Proalbumin to A l b u ~ i n Conversion by a Proinsulin Processing Endopeptidase of Insulin Secretory Granules"

A lysate of purified insulin secretory granules, which contains two types of proinsulin processing activity (type 1, Arg-Arg-directed and type 11, Lys-Arg-di-rected (Davidson, H. W., Rhodes, C. J., and Hutton, J. C. (1988) Natwre 333, 93-96), was found to process proalbumin by specific proteolytic cleavage of the COOH-terminal side of the Arg-'-Arg" sequence, The subcellular distribution of proalbumin processing activity in insulinoma tissue paralleled that for proinsu- lin conversion and occurred principally in a secretory granule fraction. Cleavage appeared to result from the Arg-Arg-directed type 1 proinsulin processing endopeptidase. It was Ca'+-dependent (K0.8 activation = 1.0-1.5 mM Ca"), unaffected by group-specific inhib- itors of serine, cysteinyl, or aspartyl proteinases, and had an acidic pH optimum (5.5). Active-site inhibitor studies showed this activity had a preference for dibasic over monobasic amino acid sequences and indi- cated that the sequence of the dibasic site was an im-portant determinant of the susceptibility of the sub-strate to cleavage. The activity did not process the proalbumin Christchurch mutant (Arg-'-Arg" to Arg-'-Gln-l). but not by other related normally co-secreted at-antitrypsin az-macroglobulin, or antithrombin 111. The insulin secretory granule proalbumin processing activity was indistinguishable from a proalbumin endopeptidase re-ported in rat liver membranes and similar to the yeast KEX-2 protease. These findings suggest that a highly conserved set of proprotein endopeptidases exists, which are specific for a dibasic sequence but broadly specific for proprotein substrates. Such enzymic ac- tivities appear to be active within both the constitutive and regulated pathways of secretion. Intraorganellar Ca2+ and pH appear to play a key role in regulating their activities.

M, az-macroglobulin, or antithrombin 111. The insulin secretory granule proalbumin processing activity was indistinguishable from a proalbumin endopeptidase reported in rat liver membranes and similar to the yeast KEX-2 protease. These findings suggest that a highly conserved set of proprotein endopeptidases exists, which are specific for a dibasic sequence but broadly specific for proprotein substrates. Such enzymic activities appear to be active within both the constitutive and regulated pathways of secretion. Intraorganellar Ca2+ and pH appear to play a key role in regulating their activities.
Many secreted proteins are synthesized as larger inactive precursors and processed post-translationally by limited proteolysis at sites marked by paired basic amino acids (1)(2)(3). This applies to both constitutively secreted molecules, e.g. albumin (4,5 ) , and proteins stored intracellularly in vesicles * This work was supported by the British Diabetic Association, the Wellcome Trust, the Medical Research Councils of Great Britain and New Zealand, and Nordisk Insulin Laboratories. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Juvenile Diabetes Foundation International Research Fellow. ll To whom correspondence should be addressed.
The major site of proteolytic conversion of proinsulin in the pancreatic P-cell is the insulin secretory granule (12)(13)(14)(15). Cleavage occurs on the carboxylic side of two dibasic sequences in the proinsulin molecule at Ar$l-Arg?2 and at LysGQ-Argfi5 (16). It is catalyzed by Ca2+-dependent endoproteolytic activities of acidic pH optima (17) followed by the action of carboxypeptidase H (18,19) that trims off the COOH-terminal basic amino acid residues generated by the endopeptidase clips. Studies on lysates of purified insulinoma secretory granules have revealed that two endopeptidase activities are involved (20). One (type I) cleaves exclusively at the Arg-Arg site, the proinsulin B/C-chain junction. The other (type 11) preferentially cleaves at the Lys-Arg site, the proinsulin A/ C-chain juction, though it will also recognize the Arg-Arg site to a lesser extent .
It is not clear whether the proteolytic processing of proproteins at dibasic sequences, exemplified by proalbumin and proinsulin, occurs as a consequence of the activity of a limited number of broadly specific widely distributed enzymes or a large group of tissue-and substrate-specific enzymes. In addition, it is not clear whether proteolytic conversion activity directed at constitutively secreted proteins differs from that directed at proteins which are segregated at the level of the trans-Golgi network for storage into secretory granules. We have addressed these questions through the investigation of the conversion of proalbumin (a constitutively secreted protein (7)) by intracellular elements of the pancreatic @-cell in which the regulated pathway predominates (6).

