| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Laboratory of Molecular Oncology, Department for Human Genetics, K.U. Leuven and Flanders Interuniversity Institute for Biotechnology, B-3000 Leuven, Belgium
1Correspondence: Laboratory of Molecular Oncology, Department for Human Genetics, University of Leuven and Flanders Interuniversity Institute for Biotechnology, Gasthuisberg O/N 6, Herestraat 49, B-3000 Leuven, Belgium. E-mail: john.creemers{at}med.kuleuven.ac.be
 |
ABSTRACT |
|---|
 TOP ABSTRACT BACKGROUNDINACTIVATION OF PROPROTEIN...PERSPECTIVES FOR THE FUTUREREFERENCES |
|---|
Key Words: knockout mice • human patients • viral and bacterial infection • Alzheimer’s disease • inhibitors
|
BACKGROUND |
|---|
|
TOPABSTRACTBACKGROUNDINACTIVATION OF PROPROTEIN...PERSPECTIVES FOR THE FUTUREREFERENCES |
|---|
and entered a new era in 1990 with the discovery of the mammalian prototype furin (2
, 3)
. Examination of the primary sequences of many secretory proteins had shown that conversion of precursor to active protein by cleavage at basic motifs (usually R/K-R or R-X-X-R, where X can be any amino acid) is a common feature in nature. Predicted substrates included neuropeptides (e.g., enkephalin, dynorphin) peptide hormones (e.g., insulin, somatostatin), growth and differentiation factors (e.g., the bone morphogenetic protein/transforming growth factor ß (BMP/TGF-ß) superfamily), receptors (e.g., notch-1 and insulin receptor), enzymes (e.g., PCs, matrix metalloproteinases), adhesion molecules (e.g., 
chains of integrins and collagens), blood coagulation factors (e.g., von Willebrand factor, factor IX), plasma proteins (e.g., albumin, 1-microglobulin), viral coat proteins (e.g., HIV-1 gp160, influenza hemagglutinin), and bacterial toxins (e.g., anthrax protective antigen, diphtheria toxin). Extensive lists of precursor proteins and their cleavage sites have been collated (4
, 5)
.
Between 1967 and 1990 many attempts were made to biochemically purify these endoproteolytic, bioactivating enzymes, eventually leading to isolation of the type I and type II proinsulin-converting activities (6)
, which where later shown to correspond to PC1/3 and PC2. But the key to their eventual identification had come earlier from genetic complementation studies in yeast resulting in the isolation of the gene encoding kexin, a subtilisin-like serine protease (7)
. This protease activates the secreted yeast -mating factor and killer toxin proproteins by cleavage at dibasic sites. It was soon shown that kexin could also correctly cleave mammalian precursors such as proopiomelanocortin (POMC) and proalbumin (8
, 9)
, implying that this was the prototype of the long-sought eukaryotic proprotein processing enzymes.
Database searches identified the newly discovered protein furin (10)
as a mammalian ortholog of the subtilisin-like serine protease kexin (11)
. Since then, six other mammalian members of the family have been cloned, PC1 (also known as PC3), PC2, PACE4, PC4, PC6 (also known as PC5), and LPC (also known as PC7 or PC8) (12
, 13)
, in addition to homologs from many other eukaryotic species (14)
.
Members of the family share a common domain structure (Fig. 1
) consisting of a propeptide followed by a catalytic domain, a middle domain (also called HomoB or P-domain), and a carboxyl-terminal domain (or domains). The catalytic domain shows the highest sequence conservation (35% identity between all seven members), followed by the middle domain (15% identity), the propeptides and carboxyl-terminal domains show little homology.
|
A model for the catalytic domains of PCs has been proposed based on the crystal structures of the subtilisins and experimentally tested (15
, 16)
. A striking feature is the large increase in negative charges in the highly conserved substrate binding region of the PCs compared with the bacterial subtilisins, which explains their specificity for basic substrate segments. The middle domain is also required for catalytic activity, and deletions or mutation of only a single amino acid can inactivate the enzyme (17
, 18)
. It is predicted to form a ß-barrel structure, which has been proposed to either stabilize the catalytic domain or assist in conferring specificity to the enzyme (19
, 20)
.
The propeptide acts as an intramolecular chaperone, assisting in proper folding of the zymogen (21)
. With the exception of PC2, cleavage of the zymogen occurs autocatalytically in the endoplasmic reticulum and is a requirement for transport beyond this compartment (22
23
24)
. The propeptide remains associated with the enzyme, however, until a second internal cleavage step occurs in the TGN/endosomal system (21)
. The carboxyl-terminal domains have been shown to be important for intracellular trafficking of most PCs, but additional functions of the cysteine-rich regions (which contain a variable number of a cysteine motif; see ref 25
) remain to be discovered.
