Delineation of proteolytic and non-proteolytic functions of the membrane-anchored serine protease prostasin

Trypsin-like serine proteases are a large family of proteolytic enzymes that reside within the extracellular and pericellular space of all vertebrate species. They are emerging as major regulators of a variety of biological processes in the context of tissue development and homeostasis owing to their involvement in the activation of hormones, growth factors and cytokines, the activation and ectodomain shedding of signaling receptors and adhesion molecules, regulation of ion fluxes and extracellular matrix remodeling (Afonina et al., 2015; Chin et al., 2008; Naldini et al., 1992; Picard et al., 2008; Scudamore et al., 1998; Shimomura et al., 1993; Vallet et al., 1997; Wang et al., 1994; Yan et al., 2000).

Prostasin [or protease, serine 8 (PRSS8); also known as CAP1] and matriptase [also known as suppression of tumorigenicity 14 (ST14), MT-SP1, PRSS14] are trypsin-like membrane-anchored serine proteases expressed in epithelial compartments of mouse and human tissues (Antalis et al., 2011; Bergum et al., 2012; List et al., 2007; Szabo and Bugge, 2011). Studies using genetically modified mouse strains suggest that the two proteases are part of a single proteolytic cascade that plays a central role in the development and homeostasis of a variety of epithelia. Mice with a systemic loss of either prostasin or matriptase display strain-dependent complete or partial embryonic lethality, and all of the matriptase or prostasin null mice that survive until term die shortly after birth due to fatal dehydration (Hummler et al., 2013; Leyvraz et al., 2005; List et al., 2002; Peters et al., 2014). Detailed phenotypic analysis revealed that the matriptase/prostasin pathway is essential for terminal epidermal differentiation as well as tight junction formation and, as a result, the loss of matriptase or prostasin function leads to a severe defect in epidermal barrier function, and an abnormal hair follicle maturation (Leyvraz et al., 2005; List et al., 2009,, 2003). Subsequent studies using tissue-specific knockout mice to bypass the postnatal lethality revealed that matriptase and/or prostasin play crucial roles in epithelial development and function in a wide variety of mouse tissues including placenta, skin, salivary gland, intestines, lungs and thymus (Frateschi et al., 2012; List et al., 2009; Malsure et al., 2014; Planes et al., 2010; Szabo et al., 2014).

Activity of the matriptase/prostasin pathway during development is controlled by two transmembrane serine protease inhibitors: hepatocyte growth factor activator inhibitor 1 (HAI-1) and HAI-2 (also known as SPINT1 and SPINT2, respectively). HAI-1 is essential for placental differentiation and overall embryonic and postnatal survival (Nagaike et al., 2008; Szabo et al., 2007; Tanaka et al., 2005). Loss of HAI-2 is associated with an embryonic lethality around embryonic day (E) 8.5-9.5, and a high frequency of defects in neural tube closure and a mid-gestational embryonic lethality due to a placental failure in matriptase heterozygous mice (Mitchell et al., 2001; Szabo et al., 2009a,, 2012). Importantly, all developmental defects in HAI-1- and HAI-2-deficient mice are rescued in whole or in part by simultaneous inactivation of either matriptase or prostasin, thus demonstrating a crucial contribution of the matriptase/prostasin proteolytic pathway to the lethality observed in these mice (Szabo et al., 2009a,, 2012).

We recently generated knock-in mice that carry a point mutation resulting in the substitution of the catalytic serine 238 in prostasin with alanine and which therefore express only a catalytically inactive variant of the protease (Peters et al., 2014; Szabo et al., 2012). Unexpectedly, these mice did not exhibit profound defects in epidermal development and barrier function, nor the perinatal lethality observed in prostasin null mice (Peters et al., 2014). Loss of prostasin enzymatic activity did result in abnormal hair growth and delayed skin wound healing, indicating that at least some of the physiological functions of prostasin are proteolytic.

