Oncogene (2007) 26, 1546–1556 & 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc ORIGINAL ARTICLE Matriptase inhibition by hepatocyte growth factor activator inhibitor-1 is essential for placental development

R Szabo, A Molinolo, K List and TH Bugge

Oral and Pharyngeal Cancer Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA

Hepatocyte growth factor activator inhibitor-1 (HAI-1) is finely tuned balance between extracellular proteases and a Kunitz-type transmembrane serine protease inhibitor their cognate inhibitors. Likely reflecting an extensive that forms inhibitor complexes with several trypsin-like developmental redundancy, the genetic ablation of most serine proteases andis requiredfor mouse placental extracellular protease inhibitors, however, either has no development and embryo survival. Here we show that the effect on organ formation (see Supplementary Material essential function of HAI-1 in placentation andall other for references) or affects only the vascularization or embryonic processes is to restrict the activity of the type II homeostasis of formed organs (Gailani et al., 1997; transmembrane serine protease, matriptase. Enzymatic Huang et al., 1997; Ishiguro et al., 2000; Oh et al., 2001; gene trapping of matriptase combinedwith HAI-1 Zeeuwen et al., 2002; Ohmuraya et al., 2005). Three immunohistochemistry revealedthat matriptase is co- notable exceptions to this are the serine protease expressedwith HAI-1 in both extraembryonic and inhibitors maspin, which is required for embryo embryonic tissues. As early as embryonic day 8.5, implantation, and hepatocyte growth factor activator matriptase andHAI-1 were expressedin a population of inhibitor (HAI)-1 and HAI-2, which are essential for, chorionic trophoblasts. Ablation of HAI-1 disrupted the respectively, placental development and neural tube epithelial integrity of this cell population, causing formation (Shimomura et al., 1997; Gao et al., 2004; disorganized laminin deposition and altered expression Tanaka et al., 2005). of E-cadherin and b-catenin. This ledto a complete loss of HAI-1, the cognate inhibitor of hepatocyte growth undifferentiated chorionic trophoblasts after embryonic factor activator, is a type I transmembrane protein day 9.5 and prevented the formation of the placental that contains two extracellular Kunitz-type domains labyrinth. Genetic ablation of matriptase activity in HAI- (Shimomura et al., 1997). Kunitz-type domains are 1-deficient embryos, however, restored the integrity of serine protease inhibitor domains that bind to the active chorionic trophoblasts andenabledplacental labyrinth site of trypsin-like serine proteases to form stable formation anddevelopment to term. Furthermore, ma- inhibitor complexes to restrict the proteolytic activity triptase/HAI-1 double-deficient mice were phenotypically of their target proteases (Ascenzi et al., 2003). Although indistinguishable from matriptase single-deficient litter- originally isolated as an inhibitor of hepatocyte growth mates. factor activator, HAI-1 also has been shown to display Oncogene (2007) 26, 1546–1556. doi:10.1038/sj.onc.1209966; potent inhibitory activity towards the membrane-bound published online 18 September 2006 trypsin-like serine proteases, hepsin, prostasin, and matriptase (Lin et al., 1999; Fan et al., 2005; Herter Keywords: placenta development; labyrinth formation; et al., 2005). membrane proteolysis; trophoblast differentiation; base- HAI-1-deficient mouse embryos display placental ment membrane; protease inhibition failure and mid-gestation resorption (Tanaka et al., 2005), suggesting that the regulation of the activity of one or more extracellular proteases by HAI-1 could be essential for the development of this critical extraem- bryonic tissue. A proteolysis-independent role of HAI-1 Introduction in embryonic development has also been proposed (Tanaka et al., 2005), because HAI-1 is an integral Cellular migration and differentiation during vertebrate transmembrane protein with a cytoplasmic tail domain morphogenesis have been hypothesized to depend on a that is potentially capable of transducing extracellular signals, and because hepatocyte growth factor activator, Correspondence: Dr TH Bugge, Proteases and Tissue Remodeling hepsin, and matriptase are all dispensable for mouse Unit, Oral and Pharyngeal Cancer Branch, National Institute of embryonic development (Wu et al., 1998; List et al., Dental and Craniofacial Research, National Institutes of Health, 30 2002; Itoh et al., 2004). Convent Drive, Room 211, Bethesda, MD 20892, USA. E-mail: [email protected] Matriptase is a type II transmembrane serine protease Received 27 April 2006; revised 3 August 2006; accepted 4 August 2006; with a strictly epithelial pattern of expression in adult published online 18 September 2006 mouse and human tissues (Oberst et al., 2003a; List Membrane protease inhibition in embryonic development R Szabo et al 1547 et al., 2006). Gene targeting studies in mice have this population during the formation of the labyrinth revealed an essential role of the protease in postnatal layer at E9.5 and E10.5 (Figure 1h and i). Matriptase oral and epidermal barrier formation and hair follicle was not expressed in other trophoblast populations, in growth (List et al., 2002, 2006). However, matriptase- the yolk sac or in the placental mesenchyme (Figure 1h deficient mice develop to term, and they display no and i and data not shown). When human placenta was obvious developmental abnormalities besides the skin examined at week 13 of gestation, HAI-1 (Figure 1k) and oral cavity. and matriptase (Figure 1l) were both expressed in villous We have recently shown that perturbing the balance trophoblasts that represent the epithelial lining of the between matriptase and HAI-1 in adult epidermis cause human placental villous trees. The population of HAI-1- severe hyperplasia, abnormal cellular differentiation and expressing cells in human placenta at week 13 or at term malignant transformation (List et al., 2005). Although was identified as villous cytotrophoblasts, rather than matriptase is dispensable for embryonic development, terminally differentiated syncytiotrophoblasts, by the this observation prompted us to examine if excess of complementary pattern of expression with a syncytio- matriptase proteolytic activity, nevertheless, could con- trophoblast marker human chorionic gonadotropin tribute to the developmental failure of HAI-1-deficient (Supplementary Figure 1). embryos. Interestingly, we found that inhibition of A tight co-expression of matriptase with HAI-1 matriptase activity is the single non-redundant function extended also to the embryo proper. Both proteins were of HAI-1 in placental development and other develop- detected in the epithelial lining of several tissues, mental processes. Thus, while loss of HAI-1 lead to the including the olfactory placode, the entrance to disruption of the epithelial integrity of a specific Rathke’s pouch, the mandibular component of the first population of matriptase-expressing chorionic tropho- branchial arch, the pharyngeal region of the foregut, the blasts and their inability to form the placental labyrinth, midgut/hindgut region, the otic pit/otic vesicle and the matriptase/HAI-1 double-deficient mice displayed nor- periderm and germinal epithelium of the skin (Supple- mal placentation, developed to term and displayed no mentary Figure 2). phenotypic abnormalities besides those associated with the absence of matriptase. HAI-1 ablation disrupts labyrinth layer morphogenesis To further analyse the functional relationship between matriptase and HAI-1 during embryonic development, Results we next generated HAI-1-deficient mice by deleting the coding region of the HAI-1 gene (Figure 2a and b), Matriptase is co-expressed with HAI-1 during placenta which lead to the loss of HAI-1 mRNA (Figure 2c) and and embryo development HAI-1 protein in embryonic tissues (Figure 2d and data To determine the expression of HAI-1 during embryonic not shown). HAI-1-deficient embryos developed pro- development, we performed an immunohistochemical perly until E9.5, but thereafter displayed severe growth staining of staged mouse embryos (Figure 1 and retardation (Figure 2e). This lead to the resorption of all Supplementary Figure 2). Weak HAI-1 expression was HAI-1-deficient embryos by E12.5, as revealed by detectable as early as embryonic day (E)8.5 in tropho- genotype analysis of staged embryos from HAI-1 blasts of the chorion (Figure 1a). At E9.5 and E10.5, the heterozygous breeding pairs (Figure 2f). As reported expression of HAI-1 extended to include both chorionic previously (Tanaka et al., 2005), the placental labyrinth trophoblasts and more differentiated trophoblasts in the of HAI-1-deficient embryos was absent (Figure 2g and labyrinth layer, identified as syncytiotrophoblasts based g0), suggesting that the demise of embryos was due to the on their proximity to fetal vessels rather than maternal failure to establish the fetal–maternal interface required blood pools (Figure 1b–e). At all stages investigated to sustain the rapid growth of the embryo that is taking (E8.0–11.5), HAI-1 expression was also detected in the place at this stage. cuboidal endoderm of the visceral yolk sac (Figure 1f Placental labyrinth formation is initiated at E9.5 and data not shown). No expression was observed in by the differentiation of chorionic trophoblasts and their the placental mesenchyme, in the fetal endothelial cells migration towards the spongial layer. This is followed or in other sub-populations of trophoblasts, includ- closely by the ingrowth of embryonic vessels ing spongiotrophoblasts and trophoblast giant cells that thereby are brought into close contact with (Figure 1a–c). maternal blood pools to form the placental labyrinth, To analyse the potential co-expression of matriptase as depicted schematically in Figure 3a (Kaufmann and with HAI-1 in these extraembryonic tissues, we took Bart, 1999). As syncytiotrophoblasts of the labyrinth advantage of a recently generated knock-in mouse that and their chorionic precursors were the only cells expresses matriptase-b-galactosidase fusion protein that expressed HAI-1 (see above), we performed a from the endogenous matriptase promoter and can be detailed analysis of chorionic trophoblast population in used as a sensitive marker for matriptase expression HAI-1-deficient and -expressing placentas. Strikingly, using 5-bromo-4-chloro-3-indolyl-beta-D-galactopyra- the loss of HAI-1 resulted in the complete failure noside (X-gal) staining (List et al., 2006). Weak of chorionic trophoblasts to form the labyrinth layer matriptase expression was first detected in chorionic (Figure 3). Thus, whereas chorionic trophoblasts trophoblasts at E8.5 (Figure 1g) and was sustained in in HAI-1-expressing embryos clearly displayed