After discovery of the PCs, a major focus of the research in this field was to identify enzyme/substrate relationships in order to explain the existence of a relatively large family of PCs, several of which are expressed simultaneously in all cells. Initially, this was studied by in vitro colocalization and coexpression experiments. This approach proved most successful for the neuroendocrine members PC1 and PC2. Human proinsulin, for instance, is first cleaved by PC1 to generate 32-33 split proinsulin and subsequently by PC2 to generate mature insulin (26)
. More striking is the selective processing of substrates such POMC and proglucagon, where cell type-specific expression of PC1 and PC2 determine which end products are generated. In anterior corticotropes, where only PC1 is expressed at high levels, adrenocorticotropic hormone (ACTH) and ß-lipotrophin are the major cleavage products of POMC whereas in intermediate melanotropes, where both convertase levels are high, predominantly ß-endorphin and -melanocyte-stimulating hormone (-MSH) are produced (27
, 28)
. Similar differential processing is found for a few other substrates, which seems to suggest that substrate specificity is determined by the sequence of the cleavage site, a model that unfortunately is too simplistic. Although subtle differences between substrate specificities of PCs have been reported in in vitro experiments, this cannot explain the selectivity observed in vivo; the factor that needs to be taken into account is the cellular context. A clear example is given by PC6A and PC6B, the two splice variants of PC6. Catalytic and middle domains are identical, but unlike PC6A, PC6B contains a transmembrane anchor. Dramatic differences in substrate selectivity between the splice variants have been shown with lefty proteins, for instance, which were found to be cleaved exclusively by PC6A (29)
. Similarly, a soluble form of the ß-secretase proBACE was cleaved exclusively by furin whereas the membrane-anchored form could be cleaved by several additional PCs (30)
. This suggests redundancy of processing, frequently observed in coexpression studies and more recently confirmed in knockout mouse models.
Proprotein convertases and disease
Correct spatio-temporal activation of proproteins is obviously important to maintain homeostasis. However, bacterial and viral pathogens have developed ways to utilize the processing machinery of their hosts to their own advantage. In pathological conditions such as cancer and neurodegeneration, enhanced processing has frequently been observed to be a crucial factor.
Bacterial toxins
PCs play a pivotal role in the activation of three different classes of bacterial toxins. The first class, which includes Pseudomonas aeruginosa exotoxin A (PEA), diptheria toxin, botulinum neurotoxin, and tetanus neurotoxin, are single-chain A/B toxins, where the A subunit (toxic subunit) and B subunit (target binding subunit) are synthesized as a single polypeptide chain. The polypeptide is cleaved at the target cell surface or in an endosomal compartment by a PC but the subunits remain joined by a disulfide bond until entry of the A subunit into the cytoplasm (31)
. A subclone of CHO cells was isolated that was insensitive to PEA (and resistant to infection by some viruses). This cell line carries mutations that result in two inactive fur alleles, and transfection with furin can restore sensitivity to PEA (and viral infection) (32)
. Transfection of PACE4 into these cells can also restore infectivity of some viruses, but not sensitivity to PEA, implying that furin is the physiological PEA convertase (33)
. However, resistance to PEA is not simply a case of defective toxin processing; the receptor for PEA (2-macroglobulin receptor/low density lipoprotein receptor-related protein) is a PC substrate and its processing is required for surface expression (34)
. Administration of a PC inhibitor with high selectivity for furin to PEA-treated mice has recently been shown to increase survival rate (35)
. The importance of these experiments is twofold: 1) it demonstrates the feasibility of inhibiting furin sufficiently to block bacterial toxicity in vivo; equally important, 2) it shows that in vivo furin inhibition does not itself have severe toxic effects.
The second class of toxins are complex A+B toxins, where the A and B subunits are synthesized as separate polypeptide chains and assemble on the target cell surface to form the active toxin. PCs activate the B subunit after binding to the target cell. Anthrax toxin is an example of a complex A+B toxin consisting of two subunits: protective antigen (PA) and a toxic factor, either lethal factor or edema factor. Protective antigen binds to the anthrax toxin receptor and is cleaved at an R-K-K-R sequence to produce the activated form of PA, PA63, which heptamerizes to form the prepore before binding the toxic factor. The resulting complex is then internalized by receptor-mediated endocytosis. Upon entering an acidic compartment (probably the endosome), low pH triggers insertion of the prepore into the membrane and translocation of the toxic factor into the cytoplasm. The rate-limiting step in this process is cleavage of protective antigen at the cell surface (36)
. Furin has been shown to cleave protective antigen in vitro, although a furin-deficient CHO cell line, FD11, can also process protective antigen to a limited extent, probably via PACE4 (37)
.
The third class is pore-forming toxins such as aerolysin. Aerolysin is produced by the bacterium Aeromonas hydrophila. Proaerolysin is secreted as a dimer and binds to the glycosylphosphatidyl inositol anchors of membrane proteins on target cells. After receptor binding, cleavage by a PC on the cell surface allows association into a heptamer pore complex. Pores can cause cell lysis or, at lower concentrations, influx of calcium ions leading to apoptosis. In vitro and in vivo studies have shown that proaerolysin can be processed by furin or PACE4 (and to a lesser extent, PC6A) and that inhibition of processing can protect cells from aerolysin toxicity (38)
.
Viral infection
Many viral fusion proteins are synthesized as proproteins that are processed by PCs into two disulfide bond-linked subunits; the precursor itself is incapable of the conformational changes needed for membrane fusion. Usually the fusion peptide is located at the end of the lumenal domain of the membrane-anchored subunit and the other subunit is involved in binding to the target cell receptor. Membrane fusion is triggered by a conformational change in the fusion protein that exposes the fusion peptide and makes it available for interaction with the cell membrane. This conformational change may be triggered by low pH (in the case of viruses, which are internalized to the endosome) or by binding to a receptor (in the case of viruses, which fuse directly with the plasma membrane).
