different cells of origin. In this study, we show that FAK regulates MaSCs/progenitor activities via both kinase-dependent and -independent mechanisms that, in turn, affect normal mammary gland development as well as tumorigenesis and the maintenance of MaCSCs in different breast cancer subtypes.

FAK Ctrl (FAK^f/f), MFCKO (FAK^f/f, MMTV-Cre), and MMTV-PyMT transgenic mice have been described previously (18, 23, 24). Mammary FAK conditional KD knockin (MFCKD) mice were created mating the FAK^KD/+ mice (25) with MFCKO mice. Mammary FAK conditional knockout (MFCKO) and MFCKD mice were mated with GFP transgenic mice (Jackson Laboratory, Stock Number: 003516) to obtain MFCKO-GFP (FAK^f/f, MMTV-Cre, GFP), MFCKD-GFP (FAK^f/KD, MMTV-Cre, GFP), and corresponding Ctrl-GFP (FAK^f/f, GFP; FAK^f/+, MMTV-Cre, GFP or FAK^f/KD, GFP) mice. They were also crossed with MMTV-PyMT mice to obtain three cohorts of MFCKO-MT (FAK^f/f, MMTV-Cre, MMTV-PyMT), MFCKD-MT (FAK^f/KD, MMTV-Cre, MMTV-PyMT), and Ctrl-MT (FAK^f/+, MMTV-Cre, MMTV-PyMT; FAK^f/KD, MMTV-PyMT, or FAK^f/f, MMTV-PyMT) mice. Monitoring of mammary tumor formation was described previously (18). All procedures using mice were carried out following the guidelines of The Unit for Laboratory Animal Medicine at the University of Michigan (Ann Arbor, MI). The genotyping is described in the Supplementary Materials and Methods.

Preparation and culture of mouse MaECs or tumor cells from the virgin glands or mammary tumors is described in the Supplementary Materials and Methods or described previously in ref. 18. Normal human breast tissues were obtained from reduction mammoplasties of premenopausal woman patients at the University of Michigan health system according to approved Institutional Review Board (IRB) protocols for research in human subjects (UM IRBMED #2001-0344). They were used to prepare human MaECs as described in the Supplementary Materials and Methods. Breast cancer cell lines SUM159 and SUM149 obtained from Dr. Stephen Ethier have been extensively characterized (26). MDA-MB231 and HCC1954 cell lines were purchased from American Type Culture Collection and maintained in culture conditions according to supplier's recommendation (see Supplementary Materials and Methods for detailed conditions).

Recombinant adenoviruses encoding FAK or its mutants and lentiviruses encoding FAK short hairpin RNA (shRNA)/GFP or scrambled sequence/GFP have been described previously (27, 28). The detailed procedure for the infection of human and mouse MaECs and cancer cell lines with the viruses is described in the Supplementary Materials and Methods.

The detailed list of antibodies and chemicals is described in the Supplementary Materials and Methods. MaECs or mammary tumor cells were suspended, labeled, and analyzed by flow cytometry as described in the Supplementary Materials and Methods.

Mammosphere culture of mouse MaECs, primary human MaECs, and human breast cancer cells was conducted as previously described (29) and see Supplementary Materials and Methods for details. Apoptosis and cell proliferation were examined using TMR red (Roche) or cleaved caspase-3 labeling and Ki67 or proliferating cell nuclear antigen (PCNA) immunofluorescence. The detailed procedures for these experiments and Western blotting are described in the Supplementary Materials and Methods.

Mouse unsorted, luminal, or basal MaECs at serial dilutions were used for transplantation assays to determine mammary repopulating unit (MRU) as described in the Supplementary Materials and Methods.

The mammary ductal elongation and outgrowth after transplantation were examined by carmine staining of mammary whole mounts and observed under a dissecting microscope. In transplantation experiments using GFP-labeled donor cells, the whole mounts were observed under a fluorescent microscope. Mammary gland of virgin or pregnancy/lactation mice were also used for histology using hematoxylin and eosin (H&E) staining and immunofluorescence. The detailed procedures are described in the Supplementary Materials and Methods.

