Behavior in the Eukaryotic Secretory Pathway of Insulin-containing Fusion Proteins and Single-chain Insulins Bearing Various B-chain Mutations*

In the secretory pathway, endoproteolytic cleavage of the insulin precursor protein promotes a change in the biophysical properties of the processed insulin product, and this may be relevant for its intracellular trafficking. We have now studied several independent point mutants contained within the insulin B-chain, S9D, H10D, V12E (called B9D, B10D, and B12E), as well as the double point mutant P28K,K29P (B28K,B29P), that have been reported to inhibit insulin oligomerization. In yeast cells, the unprocessed precursor of each of these mutants is secreted, whereas >90% of the endoproteolytically released single-chain insulin moiety is retained intracellularly; a large portion of the B9D, B10D, and B12E single-chain insulins exhibit abnormally slow mobility upon nonreducing SDS-PAGE, despite normal mobility upon reducing SDS-PAGE. Although no free thiols can be detected, each of these mutants exhibits increased disulfide accessibility to dithiothreitol. After dithiothreitol treatment, a portion of the molecules can reoxidize to a form more compact than the original single-chain insulin mutants formed in vivo (indicating initial disulfide mispairing). Disulfide mispairing of a fraction of B9D, B10D, and B12E mutants also occurs in the context of single-chain insulin and even in authentic proinsulin expressed within the secretory pathway of mammalian cells. We conclude that analyses of the intracellular trafficking of certain oligomerization-defective insulin mutants is complicated by the formation of disulfide isomers in the secretory pathway.

In the secretory pathway, endoproteolytic cleavage of the insulin precursor protein promotes a change in the biophysical properties of the processed insulin product, and this may be relevant for its intracellular trafficking. We have now studied several independent point mutants contained within the insulin B-chain, S9D, H10D, V12E (called B9D, B10D, and B12E), as well as the double point mutant P28K,K29P (B28K,B29P), that have been reported to inhibit insulin oligomerization. In yeast cells, the unprocessed precursor of each of these mutants is secreted, whereas >90% of the endoproteolytically released single-chain insulin moiety is retained intracellularly; a large portion of the B9D, B10D, and B12E single-chain insulins exhibit abnormally slow mobility upon nonreducing SDS-PAGE, despite normal mobility upon reducing SDS-PAGE. Although no free thiols can be detected, each of these mutants exhibits increased disulfide accessibility to dithiothreitol. After dithiothreitol treatment, a portion of the molecules can reoxidize to a form more compact than the original single-chain insulin mutants formed in vivo (indicating initial disulfide mispairing). Disulfide mispairing of a fraction of B9D, B10D, and B12E mutants also occurs in the context of single-chain insulin and even in authentic proinsulin expressed within the secretory pathway of mammalian cells. We conclude that analyses of the intracellular trafficking of certain oligomerization-defective insulin mutants is complicated by the formation of disulfide isomers in the secretory pathway.
Clinically relevant production of insulin in the eukaryotic secretory pathway occurs as a consequence of endogenous expression of proinsulin in pancreatic beta cells (1), as a consequence of artificial expression in heterologous mammalian cells (2), or from expression of an insulin-containing fusion protein (ICFP 1 ; encoding the yeast preproalpha factor leader peptide fused via a KR cleavage site to a single-chain insulin) in Saccharomyces cerevisiae (3). In the latter case, upon ICFP cleavage in the distal secretory pathway by Kex2p endoprotease to liberate the alpha leader peptide, a crystallizable single-chain insulin is produced with a structure essentially isomorphous with that of authentic two-chain insulin (3,4).
We have shown previously that upon expression of ICFP at modest levels, the fraction of molecules remaining unprocessed is secreted rapidly, whereas the other fraction undergoing Kex2p-mediated endoproteolysis results in a single-chain insulin that primarily is not secreted (5). Moreover, this sorting, determined by Kex2p-mediated processing, is saturable. Failure to secrete the single-chain insulin moiety cannot be explained by ICFP endoproteolysis within a "dead-end" compartment from which molecules can no longer be secreted, because simple deletion of the yeast KEX2 gene not only prevents ICFP processing but also increases the quantity of ICFP secreted. It is known that by liberating insulin from its precursor (in either mammalian cells or yeast), cleavage promotes a change in the biophysical properties of the processed product (5)(6)(7)(8)(9)(10)(11), and this might play a role in protein sorting in the lumen of the distal secretory pathway (12,13).