The HIV envelope protein gp160 requires cleavage at a highly conserved R-E-K-R sequence into gp120 and gp41 to gain fusion activity (39)
. Soon after the discovery of furin, it was demonstrated that it was capable of enhancing this conversion and that inhibitors of PCs could block fusion activity and viral replication in cultured cells (40)
. It was later shown that gp160 was correctly processed in the furin-deficient LoVo cell line (41)
. Cotransfection studies later revealed that PC6, LPC, and PACE4 were also capable of processing the glycoprotein (42)
. However, only furin, PC6, and LPC are expressed in the target lymphatic tissues and T lymphocytes and therefore are the most likely candidates (43
, 44)
.
Hemagglutinin (HA) of influenza virus A mediates binding of the virus to host cell receptors and fusion with the host cell. HA is synthesized as a precursor that is proteolytically cleaved at a variable sequence into two disulfide bond-linked subunits. Fusion activity is completely dependent on cleavage. Differences in the pathogenicity of influenza viruses have often been ascribed to differences in the processing site: nonpathogenic strains have monobasic sites cleaved by trypsin-like proteases secreted by restricted types of tissue whereas more virulent strains have cleavage sites containing multiple basic residues that are processed by the more ubiquitous members of the PC family (45)
. Some nonpathogenic isolates also have polybasic processing sites, but in these cases adjacent carbohydrate groups are thought to block access of the protease. Tissue tropism and pathogenicity in these viruses are determined primarily by the presence of the HA receptor and the ability of the target cell to cleave the HA precursor.
The fusion protein of measles virus is produced by processing of a precursor (F0) at a R-R-H-K-R sequence (46)
. Expression of correctly processed F-protein in Vero cells leads to membrane fusion and formation of syncytia. Mutant F-protein, which lacks the PC cleavage site but can still be cleaved by trypsin, induces syncytia formation only after treatment with trypsin even though it is expressed on the cell surface, indicating that fusogenic activity requires processing (47)
. This unprocessed F-protein can also form virus particles, but these are completely noninfective until activated by trypsin. Infections of trypsin-activated mutant viruses in mice were less severe than wild-type infections and spread only in regions of the lung where an endogenous protease, tryptase Clara, is secreted. This enzyme is also capable of activating respiratory viruses such as human influenza and Sendai virus.
Ebola virus envelope glycoprotein contains a stretch of basic amino acids that is highly conserved in all strains of the virus, except for the Reston subtype, which has low human pathogenicity. Cleavage at this basic motif generates the mature disulfide-linked GP1-GP2 spike protein. Wild-type virus contains only mature glycoprotein and has been shown in LoVo cells that furin is required for cleavage (48)
. Defective processing may provide an explanation for the lower pathogenicity of the Reston subtype; other studies show that mutations that abolish cleavage do not prevent assembly of the glycoprotein into virus and do not reduce infectivity in a cell culture system (49)
.
In conclusion, envelope glycoproteins from many pathogenic viruses require processing by PCs. Selective inhibition of this step may block infectivity, as was shown for human cytomegalovirus by the addition of inhibitor in a cell culture system (50)
.
Cancer
The involvement of proprotein convertases in tumorigenesis has been extensively reviewed (5)
and will be discussed here only briefly.
Early studies of furin and PC1 demonstrated it could be used as a marker to distinguish between small and non-small cell lung carcinomas, with elevated levels of furin found in non-small cell and of PC1 in small cell lung carcinomas (51
, 52)
. More recent studies have confirmed the link between elevated expression of PCs and enhanced malignancy in some tumors. Small cell lung carcinomas contain elevated levels of PC1 and PC2, which presumably process the various neuroendocrine peptides that these tumors secrete (53)
. Elevated expression of furin, PACE4, LPC, and PC1 has also been demonstrated in cell lines and tissue samples from breast tumors (54)
. Overexpression of PACE4 converts squamous cell carcinoma into the more malignant spindle cell carcinoma and is sufficient for malignant conversion of nontumorigenic keratinocytes (55)
. Furin expression levels correlate with invasiveness and metastatic potential in some tumor cell lines (56
, 57)
. Conversely, using inhibitors to reduce the activity of PCs in head and neck squamous cell carcinoma cells reduced tumorigenicity and invasiveness in vitro and in a skid mouse model (58)
. Inhibition of PCs in the colon carcinoma cell line HT29 (59)
and two astrocytoma cell lines (60)
resulted in decreases in invasiveness and cell growth and decreased vascularization of tumors in nude mice.
Increased PC expression is associated with enhancement of metastatic spread and tumor cell proliferation. Among other processes, metastatic spread requires remodeling of the extracellular matrix, a reduction in cell adhesion, and an increase in cell motility. PCs are involved in all these processes. First, PCs activate members of several groups of metalloproteinases involved in the tightly controlled process of ECM degradation (MMPs, matrix metalloproteinases; MT-MMPs, membrane-type MMPs; ADAMs, a disintegrin and metalloprotease, and ADAM-TSs, ADAM with thrombospondin motifs) (5
, 61)
. Furthermore, PCs modulate the cell adhesion and signaling capacity of integrins (62)
. Integrins are transmembrane receptors consisting of heterodimers of one and one ß subunit. Nine of the 18 known -integrin subunits have putative PC cleavage sites, several of which have been shown to be authentic (63)
. Integrins bind a variety of cell surface ligands (e.g., ICAM-1, VCAM-1) and components of the extracellular matrix (e.g., laminin 5, various collagens, fibronectin). Binding of receptor to ligand results in transfer of signal-to-cytoplasmic components that control cell migration, shape, growth, and survival.