Total RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. For quantitative real-time PCR (qRT-PCR) analyses, equal amounts of RNA were reverse-transcribed by SuperScript III first-strand synthesis system (Invitrogen) with oligo(dT) as a primer, and then the resulting cDNA templates were subjected to qRT-PCR using the SYBR Green PCR Core Reagents system (Qiagen). Primer sequences are available upon request.

Statistical significance was evaluated by paired Student t test or two-way ANOVA using P < 0.05 as indicative of statistical significance. Kaplan–Meier tumor-free survival data were compared using the log-rank test.

To define the roles of FAK in basal MaSCs and luminal progenitors, we analyzed MaECs from Ctrl and MFCKO mice (24) by flow cytometry using established markers (1–3). Three major subsets were obtained: CK8/18⁺ luminal cells (R6, Lin⁻CD24^hiCD29^lo), CK5⁺ basal/myoepithelial cells (R7, Lin⁻CD24^loCD29^mod-hi), and CK8/18⁻CK5⁻ mammary stromal cells (lower, Lin⁻CD24⁻CD29^lo, Fig. 1A and Supplementary Fig. S1A). Interestingly, in MFCKO mice, the MaSC-enriched fraction (Lin⁻CD24^loCD29^hi; R9 as a R7 subpopulation) was significantly reduced relative to that of Ctrl mice (Fig. 1B). We further verified that the R9 subpopulation contained a higher fraction of CK8⁺/CK5⁺ cells (characteristic of MaSCs) as opposed to the remainder of the R7 cells (Supplementary Fig. S1B), supporting that it is enriched in MaSC activity (2, 3, 30). Further examination of the luminal population by CD61 (1) revealed that FAK deletion also reduced luminal progenitors (R10, Lin⁻CD24^hiCD29^loCD61⁺; Fig. 1C). Together, these results show that FAK deletion in MaECs decreases the content of both basal MaSCs and luminal progenitors.

We next examined mammosphere formation from Ctrl and MFCKO MaECs to assess their self-renewal potential (29). MaECs from Ctrl mice formed mammospheres (diameter > 40 μm), whereas MaECs from MFCKO mice mostly remained as single or small clusters of cells (Fig. 1D). Quantitation of the mammospheres with diameters more than 40 μm and 80 μm verified a decrease in mammosphere formation by MFCKO MaECs compared with Ctrl MaECs (Fig. 1E). Similar results were obtained when MaECs were analyzed for spherical acini formation in three-dimensional Matrigel culture (Supplementary Fig. S2A). These results were not confined to murine MaECs, as knockdown of FAK by shRNA in human MaECs also greatly reduced mammosphere formation (Fig. 1F–H). Together, these results support a model wherein FAK deletion impaired self-renewal of basal MaSCs and luminal progenitors, thereby reducing the content of these populations in MFCKO mice.