The structure of insulin has been examined extensively (14) and has been the subject of detailed mutagenesis studies. From the crystal structure (9), residues B9S, B12V, and B28P (B referring to the insulin B-chain), among others (15), participate in the dimer-forming surface of the insulin monomer; mutations such as B9D, B12E, and B28K,B29P have been described as having diminished tendency to dimerization (16 -18), whereas a B10D mutant is thought to be impaired primarily in hexamer assembly from dimers (19). Given that some assembly-deficient B-chain point mutants are thought to exhibit folding stability comparable with that of native insulin (20), we have now examined the expression of these B-chain point mutants in the secretory pathway of yeast and in transiently transfected mammalian cells. Although the B28K,B29P double mutant was expressed generally at the lowest levels, each of the B9D, B10D, and B12E mutants produced more than one isomer differing in disulfide bonding, suggesting that these mutants have varying degrees of predisposition for disulfide mispairing during initial folding in the secretory pathway.

EXPERIMENTAL PROCEDURES
Antibodies and other Materials-Polyclonal anti-insulin was made in guinea pigs (Linco Research, St. Charles, MO). Secondary antibodies and peroxidase conjugates were from Jackson ImmunoResearch Laboratories (West Grove, PA); Zysorbin was from Zymed Laboratories Inc., and protein A-agarose was from Sigma. [  employed in this study are isogenic to PA13 (MATa,kss1,112,his3,trp1^63,ade2,GALϩ), which has a complete deletion of the PEP4 gene, marked with URA3 (5). The VPS10 gene was then disrupted by transformation with XhoI and BamHI fragment of pEM10 -103 (21) to produce strain PA11D. pMIGLC, a TRP1-marked centromeric plasmid (22) bearing the wild-type ICFP sequence is as described previously (with the clarification that the strong promoter/ terminator previously reported as belonging to the glycerol-3-phosphate dehydrogenase gene, GPD1, in fact belongs to the promoter of glyceraldehyde-3-phosphate dehydrogenase, known as TDH3, and the terminator of 3-phosphoglycerate kinase, known as PGK1) (5). In one set of experiments examining the effects of vps10, each of the insulin B-chain point mutations was prepared by PCR mutagenesis (of an ϳ0.4-kb EcoRI-HindIII PCR-amplified fragment containing the relevant mutation that was used to replace the matching EcoRI-HindIII fragment of ICFP, confirmed by DNA sequencing) in the context of the pMIGLC vector, and these were transformed into PA11D and the PA13 parental strain. In a second set of experiments, strain AHY63 was employed (the kind gift of Dr. S. Nothwehr, University of Missouri, Columbia, MO) which is isogenic with W303-1A (MATa; ura3-1; leu2-3,112; his3-11,15; trp1-1, can1-100; ade2-1) but contains pho8⌬::ADE2 and carries the unmarked pep4-⌬H3 mutation (23). The VPS10 gene was then disrupted in this strain (as described above) to produce strain YB136. In addition, the TRP1 marker of the pMIGLC series of ICFP expression vectors was replaced by the URA3 marker from a URA3-marked centromeric plasmid (22) to create a series of vectors called pGMU, bearing each of the insulin B-chain point mutations. These vectors, individually, were transformed into YB136 and the AHY63 parental strain.
For experiments examining the nonreducing SDS-PAGE mobility of insulin bearing different B-chain point mutations, the URA3 marker of PA13 was disrupted by insertion of HIS3 (pUC119-ura3::HIS3 cut with XmaI) to produce strain YB118. Then, each of the pGMU series of vectors bearing insulin B-chain point mutations was transformed into YB118.
Metabolic Labeling and Immunoprecipitation in Yeast Cells-Cells were grown to exponential phase in minimal medium supplemented with required amino acids but without methionine and cysteine, using 2% glucose as the carbon source, and then sedimented at 3000 ϫ g for 5 min and resuspended at 4 A 600 /ml in the same medium plus 1 mg/ml bovine serum albumin. For pulse labeling, [ 35 S]cysteine was added at 50 Ci/A 600 for 5 min (unless otherwise indicated). The cells were chased in synthetic complete (SC) medium plus 10 mM methionine and cysteine. At 45 min of chase (unless otherwise indicated), 1-ml aliquots were transferred to ice, and 10 mM sodium azide and a protease inhibitor mixture (aprotinin, 0.25 g/ml; leupeptin, 0.1 mM; pepstatin, 10 M; EDTA, 5 mM; E64, 10 M; diisopropyl fluorophosphate, 1 mM, bovine serum albumin 0.02%) were added immediately. Cells were then sedimented at 3000 ϫ g for 5 min, and the supernatant containing secreted proteins was collected for further analysis. The cells were resuspended in 100 l of breaking buffer containing 6 M urea, 1% SDS, 1 mM EDTA, and 50 mM Tris, pH 7.5. The cells were lysed by vortexing with glass beads and then boiling for 4 min. Samples were diluted 10-fold to a final concentration of 1% Triton X-100, 150 mM NaCl, 5 mM EDTA, 50 mM Tris, pH 7.5, precleared with 20 l of zysorbin for 30 min (Zymed Laboratories Inc., San Francisco, CA), and then immunoprecipitated with 5 l of guinea pig anti-insulin plus 20 l of zysorbin. Immunoprecipitates were digested with PNGase F (New England Biolabs, Beverly, MA) and analyzed by SDS-PAGE in the absence or presence of reducing conditions as described in the text.