PCs also control tumor cell proliferation directly by activating growth factors and receptors. TGF-ß positively regulates expression of furin and is a furin substrate (64)
. Proteolytic activation of TGF-ß and other family members is an important regulatory mechanism; enhanced processing by PCs increases tumor cell proliferation and has an immunosuppressive effect (65)
. Small cell carcinomas of the lung also produce autocrine growth factors that are activated by PCs (66
, 67)
whereas IGF-1-mediated protection from apoptosis in HT29 colon carcinoma cells requires PC processing of IGF receptor (59)
.
Neurodegeneration
Recently PCs have been linked to some neurodegenerative disorders via their direct or indirect role in the production of amyloidogenic peptides.
The principal component of senile plaques in Alzheimer’s disease is amyloid-ß (Aß), which is generated by proteolytic cleavage of its precursor by ß- and 
-secretases; ß-secretase cleaves at an internal K-L sequence, precluding further generation of Aß. ß-Secretase has been identified and named BACE (beta site amyloid precursor protein cleaving enzyme); it has been shown that cleavage of pro-BACE by PCs enhances its activity (68
, 69)
. The identity of -secretase has not been established unequivocally, but prime candidates are members of the ADAM family, many of which are activated by PCs (70)
. ADAM10 and ADAM17 (also called TACE [TNF- convertase]) are of particular interest since knockout mouse models of these ADAMs show impaired -secretase activity. Some cell lines derived from ADAM10 knockout mice display almost complete loss of -secretase activity whereas others retain full activity (70)
, providing a clear example of cell type-dependent redundancy.
A novel Alzheimer amyloid plaque component recently identified is CLAC (collagen-like Alzheimer amyloid plaque component) (71)
. This secreted protein is derived from the type II transmembrane precursor CLAC-P/collagen type XXV by cleavage at an R-I-A-R motif, which can be cleaved by recombinant furin. Although CLAC and its precursor are both capable of binding specifically to fibrillized Aß in vitro, only CLAC is detected in plaques.
Three rare autosomal dominant neurodegenerative diseases caused by aberrant PC cleavage have been described: familial amyloidosis of Finnish type (FAF), familial British dementia (FBD), and familial Danish dementia (FDD). FAF is caused by amyloid deposition of a fragment of plasma gelsolin (PG) in the peripheral and central nervous system in patients with a point mutation (D187N or D187Y) in PG. Mutant PG is unable to bind to or be stabilized by Ca2+, which makes it susceptible to cleavage at a PC cleavage site that is either never, or rarely, used in wild-type PG (72
, 73)
. Processing can be inhibited by blocking the convergence of the exocytic (PG containing) and endocytic (protease containing) recycling pathway. Furin is a good candidate for the protease, since it efficiently cleaves the PG mutant in vitro and is localized in the TGN/endosomal system.
FBD and FDD are both caused by mutations in the BRI2 gene and are characterized by unique clinical symptoms in addition to dementia (74
, 75)
. BRI protein (BRI-L in the FBD and BRI-D in the FDD patients) is a type II membrane protein that is cleaved at the carboxyl terminus at a PC cleavage site (K-G-I-Q-K-R) to release a 23 amino acid peptide. In FBD patients, the stop codon is mutated into an arginine codon resulting in an extension of 11 amino acids. In FDD patients a 10 nucleotide duplication at the carboxyl terminus also results in an extension of 11 amino acids, albeit of different sequence. Although the cleavage sites are identical to wild-type, cleavage efficiency is enhanced and, more important, the mutant peptides (named ABri and ADan) are fibrillogenic. Furin was found to cleave BRI and the two mutant forms most efficiently, although other PCs (including LPC) were capable as well.
|
INACTIVATION OF PROPROTEIN CONVERTASES |
|---|
|
TOPABSTRACTBACKGROUNDINACTIVATION OF PROPROTEIN...PERSPECTIVES FOR THE FUTUREREFERENCES |
|---|
Inhibitors
The first reported PC inhibitor was the substrate analog decanoyl-R-V-K-R-chloromethylketone (40)
, which inhibits all PCs with a Ki of 0.1-2 nM (76)
. Its chemical properties, however, preclude its in vivo use. Other peptide inhibitors include polyarginines, the most potent of which is nona-arginine with a Ki ranging from 40 nM for furin to 2.5 µM for PC1. PC2 was not inhibited by polyarginines (77)
. Hexa-D-arginine has been shown to protect cultured cells and mice from PEA toxicity as discussed earlier (35)
. The propeptides have been found to be potent, but only modestly selective inhibitors of PCs with Kis in the low nanomolar range (78
79
80)
.