To determine directly the role of FAK in maintaining MaSCs in vivo, we used limiting dilution transplantation assays to assess the regenerative potential of MaSCs from Ctrl and MFCKO mice (2, 3). Analysis of unsorted MaECs and Lin⁻CD24^loCD29^mod-hi subpopulations enriched in basal MaSCs resulted in an estimated MRU frequency of 1/634 and 1/407, respectively, in Ctrl mice, that was greatly diminished in MFCKO mice (<1/9514 and 1/11493; Table 1 and Fig. 1I). Outgrowths were also generated from the Lin⁻CD24^hiCD29^lo luminal population of Ctrl (at low frequency and reduced size), but not MFCKO mice (Table 1). To ensure that the reduced MaSC/progenitor activity of MFCKO mice was caused by intrinsic defects in MaECs, we further examined MaECs from Ctrl (i.e., FAK^f/f) mice following deletion of FAK by Cre recombinase in vitro. FAK ablation was verified in Ad-Cre–infected Ctrl MaECs, and as predicted, these cells exhibited reduced mammosphere formation relative to control infected cells (Supplementary Fig. S2B). Surprisingly, although lower than control infected cells (8/22 vs. 7/12), a fraction of the Ad-Cre–infected MaECs generated outgrowths in transplantation assays (Supplementary Table S1). Genotyping of parts of outgrowths showed that those from Ad-LacZ–infected cells retained the floxed FAK alleles (Fig. 1J, lanes 1–5; Supplementary Fig. S3A, lane 1 and 2), whereas the recipient fat pads without outgrowth only contained the wild-type (WT) alleles (Fig. 1J, lane 6 and lane 11–16 and Supplementary Fig. S3A, lanes 3–6 and 11–18), as expected. Of the eight outgrowths generated from Ad-Cre–infected cells, only two small outgrowths displayed the FAK^Δ/Δ genotype (Fig. 1J, lane 10 and Supplementary Fig. S3A, lane 10), whereas the remainder exhibited a FAK^f/Δ genotype (Fig. 1J, lanes 7–9 and Supplementary Fig. S3A, lane 7–9), suggesting that outgrowths derived from Ad-Cre–infected cells were due to incomplete deletion of the floxed alleles. Together, these results show that FAK is required for the self-renewal and regenerative potential of MaSCs/progenitor cells in a cell-autonomous manner.

To investigate the mechanisms by which FAK regulates MaSCs/progenitor cells, a series of FAK mutants were analyzed for their ability to rescue the deficient sphere/acini-forming activity of FAK-null MaECs in vitro. As shown in Fig. 2A, both FAK and the D395A (defective for binding to p85 of PI3K) mutant were autophosphorylated at Y397, whereas no autophosphorylation was detected in cells transduced with the Y397F (autophosphorylation defective) or KD (K454 to R mutation, kinase-defective) mutants. Furthermore, FAK or the D395A, but not the Y397F or KD mutant, rescued the impaired sphere-forming activity of FAK-null MaECs (Fig. 2B). Similar results were obtained when these cells were analyzed for acini formation (Fig. 2C and D). These results show that FAK kinase activity and its autophosphorylation at Y397, but not downstream PI3K activation, are important for the mammosphere/acini-forming activity of MaECs.

To further examine FAK signaling in mammary regenerative potential, MFCKO and Ctrl mice were crossed with GFP transgenic mice to generate FAK^f/f;GFP (Ctrl-GFP); FAK^f/+;MMTV-Cre;GFP (Cre Ctrl-GFP) and FAK^f/f; MMTV-Cre;GFP (MFCKO-GFP) mice, thereby facilitating identification of donor cells in transplantation assays. As expected, limiting dilution transplantation of MaECs from Ctrl-GFP or Cre Ctrl-GFP, but not MFCKO-GFP mice, generated mammary outgrowths with strong GFP signals (Supplementary Table S2). MaECs from MFCKO-GFP mice were then transduced with adenoviruses expressing FAK, KD, or GFP alone, and used in transplantation assays. Comparable expression levels of FAK and the KD mutant were observed, whereas Y397 phosphorylation was only detected in cells expressing FAK, but not the KD mutant (Fig. 2E). Reexpression of FAK but not GFP alone, rescued the mammary repopulating activity of FAK-null MaECs (Fig. 2F), generating mammary outgrowths in 6 of 10 transplants. Surprisingly, reexpression of the KD mutant also restored mammary outgrowth in 8 of 14 transplants, indicating that FAK kinase activity is disposable for the regenerative potential of MaSCs.

To better characterize the role of FAK kinase activity in MaSCs/progenitor cells in vivo, MFCKO mice were crossed with FAK^KD/+ mice (25) to produce MaEC-specific KD knockin (FAK^KD/f;MMTV-Cre; designed as MFCKD) mice. In virgin mice, comparable mammary ductal invasion and branch growth were observed in MFCKD and Ctrl mice, which is in contrast with MFCKO mice that displayed compromised ductal outgrowth (24; Fig. 3A–C). During pregnancy, however, both MFCKD and MFCKO mice display marked defects in milk production (data not shown). Whole mount staining revealed that despite normal ductal architecture, the lobulo-alveolar units in MFCKD and MFCKO mammary glands were smaller and more sparse relative to Ctrl mammary glands (Fig. 3D, left). Furthermore, while the Ctrl mammary glands were filled with large lobulo-alveolar units composed of multiple individual alveoli (arrow) lined by flattened epithelial cells, MFCKD and MFCKO mammary glands were dominated by adipose-rich stroma, and the parenchyma primarily consisted of dilated ductal networks and small clusters of alveoli (arrows; Fig. 3D, right).

