Impaired tight junction sealing and precocious involution in mammary glands of PKN1 transgenic mice

The mammary gland undergoes most of its morphological development and functional differentiation in the adult animal under the control of reproductive hormones. For this reason, it is a valuable model for studies of development and hormone regulation as well as for the in vivo analysis of signaling pathways (Daniel and Smith, 1999). During the transition from pregnancy to lactation, the mammary gland undergoes a programmed set of complex changes during which many metabolic and secretory pathways are coordinated to establish copious milk secretion in a process termed secretory activation or lactogenesis 2 (McManaman and Neville, 2003). The sealing (or closure) of the tight junctions between neighboring mammary epithelial cells has been identified as a key component of this transition, and a fall of systemic progesterone levels as well as the presence of prolactin and glucocorticoids is required for this sealing to occur (Nguyen et al., 2001b; Nguyen and Neville, 1998).

The precise mechanisms governing tight junction closure at secretory activation have not yet been elucidated. A potential signaling pathway, the Rho family of small GTPases is linked by some tantalizing evidence to the regulation of tight junctions in the mammary gland and other epithelial and endothelial cells (Matter and Balda, 2003). In con8 mammary epithelial tumor cells, it has been demonstrated that glucocorticoid downregulation of RhoA is required for the organization of the apical junctional complex and tight junction formation (Rubenstein et al., 2003). Furthermore, Jou et al. studying MDCK cells, found that both constitutively active and dominant-negative RhoA and Rac1 impaired tight junction functionality in a dose-dependent and reversible manner (Jou et al., 1998). Notably, constitutive activation of RhoA resulted in a disorganization of the tight junction complexes in these cells reminiscent of freeze-fracture studies of the tight junction complexes in the pregnant mammary gland (Pitelka et al., 1973). On the basis of these studies we hypothesize that a decrease in Rho signaling may be involved in secretory activation, and particularly in the tight junction closure that occurs at that time.

The small GTPases of the Rho family act as molecular switches that cycle between an inactive GDP-bound and an active GTP-bound state to regulate a plethora of cellular functions including transcriptional activation, the regulation of membrane trafficking, cell adhesion, migration and cell cycle progression (Etienne-Manneville and Hall, 2002). These functions are coordinated by the differential activation of effector molecules including kinases of the ROCK/Rho-kinase and the PKN/PRK family as well as a number of other proteins (Bishop and Hall, 2000). Among these targets, protein kinase N1 (PKN1) is a conserved, ubiquitously expressed Ser/Thr kinase having a catalytic domain highly homologous to PKC (Mukai, 2003). In addition to RhoA and Rac1 (Owen et al., 2003), PKN1 can be activated by unsaturated fatty acids, limited proteolysis and PDK1 (Mukai, 2003). A considerable number of molecules interacting with PKN1 in vitro have been identified, and there is accumulating evidence that some of the functions of this kinase are independent of the cytoskeletal effects of RhoA. PKN1 has been established as part of a signaling cascade that links Rho to the Jun promoter (Marinissen et al., 2001) and results from various experimental systems support a role for this kinase in the regulation of transcription (Kitagawa et al., 1998; Morissette et al., 2000). The fact that PKN1 is able to activate the androgen receptor implicates it in the progression of prostate cancer and recent data suggest that it may be involved in the action of other steroid hormones (Metzger et al., 2003). Despite a growing number of binding partners for PKN1, its physiological role remains elusive, in part because of a lack of selective inhibitors and suitable in vivo models to study its function.

To begin to dissect the role of the Rho pathway, and specifically PKN1 in secretory activation, we generated transgenic mice that express a constitutively active form of PKN1 within the mammary epithelium. PKN1 transgenic mice failed to lactate and displayed a phenotype of precocious involution, accompanied by a perturbation of tight junction sealing at parturition. We also provide evidence that PKN1 also interferes with glucocorticoid-mediated tight junction sealing in vitro.

To direct expression to the mammary gland, a constitutively active fragment of PKN1 with the Flag tag at the N-terminus was placed under the control of the MMTV-LTR promoter and the polyadenylation sequence of the SV40 virus was used to facilitate transcription (Fig. 1A). Founder animals were identified by Southern blot as well as a PCR screen using transgene-specific primers and mated with wild-type animals to establish six transgenic lines. To assess transgene expression, RNA isolated from mammary glands of each line was subjected to RT-PCR analysis. The transgenic transcript was detected in animals of lines 933 and 995 (Fig. 1B). Northern blot analysis with a probe consisting of the human PKN1 kinase domain confirmed these results (data not shown).

