The role of the carrier protein and disulfide formation in the folding of beta-lactamase fusion proteins in the endoplasmic reticulum of yeast.

We have studied the relationship between folding and secretion competence of hsp150-beta lactamase fusion proteins in Saccharomyces cerevisiae. hsp150 is a secretory protein of yeast, and beta-lactamase was chosen, since its folding can be monitored by assaying its enzymatic activity. The hsp150 pre-pro-protein consists of a signal peptide, subunit I, a repetitive region, and a unique C terminus. Fusion of beta-lactamase to the C terminus of hsp150 produced Cla-bla protein, which was secretion-competent but inactive. The Pst-bla protein, where beta-lactamase was fused to subunit I, was also inactive and mostly secreted, but part of it remained in the pre-Golgi compartment. When beta-lactamase was fused to the C-terminus of the repetitive region, the fusion protein (Kpn-bla) was translocated to the endoplasmic reticulum, acquired disulfide bonds, and adopted an enzymatically active conformation. Kpn-bla was secreted to the medium without decrease of specific activity or retention in the cell. Folding of Kpn-bla to an active and transport-competent form required co-translational disulfide formation, since treatment of cells with dithiothreitol resulted in endoplasmic reticulum-retained inactive Kpn-bla. When dithiothreitol was removed, Kpn-bla resumed transport competence but remained inactive. Reduction of prefolded Kpn-bla did not inhibit enzymatic activity or transport. The repetitive hsp150 carrier may have use in heterologous protein production by conferring secretion competence to foreign proteins in yeast.

Fax: 358-0-4346028. lation (2). Recently, a convenient method was introduced to manipulate proteins by reducing their disulfides by treating live cells with DTT (7-9). We have applied this method to yeast cells, and shown that the disulfides of the secretory hspl50 protein and the vacuolar enzyme carboxypeptidase Y (CPY) could be reduced with DTT, causing retention of the proteins in the ER, while the secretory apparatus remained functional.
The sulfhydryls could be reoxidized by removal of DTT, resulting in resumption of transport (10). Here we studied in S.
cerevisiae the relationship of folding and secretion competence following the fate of p-lactamase fusion proteins, whose folding could be monitored by assaying their p-lactamase activity.
Construction of the HSPl5O-bla Fusion Genes and the Integration Vector-From the genomic clone ofiYSF'150 (ll), a SalI-ClaI DNAfragment, which contains the HSP150 gene devoid of its four last codons, plus approximately 2 kb of upstream sequence, was cloned between the SalI and ClaI sites of Bluescript I1 SK-(Stratagene). The 450 bp ADCl terminator from pAAH5 (12) was ligated downstream from HSP150, between the HindIII and BamHI sites. The KpnI site of the vector was removed by digesting approximately 100 bp of the plasmid with Ba131 nuclease starting from the SalI site, and a BamHI linker was added. The bla gene without its signal sequence was synthesized with polymerase chain reaction using Pfu polymerase (Stratagene), and pUC8 as template. The 3' primer A (5'-GCAACCAAGCTTGAGTAAACTTG-GTCTGACAG) containing a HindIII site was used for all constructions.
