Activation of furin requires autoproteolytic cleavage of its 83‐amino acid propeptide at the consensus furin site, Arg‐Thr‐Lys‐Arg107↓. This RER‐localized cleavage is necessary, but not sufficient, for enzyme activation. Rather, full activation of furin requires exposure to, and correct routing within, the TGN/endosomal system. Here, we identify the steps in addition to the initial propeptide cleavage necessary for activation of furin. Exposure of membrane preparations containing an inactive RER‐localized soluble furin construct to either: (i) an acidic and calcium‐containing environment characteristic of the TGN; or (ii) mild trypsinization at neutral pH, resulted in the activation of the endoprotease. Taken together, these results suggest that the pH drop facilitates the removal of a furin inhibitor. Consistent with these findings, following cleavage in the RER, the furin propeptide remains associated with the enzyme and functions as a potent inhibitor of the endoprotease. Co‐immunoprecipitation studies coupled with analysis by mass spectrometry show that release of the propeptide at acidic pH, and hence activation of furin, requires a second cleavage within the autoinhibitory domain at a site containing a P6 arginine (‐Arg70‐Gly‐Val‐Thr‐Lys‐Arg75↓‐). The significance of this cleavage in regulating the compartment‐specific activation of furin, and the relationship of the furin activation pathway to those of other serine endoproteases are discussed.
Biosynthesis of proteins destined for residence within, or routing through, the secretory pathway requires an orchestrated series of events, including one or more proteolytic cleavages to yield the mature and functional molecule. The primary translation product for many bioactive proteins (e.g. growth modulators, receptors, peptide hormones and proteinases) contains, in addition to the active region, a propeptide, that is typically an N‐terminal extension of the mature protein. The functions of propeptides appear to be manifold. In the case of several precursors of growth modulators (e.g. nerve growth factor and amphiregulin) the propeptides are required for stability or secretion of the active factor (Suter et al., 1991; Thorne and Plowman, 1994), whereas the propeptide of transforming growth factor‐β participates in formation of latent complexes (Miyazono et al., 1988). The propeptide of the yeast enzyme carboxypeptidase Y is required for the receptor‐mediated sorting to the vacuole (Valls et al., 1990). Finally, propeptides are important for the activation of all classifications of proteinases (i.e. serine‐, aspartyl‐, cysteinyl‐ and metallo‐proteinases; reviewed in Baker et al., 1993).
Perhaps the most thoroughly investigated examples of propeptide‐mediated proteinase activation are those of the bacterial serine proteinases, subtilisin and α‐lytic protease. The N‐terminal propeptides of both enzymes are required for the correct folding of their catalytic domains and can facilitate the refolding of denatured enzyme in vitro (Power et al., 1986; Ikemura and Inouye, 1988; Silen et al., 1989; Zhu et al., 1989). Following translation of the nascent chain into the periplasmic space and folding of the zymogen, the N‐terminal propeptides of both bacterial endoproteases are cleaved by an intramolecular reaction (Power et al., 1986; Silen et al., 1989). The propeptides remain associated with the catalytic domains through non‐covalent interactions and act as potent autoinhibitors (Baker et al., 1992; Li et al., 1995). Structural and biochemical analyses have shown that the subtilisin propeptide binds to the enzyme primarily through multiple non‐polar interactions, with the C‐terminus extending into the active site of the enzyme, thus acting as a competitive inhibitor (Li and Inouye, 1994; Bryan et al., 1995; Gallagher et al., 1995; Li et al., 1995). When subtilisin E propeptide is degraded, by an as yet uncharacterized pathway, the enzyme is free to act on substrates in trans (Ikemura and Inouye, 1988).
The bacterial subtilisins are evolutionarily related to the eukaryotic proprotein convertases, a family of calcium‐dependent serine endoproteases. The convertases catalyze the proteolytic maturation, and hence activation, of precursor proteins within the secretory pathway by cleavage at oligo‐basic amino acid sequences (Steiner et al., 1992; Smeekens, 1993). Members of the proprotein convertase family include the yeast convertase Kex2p, which catalyzes the activation of α‐mating pheromone, and a number of Kex2p homologs expressed in metazoa including furin, PC1/3, PC2, PC4, PC5/6, LPC/PC7/8 (hereafter termed PC7) and PACE‐4 (Bruzzaniti et al., 1996; Meerabux et al., 1996; Seidah et al., 1996; for reviews, see Steiner et al., 1992 and Smeekens, 1993).
