Characterization of a nonglycosylated single chain urinary plasminogen activator secreted from yeast.

Using site-directed mutagenesis, we have changed the asparagine in human single-chain urinary plasminogen activator (u-PA) at position 302 to an alanine. This alteration removes the only known amino acid residue glycosylated in the protein. The single-chain u-PA containing an alanine residue at position 302 instead of asparagine (scu-PA(N302A] cDNA gene was expressed in the yeast Saccharomyces cerevisiae. Secretion of the protein product into the culture broth was achieved by replacing the human secretion signal codons with those from yeast invertase, adding a yeast promoter from the constitutively expressed glycolytic genes triosephosphate isomerase or phosphoglycerate kinase, and integrating multiple copies of these transcriptional units into the genome of yeast strains carrying the "supersecreting" mutation ssc1. When fermented in a fed-batch mode, these recombinant baker's yeast strains secreted scu-PA(N302A) in a strongly growth-associated manner. Greater than 90% of the u-PA found in the culture broth was in the single-chain form. Scu-PA(N302A) was purified to homogeneity using two chromatography steps. The purified protein had a molecular weight of 47,000 as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and lacked any detectable N-linked glycosylation. The in vitro fibrinolytic properties of scu-PA(N302A) were found to be essentially equivalent to those of natural single-chain u-PA derived from the human kidney cell line TCL-598. Since scu-PA(N302A) lacks the immunogenic N-linked carbohydrate pattern of yeast, it may be a useful therapeutic agent which can be produced economically by yeast fermentation.

Using site-directed mutagenesis, we have changed the asparagine in human single-chain urinary plasminogen activator (u-PA) at position 302 to an alanine. This alteration removes the only known amino acid residue glycosylated in the protein.
The single-chain u-PA containing an alanine residue at position 302 instead of asparagine (scu-PA(N302A)) cDNA gene was expressed in the yeast Saccharomyces cerevisiae. Secretion of the protein product into the culture broth was achieved by replacing the human secretion signal codons with those from yeast invertase, adding a yeast promoter from the constitutively expressed glycolytic genes triosephosphate isomerase or phosphoglycerate kinase, and integrating multiple copies of these transcriptional units into the genome of yeast strains carrying the "supersecreting" mutation sscl. When fermented in a fed-batch mode, these recombinant baker's yeast strains secreted scu-PA(N302A) in a strongly growth-associated manner. Greater than 90% of the u-PA found in the culture broth was in the single-chain form. Scu-PA(N302A) was purified to homogeneity using two chromatography steps. The purified protein had a molecular weight of 47,000 as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and lacked any detectable N-linked glycosylation.
The in vitro fibrinolytic properties of scu-PA(N302A) were found to be essentially equivalent to those of natural single-chain u-PA derived from the human kidney cell line TCL-598. Since scu-PA(N302A) lacks the immunogenic iv-linked carbohydrate pattern of yeast, it may be a useful therapeutic agent which can be produced economically by yeast fermentation.
Two immunologically distinct plasminogen activators, urinary plasminogen activator (u-PA)' and tissue plasminogen activator (t-PA), have been isolated from human tissue (l-3) and cDNA genes for both have been cloned (4, 5). Both enzymes exhibit kinetic parameters consistent with physiological roles in normal in uivo hemostasis.
Both share a number of structural features, including "kringle," epidermal growth factor-like, and serine protease catalytic domains, as well as plasmin-susceptible peptide bonds in the region between the kringle and catalytic domains. Both molecules are cleaved by plasmin to two-chain forms in which the two chains are held together by at least one disulfide bond. Both t-PA and scu-PA exhibit fibrin selectivity in their activation of plasminogen in both in vitro and in uiuo model systems and also in man (6-9). In other words, they catalyze the formation of active plasmin more efficiently in the presence of a fibrin clot than in the circulation.
Thus, both molecules spare circulating thrombogenic factors such as fibrinogen and (Yeantiplasmin.
Despite these similarities, the two molecules are also known to differ in a number of characteristics: (a) t-PA but not u-PA exhibits significant affinity for fibrin and soluble fibrin fragments (14); (b ) u-PA, through its epidermal growth factorlike domain, binds to a cell surface receptor while t-PA exhibits no measurable affinity for the u-PA receptor, but may bind to a different receptor on HUVEC cells (10, 11); (c) t-PA contains two domains not found in the u-PA molecule, a second kringle domain, and a finger domain resembling that found on fibronectin.