For the sample preparation in the experiment shown in Fig. 2, yeast cells were lysed in the absence of detergent, Triton X-100 was added to 0.9%, and the samples were immunoprecipitated with anti-insulin as described above. The immunoprecipitates were then exposed sequentially to 18 mM dithiothreitol (DTT) for 30 min at 37°C, to iodoacetamide at a final concentration of 36 mM (37°C, 5 min), and finally to SDS gel sample buffer before analysis by SDS-PAGE.
Tricine-Urea-SDS-PAGE-We modified the Tricine-urea-SDS-PAGE system (24) as follows. In addition to TEMED and ammonium persulfate as described, the resolving gel contained 11.2% acrylamide, 0.35% bisacrylamide, 10% Prosieve-50 (BioWhittaker), and a final concentration of 6.0 M urea in 0.1% SDS and 1.0 M Tris, pH 8.45. Before this was allowed to polymerize, it was overlaid with a small quantity (sufficient for adding 0.2 cm of overlay) of "spacer gel" containing, in addition to TEMED and persulfate, 9.3% acrylamide and 0.6% bisacrylamide without Prosieve or urea in 0.1% SDS and 1.0 M Tris, pH 8.45. 30 min after polymerization of the resolving and spacer gel, a stacking gel was poured containing, in addition to TEMED and persulfate, 3.7% acrylamide and 0.24% bisacrylamide in 0.075% SDS and 0.75 M Tris, pH 8.45. None of the gel solutions were degassed before use. The sample buffer we used resulted in a final concentration of 2% SDS, 12% glycerol, and 0.0025% Serva blue in 50 mM Tris, pH 6.8. Typically, insulin samples were boiled for 5 min before gel loading, although we found that omitting the boiling step had essentially no effect on the outcome of these experiments. The tank buffers were the same as those described previously: a running buffer of 0.1% SDS in 0.1 M Tris, 0.1 M Tricine, pH 8.25, and an anode buffer of 0.2 M Tris, pH 8.9 (24). This gel system was essential to the detection of mobility differences between the insulin species described.
Mammalian Cell Culture, Transfection, and Labeling-HEK293 cells were cultured in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum and 0.1% penicillin-streptomycin (Invitrogen) at 37°C with 5% CO 2 . A cDNA encoding wild-type human proinsulin was cloned into the pcDNA3 (CMV promoter-driven) expression vector (Invitrogen). Each of the insulin B-chain point mutations was prepared by PCR mutagenesis within an ϳ0.4-kb PCR-amplified fragment within the pGEM-T vector and then subcloned to replace the appropriate wild-type proinsulin sequence in pcDNA3 with successful completion of the mutagenesis confirmed by DNA sequencing. The pRSV-InsDC plasmid (encoding a preproinsulin in which the C-peptide and endoproteolytic cleavage sites are completely lacking), obtained from Dr. H. P. Moore (University of California, Berkeley), was subcloned to pcDNA3 and used as the template for further PCR mutagenesis to encode an artificial 7-amino acid C-peptide sequence comprising MGGGGGM (Met-Gly-Gly-Gly-Gly-Gly-Met). The resulting construct is referred to in this manuscript as SCI (single-chain insulin). The SCI construct was then submitted to additional rounds of PCR mutagenesis to produce the same specific B-chain point mutations as those produced in the yeast ICFP and in proinsulin.
Plasmid DNA was transfected into HEK293 cells using Lipo-fectAMINE (Invitrogen). Cells were metabolically radiolabeled with a [ 35 S]methionine/cysteine mixture for 1 h in methionine/cysteine-deficient medium and chased for 1 h in complete medium. Where indicated, brefeldin A was included during labeling and chase at a concentration of 10 g/ml. At the end of the chase, the medium was collected, and the cells were lysed in 100 mM NaCl, 1% Triton X-100, 0.1% SDS, 10 mM EDTA, and 25 mM Tris, pH 7.4. Cell lysates and chase media were treated with a proteinase inhibitor mixture (Roche Molecular Biochemicals), precleared with zysorbin (Zymed Laboratories Inc.), and then subjected to immunoprecipitation with guinea pig anti-insulin as described above. Zysorbin-bound immunocomplexes were sedimented at 12,000 ϫ g for 1 min, and pellets were washed twice with cell lysis buffer and once in high salt buffer (0.5 M NaCl, 1% Triton X-100, 10 mM EDTA, and 25 mM Tris, pH 7.4) before analysis by Tricine-urea-SDS-PAGE (24) in the absence or presence of reducing conditions as described above.