Like the bacterial subtilisins, the proprotein convertases are synthesized as inactive precursors that require proteolytic cleavage of their N‐terminal propeptides. This was first established for furin, a type I membrane protein concentrated in the TGN/endosomal system (Molloy et al., 1994; Jones et al., 1995; Schafer et al., 1995; Takahashi et al., 1995; Voorhees et al., 1995) that cleaves a large number of proprotein molecules at the consensus furin site ‐Arg‐X‐Lys/Arg‐Arg‐ in both the biosynthetic and endocytic pathways (for reviews, see Bresnahan et al., 1993; Van de Ven et al., 1993). The furin zymogen undergoes autocatalytic cleavage of its 83‐amino acid N‐terminal propeptide at the C‐terminal side of the consensus furin site ‐Arg‐Thr‐Lys‐Arg107↓ soon after deposition of the molecule into the RER (t½ <10 min) (Leduc et al., 1992; Molloy et al., 1994; Vey et al., 1994; Creemers et al., 1995). Not only is the RER‐localized propeptide cleavage a necessary step for activation of furin, it is also required for export of the protein from this compartment (Molloy et al., 1994; Creemers et al., 1995). However, this cleavage alone is not sufficient for enzymatic activation of furin. Rather, furin activation also requires exposure of the endoprotease to post‐RER compartments (Molloy et al., 1994; Vey et al., 1994). For example, a furin construct unable to be exported from the RER undergoes autoproteolytic cleavage of its propeptide, but remains proteolytically inactive against substrates in trans (Molloy et al., 1994).
To begin identification of the additional steps comprising the furin activation pathway, we have developed a simple in vitro system to study this process under defined conditions. Here, using this approach, we report the discovery of an ordered series of steps required for the activation of furin. The relationship of these steps to the compartment‐dependent activation of furin in the secretory pathway is discussed.
To determine the requirements for the activation of furin, a series of epitope‐tagged furin constructs was used (Figure 1). Introduction of the FLAG epitope tag immediately C‐terminal to the propeptide cleavage site permits the use of immunologic methods to monitor propeptide cleavage and correlation of this event with enzyme activity. Importantly, the FLAG tag at this position does not affect propeptide cleavage or enzyme activity measurably in vivo (Molloy et al., 1994). By employing two FLAG peptide‐specific antibodies, mAb M1 and mAb M2, the zymogenic form of furin (which reacts with mAb M2 only) can be distinguished from the mature forms (i.e. propeptide‐cleaved) of furin generated by autoproteolytic cleavage C‐terminal to Arg107 (reacts with both mAb M1 and mAb M2).
In initial studies, the importance of propeptide cleavage in furin activation was determined (Figure 2). Replicate plates of BSC‐40 cells were either mock‐infected (lane 1), or infected with wild‐type vaccinia virus (lane 2), with vaccinia recombinants expressing the TGN localized fur/f (lane 3), or one of two furin constructs that are concentrated in the RER; a truncated soluble form of furin containing the RER retrieval signal ‐Lys‐Asp‐Glu‐LeuCOOH, fur/fΔtc‐k (lane 4), or the active site mutant fur/fD153N (lane 5). Analysis of crude membrane preparations by Western blot (Figure 2A and B) showed that both fur/f and fur/fΔtc‐k had undergone the autoproteolytic propeptide cleavage (both recognized by mAb M2 and mAb M1) whereas fur/fD153N failed to undergo this step (reactive with mAb M2 only). Each of the samples was then assayed for furin enzymatic activity using the Q1 internally quenched peptide substrate (Figure 2C). Only the membrane sample from cells expressing fur/f was capable of efficiently cleaving the substrate.
The inability of the RER‐localized fur/fΔtc‐k to cleave in vitro the synthetic furin substrate, under conditions in which the TGN‐localized fur/f construct could, is in agreement with our previous work demonstrating that propeptide cleavage is necessary but not sufficient for furin activation (Molloy et al., 1994). Additional factors, including cellular calcium stores, are also important for furin activation. The concentration of available calcium in the RER versus the TGN is not firmly established (see Discussion); however, one measured difference between these compartments is pH. The pH of the RER is neutral (Mellman et al., 1986) whereas that of the TGN is ∼6.2 (Seksek et al., 1995). To determine directly whether calcium and/or the acidification of the TGN contribute to furin activation, fur/fΔtc‐k in crude membrane preparations was measured with or without pre‐incubation in the presence or absence of 5 mM calcium under neutral (pH 7.5) or acidic (pH 6.0) conditions (Figure 3). Consistent with the data in Figure 2, fur/fΔtc‐k failed to be activated at neutral pH, irrespective of the presence of calcium. By contrast, exposure of fur/fΔtc‐k to a TGN‐like acidic and calcium‐containing environment resulted in a striking increase in enzyme activity. Importantly, the activation of fur/fΔtc‐k required a pre‐incubation under these conditions (compare lanes 1 and 6 with lane 5).
The pH optimum for activation of fur/fΔtc‐k was determined using a range of buffers (Figure 4A). Maximal activation occurred at pH 6.0. A time‐course of fur/fΔtc‐k activation at pH 6.0 was next determined (Figure 4B). Under these conditions enzyme activity increased linearly for ∼180 min.