Several protein-based inhibitors have been generated by bioengineering the reactive site of proteinase inhibitors into a basic motif. The best characterized and most specific inhibitor is the 1-antitrypsin variant 1-PDX in which the reactive site loop contains an R-I-P-R sequence (81)
. Furin is inhibited with a Ki of 0.6 nM, PC6 with a Ki of 2.3 nM, and all other PCs at a substantially higher Ki. Depending on the experimental conditions, 1-PDX can be used as a highly furin-specific inhibitor or as a general inhibitor for PCs and has been used successfully as a therapeutic for human cytomegalovirus infection in cell culture (50
, 82)
. A naturally occurring protein-based inhibitor is the ovalbumin-type serpin human proteinase inhibitor-8 (83)
. It contains two potential furin recognition sequences in its reactive site loop (R-N-S-R-C-S-R) and inhibits furin with an overall Ki of 53.8 PM, making it the most potent inhibitor described so far. The physiological relevance of this inhibition and its specificity for furin remain to be established.
ProSAAS and 7B2 are naturally occurring inhibitors specific for PC1 and PC2, respectively. The carboxyl-terminal part of 7B2 inhibits PC2 with high potency (Ki 57 nM) (84)
; more important, the amino terminus is essential for PC2 activation (85)
. Binding of 7B2 to PC2 occurs in the ER and inactivation occurs by two sequential cleavages: the first by furin or another PC in the TGN, the second by PC2 itself in the secretory granule (86
, 87)
. The granin-like protein proSAAS was recently identified as an endogenous inhibitor of PC1 (Ki in the low nanomolar range for peptides spanning the inhibitory sequence) (88)
but, unlike 7B2, it is not required for PC1 activity or increased stability (89)
.
PC-deficient cell lines
Two methods have been used to study PC deficiency in tissue culture cell lines: generation/isolation of knockout cell lines and antisense knock-down experiments. Several cell lines harboring inactivating mutations in the furin gene have been described, the most frequently used being LoVo cells, a human colon carcinoma cell line (90)
, and RPE.40 cells, derived from Chinese hamster ovary CHO cells (32)
. Transfection experiments in these cell lines have provided insight about the necessity or redundancy of furin in the processing of numerous substrates. The limitation of this approach is that substrates are expressed in a heterologous system whose molecular microenvironment may not be physiologically representative. Antisense knockdown experiments can overcome these limitations, but efficient suppression of the targeted PC is hard to confirm due to low endogenous expression levels. It has been used successfully in e.g. HepG2 cells to demonstrate processing redundancy of proalbumin by furin, PACE4, and LPC and in primary cultures of rat sympathetic neurons to demonstrate processing of proneuropeptide Y by PC2 (91
, 92)
.
Animal models and human patients
With the exception of PC6A, all PCs have now been individually inactivated in mice, and analysis indicates redundancy in many cases and specific PC–substrate pairs in others (Table 1
).
|
Today, furin is the most extensively studied PC and is considered the workhorse of the family; it is expressed in virtually all cells and is capable of cleaving an impressive number of substrates in vitro. Furin is expressed early in the developing mouse embryo and is highly expressed at E8.5 in tissues including the heart tube, the gut endoderm, lateral plate mesoderm, allantois, and the notochordal plate. Functional inactivation of the fur gene in mice results in embryonic lethality at E10.5-11.5. Knockout embryos show multiple defects including failure of chorioallantoic fusion, abnormal yolk sac vasculature, lack of axial rotation, and severe ventral closure defects (93)
.
Failure of the allantois to fuse with the chorion and form the placenta is found in mice lacking the cell adhesion molecules VCAM1 and its receptor 4-integrin, the latter being a potential PC substrate. However, chorioallantoic fusion is abolished in BMP-5 and BMP-7 double null mice, and other aspects of the phenotype can be explained by defects in TGF-ß or TGF-ß-related (BMPs, nodal, lefty) signaling. For example, abnormalities in blood vessel formation of the yolk sac are similar to those found in TGF-ß1 knockout mice (94)
, which display defective yolk sac vasculogenesis although blood islands and endothelial precursors form. Several members of the TGF-ß superfamily have been shown to be processed by PCs in vitro and in vivo and, as some BMPs are highly unstable after cleavage of the propeptide, processing is an important modulator of activity (95
, 96)
. The ventral closure defects in fur-/- mice are most noticeable in the developing heart tube, which fails to either fuse or undergo looping morphogenesis, and in the intestine, where the sheet of endoderm that forms the primitive gut fuses only rostrally and caudally, leaving the midgut region open. These closure defects are a result of absent or delayed fusion of the definitive endoderm. In wild-type embryos, furin expression levels are particularly high in definitive endodermal cells in regions where fusion is occurring. Fur null embryos also show reduced numbers of cardiomyocytes, and it has been shown that TGF-ß/BMP-related signals are required for specification of cardiac lineages (97)
.
Although the phenotype of the PACE4 knockout mice shows some similarities to the fur-/- mice, only 
25% of homozygous knockout embryos die prenatally and, at a later stage of development, between E13.5 and E15.5. Again, the embryos have cardiac malformations that probably lead to death, but the heart defects are much more varied. The heart tube may loop in the wrong direction and have fused chambers or defective connections to the vascular system such as truncus arteriosus. Situs defects are found in other internal organs including the liver, lungs, pancreas, spleen, and gut. These left/right axis defects are preceded at e8.5 by abnormal expression of the axis determining genes nodal and lefty, the products of which are both PC substrates. These genes are normally asymmetrically expressed and involved in determining the left-right axis, but in PACE4-/- embryos expression can be absent or bilateral. Expression of nodal in the lateral plate is partially regulated by a positive feedback mechanism, so loss of expression can be explained as an effect of impaired activation by PACE4 (98)
.