Reactivity with 4-Acetamido-4Ј-maleimidylstilbene-2,2Ј-disulfonic Acid (AMS)-AMS was purchased from Molecular Probes (catalog No. A-485). Immunoprecipitates with anti-insulin, derived either from yeast or mammalian cells, were divided in 50 l of either 80 mM Tris-HCl, pH 6.8, 1% SDS, 1 mM phenylmethylsulfonyl fluoride, and 20 mM AMS or an identical buffer without AMS. The samples were then resolved by nonreducing SDS-PAGE. In some cases, identical aliquots of radiolabeled yeast or mammalian cell lysate were incubated with AMS prior to lysis and insulin immunoprecipitation, but the results were identical to those obtained when the AMS reaction was performed after immunoprecipitation.

Secretion from Yeast of Unprocessed Insulin-containing Fusion Proteins with Lack of Secretion of Processed Single-chain Insulins Containing B-chain Point
Mutations-We previously found that in yeast, after Kex2p-mediated endoproteolytic removal of the alpha leader peptide from ICFP (bearing the wild-type B-chain residues 1-29 and A-chain residues 1-21 made contiguous by an AAK tripeptide linker sequence), a single-chain insulin was produced that was not efficiently secreted and instead accumulated intracellularly in a manner dependent upon PEP4, encoding the major yeast vacuolar proteinase (5). Intriguingly, this intracellular targeting was entirely dependent upon the catalytic activity encoded by KEX2, and the sorting function provided by this gene could be replaced in trans by expression of a truncated, membraneless, secretory Kex2p. In contrast, intracellular insulin targeting was not perturbed by deletion of the VPS10 gene encoding a vacuolar sorting receptor.
The present study was initiated to examine the effect of B-chain point mutations on the expression and trafficking of insulin. Because substitution of proximal B-chain residues can provoke increased conformational changes of the insulin moiety including increased B-chain flexibility (25,26), and the B-subdomain may provide a template to guide folding of the A-chain (27,28), it seemed possible that such B-chain point mutants might be recognized as misfolded by Vps10p (29 -31). We therefore employed a pulse-chase protocol to examine the behavior of ICFP constructs bearing B9D, B10D, B12E, or B28K,B29P point mutations, driven by the strong TDH3 promoter, in strains that were (or were not) deleted for VPS10. At this level of substrate expression the processing activity of Kex2p is limiting (5). As shown in Fig. 1 (bottom left), for the wild-type insulin sequence the unprocessed ICFP was well secreted, and the processed form was stored (in the vacuole) of pep4 null cells. In agreement with our previous study, introduction of the B5T mutation greatly augmented ICFP expression, but the small amount of cleaved B5T-insulin was not retained efficiently intracellularly (Fig. 1). Each of the other point mutants was expressed (although the B28K,B29P samples had to be overloaded in order to achieve a comparable radioactive protein signal), and in each case, the fusion protein was secreted, whereas the processed insulin mutant was retained efficiently (Fig. 1). For the vps10 mutant strains, there was at most a modest effect observed for B10D, B12E, and B28K,B29P mutants, as the final ratio of unprocessed ICFP in the medium to processed insulin in the cells was not greatly affected. As shown in the enlarged image of the gel analyzing the B10D mutant (Fig. 1, right), only a small decrease in intracellular insulin, coupled with a small increase in secreted ICFP, was detectable.
Single-chain Insulins Containing B-chain Point Mutations Exhibit Increased Sensitivity to Reduction with Dithiothreitol-Upon endoproteolytic processing in yeast, the resulting singlechain insulin containing the wild-type sequence has a structure that tends to resist disulfide reduction with DTT under nondenaturing conditions (5). To further explore the stability of single-chain insulins bearing B-chain point mutations, immunoprecipitates from cells expressing various mutant constructs were exposed to 20 mM DTT under nondenaturing conditions (see "Experimental Procedures"). After such treatment, singlechain insulin without additional B-chain point mutations was largely recovered as a band with an identical mobility to that of a parallel sample not treated with DTT ( Fig. 2, open arrow). However, each of the B-chain point mutants exhibited enhanced sensitivity to DTT as demonstrated by diminished recovery of the single-chain insulin band upon SDS-PAGE. These data suggest differences in the conformational stability of the single-chain insulins bearing wild-type and mutant sequences.