These additional domains of the t-PA molecule may be responsible for the fibrin affinity exhibited by t-PA (12, 13). However, fibrin selectivity must involve factors other than fibrin affinity because u-PA in its singlechain form (scu-PA) exhibits little or no fibrin affinity but appears to lyse clots with fibrin selectivity comparable to that of t-PA (8).
Fibrin selectivity during the conversion of plasminogen to plasmin distinguishes the activity of t-PA and scu-PA from that of urokinase (tcu-PA) and streptokinase and suggests that these molecules are more suitable for thrombolytic therapy in man. The demonstrated affinity of t-PA for fibrin, and kinetic evidence that the interaction with fibrin reduces the K, of t-PA for plasminogen, suggest a plausible mechanism by which t-PA may act in a fibrin-selective manner (1). By contrast, the mechanism by which scu-PA acts in a fibrinselective manner is unclear. For example, while scu-PA lacks measurable fibrin affinity and appears to exhibit little or no activity in uitro (XI), it acts as a potent, fibrin-selective thrombolytic agent in man and animals (7,8). In this report, we describe the production of human scu-PA in the baker's yeast Saccharomyces cerevisiae.
We chose to construct strains of yeast which secrete human scu-PA rather than produce it in the cytoplasm, because secretion appears important for accurate folding and disulfide bond formation in secretory proteins (16). Since human scu-PA is a 411-amino acid residue glycoprotein with an apparent molecular weight of about 53,000 and contains up to 12 disulfide bonds, we reasoned that refolding of aggregated scu-PA from inclusion bodies produced cytoplasmically in Escherichiu coli or yeast would be inefficient. While many mammalian proteins have been secreted from yeast, secretion of proteins larger than about 20,000 M, is typically inefficient with a large fraction of the protein remaining internal (17-20). Therefore, we took advantage of the yeast mutation sscl which has proven useful for the secretion of other mammalian proteins, including bovine prochymosin and growth hormone (18), to build a yeast strain which secretes at least two-thirds of the scu-PA synthesized into the culture broth. Human scu-PA normally contains carbohydrate linked to asparagine residue 302 (4). Yeast cells also carry out N-linked glycosylation, but unlike human cells would be expected to hyperglycosylate human scu-PA by adding a heterogeneous cluster of over 50 mannose residues as has been found for other secreted glycoproteins such as yeast invertase (21), yeast acid phosphatase (22), and human cY1-antitrypsin (23). Therefore, we altered the codon for amino acid residue 302 in the scu-PA cDNA to encode alanine instead of asparagine. We describe here the secretion from yeast, purification, and in vitro characterization of scu-PA(N302A) lacking any detectable N-linked carbohydrate. Except for the absence of Nlinked carbohydrate in in vitro assays, this scu-PA derivative is indistinguishable from scu-PA isolated from mammalian sources.

MATERIALS AND METHODS AND RESULTS'
The ssc Mutations Influence the Efficiency of Scu-PA Secretion by Yeast-Secretion of scu-PA was examined from both wild-type strains of yeast and from strains carrying ssc mutations which increase the secreted levels of other proteins of comparable size such as bovine prochymosin and growth hormone (18). The amounts of scu-PA secreted into the culture broth, expressed both as units per ml and normalized to cell mass, are shown in Table I. All strains were transformed with the same autonomously replicating plasmid carrying the human scu-PA cDNA transcribed from the yeast triosephosphate isomerase promoter and carrying the yeast invertase secretion signal in place of the human secretion signal. Clearly, cells carrying the sscl mutation secrete more scu-PA, both in absolute amounts and per cell, than do cells which are wild-type for secretion. A significant increase is also observed for cells carrying the ssc2 mutation, although the effect is less striking. It is clear from a comparison of Experiments 2 and 3 of Table I that scu-PA(N302A) is secreted from an sscl strain just as efficiently as is scu-PA. Thus, the substitution of alanine for asparagine at position 302 has no measurable effect on the secretion of scu-PA from yeast. Scu-PA(N302A) Secretion Is Growth-associated-Secretion of scu-PA by yeast was examined more closely under conditions of controlled pH, dissolved oxygen, and glucose feeding. Strain CGY1891, carrying the sscl mutation and integrated copies of scu-PA(N302A) transcriptional units, was grown in a modified YPD-type medium (see "Materials and Methods") for 138 h as shown in Fig. 3. Cell density and secreted scu-PA levels increased in parallel for the entire fermentation  indicating that secretion of scu-PA is growth-associated. Secreted scu-PA(N302A) levels as high as 1800 IU/ml were attained, corresponding to about 15 mg of scu-PA(N302A)/ liter. The fact that the amidolytic activity of unfractionated culture broth prior to plasmin treatment was always less than 5% of the activity obtained after plasmin treatment indicates that very little scu-PA was converted to tcu-PA during the fermentation (Table II). In addition, secreted scu-PA was stable in the unfractionated broth for several days at 4 "C. Scu-PA was also detected within the yeast cells during fermentation both by immunoblot and by activity in a fibrin plate; however, internal scu-PA antigen levels typically represented no more than about 30% of the total scu-PA detected (data not shown).