To explore the cellular basis of the preferential effect of KD knockin mutation on alveologenesis as opposed to ductal outgrowth, the luminal and basal compartments were examined for cell proliferation in virgin females. Double-labeling immunofluorescence revealed abundant proliferating luminal (Ki67⁺CK8⁺) cells in the ducts of Ctrl, but not MFCKO or MFCKD mice (Fig. 3E and F). However, similar analysis showed that the number of proliferating basal (PCNA⁺CK5⁺) cells was significantly affected only in MFCKO, but not MFCKD, mice when compared with Ctrl mice (Fig. 3G and H). This labeling also verified the decreased luminal cell proliferation (PCNA⁺CK5⁻ cells) in the ducts of MFCKO and MFCKD mice (Fig. 3G and I). Together, these results suggest that FAK kinase activity plays a preferential role in controlling proliferation of luminal MaECs and alveologenesis, whereas the kinase-independent functions of FAK are sufficient to promote basal MaEC proliferation, ductal outgrowth, and branching morphogenesis.

The differential requirement of FAK kinase activity to promote proliferation of luminal, but not basal, MaECs raised the interesting possibility that basal MaSCs and luminal progenitors are maintained through distinct functions of FAK. To investigate this possibility, we first examined MaSC and luminal progenitor contents in MFCKD mice using phenotypic markers (Fig. 4A). Relative to Ctrl and MFCKO mice (see Fig. 1A–C), decreased content of luminal progenitors, but not basal MaSCs, was found in MFCKD mice (Fig. 4B). Genotyping of luminal and basal MaECs verified that FAK was effectively ablated in both compartments of MFCKO and MFCKD mice, whereas the KD allele was not excised in either compartment of MFCKD mice (Fig. 4C). Interestingly, while luminal MaECs from MFCKO and MFCKD mice both showed reduced acini-forming activity compared with those from Ctrl mice, basal MaECs from MFCKD, but not MFCKO, mice generated acini at levels comparable to Ctrl mice (Fig. 4D and E). Ki67 staining revealed abundant proliferating cells in Ctrl luminal acini as well as both Ctrl and MFCKD basal acini, but few were detected in MFCKO and MFCKD luminal acini or MFCKO basal acini (Fig. 4F–H). Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling staining did not reveal any apoptotic cells in the luminal cells in any of the genotypes (except those detached cells in the hollow space; arrows), but modest increases in apoptosis were found in the basal acinar cells from MFCKO, but not Ctrl or MFCKD mice (Supplementary Fig. S3B). Examination in basal MaECs for the expression of several transcription factors implicated in regulating MaSC generation (31, 32) showed that while the expression levels of Snail, Slug, and Sox9 were all reduced in basal MaECs from MFCKO mice, normal levels were maintained in MFCKD basal cells (Fig. 4I and J). Together, these results suggest that FAK kinase activity is required to promote cell proliferation to maintain colony-forming activity of luminal progenitors, but FAK kinase-independent functions are sufficient to support basal MaSC activity through regulation of both survival and proliferation and possibly by maintaining the expression of key transcription factors such as Snail, Slug, and/or Sox9.