In Yeast, Some Single-chain Insulins Containing B-chain Point Mutations Exhibit Abnormal Mobility upon Nonreducing SDS-PAGE-Although many structural analyses of insulin Bchain point mutants have begun with chemically or bacterially synthesized peptides or otherwise purified material that has been folded or refolded in vitro or purified for structural homogeneity by selection of a specific HPLC peak (32), relatively few studies have examined the folding of such mutants as they are produced in vivo within the eukaryotic secretory pathway. It is known that substitution of proximal B-chain residues, such as the B9S substitution with a Glu residue, can provoke conformational changes of the B-chain (25,26) that may influence formation of one or more interchain disulfide bonds (33). (Disulfide mispairing is also known to occur during production of recombinant IGF-1; see "Discussion".) To study this question, we examined the mobility of intracellular single-chain insulins bearing various B-chain point mutations upon reducing versus nonreducing SDS-PAGE. In short, upon reducing SDS-PAGE, all single-chain insulin constructs exhibited an identical mobility (Fig. 3A, left). However, upon nonreducing SDS-PAGE, single-chain insulins bearing B9D, B10D, or B12E point mutations consistently exhibited a slower SDS-PAGE mobility (arrows in Fig. 3A, right). The B28K,B29P double mutant was always more poorly expressed than the other insulin constructs; however, its mobility by nonreducing SDS-PAGE appeared normal (Fig. 3A). The isoelectric points of the constructs were unlikely to explain the anomalous mobility of the B-chain point mutants, as these parameters did not strictly correlate (Fig. 3A, and data not shown). To determine whether any of the mutant forms contained unpaired cysteine residues bearing free thiols, each construct was incubated (or not) in the presence of AMS (34). This compound adds Ͼ500 daltons to the molecular mass of the polypeptide upon thiol reaction with an AMS residue, and such a molecular mass increase should be readily detected as loss of the 6-kDa single-chain insulin band (with shift to a higher molecular weight). However, as shown in   Fig. 4 (upper panels), in which equal proportions of cell lysate and medium were analyzed, a nearly undetectable quantity of processed insulin was released to the medium (amounting to no more than a few percent) for any of the constructs examined. However, when markedly unequal exposures of these gels were obtained, significantly underexposing the cell lysate insulin while overexposing that recovered in the medium (Fig 4, lower panels), two interesting features became apparent. First, although the bulk of intracellular B9D, B10D, and B12E mutants exhibited abnormal mobility on nonreducing gels, a much smaller quantity of these mutant insulins exhibited a mobility that was indistinguishable from that of the single-chain insulin bearing the wild-type control sequence (Fig. 4, left). Second, for these mutants, the very little processed insulin that was secreted was enriched in a species for which mobility was indistinguishable from that of the single-chain insulin bearing the wild-type control sequence (Fig. 4,  right). Together, the data in Figs. 3 and 4 demonstrate that for single-chain insulin bearing certain B-chain point mutations, distinct isoforms are detectable upon nonreducing SDS-PAGE, which can no longer be distinguished upon disulfide reduction. Assuming that the faster species co-migrating with singlechain insulin bearing the wild-type control sequence represents a form containing three correct (A6-A11, B7-A7, B19-A20) disulfide pairs, then a substantial portion of these B-chain point mutants is likely to represent an isomeric by-product of equal molecular mass but with one or more disulfide bonds interchanged (33,(35)(36)(37)(38). Moreover, although the processed insulin moiety is very poorly secreted, in the case of these selected B-chain point mutants, the secretion tends to be enriched in the better folded isomer, whereas what is retained intracellularly tends to be enriched in the misfolded isomer.
In Vitro Disulfide Isomerizaton of Single-chain Insulin Bearing B-chain Point Mutations-In the course of these studies it was observed that unlike in authentic two-chain insulin, which, after SDS denaturation, is very sensitive to in vitro disulfide reduction because of chain dissociation, it was more difficult to obtain completely reduced and alkylated singlechain insulin in vitro. An example of this for the single-chain insulin bearing the wild-type control sequence is shown in Fig.  5 (right) in which the protein becomes divided into three bands: a fast migrating species identical to the initial mobility under nonreduced conditions and two slower migrating species, the slowest of which migrates in the approximate position of fully reduced single-chain insulin (see Fig. 3A, left panel). Although this complex pattern of bands could be attributed to incomplete initial insulin reduction with 20 mM DTT (Fig. 5), a remarkable finding was obtained when examining the behavior of processed single-chain insulin recovered from yeast cells bearing ICFP with various B-chain point mutations. Specifically, for the single-chain insulin mutants B9D, B10D, and B12E (which exhibit abnormal mobility upon nonreducing SDS-PAGE), after The processed single-chain insulin monomers recovered from lysates of cells expressing B9D, B10D, and B12E are comprised of a form of slow gel mobility, although in this experiment a very small portion of these mutants can be detected at a mobility similar to that of the control single-chain insulin without B-chain point mutations (C). Although an extremely small percentage of any of these constructs is secreted (upper panel), upon overexposure of the gel derived from media, the minor secretion of the B9D, B10D, and B12E single-chain insulin mutants appears to be enriched in the faster migrating species, which has a mobility comparable with that of the control protein.