RESULTS
The conversion of proalbumin to albumin by insulin secretory granule extracts is illustrated in Fig. 1. The loss of the positively charged propeptide causes an increase in anodal mobility of the protein on agarose gel electrophoresis and results in the ability to bind 63Ni. Binding of 63Ni requires a free a-amino group on residue 1 and a histidine in position 3, uiz. Asp-Ala-His-, and therefore provides confirmation that Portions of this paper (including "Experimental Procedures" and  cleavage was directed at the carboxylic side of the proalbumin Arg-'-Arg-' sequence. Proalbumin conversion ( Fig. 1) was abolished in the presence of EDTA. It was unaffected by proteinase inhibitors ( cy2-macroglobulin, antithrombin 111, and cyl-antitrypsin M) that are normally constitutively cosecreted with albumin from the liver. However, a mutant form of cyl-antitrypsin, cyl-antitrypsin Pittsburgh (in which the reactive center Met358 is changed to Arg3"" (21), abolished the proalbumin conversion. The variant proalbumin, proalbumin Christchurch (Arg-'-Arg" to Arg-'-Gln" (22)), was not processed, reaffirming the dibasic sequence specificity for this conversion activity.
Proalbumin processing activity in different subcellular fractions of pancreatic 8-cell tissue paralleled that for proinsulin processing activity (Fig. 2). The majority of proalbumin (and proinsulin) processing activity was found in the subcellular fraction with the highest insulin content (fraction Ib). According to marker enzyme analysis and electron microscopy this fraction consisted of highly purified insulin secretory granules (23-25). Proalbumin (and proinsulin) processing activity was also observed in fraction IC which contained insulin secretory granules and contaminating mitochondria. Little proalbumin (and proinsulin (17)) processing was found in the other @-cell subcellular fractions (including the lysosome-rich fractions (Ia)). The rate of proalbumin conversion activity by the insulin secretory granule fraction was comparable with that for proinsulin processing when determined under identical assay conditions of 5 mM Ca2+ and pH 5.5 (see "Experimental Procedures"; Fig. 2, A and B ) .
The proinsulin processing type I (Arg-Arg-directed) and type I1 (Lys-Arg-directed) endopeptidase activities can be separated by anion-exchange chromatography (20) (Fig. 3). The processing of proalbumin to albumin was associated principally with the Arg-Arg-directed type I activity and with a minor component co-eluting with the type I1 endopeptidase activity. The relative ratio of proalbumin processing by the type I and type I1 endopeptidase was also 9:1, which coincided with the relative rates for the Arg-Arg-directed cleavage of Proprotein processing activity in subcellular fractions of rat insulinoma tissue. Subcellular fractionations were prepared by Percoll density gradient centrifugation as previously described (see "Experimental Procedures"). The fractions contained the following organelles (as determined by marker enzyme analysis): la, lysosomes; Zb, insulin secretory granules; IC, insulin secretory granules/mitochondria; ZZ, mitochondria/endoplasmic reticulum; ZII, endoplasmic reticulum/Golgi apparatus; ZV, plasma membrane/endoplasmic reticulum; V, cytosol. '2sI-Proalbumin to 1251-albumin processing activity by each subcellular fraction (50 pgprotein) was assayed at 30 "C over 2 h as described and is shown in panel A as a mean f S. E. ( n = 4). Similarly, 12sII-proinsulin processing activity in each of these fractions (50 pg of protein) was assayed at 30 "C for 3 h as The proalbumin processing activity of insulin secretory granules was Ca2+-dependent ( K o .~ activation = 1.0-1.5 mM Ca2+; Fig. 4A). The pH optimum of this conversion was acidic (pH 5.5, Fig, 4B); there was little proalbumin processing below pH 4.0 or above pH 7.0 (Fig. 4B). These  proteinases (26) had no effect on proalbumin processing by the insulin secretory granule fraction, whereas the metal ion chelators EDTA, CDTA,2 and EGTA, but not 1,lO-phenanthroline, all inhibited this activity (Table I, Fig. 1). The inhibition by EDTA was fully restored by the addition of Ca2+  (Table I).