In this study, we performed a systematic analysis of physiological, pathophysiological and biochemical processes previously reported to require prostasin to determine the specific contribution of the proteolytic and non-proteolytic functions of the protease to these processes. We found that prostasin proteolytic function is essential for matriptase zymogen activation during establishment of the feto-maternal barrier in the placenta. Unexpectedly, although devoid of catalytic activity and the capacity to activate placental matriptase, prostasin caused defects in placental differentiation and prenatal lethality in mice lacking HAI-1 or HAI-2. Equally unexpectedly, the absence of HAI-1 completely compensated for the loss of all proteolytic and non-proteolytic functions of prostasin, as mice with combined deficiency in HAI-1 and prostasin displayed normal development and postnatal survival. Taken together, these findings provide important new insights into prostasin function and its mechanistic relationship with other components of the matriptase/prostasin proteolytic pathway.

The essential role of prostasin in mouse postnatal survival is well established (Leyvraz et al., 2005; Peters et al., 2014; Szabo et al., 2012). Reports of its role in embryonic development, however, show far less concordance. A study by Hummler et al. (2013) reported that loss of prostasin in an inbred C57 strain leads to defects in placental differentiation and complete embryonic lethality. By contrast, we have recently reported 50-100% viability of either matriptase or prostasin null embryos in outbred strains (Peters et al., 2014; Szabo et al., 2009a,, 2014). The importance of matriptase and prostasin during embryogenesis thus appears to be strongly dependent on the genetic background. To enable a direct comparison of the phenotypes resulting from the specific loss of prostasin proteolytic function in mice expressing catalytically inactive prostasin (prostasin knock-in, Prss8^Ki/Ki) and those resulting from a complete loss of all prostasin functions (prostasin null, Prss8^−/−), we established genetically background-matched colonies of Prss8^Ki/+ and Prss8^+/− breeder mice by first interbreeding mice carrying Prss8^Ki and Prss8⁻ alleles. The resulting Prss8^−/− and Prss8^Ki/Ki offspring of these Prss8^+/−×Prss8^+/− and Prss8^Ki/+×Prss8^Ki/+ breeding pairs were then used throughout this study.

Analysis of the prenatal survival of offspring from Prss8^+/−×Prss8^+/− breeders showed a partial but highly significant decrease in prenatal viability in mice lacking prostasin, as fewer than half of the expected number of Prss8^−/− mice were observed at birth (Fig. 1A; P<0.003, χ²). Allele distributions in embryos extracted on or before E12.5 revealed the expected numbers of Prss8^−/− mice (Fig. 1B,C). All of the Prss8^−/− embryos extracted before E12.5 appeared normal, indicating that the loss of prostasin did not affect pre-implantation and early post-implantation development (Fig. 1D). At E13.5, however, the survival of the Prss8^−/− mice decreased sharply, with only 51% of the expected number of prostasin-deficient embryos extracted between E13.5 and E14.5 being found alive (Fig. 1B,C; P<0.04, χ²). The survival rate then remained virtually unchanged for the rest of the prenatal period, with the number of living Prss8^−/− embryos reaching 49% between E15.5 and E16.5 (P<0.04, χ²) and 46% at birth, suggesting that prostasin is specifically required for the relatively narrow interval of mid-gestational development (Fig. 1A-C).

Macroscopic inspection of living and non-living Prss8^−/− embryos at E13.5 and E14.5 did not reveal any obvious growth retardation and/or developmental abnormality that would explain the sudden loss of viability (Fig. 1D′). This would be consistent with a recent report showing that defects in placental rather than embryonic development result in loss of mid-gestational viability in prostasin-deficient C57 mice (Hummler et al., 2013). Importantly, however, in contrast to Hummler et al. (2013) microscopy analysis of placental tissues from living E13.5 Prss8^−/− embryos did not reveal any obvious differences in the overall histological structure or differentiation of the prostasin- and HAI-1-expressing labyrinthine epithelium, when compared with the prostasin-expressing wild-type littermate control tissues (Fig. 1E-G′). Indeed, histomorphometric quantification of the basic structural characteristics of the placental labyrinth of Prss8^−/− E13.5 placentas, including the thickness and cross-sectional area of the labyrinth, as well as the density of fetal vessels within the labyrinth, all showed values comparable to those of wild-type littermates (Fig. 2A-C).

