Integrin-mediated adhesion regulates the development and function of a range of tissues; however, little is known about its role in glandular epithelium. To assess the contribution of β1 integrin, we conditionally deleted its gene in luminal epithelia during different stages of mouse mammary gland development and in cultured primary mammary epithelia. Loss of β1 integrin in vivo resulted in impaired alveologenesis and lactation. Cultured β1 integrin–null cells displayed abnormal focal adhesion function and signal transduction and could not form or maintain polarized acini. In vivo, epithelial cells became detached from the extracellular matrix but remained associated with each other and did not undergo overt apoptosis. β1 integrin–null mammary epithelial cells did not differentiate in response to prolactin stimulation because of defective Stat5 activation. In mice where β1 integrin was deleted after the initiation of differentiation, fewer defects in alveolar morphology occurred, yet major deficiencies were also observed in milk protein and milk fat production and Stat5 activation, indicating a permissive role for β1 integrins in prolactin signaling. This study demonstrates that β1 integrin is critical for the alveolar morphogenesis of a glandular epithelium and for maintenance of its differentiated function. Moreover, it provides genetic evidence for the cooperation between integrin and cytokine signaling pathways.
The interaction between a cell and its microenvironment is critical in determining responses to both external and internal stimuli, regulating functions that include migration, proliferation, apoptosis, and differentiation. Integrins are the major class of receptors for the ECM and link cellular adhesion to the actin cytoskeleton and signal transduction pathways (Hynes, 2002). β1 integrin heterodimers are widely expressed and control various developmental processes, including neurogenesis, chondrogenesis, myoblast fusion, and skin and hair follicle morphogenesis (Brakebusch et al., 2000; Graus-Porta et al., 2001; Aszodi et al., 2003; Schwander et al., 2003). However, the function of β1 integrins in the maintenance of a differentiated tissue such as glandular epithelium is not well understood.
The mammary gland consists of a network of luminal ductal and alveolar epithelium that is surrounded by and interacts with myoepithelial cells and is subtended by an ECM termed the basement membrane (BM). Development of the mammary gland occurs primarily after birth, with the ductal network extending to fill the mammary fat pad. During pregnancy, cells within alveoli proliferate and differentiate, achieving functional capacity with the synthesis of milk proteins at lactation. At involution, the majority of the epithelium undergoes apoptosis and remodeling to result in a tissue closely resembling that of the nulliparous gland (Hennighausen and Robinson, 2001).
Cell–matrix adhesion regulates mammary cell structure and function in culture models. Inhibition of mammary epithelial cell contact with the BM alters morphology, survival, proliferation, and differentiation (Barcellos-Hoff et al., 1989; Streuli et al., 1991; Pullan et al., 1996). Culturing cells on a reconstituted BM matrix or on laminin restores many of these functions (Streuli et al., 1995b). Interaction of mammary cells with these ECMs is mediated by integrins, and luminal epithelium expresses the α2β1, α3β1, and α6β1 heterodimers (Prince et al., 2002). The importance of β1 integrins in mammary cell function has been demonstrated by the use of function-blocking antibodies, which inhibit differentiation (Streuli et al., 1991). However, because antibody-induced integrin clustering can also result in its activation (Akiyama et al., 1994), the role of β1 integrin in mammary cell function remains unclear.
Far less is known about the role of integrins in mammary gland development and cell function in vivo. Mammary glands formed from either α3 or α6 integrin–null epithelia are fully developed and functional (Klinowska et al., 2001). The only in vivo evidence for a role of these integrins in mammary gland development comes from studies using a dominant-negative approach, where the β1 cytoplasmic domain was fused to the extracellular domain of the T cell differentiation antigen CD4 (Lukashev et al., 1994). In these studies, β1 integrin was implicated in the control of mammary cell proliferation, survival, and differentiation (Faraldo et al., 1998, 2002). However, because the transgene can act as either an agonist or an antagonist depending on the level of expression (Lukashev et al., 1994), these data are difficult to interpret.
To understand the role of αβ1 integrins in the regulation of mammary epithelial cell function and to determine whether this class of adhesion receptors is indeed involved in the differentiation response, we examined the consequence of conditionally deleting β1 integrin in mammary epithelia. Two strategies were used. First, we examined the role of β1 integrin in vivo by crossing LoxP-flanked β1 integrin (Itgβ1fx/fx) mice (Graus-Porta et al., 2001) with those expressing Cre under the control of either the β lactoglobulin (Blg) or whey acidic protein (Wap) promoter (Selbert et al., 1998; Wintermantel et al., 2002); here β1 integrin was deleted from mammary epithelia before pregnancy or after the initiation of mammary differentiation. Second, we determined the effects of deleting β1 integrin in culture by infecting primary mammary epithelial cells from Itgβ1fx/fx mice with Cre-recombinase adenovirus (AdCre).
Several recent investigations have used a conditional-null approach to study the role of β1 integrins in the development and function of specific tissues; however, none has yet identified a link with epithelial cell differentiation in vertebrate systems. In the epidermis, conditional ablation of β1 integrin leads to defects in the assembly of BM proteins, resulting in altered proliferation and the consequent defects in hair follicle morphogenesis, as well as impaired cytoskeletal dynamics, leading to reduced migration and wound reepithelialization after injury (Brakebusch et al., 2000; Grose et al., 2002; Raghavan et al., 2003). However, the spatial and temporal differentiation of keratinocytes is unaffected (Raghavan et al., 2000). In mesenchymal tissue such as bone, β1 integrin ablation in chondrocytes causes chondrodysplasia through abnormal adhesion of cells to collagen and fibronectin and decreased proliferation via defects at the G1/S boundary of the cell cycle (Aszodi et al., 2003), whereas in muscle, β1 integrin deletion leads to defective myoblast fusion and sarcomere assembly (Schwander et al., 2003). In neuronal tissues, integrin ablation also has a dramatic phenotype, e.g., in the neural crest–derived peripheral nervous system, where β1 integrins are required for accurate timing in Schwann cell migration and myelination, and for maintaining processes around axons, as well as in neuromuscular junction maturation (Feltri et al., 2002; Pietri et al., 2004).