The proteinase active-site-directed inhibitor tosyl-L-lysylchloromethane or tosyl-L-phenylalanylchloromethane had no effect on insulin secretory granule proalbumin processing activity. The corresponding dibasic tripeptide Ala-Lys-Arg-CH2Cl, however, produced over 90% inhibition (Table I) whereas the monobasic ch~oromethane tripeptide Ala-Nle-Arg-CH2C1 inhibited this conversion activity to a much lesser extent. This apparent preference for dibasic sequences was illustrated further in that the monobasic sequence inhibitors leupeptin and antipain were only partially inhibitory of millimolar concentrations ( Table I). The dibasic sulfonium salt tripeptide Ala-Arg-Arg-CH2S+(CH&, which contains the same dibasic sequence as proalbumin, inhibited conversion at a lower concentration than Ala-Lys-Arg-CH2S+(CH,)2. This reaffirms the Arg-Arg specificity of this processing endopeptidase activity inferred from the results obtained after DE52 chromatography (Fig. 3) (27).
The proalbumin processing by insulin secretory granules was inhibited by 20 mM F-but not by C1-, Br-, I-, SCN-, and PO:at equivalent con~entrations ( Table I). The activity was inhibited by Zn2+ and to a lesser extent by Mn2+ and Co2+ (Table I). Similar inhibitory effects by these cations have been observed for insulin secretory granuIe proinsulin conversion (17).
Both the proinsulin and proalbumin converting activity of insulin secretory granules were inhibited by al-antitrypsin Pittsburgh (Fig. 5 ) over a similar concentration range (KOA inhibition = 100 nM for proalbumin conversion and 126 nM for proinsulin conversion). Proinsulin conversion, like proal- Effect of various inhibitors on proalbumin conversion by an insulin secretory granule endopeptidase 1251-Proalbumin conversion by an insulin secretory granule endopeptidase was assessed over a 3-h period at 30 "C as previously described (see "Experimental Procedures"). In control samples (without any added inhibitor to the assay medium) 62.8 2 2.3% ( n = 8) of the proalbumin had been converted to albumin. The effects of inhibitors added to the assay are shown as a percentage of this control value. Each value is a mean of at least two independent determinations.
In this in vitro study, we demonstrate that a proinsulin processing endopeptidase activity of insulin secretory granules can also process proalbumin to albumin. This proalbumin processing endopeptidase activity was mainly attributed (>go%) to the Arg-Arg-specific type I proinsulin processing endopeptidase (20). The specificity of proalbumin conversion was confirmed by the ability of the reaction product to bind 63Ni (32,35). Cleavage before or within the ArgP2-Arg" dibasic site and subsequent NHz-terminal trimming are excluded as albumin production would then require the additional involvement of an aminopeptidase which would have been inhibited by 1,lO-phenanthroline (Table I). The pro~bumin processing activity of insulin secretory granules has almost identical biochemical characteristics to the Ca2+-dependent p r o a l b~i n conversion endopeptidase present in liver membrane preparations (11). Both these activities have the same CaZc and pH requirements, the same dibasic specificity, will not process the proalbumin Christchurch mutant (22), are insensitive to the same group-specific proteinase inhibitors, and are inhibited by the al-antitrypsin Pittsburgh mutant (21, 36). Proalbumin conversion in hepatocytes and proinsulin conversion in the pancreatic @-cell therefore appear to be cleaved by very similar Arg-Arg-directed endopeptidase activities.
The major intracellular site of proinsulin conversion is the secretory granule f 12-15), which with its acidic intragranular pH (5)(6) (12,24) and its free Ca2* concentration of 1-10 m M (25) constitutes an optimal environm~nt for the proinsulin processing endopeptidase types I and I1 (20). The constitutive transport vesicle, involved in proalbumin processing (4, 5, 7) has an acidic inner environment (37), and one would predict that on the basis of our present findings the intravesicular Ca2+ concentration is in the millimolar range. The formation of a secretory granule in pancreatic P-cells or of a constitutive transport vesicle in hepatocytes occurs in the region of the trans-Golgi network (4,5,7,12). It is generally believed that this particular compartment has a relatively low Ca2+ concentration (38) and a near neutral pH (12). It follows, therefore, that the formation of either a secretory granule or constitutive transport vesicle is accompa~ed by the insertion or activation of proteins which result in intraorganellar acidification and Ca" accumulation. This common feature, together with the findings that proprotein processing occurs by a similar enzymic activity in both the constitutive and regulated secretory pathways, suggests that these compartments contain many molecular components in common.