Loss of HAI-2 in mice with decreased prostasin activity leads to an early-onset intestinal failure resembling congenital tufting enteropathy

Prostasin (CAP1/PRSS8) is a glycosylphosphatidylinositol (GPI)-anchored serine protease that is essential for epithelial development and overall survival in mice. Prostasin is regulated primarily by the transmembrane serine protease inhibitor, hepatocyte growth factor activator inhibitor (HAI)-2, and loss of HAI-2 function leads to early embryonic lethality in mice due to an unregulated prostasin activity. We have recently reported that critical in vivo functions of prostasin can be performed by proteolytically-inactive or zymogen-locked variants of the protease. Here we show that the zymogen form of prostasin does not bind to HAI-2 and, as a result, loss of HAI-2 does not affect prenatal development and survival of mice expressing only zymogen-locked variant of prostasin (Prss8 R44Q). Indeed, HAI-2-deficient mice homozygous for R44Q mutation (Spint2-/-;Prss8R44Q/R44Q) are born in the expected numbers and do not exhibit any obvious developmental abnormality at birth. However, postnatal growth in these mice is severely impaired and they all die within 4 to 7 days after birth due to a critical failure in the development of small and large intestines, characterized by a widespread villous atrophy, tufted villi, near-complete loss of mucin-producing goblet cells, loss of colonic crypt structure, and bleeding into the intestinal lumen. Intestines of Spint2-/-;Prss8R44Q/R44Q mice showed altered expression of epithelial junctional proteins, including reduced levels of EpCAM, E-cadherin, occludin, claudin-1 and -7, as well as an increased level of claudin-4, indicating that the loss of HAI-2 compromises intestinal epithelial barrier function. Our data indicate that the loss of HAI-2 in Prss8R44Q/R44Q mice leads to development of progressive intestinal failure that at both histological and molecular level bears a striking resemblance to human congenital tufting enteropathy, and may provide important clues for understanding and treating this debilitating human disease.

The activity of the matriptase-prostasin pathway during development is controlled by two transmembrane serine protease inhibitors, hepatocyte growth factor activator inhibitor (HAI)-1 and HAI-2. In mice, HAI-1 is essential for placental development and overall embryonic as well as postnatal survival [19][20][21]. Loss of HAI-2 is associated with an early embryonic lethality on or before embryonic day (E) 8.5 in mice expressing normal levels of matriptase and prostasin, and with high frequency of neural tube defects, including exencephaly, spina bifida, and curly tail, as well as a mid-gestational embryonic lethality, due to a placental failure in matriptase-heterozygous mice [22][23][24]. All of these developmental defects in HAI-1-and HAI-2-deficient mice are rescued by simultaneous inactivation of either matriptase or prostasin, thus demonstrating a critical contribution of matriptase-prostasin proteolytic pathway to the developmental defects observed in these mice [22,24]. Although HAI-1 and HAI-2 were shown to efficiently inhibit proteolytic activity of both matriptase and prostasin in vitro and in cell culture, epistatic analysis in mouse strains lacking either of the two proteases indicate that, at least during embryonic development, the two inhibitors play distinct roles, with HAI-1 acting predominantly as a direct inhibitor of matriptase, whereas HAI-2 regulates the activity of the pathway by targeting prostasin [25,26]. Phenotypes of mice with genetic modification in matriptase, prostasin, HAI-1, and HAI-2 genes and their relative interactions are summarized in S1 Table [3-6, 11, 20-25, 27-32].
Mutations in the SPINT2 gene, encoding HAI-2, have recently been described in a subset of patients with congenital tufting enteropathy (CTE) [33]. CTE presents in infants as a severe intestinal insufficiency associated with watery diarrhea, dehydration, and failure to thrive in the absence of parenteral feeding [34]. Histologically, CTE is characterized by epithelial dysplasia, varying degrees of villous atrophy and a compromised intestinal epithelial barrier. In over 70% of patients, the underlying mutation is in the EPCAM gene, encoding the epithelial cell adhesion molecule (EpCAM), a highly conserved cell surface glycoprotein involved in regulation of epithelial cell physiology [33, 35,36]. In the absence of HAI-2, matriptase has been shown to cleave EpCAM in cultured intestinal epithelial cells, causing premature degradation of the tight junction protein claudin-7. Therefore, an increase in the activity of the matriptaseprostasin pathway, leading to an excessive cleavage of EpCAM protein and destabilization of tight junctions, has been proposed as the etiology of CTE in patients with SPINT2 mutations [37].
We previously generated a knock-in mouse strain that carries a point mutation resulting in the substitution of arginine 44 in the activation cleavage site with glutamine (Prss8 R44Q ) [5,31]. The intent was to generate a prostasin variant that is "locked" in the zymogen conformation by being resistant to activation site cleavage. Compatible with achieving this aim, R44Q prostasin in tissues from Prss8 R44Q/R44Q mice displayed a mobility in SDS/PAGE that was similar to the zymogen form of prostasin [5,31]. Furthermore, the phenotype of Prss8 R44Q/R44Q mice was identical to a knock-in mouse strain that carries a point mutation resulting in the substitution of the catalytic serine 238 with alanine (Prss8 S238A ) [5,31], indicating that R44Q prostasin does not attain full biological activity. Nevertheless, it cannot be excluded that R44Q prostasin in some tissues may undergo activation site cleavage by certain proteases with chymotryptic specificity [38]. Surprisingly, the phenotypes of Prss8 R44Q/R44Q or Prss8 S238A/S238A mice do not match those of prostasin null mice (Prss8 -/-). Thus, both Prss8 R44Q/R44Q and Prss8 S238A/S238A mice are fully viable and only exhibit a mild defect in skin and hair development [5,25,31], indicating that critical prostasin biological functions are independent of its proteolytic activity.
In this study, we show that, unlike the wildtype and proteolytically-inactive prostasin, the zymogen-locked (R44Q) variant of prostasin is not an effective target for inhibition by HAI-2. As a result, loss of HAI-2 does not affect embryonic development and prenatal survival of Prss8 R44Q/R44Q mice. However, loss of HAI-2 leads to an inability to gain weight and death within 4 to 7 days after birth. Spint2 -/-;Prss8 R44Q/R44Q pups present with distended colon and highly abnormal structure of epithelial compartments of both small and large intestine. Furthermore, the mice exhibit a progressive loss of EpCAM, E-cadherin, and claudin-7, indicating that the phenotype of Spint2 -/-; Prss8 R44Q/R44Q mice recapitulates early events in the development of CTE in humans and provides an excellent model for study of the etiology and treatment of this debilitating disease.

