Two Novel Streptomyces Protein Protease Inhibitors PURIFICATION, ACTIVITY, CLONING, AND EXPRESSION*

a of In an effort to understand the synthesis and secretion of proteins by Streptomyces, we identified and characterized two naturally occurring abundantly produced proteins in culture supernatants of Streptomyces lividam Streptomyces longisporus. We purified these 10-kDa proteins and obtained partial amino acid sequence in- formation which was then used to design oligonucleotide probes in order to clone their genes. Analysis of the sequence data indicated that these proteins were related to each other and to several other previously characterized Streptomyces protein protease inhibitors. We demonstrate that both proteins are protein protease inhibitors with specificity for trypsin-like en- zymes. The presumptive signal peptidase cleavage sites and subsequent aminopeptidase products of each pro- tein are characterized. Finally, we show that the cloned genes contain all of the information necessary to direct synthesis and secretion of the proteins by Escherichia

ment (1). Unlike the Gram-negative bacteria, e.g. Escherichia coli, Gram-positive bacteria lack a cell wall and are therefore able to secrete proteins directly into the external milieu. Accordingly, a variety of extracellular degradative enzyme activities produced by Streptomyces have been identified, which can hydrolyze proteins, lipids, or complex carbohydrates (2-8). Besides hydrolytic enzymes, Streptomyces spp. secrete a variety of enzyme inhibitors such as an inhibitor of a-amylase (9, 10) and both peptide (11) and protein protease inhibitors (12, 13). These proteins range in size from <lo kDa, e.g. tendamistat (9), to >lo0 kDa, e.g. @-galactosidase (5).
In the present study, we looked for proteins which were efficiently produced and secreted into the growth medium by Streptomyces lividans and Streptomyces langi-sporus. For both species we identified abundant proteins with apparent molecular weights of -10,000 by SDS-PAGE' analysis. We purified both proteins, determined amino acid sequences of tryptic peptides, and used this information to clone the genes for the proteins. The proteins and their genes exhibit a high degree of homology and as expected, the genes encode presumptive signal peptides required for secretion of the proteins. The proteins are protein protease inhibitors related to the Streptomyces subtilisin inhibitor isolated from Streptomyces albo- griseolus (14,15) and the plasmin inhibitor, plasminostreptin, isolated from Streptomyces arztifibrimZyticus (16, 17). The proteins exhibit extensive NHz-terminal heterogeneity which, at least in part, is related to the Iength of time in culture. By comparison of the NH2-terminal sequences of the proteins obtained at different times of culture we have been able to define the apparent signal peptidase processing site for each protein as well as some subsequent extracellular processing products. Finally we show that expression of both proteins in E. coli leads to accumulation of the products in the periplasmic space. Apparently, the Streptomyces-derived translation initiation and signal sequence elements function in this heterologous bacterium. The abbreviations used are: SDS-PAGE, sodium dodecyl sulfatepolyacrylamide gel electrophoresis; SSI, Streptomyces subtilisin inhibitor; PSN, plasminostreptin; STI1, Streptomyces trypsin inhibitor 1; stil, the gene for STI1; STI2, Streptomyces trypsin inhibitor 2; sti2, the gene for STI2; API-~c, alkaline protease inhibitor-2c; BPTI, bovine pancreatic trypsin inhibitor; TBST, tris-hydroxymethylaminomethane-buffered saline with Tween-20; RP-HPLC, reverse-phase high performance liquid chromatography; kb, kilobase paids); bp, base pair(s).