Our work extends these previous studies to demonstrate a key role for β1 integrin in epithelial differentiation and expression of tissue-specific genes. Moreover, it identifies a mechanism for this role, which is through the integration of signaling networks between the integrin and cytokine pathways. Thus, we provide genetic evidence for an integrin-dependent effect on gene expression–related differentiation processes in a mammalian system in vivo. Our data support previous studies on integrin dependence of gene expression in other models (Martin-Bermudo and Brown, 1999) and strengthen the notion that in certain cell types, signaling pathways driven by soluble factors are dependent on their integrin-mediated positional identity.
Conditional deletion of β1 integrin in mammary epithelia in vivo
β1 integrin knockout mice have an embryonic lethal phenotype, thus preventing assessment of mammary gland development (Fassler and Meyer, 1995; Stephens et al., 1995). To determine the specific role of β1 integrin during the proliferative and differentiation stages of mammopoiesis, we crossed Itgβ1fx/fx mice with transgenic lines expressing Cre under the control of the Blg and Wap promoters (Fig. 1 A). The specificity of Cre-mediated recombination in Blg-Cre and WapiCre mice was determined by crossing these mice with Rosa26R mice, which express LacZ after Cre recombination (Soriano, 1999). Blg-Cre–mediated recombination was detected in the ductal epithelium of nulliparous mice from the age of 12 wk (i.e., after pubertal development), with the majority of epithelial cells positive for X-gal staining (Fig. 1 B). Cre recombination had occurred by all stages of pregnancy and lactation in Blg-CreTg/+;Rosa26Rfx/fx mice with a mean of 80% (range 50–100%) of epithelial cells positive for X-gal staining (Fig. 1 D). In contrast, WapiCre mice demonstrated a temporally restricted Cre-activity pattern, with X-gal staining detected in ∼60% of epithelia from midpregnancy and throughout lactation but not in nulliparous mice up to the age of 18 wk or until the 12th d of pregnancy, when alveoli have already begun to form (Fig. 1, C and E). Expression of both Blg-Cre and WapiCre recombination was restricted to luminal mammary epithelial cells.
These data define Blg-Cre mice as a good model for examining gene function throughout pregnancy and lactation, as the targeted gene will be deleted before the initiation of pregnancy. Because the WapiCre promoter is only expressed after differentiation has been initiated in midpregnancy, we also crossed these mice with Itgβ1fx/fx animals to examine the effects of β1 integrin deletion in differentiated luminal epithelial cells. Using these two models, we assessed the role of β1 integrin in mammary epithelial cell function during the proliferative and differentiation phases of mammary gland development.
Analysis of β1 integrin deletion in Itgβ1fx/fx;Blg-CreTg/+ mice (Fig. 1 and see Fig. 3) or Itgβ1fx/fx;WapiCreTg/+ mice (see Fig. 7) was performed at the DNA and protein levels. PCR analysis of mammary glands at day 2 of lactation identified a recombined 1.3-kB product in mutant mice along with the unrecombined 2.1-kB allele, which is present in the stromal and myoepithelial cells that do not undergo Cre-mediated recombination. Wild-type mice only contained the unrecombined 2.1-kB allele (Fig. 1 F). Examination of β1 integrin protein by immunofluorescence demonstrated lateral and basal expression in the luminal epithelium and basal expression in myoepithelial cells, which is consistent with previous findings (see Fig. 3 A and Fig. 7, K and M; Prince et al., 2002). In mutant mice, β1 integrin was lost from the lateral and basal surfaces of the luminal epithelium (see Fig. 3 B and Fig. 7, L and N, white arrows) but could still be detected in myoepithelial cells, which subtend the luminal epithelium. These data demonstrate temporal deletion of β1 integrin from luminal epithelial cells during mammary gland development.
Ablation of β1 integrin expression results in impaired lobuloalveolar development and lactogenesis
Itgβ1fx/fx;Blg-CreTg/+ females displayed a lactation phenotype that varied from a total failure of lactation to being able to support litters, although pups from these dams displayed dramatically reduced growth rates and were much smaller than age-matched controls (Fig. 2, A and B). A lack of milk was evident in the stomachs of pups from the most severely affected females, with these pups dying from dehydration within 72 h despite repeated suckling (unpublished data). Examination of two key components of milk, β casein and milk fat, confirmed that lactogenesis was compromised in the absence of β1 integrin in vivo (Fig. 2, C–E).
Normal ductal elongation and side branching occurred in Itgβ1fx/fx;Blg-CreTg/+ mice (not depicted), but there was a significant decrease in lobuloalveolar development during pregnancy (not depicted) and at the second day of lactation (Fig. 2, F–K). 5 of 10 females were unable to support pups, and these mice displayed the most severe mammary defects, with a dramatic inhibition in the formation of lobuloalveoli. For example, lactation failed in the mammary gland of the mouse shown in Fig. 2 J and was partial in the gland shown in Fig. 2 K. The defect in alveolar development was even more apparent in histological sections at the second day of lactation, when Itgβ1fx/fx;Blg-CreTg/+ mice displayed decreased alveolar density (Fig. 2, L–Q).