The 5' primer B (5'-GCTTATATCGATGGTACCTGCAGTCACCCA- GAAACGCTGGTG) was used for CZa-bla and Kpn-bla constructions, and the 5' primer C (5'-GCTTATCTGCAGCTCACCCAGAAAC- GCTGG) for the Pst-bla construction. The suitably digested polymerase chain reaction products were ligated between the ClaI and HindIII, KpnI and HindIII, and PstI and HindIII sites to create the Cla-bla, Kpn-bla, and Pst-bla constructions, respectively, which were trans-  1. A, schematic presentation of the products of the HSP150-bla genes. The primary translation products of the HSP150 gene, the bla gene, and the Cla-bla, Kpn-bla, and Pst-bla genes, where bla was fused to different fragments ofHSP150, are shown. The numbers refer to the last amino the linkers between the hspl50 and p-lactamase moieties. hspl50 has cysteines Cys3I6, Cys3", Cys4", and Cys4I3. Kpn-bla has cysteines Cys3I6 acid residue of each domain. SS, signal peptide; SUI, subunit I; SUII, subunit 11; rr, repetitive region. Uppercase letters denote the amino acids of (hspl50-fragment), Cys322 (linker), and CYS"~ and C Y S~~' (p-lactamase). B, Northern analysis of chimeric mRNA molecules. RNA was isolated from strains H337 (Cla-bla, C ) , H335 (Kpn-bla, K), H395 (Pst-bla, P), and H1. Prior to RNA extraction, the cells were grown overnight a t 24 "C and incubated for 30 min a t 24 "C (HS-) or 37 "C (HS+). 10 pg of each RNA sample was subjected to Northern analysis, probing the membrane successively with HSP150 ( a ) , bla (b), and ACT1 (c) DNA. The estimated sizes (kb) of the RNA molecules are indicated.
Assay of p-Lactamase Activity-The p-lactamase activity was assayed using Nitrocefin (Glaxo), according to Ref. 16, but a t room temperature. Culture medium samples were assayed after removal of the cells by centrifugation. For the determination of p-lactamase activity of whole cells, the pelleted cells were lysed and suspended in 100 m M potassium phosphate, pH 7, to the original sample volume. To determine the p-lactamase activity entrapped in the cell wall (Nitrocefin is non-penetrable), washed cells were suspended in 1 ml of the potassium phosphate buffer containing 10 m M NaN,, to the original volume. The intracellular activity was calculated by subtracting the cell wall activity from the whole cell activity. Secretory p-lactamase was not nonspecifically bound to the cells, as shown by assaying the p-lactamase activity of H1 cells incubated with Kpn-bla-containing growth medium.
In Vitro Dunslation-The Pst-bla construct was placed under the control of the lac promoter in pUC18, a t the BamHI site. The plasmid was digested with PvuII, and a 1.9-kb fragment containing both the promoter and the fusion gene was isolated to serve as a template for in vitro transcriptiodtranslation, using the E. coli S30 Extract System for Linear Templates (Promega) according to manufacturer's instructions.
Other Procedures-Northern analysis of total RNA was performed, as described (17). The stability of the transformants was studied by growing them to early, middle, and late logarithmic phase in YPD medium, followed by plating on YPD plates. The p-lactamase activity of 60 colonies from each plate was assayed by mixing them with Nitrocefin solution on microtiter plates; all produced p-lactamase activity and grew on SC-URA plates. SDS-PAGE was in reducing 8% gels unless otherwise stated. Non-reducing SDS-PAGE and in vivo protein reduction were performed as before (10). Precipitation was with 14% trichloroacetic acid for 1 h on ice. Zymolyase lOOT was from Seikagaku; cycloheximide and D'IT were from Sigma. For scanning, an LKB Ultroscan XL laser densitometer was used.

RESULTS
Fusion of the p-Lactamase Gene to Fragments of the HSP150 Gene-"he product of the HSP150 gene is composed of four domains: (i) the signal peptide, (ii) subunit I, (iii) the repetitive region (rr) of subunit 11, and (iv) the C-terminal region of subunit I1 (Fig. IA). The repetitive region consists of an 11-fold repeat of a 19-amino acid peptide. The signal peptide is cleaved off in the ER, and subunit I is detached from subunit I1 a t a kex2 recognition site. The subunits remain noncovalently associated and are efficiently secreted to the culture medium (11).