We next conducted transplantation assays to evaluate the role of FAK kinase-dependent and -independent activity in maintaining the regenerative potential of the basal and luminal MaECs. As expected, MRU activity in the luminal compartment of MFCKD-GFP mice was decreased to levels similar to those found in MFCKO mice (Supplementary Table S3). Interestingly, while very low MRU activity was detected in basal MaECs of MFCKO-GFP mice (<1/12818), the MRU activity found in basal MaECs of MFCKD-GFP mice (1/530) was comparable with that detected in Ctrl-GFP mice (1/616, Fig. 4K and Supplementary Table S3), thereby confirming the kinase-independent function of FAK in maintaining the regenerative potential of basal MaSCs. However, in contrast with the expanded mammary outgrowth and large lobulo-alveolar units observed in Ctrl-GFP transplants during pregnancy, the recipient fat pads transplanted with basal MaECs from MFCKD-GFP mice displayed little expansion of mammary outgrowths, with only small numbers of lobulo-alveolar units having been formed (Fig. 4K). Hence, basal MaSCs harboring FAK KD mutation are unable to support pregnancy-induced alveologenesis, despite their ability to generate complete mammary outgrowths.

The finding that FAK kinase activity is specifically required to maintain luminal progenitors, but not basal MaSCs, prompted us to examine whether this particular FAK function is important for tumorigenesis and maintenance of MaCSCs in tumors arising from luminal progenitors. Thus, we generated MaEC-specific FAK^KD knockin mutation in the MMTV-PyMT breast cancer model that specifically targets luminal progenitors (33). Three cohorts of female mice with the genotype FAK^f/KD;MMTV-Cre;MMTV-PyMT (designated MFCKD-MT mice), FAK^f/KD;MMTV-PyMT, FAK^f/f;MMTV-PyMT or FAK^f/+;MMTV-Cre;MMTV-PyMT (collectively designated Ctrl-MT mice), and FAK^f/f;MMTV-Cre;MMTV-PyMT (designated MFCKO-MT mice) were established and examined for mammary tumorigenesis. Interestingly, similar to MFCKO-MT mice (18), MFCKD-MT mice displayed an increased tumor-free interval compared with Ctrl-MT mice (Fig. 5A). Furthermore, decreased number of tumors (Fig. 5B) and reduced tumor growth (Fig. 5C) were also found in both MFCKO-MT and MFCKD-MT mice relative to Ctrl-MT mice. Consistent with these results, tumor cell proliferation was markedly decreased in MFCKD-MT and MFCKO-MT mice when compared with Ctrl-MT mice (Fig. 5D). Western blot analysis of lysates from primary tumors showed that FAK expression was abolished in MFCKO-MT tumors and also decreased in MFCKD-MT tumors (likely due to the remaining KD allele) relative to Ctrl-MT tumors (Fig. 5E). Moreover, activated FAK (as measured by pY397) was not detected in MFCKO-MT or MFCKD-MT tumors. As reported previously for MFCKO-MT tumors (18), cyclin D1 expression was also decreased in MFCKD-MT tumors. Together, these results show that loss of FAK kinase activity in MFCKD-MT mice suppressed MMTV-PyMT–induced mammary tumorigenesis and tumor growth.

To explore whether the loss of FAK kinase activity affects the MaCSCs maintenance, we first conducted flow cytometry for tumor cells and corresponding normal MaECs from various mouse models. As expected (see Fig. 1A), the luminal and basal MaECs from Ctrl, MFCKO, and MFCKD mice segregated in gate R5 (luminal) and R6 (basal), respectively (Supplementary Fig. S4A, top). Interestingly, using these gating criteria, tumor cells from Ctrl-MT, MFCKO-MT, and MFCKD-MT mice were confined entirely to the luminal gate R5 (Supplementary Fig. S4A, bottom), supporting the notion that tumors arising in the MMTV-PyMT model originate in the luminal compartment (33), thereby defects in luminal progenitors upon loss of FAK kinase activity reduced tumorigenesis in MFCKD-MT mice. We next assessed MaCSC content directly using the Lin⁻CD24⁺CD29⁺CD61⁺ signature as described previously for MMTV-PyMT model (27, 34). The fraction of Lin⁻CD24⁺CD29⁺CD61⁺ tumor cells in MFCKD-MT and MFCKO-MT mice was significantly lower than that found in Ctrl-MT mice (Fig. 5F and G). Hence, loss of FAK kinase activity suppressed mammary tumorigenesis in association with a dramatic decrease in MaCSC content via depletion of luminal progenitors (i.e., the proposed cell of origin for MaCSCs) in the MMTV-PyMT breast cancer model.