The results in Figures 2 and 3 demonstrated a requirement for calcium and an acidic pH for the activation of furin following the autoproteolytic cleavage of its 83‐amino acid propeptide. However, because furin undergoes an autoproteolytic cleavage of its N‐terminal propeptide in the RER, the endoprotease must be intrinsically active. These findings suggest that the acidic environment of the TGN facilitates the removal of a furin inhibitor. To examine this possibility, a crude membrane fraction was prepared from BSC‐40 cells expressing fur/fΔtc‐k. An aliquot of the membrane sample was then subjected to limited digestion with trypsin to facilitate degradation of the furin inhibitor, followed by an assay for furin activity (Figure 5). A sequential treatment of the sample with trypsin followed by addition of soybean trypsin inhibitor (to block trypsin selectively) resulted in a large increase in furin activity compared with control samples (compare column 3 with columns 1 and 2). These results are consistent with the hypothesis that fur/fΔtc‐k is inactive at neutral pH because of its association with a trypsin‐labile inhibitor.
Because the N‐terminal propeptides of the structurally related bacterial subtilisins are potent autoinhibitors, we examined the ability of the furin propeptide to inhibit furin's activity in vitro. A GST fusion protein containing residues encoding the entire furin propeptide (Gln25→Arg107), GSTpro, was prepared. Increasing amounts of the purified fusion protein were then pre‐incubated for 1 h with furin enzyme in vitro prior to addition of an AMC substrate (Figure 6). Furin activity was inhibited with near‐stoichiometric quantities of added propeptide. The inhibitory effect was specific for sequences within the furin propeptide since GST showed no inhibitory activity.
Together, the results in Figures 2 and 6 argue that the furin propeptide functions as a tightly bound autoinhibitor and suggest that removal of this fragment requires exposure of the complex to an acidic pH. Therefore, to monitor the fate of the furin propeptide during enzyme activation, a dual epitope tag construct was generated (Figure 1). In this construct, fur/f/haΔtc‐k, the hemaglutinin (HA) epitope tag was inserted by loop‐in mutagenesis C‐terminal to the predicted signal peptidase cleavage site (Ala24↓Gln25; see Figure 1). Western blot analysis and time‐course of activation studies showed that fur/f/haΔtc‐k underwent correct propeptide cleavage and was activated with similar kinetics as fur/fΔtc‐k (data not shown). To show that the HA tag did not affect the furin inhibitory properties of the propeptide, a second GST fusion protein, GSTpro/ha, was constructed containing the HA‐tagged furin propeptide. Enzyme inhibition studies showed that GSTpro/ha attenuated furin activity at concentrations very similar to GSTpro (Figure 6).
The potent autoinhibitory properties of the furin propeptide suggest that the lack of activity of fur/fΔtc‐k when incubated at neutral pH may be the result of a tight‐binding interaction between the mature enzyme and the cleaved propeptide domain. To demonstrate directly the association of the furin propeptide with the enzyme in vivo, a co‐immunoprecipitation experiment was performed (Figure 7). Replicate samples of a crude membrane preparation from BSC‐40 cells expressing fur/f/haΔtc‐k were incubated at either pH 7.5 or pH 6.0 and the furin construct was immunoprecipitated with mAb M1 either immediately or following preincubation in either of the two buffers. Analysis of the mAb M1 immunoprecipitates by Western blot using the mAb 12CA5 showed that the HA‐tagged propeptide is associated with the mature enzyme (Figure 7A). At pH 7.5, the propeptide–enzyme complex remained stable during an extended incubation (Figure 7A, lane 2). By contrast, incubation of the complex at pH 6.0 resulted in release of propeptide from the endoprotease (Figure 7A, lane 4). The loss of co‐immunoprecipitating propeptide was coincident with a marked increase in furin activity (Figure 7B). Interestingly, Western blot analysis of the whole extract showed that the propeptide underwent a second proteolytic cleavage, generating an ∼6 kDa HA‐tagged peptide, that was coincident with its release from the enzyme and furin activation (Figure 7C).
Identification of the ∼6 kDa HA‐tagged propeptide fragment was accomplished using mass spectrometry (MS) (Figure 8). A crude membrane sample prepared from BSC‐40 cells expressing fur/f/haΔtc‐k was incubated at pH 6.0 and the extract fractionated by reversed‐phase HPLC. The column fractions containing the HA‐tagged ∼6 kDa band (determined by Western blot) were then subjected to MALDI–TOF MS and electrospray LC/MS. Analysis by MALDI showed that the ∼6 kDa HA‐tagged band was composed of three species with masses of: (i) 7032 ± 2.7 a.m.u.; (ii) 6878.3 ± 4.1 a.m.u.; and (iii) 6753.9 ± 3.3 a.m.u. Analysis by LC/MS allowed refinement of the mass measurements for each species to 7035.5 a.m.u., 6880.0 a.m.u. and 6752.0 a.m.u., respectively. Using mass spectrometric methods, we showed that the 83‐amino acid propeptide was cleaved to generate three distinct N‐terminal fragments with masses of 7036, 6880 and 6752 a.m.u. These masses are 17 a.m.u. less than the predicted HA‐tagged‐Gln25→Arg75, HA‐tagged‐Gln25→Lys74 and HA‐tagged‐Gln25→Thr73, respectively. However, because N‐terminal glutamine residues readily cyclize, generating products that are 17 a.m.u. smaller than predicted (Krishna and Wold, 1993), we conclude that the N‐terminus of profurin initiates at Gln25, which becomes cyclized to pyroglutamic acid (pGlu). Thus, these peptides were identified as: (i) HA‐tagged‐pGlu25→Arg75; (ii) HA‐tagged‐pGlu25→Lys74; and (iii) HA‐tagged‐pGlu25→Thr73. The identification of these three peptide shows that the internal cleavage of the furin propeptide occurs after the ‐Lys74‐Arg75‐ doublet and that a carboxypeptidase B‐like activity in the extract degrades slightly the C‐terminus.