Purification of Extracellular Proteins-Initially the small protein produced by S. liuidans, herein designated ST11 (Streptomyces irypsin inhibitor l), was identified as an abundant, exported protein in the culture supernatants of S. liuidans grown in SL-glycerol medium for 30 h at 28 "C. Culture supernatants were isolated and filtered, first through cheesecloth and then through a 0.45-pm Nalge filter. The filtrate was concentrated 10-fold in a hollow-fiber filter apparatus (Amicon,. Solid (NH4)&04 was added to 60% of saturation and the resulting slurry was stirred overnight at 4 "C. Precipitated proteins were recovered by centrifugation (12,000 X g, 20 rnin), redissolved in 0.01 M sodium phosphate buffer, pH 7.0, and dialyzed against the same buffer. The dialysate was then subjected to electrophoresis on a 3-mm-thick 15% preparative SDS-polyacrylamide gel (27). A guide lane was stained to localize the protein, the corresponding section of the unstained gel was excised, and the protein was electroeluted. This procedure yielded approximately 1 mg of STIl/ liter of culture supernatant. The majority of the STIl sequence data were obtained with protein isolated in this fashion.
For functional assays, nondenatured ST11 was isolated from unconcentrated medium by the addition of (NH&S04 to 65% of saturation. The resulting slurry was stirred for 30 min at 25 "C, and precipitated proteins were recovered by centrifugation (5,000 X g, 30 rnin). The precipitate was redissolved in 10 mM ammonium acetate buffer, pH 5.0, and dialyzed against the same buffer. The dialysate was applied to an S-Sepharose Fast Flow column (Pharmacia LKB Biotechnology Inc.), 2.5 X 10 cm, equilibrated in 10 mM NH4Ac, pH 5.0, at 2.5 ml/min. The column was eluted with a linear gradient from 10 to 500 mM NH,Ac, pH 5.0. ST11 eluted at about 300 mM NH,Ac. Further purification of the protein could be obtained, when necessary, by reverse-phase high performance liquid chromatography (RP-HPLC).
The small protein observed in culture supernatants of S. longispopus and designated ST12 (Streptomyces pypsin inhibitor 2) was purified by a two step procehure involving (NH4),S04 induced aggregation (65% of saturation) followed by cation exchange chromatography. S. longisporus was grown in trypticase-soy broth medium for 72 h at 28 "C. Under these conditions, the organism produced a large amount of a melanin-like pigment. On addition of (NH4)2S04, this pigment polymerized to form a dense mat of black fibrous material. Precipitated ST12 was caught in this material and the entire aggregate floated. The insoluble material was collected by centrifugation (5,000 X g, 30 min), both from the precipitate and from the mat floating on top of the supernatant solution. The aggregate was resuspended in 0.05% trifluoroacetic acid and dialyzed for 16 h at 4 "C against 10 mM NHIAc, pH 6.0. Insoluble material was removed by filtration and the clarified sample was applied to an S-Sepharose Fast Flow column (Pharmacia), 2.5 X 15 cm, equilibrated in the dialysis buffer. The column was eluted with a linear gradient from 10 to 250 mM NH4Ac, pH 6.0. ST12 eluted at about 100 mM NH4Ac. From 50 to 100 mg of ST12 could be recovered from 1 liter of culture supernatant. Recombinant ST12 expressed in S. liuidans, which does not produce melanin, was purified in the same way, except that during the ammonium sulfate fractionation step, the protein precipitated in the usual way.
Production of Antibodies-New Zealand White rabbits were inoculated subscapularly with 100 pg of either gel-purified STIl or chromatographically-purified STI2. The rabbits were boosted 10 or 14 days later using another 100 pg of the same protein in incomplete Freund's adjuvant. Sera were collected 2 weeks after the boost and tested for reactivity by Western blotting. Anti-ST11 antibodies crossreacted with STI2; however, we found that we could obtain minimal reaction of anti-ST12 antibodies against STIl by using a 1:500 dilution of the antiserum and low amounts of antigen.
Amino Acid Sequence Analysis-The Streptomyces proteins were denatured and reduced in a solution of 6 M guanidinium chloride; 500 mM Tris-HC1, pH 8.1, which had been saturated with CH&N (-33% v/v). Cysteine residues were then blocked with dansylaziridine (28) to yield a fluorescent derivative. The proteins were recovered by * M. Brawner, personal communication.
precipitation with ethanol. The precipitated proteins were redissolved in freshly deionized 10 M urea and then diluted 1:l with 100 mM NH,HC03. Trypsin (L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated, Cooper Biomedical) was added at 0 and 2 h of incubation at 37 "C to yield a final enzyme to substrate ratio of 4100 (w/w). The resulting digests were fractionated by RP-HPLC on a Brownlee RP300 octyl-silica column (4.6 X 250 mm). The effluent from the column was monitored for absorbance at 214 nm and for fluorescence (excitation at 230 nm, emission above 440 nm). Peptide peaks were collected manually and lyophilized.