Mouse strains
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. HAI-2-deficient (Spint2 -/-) and knockin mice expressing zymogen-locked prostasin (Prss8 R44Q/R44Q ) have been described in detail previously [24,31]. 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 to two-week-old mice were subjected to genomic DNA extraction and genotyped by PCR as described elsewhere [3,24,31].

Immunohistochemistry
Embryonic day (E) 18.5 to post-natal day (P) 4 old mice were euthanized, intestinal tissues extracted and immediately fixed in aqueous-buffered zinc formalin fixative (Z-Fix, Anatech Ltd., Battle Creek, MI) for 24 hours at room temperature and paraffin embedded (Histoserv Inc., Germantown, MD). 5 μm paraffin sections were stained for hematoxylin&eosin (H&E), alcian blue/PAS, and terminal deoxynucleotidyl transferase dUTP Nick-End Labeling (TUNEL) (all performed by Histoserv Inc.). Alternatively, the sections were immunostained after antigen retrieval by incubation for 20 min at 100˚C in 0.01 M sodium citrate buffer, pH 6.0 essentially as described previously [25] (see S2 Table for information on antibodies). Three to five mice per each genotype and time point were analyzed. Number of proliferating and apoptotic cells was determined by manually counting Ki67-positive, and cleaved caspase-3-and TUNEL-positive cells, respectively, in five non-overlapping fields of small intestine in each sample.

Protein extraction and Western blot analysis of mouse intestinal tissues
Small and large intestines were collected from E18.5, P2, and P4 mice, snap-frozen in liquid nitrogen, and stored at -80˚C until further use. For Western blot analysis, the tissues were homogenized in buffer containing 2% SDS and 10% glycerol in 62.5 mM Tris/Cl pH 6.8. The lysates were cleared by centrifugation at 20,000 g for 10 min at 4˚C to remove the tissue debris and the protein concentration in supernatant was determined by BCA assay (Pierce, Rockford, IL). 80 μg of total protein was loaded on 4-12% reducing SDS-PAGE and analyzed by Western blotting, incubating with primary antibody overnight at 4˚C, followed by incubation with secondary antibody conjugated to alkaline phosphatase for 1.5 h at room temperature (see S2  Table for information on antibodies). Alkaline phosphatase activity was visualized using nitroblue tetrazolium and 5-bromo-4-chloro-3'-indolylphosphate substrates (Sigma-Aldrich, St. Louis, MO). Data shown are representative of at least two independent Western blot experiments run on tissue lysates from three separate mice per each genotype and time point.