Unlike the initial furin propeptide cleavage site (‐Arg‐Thr‐Lys‐Arg107‐) processed in the RER, the ‐Lys74‐Arg75‐ doublet is part of a site that lacks a P4 arginine typical of furin substrates, yet possesses a P6 arginine (‐Arg70‐Gly‐Val‐Thr‐Lys‐Arg75‐). This kind of cleavage site is also present in a subset of furin substrates preferentially cleaved at acidic pH (Brennan and Nakayama, 1994a,b). To determine the importance of cleavage of the propeptide at Arg75 in the activation of furin, as well as to assess the contribution of Arg70 to this step, site‐directed mutagenesis was employed. Replicate plates of BSC‐40 cells were infected with vaccinia recombinants expressing either fur/f/haΔtc‐k or constructs containing either Arg75→Ala (R75A:fur/f/haΔtc‐k) or Arg70→Ala (R70A:fur/f/haΔtc‐k) mutations. Crude membrane preparations from each sample were incubated in pH 7.5 or 6.0 buffers, or subjected to limited trypsin proteolysis at pH 7.5 to release the autoinhibitory action of the propeptide (Figure 9). Each construct was proteolytically inactive following incubation at pH 7.5. After incubation at pH 6.0, activation of fur/f/haΔtc‐k was observed. By contrast, furin activity was suppressed in samples containing R75A:fur/f/haΔtc‐k or, to a slightly lesser extent, R70A:fur/f/haΔtc‐k. Following limited trypsinization at neutral pH, however, each of the constructs could be activated. These data show that acid pH‐dependent cleavage of the ‐Lys74‐Arg75‐ doublet is necessary for activation of furin. Furthermore, efficient cleavage at this site requires a P6 arginine (Arg70).
Earlier studies revealed that activation of furin requires one or more steps subsequent to the initial cleavage of its N‐terminal propeptide (Rehemtulla et al., 1992; Molloy et al., 1994; Vey et al., 1994; Thomas et al., 1995). Here, using an in vitro approach, we identify a sequence of events necessary for furin activation. Following translocation of furin into the RER, and concomitant removal of the signal sequence, the proregion is autoproteolytically cleaved at ‐Arg‐Thr‐Lys‐Arg107↓ (Leduc et al., 1992; Molloy et al., 1994; Vey et al., 1994; Creemers et al., 1995). In the neutral environment of the RER, the propeptide remains associated with the endoprotease and functions as an autoinhibitor. Exposure of the furin–propeptide complex to a mildly acidic (pH 6.0) and calcium‐containing (low millimolar) environment characteristic of the TGN results in a second cleavage within the propeptide at ‐Arg70‐Gly‐Val‐Thr‐Lys‐Arg75↓ as determined by mass spectrometry. Mutational analyses showed that both the P1 (Arg75) and P6 (Arg70) residues are essential for the acid pH‐dependent activation of furin. Following the internal cleavage of the propeptide, the fragments dissociate from furin, permitting the enzyme to cleave substrates in trans.
The furin activation pathway determined using the in vitro approach reported here is consistent with earlier studies, performed in vivo, that described the requirements for activation of this endoprotease (Molloy et al., 1994; Vey et al., 1994). Not only is propeptide cleavage upon entry into the RER a necessary step for activation of the enzyme, it is also a requirement for subsequent export of the membrane‐tethered protein (Molloy et al., 1994; Creemers et al., 1995). This requirement suggests that in vivo cleavage at Arg107 is necessary for folding of the molecule into a transport‐competent form or perhaps facilitates the generation of furin oligomers. None the less, this maturation step, which requires calcium and is sensitive to reducing agents, is alone not sufficient for furin activation. Furthermore, treatment of cells with either brefeldin A or monensin show that furin activation requires exposure to a late Golgi compartment (Vey et al., 1994). Our results are consistent in showing the importance of a mildly acid pH to the second site cleavage at ‐Arg70‐Gly‐Val‐Thr‐Lys‐Arg75↓‐ and activation of furin. Indeed, by comparing the relative activity of furin constructs containing deletions of the enzyme's cytoplasmic tail, full activation of furin was found to depend on the correct localization and routing within the TGN/endosomal system (Molloy et al., 1994). Whether the need for the correct localization of furin to the TGN is important specifically for the time‐dependent internal cleavage, conformational changes in the endoprotease and/or the propeptide, or secondary modification of the protein is not known. However, we have observed that truncated forms of furin are less efficiently sialylated, implying that dwell time in the TGN (including possibly transit through communicating endosomal compartments) affects the maturation of furin (S.S.Molloy and G.Thomas, unpublished results).