Closer examination also revealed defects in the integrity of epithelial structures. Whereas the alveoli of wild-type females contain a simple layer of luminal epithelia surrounding a hollow lumen (Fig. 2 O), the alveoli of Itgβ1fx/fx;Blg-CreTg/+ mice were disorganized and contained clumps of epithelial cells bulging into what would normally be luminal space (Fig. 2, P and Q). These protrusions often spanned the width of the alveoli and consisted of multiple layers of cells (Fig. 2 Q). Interestingly, the protuberant cells remained attached to other cells and had not completely “fallen” into the lumens.
Immunofluorescent staining of sections from glands of Itgβ1fx/fx;Blg-CreTg/+ mice at day 2 of lactation demonstrated that, despite deletion of β1 integrin (Fig. 3, A and B), epithelial polarity of mammary alveoli was essentially normal. The tight junction marker ZO-1 was located on the apical surface of luminal epithelia (Fig. 3, A and B), whereas laminin was present at the basal surface (Fig. 3, C and D). β catenin was predominately on the lateral surface at the site of cell–cell contacts, although some basal and some apical staining was also evident (Fig. 3, E–G). As indicated by the histological sections (Fig. 2, P and Q), the majority of alveoli had epithelial cells protruding into the lumen and retained cell–cell contacts as determined by β catenin staining (Fig. 3, D and F, arrows); even in these morphologically disrupted alveoli, the basal markers laminin and β4 integrin were located normally (Fig. 3 D and not depicted). Interestingly, these protruding cells were not apoptotic as demonstrated by normal nuclear appearance (Fig. 3 F) and caspase 3 staining (not depicted), retained cell–cell adhesions as demonstrated by β catenin, and located ZO-1 to the apical surface adjacent to the lumens (Fig. 3 B and not depicted). However, BM cell adhesion had clearly been disrupted in the cells given their failure of attachment to the laminin substratum; this was more severe in some cases (Fig. 3, D and F) than in others (Fig. 3, B and G), although in each case some of the most apically located luminal cells were displaced from the BM. These data demonstrate that β1 integrin is essential for normal lobuloalveolar development and lactation. It is also required for proper alveolar epithelial morphogenesis.
Deletion of β1 integrin results in disruption of focal adhesions and multicellular morphogenesis
To confirm that β1 integrin has a key role in alveolar morphogenesis, we assessed focal adhesion function and multicellular morphogenesis of mammary epithelial cells in the absence of β1 integrin in culture using our established culture model. Cells were harvested from Itgβ1fx/fx mice and infected with either AdCre or β-galactosidase adenovirus (Adβgal) before being cultured on a stromal ECM (type I collagen) or a BM matrix. 24 h after infection with either AdCre or Adβgal, no differences in morphology were observed between cells plated on collagen, which is consistent with the time required for adenoviral-expressed proteins to be translated (Fig. 4, A and B): at this time, normal levels of β1 integrin were present (not depicted). Between 36 and 48 h after infection, defects in cell morphology became apparent, with AdCre-infected cells displaying a rounded phenotype compared with Adβgal-infected controls (Fig. 4, C and D).
β1 integrin is normally localized to focal adhesions, which are protein complexes that connect the ECM with the actin cytoskeleton (Fig. 4 M). However, in AdCre-infected cells, these focal adhesions largely disappeared, corresponding to the loss of β1 integrin, although some small adhesion complexes containing integrin-linked kinase (ILK) and talin remained (Fig. 4 N). Despite the considerable reduction of focal adhesions, AdCre-infected cells continued to express markers for other forms of adherent junctions, such as tight junctions (e.g., ZO-1; PFig. 4, O and P), indicating that deletion of β1 integrin does not simply result in a global disruption of all cellular adhesions.
When AdCre-infected cells that showed no apparent phenotype 24 h after infection were treated with trypsin and replated back onto collagen (Fig. 4 F), they displayed focal adhesion defects similar to infected cells that were maintained continuously on collagen for 48 h; although these cells adhered to the collagen substratum, they were unable to spread on it. The infected cells were also unable to spread on collagen IV (Fig. 5 C). Together, these data demonstrate that without β1 integrin, primary mammary epithelial cells cultured in two-dimensions on collagen fail to either initiate or maintain focal adhesion complexes, resulting in defective form and structure of the cell. β1 integrin–null cells did not lose their intrinsic capacity to spread on ECM proteins because when they were plated onto vitronectin, they spread well and assembled peripheral adhesion complexes containing αv (Fig. 4, Q and R) and β3 integrin subunits (not depicted).
Mammary epithelial cells cultured on a BM matrix form three-dimensional acini, which are polarized spheres with hollow lumina that resemble alveoli formed during the later stages of pregnancy (Aggeler et al., 1991). The requirement of β1 integrin for the maintenance and initiation of this process was tested by plating cells either directly onto the BM matrix after virus infection or onto collagen for 24 h and then replating them onto the BM matrix. AdCre- and Adβgal-infected cells began to form acini within 18–24 h after infection, and no differences were observed between the two populations at this time (Fig. 4, G and H). From 24 to 36 h after infection, the acini formed from β1 integrin–null cells began to disrupt and individual cells could be seen dissociating from the polarized structure (Fig. 4 J). Cells lacking β1 integrin before culture on BM displayed a more severe phenotype; instead these cells remained as either individuals or groups of loosely clustered cells for 48 h (Fig. 4 L).
Acini formed from Adβgal-infected cells localized the tight junction protein ZO-1 to the apical surface, whereas focal adhesion proteins such as β1 integrin or ILK were predominately on the basal and, to a lesser extent, lateral surfaces (Fig. 4 S and see Fig. 8 B). Adβgal and AdCre infection was confirmed by antibody staining (PFig. 4, N and P) and by Western blot (Fig. 5). In experiments where 100% of cells were infected with AdCre, β1 integrin–null cells failed to polarize properly: ZO-1 and ILK were mislocalized from the apical and basal surfaces and were instead present on all cell surfaces (Fig. 4 T). The total disruption of acini formation was reduced by decreasing the percentage of AdCre-infected cells to 50–75% (Fig. 4 V). In these cultures, acini were formed, and although ZO-1 appeared to be slightly reduced in intensity, it was correctly located on the apical surface (Fig. 4 V). Basal markers of polarity such as ILK were also correctly located (unpublished data).