initial reduction with 20 mM DTT followed by treatment with 40 mM iodoacetamide, intended to alkylate all free sulfhydryls, the same three-species banding pattern was obtained in proportions similar to that observed for the wild-type control protein (Fig. 5). Indeed, the faster migrating isoform (similar to the mobility of the nonreduced wild-type control protein) was actually faster after DTT/iodoacetamide treatments than before. Although these findings do not establish whether initial reduction was complete, the data do establish that (for at least a portion of the molecules) the disulfide reduction is followed by in vitro reoxidation of single-chain insulins. For certain Bchain point mutants, this reoxidation clearly matches a better folded species than that which accumulated in living cells in the first place. Similar reoxidation results were obtained by combining reduced single-chain insulin with glutathione containing mixtures of reduced (GSH) and oxidized (GS-SG) species (not shown). These results are consistent with previously published studies of single-chain insulins in vitro (33).

Folding and Transport of Single-chain Insulin Bearing Bchain Point Mutations in the Secretory Pathway of Mammalian
Cells-The foregoing data are consistent with the possibility that B9D, B10D, and B12E point mutations may have placed conformational stress on the single-chain insulin structure (25), which might lead to disulfide mispairing in the oxidizing environment of the endoplasmic reticulum (ER). As the overall oxidoreductase capability of the yeast and mammalian ER may differ (39), it seemed of interest to examine the folding of single-chain insulin constructs in the context of the mammalian secretory pathway. A cDNA encoding a single-chain insulin (SCI) bearing the wild-type preproinsulin signal peptide, B-chain and A-chain, in which the chains are linked by the 7-amino acid noncleavable sequence MGGGGGM, was constructed. When driven by a CMV promoter and introduced by transient transfection in HEK293 cells, an insulin-immunoprecipitable band was produced that migrated upon nonreducing SDS-PAGE slightly slower than an authentic iodoinsulin standard (Fig. 6A). This slightly slower mobility was attributable solely to the presence of the linker peptide, as cyanogen bromide excision of the linker caused the SCI to co-migrate precisely with the two-chain insulin standard. 2 The SCI was then mutagenized to include B9D, B10D, or B12E point mutations. As shown in the nonreducing gel of Fig. 6A, immunoprecipitation of B9D or B12E point mutants produced two monomeric isoforms: a faster species that co-migrated with the wildtype control SCI sequence and a second, slower form (in addition small quantities of dimeric B9D mutant insulin were observed). As in yeast (compare with Fig. 4), less than half of the B9D or B12E mutant monomers migrated with the normal mobility. The B10D point mutant exhibited even greater perturbation of mobility upon nonreducing SDS-PAGE. Additionally, as has been reported for a known disulfide-mispaired isomer of insulin (38), none of the bands was changed by incubation with alkylating agent (ϩAMS, Fig. 6B). Once again, these data suggest disulfide mispairing of the B9D, B10D, and B12E mutant SCIs.
Remarkably, despite large-scale generation of a single-chain insulin by-product with disulfide interchange, which must constitute significant misfolding, the mutant SCI constructs were secreted with an efficiency comparable with that of SCI bearing the wild-type insulin sequence. Moreover, unlike what was observed in yeast, secretion efficiency was very high and was comparable for both disulfide isomers of mutant SCI. An example of this is shown for the B9D mutant in Fig. 7 (but this was also observed for B10D and B12E mutants; not shown). Following a 1-h pulse labeling plus an additional 1.5-h chase, the majority of the wild-type SCI and both monomeric disulfide isomers of the B9D mutant were released to the medium. This release represents genuine secretion rather than nonspecific release (such as due to impaired cell viability) because it was abolished by treatment of the cells with brefeldin A, which blocks intracellular protein transport through the secretory pathway (Fig. 7). Thus, at least in HEK293 cells, misfolded disulfide isomers of single-chain insulin are apparently invisible both to ER retention/quality control machinery (40) and to Golgi quality control machinery (13).
Introduction of B-chain Point Mutations within the Coding Sequence of Authentic Proinsulin-The ICFP expressed in yeast employed a short (3-amino acid) linker peptide between B-and A-chain sequences, and the SCI expressed in mammalian cells employed a short (7-amino acid) linker. By contrast, authentic human proinsulin employs a 35-amino acid linker that includes the 31-amino acid C-peptide plus two sets of flanking dibasic residues. B10D is a naturally occurring point mutation within proinsulin, associated with autosomal dominant hyperproinsulinemia and glucose intolerance (41). Steiner and colleagues (42) have raised the suggestion that some B10D proinsulin molecules might undergo misfolding, including incorrect sulfhydryl oxidation. Moreover, the proinsulin cDNA bearing the B9D point mutation has been expressed in AtT-20 In the far left lane, the mobility of iodinated two-chain insulin standard is shown. Note that a major fraction of the B9D, B10D, and B12E point mutants migrate with anomalously slow monomer mobility. Additionally, a very small fraction of the B9D mutant is reproducibly recovered as an even slower mobility species interpreted as a dimer. All of the single-chain insulin constructs shown here collapse to a single band of identical mobility under reducing conditions (not shown). B, in an experiment similar to that shown in Fig. 3B, immunoprecipitates were incubated in a buffer containing (ϩ) or not containing (Ϫ) 20 mM AMS as described under "Experimental Procedures." However, no band shift indicative of free thiols was detected for any of the constructs.