Formation of prostasin inhibitory complexes in vitro
The preparation of wildtype, S238A, and R44Q mutated variants of human prostasin has been described in detail before [22,39,40]. Recombinant human HAI-1, HAI-2, and serpin E2/PN-1 were purchased from R&D Systems (Minneapolis, MN). To detect prostasin/HAI and prostasin/PN-1 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 minutes at 37˚C. 100 ng of prostasin zymogen or matriptaseactivated prostasin in 50 mM Tris/HCl, pH 8.0, 100 mM NaCl buffer was then incubated with 200 ng of human recombinant PN-1, HAI-1, or HAI-2 (all R&D Systems) for 30 min at room temperature. Reduced/non-boiled samples were analyzed by Western blotting as described above. Presented data are representative of two independent experiments.

Analysis of HAI-2/prostasin interaction in mouse embryonic tissues
Protein extraction from the embryonic portion of mouse placenta was performed as described in detail in [22]. Combined lysates from three placentae of the same genotype, corresponding to 5 mg of total protein as determined by BCA assay (Pierce, Thermo Scientific, Rockford, IL), were diluted to 5ug/ul in 50 mM Tris/HCl, pH 8.0; 1% NP-40; 500 mM NaCl and pre-incubated with 50 μl GammaBind G Sepharose beads (GE Healthcare Bio-Sciences, Uppsala, Sweden) for 30 minutes at 4 o C with gentle agitation. The samples were spun at 1,000 g for 1 min to remove the beads, and the supernatant was then incubated with 3 μg goat anti-mouse HAI-2 or goat anti-mouse prostasin antibody (both R&D Systems) and 50 ul of GammaBind G Sepharose beads for 3 hours at 4 o C. The samples were spun at 1,000 g for 1 min, the supernatant was removed, and the beads were washed 3 times with 1 ml ice-cold 50 mM Tris/HCl, pH 8.0; 1% NP-40; 500 mM NaCl buffer. The beads were then mixed with 40 ul of 1x SDS loading buffer (Invitrogen, Carlsbad, CA) with 0.25 M β-mercaptoethanol, incubated for 5 min at 99 o C, and cooled on ice for 2 minutes. Western blot analysis using mouse anti-human prostasin (1:250, BD Transduction Labs) or goat anti-mouse HAI-2 (1:500, R&D Systems) primary antibodies was performed as described above. Data shown are representative of two independent Western blot experiments run on tissue lysates from three separate placentae per each genotype. Western blot analysis of the starting material shown in S1 Fig was performed using 60 ug of tissue lysate.

RNA preparation and quantitative RT-PCR
Small and large intestines collected postnatal day (P)2 mice were homogenized in Trizol reagent (Life Technologies, Grand Island, NY) and total RNA was extracted according to the manufacturer's instructions. 1 ug of RNA was reverse transcribed with oligo dT primer using RETROscript kit (Invitrogen, Carlsbad, CA). Real-time PCR was conducted on 0.5 ul of cDNA template using iQ SYBR Green Supermix (Bio-Rad Laboratories, Hercules, CA) and 7500 Real-Time PCR System with 7500 Software v2.3 (Applied Biosystems, Foster City, CA). Primer sets used for cDNA amplification are listed in S3 Table. The expression of each gene was normalized to expression of ribosomal protein S15 with the ΔCt method. The assay was performed in triplicate. Results shown represent means and standard deviations from three independent samples per genotype.

Enzymatic de-glycosylation assay
Wildtype recombinant human prostasin zymogen was prepared and activated by matriptase as described above. 100 ng of zymogen or of activated, double-chain prostasin was then subjected to removal of N-and O-linked carbohydrates using Enzymatic Protein Deglycosylation Kit (EDEGLY, Sigma-Aldrich, St. Louis, MO) according to manufacturer's instructions. Resulting products were then analyzed by reducing SDS-PAGE and anti-prostasin Western blot analysis as indicated above.
Initial body weight at birth (N ! 13 for each genotype) and the length of small and large intestines (N ! 5 for each genotype and time point), as well as the immunohistological, Western blot, and qPCR analysis were statistically evaluated using a two-sample Student's t-test, two-tailed.
Number of proliferating cells was determined by manually counting Ki67-positive cells in five non-overlapping fields of small intestine in each sample and the observed values were statistically evaluated using a two-sample Student's t-test, two-tailed.
Western blot protein signal quantification was performed using ImageJ 1.46r software and statistically analyzed using a two-sample Student's t-test, two-tailed.