Thus, integrin deletion both in vivo and in the culture model results in acini containing incorrectly localized cells and disrupted morphogenesis. β1 integrins are therefore required for the multicellular glandular morphogenesis that normally occurs in association with BM.
Disruption of signaling pathways in response to β1 integrin deletion
To address the perturbed focal adhesion formation and epithelial morphogenesis further, we investigated the requirement for β1 integrin in the activation of selected downstream signaling mediators, including the protein tyrosine kinase FAK and the adaptor protein paxillin. Infection of primary epithelial cells with AdCre virus resulted in knockdown of β1 integrin by 48 h after infection in cells plated on collagen and the BM matrix (Fig. 5). Loss of β1 integrin in cells cultured continuously on collagen in growth medium (similar to those cells shown in Fig. 4, C and D) resulted in decreased levels of tyrosine phosphorylation at residue 397 of FAK, which is critical for its full activation (Fig. 5 A; Schlaepfer and Hunter, 1998). Similarly, tyrosine phosphorylation of paxillin was reduced. Erk2 and calnexin were used as protein loading controls (Fig. 5, A and B; and not depicted). Thus, phosphorylation of FAK and paxillin were dramatically reduced in the absence of β1 integrin. The primary integrin receptors for collagen are α1β1 and α2β1, whereas α6β1 and α6β4 are the main laminin receptors (Taddei et al., 2003). Although the β1-null cells were unable to spread well on collagen I or IV, the small number of adhesion complexes that did form retained the ability to initiate proximal signaling because they contained phosphorylated paxillin (Fig. 5 C, top). Moreover, the β1-null cells spread on vitronectin and had large peripheral adhesion complexes showing robust phosphopaxillin staining (Fig. 5 C, bottom), although interestingly the smaller, more centrally located phosphopaxillin adhesions are not visible in the absence of β1 integrin (yellow arrows).
We next investigated the phosphorylation status of FAK and paxillin in cells cultured on laminin-rich BM matrix. The total levels of both these proteins were slightly lower, whereas their tyrosine phosphorylation was reduced dramatically, even in the presence of lactogenic hormones (Fig. 5 B). Thus, deletion of β1 integrin diminishes the activation of some of the immediate proximal downstream signaling mediators required for normal cell morphology, although the cells do retain a capacity for proximal integrin signaling when they adhere to the ECM through alternative β integrin subunits. A reduction of FAK and paxillin tyrosine phosphorylation also occurs after integrin deletion in vivo (unpublished data).
The impairment of adhesion signaling in β1 integrin–null epithelia raised the possibility of a general inability of the cells to respond to external stimuli. To test whether MAPK signaling can be activated after reducing β1 integrin–null levels, cells were stimulated with EGF and the phosphorylation of Erk1/2 was measured (Gilmore et al., 2002). This pathway was activated in cells plated on the BM matrix and stimulated with lactogenic hormones, regardless of the presence or absence of β1 integrin (Fig. 5 B). However, mammary cells grown on the BM matrix do not undergo significant proliferation (unpublished data), so subsequent experiments were performed on collagen. To test rapid activation of the MAPK pathway, cells were serum starved and treated with EGF for 15 min. Although AdCre infection resulted in the knockdown of β1 integrin (Fig. 6 A), both these and control cells responded to EGF by increasing Erk1/2 phosphorylation compared with nonstimulated cultures. Furthermore, similar levels of pErk were observed after continuous exposure to serum and EGF (Fig. 6 B). Thus, β1 integrin is not required for the phosphorylation of Erk in response to either pulse or steady-state exposure to EGF.
Because other β integrins may substitute for the β1 subunit, and to rule out the possibility that the residual β1 integrin in Fig. 6 (A and B) contributes to EGF signaling, primary wild-type cultures were trypsinized and kept in suspension (where integrins are unligated) or cultured on collagen before a 15-min stimulation with EGF (Fig. 6 C). EGF induced Erk1/2 phosphorylation in both attached and suspended cells, demonstrating activation of the MAPK-signaling cascade in the complete absence of integrin ligation. Thus, unlike some other cell systems (Bill et al., 2004), EGF-dependent activation of this pathway in primary mammary epithelial cells is not dependent on β1 integrin or integrin ligation.
Deletion of β1 integrin from differentiated epithelium results in impaired lactogenesis
The use of Itgβ1fx/fx;Blg-CreTg/+ mice identified a critical requirement for β1 integrin in lactogenesis. However, because of this impaired development we could not determine if β1 integrin is required directly for mammary differentiation or if the defective lactation occurs as a result of abnormal lobuloalveolar development. To address this, we used Itgβ1fx/fx;WapiCreTg/+ mice to examine the role of β1 integrin in vivo after luminal epithelial cells had become differentiated.