The ordered, pH‐dependent cleavages of the furin propeptide suggest a mechanism for the compartment‐specific autoproteolytic activation of furin in vivo which is based on furin's cleavage site specificity. The autoproteolytic and RER‐specific initial cleavage occurs at the consensus furin site ‐Arg104‐Thr‐Lys‐Arg107↓‐ (Leduc et al., 1992). Typical of such sites (‐Arg‐X‐Lys/Arg‐Arg‐) that are cleaved at either neutral or acidic pH (Hatsuzawa et al., 1992a,b; Molloy et al., 1992), both the P1 (Arg107) and P4 (Arg104) arginine residues are required (Leduc et al., 1992). By contrast, furin is able to cleave some substrates containing the motif ‐Arg‐X‐X‐X‐Lys/Arg‐Arg↓‐ only in an acidic environment. For example, proalbumin is cleaved most efficiently at pH 5.5–6.0 at a site requiring both P1 and P6 arginine residues (Brennan and Nakayama, 1994a,b). Here, we show that the measured pH optimum for furin activation in vitro (pH 6.0; Figure 4), by virtue of cleavage of the propeptide at ‐Arg70‐Gly‐Val‐Thr‐Lys‐Arg75↓‐, is in close agreement with the pH of the TGN (pH ∼6.2; Seksek et al., 1995). Furthermore, mutation of either the P1 (Arg75) or P6 (Arg70) residues to Ala blocks furin activation by acidification but not by trypsinization (Figure 9), demonstrating that cleavage of the propeptide has been blocked, but folding of the enzyme has not been affected. Taken together, these data suggest that furin autoproteolytically cleaves its propeptide at Arg75 within the acidic environment of the TGN, either in trans by an associating active furin protein, or by the cognate furin molecule within the propeptide–enzyme bimolecular complex. Whether another protease(s) may be required, as is the case with carboxypeptidase Y activation (Sorensen et al., 1994), is not currently known.
The acidic environment of the TGN may also promote dissociation of the cleaved propeptide fragments from furin. Comparison of the propeptide sequences of the proprotein convertase family shows the presence of multiple conserved histidine residues (reviewed in Siezen et al., 1995, see also legend to Figure 10B). The histidine residues may coordinate a metal atom necessary for enzyme activation or participate in ionic interactions with the mature domain. The pKa of the ionizable second amino group on the histidine imidazole ring is ∼7.0, although this could vary depending on the local microenvironment within the protein (Creighton, 1993). However, within the neutral environment of the RER, these histidine residues should be neutral, while in the acidic environment of the TGN they should become positively charged. This suggests that the acidification necessary to activate furin may contribute to changes in ionic interactions between the propeptide and a metal atom or the propeptide and the enzyme. These changes, together with cleavage of the propeptide at Arg75, could lead to the dissociation of the propeptide fragments from furin, thereby unmasking the protease to act on substrates in trans.
In addition to the requirement for acidification, activation of furin may also be regulated by changes in calcium concentration. Whereas maximal activation of furin in vitro requires millimolar levels of calcium, propeptide cleavage in the structurally related enzyme PC1/3 in vivo requires calcium at only micromolar levels (Vey et al., 1994; J.VanSlyke and G.Thomas, unpublished results). In the RER, the concentration of total lumenal calcium is 3 mM (Sambrook, 1990); however, the free calcium concentration in this compartment is predicted to be ∼1 μM (Kendall et al., 1994). By contrast, the free calcium concentration in the TGN is believed to be in the millimolar range (Roos, 1988; Chanat and Huttner, 1991; Chandra et al., 1991; Song and Fricker, 1995). Interestingly, a homology model of furin based on the crystal structures of subtilisin and thermitase has suggested the presence of a high‐affinity calcium‐binding site (Ca1) and a medium‐affinity binding site (Ca2) (Siezen et al., 1994). It is possible that filling of the Ca1 site alone is required for cleavage at Arg107 in the ER, whereas filling of both Ca1 and Ca2 is required for cleavage at Arg75, and hence full activation of furin.
Although activation of the evolutionarily related furin and subtilisin endoproteases shares several properties—including a requirement for autoproteolytic cleavage of propeptides—these processes differ significantly in at least three respects. First, furin activation is pH‐ and thus compartment‐dependent, while subtilisin shows no such dependency. The need for this regulation is currently being investigated, but it seems likely that the delay of proteolytic activation prevents the enzyme from cleaving substrates prematurely or inappropriately. Alternatively, the compartment‐specific activation of furin may facilitate either the adoption of secondary conformations or oligomerization necessary to the generation of the active species. Second, the fate of the subtilisin propeptide appears to be different from that of furin. The rapid and complete degradation of the subtilisin E propeptide during activation, by an as yet uncharacterized pathway, is likely a result of the broad substrate specificity of this endoprotease. By contrast, we show that dissociation of the furin propeptide is achieved with only limited proteolysis, presumably a result of the restricted cleavage site specificity of this proprotein convertase. Third, the propeptides of proprotein convertases share no sequence similarity with the subtilisin propeptide, despite the homologies among their catalytic domains (Fuller et al., 1989; Seidah et al., 1994; Siezen et al., 1995). This suggests that the roles the furin and subtilisin propeptides play in enzyme maturation may be significantly different. Whether or not the furin propeptide directs folding of the enzyme, as it does with subtilisin, remains to be determined. This role for the propeptide is implied by the finding that truncation of the furin propeptide results in the production of an inactive protease (Rehemtulla et al., 1992).