To determine whether the crucial contribution of prostasin to development after mid-gestation in our genetic background is mediated by prostasin proteolytic activity, or by its non-proteolytic functions, we next analyzed prenatal survival of Prss8^Ki/Ki mice. Contrary to the prostasin full knockouts, no decrease in survival was observed in Prss8^Ki/Ki mice at any stage of embryonic development, and these mice were born in the expected Mendelian ratio (Fig. 2D-F). To further validate this finding, mice carrying Prss8^Ki and Prss8⁻ alleles were also directly interbred to generate Prss8^Ki/− breeding pairs capable of producing Prss8^Ki/Ki and Prss8^−/− offspring within the same litter. Consistent with the findings presented above, analysis of the newborn offspring of the Prss8^Ki/− breeders revealed a highly significant decrease in the embryonic survival of Prss8^−/− mice compared with their Prss8^Ki/Ki littermates (Fig. 2G). At birth, Prss8^Ki/Ki mice vastly outnumbered Prss8^−/− newborns (Fig. 2G; observed ratio 67:23, expected ratio 1:1, P<0.0001, χ²). Survival of Prss8^Ki/− mice also appeared to be affected, with an observed ratio between Prss8^Ki/Ki and Prss8^Ki/− newborns of 67:88 (Fig. 2G; expected 1:2, P<0.01, χ²). These data show that prostasin supports embryonic survival after a mid-gestational checkpoint and that this is independent of its proteolytic activity.

We have previously reported that the expression level of matriptase protein and its tissue localization are unaltered in Prss8^−/− embryos, but these embryos display a complete absence of the active two-chain form of matriptase in the placenta (Szabo et al., 2012). We have also recently reported that catalytically inactive prostasin stimulates matriptase autoactivation in a reconstituted cell-based system (Friis et al., 2013), suggesting that placental matriptase activation could be either proteolysis dependent or independent. To investigate this, placental tissues from Prss8^−/−, Prss8^Ki/Ki and their respective wild-type littermate controls were analyzed for the presence of the active two-chain form of matriptase. Expression analysis of placental tissues from Prss8^Ki/Ki embryos at the time of active labyrinth differentiation (E12.5) by immunohistochemistry and western blot did not reveal any obvious effect of the loss of prostasin enzymatic activity on the levels of total prostasin and matriptase proteins (Fig. 3A,B), or their respective patterns of expression in placental epithelium (Fig. 3C-F). Direct detection of active matriptase in placental tissues by western blot is not feasible owing to very low signal intensity (Szabo et al., 2012). Therefore, we determined the amount of active matriptase that formed inhibitor complexes with its cognate inhibitor HAI-1, which selectively binds the active two-chain form of matriptase (Benaud et al., 2001; Lee et al., 2005). Placental lysates from control mice revealed detectable amounts of isolated matriptase serine protease domain (SPD), which is formed by activation site cleavage (Fig. 3G, lanes 2 and 3). By contrast, only minimal amounts of active matriptase were detected in placentas expressing catalytically inactive prostasin, with no obvious difference in the levels of detectable matriptase SPD between Prss8^Ki/Ki and Prss8^−/− mice (Fig. 3G, compare lanes 4 and 5 with 6 and 7), suggesting that activation of matriptase in the developing placental labyrinth is dependent on prostasin proteolytic activity.

Loss of the transmembrane serine protease inhibitors HAI-1 and HAI-2 in mice results in complete embryonic lethality (Fan et al., 2007; Mitchell et al., 2001; Szabo et al., 2009a; Tanaka et al., 2005). Prenatal survival of both HAI-1- and HAI-2-deficient mice can, however, be fully restored by genetic inactivation of either matriptase or prostasin, indicating a crucial contribution of the prostasin/matriptase pathway to the demise of mice lacking either of the two HAI proteins (Szabo et al., 2009a,,b,, 2007,, 2012). To establish whether the prostasin-dependent loss of viability of HAI-deficient mice requires prostasin proteolytic activity, we interbred mice carrying HAI-1 or HAI-2 null alleles (Spint1^+/− or Spint2^+/−, respectively) with mice carrying both prostasin null and catalytically inactive knock-in alleles and analyzed their offspring at birth.