We next evaluated transgene expression at the protein level in the positive lines by immunoprecipitation using an antibody against the Flag epitope. We were able to demonstrate the presence of a protein of ∼53 kDa that reacted with an antibody directed against the C-terminus of the human PKN1 in animals of both these lines. The mass of the protein is consistent with the reported molecular mass of the PKN1 kinase domain (Ueyama et al., 2001). By contrast, no signal was obtained in other transgenic lines or wild-type animals (Fig. 1C).

The temporal expression pattern of the transgene with respect to distinct phases of mammary gland development was analyzed by northern blot and a band of the expected size could be detected at all developmental stages. Fig. 1D demonstrates that compared with lactation, expression of the transgene was low in virgin animals and during the first half of pregnancy and more RNA was needed to obtain a hybridization signal. Strong signals were obtained from lactating tissues, whereas during involution, the expression level of the transgene decreased again. This temporal expression pattern is consistent with the previously reported activity of the MMTV-LTR promoter (Wagner et al., 2001). Transgenic animals of both lines showed no gross phenotypic abnormalities and developed normally to the adult stage. They were fertile and transgenic females gave birth to normal sized litters.

To determine whether the expression of constitutively active PKN1 interferes with the functionality of the mammary gland, mortality and growth rate of the pups were monitored. With litter sizes not significantly differing between transgenic (11.4 pups) and wild-type animals (12.9 pups, P=0.29), the perinatal mortality determined as the fraction of pups dying within the suckling period was significantly higher in transgenic animals (19.3%) compared with wild-type animals (0.35%, P=0.003). Of the remaining pups, offspring suckled by transgenic mothers of either line gained significantly less weight than pups suckled by wild-type dams (Fig. 2). To determine whether this phenotype was primarily due to lactational failure in transgenic mothers or related to the inability of their offspring to suckle, foster mother experiments were conducted. Pups of transgenic mothers fostered to wild-type dams had completely normal growth rates and mortality. By contrast, wild-type pups fostered to transgenic mothers displayed decreased weight gain and increased mortality (data not shown) showing that this phenotype depends solely on the genotype of the mother. We therefore analyzed mammary gland morphology in transgenic animals to determine whether structural changes in the mammary gland accompanied the observed functional deficiency.

Neither whole mount preparations nor conventional histology demonstrated morphological abnormalities in transgenic mice of line 933 in either the virgin state or the first half of pregnancy (data not shown). Toward the end of pregnancy, transgenic mammary glands were indistinguishable from wild-type organs with respect to side branching and alveolar budding (Fig. 3A). As in their wild-type counterparts, epithelial cells in transgenic animals were organized in alveoli and displayed large cytoplasmic lipid droplets indicative of a proper differentiation in late pregnancy (Fig. 3B, arrow) (Richert et al., 2000). However, with the onset of lactation, striking differences were observed between transgenic and wild-type animals. Glands from transgenic mice appeared sparse in whole mounts, in contrast to wild-type glands, which nearly filled the fat pad (Fig. 3C). Histological sections revealed that transgenic alveoli had failed to expand at the onset of lactation as indicated by the small size of their lumina. In addition, the contents of the alveolar lumina appeared condensed and stained intensely with eosin (Fig. 3D). This finding became more pronounced at the second day of lactation, when a dense amorphous eosinophilic substance resembling thickened secretion product could frequently be observed in condensed transgenic alveoli. In addition, pyknotic nuclei were seen in transgenic mammary glands at this stage and frequently, apoptotic bodies appeared in the lumen of mammary milk ducts and alveoli (Fig. 3E, arrows). Consistent morphological findings were obtained in both transgenic lines at this stage. In addition to alveoli exhibiting this altered phenotype, alveoli with a normal appearance were also present in the mammary glands of transgenic animals in early lactation.