Itgβ1fx/fx;WapiCreTg/+ mice were able to support their litters during lactation. However, close examination of pup growth rates revealed a significant decrease in pup weight from Itgβ1fx/fx;WapiCreTg/+ dams (Fig. 7 A), demonstrating a defect in the functional capacity of the mammary glands. Whole mount analysis demonstrated no obvious differences in lobuloalveolar development (Fig. 7, C–F), although histological sections indicated that alveoli in Itgβ1fx/fx;WapiCreTg/+ mice were smaller than wild-type controls (Fig. 7, G–J). Consistent with a role for β1 integrin in the regulation of alveolar structure, increased numbers of epithelial cells were present in the lumen of alveoli from Itgβ1fx/fx;WapiCreTg/+ mice, although this phenotype was less dramatic than that observed in Itgβ1fx/fx;Blg-CreTg/+ mice. The alveoli retained a more open morphology at lactation day 2 (Fig. 7 J) than in the Itgβ1fx/fx;Blg-CreTg/+ mice (Fig. 2, P and Q). The defect in the functional capacity of alveoli was not rescued by repeated pregnancies, with pups from Itgβ1fx/fx;WapiCreTg/+ mice after the second pregnancy displaying a dramatic reduction in growth rates compared with wild-type controls (Fig. 7 B).
Immunofluorescent staining revealed that luminal cells within alveoli of the Itgβ1fx/fx;WapiCreTg/+ mice were correctly polarized because they positioned ZO-1 in an apical location (Fig. 7, K–N), β catenin was located at cell–cell junctions (Fig. 8 D, bottom right), and laminin was deposited at the basal cell surface (not depicted). Despite this, the cells were unable to synthesize milk efficiently. They made low levels of milk fat, as determined by Oil Red O staining (PFig. 7, O and P), and milk proteins, as disclosed by immunostaining with an antibody to β casein (Fig. 7, Q–T). Thus, the WapiCre alveoli develop normally and the epithelial cells retain their polarity, but they fail to differentiate efficiently in vivo.
Together, these data suggest that deletion of β1 integrin after the initiation of differentiation in vivo results in an inability to maintain a functional phenotype in late pregnancy and lactation. This indicates that β1 integrin may be needed directly for the signaling pathways that maintain the differentiation response. We therefore tested this possibility using a culture model.
β1 integrin is necessary for prolactin (Prl)-induced mammary epithelial cell differentiation
A direct role for β1 integrin in Prl signaling and milk protein gene expression has not been explored using a null mutation, though it has been suggested (Streuli et al., 1995b; Edwards et al., 1998; Faraldo et al., 1998, 2002). To directly assess whether β1 integrin is required for epithelial differentiation in response to a Prl signal, we infected primary cultures from late-pregnant Itgβ1fx/fx mice with AdCre or Adβgal virus and plated them on the BM matrix (Fig. 8, A and B). After 24 h under growth conditions, cultures were exposed to differentiation medium with or without Prl for another 24 h, after which the production of β casein, a differentiation endpoint marker, was assessed. Because primary mammary epithelial cells undergo little proliferation on the BM matrix in the presence of differentiation medium, this assay allowed the role of β1 integrin in differentiation to be evaluated in the absence of proliferation (which does occur in the in vivo context).
β casein was not synthesized in the absence of Prl, whereas the addition of Prl resulted in the synthesis of β casein in both noninfected and Adβgal-infected cells (Fig. 8 A). In contrast, AdCre-infected cells, in which β1 integrin levels are significantly reduced (Fig. 8 A and Fig. 5 B), failed to undergo differentiation. The lack of β casein synthesis and secretion into the lumen of acini was confirmed by immunofluorescent staining (Fig. 8 B). Acini that maintained a polarized structure because of a heterogenous population of AdCre-infected and noninfected cells produced β casein in ∼25% of acini compared with 75% of βgal-infected controls. AdCre infection of 100% of cells resulted in a complete failure of differentiation. These data demonstrate that β1 integrin is essential for Prl-mediated differentiation in mammary epithelia.
β1 integrin–null cells fail to activate Stat5
The failure of Prl-induced differentiation in β1 integrin–null cells suggested that this ECM receptor is required for the transduction of the Prl signaling pathway. Whole mammary gland lysates from Itgβ1fx/fx;WapiCreTg/+ and Itgβ1fx/fx;WapiCre+/+ mice at days 3 and 14 of lactation were probed with phosphospecific antibodies to Stat5 (Fig. 8 C). A reduction in the levels of activated Stat5 was observed during both early and late lactation in mutant mice compared with wild-type controls; moreover, impaired nuclear translocation of Stat5 occurred at day 18 of pregnancy (i.e., just before lactation) in both Itgβ1fx/fx;Blg-CreTg/+ and Itgβ1fx/fx;WapiCreTg/+ mice (Fig. 8 D). These results confirm that loss of β1 integrin prevents Stat5 signaling and that a similar phenotype results in glands of mice generated with both the Blg and Wap promoters. β1 integrin is therefore required for the activation of a transcription factor essential for the induction of milk gene synthesis in vivo.
To assess whether β1 integrin acutely mediates Stat5 activation, we performed a nuclear translocation assay on primary mammary epithelial cells infected with AdCre or Adβgal (Fig. 8, E and F). After activation, Stat5 translocates into the nucleus, where it initiates transcription of target genes such as β casein and Wap (Gouilleux et al., 1994; Liu et al., 1996). In the absence of Prl, only 2.9 ± 0.96% of the noninfected cells displayed Stat5 nuclear translocation; this increased to 49.2 ± 0.15% after 15 min of stimulation with Prl and was not affected by infection with Adβgal virus (51.6 ± 2.36% of cells with nuclear localization). However, the infection of cells with AdCre virus, which ablates β1 integrin, resulted in a failure of Prl to initiate Stat5 activation, with only 6.8 ± 2.64% of cells demonstrating Stat5 nuclear localization. The use of suboptimal levels of AdCre virus during infection provided a further internal control, resulting in a proportion of the cells that were uninfected as assessed by Cre antibody staining. These Cre-negative cells behaved in a manner similar to noninfected or Adβgal-infected controls, displaying Stat5 activation in 42.2 ± 8.15% of cells. Thus, our data show that β1 integrin is critical for the Prl-induced activation of Stat5, a transcription factor essential for mammary epithelial differentiation, and they provide genetic evidence for cooperation between integrin and cytokine signaling pathways.