We wanted to study whether fusion of p-lactamase to fragments of hspl50 would allow folding of the p-lactamase portion to an enzymatically active conformation, and whether the fusion protein would be secreted to the growth medium. "he E. coli TEM p-lactamase gene, bla, lacking its signal sequence, was fused to various portions of the HSP150 gene. p-Lactamase was chosen as a reporter, because its folding can be monitored by determining its enzymatic activity (18) and because it is not secreted in S. cereuisiae in its authentic form (19, 20). The largest construct, Clu-bla, lacked the last four codons of HSP150. Kpn-bla contained the signal sequence, and codons for SUI and the repetitive region. Pst-bla contained the signal sequence and 45 codons of SUI. The fusions were introduced in an integrative derivative of yEp24 to strain H1, to create strains H337 (Clu-bla), H335 (Kpn-blu), and H395 (Pst-blu). Southern analysis showed that the plasmids had integrated as single copies into the URA3 locus (not shown), and integration appeared stable (see "Materials and Methods").
Dunscription of the Chimeric Genes-The chimeric genes were under the control of the heat-regulated HSP150 promoter (17). Two RNA samples were extracted: one from cells incubated at 24 "C (HS-) and the other from cells heat-shocked for 30 min a t 37 "C (HS+). Northern analysis using an HSP150 probe (Fig. lL3, panel a ) 6 ) .
Immunoprecipitation of the culture medium of H335 cells (Kpn-bla), 35S-labeled like the H337 cells above, revealed a new protein of about 135 kDa which was recognized by anti-hspl50 ( Fig. 2 B , lane 1 ) and anti-p-lactamase (lane 3 ) . I t was designated Kpn-bla. Very little of Kpn-bla-related proteins could be detected in the lysate (lanes 2 and 41, indicating that Kpn-bla was secreted efficiently to the growth medium. When H395 cells (Pst-bla) were analyzed, anti-hspl50 detected only hspl50 (Fig. 2C,panel a , lanes l and 2), because the antiserum was raised against subunit I1 (see Fig. lA). Anti-plactamase detected a 48-kDa protein from the culture medium (lane 3 ) , and a 44-kDa protein from the cell lysate (lane 4 ) . According to densitometric scanning, about 35% of Pst-bla (32 and 38% in two experiments) remained cell-associated, and the rest was secreted. The 44-kDa protein was the pre-Golgi form, since Pst-bla labeled at 37 "C in H421 cells migrated like a 42-kDa protein (Fig. 2C, panel b, lane I). In  To compare the relative amounts of the three fusion proteins secreted to the culture medium, they were 35S-labeled and subjected to immunoprecipitation and SDS-PAGE analysis in parallel. According to scanning of the bands, similar amounts of Cla-bla (10 Met, 6 Cys) and Kpn-bla (9 Met, 4 Cys) were detected in the media (Fig. 3 A , lanes l and 2, open arrowhead), whereas the amount of Pst-bla (9 Met, 2 Cys) was roughly one-fourth of that of Cla-bla (lane 3 , black arrowhead). This was confirmed by Western analysis of culture medium samples of strains H1, H337 ( C ) , H335 ( K ) and H395 (PI, using anti-plactamase (Fig. 3B, panel a ) and anti-hspl50 (panel b). Thus, Kpn-bla was efficiently secreted to the culture medium, apparently without significant retardation in secretory organelles, or in the cell wall or periplasm. Cla-bla was transport-competent, part of it remaining cell-associated. Pst-bla was translocated to the ER, but only small amounts were secreted. Pst-bla was evidently harmful for the cells, since the generation time of strain H395 was 3 h, 25 min, whereas that of the other transformants and the parental strain was 2 h. p-lactamase Activity of the Fusion Proteins-To monitor the folding of the fusion proteins, their p-lactamase activities were measured. When strain H335 (Kpn-bla), was incubated at 24 "C, 0.31 unitlml p-lactamase activity was in the culture medium after 3 h (Fig. 4C, triangles). Much less activity was found in the cell walVperiplasm (circles) and intracellularly (diamonds). When the incubation was performed at 37 "C, to enhance synthesis by heat shock, 1.62 unitdm1 p-lactamase activity was detected in the culture medium, 0.29 unitlml in the cell wall, and only traces intracellularly (Fig. 40). Very little p-lactamase activity could be detected at either temperature in the media or cells of strains H337 (Cla-bla) (Fig. 4, A andB) or H395 (Pst-bla) (Fig. 4, E and F).