Given our findings that FAK kinase activity is required for luminal progenitors, but not basal MaSCs, we next tested whether FAK kinase inhibitors differentially affect human breast cancer cells that share gene signatures with luminal progenitors (luminal progenitor-like or basal subtype) and basal MaSCs (MaSC-like or claudin-low subtype), respectively (5, 12). Western blotting revealed that MaSC-like SUM159 cells have higher level of FAK phosphorylation at Y397 than the luminal progenitor-like, SUM149 cells (Fig. 6A; refs. 12, 35). PF573228, a FAK-specific inhibitor (27, 36), decreased FAK kinase activity as measured by its phosphorylation at Y397, but not affected Pyk2 phosphorylation (i.e., no Pyk2 compensation), in both SUM149 and SUM159 cells. PF562271, the dual FAK/Pyk2 inhibitor (27, 36), inhibited both FAK and Pyk2 phosphorylation in SUM149 and SUM159 cells (Fig. 6B). Treatment with either inhibitor did not significantly affect the survival of either cell line (Supplementary Fig. S4B), but suppressed the proliferation of SUM149, but not SUM159, cells (Fig. 6C and D). These inhibitors also decreased proliferation of HCC1954 (another luminal progenitor-like breast cancer), but not MDA-MB-231 cells (another claudin-low subtype; Fig. 6D). Similar to the differential effects observed on cell proliferation, tumorsphere formation by SUM149 and HCC1954 cells, but not SUM 159 and MDA-MB-231 cells, was significantly decreased in the presence of PF573228 or PF562271 (Fig. 6E and F).

Finally, we examined the effect of FAK knockdown (i.e., blocking both kinase-dependent and -independent functions of FAK) on SUM149 and SUM159 cells. As shown in Fig. 6G, treatment of either SUM149 or SUM159 cells with FAK shRNA/GFP suppressed FAK expression relative to cells treated with control scrambled sequence/GFP. Analysis of the GFP-positive cells revealed that knockdown of FAK in either cell line blocked their proliferation (Fig. 6H and I) and tumorsphere formation (Fig. 6J and K). Together, these results suggest that while FAK kinase inhibitors are more effective in targeting luminal progenitor-like, basal breast cancer cells, new approaches will be required to target the kinase-independent functions of FAK that drive the proliferation programs operative in the MaSC-like, claudin-low breast cancer subtype.

Here, we have established distinct roles for FAK kinase-dependent and -independent functions in the regulation of luminal progenitors and basal MaSCs and in promoting tumorigenesis and breast cancer heterogeneity. As summarized in Fig. 6L, in the normal mammary gland, FAK ablation in MFCKO mice decreases (broken lines) the colony-forming activity of luminal progenitors (red to pink) and the self-renewal potential of basal MaSCs (green to light green). In MFCKD mice, loss of FAK kinase activity impairs the maintenance of luminal progenitors, but does not affect basal MaSCs, accounting for their normal ductal outgrowth. Both luminal progenitors and basal MaSCs might serve as targets for transformation by oncogenes such as PyMT to form MaCSCs (brown) with deregulated self-renewal (more lines). The decreased content of MaCSCs and their compromised tumorigenicity (broken lines; orange) in MFCKO-MT mice suggest either luminal progenitors or basal MaSCs (both depleted in MFCKO mice) could be the cells of origin in PyMT-induced mammary tumors. However, the reduced tumorigenesis, decreased MaCSC formation, and self-renewal in MFCKD-MT mice, where the pool of luminal progenitors but not basal MaSCs was depleted after loss of FAK kinase activity, provide direct support that luminal progenitors (but not basal MaSCs) serve as the tumorigenic cell origin in the MMTV-PyMT mouse model and that FAK kinase activity is required for propagation of luminal progenitor-like MaCSCs. Importantly, the differential role of FAK kinase-dependent and -independent functions in regulating luminal progenitors and basal MaSCs in mouse models are recapitulated in distinct human breast cancer cells (i.e., luminal progenitors -like, basal subtype, and basal MaSC-like, claudin-low subtype of breast cancer, respectively).