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1548

Figure 1 Matriptase colocalizes with HAI-1 in chorionic trophoblasts during placental development. (a) Expression of HAI-1 at E8.5 in the chorionic trophoblasts (brown staining, examples with arrows) prior to the initiation of placental labyrinth formation. (b–e) Expression of HAI-1 in both chorionic trophoblasts (arrows) and differentiated syncytiotrophoblasts (arrowheads) at E9.5 (b, d and e) and E10.5 (c), and absence of HAI-1 in placental mesenchyme (‘M’ in (b) and (c)), cytotrophoblasts (surrounding maternal blood lacunae; curved arrows in (c)), spongiotrophoblasts (stars in (b)) and trophoblast giant cells (open arrowheads in (b)). (d and e) High- magnification micrographs showing the expression of HAI-1 in undifferentiated chorionic trophoblasts (arrows in (d)) and in syncytiotrophoblasts (arrowheads in (e)) but not in the cytotrophoblasts of the labyrinth layer (arrows in (e)). (f) Expression of HAI-1 in the cuboidal endoderm of the yolk sac (arrowheads) at E8.5. (g–j) Matriptase expression in extraembryonic tissues from mouse embryos. (g) At E8.5, weak matriptase expression (cyan staining, arrows) is restricted to the chorionic trophoblasts. The selective expression of matriptase in chorionic trophoblasts (arrows in (h) and (i)), but not in differentiated trophoblasts of the labyrinth, spongiotrophoblasts, trophoblast giant cells or emerging fetal vessels (arrowheads) at E9.5 (h) and E10.5 (i). High magnification (j)ofa section double-stained with HAI-1 (brown) and X-gal (cyan) showing colocalization of matriptase and HAI-1 in chorionic trophoblasts (arrows). Expression of HAI-1 (k) and matriptase (l) in villous cytotrophoblasts of week 13 human placenta (arrows), but not in villous mesenchyme (‘M’ in (k) and (l)). A layer of HAI-1-negative syncytiotrophoblasts is indicated by arrowheads in (k). Sections in (a)–(f) and (j)–(l) were counterstained with hematoxylin, and sections in (g)–(i) with nuclear fast red. Size bars: a and g, 50 mm; b, c, h and i, 100 mm; d–f,15mm; and j–l 25 mm.