To study whether the fusion proteins were enzymatically active in the ER, the Kpn-bla and Pst-bla constructs were expressed in a s e d 8 mutant (strains H393 and H421, respectively). When H393 (Kpn-bla) was incubated for 90 min at 37.5 "C, p-lactamase activity accumulated inside the cells (

5A, triangles), and very little was in the medium (closed circles).
When cycloheximide was added (arrowhead), and the cells were shifted to 24 "C, the cell-associated activity decreased (closed triangles), while the activity in the medium increased, showing that the accumulated fusion protein was secreted. Thus, the p-lactamase portion of the Kpn-bla fusion protein folded in the ER to an active conformation, and the specific activity of the fusion protein did not change significantly after exit from the ER. In contrast, the Pst-bla molecules accumulating in the ER in strain H421 had very little enzymatic activity (Fig. 5B 1. The Effect of Reduction of the Disulfides of Kpn-bla on Its Z'kansport Competence-Next we wanted to manipulate the conformation of intracellular Kpn-bla by reducing its possible disulfides (lo), to study the consequences on transport. H335 cells were metabolically labeled in the presence of DTT. Immunoprecipitations with anti-hspl50 (Fig. 6A, panel a ) and antibla (panel b ) showed that the fusion protein was not secreted (lanes 3 ) . Cell-associated Kpn-bla was barely visible (lanes 41, probably due to heterogeneity of the glycans. Nevertheless, it could be efficiently chased to the growth medium after the removal of DTT (lanes 5-12). Thus, under normal conditions, Kpn-bla acquired disulfides, which were reduced by DTT treatment, leading to reversible intracellular retention.
When Kpn-bla was synthesized in the presence of DTT, P-lactamase activity could not be detected in the culture medium (Fig. 6B,panelb,

closed circles) or intracellularly (open circles).
Removal of DTT (arrowhead) did not increase p-lactamase activity. In vitro reduction of secreted Kpn-bla did not inactivate the enzyme (not shown). Thus, synthesis under reducing conditions produced inactive Kpn-bla, and removal of DTT resulted in resumption of transport competence of the molecules. However, the reoxidized structure was different from the native one, since it had no enzymatic activity.
Kpn-bla was then allowed to fold in the absence of DTT, while retaining it at 37.5 "C in the ER in strain H393 (sec18) (Fig. 5A). Part of the cells were incubated further a t 24 "C in the presence of cycloheximide only, which led to secretion of the molecules, as described above (arrowhead, filled symbols). Addition of DTT with cycloheximide to part of the cells had no effect on the se- cretion or enzymatic activity of Kpn-bla (open symbols). Native and in vivo reduced, ER-trapped Kpn-bla migrated differently in non-reducing SDS-PAGE, the latter comigrating with in vitro completely reduced ER-trapped Kpn-bla (not shown). This demonstrated that prefolded Kpn-bla was not resistant to DTT.

Folding of P-Lactamase Fusion Proteins in the
Since the reduction of prefolded Kpn-bla did not inhibit its enzymatic activity or secretion, we assume that co-translational disulfide formation was required for the folding of the fusion protein into an enzymatically active conformation and that the site of retention of reduced Kpn-bla was the ER. Continuous Secretion of Active Kpn-bla-The above results suggest that the Kpn fragment of hspl50 may have biotechnological potential for production of heterologous proteins, which, without a carrier, are retained in the yeast cells. Therefore we studied the production of Kpn-bla more closely. When strain H335 (Kpn-bla) was cultivated at 24 "C, p-lactamase activity in the medium increased as long as the cells were growing, reaching a maximum of 18 units/ml in the stationary phase (Fig. 7A,  panel a, triangles). When the cells were grown to a density of A,,, = 1 at 24 "C and cultured further at 37 "C, the activity reached a maximum of 29 unitdm1 (Fig. 7A, panel b, triangles). The fusion protein appeared stable at both temperatures.