Although many factors are likely to regulate MaSC activities, recent work shows that Slug and Sox9 act cooperatively to maintain the MaSC state (31). Interestingly, the expression levels of Slug and Sox9 as well as Snail were reduced in basal cells from MFCKO mice, but were rescued in MFCKD mice, suggesting that kinase-independent functions of FAK may support basal MaSC activity by maintaining the expression of these critical transcription factors. Additional studies will be required to clarify the specific roles of these potential targets and the mechanisms by which KD FAK regulates their expression.

Our studies revealed important differences in the regulation of normal MaSCs/luminal progenitors and MaCSCs by FAK signaling. We showed previously that the PI3K/Akt pathway, which is dependent on FAK kinase activity, was selectively activated in WT, but not FAK-null MaCSCs, and that the D395A mutant defective in PI3K binding failed to restore tumorsphere formation as well as tumorigenicity of FAK-null MaCSCs (27). Here, however, we found that the D395A mutant fully rescued sphere/acini-forming capacity of FAK-null MaECs, suggesting that PI3K pathways downstream of FAK are not required to maintain normal MaSC/luminal progenitors. Although more studies are needed to decipher the underlying mechanisms operative in this system, the differential requirement of PI3K pathway for MaCSCs, but not normal MaSCs/luminal progenitors, may be exploited to design specific strategies to target MaCSCs while sparing their normal counterparts.

The differential requirement for FAK kinase activity in distinct breast cancer subtypes has important implications for the development of therapies that take breast cancer heterogeneity under consideration. Recent studies in mouse models, as well as human patients with Brac1 mutations, provide strong support for the proposal that basal breast cancers originate from luminal progenitors (5, 37–39). Thus, our finding that FAK kinase activity plays a differential role in regulating luminal progenitors, as well as in tumorigenesis and the maintenance of MaCSCs in tumors derived from luminal progenitors, suggests that FAK kinase inhibitors could be an effective therapy for this breast cancer subtype. However, the recently identified claudin-low subtype of breast cancer exhibits mesenchymal characteristics (8) and is closely associated with the gene signature of basal MaSC-enriched cells (5, 12), suggesting that these cancers may be derived from basal MaSCs. Furthermore, following neoadjuvant chemotherapy or hormonal treatment, gene expression profiles of residual cancer cells (of both luminal or basal tumors) become more closely related to those of claudin-low subtype tumors (40). Our findings that kinase-independent functions of FAK support basal MaSCs and that FAK kinase inhibitors do not effectively target MaSC-like, human breast cancer cells, suggest that novel strategies may be required to inhibit the claudin-low breast cancer subtype and relapsed breast cancers that display similar molecular signatures.

M.S. Wicha has a commercial research grant from MedImmune and Dompe Phamaceuticals, has ownership interest in OncoMed Pharmaceuticals, and is a consultant for MedImmune, Verastem, Paganini, and Cerulean. J.-L. Guan had a commercial research grant from Verastem and was a consultant for Guidepoint Global. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Luo, X. Zhao, J.-L. Guan

Development of methodology: M. Luo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Luo, X. Zhao, S. Liu, M.S. Wicha

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Luo, S. Chen, J.-L. Guan

Writing, review, and/or revision of the manuscript: M. Luo, M.S. Wicha, J.-L. Guan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Luo, S. Chen, S. Liu, J.-L. Guan

Study supervision: J.-L. Guan

Other: performed some studies with human breast cancer cells: S. Chen

The authors thank Drs. David Schlaepfer for FAK shRNA lentiviral construct, Alan Kraker and Donnie Owens for PF-573228 and PF-562271, Stephen Ethier for SUM149 and SUM159 cells, UMCCC Flow Cytometry Core and Christine Bian for assistance, and Steve Weiss and our laboratory members for helpful comments.

This research was supported by NIH grants (J.-L. Guan).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.