Some embryos deficient in PACE4 show severe craniofacial defects such as cyclopism and truncation of the head, demonstrating that PACE4 activity is necessary for anterior patterning. This aspect of the phenotype may also be explained by a reduction in nodal processing as nodal has been shown to be required for correct patterning of the anterior-posterior axis (99)
.
The defects found in fur-/- and PACE4-/- mice partly overlap with the phenotypes of knockout models of specific TGF-ß/BMP members; however, both are less severe than any of these candidate substrates, indicating a (limited) degree of redundancy. It may be that other PCs such as LPC, PC6A, and PC6B are also involved in processing of these substrates. This is further supported by in vitro studies demonstrating that BMP-4 can be cleaved by furin, PACE4, PC6B, and LPC (100)
.
Functional inactivation of PC6B results in embryonic lethality at E10.5-11.5 (Y. Lu, A. Franzusoff et al., unpublished results), although no details are yet available. PC6 is locally expressed at high levels during early stages of development; the highest expression was observed at e7.5-8.5 in some tissues including the extraembryonic primitive endoderm, the amnion, the nascent mesoderm, and the yolk sac and between e9.5-11.5 in the apical ectodermal ridge or limb buds (96
, 101)
. Staining is prominent at e8.5 in the most posterior somite; staining moves caudally as the embryo develops, suggesting a role in somatogenesis, a process known to involve a variety of PC substrates such as FGFs, PDGF, Notch, and BMP-4. Based on overlapping expression patterns, PC6 might play a role in processing of BMP-2 and BMP-7 (96)
. PC6 has been implicated in the processing of lefty proteins, although isoform A appeared to be the prime candidate (29)
. Characterization of the lethal phenotype will reveal which of these substrates require PC6B and for which there is redundancy.
LPC appears to have a redundant, nonessential function, as its absence does not result in any detectable phenotype. The knockout mice are born in Mendelian proportions, and appear perfectly healthy and fertile (B. Young, personal communication). Histochemical analysis of internal organs has not revealed any obvious abnormalities. Candidate substrate processing analysis so far has been performed only for pro-neurotensin/neuromedin N, which is unaffected (102)
. The absence of a phenotype is surprising in light of its ubiquitous expression at all development stages and in adult tissues including brain, lung, liver, kidney, spleen, and thymus (96
, 103)
. In in vitro experiments LPC is capable of cleaving a number of substrates with variable cleavage sites (104)
. Taken together, these preliminary data suggest that LPC is either involved in the processing of a set of nonessential substrates or is active in cells where sufficient processing redundancy is present.
A similar observation is made in the conditional furin knockout (A. Roebroek et al., unpublished results) where tissue-specific inactivation of furin results in only a mild phenotype that is unmasked only after specific functional studies or analysis of the proteome. Analysis of substrates revealed mainly increased levels of precursor without a severe effect on the amount of processed product.
PC4 expression is restricted to ovarian and testicular germ cells and PC4 knockout mice were found to have reduced fertility. Sperm from male mice were less able to fertilize eggs in vitro; eggs that were fertilized were not viable, failing to develop to the blastocyst stage (105)
. The ovaries of female PC4 knockout mice show delayed folliculogenesis, which may contribute to the reduced fertility (106)
. One PC4 substrate has been identified that may largely explain the phenotype. Pituitary adenylate cyclase-activating polypeptide (PACAP) is expressed in many tissues, but highest expression is found in the brain and testis. In ovaries, PACAP induces production of steroid hormones in granulosa cells; in testis it induces testosterone production. High levels of PACAP are present in the cap and acrosome of spermatids, suggesting a function in fertility. Mature PACAP consists of two peptides, PACAP38 and PACAP27, which are produced by PC cleavage of proPACAP. In neuronal cells PC2 and PC1 appear to be the relevant convertases (107)
, but testis and ovaries from PC4 null mice contain no processed PACAP, indicating that in germ cells processing is performed by PC4 (108)
. Indeed, female PACAP knockout mice show a decrease in fertility; however, this appears to be mainly due to a decrease in mating frequency and only to a limited extent to reduced litter size (109)
. The PC4 null mice show larger reductions in litter size, suggesting that additional substrates contribute to the phenotype. Sperm/egg interactions involve a number of other potential PC substrates, including the ADAMs fertilin- and -ß (110)
and their receptor integrin-6. Another potential substrate is the testis-specific isoform of angiotensin-converting enzyme (111)
, where male knockout mice have reduced fertility (112)
.
Two animal models are available that lack activity of the neuroendocrine-specific enzyme PC2: the PC2 knockout and the knockout of its chaperone/inhibitor, 7B2. PC2 knockout mice appear normal at birth and growth postpartum is only slightly retarded in contrast to PC1/3 nulls. They have chronic hypoglycemia and show reduced elevation of blood glucose in response to peritoneal glucose infusion (113)
. The effects on blood sugar levels suggested reduced circulating levels of glucagon. Analysis of pancreatic and circulating islet hormones showed a severe reduction in the levels of mature insulin, glucagon, and somatostatin and increased levels of precursors. Electron microscopic analysis of islet endocrine cells showed increased numbers of immature secretory granules, confirming the blockade in processing (114)
.