Proteins from the culture medium of H335 cells, grown at 30 "C to different densities, were trichloroacetic acid-precipi-tated and resolved in SDS-PAGE. Kpn-bla appeared to be the major protein detected by Coomassie Blue staining (Fig. 7B, large arrowhead). Its amount at A,,, = 11 was estimated to be about 2 gg/ml, as compared to the staining of BSA (Fig. 7B). This may be an underestimation, since hspl50 (small arrowhead) is very poorly stained by Coomassie Blue. According to Western analysis, its amount in the medium was similar to that of Kpn-bla (Fig. 3B, panel b).

DISCUSSION
We fused the TEM p-lactamase of E. coli, lacking its own signal sequence, to the C termini of various portions of the yeast secretory glycoprotein hspl50. The chimeric proteins were used to study the effect of the hspl50-carrier sequence on the folding of the p-lactamase portion and the effect of folding on the secretion competence of the fusion proteins. In the largest chimera, Cla-bla, hspl50 lacked its 4 C-terminal amino acids. In Kpn-bla, 0-lactamase was fused to the repetitive region. The smallest fusion protein, Pst-bla, contained the signal peptide and two-thirds of subunit I (see Fig. lA). The folding of the p-lactamase portion was monitored by assaying its enzymatic activity (18). The authentic structural bla gene produces in S. cerevisiae enzymatically inactive, cell-associated pre-Plactamase (19, 20).
The HSPl50-bla fusion genes were transcribed similarly. Their products were translocated to the ER but differed in p-lactamase activity and secretion competence. Pst-bla did not gain an enzymatically active structure in the ER. Less Pst-bla was detected in the culture medium than Cla-bla or Kpn-bla, and one-third of it appeared to remain in the pre-Golgi compartment. The almost entire hspl50 fragment rendered Cla-bla secretion-competent, some remaining in the Golgi, periplasm, or cell wall. However, Cla-bla had virtually no p-lactamase activity. It seems that the incomplete subunit I and the Cterminal domain of hspl50 interfered somehow with the folding of the p-lactamase portion, causing malfolding.
In contrast to Pst-bla and Cla-bla, Kpn-bla adopted an enzymatically active conformation and acquired disulfide bonds in the ER. It was efficiently secreted to the culture medium, without decrease of specific activity, or retention in the secretory route or the cell walllperiplasm. Partial secretion of active, unstable p-lactamase has been achieved using the invertase signal sequence (21), but the secreted activity was 1/30 that of Kpn-bla.
We were able to distort the conformation of the fusion proteins by reducing their disulfides in vivo. When disulfide formation of Kpn-bla was prevented by DTT treatment of cells, the protein was inactive and retained in the ER. Removal of DTT allowed Kpn-bla to adopt a secretion-competent structure, which was different from the native form, since the enzymatic activity was not resumed. DTT treatment distorted the folding of Kpn-bla severely only co-translationally. Once Kpn-bla was allowed to fold normally, subsequent reduction did not affect its enzymatic activity or exit from the pre-Golgi compartment. The role of protein folding in secretion in yeast has been observed before in the case of human a-amylase, where substitution of Cys,,, abolished secretion and activity (22). In the case ofyeast acid phosphatase, the removal of N-glycosylation sites caused malfolding and irreversible ER retention (23). Reduction of hspl50 and carboxypeptidase Y by DTT treatment resulted in their retention in the ER. This must have been due to malfolding and not free sulfbydrylsper se, since sulfhydryl-containing invertase was secreted normally in the presence of DTT (10). Thus, yeast clearly has a quality control apparatus, which monitors proteins and retains them in the ER, unless they display acceptable structural features. For foreign proteins, the transport-competent conformation does not need to be the biologically active structure, since Cla-bla, Pst-bla, and reoxidized Kpn-bla were se- leader (6, 31-37). Pro-a-factor is a secretory protein of yeast.