Because transgenes driven by the MMTV promoter have been shown previously to be expressed in patches (Wagner et al., 2001), we attempted to determine whether transgene expression was correlated with the altered morphology observed in parts of the transgenic mammary glands employing in situ hybridization. As Fig. 3F shows, constitutively activated PKN1 was expressed in a patchy manner throughout the gland. Notably, small alveoli that had failed to expand at secretory activation were characterized by a higher rate of transgene expressing cells compared with alveoli that had a more normal morphology, where only a few cells expressed the transgene. Thus, the morphological abnormalities observed in mammary glands of transgenic animals correlated with the expression of the transgene.

As lactation progressed, wild-type mammary glands maintained a fully differentiated state and stromal adipocytes were largely regressed until lactation ceased after 21 days. By contrast, a progressive loss of epithelial tissue was observed in transgenic mammary glands over the course of lactation. After 5 days, focal loosening of the epithelium was seen resulting in widely regressed glands after 10 days of lactation (Fig. 3G). Taken together, the appearance of apoptotic bodies and the morphological findings over the course of lactation are consistent with precocious involution in transgenic mice. No differences were observed in fully involuted glands between wild-type and transgenic animals (data not shown).

In addition to the these morphological findings in lactation, common to both transgenic lines, transgenic mice of line 995 frequently exhibited a defect in ductal elongation and side branching during their pubertal development, resulting in a very inhomogeneous morphology in pregnancy and lactation. As these additional defects were only observed in one of the transgenic lines, the contribution of an effect related to the integration site of the transgene can not be investigated using the existing transgenic lines. However, because the morphological findings in lactation were common to both independent lines, they probably reflect specific effects related to the expression of the transgene. For this reason we focused on line 933 for further analysis,

The ability to synthesize and secrete milk proteins is one of the defining characteristics of the mammary epithelial cell during lactation. To further characterize the developmental defect in PKN1 transgenic mice, we therefore assessed the expression level of the major milk proteins WAP, β-casein, α-lactalbumin and WDNM in transgenic mice at the onset of lactation using microarray technology and, as shown in Fig. 4A, none of these genes was significantly altered in its expression. Northern blot hybridizations using a probe specific for the murine WAP transcript confirmed these results (data not shown) suggesting that milk protein gene transcription and turnover are normal in transgenic mice on a global level. To determine whether a secretory defect was present and whether individual cells in alveoli that had failed to expand at parturition were capable of milk protein synthesis, we examined the localization of β-casein by immunohistochemistry. As Fig. 4B demonstrates, a strong signal for β-casein could be detected in the lumen of morphologically abnormal alveoli by immunohistochemistry (upper panel). Remarkably, whereas at high power (lower panels) immunoreactivity for β-casein was localized mainly to the alveolar lumina and the very apical portions of the cell in wild-type glands, casein localization appeared to be altered in transgenic animals (Fig. 4B, lower panel). In addition to dense staining in the alveolar lumen, large pools of β-casein were present throughout the epithelial cell and not restricted to the apical part of the cells as in the wild-type animals. At the ultrastructural level, large vesicles with retained casein micelles within the cells were observed (Fig. 4C). These observations suggest a secretory defect that prevents efficient movement of secretory vesicles and possibly lipid droplets to the apical membrane. As such defects are often accompanied by premature involution (Schwertfeger et al., 2003), we next sought evidence of precocious involution in transgenic mammary glands.

During the first phase of mammary gland involution, mammary epithelial cells undergo apoptosis followed by the clearance of these cells by neighboring epithelial cells or shedding into the ductal lumen (Furth et al., 1997). The TUNEL method was used to detect apoptotic cells in various developmental stages and, as shown in Fig. 5A,B, the fraction of apoptotic cells was low in late pregnancy in the wild type as well as in transgenic animals. However, an increase in the percentage of apoptotic cells was observed in transgenic mice at the onset of lactation reaching statistical significance at the second day postpartum (Fig. 5B). At later stages of lactation, the number of apoptotic cells in transgenic mammary glands returned to the low level of their wild-type counterparts.

In addition to abundant apoptosis, the early phase of mammary gland involution is characterized by the induction of numerous genes, of which Tgfb3 and Igfbp5, in particular, have been shown to be induced solely by local factors independently of systemic hormone levels (Nguyen and Pollard, 2000; Tonner et al., 1997). We therefore analyzed the expression of these genes at the onset of lactation using microarray technology and, as depicted in Fig. 5C, both Igfbp5 and Tgfb3 were significantly upregulated on the first day of lactation in transgenic mice. Furthermore Tgfb3 expression was virtually absent in wild-type animals at the second day of lactation when analyzed by northern blot, whereas clear signals were obtained in transgenic animals (Fig. 5D). Taken together, these data indicate that precocious involution occurs in PKN1 transgenic mice.

