Catalysis and Substrate Selection

The catalytic domain of all membrane-anchored serine proteases belongs to the S1 peptidase family (Rawlings & Barrett 1993). The biochemistry of this family, which includes the prototypic proteases chymotrypsin and trypsin, is exceptionally well characterized, and the mechanism of catalysis is understood in great detail (for reviews, see Hedstrom 2002, Page & Di Cera 2008). A catalytic triad of serine, aspartate, and histidine residues with a conserved spatial orientation executes peptide bond hydrolysis. All membrane-anchored serine proteases show a strong preference for cleavage of substrates after arginine or lysine residues. Target substrate selection in vivo, however, is determined by a host of additional factors, including substrate binding to other parts of the catalytic domain and to accessory protein domains, control of proenzyme activation, protease and substrate compartmentalization, macromolecular and small-molecule cofactors, and general and specific protease inhibitors.

Membrane Anchorage

Membrane anchorage is achieved by one of three means (Figure 1): by a single-pass transmembrane domain located close to the C terminus that orients the protease as a type I transmembrane protein (as in tryptase γ1), by a glycosylphosphatidylinositol (GPI) anchor that is added posttranslationally to the C terminus and attaches the protease to the outer leaflet of the plasma membrane (as in prostasin and testisin), or by a signal anchor that is located near the N terminus and orients the protease in the membrane as a type II transmembrane protein with a large extracellular C terminus containing the serine protease domain (all type II transmembrane serine proteases).

Orchestrating Neural Tube Closure

The neural tube forms from a thickened region of ectoderm that is termed the neural plate or neuroepithelium. From this structure, bilateral neural folds form at the junction with the surface ectoderm. These folds elevate, come into contact, and fuse at the midline to create the neural tube (Figure 2). Failure to complete the closure of the neural tube is among the most common and serious human birth defects, affecting approximately 1 in 1,000 established pregnancies and causing anencephaly and spina bifida (reviewed in Copp et al. 2003). Owing to the clinical importance of this birth defect, considerable effort has been put into determining the molecular basis of neural tube closure. Genetic analysis has identified a diverse array of molecules as critical to the process, including cytoskeleton-associated proteins, cell cycle and neurogenesis proteins, cell death–associated proteins, cell signaling molecules, extracellular matrix proteins and their receptors, and transcription factors. Of the more than 150 genes identified over the years by loss-of-function studies as critical to neural tube closure, not one has encoded an extracellular protease, which suggests that pericellular proteolysis might not play a role in this process (reviewed in Copp & Greene 2010, Zohn & Sarkar 2008). However, a recent landmark study by Camerer et al. (2010) changed this perception by showing that protease-activated receptor (PAR)-1 or PAR-2 activation in the nonneural surface ectoderm that directly overlies the forming neural tube is essential for late events of neural tube closure. Proteolytic activation of either of the PARs in this tissue induces fusion of the neural folds by activating a cell motility-associated molecule, Rac1, through the G_i/o/z subfamily of G proteins (Figure 2). Thrombin, the principal activator of PAR-1, is widely expressed in the developing central nervous system (Dihanich et al. 1991, Szabo et al. 2009a) and likely accounts for PAR-1 activation during neural tube closure. PAR-2, however, is not efficiently activated by thrombin, but systematic profiling revealed that a surprisingly large number of membrane-anchored serine proteases with PAR-2-activating potential are expressed at the site of neural tube closure. These include matriptase; matriptase-3; transmembrane protease, serine 2 (TMPRSS2); TMRPSS4; hepsin; and prostasin (Camerer et al. 2010, Szabo et al. 2009a). Of these, matriptase is a particularly potent activator of PAR-2 and, interestingly, can sensitize cultured cells to PAR-2 activation by prostasin and hepsin, which suggests that a local membrane-anchored serine protease cascade mediates PAR-2 activation in the context of neural tube closure (Camerer et al. 2010).

Open in a separate window

Figure 2

Membrane-anchored serine protease signaling in the surface ectoderm facilitates neural tube closure. Protease-activated receptor (PAR)-dependent Rac1 activation in the surface ectoderm through G_i/o/z is required for the fusion of the neural folds. PAR-1 is activated by thrombin. PAR-2 may be activated by prostasin, acting through matriptase, or by other locally expressed membrane-anchored serine proteases including hepsin; transmembrane protease, serine 2 (TMPRSS2); TMPRSS4; and matriptase-3. Membrane-anchored serine protease activity in surface ectoderm cells is negatively regulated by hepatocyte growth factor activator inhibitor 2 (HAI-2).