ingrowth towards the spongial layer at E9.5 (Figure 3b), placentas all showed that the chorionic trophoblast no such process was observed in the HAI-1-deficient layer was missing. This difference cannot be attributed placentas (Figure 3b0). Furthermore, in HAI-1- to the diminished ability of chorionic trophoblast cells expressing placentas, the chorionic trophoblasts were to proliferate, since at E9.0, the last time point detectable as a keratin- and p57/Kip2-negative (marker when these cells are still detected in HAI-1-deficient for terminally differentiated trophoblasts; Riley et al., tissue, immunostaining revealed high proportion of 1998) sheath of cells in contact with the fetal mesen- Ki67-positive cells in both HAI-1-expressing and HAI- chyme during the labyrinth formation (Figure 3c–e). 1-deficient chorionic cells (Supplementary Figure 3 Strikingly, hematoxylin and eosin (H&E) staining and data not shown). No gross abnormalities were (Figure 3b0), keratin staining (Figure 3c0 and d0)and detected in trophoblast cell lineages not expressing p57/Kip2 staining (Figure 3e) of HAI-1-deficient HAI-1, including trophoblast giant cells and

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1549

Figure 2 HAI-1 is required for placental labyrinth formation and embryonic development beyond E9.5. (a) Structure of the wild-type HAI-1 allele (top), HAI-1 targeting vector (middle) and the targeted HAI-1 allele (bottom). Exons of the HAI-1 coding region are indicated as black boxes, long and short arms of the targeting vector are shown as cross-hatched boxes. Neomycin- phosphoribosyltransferase (Neo, blue box) and the thymidylate kinase genes (TK, gray box) are indicated. (b) Southern blot of DrdI-digested genomic DNA from targeted (lane 1) and wild-type (lane 2) ES cell clones. (c) Northern blot (top panel) of total RNA extracted from the E10.5 wild type ( þ / þ , lane 1), HAI-1-heterozygous ( þ /À, lane 2) and HAI-1-deficient (À/À, lane 3) embryos hybridized to a mouse HAI-1 cDNA probe. The corresponding RNA gel (bottom panel) shows equal loading in all three lanes. HAI-1 mRNA is absent in HAI-1-deficient embryos and reduced in HAI-1-heterozygous embryos. (d and d0) Absence of HAI-1 immunoreactivity in E9.5 placental tissues of HAI-1-deficient embryos (d), compared to the wild-type littermate controls (brown staining, arrows in (d0)). (e) Macroscopic appearance of HAI-1-deficient (top row) and wild-type littermate (bottom) embryos at E9.5 (left), E10.5 (middle) and E11.5 (right) reveals severe growth retardation of HAI-1-deficient embryos after E9.5. (f) Distribution of HAI-1 þ / þ , HAI-1 þ /À and HAI-1À/À genotypes in the offspring from HAI-1 heterozygous mating pairs at E9.5 (top), E12.5 (middle) and postnatal day 1 (bottom) shows Mendelian representation of HAI-1À/À embryos at E9.5 and absence of HAI-1À/À embryos after E12.5. (g and g0) H&E-stained littermate HAI-1-expressing (g) and an HAI-1-deficient (g0) placenta at E10.5 reveals the absence of the placental labyrinth layer (‘Lb’ in (g)) in HAI-1À/À placentas, resulting in a complete absence of the contact between fetal vessels (arrowheads in (g) and (g0)) and maternal blood pools (open arrowheads in (g) and (g0)). Sections in (d) and (d0) were counterstained with hematoxylin. Size bars: d and d0,50mm; g and g0, 100 mm. spongiotrophoblasts, or in other parts of the developing Interestingly, in situ hybridization for spongiotropho- placenta, as assessed by H&E staining for necrosis or blast marker Tpbp showed that the space between the abnormal tissue architecture, TdT-mediated dUTP nick allantoic mesenchyme and the spongial layer in these end labelling assays for apoptosis, Ki67 staining for placentas is filled with a population of non-spongio- proliferation or p57/Kip2 staining (Figure 3b0–e0 and trophoblast cells (Supplementary Figure 4), suggesting data not shown). that loss of HAI-1 does not ablate chorionic trophoblast