Binding of HAI-1 and HAI-2 to prostasin requires zymogen conversion but not proteolytic activity of the protease
We have previously reported that the developmental abnormalities and pre-natal lethality associated with embryonic loss of HAI-2 are averted by genetic inactivation of either matriptase or prostasin, thus identifying the matriptase-prostasin pathway as the only critical target of HAI-2 in pre-natal development [22,24]. Specifically, as the phenotypes associated with prostasin deficiency are largely unaffected by loss of HAI-2 expression, we hypothesized that prostasin, rather than matriptase, serves as the primary target for HAI-2 inhibition [25]. Interestingly, most of prostasin's physiological functions do not appear to require prostasin proteolytic activity or the proteolytic conversion of prostasin zymogen into its catalytically active double-chain form [5,25,31]. In order to further characterize the interaction of prostasin with HAI-2, we first tested the ability of different variants of prostasin to form inhibitory complexes with HAI-2, its closest homolog HAI-1, and a previously reported cognate serpin-type inhibitor of prostasin, protease nexin (PN)-1 [41,42]. Thus, recombinant wildtype, S238A, and R44Q prostasin variants in their native zymogen form or converted into the double-chain forms by matriptase were incubated with the three inhibitors, followed by detection of inhibitory complexes by Western blot using prostasin antibody. As previously reported, wildtype double-chain (dc)-prostasin readily formed SDS-stable complexes with all three inhibitors ( Fig  1A). Binding to HAI-2 appeared to be particularly effective, with virtually all of the detectable prostasin protein sequestered into the complex (Fig 1A, lane 6). Loss of proteolytic activity, as expected, prevented binding of S238A prostasin to the serpin-type inhibitor PN-1, but did not significantly affect binding to either of the HAI proteins ( Fig 1B, lanes 4, 6, and 8). Small amounts of complexes were also detected with unactivated wildtype and the S238A prostasin ( Fig 1A and 1B, lanes 3, and 5), most likely due to a low-level conversion of prostasin zymogen in the absence of matriptase, as evidenced by a presence of a prostasin species with an apparent molecular weight corresponding to activated dc-prostasin in prostasin zymogen sample after de-glycosylation (Fig 1A, lanes 11 and 12). This is consistent with a near-complete inability of R44Q prostasin variant that is unable to undergo proteolytic cleavage by matriptase (and referred to herein as zymogen-locked prostasin) to form inhibitory complexes with HAI-1, HAI-2, or PN-1 (Fig 1C).
HAI-2 is essential for inhibition of prostasin activity during mouse embryogenesis [22]. In order to test whether conversion into the double-chain form is indeed necessary for prostasin to become a target for inhibition by HAI-2 in vivo, we next performed a co-immunoprecipitation assay to analyze the interaction between the two proteins in embryonic tissues extracted from mice expressing wildtype (control) or zymogen-locked (R44Q) variants of prostasin (for analysis of starting material see S1 Fig). Consistent with its role as HAI-2 target, prostasin was readily detectable in placental extracts from control mice after immunoprecipitation with an anti-HAI-2 antibody (Fig 1D, top panel, lane 1). In contrast, no prostasin co-immunoprecipitated with HAI-2 in tissues that only express R44Q prostasin ( Fig 1D, top panel, lane 2), despite expressing prostasin protein at levels comparable to their littermate controls ( Fig 1D, top panel, compare lanes 4 and 5). Similarly, although placental tissues from mice expressing wildtype and R44Q prostasin express comparable levels of HAI-2 protein (Fig 1D, bottom panel, compare lanes 1 and 2), the amount of the inhibitor that was detected after immunoprecipitation with an anti-prostasin antibody was dramatically reduced in tissues expressing only R44Q prostasin (Fig 1D, bottom panel, compare lanes 4 and 5). Thus, our data indicate that zymogen-locked prostasin is unable to efficiently form stable complexes with HAI-2 in vitro or in vivo and is therefore not likely to be a target for HAI-2 inhibition.