The Specific Activity of scu-PA(N302A)-Several properties of scu-PA(N302A) purified from yeast culture broth were compared to those of native scu-PA derived from human kidney cell line TCL-598. Neither purified preparation contained significant levels of plasmin-independent amidolytic activity, consistent with the fact that these preparations contain only trace levels of contaminating tcu-PA. In fact, the amidolytic activity of both scu-PA preparations prior to plasmin treatment was less than 1000 IU/mg, well within the range observed previously for scu-PA (15). After treatment with plasmin to activate any latent amidolytic activity, the specific activities of both preparations were virtually identical, in the range of lOO,OOO-120,000 IU/mg (data not shown).

Electrophoretic
Mobilities of Scu-PA and Scu-PA(N302A)- The electrophoretic mobilities of scu-PA(N302A) secreted from yeast as well as scu-PA secreted from both yeast and from Chinese hamster ovary cells were examined on polyacrylamide gels containing SDS both with and without prior treatment with N-glycanase to remove N-linked carbohydrate (Fig. 4) Therefore, as expected from the asparagine-to-alanine codon - 46,500 change, scu-PA(N302A) is secreted from yeast cells without the addition of N-linked carbohydrate. 36,500 Scu-PA(N302A) preparations contained traces of a singlechain form of u-PA having an M, of about 27,000 (Fig. 4). 26,600 This species is most likely the result of cleavage of scu-PA(N302A) after Glu-143, as has been observed previously (55).  (Table II) yeast and scu-PA obtained from Chinese hamster ovary cells were compared in vitro (Fig. 5). Equivalent amounts of the two plasminogen activators were incubated for up to 5 h in tubes containing 'Z51-fibrin-labeled plasma clots. The extent of clot lysis was judged by the amount of radioactivity released ("Materials and Methods"). At concentrations of 2.5, 1.25, and 0.63 pg/ml, scu-PA(N302A) and scu-PA exhibited comparable kinetics of clot lysis (Fig. 5, A and B). In addition, the levels of a*-antiplasmin remaining after lysis initiated by scu-PA and scu-PA(N302A) were comparable and significantly higher than the levels remaining after lysis initiated by tcu-PA (Fig. LX). Thus, by these assays, scu-PA(N302A) and scu-PA are virtually indistinguishable in their potency and fibrin selectivity of clot lysis in uitro.

DISCUSSION
These studies indicate that secretion by yeast of human urinary plasminogen activator modified to contain an alanine residue in place of the normally glycosylated asparagine residue at position 302 results in a fully active, fibrin-selective, scu-PA molecule which lacks any detectable N-linked carbo-hydrate. Because of the differences between yeast and mammalian glycosylation patterns, production of a nonglycosylated form of scu-PA may be the only practical way to avoid immunogenicity of this and other mammalian glycoproteins secreted from yeast. While both yeast and mammalian cells add identical preformed "core" mannose clusters from carrier lipids to asparagines, the sequence of events following this transfer differs considerably. Yeast cells process the core unit mainly by adding large numbers of mannose residues to the (Y~,~ backbone, while mammalian cells usually remove most of the mannose residues of the core and replace them with other sugars such as galactose, fucose, and sialic acid (47).