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1550

Figure 3 HAI-1 deficiency perturbs normal differentiation of chorionic trophoblasts. (a) Schematic depiction of matriptase and HAI-1 expression during placental labyrinth formation in mice. After E9.0, HAI-1- and matriptase-positive chorionic trophoblasts (blue) start to differentiate into HAI-1-positive, matriptase-negative syncytiotrophoblast cells, followed closely by the ingrowth of embryonic vessels to form the mature placental labyrinth. (b–e0) Micrographs of H&E (b and b0), keratin (c, d, c0 and d0) and p57/Kip-2 staining (e and e0) of the mesenchymal–trophoblast interface of HAI-1-expressing (b–e) and HAI-1-deficient (b0–e0) placentas at E9.5. (d) and (d0) are high-magnification micrographs of the boxed area in (c) and (c0), respectively. HAI-1-expressing placentas display ingrowth of chorionic trophoblasts towards the spongial layer (sites of ingrowth indicated by stars in (b)–(e)). Fetal vessels within the forming labyrinth layer are indicated with arrowheads in (b)–(e). Furthermore, HAI-1-expressing placentas display a well-defined layer of keratin- and p57/Kip-2-negative chorionic trophoblasts (arrows in (b)–(e)) that separate the mesenchyme (located above the dashed line in (b)–(e0)) from the labyrinth layer. In sharp contrast, fetal vessels are confined to the allantoic mesenchyme (arrowheads in (b0)–(e0)), and the keratin- and p57/Kip-2-negative chorionic trophoblasts are completely absent from the HAI-1-deficient placentas. Pools of maternal blood present in the fetal part of the placenta are indicated by curved arrows in (b)–(e0). Sections in (c)–(e0) were counterstained with hematoxylin. Size bars: b, b0, d, d0, e and e0,25mm; c and c0,75mm.

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1551 population, but rather perturbs the differentiation chorionic trophoblasts of HAI-1-deficient placentas program of these cells. appeared normal at E8.5 when compared to HAI-1- expressing chorionic trophoblasts (Figure 4a and 4a0), HAI-1 ablation disrupts the integrity of chorionic already at E9.0, this cell population was grossly trophoblasts prior to the initiation of labyrinth formation abnormal in HAI-1-deficient placentas. Thus, whereas Because chorionic trophoblasts expressed both HAI-1 chorionic trophoblasts of HAI-1-expressing placentas and matriptase already at E8.5 (see above), we presented as a distinct epithelial sheath of cells that investigated if the failure of placental development was deposited a well-organized, laminin-containing mem- caused by the perturbance of this cell population prior brane (Figure 4b and c), this cell layer was completely to the initiation of labyrinth formation at E9.5. Whereas disrupted in HAI-1-deficient placentas and displayed

Figure 4 Loss of HAI-1 disrupts the integrity of the chorionic trophoblast layer prior to the initiation of placental labyrinth formation. (a and a0) H&E staining of chorionic trophoblasts (arrows) from HAI-1-expressing (a) and HAI-1-deficient placentas (a0)at E8.5. (b and b0) Deposition of a laminin-containing membrane (brown staining, arrowheads in (b)) by matriptase-expressing chorionic trophoblasts (cyan staining, arrows in (b)) in HAI-1-expressing placentas, and disrupted tissue architecture (arrows in (b0)) and disorganized laminin deposition (brown staining, arrowheads in (b0)) in HAI-1-deficient placentas at E9.0. (c and c0) Well-developed laminin-containing membrane (brown staining, arrowheads in (c)) at the mesenchyme–trophoblast interface in HAI-1-expressing placentas at E9.5, and absence of laminin-containing membrane in HAI-1-deficient placentas (c0). (d and d0) Well-developed epithelium-like sheath of matriptase-expressing chorionic trophoblasts (cyan, arrows in (d)) in HAI-1-expressing placentas, and scattered clusters of matriptase-expressing cells (cyan, arrows in (d0)) in HAI-1-deficient placentas. (e and e0) Prominent membrane- localized E-cadherin immunostaining by HAI-1-expressing placentas (brown staining, arrows in (e)), and diminished, often punctate, immunostaining of chorionic trophoblasts cells of HAI-1-deficient placentas (brown, arrows in (e0)). Insets in (e) and (e0) are high magnification of the boxed area. (f and f00) Prominent membrane-confined b-catenin immunostaining by HAI-1-expressing placentas (brown, arrows in (f)), and diminished (brown, arrows in (f0)), or absence of membrane-confined b-catenin immunostaining (f00)in matriptase-expressing chorionic trophoblasts (cyan, examples with arrowheads in (f0)) of HAI-1-deficient placentas. (g) Western blot of protein lysates from E9.5 placentas showing reduced E-cadherin and b-catenin in HAI-1-deficient placentas. Sections in (b)–(c0) and (e)–(f00) were counterstained with hematoxylin, and (d) and (d0) with nuclear fast red. Sections in (b) and (f0) were X-gal stained prior to immunohistochemistry. Size bars: a, a0, d, d0, e, e0,50mm; b, b0, c, c0, f, f0, f00,20mm.