Although the propeptides of members of the convertase family share no sequence similarity with the propeptide of subtilisin, they are similar to one another (Figure 10B; Siezen et al., 1995). Strikingly, the convertase propeptides are highly charged and contain multiple doublets or clusters of basic amino acids in two distinct regions (Siezen et al., 1995). One basic residue cluster (‐Arg‐X‐Lys‐Arg↓‐) marks the predicted initial sites of autoproteolytic cleavage of the propeptides (Figure 10B). Indeed, biochemical studies show that, as for furin, the propeptides of Kex2p and PC1/3 are cleaved rapidly at this site following deposition of the recently synthesized proenzymes into the RER (Vindrola and Lindberg, 1992; Gluschankof and Fuller, 1994; Goodman and Gorman, 1994; Lindberg, 1994). The second cluster of oligo‐basic residues present in the propeptides of several other convertases—including those for PC1/3, PC2, PC5/6 and PACE‐4—can be aligned with the furin internal cleavage site (Figure 10B). This suggests that the multi‐step, pH‐dependent activation pathway described here for furin may similarly be used by other members of the family. However, not all proprotein convertases are expected to share all features of the furin activation pathway. For instance, the PC4 and PC7 propeptides lack an internal cluster of basic residues marking the second cleavage site, suggesting that these members of the family may be activated by an alternate pathway. Additionally, although proPC2 undergoes multiple cleavages of its proregion, it also requires the participation of an associating proPC2‐binding protein, 7B2 for activation (Braks and Martens, 1994; Matthews et al., 1994). None the less, our results delineating the multi‐step activation of furin in vitro provide a basis for further work on furin as well as creating a paradigm with which to evaluate the activation pathways of other enzymes in this important converting enzyme family.
Materials and methods
The protease inhibitors PMSF, pepstatin‐A, and soybean trypsin inhibitor (SBTI) were from the Sigma Chemical Company (St Louis, MO). E‐64 was from Boehringer‐Mannheim Biochemicals (Indianapolis, IN). The peptide substrate Boc‐Arg‐Val‐Arg‐Arg‐4‐methyl‐coumaryl‐7‐amide was from Peptides International (Louisville, KY). The internally quenched substrate Q1, Abz‐Arg‐Val‐Lys‐Arg‐Gly‐Leu‐Ala‐Tyr(NO2)‐Asp‐OH, was a gift from Dr Herbert Angliker (Friedrich Miescher‐Institut, Basel). The secreted soluble furin construct Fur713t was purified as described previously (Molloy et al., 1992). Alkaline phosphatase‐ and horseradish peroxidase‐conjugated goat anti‐mouse secondary antibodies were obtained from Southern Biotechnology Associates (Birmingham, AL). The mAb 12CA5 directed against the HA epitope was obtained from Boehringer‐Mannheim Biochemicals. Purified mAbs M1 and M2 directed against the FLAG epitope were obtained from Kodak‐IBI (Rochester, NY).
BSC‐40 cells were maintained in minimal essential medium (MEM; Gibco‐BRL, Gaithersburg, MD) containing 10% fetal bovine serum (FBS; HyClone, Logan, UT) and 25 μg/ml gentamicin as described previously (Thorne et al., 1989; Bresnahan et al., 1990).
Furin constructs and vaccinia virus (VV) expression
The FLAG epitope‐tagged furin construct fur/f was generated previously (Molloy et al., 1994). HA‐tagged furin constructs were based on the EcoRI–KpnI fragment of furin in Bluescript (SK−) as a template, and generated by single‐primer (Kunkel) mutagenesis techniques with the oligonucleotide FURIN/HA (5′‐CTG CTA GCA GCT GAT GCT CAA GGA TAC CCC TAC GAC GTG CCC GAC TAC GCC CAG GGC CAG AAG GTC TTC‐3′). Underlined sequences correspond to native furin sequences between Leu19→Ala24 and Gln25→Phe30. This oligonucleotide introduced the sequence encoding QGYPYDVPDYA (the HA tag is underlined) between the end of the signal sequence and the beginning of the furin propeptide (A24↓Q25G26; see Results). The additional two residues that are not part of the HA epitope (QG) were introduced to preserve the sequence C‐terminal to the signal sequence cleavage site. The full‐length furin construct with the HA tag (fur/f/ha) was then generated using the cloning strategy described previously for the construction of fur/f (Molloy et al., 1994). The furin mutants R70A:fur/f/haΔtc‐k and R75A:fur/f/haΔtc‐k were also generated by single‐primer mutagenesis using the primers R70A (5′‐CAC TTC TGG CAT GCA GGA GTG ACG AAG CGG‐3′ and R75A (5′‐GTG ACG AAG GCC TCC CTG TCG‐3′). The nucleotides in each oligo that are complementary to the furin sequence are underlined. The R70A mutation introduced a diagnostic SphI site and the R75A mutation introduced a diagnostic StuI site (shown in bold). The ER‐retained furin construct fur/fΔtc‐k was generated previously (Molloy et al., 1994) and used to construct R70A:fur/f/haΔtc‐k and R75A:fur/f/haΔtc‐k by swapping the truncated C‐terminus with the KDEL sequence from pZVneo:fur/fΔtc‐k in place of the full‐length furin tail. All furin constructions were ultimately cloned into the pZVneo vector for generating recombinant vaccinia virus by marker transfer as previously described (VanSlyke et al., 1995).