Ballou and co-workers (48) have determined several antigenie features of the yeast carbohydrate pattern. In addition, mannose receptors have been found on several cell types (49), and the presence of mannose-rich sugar on asparagines of human t-PA significantly reduces its circulating half-life (50). Therefore, mammalian proteins bearing a yeast pattern of carbohydrate may be antigenic and may also be cleared more rapidly from the mammalian circulation than their natural forms. In this regard, it is interesting to note that preliminary studies of the half-life of nonglycosylated scu-PA(N302A) in rabbits and dogs has demonstrated that the clearance of scu-PA(N302A) is very similar to that of native scu-PA isolated from human kidney cells3 The approach taken here of altering the amino acid sequence of the N-linked glycosylation sequon Asn-X-Thr/Ser should have general application for secretion of mammalian glycoproteins by yeast. Indeed, there have been reports of nonglycosylated murine and human GM-CSF secreted from yeast (51,52). In those cases, serine was substituted for asparagine at the first position of the Asn-X-Thr glycosylation sequon or valine was substituted for threonine at the third position.
There are no reliable rules for choosing a substitute amino acid residue, and certainly some residues may be detrimental to protein folding or stability. We chose alanine in this case because of its small side chain and its inability to be glycosylated.
Several factors are involved in obtaining efficient secretion of scu-PA from yeast. As observed previously for other mammalian proteins, the yeast invertase secretion signal is slightly more efficient for scu-PA secretion from yeast than is its natural mammalian secretion signal.4 Similarly, host strains carrying sscl or ssc2 mutations secrete scu-PA severalfold more efficiently than wild-type hosts. Interestingly, the presence of glycosylated asparagine or nonglycosylated alanine at position 302 had no detectable effect on the efficiency of secretion of scu-PA from yeast. Others have suggested that glycosylation is important for secretion of various mammalian proteins (52,53); but, in this case of heterologous protein secretion, glycosylation appears to be irrelevant. The use of a yeast secretion signal in place of the human signal normally found on scu-PA is important for obtaining efficient secretion from yeast, but it must be removed precisely if the resulting scu-PA is to be used therapeutically in man. In this case and also in the case of a-interferon (54), the invertase secretion signal was removed accurately during secretion from yeast. The fact that the junction between the invertase secretion signal and scu-PA, alanine-serine, is fortuitously the same as the junction between the invertase signal and invertase itself may be a factor in its accurate removal. The strain described in this report secretes seu-PA at a level corresponding to about 0.2% of the yeast-soluble cell protein. This represents about 2.5% of the protein in the rich culture medium used for these fermentations.
Purification required only two column chromatography steps and was accomplished with the relatively high yield of 43%. While scu-PA is quite sensitive to proteolytic cleavage by plasmin and thrombin in the "hinge" region between the two chains, scu-PA produced by yeast cells remained in the single-chain form even after storage of the crude fermentation broth for several days at 4 "C.
Nonglycosylated scu-PA molecules secreted by yeast may find application as therapeutic agents in man. Indeed, with the exception of its carbohydrate content and the amino acid residue at position 302, scu-PA(N302A) appears indistinguishable in its in vitro properties from native scu-PA produced by mammalian cells. The preservation of the molecule in its single-chain form during yeast fermentation is encouraging because mammalian cell processes frequently produce mixtures of single-and two-chain u-PA in which the twochain molecules may constitute over 50% of the product. The yeast produced single-chain u-PA appears fibrin-selective in its activation of plasminogen (Fig. 5).  in our preparations (Fig. 4) is unlikely to affect interpretation of these results because it is a quite minor component in our preparation and because Stump et al. (55) have shown that a similar low molecular weight scu-PA is as fibrin-selective as full-length native scu-PA. Studies in the rabbit venous thrombosis model and in the dog arterial thrombosis model have confirmed the fibrin selectivity of our preparations of scuPA(N302A) in viva.' Finally, the successful construction of yeast strains which secrete detectable amounts of scu-PA now allows dissection of the structurefunction relationships of this protein by coupling powerful molecular genetic mutagenic techniques with a Petri plate activity assay for scu-PA secreted from individual yeast colonies.