Although processing of proinsulin occurred with reduced efficiency, the production of glucagon in cells and somatostatin in 
cells was essentially absent. The residual processing of proinsulin is due to the presence of PC1, which is also expressed in ß cells and cleaves proinsulin efficiently at one of the two sites and inefficiently at the second. In and cells PC1 is not expressed; thus processing of proglucagon to glucagon and prosomatostatin to somatostatin is entirely dependent on PC2. An unresolved issue is the observation that in the PC2 knockout mice prosomatostatin is normally processed at the monoarginine site to generate SS-28 (113)
. The enzyme responsible for this cleavage remains to be identified.
Histochemical analysis of PC2 knockout mice showed normal islets at birth but dramatic increases in and cell mass by 3 months. This increase was due to increases in cell number as well as individual cell size. Decreased circulating glucagon or glucose levels or the increased relative insulin level probably causes this hyperplasia. Hyperplasia of ß cells can be caused by hyperglycemia or, through a negative feedback mechanism, by hypoinsulinemia; it is likely that similar mechanisms control and cell mass. Glucagon replacement therapy fully restored normoglycemia and corrected cell hyperplasia by induction of apoptosis (115)
.
Other studies have shown that processing of POMC is blocked at a number of sites whereas other sites are processed correctly, presumably by PC1, leading to the accumulation of ACTH and ß-lipotrophin (116
, 117)
. Anterior pituitary corticotropic cells, which contain mainly PC1, normally process POMC to produce ACTH. Intermediate pituitary melanotropes, which have higher levels of PC2, process ACTH further to -MSH and CLIP; this processing is inhibited in PC2 null mice.
Processing of proislet amyloid polypeptide in ß cells is partially blocked (118)
and the brains of PC2 knockout mice accumulate precursors of enkephalin (119)
, cholecystokinin (120)
, melanin-concentrating hormone (121)
, and orphanin FQ/nociceptin (116)
, as well as precursors of dynorphin that are unprocessed at monobasic sites (122)
, implicating PC2 in all these processes.
Processing of proneurotensin/neuromedin N (proNT/NN) provides an interesting example of redundancy (102)
. Maturation of proNT/NN to NT and NN was reduced in whole brain extracts from PC2 knockout mice by 15 and 50%, respectively. However, immunocytochemistry showed differences in the efficiency of maturation in different regions of the brain, implying some degree of region-specific redundancy in convertase activity. Colocalization and in vitro coexpression studies suggest that this redundancy is provided by PC1 or PC6A.
The 7B2 knockout mouse has no detectable PC2 activity and shows the same defects in islet hormone processing and islet morphology as the PC2 knockout mouse, but in addition has increased ß cell mass (85)
. These mice die within 9 wk of birth from a severe form of Cushing’s disease, a syndrome caused by elevated circulating levels of ACTH. The mice are similar to the PC2 knockout mice with respect to their elevated production of ACTH in melanotropic cells, but differ in their dramatically increased ACTH secretion. Hypersecretion may be caused by dopaminergic deficiency in 7B2 null mice as dopaminergic input represents the primary inhibitory control of peptide release from the intermediate lobe. This deficiency may be a direct result of 7B2 inactivation or, indirectly of the high corticosterone production (117)
. Taken together, this indicates that 7B2 has functions in additional to its role as a chaperone/inhibitor of PC2, a conclusion that is supported by the observation that 7B2 is sometimes expressed in tissues that do not express PC2 (123)
. Furthermore, processing of the PC2 substrate progastrin is more severely affected in the 7B2 knockout mice (124)
.
PC1 is the only member of the family for which deficiency has been described in both mouse (two independent models) and human with remarkable differences in phenotype. In one mouse model (125)
, 40% of PC1 null embryos die before birth, and another 40% within 6 days postpartum. The remaining pups appear normal at birth but are only 60% of the size of heterozygous or wild-type littermates after 6 wk and suffer from chronic mild diarrhea associated with bulky moist stools. PC1 null mice show decreased levels of growth hormone (GH) mRNA and decreased circulating GH; GH-secreting cells, the anterior pituitary somatotrophs, are shrunken and inactive. This is consistent with the low or undetectable circulating growth hormone-releasing hormone (GHRH) along with increases in GHRH mRNA and unprocessed proGHRH. The phenotypic dwarfism is reminiscent of mice homozygous for an inactivating mutation in the GHRH receptor (126)
. Although circulating corticosterone levels are normal, the mice have increased levels of POMC mRNA and unprocessed POMC, whereas ACTH biosynthesis was undetectable in the pituitary. Processing of proglucagon to GLP1 and GLP2 in the L cells in the gut is also completely inhibited. Blood glucose levels are normal despite a severe block in proinsulin processing, which results in accumulation of immature secretory granules in the pancreatic ß cells (127)
.
Targeted inactivation of PC1 in the second mouse model resulted in embryonic lethality (102)
. The reason for this remarkable discrepancy with the other model is unknown, but differences in targeting strategies could be the answer. In the first model, <1 kb is deleted, encompassing the signal peptide-encoding first exon and several upstream transcriptional control elements (125)
. No PC1 mRNA or PC1 immunoreactive protein was detected in the knockout mice. The second model was generated by deletion of exons 3-7 that encode part of the catalytic domain and span a genomic region of 15 kb (102)
. This relatively large deletion might affect transcription of genes other than PC1. Alternatively, the remaining exons might encode a toxic protein, although the unaffected heterozygotes make this explanation unlikely.