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1552 disorganized laminin deposition (Figure 4b0). As a newborn matriptase/HAI-1 double-deficient pups did consequence, no laminin membrane was observed at not reveal any obvious abnormalities besides the E9.5 (Figure 4c0). Owing to their expression of epidermal defects associated with loss of matriptase, matriptase, the aberrantly developing chorionic tropho- including abnormally thin, reddish, wrinkled and dry blasts in HAI-1-deficient placentas could still be epidermis (Figure 5c), and the absence of erupted identified at E9.5, where they appeared as scattered whiskers (data not shown). All matriptase single- clusters of matriptase-positive cells or as single matrip- deficient and matriptase/HAI-1 double-deficient pups tase-positive cells (compare Figure 4d0 with d). More- died within the first day after birth with no obvious over, E-cadherin displayed an abnormal punctate difference in survival time. Also, milk spots were missing appearance in this cell population (compare Figure 4e0 from the stomachs of about 75% of the matriptase and e), and b-catenin immunostaining was either single-deficient and matriptase/HAI-1 double-deficient strongly reduced or largely absent from the periphery pups (examples in Figure 5c). Histological analysis of of the cell membrane (compare Figure 4f0 and f00 with the skin of newborn matriptase/HAI-1 double-deficient Figure 4f). In accordance with this, Western blot pups revealed an abnormal structure of the stratum analysis of protein lysates from HAI-1-deficient placen- corneum with tightly packed layers of immature tas at E9.5 showed reduced levels of both E-cadherin corneocytes (compare Figure 5d and d00), identical to (Figure 4g, top panel) and b-catenin (Figure 4g, bottom the skin of matriptase single-deficient littermate mice panel), reflecting a quantitative loss of chorionic (Figure 5d0). Also, similar to matriptase single-deficient trophoblast precursor cells and/or the aberrant expres- pups, histological examination of a number of other sion of the two epithelial markers by these cells. Taken organs and tissues of matriptase/HAI-1 double-deficient together, these results suggest that excess matriptase pups did not reveal any overt abnormalities (data not activity in HAI-1-deficient placentas caused a disruption shown). of the epithelial integrity of a population of matriptase- We next analysed the development of HAI-deficient expressing chorionic trophoblasts, leading to the per- placentas in the absence of matriptase. Notably, the turbance of their normal differentiation and subsequent failure of HAI-1-deficient chorionic trophoblasts to failure to form the placental labyrinth. migrate towards the spongial layer was completely restored by the added deficiency of matriptase. This led to normal labyrinth layer formation (Figure 5e–e00) Combined loss of both matriptase and HAI-1 results in and reversion of the growth retardation of the embryo normal placentation and term development proper (data not shown). Moreover, the chorionic Based on the above findings, we hypothesized that the trophoblasts presented normally in matriptase/HAI-1 functional defects in HAI-1-deficient chorionic tropho- double-deficient placentas (Figure 5f and g) despite the blasts could be alleviated by the simultaneous removal loss of HAI-1 expression (compare Figure 5f and h). of matriptase, and that matriptase/HAI-1 double- Taken together, these data show that loss of both the deficient embryos, therefore, would be capable of membrane protease and its inhibitor are compatible placental labyrinth formation. To challenge this hypoth- with development, whereas single deficiency in HAI-1 esis, we crossed HAI-1-heterozygous mice to previously disrupts embryonic development as a consequence of generated mice carrying a matriptase allele disrupted excessive matriptase activity. either by conventional gene targeting (List et al., 2002) or by insertion of the b-galactosidase marker gene (List et al., 2006). The resulting matriptase/HAI-1 double- heterozygous mice were then interbred, and their Discussion offspring was genotyped at birth. As expected from the above experiments, out of 130 matriptase-expressing The type II transmembrane serine protease matriptase is pups (matriptase þ / þ or matriptase þ /À) born from essential for postnatal survival, but dispensable for matriptase/HAI-1 double-heterozygous breeding pairs, development to term. In this paper, we reached the 45 were wild type and 85 were heterozygous for the unexpected conclusion that matriptase activity, never- targeted HAI-1 allele, while no HAI-1-negative pups theless, needs to be tightly regulated for embryonic were detected (Figure 5a). Strikingly, however, among development to occur, and that the Kunitz-type the matriptase-deficient offspring born from these transmembrane serine protease inhibitor HAI-1 is one crosses, the distribution of the HAI-1 wild type, essential matriptase inhibitor. This conclusion was heterozygous and null alleles in newborn pups displayed based on the following observations: (a) Matriptase a normal Mendelian distribution (Figure 5a and b). displayed colocalization with HAI-1 during the deve- Thus, of 57 matriptase-deficient pups, 15 were wild type, lopment of both extraembryonic and embryonic tissues. 29 were heterozygous and 13 were homozygous for the (b) Loss of HAI-1 specifically disrupted the integrity of a targeted HAI-1 allele (Po0.001 relative to matriptase- population of matriptase-expressing chorionic tropho- expressing pups, w2 analysis, two-tailed). As expected, blasts, preventing the formation of the placental immunostaining of the skin (Figure 5d00) and other labyrinth. (c) HAI-1-deficient mice displayed normal organs (data not shown) of newborn matriptase/HAI-1 placentation and development to term when placed on a double-deficient mice showed the complete absence of matriptase-deficient background. The co-expression of HAI-1 immunoreactivity. Macroscopic examination of matriptase and HAI-1 in the developing placenta was

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1553

Figure 5 Term delivery and normal placental development of matriptase/HAI-1 double-deficient embryos. (a) Genotype distribution of newborn pups from intercrosses of matriptase/HAI-1 double-heterozygous mice. No HAI-1-deficient pups were detected among 130 matriptase-expressing (blue bars) offspring. In contrast, HAI-1-deficient pups were detected in the expected Mendelian frequency in the 57 matriptase-deficient pups (green bars). (b) Southern blot analysis of a litter of newborn pups from matriptase/HAI-1 double- heterozygous parents demonstrating the deliverance to term of HAI-1-deficient pups (lanes 4 and 5). (c) External appearance of newborn matriptase and HAI-1-expressing (left), matriptase/HAI-1 double-deficient (middle) and matriptase single-deficient (right) littermates. Matriptase/HAI-1 double-deficient pups are outwardly indistinguishable from matriptase single-deficient pups. (d and d00) Histological appearance of wild-type (d), matriptase-deficient (d0) and matriptase/HAI-1 double-deficient (d00) skin from newborn littermates after HAI-1 immunostaining. Matriptase/HAI-1 double-deficient skin displays absence of HAI-1 immunoreactivity (brown staining, arrows in (d) and (d0)) in the upper layers of the epidermis and hair follicles, and an abnormally compact stratum corneum similar to that of matriptase-deficient skin (arrowheads in (d0) and (d00)). (e and e00) Micrographs of H&E staining of the placental tissue of wild type (e), HAI-1-deficient (e0) and matriptase/HAI-1 double-deficient (e00) placentas at E10.5 showing restoration of labyrinth layer formation in matriptase/HAI-1 double-deficient placentas. HAI-1-expressing and matriptase/HAI-1 double-deficient placentas both display fetal vessels (arrowheads in (e) and (e00)) in the labyrinth layer. In sharp contrast, no chorionic trophoblasts are present in the HAI-1 single-deficient placentas (e0). (f–h) Combined X-gal and HAI-1 immunostaining (f), X-gal and keratin immunostaining (g) and HAI-1 immunostaining (h) of matriptase/HAI-1 double-deficient (f and g) and matriptase and HAI-1-expressing placentas (h) showing the presence and normal appearance of HAI-1-negative (compare brown staining in (f) and (h)), keratin-negative (brown staining in (g)) chorionic trophoblasts (arrows in (f) and (g)) in matriptase/HAI-1 double-deficient placentas that transcribe the matriptase alleles (cyan staining, arrows in (f) and (g)). Sections in (d) and (d00) and (f)–(h) were counterstained with hematoxylin. Size bars: d, d0, d00,25mm; e, e0,e00,75mm; f–h,25mm. observed not only in mice but also in humans, indicating HAI-1 during development. Thus, our data are not that inhibition of matriptase by HAI-1 may be required incompatible with HAI-1 having several other impor- also for placental development in other mammals, and tant protease targets during development, but the perturbation of the matriptase-HAI-1 balance could inhibition of these proteases by HAI-1 then must be merit investigation as a cause of early miscarriage. redundant with other inhibitors. Besides matriptase, It is important to emphasize that our data do not HAI-1 has been reported to form inhibitor complexes show that matriptase necessarily is the only target for with the serine proteases hepatocyte growth factor