Consistent with our previous reports, expression of prostasin wild-type protein led to a complete loss of prenatal viability in mice lacking HAI-1. Thus, no Spint1^−/− mice carrying two or one wild-type allele of prostasin (Spint1^−/−;Prss8^+/+ or Spint1^−/−;Prss8^+/−) were observed at birth (Fig. 4A; both P<0.0001, χ²). As anticipated, complete loss of prostasin was able to restore embryonic survival in Spint1^−/− mice (Fig. 4A; Spint1^−/−;Prss8^−/−, P=0.51, χ²). Surprisingly, however, breeding of Spint1^+/−;Prss8^Ki/+ mice did not produce any viable Spint1^−/−;Prss8^Ki/Ki mice at birth, indicating that the elimination of prostasin proteolytic function is not sufficient for the completion of prenatal development in the absence of the HAI-1 protein (Fig. 4B; P<0.0005, χ²).

Collection of embryos from Spint1^+/−;Prss8^+/− or Spint1^+/−;Prss8^Ki/+ breeding pairs between E12.5 and E15.5 showed a complete absence of living Spint1^−/−;Prss8^+/+ mice (Fig. 4C,D, P<0.02 and P<0.05, respectively, χ²). As expected from the analysis of newborn mice (Fig. 4A), the viability of HAI-1-deficient mice was completely restored at both embryonic stages in a prostasin null background (Fig. 4C; Spint1^−/−;Prss8^−/−). Importantly, viability was also improved by the selective loss of prostasin proteolytic activity, with about half of the expected Spint1^−/−;Prss8^Ki/Ki embryos still alive at E12.5-E13.5 (Fig. 4D, left; 52%, P=0.41, χ²). However, this effect was transient, and no living Spint1^−/−;Prss8^Ki/Ki embryos were detected after E14.5 (Fig. 4D, right; P<0.02, χ²).

To investigate whether the extended lifespan of HAI-1-deficient mice in the prostasin knock-in background was caused by improved placental function, we next performed a histological analysis of placental tissues extracted from Spint1^−/− embryos that were Prss8^+/+, Prss8^Ki/Ki or Prss8^−/− at E12.5. Control tissues exhibited well-formed placental labyrinth with a highly branched fetal vasculature (Fig. 4E), whereas placentas from Spint1^−/−;Prss8^+/+ mice showed a complete absence of the labyrinth structure and no invading fetal vessels (Fig. 4F,I,J). Prostasin full deficiency was able to restore placental differentiation, as evidenced by the thickness and area of the labyrinth layer in the midline cross-section of the tissue as well as the density of the fetal vasculature within the labyrinth layer of Spint1^−/−;Prss8^−/− mice, which were all comparable to those of wild-type controls (Fig. 4G,I,J). Even though the loss of prostasin proteolytic activity also partially restored labyrinth differentiation in Spint1^−/−;Prss8^Ki/Ki mice, the structure of the labyrinth layer remained abnormal with strongly reduced cross-sectional area and branching of fetal vasculature (Fig. 4H-J).

Loss of HAI-2 leads to an early embryonic lethality with no surviving Spint2^−/− mice identified at or after E10.5 (Mitchell et al., 2001; Szabo et al., 2012) (Fig. 5C). The complete loss of prostasin protein compensates for the loss of HAI-2 during embryogenesis, as evidenced by the presence of living Spint2^−/−;Prss8^−/− mice at birth (Mitchell et al., 2001; Szabo et al., 2012) (Fig. 5A). By contrast, a similar analysis did not reveal any surviving Spint2^−/−;Prss8^Ki/Ki embryos, suggesting that the selective loss of prostasin proteolytic activity is not sufficient to prevent embryonic lethality in HAI-2-deficient mice (Fig. 5B). Furthermore, the loss of prostasin proteolytic activity was not sufficient to extend the life of Spint2^−/− mice when compared with mice expressing wild-type alleles of prostasin, as suggested by a complete lack of living Spint2^−/−;Prss8^Ki/Ki mice among the embryos extracted at E10.5 and E11.5 (Fig. 5C; P<0.01, χ²). As expected from the analysis of the newborn mice (Fig. 5A), Spint2^−/−;Prss8^−/− embryos were found alive and in the expected Mendelian ratio (Fig. 5C). These data indicate that the prostasin-dependent embryonic lethality observed in mice lacking either of the two HAI proteins is mediated by non-proteolytic functions of prostasin, although its proteolytic activity does contribute to placental defects in mice lacking HAI-1.