Glutathione S‐transferase (GST) fusion proteins
Sequences encoding the native and HA‐tagged furin propeptides were produced by PCR amplification of the appropriate furin construct in pZVneo and inserted between the EcoRI and BamHI sites of pGEX 3X (Pharmacia, Piscataway, NJ). GSTpro was constructed using the 5′ primer NNAPROG: 5′‐GCG GGA TCC AGG GCC AGA AGG TCT TC‐3′. GSTpro/ha was constructed using the 5′ primer NHAPROG: 5′‐GCG GGA TCC TCT ACC CCT ACG ACG TGC CC‐3′. The 3′ primer used for the construction of both GSTpro and GSTpro/ha was CPROG: 5′‐GCG GAA TTC ACC GTT TAG TCC GTC GCT T‐3′. The sequences complementary to wild‐type or HA‐tagged furin are underlined and newly introduced restriction sites are shown in bold. GSTpro and GSTpro/ha were expressed in bacteria and purified according to the manufacturer's instructions (Pharmacia).
Crude membrane preparations were made for the analyses of furin constructs by immunoblotting, activation assays, HPLC and mass spectrometry. Confluent BSC‐40 cells on 10 cm or 15 cm plates (1 or 2×107 cells per plate, respectively) were infected with recombinant vaccinia virus at a multiplicity of infection (m.o.i.) of 5 and incubated for 14–16 h at 37°C in a defined serum‐free medium (MCDB202; McKeehan and Ham, 1976). For harvesting, the plates were placed on ice and the medium aspirated. The cells were then removed with a rubber policeman in 2.5 ml (for 10 cm plates) or 5 ml (for 15 cm plates) of 10 mM HEPES, pH 7.2, supplemented with protease inhibitors (0.2 mM PMSF, 1 mM EDTA and 0.01 mM pepstatin‐A, and 5 mM E‐64). The cells were lysed by repeated passage through a 25‐gauge needle followed by centrifugation at 5000 g for 5 min at 4°C to remove unbroken cells and nuclei. Membranes were pelleted from 400–500 μl aliquots of the low‐speed supernatants by centrifugation at 55 000 r.p.m. in either a TLA 100.1 or TLA 100.3 fixed‐angle rotor (Beckman Instruments, Inc., Fullerton, CA) for 1 h at 4°C. Pellets were resuspended in buffers as indicated.
Furin activity assays
All assays were carried out in 100 mM HEPES, pH 7.5, 1 mM 2‐mercaptoethanol, 0.5% Triton X‐100 and 1 mM CaCl2. Fluorometric assays with 4‐methyl‐coumaryl‐7‐amide‐containing peptides (AMC substrate) were performed as described previously (Molloy et al., 1992) except when GSTpro or GSTpro/ha were added as indicated. Assays with the internally quenched Q1 substrate were performed essentially as described previously (Angliker et al., 1995) except that a 3 ml reaction volume with 3.8 μM Q1 substrate was used.
Furin activation assays
Membrane pellets were resuspended by trituration into 100 μl of 10 mM Bis–Tris (pH 6.0 or 7.5), 0.5% Triton X‐100, 5 mM CaCl2, 1 mM 2‐mercaptoethanol. Protease inhibitors were added as indicated. Appropriate substitutions or omissions were made to this buffer to analyze the effect of pH and calcium on the activation of the ER‐retained furin constructs. Following resuspension the samples were transferred to 1.5 ml Eppendorf tubes and pre‐incubated in a 30°C water bath for the indicated lengths of time. Following incubation, two 30 μl aliquots of the samples were assayed for enzymatic activity against the Q1 substrate.