Independent support for a membrane-anchored serine protease signaling axis acting in neural tube closure arose from studies of the effect of ablation of hepatocyte growth factor activator inhibitor 2 (HAI-2) (Szabo et al. 2009a). This transmembrane serine protease inhibitor is expressed in neural and nonneural ectoderm precisely at the future point of contact between the neural folds, and its deletion from mice causes neural tube closure defects that are strikingly similar to those observed after combined PAR-1 and PAR-2 ablation. Furthermore, this phenotype was partially rescued by simultaneous ablation of matriptase, indicating that both insufficient and excessive local protease-initiated cell signaling derails the closing of the neural tube, with the latter possibly due to the well-established desensitization of PARs that occurs after constitutive stimulation.

Excess local membrane-anchored serine protease activity is detrimental not only to neural tube closure but also to mouse placental labyrinth formation. This structure, which is vital to the fetomaternal exchange of nutrients and oxygen, is formed by branching morphogenesis of chorionic trophoblasts. Matriptase, HAI-2, and the related inhibitor HAI-1 all are expressed by this epithelial cell population. Ablation of either HAI-1 or HAI-2 or the combined haploinsufficiency for HAI-1 and HAI-2 leads to the failure of the placental labyrinth to form. However, when matriptase is simultaneously reduced or eliminated from the placenta, placental development proceeds normally (Fan et al. 2007; Szabo et al. 2007, 2009a,b; Tanaka et al. 2005), which indicates that matriptase is not necessary for placental development but that its activity in the placenta nevertheless needs to be regulated by both HAI-1 and HAI-2.

HAI-2 deficiency in humans does not cause placental or neural tube defects but rather a wide spectrum of other developmental abnormalities such as duplication of internal organs, duplication and abnormal location of digits, craniofacial dysmorphisms, anal and choanal atresia, fistulas, and hamartoma (Heinz-Erian et al. 2009). It remains to be established if these are caused by excess matriptase activity or by the loss of inhibition of other membrane-anchored or secreted serine proteases.

Erecting Barriers Between Cells

Sealing the epidermis

Prior to birth, terrestrial mammals develop an external permeability barrier that prevents excessive water loss and the entry of noxious chemicals and microbes. This barrier function is provided by the epidermis, a multilayered epithelium in which keratinocytes arise from proliferating basal cells, move outward, and undergo a series of distinct differentiation events to form the stratum corneum, a two-compartment structure consisting of a lipid-enriched extracellular matrix in which an interlocking meshwork of dead keratinocytes (corneocytes) is embedded. The stratum corneum and the tight junctions formed between the underlying living keratinocytes constitute the epidermal barrier. The thickness of the stratum corneum is regulated by the shedding (desquamation) of the outermost layer of corneocytes through proteolytic degradation of their desmosomal junctions by epidermal kallikrein-related peptidases (Fuchs & Raghavan 2002, Furuse et al. 2002, Nemes & Steinert 1999, Presland & Dale 2000, Roop 1995, Segre 2003).

The epidermis has long been studied as a convenient model of epithelial cell differentiation, but the essential role of membrane-anchored serine proteases in epidermal development was uncovered only recently by the fortuitous discovery that mice deficient in either matriptase or prostasin fail to develop epidermal barrier function sufficient for postnatal survival (Leyvraz et al. 2005, List et al. 2002). Subsequent studies showed that both proteases are selectively expressed in terminally differentiating keratinocytes and act as part of a single proteolytic cascade in which the autoactivating matriptase serves to activate the prostasin zymogen (Netzel-Arnett et al. 2006). This proteolytic cascade regulates at least four key steps of terminal epidermal differentiation that are shown schematically in Figure 3: (a) formation of claudin-1-based tight junctions between living suprabasal keratinocytes (Furuse et al. 2002, Leyvraz et al. 2005, List et al. 2009); (b) activation of an intracellular μ-calpain-caspase-14-bleomycin hydrolase proteolytic and hydrolytic cascade that converts profilaggrin into modified free hygroscopic amino acids that hydrate the skin and absorb UV light (Denecker et al. 2007, Kamata et al. 2009, Leyvraz et al. 2005, List et al. 2003, Yamazaki et al. 1997); (c) initiation of intercorneocyte lipid manufacturing and deposition, possibly by enhancing keratinocyte sodium uptake by activation of the epithelial sodium channel (ENaC) (see below) (Charles et al. 2008, Leyvraz et al. 2005, List et al. 2003); and (d ) induction of desquamation by kallikrein-related peptidases by promoting their expression or activation (List et al. 2003, Sales et al. 2010). The matriptase-prostasin cascade also plays a critical role in the development of hair follicles and in hair growth, but very little as yet is known about the differentiation and growth pathways upon which the protease cascade acts (Alef et al. 2009; Avrahami et al. 2008; Basel-Vanagaite et al. 2007; List et al. 2002, 2006, 2007a, 2009; Spacek et al. 2010).