Loss of prostasin function leads to a catastrophic loss of the epidermal barrier, the absence of whiskers at birth, and a complete perinatal lethality (Leyvraz et al., 2005; Peters et al., 2014; this study). Unexpectedly, analysis of postnatal survival in the offspring from Spint1^+/−;Prss8^+/− breeding pairs showed that, whereas all of the HAI-1-expressing Prss8^−/− mice (Spint1^+/+;Prss8^−/− or Spint1^+/−;Prss8^−/−) died within the first few days after birth, mice deficient in HAI-1 and prostasin (Spint1^−/−;Prss8^−/−) did not exhibit any postnatal lethality and were all found alive at weaning and when followed for up to 4 months (Fig. 5D). Thus, the essential role of prostasin in postnatal survival is completely circumvented in mice lacking HAI-1. At birth, mice lacking both prostasin and HAI-1 did not display any gross developmental abnormalities (Fig. 5E). In contrast to Prss8^−/− mice, the Spint1^−/−;Prss8^−/− newborn mice did present with visible whiskers, although these were shorter than those of wild-type littermate controls (Fig. 5F-H). The initial eruption of body hair was also markedly delayed in Spint1^−/−;Prss8^−/− mice (Fig. 5I). The body size, whisker and body hair growth all normalized between 2 and 3 weeks of age, and the adult double-deficient mice did not exhibit any obvious defects compared with the HAI-1- and prostasin-sufficient littermate controls (Fig. 5J). Similar analysis of the offspring from Spint2^+/−;Prss8^+/− breeding pairs did not reveal any effect of the loss of HAI-2 on the survival of prostasin-deficient mice, as all of the 33 born Prss8^−/− mice (Fig. 5A) died within 48 h after birth, irrespective of their Spint2 genotype. These findings indicate that the absence of HAI-1, but not of HAI-2, can completely compensate for the loss of essential functions of prostasin during postnatal development.

The above findings could be explained by HAI-2, unlike HAI-1, directly inhibiting prostasin rather than matriptase activity, so that only the loss of HAI-1, but not of HAI-2, can compensate for the loss of prostasin by removing inhibition of matriptase and increasing its activity to a level required to sustain development and survival. However, since HAI-2 is also essential for survival in Prss8^Ki/Ki mice (Fig. 5B), in this scenario the inhibitor would also need to regulate non-proteolytic functions of prostasin, predicting an effective binding of HAI-2 to the catalytically inactive protease. To test this hypothesis, we incubated zymogen or matriptase-activated double-chain forms of wild-type and S238A prostasin with recombinant HAI-2 or PN-1, which is a serpin-type (SERPINE2) cognate inhibitor of prostasin. As expected, PN-1, which requires proteolytic cleavage of its reactive center loop to form a stable complex with a protease, only bound the active double-chain form of wild-type prostasin (Fig. 5K). By contrast, double-chain forms of both wild-type and catalytically inactive prostasin were very efficiently sequestered by HAI-2 (Fig. 5K). HAI-2 also formed complexes with the zymogen forms of wild-type and S238A prostasin, although at reduced levels. These findings are consistent with the ability of HAI-2 to regulate both proteolytic and non-proteolytic functions of prostasin.