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1554 activator, hepsin and prostasin, and may have addi- precursor cells, triggering their demise or aberrant tional unidentified targets (see Introduction). It is differentiation. Alternatively, excess cell surface matrip- therefore formally possible that HAI-1 does not act tase activity may lead to cleavage of substrates that are directly on matriptase during development, but rather not normally cleaved by matriptase, such as components on other proteases that act either upstream or down- of the emerging basement membrane that is being stream of matriptase in a matriptase-dependent proteo- organized by chorionic trophoblasts at this develop- lytic pathway. However, for the following reasons this mental stage or components of their cell–cell junctions. does not seem likely: First, biochemical studies have Equally plausible, loss of HAI-1 may lead to patholo- shown that HAI-1 binds directly to matriptase with high gically increased cleavage of substrate(s) that are affinity and efficiently inhibits matriptase proteolytic normally cleaved by matriptase. The epithelial sheath activity (Oberst et al., 2003b, 2005). Second, shed of chorionic trophoblasts that express matriptase prior endogenous matriptase has been isolated in a complex to labyrinth formation is quite small (roughly 0.5 mm2). with shed endogenous HAI-1 in breast milk, indicating A direct biochemical/proteomic analysis of this tissue that a direct interaction occurs between matriptase and from HAI-1-expressing and HAI-1-deficient embryos to HAI-1 when co-expressed in vivo (Lin et al., 1999). Third identify matriptase proteolytic targets therefore may and most importantly, we found no evidence of prove challenging. A systematic screening for the expression of mRNA for hepatocyte growth factor expression of known candidate matriptase substrates activator in mouse placentas by reverse transcription– in these cells, followed by a genetic analysis to test if the polymerase chain reaction (PCR), and prostasin and specific ablation of each of these candidate substrates matriptase were expressed in different populations of can rescue placental formation in HAI-1-deficient cells in the developing placenta (R Szabo and TH embryos, appear likely to be more productive. Bugge, unpublished data). Taken together, our studies have identified matriptase Why would matriptase and HAI-1 be expressed as the single non-redundant target for HAI-1 during during placental development if the combined deficiency mouse embryonic development and have provided the in both is compatible with normal placentation, and mechanistic basis for the failure of HAI-1-deficient mice excessive expression of the protease could place the to form a functional placenta. embryo at risk for developmental failure? Three scenarios appear equally possible: The regulatory elements that confer matriptase expression in the Materials andmethods developing embryo may be identical to those required to confer expression of the protease in tissues where it is Gene targeted mice needed (e.g. epidermis and oral epithelium) and the The generation of conventional matriptase knockout and embryonic matriptase expression will be evolutionarily b-galactosidase-tagged matriptase knock-in mice has been conserved despite providing no benefit. Alternatively, described (List et al., 2002, 2006). To generate HAI-1 knockout mice, a 4.8 kb fragment of the 50-untranslated region matriptase has a critical function in development, but 0 this function is redundant with closely related proteases (long arm) and a 2.5 kb fragment of the 3 -untranslated region (short arm) of the mouse HAI-1 gene were amplified from such as matriptase-2 (Velasco et al., 2002; Hooper et al., mouse 129/SvEvTac genomic DNA by long-range PCR, using 2003) and matriptase-3 (Szabo et al., 2005). Finally, the Expand Long DNA Template kit (Roche, Indianapolis, embryonic matriptase expression may be important for IN, USA) according to the manufacturer’s instructions. The development only under environmental stress conditions primers used for the amplification were L51 (50-CTAGCTC (infection, intoxication, malnutrition, heat/cold expo- GAGTGTGTTCTCACCACTGAACC-30) and L31 (50-TCA sure, other) that do not present under standard animal GCTCGAGTGAACACCAGAAGTGCGTCCC-30) for the housing conditions. long arm, and S52 (50-TAGACCGCGGCATATCTCTA Matriptase and HAI-1 also are co-expressed in a large CCTTCCTGTCC-30) and S32 (50-TGACGCGGCCGCATT 0 number of tissues of the embryo proper, including the CTGTAGAGTAGGCTTGCC-3 ) for the short arm. The developing oropharynx, olfactory system, intestine and HAI-1 targeting vector was generated by cloning the long arm DNA into the XhoI restriction site of the pKO1907 skin (this study), as well as in a number of adult tissues Scrambler vector (Stratagene, La Jolla, CA, USA) followed by in mice (K List and T Bugge, unpublished data) and the cloning of the short arm into the NotI/SacII sites of the humans (Oberst et al., 2003a). It therefore seems likely same vector. The targeting vector was linearized with PvuI, that HAI-1 also may have an essential role in restricting transfected into the 129/SvEvTac embryonic stem (ES) cells by matriptase activity in these tissues, and that even if electroporation, and the positive clones were identified by placentation could be restored, HAI-1 deficiency would Southern blot hybridization of DrdI-digested genomic DNA not be compatible with term development or adult life. using a 32P-labeled DNA probe corresponding to the 30- Multiple molecular mechanisms by which excessive untranslated region of the HAI-1 gene, which was external to matriptase activity in HAI-1-deficient chorionic tropho- the targeting vector sequence. The HAI-1-targeted ES cells blast precursors may lead to the disruption of epithelial were injected into the blastocoel cavity of C57Bl/6J-derived blastocysts and implanted into pseudopregnant females. The integrity, aberrant differentiation and subsequent failure resulting chimeric male offspring was bred to NIH Black Swiss of these cells to produce a progeny of labyrinth-forming females (Taconic Farms, Germantown, NY, USA) to generate cells are possible. The premature activation of matrip- offspring carrying one HAI-1-targeted allele. Genotyping of tase in the secretory pathway caused by HAI-1 animals from ear and yolk sac biopsies was performed by PCR deficiency may be toxic to the chorionic trophoblast with primers HAI-54 (50-CTGTCTGTTTACCTTGTCTCG-30)