Open in a separate window

Figure 3

Epidermal differentiation processes controlled by the matriptase-prostasin proteolytic cascade. (a) Formation of tight junctions between granular-layer keratinocytes. (b) Cytoplasmic processing of profilaggrin to filaggrin monomers by μ-calpain and complete processing to free hygroscopic and UV light–absorbing amino acids by caspase-14 and bleomycin hydrolase. (c) Epidermal lipid synthesis by transitional cells and its deposition into the intercellular space between corneocytes as lipid lamellae. (d ) Proteolytic shedding (desquamation) of the stratum corneum by kallikrein (KLK)-mediated degradation of desmogleins in desmosomes linking corneocytes. If the four processes are independent or interrelated, and what target substrate(s) is cleaved in each process by the matriptase-prostasin cascade, are unknown.

As demonstrated by recent genetic interaction studies, the local action of cognate serine protease inhibitors must tightly regulate membrane-anchored epidermal serine proteases. HAI-1 is coexpressed with matriptase and prostasin in the terminally differentiating layers of the epidermis as well as in hair cortex and cuticle cells of the hair follicle. HAI-1 ablation from keratinocytes leads to fatal ichthyosis, a hallmark of compromised epidermal barrier function, and to defects in hair structure and growth (Nagaike et al. 2008). However, both barrier and hair follicle defects are eliminated completely by simultaneous reduction in the level of epidermal matriptase (Nagaike et al. 2008, Szabo et al. 2009b). Netherton syndrome, a second example of the detrimental effects of dysregulated epidermal membrane-anchored serine protease activity, is characterized by premature stratum corneum shedding with direct exposure of the living layers of the epidermis to the external environment. It is caused by genetic deficiency of the lympho-epithelial Kazal-type-related serine protease inhibitor (LEKTI) (Chavanas et al. 2000a,b; Deraison et al. 2007; Descargues et al. 2005, 2006; Hachem et al. 2006; Komatsu et al. 2002; Magert et al. 1999; Schechter et al. 2005; Suzuki et al. 1994). At the molecular level, LEKTI deficiency causes a matriptase-initiated runaway kallikrein-related peptidase cascade, which leads to premature desmosome degradation (Sales et al. 2010).

Developing the Inner Ear

At least two and likely three membrane-anchored serine proteases participate in auditory development, which was realized recently when mutations in the TMPRSS3 gene were identified as a cause of congenital and childhood onset autosomal recessive deafness (Ben-Yosef et al. 2001, Scott et al. 2001). Spurred by these findings, subsequent systematic screening of human deafness-associated loci and functional analysis of mouse knockouts additionally identified hepsin and likely spinesin as critical for hearing function (Guipponi et al. 2007, 2008). The cause of spinesin- and TMPRSS3-associated deafness has not been explored. Hepsin deficiency, however, has been linked to aberrant tectorial membrane development and decreased myelination of the auditory nerve. These defects appear to be secondary to either low circulating levels of thyroid hormone or low responsiveness of inner ear tissues to thyroid hormone, but how hepsin would control thyroid hormone action is not evident and as yet unstudied (Guipponi et al. 2007, 2008; Hanifa et al. 2010).

Regulating Apical Sodium Entry and Fluid Volume

The highly amiloride-sensitive epithelial sodium channel (ENaC) has a central role in determining extracellular fluid composition and volume by regulating the apical entry of sodium into polarized epithelial cells. ENaC is expressed in high-resistance epithelia of the distal colon, renal collecting tubes, airway epithelium, suprabasal epidermis, and sweat glands. Loss of ENaC function reduces fluid clearance from the alveolar space, causes renal sodium loss, and causes failure to develop epidermal barrier function (Barker et al. 1998, Hummler et al. 1996, Mauro et al. 2002, McDonald et al. 1999, Rossier et al. 2002). In its most active form, ENaC is a heterotetramer composed of homologous α, β, and γ subunits that combine to form a channel (Firsov et al. 1998, Garty & Palmer 1997, Rossier et al. 2002). The potential relevance of proteolysis to apical sodium transport became evident with the discovery that exposure of epithelial cells to trypsin or aprotinin increased or decreased, respectively, the open probability of ENaC. Later studies showed that proteolysis indeed provides one of the principal means to rapidly increase cellular sodium transport in response to a homeostatic challenge such as dietary sodium restriction or pulmonary edema (Ergonul et al. 2006, Masilamani et al. 1999, Narikiyo et al. 2002, Planès et al. 2009).