HAI-2-deficient Prss8 R44Q mice suffer from postnatal growth retardation and loss of viability due to intestinal failure
Although HAI-2-deficient Prss8 R44Q/R44Q mice were born in the expected numbers and were macroscopically otherwise unremarkable, they typically presented with about 20% reduction in body weight at birth compared to control and HAI-2-expressing Prss8 R44Q/R44Q littermates (Fig 2A, P<0.0001, Student's t-test, 2 tailed). Furthermore, the postnatal growth of Spint2 −/− ; Prss8 R44Q/R44Q mice was severely impeded, despite presenting with a milk spot indicating ability to ingest food they consistently failed to gain any weight and all died 4 to 7 days after birth (Fig 2B-2D). Macroscopic inspection of the outward appearance of Spint2 −/− ;Prss8 R44Q/R44Q pups prior to their demise did not reveal any gross developmental abnormality and thus failed to provide an explanation for the growth retardation and early postnatal lethality (Fig 2B). Similarly, upon detailed macroscopic and histological examination, most major internal organs previously reported to express HAI-2, including brain, lungs, kidney and stomach, extracted from 4 days old Spint2 −/− ;Prss8 R44Q/R44Q mice appeared unremarkable. Loss of HAI-2 did, however, lead to a severe defect in the development of the lower gastrointestinal tract. The overall appearance of small and large intestines from Spint2 −/− ;Prss8 R44Q/R44Q mice before birth (embryonic day 18.5) was largely comparable with that of control or HAI-2-expressing Prss8 R44Q/R44Q littermate controls (Fig 2E-2E"). However, as early as on postnatal day 2, intestines from HAI-2-deficient mice appeared noticeably distended and the relative length of the large intestine was significantly reduced (Fig 2F-2F" and 2H). On postnatal day 4, intestines of most HAI-2-deficient mice were filled with dark material that contained high number of red blood cells, indicating bleeding into the lumen (Fig 2G-2G", 2I and 2I'). No obvious abnormality was observed in the intestines from HAI-2-expressing Prss8 R44Q/R44Q mice at any point.
Histological analysis of intestinal tissues from Spint2 −/− ;Prss8 R44Q/R44Q mice revealed no obvious abnormalities in the villous structure of small intestines at embryonic day 18.5 ( Fig 3A  and 3B). However, already at this time these mice presented with general disorganization of intestinal epithelium within crypts of the large intestine, associated with an increased shedding of cellular material into the lumen, decreased number of mucin-producing goblet cells, and lack of well-organized crypt structure present in HAI-2-expressing littermate control mice ( Fig 3A'and 3B'). Analysis of intestinal tissues after birth revealed progressive changes in epithelia of both small and large intestine. On postnatal day 2, the small intestine showed signs of villous atrophy, increased dyslocalization of nuclei within the epithelial layer indicating loss of epithelial cell polarity, along with a substantial number of epithelial cells containing very large vacuoles and (Fig 3C and 3D). In the large intestine, this was associated with a disorganization of surface epitheliumand overall loss of crypt structure (Fig 3C' and 3D'). The severity of the phenotypes increased over time, with widespread villous atrophy, abnormal invaginations between the enterocytes, tufted villi, enterocyte crowding, and essential loss of normal tissue architecture of both small and large intestines all manifesting by postnatal day 4 (Fig 3E-3F'). Abnormal differentiation of intestinal epithelium also manifested by, respectively, a significant reduction and a near complete absence of mucin-producing goblet cells in small and large intestines at P2 and P4 (Fig 3G-3I). Intestines from HAI-2-expressing Prss8 R44Q/R44Q mice did not exhibit any obvious histological abnormality at any point.
Immunohistochemical analysis of the expression of the cell proliferation marker Ki67 showed that loss of HAI-2 did not affect epithelial cell proliferation during pre-natal stages of development, as documented by a comparable number of Ki67-positive epithelial cells in control and Spint2 −/− ;Prss8 R44Q/R44Q mice at E18.5 (Fig 4A-4B' and 4G). However, analysis of the postnatal tissues revealed a strong reduction in the proliferation rates of intestinal epithelial cells in Spint2 −/− ;Prss8 R44Q/R44Q mice after birth, with approximately 70 and 80% decrease in the number of Ki67-positive epithelial cells in both small and large intestines at postnatal day 2 and 4, respectively (Fig 4C-4G). On the other hand, despite widespread villous atrophy, substantial shedding of epithelial cells, and a general disorganization of intestinal epithelium, no apoptotic cells were detected by TUNEL or anti-p-caspase-3 stain in the intestinal tissues from two-or four-days-old Spint2 −/− ;Prss8 R44Q/R44Q mice, as well as from their control (Spint2 + ; Prss8 + ) and HAI-2-expressing prostasin zymogen-locked (Spint2 + ;Prss8 R44Q/R44Q ) littermates, suggesting that epithelial demise in these animals is not associated with the activation of an apoptotic program. No histological abnormalities were detected in intestines from HAI-2-expressing Prss8 R44Q/R44Q mice at any stage, suggesting that the observed changes result from the loss of HAI-2 expression rather than from the inability of R44Q prostasin to be converted into its double-chain form. Taken together, our histopathological analysis indicates that loss of HAI-2 in Spint2 −/− ;Prss8 R44Q/R44Q mice leads to severe defects in intestinal epithelial integrity consistent with those observed in CTE patients.