Protein or peptide samples were subjected to vapor-phase hydrolysis with 5.7 N constant boiling HC1 (Pierce Chemical Co.) for 16-18 h at 110 "C in uacuo. The hydrolysates were analyzed on a Beckman 6300 Amino Acid Analyzer. For sequence analysis, samples were dissolved in either 25% trifluoroacetic acid or 0.1% SDS and subjected to automated Edman degradation in a Beckman model 890 M Sequenator. Released phenylthiohydantoin-derivatives were identified by RP-HPLC on a Beckman Ultrasphere ODS column (4.6 X 250 mm) .
Trypsin Assay-Trypsin activity was measured using benzoylarginine-p-nitroanilide as substrate. Assays were performed in microtiter plates in a total volume of 200 pl. The buffer was 50 mM Tris-HC1, pH 8.0, 1 mM EDTA and contained 1 mg/ml bovine serum albumin. The enzyme concentration was held constant at 0.2 pM, and the inhibitor and substrate concentrations were varied over the range 0.025-0.5 p~ and 0.25-2 mM, respectively. The plate was read at 405 nm in a V,, plate reader and the data were collected using an IBM PS/2 model 60. Only assays for which linear rates could be determined were used for subsequent analyses. Aprotinin (Sigma) and bovine pancreatic trypsin inhibitor (BPTI, Sigma) were assayed in the same system for comparison with STIl and STI2.
Charon Phage Cloning of the Gene Encoding STIl (sti1)"A genomic library of S. liuidans 1326 was prepared using the X phage vector Charon 30. Total DNA from S. liuidans 1326 was partially digested with Sau3A, size-fractionated on sucrose gradients, ligated to BamHI-digested Charon 30 DNA, and packaged in uitro using standard techniques (29). The Sau3A fragments used in the cloning were in the size range of 10 to 30 kb. The library was screened for STIl sequences by plaque hybridization using a 32P-labeled mixed oligonucleotide probe. The probe was synthesized as a mixture of 24 degenerate sequences complementary to the mRNA sequence predicted from amino acids 92-99 of the mature STIl protein sequence and taking into consideration the Streptomyces codon usage bias (30). The sequence of the probe was as follows.
Protein Localization in E. coli-E. coli strain JMlOl containing ST11 plasmid constructions were grown at 37 "C in LB medium containing ampicillin (100 pg/ml) and isopropylthiogalactoside (1 mM). At various times during growth, 1-ml samples were removed, and the cells were shocked with chloroform to release periplasmic proteins as described (31). Cytoplasmic proteins were released from chloroform-shocked cells by sonication. Proteins were then examined by Western blot for the presence of STI1.
Construction of End-labeled Probes for SI Nuclease Mapping-An STIl probe was prepared by end labeling the FokI site within the stil gene contained on the plasmid pHN1. pHNl was constructed by

RESULTS AND DISCUSSION
We examined culture supernatants of Streptomyces spp. for abundantly produced proteins by precipitation with trichloroacetic acid followed by electrophoresis on SDS-polyacrylamide gels and Coomassie Blue staining (Fig. 1). Culture supernatants from both S. liuidans and S. longisporus contained 10-kDa polypeptides as major secreted protein species? We decided to clone the genes for these polypeptides in order to obtain DNA sequences encoding the proteins, their leader (secretory signal) sequences, and the transcriptional and translational elements responsible for their expression.