The mechanism by which serine proteases regulate ENaC activity is complex and far from fully elucidated. A tentative model is shown in Figure 4. During biosynthesis of the channel, furin cleaves the α subunit after the sequences RSTR²⁰⁵ and RSAR²³¹ and the γ chain after RKRR¹⁴⁴ (Hughey et al. 2003, 2004). α-Chain cleavage increases channel activity, but these furin cleavage events are constitutive and therefore unlikely to be the point of action for homeostatic regulation of ENaC by proteases. A second cleavage event in the ENaC γ subunit, which can enhance sodium transport, occurs after the sequence RKRK¹⁸⁶ (Bruns et al. 2007). Cleavage at RKRK¹⁸⁶, when combined with the preceding furin cleavage after RKRR¹⁴⁴, increases ENaC activity several-fold by removing an inhibitory loop of the γ chain (Bruns et al. 2007). Importantly, cleavage at this site in the γ chain is enhanced by sodium restriction in vivo (Masilamani et al. 1999), and is consistent with the activation of ENaC by serine proteases with trypsin-like specificity in various model systems, such as Xenopus oocytes reconstituted with all three ENaC subunits and cell lines expressing endogenous ENaC (Garcia-Caballero et al. 2008; Vallet et al. 1997; Vuagniaux et al. 2000, 2002; Wakida et al. 2006).

Open in a separate window

Figure 4

Proteolytic regulation of epithelial sodium channel (ENaC) activity. Only one α subunit is shown, and the β subunit is omitted for simplicity. (a) The unprocessed channel has low sodium currents. (b) Furin removes an autoinhibitory disulfide-bonded loop in the α subunit during ENaC trafficking to the plasma membrane to increase the open probability. Furin also partially removes a second autoinhibitory loop in the γ subunit. (c) Full ENaC activation is achieved by complete removal of the γ subunit autoinhibitory loop by prostasin. Other membrane-anchored serine proteases may activate furin-processed ENaC in a similar manner.

TMPRSS3, TMPRSS4, matriptase, and prostasin can activate ENaC in model systems (Guipponi et al. 2002, Vallet et al. 1997, Vuagniaux et al. 2002; reviewed in Rossier & Stutts 2009). However, evidence for a physiological role in channel activation is limited to prostasin thus far. Most compelling in this regard, ENaC activity is reduced in prostasin-silenced cultured epithelial cells, and fluid clearance from the lungs is impaired in mice with alveolar epithelium-specific prostasin ablation (Narikiyo et al. 2002, Planès et al. 2009, Tong et al. 2004). ENaC activation by prostasin requires membrane anchorage, is inhibited by aprotinin, and requires an intact RKRK¹⁸⁶ γ chain sequence. Paradoxically, however, the catalytic activity of prostasin appears to be dispensable for its capacity to activate ENaC, at least in the Xenopus model system, where catalytically inactive prostasin can induce both ENaC γ subunit cleavage and channel activation (Andreasen et al. 2006, Bruns et al. 2007). Tentatively, inactive prostasin may activate ENaC by titrating out a membrane-associated serine protease inhibitor, thereby increasing the activity of an endogenous ENaC-activating protease. Alternatively, inactive prostasin may serve as a scaffold for assembly of an ENaC-activating protease complex on the cell surface. Whichever the case, the above findings make it clear that mechanistic insight into the complex association between membrane-anchored serine proteases and epithelial sodium transport is still limited.