Loss of HAI-2 function leads to altered expression of epithelial junctional proteins
Improper intestinal barrier function has been implicated in the etiology of CTE [43]. Furthermore, unregulated matriptase activity leading to increased turnover of the epithelial junctional proteins EpCAM and claudin-7 has recently been implicated in the etiology of intestinal failure in a subset of congenital tufting enteropathy (CTE) patients carrying mutation in the SPINT2 gene encoding HAI-2 [37]. As the developmental defects observed in HAI-2-deficient mice appear to closely mimic clinical features of CTE, we next analyzed expression of main components of epithelial tight and adherens junctions in intestines from Spint2 −/− ;Prss8 R44Q/R44Q mice and their healthy littermate controls at different developmental time points. At E18.5, before the mice are born and begin to ingest food, expression of most epithelial junctional proteins, including claudin-2, 4, 6, and 7, E-cadherin, and occludin appeared to be unaffected in the intestines from Spint2 −/− ;Prss8 R44Q/R44Q mice (Fig 5A). Loss of HAI-2 did, however, lead to a 56% and 37% reduction in the expression of EpCAM and claudin-1, respectively (Fig 5A and  HAI-2 maintains intestinal epithelial integrity 5D). The levels of both proteins continued to decline after birth, with only 18% and 5% of EpCAM and 49% and 25% of claudin-1 protein detected at postnatal day 2 and 4, respectively, compared to HAI-2-expressing littermate controls (Fig 5B-5D). This was followed by a reduced expression of E-cadherin (50% on P2, 55% on P4), claudin-7 (60% on P2, 44% on P4) and occludin (unchanged on P2, 54% on P4) (Fig 5C and 5D). On the other hand, expression of claudin-2 was not affected during the progression of the disease, whereas the expression of claudin-4 gradually increased in HAI-2-deficient intestines compared to Spint2 + ;Prss8 + littermate controls (185% on P2, 248% on P4) (Fig 5A-5D). Further analysis of the gene expression showed no significant changes in mRNA levels in P2 intestinal tissues for any of the proteins down-regulated in the absence of HAI-2 ( Fig 5E). Interestingly, however, mRNA for the sole junctional protein found upregulated in Spint2 −/− ;Prss8 R44Q/R44Q tissues, claudin-4, was increased 2.75-fold, compared to Spint2 + ;Prss8 + littermate controls (Fig 5E).
These data suggest that the loss of HAI-2 leads to a specific sequence of molecular events that are expected to affect integrity of epithelial cell-cell junctions and intestinal barrier function.