In this paper, we report on two phenotypes that result from ablation of the β1 integrin gene from luminal cells of mammary epithelium. The first is that mammary gland development during pregnancy is perturbed, resulting in malformed alveoli. The second is that lactational differentiation is compromised, leading to the inability of dams to nurse their pups efficiently.
β1 integrin is essential for alveolar development
Conditional deletion of β1 integrin resulted in normal elongation and tertiary side branching of mammary ducts after pregnancy. This is in concordance with a recent study where normal ductal outgrowth during puberty was observed in Itgβ1fx/fx;MMTV-CreTg/+ mice (White et al., 2004). In our work, the lack of β1 integrin compromised subsequent lobuloalveolar development. Sparse alveoli were observed in histological sections of mammary glands from Itgβ1fx/fx;Blg-CreTg/+ mice, and alveoli failed to expand into fully formed lobuloalveoli. Variation in phenotype severity (and indeed the number of epithelial cells expressing Cre, ranging from 50 to 100% in the Blg-Cre line) was observed between mice, most likely because of the mosaic expression of Cre, as has been observed in other lines (White et al., 2004). Consistent with this, both β1 integrin–null and normal alveoli were observed in the same glands, with only those lacking β1 integrin displaying a phenotype.
The retarded lobuloalveolar development in Itgβ1fx/fx;Blg-CreTg/+ mice is suggestive of a defect in epithelial proliferation. In a separate study, we found that the loss of β1 integrin leads to decreased epithelial cell proliferation both in culture and in Itgβ1fx/fx;WapiCreTg/+ mice (at the beginning of lactation, when a proliferative burst normally occurs) and that this is attributable to increased levels of the cell cycle inhibitor p21 and the consequent effects on cell cycle machinery (Li et al., 2005). This phenotype may reflect a general requirement for β1 integrin in vivo because related proliferation defects have been noted in other β1 integrin–null cell types such as chondrocytes (Aszodi et al., 2003). Reduced proliferation has also been noted in the mammary glands of mice expressing a dominant-negative chimeric form of β1 integrin (Faraldo et al., 1998).
A novel phenotype that we observed in vivo is disrupted alveolar morphogenesis. In some cases, protuberant networks of cells formed within alveoli; these cells are wholly detached from the subtending laminin-rich BM, but retain cell–cell interactions as judged by β catenin staining, and the tight junction protein ZO-1 was localized apically, adjacent to the residual lumens. Remarkably, these cells do not undergo overt apoptosis. In contrast to ablation of β1 integrin in skin, deletion in the mammary gland does not appear to cause defects in BM formation (Raghavan et al., 2000); this may be because β1 integrin was targeted in luminal epithelium rather than myoepithelial cells, which are the main source of BM proteins (Gudjonsson et al., 2002).
Although we are currently trying to understand the mechanism for the protuberant phenotype, the displacement of luminal cells from the BM reflects a key role for β1 integrin heterodimers in mediating cell–ECM adhesion in luminal mammary epithelia. This is supported by the observations that loss of β1 integrin in the culture model resulted in defective focal adhesion function and cell rounding on two-dimensional substrata, as occurs after perturbation of β1 integrin in other cell types (Lukashev et al., 1994), and that polarized acini could not form or be maintained on a three-dimensional laminin-rich matrix. Both β3 and -4 integrins are expressed in many mammary carcinoma lines and become up-regulated in primary cultures after an extended period in culture, but our data suggest that they may not substitute in vivo. β4 integrin is normally restricted to myoepithelial (as opposed to luminal) cells in vivo (Runswick et al., 2001), and we have not been able to detect the β3 subunit in vivo (unpublished data). Furthermore, transplanted α6-null mammary epithelium lacking the α6β4 heterodimer develops normally in vivo (Klinowska et al., 2001).
β1 integrin controls mammary epithelial cell differentiation
Although it is possible that altered proliferation and morphogenesis of mammary alveoli lead to the failed lactogenesis of β1 integrin–null mammary glands, two lines of evidence indicate that this effect is more direct. First, we have performed experiments similar to those with the Blg promoter, using instead the Wap promoter, which is expressed at a later time in development. Our rationale for this is that the Blg promoter is expressed in prepregnant ductal epithelium; thus, integrin ablation before pregnancy might inhibit the formation of proper alveoli. However, by deleting β1 integrin after the advent of alveolar formation via the Wap promoter (which is not expressed until midpregnancy), we demonstrate that lactation is still disrupted in cells that are already committed to an alveolar fate and have begun to differentiate. Moreover, our data using the Itgβ1fx/fx;WapiCreTg/+ mice indicate that milk production is compromised in alveoli where the epithelial cells retain the ability to polarize (as judged by apical distribution of ZO-1, lateral distribution of β catenin, and basal localization of laminin) and maintain a more open morphology than alveoli from the Itgβ1fx/fx;Blg-CreTg/+ mice. Second, we have extracted fully differentiated cells from the mammary glands of late-pregnant Itgβ1fx/fx mice and infected them with AdCre virus. Under these conditions, and using well-established assays for mammary differentiation, the loss of β1 integrin prevents Prl-induced milk protein expression.
Together, these experiments demonstrate a central role for β1 integrin in the maintenance of a differentiation signal in mammary epithelium: indeed, they provide key evidence using gene deletion that integrins are required to maintain tissue-specific function in glandular epithelium. β1 integrin has also been implicated in the control of epidermal differentiation; however, unlike the mammary gland, where β1 integrin signaling is permissive for differentiation, in the epidermis the β1 integrin signaling is inhibitory (Watt, 2002).