Immunoprecipitations and immunoblotting
A portion of the membrane pellets resuspended in Bis–Tris buffer was diluted with mRIPA/Ca2+ buffer (50 mM Tris–HCl, pH 8.0, 150 mM NaCl, 1% NP‐40, 1% sodium deoxycholate, 1 mM CaCl2). The monoclonal antibody M1 was added (50 μg/ml) and the reaction mixture rotated for 2 h at 4°C. Protein G–Sepharose (Zymed, South San Francisco, CA) was added (30 μl of a 50% slurry) to the immunoprecipitation reactions and incubated 1 h. The Sepharose beads were washed four times with mRIPA/Ca2+ buffer before resuspending them in SDS sample loading buffer. Samples were separated on a 15% peptide gel using a SDS–Tris–tricine buffer system (Schägger and von Jagow, 1987; Dayhuff et al., 1992). The proteins were then transferred to nitrocellulose membranes for 30 min at 36 V. The membranes were first blocked in TBST (50 mM Tris–HCl, pH 8.0, 150 mM NaCl, 0.05% TX‐100, 0.01% NaN3) containing 5% milk and then probed with mAb 12CA5 (0.5 μg/ml final concentration) overnight at 4°C. The secondary antibody (goat anti‐mouse–alkaline phosphatase conjugate) was incubated with the blot for 1 h at room temperature and the immunostaining pattern was developed with BCIP and NBT solutions (Zymed). When extracts were to be analyzed directly, portions of the membrane preparation suspensions were mixed with SDS sample loading buffer and subjected to peptide gel electrophoresis and Western blot analysis as described above.
Limited trypsinization of propeptide
Confluent BSC‐40 cells were infected with VV:hFur/fΔtc‐k and cell membranes were prepared as described above. Membrane pellets were resuspended by trituration in 100 μl of 100 mM HEPES, pH 7.0, 0.5% Triton X‐100, 5 mM CaCl2, 1 mM 2‐mercaptoethanol. Controls had no subsequent additives, except when 5 μl of 1 mg/ml soybean trypsin inhibitor (SBTI; final concentration = 2.5 μM) along with 2 μl of 1 μg/ml bovine trypsin (for final concentration of 0.83 nM) were added. Trypsin alone was added to the experimental samples and all samples were incubated 1 h at 30°C, after which 5 μl of 1 mg/ml SBTI was added to the experimental samples and all samples were incubated a further 30 min. Enzymatic activity was determined using the Q1 substrate as described above.
Mass spectrometry of propeptide fragment
Forty 15 cm plates of confluent BSC‐40 cells (total of 8×108 cells) were infected at an m.o.i. of 5 with VV:hFur/f/haΔtc‐k construct. At 18 h post‐infection, the cells were harvested and a crude membrane preparation made (see above). The membrane preparation was incubated for 8 h at 30°C in 5 ml of 10 mM Bis–Tris, pH 6.0, 5 mM CaCl2, 0.5% Triton X‐100, with 0.2 mM PMSF, 0.01 mM pepstatin‐A and 5 mM E‐64. The clarified supernatant from this incubation was acidified with TFA (final concentration = 0.1%) and run on reversed‐phase HPLC using a Vydac C4 column, developing the following gradient: 16 to 25% B over 2 min, 25 to 49% B over 75 min, 49 to 90% B over 10 min, where buffer A is 0.1% TFA in H2O and buffer B is 0.1% TFA in 80% acetonitrile, with a flow rate of 1 ml/min. 1 ml fractions were collected, dried down and resuspended in 100 μl water; 10 μl of each fraction was then run on 15% peptide gel using a SDS–Tris–tricine buffer system, transferred to PVDF membrane and immunoblotted using mAb 12CA5. Fractions containing the ∼6 kDa fragment were acidified with acetic acid (final concentration = 10%). 1 μl of each was used for MALDI–TOF analysis on a Voyager Elite (PerSeptive Biosystems, Cambridge, MA) in linear mode with the sample embedded in a sinapinic (3,5‐dimethoxy‐4‐hydroxy cinnamic) acid matrix. Electrospray capillary LC/MS was performed on a Perkin‐Elmer/Sciex API‐III triple quadrupole with an ionspray source on 25 μl of each sample using R1 resin (PerSeptive Biosystems). At a flow rate of 15 μl/min the chromatography was developed with 0.1% TFA and a gradient of 0 to 60% isopropanol over 60 min, followed by 60 to 100% isopropanol over 5 min. Masses from MALDI–TOF were used to scan the LC/MS data for multiply charged ions of molecular species in the range of interest. The calculation of expected molecular weights was facilitated with the Sherpa data analysis program written by J.Taylor (University of Washington), and all calculations employed average isotope abundance masses.
We are grateful to L.Thomas for expert technical assistance throughout all stages of this work. We thank H.Angliker for the Q1 substrate, S.Arch and S.Molloy for critical reading of the manuscript, and members of the Thomas laboratory for helpful discussions. This work was supported by NIH grants DK37274, DK44629 and HD30236 (G.T.). E.D.A. was the recipient of a Tartar Trust Fellowship. J.V. is the recipient of fellowship NRSA DK09069. F.J. is the recipient of an MRC (Canada) fellowship. C.D.T. was the recipient of fellowship NRSA DK09370.
↵† E.D.Anderson and J.K.VanSlyke contributed equally to these studies
- Copyright © 1997 European Molecular Biology Organization