Discussion
Recent genome editing studies in mice unexpectedly proposed that key developmental and postnatal functions of the membrane-anchored serine protease, prostasin, can be executed by prostasin variants that either are locked in the one-chain zymogen conformation or can undergo conversion to the active two-chain form, but are rendered catalytically-inactive by mutation of the active site serine residue [5,31]. The current study, when combined with these and other previous studies [22,24], supports the notion that the one-chain prostasin zymogen is refractory to HAI-2 regulation, while the two-chain prostasin is susceptible to HAI-2 regulation and is an essential developmental inhibitory target for the protease inhibitor. Although rendering prostasin zymogen-locked allowed HAI-2-deficient embryos to successfully pass previously established HAI-2-dependent early and mid-term developmental milestones, these mice all developed a congenital tufting enteropathy (CTE)-like phenotype leading to an inability to gain weight after birth and an early postnatal lethality. Furthermore, we found that reduction of EpCAM protein expression was an early event in the HAI-2-deficient intestine, and that this was followed by a progressive loss of claudin-7, claudin-1 and E-cadherin, but an increase in gene and protein expression of claudin-4. Given the essential role of junctional proteins in the development and maintenance of epithelial function [44][45][46], it is reasonable to propose that these molecular changes are, at least in part, responsible for the intestinal demise observed in Spint2 −/− ;Prss8 R44Q/R44Q mice.
Our analysis, thus, lends critical in vivo support to the central hypothesis generated from a recent cell-based study addressing the molecular consequences of HAI-2-insufficiency on intestinal epithelium. This study proposed EpCAM as a critical pathogenic substrate for   While our study does not directly address the identity of the protease or proteases responsible for the intestinal abnormalities observed in Spint2 −/− ;Prss8 R44Q/R44Q mice, it is reasonable to assume that, again, matriptase activity may play an important role. In addition to prostasin, matriptase has previously been identified as a very efficient target for HAI-2 inhibition in vitro [47]. As mentioned above, early loss of EpCAM protein, followed by a decrease in claudin-7 expression in Spint2 −/− ;Prss8 R44Q/R44Q mice mimic closely molecular changes induced by unregulated matriptase activity in HAI-2-silenced intestinal cell monolayers [37]. Furthermore, as HAI-2 does not appear to efficiently target prostasin in Prss8 R44Q/R44Q tissues, we do not expect that the loss of the inhibitor would directly trigger critical mis-regulation of prostasin activity. Analysis of the contribution of matriptase to the development of CTE-like phenotype in HAI-2-deficient mice is therefore of paramount importance and should be addressed in future studies.
Previous studies have revealed that insufficient intestinal matriptase proteolytic activity, like excessive activity implied for CTE patients, results in an increased paracellular permeability of intestinal epithelium followed by an epithelial demise [8,28]. The superficially similar effects of blunted and excessive matriptase activity on intestinal epithelium and intestinal function raises the obvious possibility of a common etiology; that is the perturbation of EpCAMdependent tight junction formation. In its most attractive scenario, matriptase would function as a processing protein in intestinal epithelium that converts EpCAM to its biologically active tight junction-inducing form, with HAI-2 serving to prevent excessive or aberrant proteolysis of EpCAM. Although this hypothesis is attractive and warrants further investigation, some observations indicate that this scenario may be too simplistic: First, we have failed to detect a change in EpCAM protein expression in mice with intestinal ablation of matriptase or prostasin (S1 Fig). Second, expression of claudin-2, a key regulatory target for matriptase in the normal intestine that accumulates under conditions of matriptase deficiency [28], was unaffected by the loss of HAI-2. Altogether, these findings indicate that the effects of the matriptase-prostasin system on epithelial cell-cell adhesion are complex and dynamic, and that, importantly, the target substrates for the system may differ under normal physiological conditions, and under conditions where this proteolytic system is dysregulated, such as complete or partial deficiency for HAI-2.
HAI-2 deficiency induced the loss of intestinal EpCAM already at late stages of embryonic development, showing that no environmental trigger, such as microbial colonization or food exposure, is required for initiation of the chain of molecular events leading to CTE. This suggests that strategies aimed at reshaping the microbiome to control intestinal permeability in individuals with CTE caused by HAI-2 deficiency are likely to be nonproductive. Rather, efforts should be focused on controlling the aberrant activity of the matriptase-prostasin axis. An obvious and attractive strategy would entail the oral administration of non-absorbable small molecule trypsin-like serine protease inhibitors with high activity towards matriptase and/or prostasin, and, correspondingly, a low activity towards trypsin and chymotrypsin. This approach may prove particularly useful for treatment of human patients, as matriptase deficiency in humans, unlike other species examined, has not been reported to be associated with overt gastrointestinal abnormalities [16][17][18].
In conclusion, we have shown that mice carrying a zymogen-locked version of prostasin are permissive for development in the absence of HAI-2, but that these HAI-2-deficient mice develop a CTE-like syndrome, possibly initiated by excessive proteolytic cleavage of EpCAM. Furthermore, our study provides an excellent animal model for the development of treatment strategies for this debilitating congenital disease.