One important change that occurs during secretory activation is the sealing of the tight junctions between mammary epithelial cells (Nguyen and Neville, 1998). Injection of radiolabeled sucrose into the mammary milk duct has proved to be a valid tool to assess the functionality of tight junctions in the mouse mammary gland (Nguyen et al., 2001b). We therefore used this technique to determine whether tight junction closure at secretory activation is impaired in PKN1 transgenic mice.

When injected into the milk duct of wild-type mice at the first day of lactation, only a small fraction of the injected tracer could be recovered in blood samples 5 minutes after injection. By contrast, transgenic animals of line 933 injected at the same developmental stage displayed a significantly higher concentration of [¹⁴C]sucrose in their circulation indicative of an increased permeability of their tight junctions at the onset of lactation (Fig. 6A).

We next investigated whether this functional impairment was accompanied by structural changes of the tight junctions in PKN1 transgenic mice. Using transmission electron microscopy, we were not able to detect any differences in the ultrastructural appearance of the tight junction complexes between wild-type animals and PKN1 transgenic mice (data not shown). We then analyzed the localization of the tight junction associated protein ZO-1 by immunohistochemistry (Fig. 6B). In wild-type animals, ZO-1 staining was restricted to a narrow band at the cell border in the plane of the tight junction. In PKN1 transgenic mice, ZO-1 localized to the apical borders of the cells as well, although focally broadened ZO-1 staining was observed in these animals similar to that seen in wild-type mice during pregnancy (arrow in Fig. 6B). When assessed by immunohistochemistry, the distribution of occludin was not altered in transgenic animals. Similarly, no difference was observed with respect to the localization of the adherens-junction-associated molecules E-cadherin and β-catenin (data not shown). In sum, tight junction formation, as indicated by localization of tight junction proteins to the junctional complexes of the alveolar cells, was not altered by expression of PKN1, but tight junction sealing, indicated by sucrose permeability, was at least partially inhibited in the transgenic mice.

To determine whether similar effects could be seen in vitro, we examined the effect of expression of PKN1 and dominant-negative PKN1 in a mammary cell culture system.

The EPH4 cell line was subcloned from the Ip-1 cell line (Fialka et al., 1996), in turn a derivative of the lactogenic-hormone-sensitive parental line IM-2 originally cultured from the fourth mammary gland of mid-pregnant BALB/c mice (Reichmann et al., 1992). The line produces a monolayer of uniform cuboidal cells with a transepithelial resistance (TER) of up to 7000 ohm cm² when grown on a Transwell filter (Vietor et al., 2001). In our hands these high resistances were achieved much more rapidly in the presence of glucocorticoids as can be seen in Fig. 7A. In the absence of hydrocortisone, the cells achieved a TER of about 1000 ohm cm² after 4 days in culture. Addition of hydrocortisone initiated a rapid increase in resistance and as depicted in Fig. 7B, a reduction in the paracellular permeability of radiolabeled mannitol paralleled the observed increase in the TER. Immunohistochemistry for the tight junction proteins occludin and ZO-1 showed that even at subconfluent density these proteins were present in these cultures at the appropriate apical localization (Fig. 8). With time in culture, the cell borders became more regular but there was no difference in the localization of the tight junction proteins. Thus hydrocortisone is not necessary for tight junction formation in this cell line, but does promote tight junction sealing, similarly to its action on tight junctions during the transition from pregnancy to lactation (Nguyen et al., 2001a). Therefore, the EpH4 system represents a valid model system to investigate the sealing of the mammary tight junctions at secretory activation.

To investigate the effect of PKN1 activation on mammary tight junction function in vitro, EpH4 cells were stably transfected to express constitutively activated PKN1 under the control of the RSV promoter (Fig. 9A). Transfected cells grew with kinetics similar to those of untransfected cells and showed no morphological abnormalities. Tight junction permeability was monitored by measuring the TER and as demonstrated in Fig. 9B, both PKN1 and mock-transfected cells exhibited a gradual increase in the TER when grown for 5 days in the absence of glucocorticoids. In response to hydrocortisone, monolayers of mock-transfected cells displayed a rapid increase in TER as demonstrated before; this response was dramatically delayed in PKN1-transfected cells.