FIG. 7. Single-chain insulin bearing the B9D point mutations is efficiently secreted from HEK293 cells.
Cells transiently transfected to express SCI bearing the wild-type B-chain sequence (ϩ) or mutant sequence were pulse-labeled for 60 min and chased for 1.5 h. Where indicated, brefeldin A (BFA, 10 g/ml) was included during the labeling and chase periods. The media were collected, and cells were lysed. Finally, all samples were immunoprecipitated with anti-insulin and analyzed by nonreducing SDS-PAGE. Although the B9D mutant existed as two monomeric isomers with different SDS-PAGE mobility (marked by carets (Ͼ)), both forms were efficiently released to the medium, and the secretion of all forms was blocked by brefeldin A treatment. Identical results were obtained for SCI bearing B10D or B12E mutations (not shown). C, cell lysate; M, chase medium. cells, where it was found to be secreted, and this was taken as evidence suggesting that B9D proinsulin is not misfolded in the secretory pathway (43). It was therefore of interest to re-examine the folding of proinsulin bearing B-chain point mutations. As shown in the nonreducing gel of Fig. 8, each of the B9D, B10D, and B12E proinsulin point mutants showed some slower mobility isoforms consistent with disulfide mispairing. Moreover, all of the mutant proinsulin constructs ran as a single fully reduced species under reducing conditions. However, unlike for the ICFP and SCI constructs, the majority of the B9D, B10D, or B12E mutant proinsulins migrated upon nonreducing SDS-PAGE with the normal gel mobility (as exhibited by wildtype proinsulin). Thus, these B-chain point mutations are evidently better tolerated within the context of the full 35-amino acid linker of proinsulin than within single-chain insulins expressed in yeast or mammalian cells (32). DISCUSSION Many previous structural analyses of proinsulin, insulin, and insulin mutants have combined to provide an extremely valuable contribution to the biotechnology of insulin production (3). However, relatively few studies have examined the folding of such proteins within the context of the eukaryotic secretory pathway. Our laboratory has long been interested in the relationship between insulin self-assembly and insulin storage in the secretory granule compartment of mammalian cells (10, 44 -46). Recently, we have expanded our interest to include the possible relationship between the assembly of insulin and its trafficking in the secretory pathway of yeast (5), using a singlechain insulin like that which forms the basis for pharmaceutical insulin production (47). For both of these systems, it seemed of interest to study point mutants with reported defects in insulin assembly, although it would need to be excluded that such point mutations might interfere with proper disulfide bond maturation. Such a concern might seem unlikely in the case of a mutant such as B10D, which in vitro is known to stabilize insulin structure and in vivo generally improves insulin expression. However, it is already known that thermodynamic stability does not necessarily correlate with biological activity (48,49), and it is equally important to consider that thermodynamic stability may have incomplete predictive value with regards to the co-translational folding pathway that occurs during translocation into the specialized environment of the ER. In the cell, the polypeptide may fold as it enters the ER from N terminus to C terminus (50), in the immediate proximity of the protein-conducting channel (51), timed in relation to signal peptide cleavage (52), and in the presence of ER-specific thiol-interactor proteins, which can make mixed disulfides with newly made thiol-containing secretory proteins (53). These compartment-specific factors may be reconstituted using isolated ER microsomes (54), but so far such systems are not easily transferred to current biophysical methodologies necessary for detailed analyses of protein structure and conformational stability.
Before discussing the particular results obtained in the current study, it is worth considering the relationship of the present analysis of single-chain insulin and proinsulin to that of insulin-like growth factor-1 (IGF-1), two homologous proteins that encode similar but nonhomologous structures (4). One suggestion for the structural difference is that whereas the intra-A-chain disulfide bond of insulin is very stable, the Cys 47 -Cys 52 pairing of IGF-1 is an unfavorable high energy bond that is under strain (55). Like the situation for insulin, IGF-1 bioactivity and normal half-life requires the presence of all three native disulfide bonds (56 -58). Although disulfide mispairing in vivo has not previously been reported for singlechain insulin or proinsulin, it is well known to occur during production of IGF-1 in the secretory pathway (36,37) and upon IGF-1 refolding in vitro (55,59,60). Mispaired disulfide isomers of IGF-1 maintain some biological activity and elements of secondary structure (56,59). In one report, two such disulfide isomers were clearly recovered within the same reversedphase high-performance liquid chromatography peak (61). Thus, such molecules may exhibit a sufficiently folded structure to escape ER-based quality control, resulting in secretion (37). Consistent with the idea that the B-chain provides a template to guide folding of the A-chain (27), during in vitro refolding an insulin B-chain/IGF-1 A-chain chimera forms stable native disulfides and a single thermodynamically stable tertiary structure, whereas an IGF-1 B-chain/insulin A-chain chimera cannot form/maintain a single set of native disulfide bonds but instead folds into two distinct disulfide isomers (28,62). From these data, it seems reasonable to be concerned about the possibility that certain mutations in the insulin B-chain sequence might impair the fidelity of native disulfide bond formation within single-chain insulins and in proinsulin as well.