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1555 and HAI-34 (50-TAAAGGGATGTTGTCCTGGG-30) for the 2% bovine serum albumin in PBS, and incubated overnight at wild-type allele and HAI-KOR51 (50-GAGAATTCCTA 41C with the primary antibody (see Supplementary Methods). GAGGTACCC-30) and HAI-KOR31 (50-GGTAGTTGAGG Bound antibodies were visualized using biotin-conjugated anti- GAACATGGC-30) for the HAI-1-targeted allele. mouse, -rabbit, rat or -goat secondary antibodies (Vector Laboratories, Burlingame, CA, USA) and a Vectastain ABC Extraction of embryonic and perinatal tissues kit (Vector Laboratories) using 3,30-diaminobenzidine as the Breeding females were checked for vaginal plugs in the substrate (Sigma Chemical Co, St Louis, MO, USA). All morning and the day on which the plug was found was microscopic images were acquired on Olympus BX40 micro- defined as the first day of pregnancy (E0.5). The pregnant scope using Olympus DP70 digital camera system (both females were euthanized in the mid-day at designated time Olympus, Melville, NY, USA). points by CO2 asphyxiation. Embryos were extracted by Caesarian section and the individual embryos and placentas Western blotting were dissected and processed as described below. Visceral yolk The fetal part of E9.5 placentas was separated from the sacs of individual embryos were washed twice in phosphate- decidua by manual microdissection. The proteins were buffered saline (PBS), subjected to genomic DNA extraction extracted by homogenizing the microdissected tissue in lysis and genotyped by PCR. Paraffin sections of 13th week human buffer (62.5 mM Tris/HCl, pH 6.8, 2% SDS, 10% glycerol) and placenta were purchased from BioChain (Hayward, CA, subjected to a Western blot analysis as described previously USA). Newborn pups were euthanized by CO2 inhalation at (Szabo et al., 2005) using primary anti-E-cadherin (Invitrogen- 01C. Skins or whole pups were fixed for 24 h in 4% Zymed, San Francisco, CA, USA) or anti-b-catenin (Sigma) paraformaldehyde (PFA) in PBS, processed into paraffin, antibodies. sectioned into parallel sagittal sections and stained with H&E, or used for immunohistochemistry as described below.

Whole-mount histological analysis and b-galactosidase staining Abbreviations Tissue samples were fixed for 30 min in 4% PFA in PBS, rinsed in PBS and stained with a b-galactosidase staining kit (Roche) E, embryonic day; ES, embryonic stem (cells); HAI, hepato- for 6h at 37 1C using X-gal as the substrate. The tissues were cyte growth factor activator inhibitor; X-gal, 5-bromo-4- post-fixed overnight in 4% PFA, embedded in paraffin and chloro-3-indolyl-beta-D-galactopyranoside. sectioned. Acknowledgements Immunohistochemistry Tissues were fixed overnight in 4% PFA, embedded in paraffin We thank Drs Chen-Yong Lin and Robert Dickson for and cut into 5-mm-thick sections. Antigens were retrieved by generously providing human matriptase antibody, and incubation for 10 min at 371C with 10 mg/ml proteinase K Drs Silvio Gutkind and Mary Jo Danton for critically review- (Fermentas, Hanover, MD, USA) for HAI-1 staining, or by ing this manuscript. This work was supported by the NIH incubation for 20 min at 1001C in 0.01 M sodium citrate buffer, Intramural program and by a grant from the Department of pH 6.0, for all other antigens. The sections were blocked with Defense (DAMD-17-02-1-0693) to Dr Thomas H Bugge.