It is well known that members of the Prl signaling pathway have a fundamental role in both alveolar development and mammary differentiation (Hennighausen and Robinson, 2001). The defect in alveolar development and differentiation in β1 integrin–null mammary glands suggests a relationship between integrin and Prl signaling. We have demonstrated that this is indeed the case because Stat5 activation is compromised in vivo during lactation and it does not translocate to the nucleus in response to Prl in the culture model. In previous work, we have argued that matrix-specific integrins are essential to control the Prl signaling cassette in mammary cells because it is not activated during culture on stromal ECM, and we have suggested that this is attributable to a dominant-negative inhibitory mechanism that prevents differentiation on inappropriate ECM (Streuli et al., 1995a; Edwards et al., 1998). Our study now provides genetic evidence that integrins have at least a permissive role in Prl receptor signaling, and we are now trying to identify a mechanism to explain the requirement of integrins. Importantly, however, the lack of integrin does not compromise all ligand-activated signaling pathways. Steady-state Erk1/2 phosphorylation in β1 integrin–null cells cultured in growth medium is no different than that in integrin-containing cells, and acute stimulation with EGF in both integrin-null cells and wild-type cells deprived of all integrin signaling (by detachment from the ECM) leads to similar levels of Erk1/2 phosphorylation.
In summary, we have shown that β1 integrin is critical for the alveolar morphogenesis of a glandular epithelium and for the maintenance of its differentiated function. Our data highlight a central role for integrins in cytokine signaling, thus demonstrating that some signaling pathways that were once believed to result from simple ligand-dependent activation actually depend on more complex intracellular networks that take into account the integrin-mediated positional identity of cells.
Materials And Methods
Mouse strains and breeding
The Itgβ1fx/fx, Blg-CreTg/+, WapiCreTg/+, and Rosa26R mice have been described previously (Selbert et al., 1998; Soriano, 1999; Graus-Porta et al., 2001; Wintermantel et al., 2002). The WapiCre mouse contains an improved version of the Cre recombinase under the control of the Wap promoter that minimizes gene silencing and mosaicism in cells where Wap is normally expressed. ICR mice were purchased from Harlan Olac Ltd. Cre-mediated deletion was achieved by crossing fx/fx mice with Cre Tg/+ mice to produce fx/+;CreTg/+ mice. fx/+;CreTg/+ mice were then crossed with fx/fx mice to produce fx/fx;Cre Tg/+ (mutants) or fx/fx;Cre +/+ (wild type) used for analysis. Genotyping was preformed as previously described (Selbert et al., 1998; Soriano, 1999; Graus-Porta et al., 2001; Wintermantel et al., 2002) from DNA prepared from ear clips.
All mice were mated at ∼12 wk of age. Growth-rate analysis of pups was performed by standardizing litter sizes to eight pups/litter and weighing pups at least every 2 d throughout lactation. Pups that died because of dehydration were not replaced from Itgβ1fx/fx;Blg-CreTg/+ mice that displayed a lactation phenotype, and average pup weight was calculated from the maximum number of pups supported from these mice. Note that the pups nursing from mice with integrin-null mammary glands are heterozygous for the integrin flox allele, and because Cre is only expressed in mature female mammary tissue, the pups are all phenotypically normal.
Mice were housed and maintained according to the University of Manchester and UK Home Office guidelines for animal research.
Morphological and histological analysis
Whole mount analysis was performed by spreading inguinal mammary glands on polysine slides, fixing in 10% neutral buffered formalin overnight, and defatting in acetone before staining with carmine alum (0.2% carmine and 0.5% aluminium sulfate) overnight. The whole mount was dehydrated using a graded ethanol series followed by immersion in Slide Bright for 1 h and stored in methyl salicylate before photography using a dissecting microscope (Leica; Naylor and Ormandy, 2002). Quantification of ductal side branching was performed as previously described (Naylor et al., 2003). Whole mounts were subsequently paraffin embedded before sectioning (5 μm) and subjected to standard hematoxylin and eosin staining. Histology was detected with a microscope (Axioscop2; Carl Zeiss MicroImaging, Inc.) using PlanNeofluar 10× (NA 0.3) and 40× (NA 0.75) lenses at RT. Images were captured with a color camera (AxioCam; Carl Zeiss MicroImaging, Inc.) and analyzed with Openlab 4 software (Improvision).
X-gal staining was performed by fixing inguinal mammary glands with 4% PFA for 1–2 h at 4°C and washing three times in 100 mM sodium phosphate, pH 7.3, 2 mM MgCl2, 0.01% sodium deoxycholate, and 0.02% NP-40 for 30 min before adding 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, and 1 mg/ml X-gal and incubating at RT overnight. Tissue was dehydrated with an ethanol series and cleared in Slidebright before being photographed.
Oil Red O staining was performed by staining 6-μm mammary gland cyrosections in freshly diluted Oil Red O solution (six parts 0.5% Oil Red O stock solution [Sigma-Aldrich] and four parts H2O) for 15 min. Sections were rinsed twice with 60% isopropanol and once with water and then counterstained with hematoxylin for 1 min before being photographed.