An overwhelming body of evidence suggests that prostasin, matriptase, HAI-1 and HAI-2 constitute a system of proteases and protease inhibitors that acts on the surface of epithelial cells to regulate multiple aspects of development, postnatal homeostasis and tissue remodeling. First, loss-of-function studies in matriptase- and prostasin-deficient mice reveal virtually identical effects of the elimination of the two proteases on overall survival, epidermal differentiation and barrier formation, hair growth as well as lethality of HAI-1- and HAI-2-deficient embryos (Leyvraz et al., 2005; List et al., 2002,, 2003; Szabo et al., 2012). Second, several recent reports document the ability of prostasin to stimulate matriptase activity in cell culture as well as in tissues, suggesting that prostasin is likely to act upstream of matriptase (Buzza et al., 2013; Camerer et al., 2010; Friis et al., 2013; Szabo et al., 2012). Third, prostasin and matriptase form protein complexes when co-expressed in epithelial cells (Friis et al., 2013). Finally, defects in epidermal development and barrier formation in prostasin null mice that express near-normal levels of the activated two-chain form of matriptase in the epidermis show that, even in its active form, matriptase is unable to perform its essential skin functions in the absence of prostasin (Friis et al., 2016; Szabo et al., 2012).

Surprisingly, our recent work indicates that at least some of the reported physiological functions of prostasin during development are preserved in mice that express a catalytically inactive variant of the protease (Peters et al., 2014). Here, we performed a comprehensive analysis of previously reported functions of prostasin during embryonic and postnatal development, and show that specific prostasin-mediated processes vary in their degree of dependence on its proteolytic activity (Table 1). Thus, catalytically inactive prostasin fully supports embryonic development, postnatal survival and prostasin-dependent aspects of terminal epidermal differentiation and epidermal barrier formation, but is unable to mediate prostasin-dependent zymogen activation of matriptase during placental differentiation. Surprisingly, this inability to promote activation of placental matriptase only transiently extends the survival of mice lacking the cognate matriptase/prostasin inhibitor HAI-1 and shows no effect on the embryonic lethality observed in HAI-2-deficient mice. These paradoxical findings could most easily be explained by an ability of prostasin to allosterically stimulate the activity of the one-chain zymogen, in addition to the two-chain ‘active' form of matriptase (Fig. 6). This is consistent with our recent report that catalytically inactive prostasin is found in complex with a zymogen-locked matriptase in cultured epithelial cells (Friis et al., 2016).

An explanation for the varying degree of dependence of prostasin biological functions on its proteolytic activity remains to be established. It is clear that, at least in one tissue, prostasin mediates the proteolytic activation of matriptase zymogen, as evidenced by a loss of the two-chain form of the matriptase in prostasin null placentas. However, the presence of active matriptase in the epidermis of prostasin-deficient mice suggests that additional proteases are likely to contribute to matriptase zymogen activation in other tissues and/or biological processes (Friis et al., 2016; Szabo et al., 2012). As a result, matriptase zymogen activation and the full activity of the matriptase/prostasin pathway may, depending on circumstances, exhibit complete, partial or no dependence on prostasin proteolytic function. Also, various biological processes might require different levels of matriptase activity for successful completion. It is therefore plausible that although non-proteolytic functions of prostasin are sufficient to stimulate the enzymatic activity of single- or two-chain matriptase to levels that support some processes, such as epidermal development, its full activation, dependent also on prostasin proteolytic functions, is required during other processes, such as hair growth and wound closure. Alternatively, prostasin could have some proteolytic and/or non-proteolytic functions that are independent of matriptase activity, including acting as a non-enzymatic co-factor to other trypsin-like serine proteases or directly cleaving as yet unidentified downstream targets. Although we cannot formally exclude this possibility, the fact that the loss of prostasin function leads to phenotypes that are virtually identical to those observed in the absence of matriptase would suggest that the two proteases regulate developmental functions as part of a single proteolytic system.

Genetic background appears to have a considerable effect on the phenotypes associated with the loss of prostasin or matriptase, in particular during embryonic development, as documented by complete embryonic lethality in an inbred C57BL/6J background but partial or complete prenatal survival in several prostasin- or matriptase-deficient strains of mixed background (Hummler et al., 2013; Leyvraz et al., 2005; List et al., 2002; Peters et al., 2014; Szabo et al., 2009a,, 2014; R.S. and T.H.B., unpublished). This could also explain our inability to observe previously described structural defects in the development of the placental labyrinth as the cause of death in C57BL/6J Prss8^−/− embryos (Hummler et al., 2013). However, it should be noted that our current findings do not eliminate defects in placental function as the most likely reason for the lethality in mice lacking prostasin.