The human syndrome associated with PC1 deficiency is comprised of extreme childhood obesity, abnormal glucose homeostasis, hypocortisolism, amenorrhea, and impaired gastrointestinal function (17
, 128)
. The first patient, a 50-year-old Caucasian woman, was reported to secrete extremely high levels of proinsulin but no significant amounts of insulin. The reduction in insulin-like activity of proinsulin appears to be compensated for by its high circulating concentrations whereas its prolonged half-life explains the postprandial hypoglycemia. Impaired processing of POMC to ACTH probably underlies the probands impaired adrenal function. Ovulation could be induced by administration of gonadotropins, suggesting that hypogonadotrophic hypogonadism arises from impaired processing of gonadotropin-releasing hormone or upstream factors. The molecular mechanism causing obesity in this syndrome is unknown, but there are a number of candidate anorexogenic substrates (e.g., -MSH, GLP-1, CART). However, processing of proCART (cocaine-amphetamine-related transcript), a downstream target of leptin that suppresses appetite, does not appear to be affected in PC1 knockout mice (127)
. Abnormal gastrointestinal function was studied only after identification of the second patient, a Caucasian female infant that died at the age of 18 months (R. S. Jackson et al., unpublished results). Both patients displayed marked intestinal dysfunction and impaired processing of gut-derived prohormones, although this aspect of the syndrome was more severe in the latter patient, requiring total parenteral nutrition throughout her life. Inactivation of the PC1 gene in both patients was the consequence of unique compound deleterious mutations. No residual activity was found in in vitro analysis of all four mutant alleles, suggesting that both patients are true PC1 nulls.
Although there is clear overlap of the human syndrome and the first mouse model, there are also distinct differences. In both cases. proinsulin, POMC to ACTH, and proglucagon to GLP1 and GLP2 processing are impaired. On the other hand, obesity and amenorrhea are unique for human, and dwarfism for mice. Processing of GHRH has not been determined in the patients, but growth hormone was in the low-normal range in the first patient (17
, 128)
.
|
PERSPECTIVES FOR THE FUTURE |
|---|
|
TOPABSTRACTBACKGROUNDINACTIVATION OF PROPROTEIN...PERSPECTIVES FOR THE FUTUREREFERENCES |
|---|
Further in vivo studies will undoubtedly extend the list of diseases for which PC inhibition might provide a therapeutic strategy. The knockout mice provide a powerful tool to examine the progression of diseases in the absence of the PC predicted to be relevant by in vitro experiments. In the case of lethal general knockouts, this will require the generation of new conditional mouse models. The viability of tissue-specific furin knockout mice indicates that embryonic indispensability does not preclude (limited) redundancy during adult life (A. Roebroek et al., unpublished results). For the animals with mild or undetectable phenotypes, analysis of the proteome or crossing with other knockout strains might provide new insights into redundancy.
Inhibitor design will remain an important area of research. The availability of crystal structures of different convertases will facilitate the rational design of specific inhibitors. All inhibitors described to date are substrate analogs, and with the overlapping specificity of all PCs it might be difficult to obtain inhibitors entirely specific for one PC using this approach. However, recent results with synthetic peptides and peptides derived from the naturally occurring inhibitors 7B2 and proSAAS look promising (77
, 88
, 129
130
131)
.
Another problem that can be addressed only in an animal model is the delivery of inhibitor to the correct site, in both the organism and in the cell. The tissue-specific redundancy revealed for some substrates implies it will be impossible to completely inhibit processing of such substrates. However, if the contributions of different convertases to the processing of, for instance, viral coat proteins in epithelial cells, are known, an inhibitor can be selected that will find a balance between maximizing viral inhibition and minimizing interference in normal physiological processes. A recent study where mice administered with PEA were treated with the PC inhibitor hexa-D-arginine suggests, at least for toxins activated on the cell surface, that this may not be a problem (35)
.
The 7B2 knockout mouse has provided an interesting counterpoint to the PC2 knockout mouse and it will be intriguing to compare the PC1 knockout with a proSAAS knockout mouse. ProSAAS has a broader expression pattern than PC1 and therefore might be expected to have additional functions (132
133
134)
. The evolution of specific inhibitors like 7B2 and proSAAS indicates that control of PC activity is a tightly controlled and dynamic process.
Animal models and human patients can provide complementary information, and animal models allow experiments that are impossible in humans. Animal models will provide clues to what to look for in patients. A comparison of the phenotypes of the PC1 knockout mouse and the human PC1-deficient patients, however, shows the danger of extrapolating results from animal models to humans and the importance of finding human subjects (17
, 102
, 125)
.
Differences between the PC1 null mouse and the human patients are intriguing, and show that redundancy of cleavage has allowed diversion between species. This is seen clearly in the case of proinsulin; rodents have two proproteins that differ from human insulin at all four cleavage sites. One of these is a substitution of lysine for methionine that changes the cleavage site from KSRR to MSRR and presumably alters the specificity (135)
. This should serve to emphasize that, no matter how informative animal models may be, investigation of human diseases requires human subjects.
|
ACKNOWLEDGMENTS |
|---|
Received for publication December 10, 2002. Accepted for publication March 10, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTBACKGROUNDINACTIVATION OF PROPROTEIN...PERSPECTIVES FOR THE FUTUREREFERENCES |
|---|