Controlling the Natriuretic Peptide System

Atrial natriuretic peptide (ANP) and brain natriuretic peptide (BNP) are polypeptide hormones released from secretory storage granules of atrial cardiocytes in response to high blood pressure and volume overload. They promote urine production, urinary sodium excretion, and widening of blood vessels by exerting pleiotropic effects on kidneys, adrenals, vasculature, and the peripheral and central nervous systems through binding and activation of a guanylyl cyclase-coupled receptor, natriuretic peptide receptor-A. ANP and BNP also display antiproliferative effects on vascular smooth muscle cells, fibroblasts, and cardiocytes (reviewed in de Bold 1985, McGrath et al. 2005). Both hormones are synthesized as inactive precursors, proANP and proBNP, that are converted to their biologically active forms by proteolytic removal of 98 and 76 amino acids, respectively, from their N termini. Furin constituitively activates proBNP during biosynthesis (Semenov et al. 2010) and it is stored in secretory granules in its biologically active form. ProANP, however, lacks the RXXR furin cleavage motif found in proBNP. It is proteolytically converted to ANP only during secretion from storage granules (reviewed in McGrath et al. 2005, Wu et al. 2009). The identity of the proANP-converting protease was long uncertain and was finally revealed to be the complex cardiac-specific membrane-anchored serine protease, corin (Hooper et al. 2000, Yan et al. 1999). In model systems, corin proteolytically processes proANP to ANP by cleavage after R⁹⁸ (Knappe et al. 2003, Yan et al. 2000). In Corin-ablated mice, ANP is undetectable, and proANP is elevated. Furthermore, these mice display spontaneous and salt-sensitive hypertension similar to proANP-and natriuretic peptide receptor-A-deficient mice, as well as cardiac hypertrophy and reduced cardiac function in later life (Chan et al. 2005 and references therein).

Corin is found on the surface of proANP-producing atrial cardiocytes as a mixture of zymogen and active forms (Gladysheva et al. 2008). Productive interaction of corin with proANP requires at least the first frizzled domain and the first four LDLR repeats, which likely bind directly to proANP (Knappe et al. 2004). The corin zymogen itself is activated by an unidentified serine protease in a process that involves the second frizzled domain and N-linked glycans within the serine protease domain (Liao et al. 2007, Wang et al. 2008). Importantly, corin zymogen activation is a rate-limiting step for the entire natriuretic peptide system, as revealed by recent genome-wide association studies. Thus, a minor CORIN allele that is characterized by two nonsynonymous nonconservative single nucleotide polymorphisms is present in approximately 10% of surveyed African Americans and gives rise to a corin variant protein with T555I and Q568P substitutions in the second frizzled domain (Dries et al. 2005). ProANP processing by this corin variant is unaltered, but zymogen activation is 60% reduced (Wang et al. 2008). This reduction of corin zymogen activation puts carriers of the variant CORIN allele at increased risk for both hypertension and cardiac hypertrophy (Dries et al. 2005, Rame et al. 2007).

Adjusting Cellular Iron Export to Iron Need

Iron homeostasis is maintained by regulating intestinal iron absorption and the release from macrophages of iron recycled from erythrocytes (Figure 5). Both processes are controlled systemically by a hepatic peptide hormone, hepcidin (reviewed in Andrews 2008, Nemeth & Ganz 2006). Hepcidin binds directly to the major iron export protein, ferroportin, which is located on macrophages and on the basolateral membrane of iron-absorbing enterocytes of the duodenum. This binding triggers the internalization and ubiquitin-mediated lysosomal degradation of ferroportin. Loss of ferroportin thus prevents both intestinal iron uptake and release of iron from intracellular stores of iron-recycling macrophages, with the net result of decreased serum iron (De Domenico et al. 2007, 2008; Nemeth et al. 2004b, 2006). Hepcidin is regulated at the level of transcription, as it is induced by iron overload and suppressed by iron deficiency, anemia, and hypoxia (Nemeth et al. 2004a, Nicolas et al. 2002). The GPI-anchored protein hemojuvelin is a major regulator of hepcidin production that functions as a coreceptor for bone morphogenetic protein (BMP)6, which signals through BMP receptors and SMAD proteins to stimulate hepcidin gene (HAMP) transcription (Figure 5; Andriopoulos et al. 2009, Babitt et al. 2006, Huang et al. 2005, Meynard et al. 2009, Niederkofler et al. 2005, Roetto et al. 2003, Wang et al. 2005).

Open in a separate window

Figure 5

Matriptase-2 regulates cellular iron export. ➀ Dietary iron and iron recycled from erythrocytes are stored in enterocytes and splenic macrophages and are released to the circulation through ferroportin. ➁ Hepcidin produced by hepatocytes binds ferroportin and targets the channel for degradation. ➂ Hepcidin gene (HAMP) expression is positively regulated by bone morphogenic protein (BMP)6, which signals through the BMP receptor (BMPR)-SMAD pathway in a hemojuvelin-dependent manner. Matriptase-2 increases cellular iron export by degrading hemojuvelin to reduce hepcidin production and thus increase ferroportin levels.