Cloning the stil Gene of S. liuidans-The 10-kDa protein defined by SDS-PAGE analysis was purified from S. lividans conditioned medium after 30 h in culture by a combination of (NH4),S04 precipitation and preparative SDS-PAGE as described under "Experimental Procedures." The reduced denatured protein was digested with trypsin and the fragments isolated by RP-HPLC. Sequence analysis of the resulting peptides provided sufficient information to account for >80% of the protein. The sequence of one of the peptides, Ala-Phe-Ala-Asn-Glu-Cys-Val-Lys, was used in conjunction with a Streptomyces codon usage bias table (30) to design a 24-bp 24-fold degenerate oligonucleotide probe for the mRNA en-Two points of interest should be noted with regard to this figure. First, we were not concerned with absolute amounts, Le. we did not determine whether or not we were quantitatively precipitating every protein in the culture supernatant. Our purpose was merely to try to identify major secreted species. Second, the reader should bear in mind that the sample derived from S. longisporus represents a 72-h time point. Other proteins are evident in the culture supernatant at earlier times; however, these proteins appear to turnover without being replaced, whereas the 10-kDa species continues to accumulate. The net result is that by 72 h, the 10-kDa species accounts for >80% of the total protein in the sample. Other protein bands were evident on the original gel, but were lost during photographic reproduction. coding this peptide. This oligonucleotide mixture was used to probe a Charon 30 phage genomic library from S. liuidans.
Putative clones were detected with a frequency of 3 X One recombinant clone, Ch25.5, containing an insert of approximately 18 kb, was selected for further study. A 15-kb BglII-EcoRI fragment of this insert, containing sequences which hybridized with the probe, was subcloned into pUC18 to give the plasmid pD1.
The gene encoding the protein was localized to a 3-kb PstI-BamHI DNA fragment from pD1 and then to a 180 bp RsaI fragment by probing Southern blots with the mixed oligonucleotide probe. The RsaI fragment was subcloned into M13 mplO and sequenced (32). Analysis of this sequence indicated an open reading frame which agreed with tryptic peptide amino acid sequence data. Additional nucleotide sequence was obtained by cloning and sequencing restriction fragments flanking the 180-bp RsaI fragment. A restriction map for the gene region indicating the sequencing strategy is shown in Fig. 2A. The complete nucleotide and amino acid sequences are shown in Fig. 3A. Amino acid sequence analysis of the full-length protein isolated as described, revealed marked heterogeneity at the NH, terminus. Sequences were determined consistent with starts at Ser' (-48%, nucleotide +105) and Tyr3 (-52%, nucleotide +111). We have assigned Ser as the NH2 terminus of the mature protein since it defines the longest sequence found. The nucleotide sequence also predicts a 35-amino acid stretch that resembles an amino-terminal signal peptide. This sequence contains two potential ATG initiation codons. We have assigned the ATG at position +1 as the initiation codon based on its spacing relative to the putative ribosome binding site AAGGA found 13 bp upstream (underlined, Fig. 3A); however, we cannot rule out the possibility that the ATG at position -9 is used. We designate this protein ST11 and the gene stil.
The 5'-end of the stil transcript was determined by S1 nuclease mapping using a 159-bp EcoRI-FokI DNA fragment (see "Experimental Procedures") and RNA purified from S.  identified at position -70 (Fig. 3A).
Cloning the sti2 Gene of S. longisporus-The analogous 10-kDa protein produced by S. longisporus was purified from conditioned medium after 72 h of culture as described under "Experimental Procedures." Partial sequence analysis of the protein indicated that it was homologous to STI1, and thus, a 180-bp fragment of the stil gene was used to probe a Southern blot of BamHI digested S. longisporus DNA. A hybridizing fragment was localized in the 2.1-kb region. S. longisporus DNA was then cut with BamHI, and a limited library of fragments of -2 kb were cloned into the Streptomyces vector pIJ703 (20). The resulting library was transformed into S. lividans 1326/12K. In order to identify clones encoding the S. longisporus 10-kDa protein, the transformants were screened for direct expression of the protein by colony immunoblotting using a rabbit antiserum which had been raised against the purified protein (see "Experimental Procedures"). Under the immunoblotting conditions used, the antiserum did not cross-react with STI1, and hence positive clones were readily selected. One of these, designated pSTI2-5642, was further characterized. We point out that the expression cloning strategy allowed us to not only obtain the gene for the protein but also demonstrated that the cloned DNA fragment contained all of the information necessary to direct expression and secretion of the protein in a heterologous Streptomyces species. We have designated this protein ST12 and its gene sti2.