Tight junction formation did not appear to be altered in PKN1-transfected cells because ZO-1 and occludin localized correctly to the apical cell borders in both transfected and control cells after hydrocortisone treatment for 5 days (Fig. 9C). Taken together, PKN1 activation highly diminished the hydrocortisone effect on transepithelial resistance without affecting ZO-1 and occludin targeting to the tight junction.

As the data obtained so far suggest a negative role for PKN1 in the regulation of mammary tight junction sealing, we next tested the effect of a dominant-negative PKN1 mutant on TER development in EpH4 cells. Stable transfection of these cells with PKN1 T774A, a point mutant rendered catalytically inactive by an amino acid exchange within its activation loop (Mukai, 2003), again did not significantly affect growth kinetics or cell morphology. However, when grown on Transwell filters, tight junction sealing was greatly accelerated in these cells as reflected by significantly higher TERs compared with control cells either in the absence or presence of hydrocortisone stimulation (Fig. 10). Notably, cells expressing the dominant-negative form of PKN1, when grown without hydrocortisone, displayed a TER pattern very similar to control cells stimulated with steroids, suggesting that inhibition of PKN1 signaling is sufficient to mimic the stimulatory effects of hydrocortisone on tight junction sealing in this model system. Stimulation of EpH4 cells expressing PKN1 T774A with hydrocortisone resulted in an additional increase of the TER, furthermore supporting the notion that PKN1 negatively regulates glucocorticoid-mediated tight junction sealing in vitro.

To test the hypothesis that signaling through the Rho pathway plays a role in secretory activation, transgenic mice expressing a constitutively activated form of the Rho effector kinase PKN1 in the mammary epithelium were used. We found that activated PKN1 had no effect on differentiation in late pregnancy, but that it inhibited both secretion of milk components and tight junction closure on day 1 of lactation, leading to a phenotype of precocious involution that diminished the ability of the dams to support a litter.

Although the developmental cycle of the mouse mammary gland is highly complex (Richert et al., 2000), consistent morphological alterations in two independent transgenic lines were observed only at the onset of lactation. Although the finding of a defect largely confined to this stage could reflect the hormone dependence of the MMTV promoter, numerous models using the same promoter display defects in either the virgin state or during pregnancy (Jager et al., 2003; Jiang and Zacksenhaus, 2002). It therefore seems reasonable to argue that PKN1 activation selectively interferes with cellular processes occurring at secretory activation. Our observation that cells in morphologically abnormal alveoli were capable of milk protein synthesis furthermore suggests that PKN1 does not block differentiation per se but rather interferes with a more defined set of processes that lead to a delayed switch between the pregnant and lactating state upon parturition.

Among the processes occurring at secretory activation, tight junction sealing has been postulated to be a prerequisite for successful lactation (Itoh and Bissell, 2003). Using intraductal injection of a radiolabeled tracer as well as determining the TER in cultivated mammary epithelial cells, we demonstrated that PKN1 activation results in impaired tight junction sealing in vitro as well as in vivo. This is of particular interest, because Rho activation has previously been linked to a perturbation of tight junction function in various experimental systems (Hopkins et al., 2003; Jou et al., 1998; Wojciak-Stothard et al., 2001). In Con8 cells, expression of constitutively activated RhoA prevented tight junction formation upon glucocorticoid treatment (Rubenstein et al., 2003) and expression of constitutively activated RhoA increased tight junction permeability in MDCK cells (Jou et al., 1998). However, Rho inhibition has also been demonstrated to perturb tight junction function, suggesting that the activity of these molecules is carefully balanced and highly dependent on cellular context (Matter and Balda, 2003). One reason for this complexity may be that a plethora of molecules is involved in the downstream transduction of Rho signals within the cell. A number of previously published reports support a role for PKN1 in the regulation of tight junctions, although there has been no direct investigation of its involvement. For example, an increase in paracellular permeability in response to Rho activation was reversed by the kinase inhibitor Y-27632 in a number of experimental systems (Stamatovic et al., 2003; Wojciak-Stothard et al., 2001). This compound has been widely used as an inhibitor for Rho kinase, but a detailed analysis revealed that it equally inhibits the PKN1 isoform PKN2 (Davies et al., 2000) as well as PKN1 itself (Y.O., unpublished results). In MDCK II cells, RhoA activation increased the TER which could be reversed by expression of either of two Rho mutants RhoAV14/L40 and RhoAV14/C42 (Fujita et al., 2000), both of which have been shown to be defective in their binding to PKN1 (Sahai et al., 1998). These findings are consistent with our notion that PKN1 might be involved in the regulation of tight junction permeability downstream of Rho GTPases.