Primary cell assays
Extraction and adenoviral infection of primary mammary epithelial cells from 15.5- to 17.5-d pregnant mice was performed as described previously (Watkin and Streuli, 2002). Cells were infected with either Cre recombinase (AdCreM1; Microbix Biosystems, Inc.) or Adβgal (a gift from A. Ridley, University College London, London, UK) in suspension for 45 min before being plated at equal density onto either collagen I or the BM matrix (Matrigel; Becton Dickinson; Pullan et al., 1996). Cells were cultured in growth medium containing 5 μg/ml insulin, 1 μg/ml hydrocortisone, 3 ng/ml EGF, 10% FCS, 50 U/ml Pen/Strep, 0.25 μg/ml fungizone, and 50 μg/ml gentamycin in F12 medium or differentiation medium containing the same concentrations of insulin, hydrocortisone, fungizone, Pen/Strep, and gentamycin in DME/F12 medium (GIBCO BRL) with or without 3 μg/ml Prl (ovine Prl) for the induction of differentiation as indicated (reagents were purchased from Sigma-Aldrich unless otherwise indicated). Replating experiments were performed after 24 h growth on collagen. Growth medium was removed, and cells were incubated for 10 min in HE buffer (9.5 g/L HBSS-Ca2+ free, 0.35 g/L NaCO3, and 2g/L EDTA; Sigma-Aldrich) before trypsinizing (GIBCO BRL) for 2–3 min, spinning at 12,000 rpm for 4 min, and plating onto either collagen or the BM matrix as described. 24 h after infection, zVAD was added to all cultures to minimize apoptosis.
Stat5 nuclear translocation assay was performed by plating adenovirus-infected cells on collagen-coated coverslips (collagen coated for 2 h at RT, followed by three washes of PBS) in growth medium. 12 h after infection, growth medium was replaced with differentiation medium containing BM matrix (1:50) without Prl (Streuli et al., 1995b). 48 h after infection, Prl was added, and after a 15-min stimulation, cells were fixed with 4% PFA before immunostaining with Stat5a (Santa Cruz Biotechnology, Inc.) and Cre (Chemicon) antibodies.
Protein was extracted from cell lysates using 1× net buffer (50 mM Tris-Cl, 150 mM NaCl, 1% NP-40, and 2 mM EDTA, pH 7.6) for cells on collagen or 2× net buffer for cells on BM matrix with 1 mM sodium orthavanadate, 10 mM sodium fluoride, and 1:100 dilution of Protease cocktail inhibitor set 1 (Calbiochem). Lysates from mammary glands were prepared by homogenizing frozen mammary tissue in 10 mM Hepes, pH 7.5, 100 mM KCl, 5% glycerol, 1% Triton X-100, 0.1% SDS, 10 mM EDTA, 1 mM DTT, 0.5 mM PMSF, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 mM Na3VO4, and 10 mM sodium molbdate. Lysates from mammary glands used for casein analysis were extracted in 2× net nuffer. Immunoblotting was performed as described previously (Naylor et al., 2003). For cell cultures, equal amounts of protein were used and equivalent loading was assessed by referral to controls such as Erk2 and calnexin. For mammary gland samples, equivalent amounts of protein were determined by quantification using a protein assay (Bio-Rad Laboratories). Specific binding was detected using HRP-conjugated secondary antibodies (GE Healthcare) with Chemiluminescence reagent (PerkinElmer) and Biomax light film (Eastman Kodak Co.). Primary antibodies were as follows: β1 Integrin, paxillin (BD Biosciences), βgal (Promega), FAK (a gift from A. Ziemiecki, University of Bern, Bern, Switzerland), calnexin (StressGen Biotechnologies), cytokeratin 8/18 (Progen Biotechnik), β casein (Streuli et al., 1995b), phospho-FAK Y397, phospho-Jak2 Y1007/1068, phosphopaxillin Y31, Jak2 (Biosource International), E-cadherin (Qbiogene), phospho-Stat5 (Upstate Biotechnology), phospho-Erk1/2, phospho-PKB S473, PKB (Cell Signaling Technology), β actin (Sigma-Aldrich), and Erk2 (Santa Cruz Biotechnology, Inc.).
Cells plated on either collagen- or BM matrix–coated coverslips were stained as described previously (Streuli et al., 1991) and photographed after mounting with Prolong anti-fade (Invitrogen). Immunostaining of mammary glands was performed on tissue that had been frozen in liquid nitrogen and stored at −80°C. 6-μm sections were then fixed and immunostained as described previously (Klinowska et al., 2001). Immunofluorescence was detected with either a confocal microscope (for tissue sections; SP2 AOBS; Leica) or a microscope (for cultured cells; Axioplan2; Carl Zeiss MicroImaging, Inc.) using PlanApochromat 100× and 63× lenses (NA 1.40) with Immersol 518F oil at RT. 0.2-μm z sections were captured with a camera (Orca ER; Hamamatsu) and analyzed with confocal software (Leica) or Openlab 4 software.
Antibodies used for immunostaining that were different to those used in Western analysis were as follows: α4β1 integrin and laminin (Klinowska et al., 1999), GoH3 (ICN Biomedicals), ZO-1 (Chemicon), cleaved Caspase 3 (Cell Signaling Technology), ILK (Upstate Biotechnology), α–smooth-muscle actin (Sigma-Aldrich), β4 integrin, β catenin (BD Biosciences), secondary α-mouse Cy2 or Rhrx, α-rabbit Cy2 or Rhrx, and α-rat Rhrx (Jackson ImmunoResearch Laboratories).
We thank Jane Kott for assistance with the confocal microscopy; Martin Humphries, Keith Brennan, and Andrew Gilmore for critical review of the manuscript; and Nasreen Akhtar, Jan Prince, and the other members of the Streuli Laboratory for their help and support throughout the course of this work.
The study was supported by a program grant from the Wellcome Trust.
M.J. Naylor's present address is Developmental Biology Program, Victor Chang Cardiac Research Institute, Darlinghurst, New South Wales 2010, Australia.
Abbreviations used in this paper: Adβgal, β-galactosidase adenovirus; AdCre, Cre-recombinase adenovirus; Blg, β lactoglobulin; BM, basement membrane; ILK, integrin-linked kinase; Prl, prolactin; Wap, whey acidic protein.