A restriction endonuclease map for the region containing sti2 and an outline of the sequencing strategy are shown in Fig. 2B. The nucleotide and amino acid sequences of ST12 are presented in Fig. 38. As with STII, NHz-terminal sequence analysis of ST12 showed some heterogeneity. Sequences were obtained consistent with starts at Ala' (SO%, nucleotide +103), Ser' (30%, nucleotide +106), and Leu3 (lo%, nucleotide +109). Again, we have assigned the longest sequence as the mature NHz terminus. Upstream of the predicted NH2 terminus of mature ST12 is a 34-amino acid sequence consistent with a signal peptide. As with stil, two in-frame ATGs are encoded within the sti2 sequence, and based on its spacing relative to the putative ribosome binding site at -13 (underlined, Fig. 3B), we propose the ATG at position +1 as the translation initiation site.
The 5'-end of the sti2 transcript was determined by S1 nuclease mapping using a 421-bp BssHII-EcoRI DNA fragment (see "Experimental Procedures") and RNA purified from S. longkporus grown for 72 h in culture. A unique 5'end was identified at position -56 (Fig. 3 8 ) .
Identity of STZl and STZ2"The sequences of both ST11 and ST12 were used to search the National Biomedical Research Foundation (NBRF) protein sequence database for other homologous proteins. Two proteins were found to have a high degree of homology, the Streptomyces subtilisin inhibitor (SSI) from S. albogriseolus (Refs. 14,15,33, and see also Ref. 12) and a potent inhibitor of plasmin, plasmino$reptir (PSN), isolated from S. antifibrinolyticus (16,17). Fig. 4 shows the alignment of ST11 and ST12 with these protein protease inihibitors as well as a third inhibitor, alkaline protease inhibitor (API-2c) from S. griseoincarnatus (34,35). Protein protease inhibitors have been classified based on their size, structure, and disulfide bond arrangement (36). The SSI-like class of inhibitors have two nonoverlapping disulfide bonds, have monomeric molecular masses of -10 kDa, and form dimers containing two protease binding sites. STIl and ST12 exhibit 66% identity with SSI and 73.4% identity with each other and with PSN, suggesting that they belong to this class. Interestingly, the site of greatest divergence among the different inhibitors lies in residues 60-73 (ST11 numbering, Fig. 4 FIG. 5. Inhibition of trypsin activity by STIl and STIB. The trypsin assay was performed as outlined under "Experimental Procedures." A, the enzyme concentration was 0.5 p~ and the substrate concentration was 2 mM. PSN was purified from S. antifibrinolyticus ATCC 21870 culture supernatants essentially as described (16). The concentration of inhibitor is calculated based on the monomer molecular weight for STI1, STI2, and PSN. PSN has been shown to be a dimer while STIl and ST12 are presumed to be dimers; however, there are two enzyme binding sites/dimer for PSN thus the concentration of monomer equals the concentration of active sites. 0, PSN V, STIP; V, BPTI; 0, aprotinin. B, for this experiment, the enzyme concentration was 0.2 p~, and the substrate concentration was 1 mM. 0, aprotinin; V, PSN; V, STI1.
residues between positions 60 and 73, inclusive, only three amino acids (Vala, Cys6', and are absolutely conserved among the five proteins. Based on the three-dimensional structures of SSI (37,38), PSN (39), and STIs 1 and 2,4 the residues corresponding to 60-73 are part of a loop extending above and away from the packed @-sheet structure of the rest of the molecule. The loop is restrained by a disulfide bond between CytP7 and Cysg7 and terminates with which starts @-strand 4. In the SSI-subtilisin complex, this loop contains most of the residues in contact with the proteolytic enzyme (40). In addition to these structural data, mutation of the P1 residue in SSI has been shown to change the specificity of the inhibitor (41), thus demonstrating the major influence the PI residue has in determining specificity.