Probably the most surprising finding of this study is the restoration of prostasin-dependent functions and the full viability of prostasin-deficient mice when they also lack HAI-1. This strongly suggests that the regulation of processes such as epidermal barrier formation, hair growth and overall survival by prostasin is mediated by modulation of the activity of heterologous proteases, and is consistent with the proposed role of prostasin as a co-factor in the stimulation of matriptase activity (Fig. 6). Restored epidermal barrier and survival of prostasin/HAI-1 double-deficient mice can then be explained by the compensation of a decreased activity of matriptase in the absence of prostasin-mediated ‘activation' by a sufficient increase in HAI-1-free, and presumably more active, matriptase. This is supported by the observation that loss of essential functions of the matriptase/prostasin pathway in skin development, when caused by loss of matriptase rather than loss of prostasin, cannot be restored by a removal of pathway inhibition through the elimination of HAI-1 (Szabo et al., 2007), identifying matriptase as the protease that executes the function of the pathway (Fig. 6).

Loss of HAI-2 did not have a visible impact on phenotypes associated with prostasin deficiency. It is therefore likely that although both HAI-1 and HAI-2 are required for the regulation of matriptase/prostasin activity during development, their mechanisms of action differ. Thus, while it is widely accepted that HAI-1 forms inhibitory complexes with matriptase in vivo and is likely to directly regulate matriptase activity at the cell surface, HAI-2 might exert its function by primarily regulating prostasin activity (Fig. 6). This would explain why the loss of HAI-2, unlike that of HAI-1, cannot compensate for the loss of prostasin by removing inhibition of matriptase. Indeed, we have previously documented that, in a cell culture-based system, prostasin efficiently co-immunoprecipitates with HAI-2, and the effect of HAI-2 on matriptase activation and shedding in this system is largely prostasin dependent (Friis et al., 2014). Furthermore, the isolation of prostasin complexes with HAI-2 from human milk was recently reported, showing a direct interaction between the two proteins in vivo (Lai et al., 2016). Interaction of HAI-2 with catalytically inactive prostasin, as shown in our current study, suggests that HAI-2 can inhibit both proteolytic and non-proteolytic functions of prostasin and can explain why HAI-2 remains essential for development and survival in the Prss8^Ki/Ki background. Although we cannot formally exclude the possibility that, in addition to its effect on prostasin, HAI-2 can also contribute to the inhibition of other serine proteases, including matriptase, this does not appear to be crucial for mouse embryonic survival.

It is important to underscore that the current study focuses on functions of the matriptase/prostasin pathway in placental differentiation and skin development and its function in embryonic and postnatal survival. Additional studies will be needed to determine the extent to which our model applies to the reported functions of matriptase and prostasin in other tissues.

All experiments were performed in an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited vivarium following standard operating procedures and were approved by the NIDCR Institutional Animal Care and Use Committee. Matriptase-deficient (St14^−/−), prostasin-deficient (Prss8^−/−), HAI-1-deficient (Spint1^−/−) and HAI-2-deficient (Spint2^−/−), as well as prostasin knock-in mice expressing proteolytically inactive protein (Prss8^Ki/Ki) have been described previously (List et al., 2002; Peters et al., 2014; Szabo et al., 2009a,, 2007). All studies used mice of mixed 129S6/Sv;NIH BlackSwiss;FVB/NJ;C57Bl/6J genetic background and were littermate controlled. Ear or tail clips of newborn or 2-week-old mice were subjected to genomic DNA extraction and genotyped by PCR (see Table S1 for primer sequences).