References

Ascenzi P, Bocdi A, Bolognesi M, Spallarose A, Colletta M, Huang ZF, Higuchi D, Lasky N, Broze Jr GJ. (1997). Tissue De Cristofaro R et al. (2003). The bovine basic pancreatic factor pathway inhibitor gene disruption produces intrau- trypsin inhibitor (Kunitz inhibitor): a milestone protein. terine lethality in mice. Blood 90: 944–951. Curr Protein Pept Sci 4: 231–251. Ishiguro K, Kojima T, Kadomatsu K, Nakayama Y, Takagi A, Fan B, Wu TD, Li W, Kirchhofer D. (2005). Identification of Suzuki M et al. (2000). Complete antithrombin deficiency in hepatocyte growth factor activator inhibitor-1B as a mice results in embryonic lethality. J Clin Invest 106: potential physiological inhibitor of prostasin. J Biol Chem 873–878. 280: 34513–34520. Itoh H, Naganuma S, Takeda N, Miyata S, Uchinokura S, Gailani D, Lasky NM, Broze Jr GJ. (1997). A murine Fukushima T et al. (2004). Regeneration of injured model of factor XI deficiency. Blood Coagul Fibrinolysis 8: intestinal mucosa is impaired in hepatocyte growth 134–144. factor activator-deficient mice. Gastroenterology 127: Gao F, Shi HY, Daughty C, Cella N, Zhang M. (2004). 1423–1435. Maspin plays an essential role in early embryonic develop- Kaufmann MH, Bart S. (1999). The Anatomical Basis of ment. Development 131: 1479–1489. Mouse Development, revised edn. Academic Press: San Herter S, Piper DE, Aaron W, Gabriele T, Cutler G, Cao P Diego, CA. et al. (2005). Hepatocyte growth factor is a preferred in vitro Lin CY, Anders J, Johnson M, Dickson RB. (1999). substrate for human hepsin, a membrane-anchored serine Purification and characterization of a complex containing protease implicated in prostate and ovarian cancers. matriptase and a Kunitz-type serine protease inhibitor from Biochem J 390: 125–136. human milk. J Biol Chem 274: 18237–18242. Hooper JD, Campagnolo L, Goodarzi G, Truong TN, List K, Haudenschild CC, Szabo R, Chen W, Wahl SM, Stuhlmann H, Quigley JP. (2003). Mouse matriptase-2: Swaim W et al. (2002). Matriptase/MT-SP1 is required for identification, characterization and comparative mRNA postnatal survival, epidermal barrier function, hair expression analysis with mouse hepsin in adult and follicle development, and thymic homeostasis. Oncogene embryonic tissues. Biochem J 373: 689–702. 21: 3765–3779.

Oncogene Membrane protease inhibition in embryonic development R Szabo et al 1556 List K, Szabo R, Molinolo A, Nielsen BS, Bugge TH. (2006). Riley P, Anson-Cartwright L, Cross JC. (1998). The Hand1 Delineation of matriptase protein expression by enzymatic bHLH transcription factor is essential for placentation and gene trapping suggests diverging roles in barrier function, cardiac morphogenesis. Nat Genet 18: 271–275. hair formation, and squamous cell carcinogenesis. Am J Shimomura T, Denda K, Kitamura A, Kawaguchi T, Kito M, Pathol 168: 1513–1525. Kondo J et al. (1997). Hepatocyte growth factor activator List K, Szabo R, Molinolo A, Sriuranpong V, Redeye V, inhibitor, a novel Kunitz-type serine protease inhibitor. J Murdock T et al. (2005). Deregulated matriptase causes Biol Chem 272: 6370–6376. ras-independent multistage carcinogenesis and promotes Szabo R, Netzel-Arnett S, Hobson JP, Antalis TM, Bugge TH. ras-mediated malignant transformation. Genes Dev 19: (2005). Matriptase-3 is a novel phylogenetically preserved 1934–1950. membrane-anchored serine protease with broad serpin Oberst MD, Chen LY, Kiyomiya K, Williams CA, Lee MS, reactivity. Biochem J 390: 231–242. Johnson MD et al. (2005). HAI-1 regulates activation Tanaka H, Nagaike K, Takeda N, Itoh H, Kohama K, and expression of matriptase, a membrane-bound serine Fukushima T et al. (2005). Hepatocyte growth factor protease. Am J Physiol Cell Physiol 289: C462–C470. activator inhibitor type 1 (HAI-1) is required for branching Oberst MD, Singh B, Ozdemirli M, Dickson RB, Johnson MD, morphogenesis in the chorioallantoic placenta. Mol Cell Biol Lin CY. (2003a). Characterization of matriptase expression 25: 5687–5698. in normal human tissues. J Histochem Cytochem 51: Velasco G, Cal S, Quesada V, Sanchez LM, Lopez-Otin C. 1017–1025. (2002). Matriptase-2, a membrane-bound mosaic serine Oberst MD, Williams CA, Dickson RB, Johnson MD, proteinase predominantly expressed in human liver and Lin CY. (2003b). The activation of matriptase requires showing degrading activity against extracellular matrix its noncatalytic domains, serine protease domain, and its proteins. J Biol Chem 30: 37637–37646. cognate inhibitor. J Biol Chem 278: 26773–26779. Wu Q, Yu D, Post J, Halks-Miller M, Sadler JE, Morser J. Oh J, Takahashi R, Kondo S, Mizoguchi A, Adachi E, (1998). Generation and characterization of mice deficient in Sasahara RM et al. (2001). The membrane-anchored MMP hepsin, a hepatic transmembrane serine protease. J Clin inhibitor RECK is a key regulator of extracellular matrix Invest 101: 321–326. integrity and angiogenesis. Cell 107: 789–800. Zeeuwen PL, van Vlijmen-Willems IM, Hendriks W, Ohmuraya M, Hirota M, Araki M, Mizushima N, Matsui M, Merkx GF, Schalkwijk J. (2002). A null mutation in the Mizumoto T et al. (2005). Autophagic cell death of cystatin M/E gene of ichq mice causes juvenile lethality pancreatic acinar cells in serine protease inhibitor Kazal and defects in epidermal cornification. Hum Mol Genet 11: type 3-deficient mice. Gastroenterology 129: 696–705. 2867–2875.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

Oncogene