It is important to note that two distinct processes are involved in tight junction regulation: assembly or formation of tight junctions and their sealing or closure, both of which have been shown to be independently regulated (Walsh et al., 2001; Woo et al., 1999). Assembly involves the movement of tight junction components to the apical borders of the cells. In Con8 cells, it is this process that is primarily regulated by glucocorticoids (Woo et al., 1999). However, in the in vivo mammary gland and 31EG4 cells (Zettl et al., 1992), proteins such as ZO1 and occludin are normally localized at the tight junction in the absence of glucocorticoids. We show here that the same is true for EpH4 cells and that glucocorticoids bring about tight junction sealing. We observed that this process was dramatically diminished in EpH4 cells transfected with active PKN1, whereas no changes could be detected in the absence of glucocorticoids, arguing against the assumption that a non-specific effect of activated PKN1 on cell monolayer characteristics is responsible for the observed perturbation of paracellular permeability. This finding was further confirmed by our observation that expression of a dominant-negative PKN1 mutant resulted in accelerated tight junction sealing in EpH4 cells. The fact that the distribution of ZO-1 and occludin remained unchanged suggests that PKN1 rather regulates sealing but not formation of tight junctions. Taken together, both our in vivo and in vitro experiments support the hypothesis that activated PKN1 inhibits tight junction closure in the mammary gland, suggesting a role for this molecule in maintaining the open tight junctions of pregnancy. In addition, it has been recently reported that the balance between the two isoforms of the Ser/Thr kinase ROCK participates in the glucocorticoid regulation of tight junction permeability downstream of RhoA in con8 cells (Rubenstein et al., 2007). Our observations point to an additional pathway, by which Rho signaling can modulate the regulation of paracellular permeability by glucocorticoids.

As PKN1 has been implicated in transcriptional activation in various systems (Mukai, 2003), its action on tight junction permeability potentially might involve changes in the transcription of tight-junction-associated proteins, possibly by interfering with glucocorticoid-regulated transcriptional activation. In this regard, it seems noteworthy that microarray analyses revealed a significant downregulation of the glucocorticoid receptor in PKN1 transgenic mice. Furthermore, significant alterations in the expression of claudin 3 and claudin 8 were found by microarray analysis in these animals (unpublished data). Claudins represent a large class of molecules implicated in the regulation of tight junction function (Schneeberger and Lynch, 2004), yet their significance in the mammary gland is presently unclear. Determining the localization of the various claudin family members in the mammary gland could significantly advance our understanding of the permeability switch occurring at secretory activation and experiments addressing this question are currently underway in our laboratory.

Accompanying impaired tight junction sealing, mammary glands of transgenic animals displayed numerous morphological abnormalities in early lactation and transgene expression correlated with these abnormalities as shown by in situ hybridization. Over the course of lactation, we demonstrated a phenotype of precocious involution in PKN1 transgenic mice reflected in increased apoptosis in early lactation followed by a progressive loss of epithelium. Whereas the precise mechanism underlying this observation remains unclear, the fact that the relative expression level of the transgene remained unchanged during lactation suggests that programmed cell death does not selectively affect transgene-expressing cells and is therefore unlikely to be a direct consequence of PKN1 activation, consistent with previous reports (Ueyama et al., 2001). As the tight junction has been identified as a central signaling interface within the cell (Matter and Balda, 2003), it is possible that a signaling pathway from the mammary tight junction initiates or maintains secretion and cell survival. Indeed, recent results from our laboratory suggest that disruption of the mammary tight junctions results in increased apoptosis under various conditions in vitro (N. E. Beeman and M.C.N., unpublished results), thereby further supporting the idea of a causal connection between impaired tight junction sealing and the induction of involution in the mammary gland.

Taken together, our observations suggest that the Rho signaling pathway and PKN1 are part of the program that maintains the state of pregnancy including open tight junctions in the mammary gland and that prolonging this program by activation of PKN1 results in precocious involution. Thus, whereas the precise molecular mechanism by which PKN1 elicits this effect remains to be elucidated, our data demonstrate for the first time a role for the Rho signaling pathway in the regulation of secretory activation.