The homeostatic circuit outlined above, although extremely elegant, does not account for how iron levels regulate hepcidin synthesis. A surprising explanation involving cell surface proteolysis emerged recently from two independent lines of research: the study of individuals with the rare autosomal recessive disorder, iron-refractory, iron-deficiency anemia (IRIDA), and the generation of mice deficient in Tmprss6, which encodes matriptase-2. Individuals with IRIDA present with an iron deficiency anemia that is unresponsive to oral iron administration. Paradoxically, hepcidin levels are high in IRIDA patients, indicating a systemic failure to suppress hepcidin gene transcription in response to iron deficiency. Tmprss6-deficient mice displayed a strikingly similar phenotype, which prompted the identification of TMPRSS6 loss-of-function mutations in IRIDA patients (Du et al. 2008, Finberg et al. 2008, Folgueras et al. 2008, Hooper et al. 2003, Velasco et al. 2002). Independent support for a role of matriptase-2 in regulating iron homeostasis was gained subsequently from several genome-wide association studies that all strongly linked serum iron, erythrocyte volume, and hemoglobin levels to common nonsynonymous coding single nucleotide polymorphisms in TMPRSS6 (Benyamin et al. 2009, Chambers et al. 2009, Ganesh et al. 2009, Tanaka et al. 2010).

Matriptase-2 is hepatocyte specific, and its expression is negatively regulated by iron, most likely at the level of mRNA translation (Hooper et al. 2003, Velasco et al. 2002, Zhang et al. 2010). This location and mode of regulation implies that the protease regulates hepcidin production by cleavage and inactivation of a substrate on the hepatocyte plasma membrane that is part of the BMP/SMAD pathway. Three lines of evidence strongly suggest that this substrate is hemojuvelin: (a) Hemojuvelin binds to the stem region of both the matriptase-2 zymogen and the active protease, (b) forced overexpression of matriptase-2 dramatically reduces cell surface hemojuvelin levels (Altamura et al. 2010, Silvestri et al. 2008), and (c) iron overload disease and low hepcidin levels are observed in hemojuvelin/matriptase-2 double-deficient mice as well as in BMP6/matriptase-2 double-deficient mice (Lenoir et al. 2011, Truksa et al. 2009).

Orphan Membrane-Anchored Serine Proteases

The unraveling of the cellular and developmental functions of membrane-anchored serine proteases has been rapid. Nevertheless, approximately half of the family members (HAT, DESC1, TMPRSS11A, HAT-like 4, HAT-like 5, TMPRSS2, TMPRSS4, MSPL, matriptase-3, polyserase-1, and tryptase γ1) remain orphans in the sense that no functions or substrates are known. HAT, TMPRSS11A, or TMPRSS2 deficiency in mice is not associated with an obvious spontaneous phenotype, which suggests subtle or partially redundant functions (Kim et al. 2006; K. Sales & T.H. Bugge, unpublished data). The effects of loss-of-function mutations still have not been determined for DESC1, HAT-like 4, HAT-like 5, TM-PRSS4, MSPL, matriptase-3, polyserase-1, and tryptase γ1, and these may prove immediately insightful. On the basis of structural similarities and overlapping expression, partial functional redundancies may be anticipated between members of the TMPRSS subfamily and between members of the HAT/DESC subfamily, but these redundancies may be challenging to delineate genetically owing to the tight clustering of their cognate genes.

Target Substrates

Some membrane-anchored serine proteases (hepsin, TMPRSS3, spinesin, testisin) have established physiological functions, but the substrates that are cleaved to exert these functions are as yet unknown. For still other proteases (prostasin and matriptase), multiple functions have been identified, but target substrates are known for only some of these. Connecting the cellular and developmental functions of this protease family to the cleavage of specific target substrates will be a particularly challenging enterprise because of the marked reliance of membrane-anchored serine proteases on microenvironmental factors for target substrate selection, the general promiscuity of proteases in purified systems, and the lack of stringent ex vivo or cell-based assays for protease substrate identification.

The authors thank Drs. Mary Jo Danton, Silvio Gutkind, Kelly Ten Hagen, Karin List, and Larry Wahl for critically reviewing this manuscript. The National Institute of Dental and Craniofacial Research intramural program supported this work. We apologize to our many colleagues whose research could not be appropriately cited owing to space constraints.