The hypervariability seen in the enzyme contact region observed with these inhibitors is similar to that observed by Laskowski and colleagues (42,43) in their exhaustive study of the ovomucoid third domain. As with ovomucoid, residues contributing to the basic structure of the inhibitors are highly conserved (e.g. the 4 Cys, P r~~~-V a l~~, etc.), whereas those that influence the function of the molecule vary widely. In the Streptomyces inhibitors, it seems likely that the sequence variation within this region results from evolutionary "tailoring" of the inhibitors for the best fit with their cognate proteases. We do not know the natural protease targets of STIl and STIS; however, as in PSN, these proteins have a basic amino acid (Arg in STIl, Lys in STI2) in the P1 position.
Thus, we would expect them to be inhibitors of trypsin-like proteases. To test this, we compared the ability of STI1, STI2, BPTI, aprotinin, and PSN to inhibit trypsin (Fig. 5) 6. Analysis of the products of induction of the stil and sti2 genes transformed into E. coli. The genes for ST11 and ST12 were transformed into E. coli under control of the PL promoter. Following heat induction, the cells were divided into cellular (C) and periplasmic (P) fractions as described under "Experimental Procedures." These fractions were subjected to SDS-PAGE, the gels were blotted onto nitrocellulose, and the ST1 proteins labeled with specific antisera. Lane 1 contains authentic ST11 from S. lividam. Lane 2 is the cellular and lane 3 the periplasmic fractions from E. coli expressing the stil gene. Lane 5 is the cellular and lane 6 the periplasmic fractions from E. coli expressing the sti2 gene. Lane 4 contains molecular mass markers.
NH, terminus of ST12 produced by S. liuidans is consistent with the Ser NH, terminus which we identified for the S. liuidans homolog STIl (see Fig. 3A). An additional 18% of the ST12 produced by S. liuidans began with Leu3 (nucleotide +109) analogous to the native protein. Surprisingly, a small but reproducible proportion (-5% of the protein) had three additional amino acids (Ala-Ala-Pro-Ala'-Ser'-) at the NH1 terminus. This result suggested that the NH, termini defined by the proteins we isolated after extended culture (72 h) might be the result of extracellular degradation by aminopeptidases subsequent to the initial signal peptidase cleavage event upstream of the putative +1 position. To examine this possibility, we analyzed samples of S. longisporus conditioned medium after various times in culture by SDS-PAGE. At early times (24 h), we observed a major species of ST12 migrating with an apparent molecular weight -1,000 higher than that observed for ST12 a t later times. After 48 h, this 11-kDa species and the normal 10-kDa species were present in approximately equal amounts. By 72 h, only the smaller 10-kDa species was present (data not shown). We isolated the higher molecular mass species from a 24-h culture and determined its NH2 terminal sequence. This protein had an additional 6 residues, Thr-Pro-Ala-Ala-Ala-Pro-Ala'-Se3-Leu3-Tyr4-(starting at nucleotide +85, Fig. 3B). We propose that this species is a precursor of the shorter molecules and is the initial product of the signal peptidase. The NH, termini found at later times result from the activity of aminopeptidase(s) present in the medium. We cannot formally rule out, however, the possibility that the other NH2 termini are generated by imprecise processing by the signal peptidase.
Finally, in order to test whether the Streptomyces secretion signals would function in a heterologous bacterial system, the stil and sti2 genes, including their putative ribosome binding sites and transcription initiation start sites, were subcloned downstream of the Plae promoter in the E. coli expression vector pUC18. Induction of the stil-containing plasmid in E. coli resulted in the synthesis of STIl as demonstrated by Western blot analysis (Fig. 6). Furthermore, the size of the STIl was comparable in SDS-PAGE mobility to the authentic 10-kDa STIl found in culture supernatants of S. liuidans. Similarly, induction of the sti2 containing plasmid in E. coli resulted in the production of ST12 which comigrated on SDS-PAGE with authentic ST12 isolated from S. longisporus cul-