Females were euthanized at designated time points by CO₂ asphyxiation and embryos were extracted by Caesarian section. Tail clips or yolk sacs of individual embryos were washed in PBS, subjected to genomic DNA extraction and genotyped by PCR (Table S1). To analyze development of the placental labyrinth, placental tissues were fixed for 24 h in 4% paraformaldehyde (PFA), processed into paraffin, sectioned, and stained with Hematoxylin and Eosin (H&E), or used for immunohistochemistry as described below. Overall thickness of the placental labyrinth was measured as the distance between the undifferentiated chorionic epithelium and the labyrinth-supporting spongiotrophoblast layer on a single midline cross-section. To evaluate branching of the fetal vasculature in the E12.5 and E13.5 placentas, a midline cross-section of each placenta was immunostained with anti-CD31 (PECAM1) antibody (see below), followed by the manual counting of individual profiles of CD31-stained vessels within the placental labyrinth. All tests were performed using tissues from at least five mice of each genotype. Only living embryos with detectable heartbeat were used for the morphometric analysis. The observed values were statistically analyzed using a two-sample, two-tailed Student's t-test.

Antigens from 5 µm paraffin sections were retrieved by incubation for 10 min at 100°C with 1 mM EDTA pH 8.0 for CD31 staining, or by incubation for 20 min at 100°C in 0.01 M sodium citrate buffer pH 6.0 for all other antigens. The sections were blocked with 2.5% bovine serum albumin (fraction V, MP Biomedicals) in PBS, and incubated with primary antibody overnight at 4°C (for detailed information on antibodies, see Table S2). Bound antibodies were visualized using biotin-conjugated secondary antibodies (Table S2) and a Vectastain ABC Kit (Vector Laboratories) using 3,3′-diaminobenzidine as the substrate (Sigma-Aldrich). All microscopy images were acquired on an Olympus BX40 microscope using an Olympus DP70 digital camera system.

Fresh tissues were homogenized and lysed in buffer containing 1% Triton X-100 and 0.5% sodium deoxycholate in PBS. The homogenates were cleared by centrifugation at 12,000 g for 10 min at 4°C and the protein concentration determined by BCA assay (Pierce). 80 µg total protein was loaded on 4-12% SDS-PAGE under reducing conditions and analyzed by western blotting, with incubation with primary antibody overnight at 4°C followed by secondary antibody conjugated to alkaline phosphatase for 1 h at room temperature (Table S2). Alkaline phosphatase activity was visualized using nitro-blue tetrazolium and 5-bromo-4-chloro-3′-indolylphosphate substrates.

Lysates from three placentae of the same genotype were combined and pre-incubated with 100 µl GammaBind G Sepharose beads (GE Healthcare Life Sciences) for 30 min at 4°C with gentle agitation. The samples were spun at 1000 g for 1 min to remove the beads, and the supernatant then incubated with 3 µg goat anti-mouse HAI-1 antibody (R&D Systems, Table S2) and 80 µl GammaBind G Sepharose beads for 3 h at 4°C. The samples were spun at 1000 g for 1 min, the supernatant removed, and the beads washed three times with 1 ml ice-cold PBS containing 1% Triton X-100 and 0.5% sodium deoxycholate. The beads were then mixed with 30 µl SDS loading buffer (Invitrogen) containing 0.25 M β-mercaptoethanol, incubated for 5 min at 99°C, and cooled on ice for 2 min. The samples were spun at 1000 g for 1 min and the released proteins then resolved by 4-12% polyacrylamide SDS-PAGE and analyzed by western blot using sheep anti-human matriptase (R&D Systems) primary antibody and peroxidase-conjugated donkey anti-sheep (Sigma-Aldrich) secondary antibodies (Table S2). The signal was visualized using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific).

The preparation of wild-type and S238A variants of human prostasin has been described (Netzel-Arnett et al., 2006; Szabo et al., 2012). To detect prostasin-HAI inhibitory complexes, PI-PLC-released pro-prostasin variants were left untreated or first activated by incubation with 10 nM human recombinant matriptase serine protease domain (R&D Systems) for 20 min at 37°C. 100 ng prostasin zymogen or matriptase-activated prostasin in 50 mM Tris-HCl pH 8.0, 100 mM NaCl buffer was then incubated with 300 ng human recombinant PN-1 or HAI-2 (both R&D Systems) for 30 min at room temperature. Reduced but non-boiled samples were analyzed by western blotting as described above (for antibodies see Table S2).

We thank Dr Mary Jo Danton for critically reviewing this manuscript.