In this report, we find that each of the single-chain insulin mutants produced in yeast is more susceptible to reduction under nondenaturing conditions (Fig. 2). Although the vps10 mutation has only a slight effect on the intracellular retention of these mutants (Fig. 1), the B9D, B10D, and B12E mutants form two different gel bands representing disulfide isomers (neither of which is appreciably secreted, although the isomer matching the predominant conformation assumed by the normal B-chain sequence shows a slight secretion; Fig. 4). For the misfolded isoform of these mutants, which predominates intracellularly, a second chance at disulfide oxidation in vitro actually appears to improve the disulfide pairing from that of the single-chain insulin product originally synthesized in vivo (Fig.  5). Thus, the trafficking of these three point mutants probably cannot be used to analyze the effects of oligomeric and multimeric assembly on wild-type insulin trafficking in the yeast secretory pathway because, although they do undergo anterograde transport, the bulk of the mutant molecules is not properly folded. By contrast, the B28K,B29P double mutant does have apparently normal disulfide maturation (Figs. 3 and  4). However, it is much more poorly expressed under identical conditions (and needs to be overloaded in our gels to see the band). FIG. 8. Anomalous mobility by nonreducing SDS-PAGE of human proinsulin bearing selected B-chain point mutations expressed in mammalian cells. HEK293 cells were transiently transfected with pcDNA3 encoding wild-type human preproinsulin (ϩ) or bearing B9D, B12E, or B10D point mutations as indicated. Cells were pulse-labeled for 30 min, lysed, and immunoprecipitated with antiinsulin as described under "Experimental Procedures" before analysis by SDS-PAGE under nonreducing conditions or after reduction with 20 mM dithiothreitol. Unlike the SCI construct, the presence of the B9D, B10D, or B12E point mutants causes only a small fraction of the protein to migrate with anomalously slow mobility (upper two carets (Ͼ)) with a majority migrating in the fast mobility position equivalent to that of wild-type proinsulin (lower caret).
A similar misfolding problem occurs when the B9D, B10D, or B12E point mutants are introduced into single-chain insulin expressed in mammalian cells (Fig. 6A). In both yeast and mammalian cells, however, we were unable to obtain positive evidence of unpaired cysteine residues for any of the B-chain point mutants studied, suggesting not a failure to oxidize but a failure to make the proper disulfide pairing (Figs. 3B and 6B). However, in this case, a major difference is observed between the two systems. Whereas in yeast, almost no processed singlechain insulin is secreted (secretion being composed primarily of the unprocessed ICFP), in mammalian cells (at least in HEK293 cells), the SCI is efficiently secreted regardless of the presence of B-chain point mutations or whether the disulfide pairing is correct or incorrect (Fig. 7). Remarkably, in these cells, even though single-chain insulins bearing Bchain point mutations are conformationally distinct from the wild type (e.g. Fig. 2), the novel conformations that they assume, including disulfide mispairing, are reasonably well folded (38,63) such that they are not recognized by the ER quality control machinery and they are allowed to be exported from both ER and Golgi compartments. Nevertheless, it is extremely unlikely that single-chain insulin isomers bearing improper disulfide pairing maintain full biological activity (35,38). Such mispairing will need to be considered in the further development of SCI constructs for potential use in gene therapy of type 1 diabetes (64).
Finally, we have examined the effects of the B9D, B10D, and B12E mutations in the context of the wild-type human proinsulin sequence and found that these mutations provide a similar tendency toward disulfide mispairing (Fig. 8), and similar to SCIs, the misfolded forms are nevertheless secreted from 293 cells (data not shown). It remains unknown whether such mutations cause the same degree of proinsulin misfolding when expressed in pancreatic beta cells (19) or in other specialized secretory cell types. However, it seems clear from these studies that the possibility of misfolding cannot be excluded simply by virtue of the fact that the proinsulin/insulin is secreted (43). It is likely that the content of protein disulfide isomerase, other ER oxidoreductases, and other ER molecular chaperones may vary between HEK293 cells and specialized secretory cells, and such activities may affect the ratio of disulfide isomers in the context of the insulin sequence (65)(66)(67).