NMRI outbred mice were obtained from Harlan-Winkelmann (Paderborn, Germany). Pregnancies were staged by observing a vaginal plug after mating and this day was subsequently considered the first day of pregnancy (P1). Intraductal injection of tracer substances into the murine mammary gland has been described in detail before (Nguyen et al., 2001b; Nguyen et al., 2000). The Institutional Animal Care and Use Committee of the University of Colorado Health Sciences Center and German authorities approved all procedures, which follow the guidelines of the United States Department of Agriculture and the European Communities Council Directive for care of laboratory animals.

A 2.3 kb fragment encoding the MMTV-LTR promoter (kindly provided by Diego Walther, Max Planck Institute for Molecular Genetics, Berlin, Germany) and a 900 bp fragment encoding the SV40 small intron and polyadenylation site (kindly provided by Yvan de Launoit, Department of Molecular Virology, University of Brussels, Belgium) were ligated into the pBluescript II SK+ vector (Stratagene, La Jolla, CA). The C-terminus of human PKN1 (a.a. 511-942), acting as constitutively active form of this molecule (Yoshinaga et al., 1999), was tagged with the Flag epitope and cloned between these fragments (Fig. 1A). Transgenic mice were generated as previously described (Theuring et al., 1990) and identified by PCR (annealing temperature 60°C, 30 cycles) using the transgene-specific primers PKN.Flag (5′-TGGACTACAAGGACGACGACG-3′) and PKN.rev (5′-GCCGCCAATATCCGCTTCTCAC-3′).

Total RNA was extracted from snap-frozen tissue using Trizol Reagent (Invitrogen, Life Technologies, Karlsruhe, Germany) whereas the RNeasy kit (Qiagen, Valencia, CA) was used to isolate RNA from EpH4 cells. Standard northern blot hybridization was performed using the radiolabelled human PKN1 kinase domain to detect transgene expression. Probes for Tgfb3 and the murine whey acidic protein (WAP) were kindly provided by Jeffrey W. Pollard (Department of Developmental and Molecular Biology, Albert Einstein College of Medicine, New York, NY) and Akihiko Yoshimura (Department of Immunobiology and Neuroscience, Kyushu University, Fukuoda, Japan), respectively.

Blots were analyzed using a STORM PhosphorImager and ImageQuant TL software (Amersham Biosciences, Freiburg, Germany). Transgene expression was assessed by standard RT-PCR using the primers PKN.Flag and PKN.rev. As a control, samples in which the reverse transcriptase was omitted were analyzed in parallel.

In situ hybridization using digoxigenin-labeled probes of the SV40 polyadenylation sequence was performed as previously described (Ansorge et al., 2004). The procedures for microarray hybridizations to Mu74AV2 chips (Affymetrix, Santa Clara, CA) have been described in detail elsewhere (Rudolph et al., 2003).

For protein isolation, snap-frozen tissue was pulverized and extracted in TSA (10 mM Tris-HCl, pH 8.0, 140 mM NaCl, 0.025% NaN₃) containing 0.5% Triton X-100 and proteinase inhibitors (Complete proteinase inhibitor cocktail, Roche Diagnostics, Mannheim, Germany) on ice for 60 minutes. Immunoprecipitation was carried out by incubation of protein extracts with 2.5% anti FLAG M2 affinity gel (Sigma, Taufkirchen, Germany) and western blots were performed using the rabbit polyclonal antibody PKN-H234 (Santa Cruz Biotechnology, Santa Cruz, CA) at a concentration of 1 μg/ml.

Formalin-fixed tissue was paraffin embedded and 5 μm sections were stained with hematoxylin and eosin according to standard procedures. Mammary whole mount preparations were obtained as previously described (Binas et al., 1995). For electron microscopy, mammary tissue was dissected into cubes of approximately 1 mm³ size and fixed in 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer overnight. Standard electron microscopy was then performed using EPON 812 as embedding medium and an EM 900 transmission electron microscope (LEO, Oberkochen, Germany).

Immunohistochemistry was performed as previously described (Russell et al., 2003). Primary antibodies against ZO-1 (Chemicon, Temecula, CA), occludin (Zymed, San Francisco, CA) and β-casein (Ab 7781, MCN) were diluted 1:100 and secondary antibodies coupled to Cy3 or fluorescein isothiocyanate (FITC) (Jackson ImmunoResearch Laboratories, West Grove, PA) were applied at a dilution of 1:150. Oregon-Green-488-conjugated wheat germ agglutinin (Molecular Probes, Eugene, OR) was used to visualize the alveolar lumen. All slides were mounted using the ProLong antifade kit (Molecular Probes). Images were collected using SlideBook software (Intelligent Imaging Innovations, Denver, CO) on a Nikon Diaphot TMD microscope equipped for fluorescence with a Xenon lamp and filter wheels (Sutter Instruments, Novato, CA), fluorescent filters (Chroma, Brattleboro, VT), cooled CCD camera (Cooke, Tonawanda, NY), and stepper motor (Intelligent Imaging Innovations). Images were digitally deconvolved using the No Neighbors algorithm (Slidebook software, Intelligent Imaging Innovations).

Apoptotic rates were determined from 10 μm paraffin sections of formalin-fixed tissue using the Fluorescein in situ cell death detection kit (Roche Diagnostics, Indianapolis, IN) according to the manufacturer's instructions. Nuclei were counterstained with DAPI (Sigma) and images were taken as described above. At least four animals were analyzed for each time point and genotype and 1000-2500 nuclei per animal were counted to determine the apoptotic rate calculated as the percentage of TUNEL-positive epithelial nuclei of all epithelial nuclei.

EpH4 mammary epithelial cells (kindly provided by Irene Fialka, Research Institute of Molecular Pathology, Vienna, Austria) were grown in DMEM (GibcoBRL, Grand Island, NY) supplemented with 5% FCS (Gemini Bioproducts, Calabasas, CA), 1% penicillin-streptomycin (GibcoBRL, Grand Island, NY) and 10 mM HEPES (Sigma) and subcultured every 3-4 days. For measurement of the TER, cells were seeded at confluent density on Transwell inserts with a pore size of 0.4 μm (Corning Costar Corporation, Cambridge, MA). For the induction of tight junction sealing, hydrocortisone (Sigma) was added to the growth medium at a concentration of 5 μmol/l. For transfection of EpH cells, the Flag-tagged contitutively activated PKN1 fragment described above was cloned into the pRC/RSV vector (Invitrogen, Carlsbad, CA), whereas a construct encoding the Flag-tagged catalytically inactive point mutant PKN1 T774A was the kind gift of Hideyuki Mukai (Biosignal Research Center, Kobe University, Japan). The Lipofectamine Plus system (Invitrogen) was used according to the manufacturer's recommendations and control cells were obtained by transfection of the respective empty vectors alone. Expression was validated by standard RT-PCR using the primers PKN.FLAG and PKN.rev for the constitutively active fragment and PKN.for (5′-GTCTTCGACAGCATCGTCAA-3′) and Flag.rev (5′-CTTATCGTCGTCGTCCTTGT-3′) for the catalytically inactive mutant, respectively. For immunocytochemistry, Transwell inserts were fixed in 2% paraformaldehyde for 10 minutes at 4°C and processed as described above. The TER was measured with an EVOM-G voltohmmeter (World Precision Instruments, New Haven, CT). To monitor [³H]mannitol permeability (Toddywalla et al., 1997), approximately 5×10⁵ d.p.m. [³H]mannitol were added to the top of the Transwell. The filters were maintained at 37°C on a rocker at low speed and 50 μl and 150 μl samples were taken from the top and bottom wells, respectively, 5 hours later. Radioactivity was determined with β-scintillation counting using Budget Solve (Research Products International, Mount Pleasant, IL).

This work contains experimental data that are part of the M.D. thesis of A.F., which has been submitted to the Faculty of Medicine of the Charité University Medicine (Berlin, Germany). The authors would like to thank Petra Schrade and Julia Foo for technical assistance, B. J. Davies and J. L. McManaman for helpful discussion, and J. W. Pollard, A. Yoshimura, I. Fialka, H. Mukai, D. Walther, Y. de Launoit and E. Hara for providing materials. This work was supported by a fellowship from Charité University Medicine, Berlin, and from the Biomedical Sciences Exchange Program between North America and Europe Inc to A.F. and NIH grants PO1-HD38129 and R37 HD19547 to M.C.N.