Influence of Specific Signal Peptide Mutations on the Expression and Secretion of the (cid:97) -Amylase Inhibitor Tendamistat in Streptomyces lividans

The Streptomyces (cid:97) -amylase inhibitor tendamistat is secreted by a signal peptide with an amino-terminal charge of (cid:49) 3. To elucidate the influence of the charged residues on protein secretion in Streptomyces , the ami- no-terminal charge was varied from (cid:49) 6 to neutral net charge. The effects of charge variation were analyzed in combination with three Streptomyces promoters and two transcriptional terminators. Introduction of addi- tional positive charges significantly decreased the amount of secreted tendamistat. On the contrary, a charge reduction to (cid:49) 2 resulted in the doubling of inhibitor production. After exclusion of transcriptional effects, the observed alterations of inhibitor secretion by the mutants with a charge of (cid:49) 6 to (cid:49) 2 were attributed to a modulation of precursor synthesis. Further- more, a tight coupling of synthesis and export was stated. Charge reduction to (cid:49) 1 or neutral charge gener-ally reduced the yield of secreted tendamistat, yet re- markable differences were found for mutants with identical net charge. Elimination of the positive charge at a defined position resulted in the release of tendamistat precursor protein, which suggested a specific uncou- pling of synthesis and translocation. The majority of secreted proteins are synthesized in the cytoplasm in a precursor form with an N-terminal extension termed the signal peptide. Although little homology in the overall amino acid sequence is found, signal peptides show a common tripartite structure irrespective of their origin. A cen-tral core of hydrophobic residues is flanked by an amino-ter-minal region, also termed the N-domain,

The Streptomyces ␣-amylase inhibitor tendamistat is secreted by a signal peptide with an amino-terminal charge of ؉3. To elucidate the influence of the charged residues on protein secretion in Streptomyces, the amino-terminal charge was varied from ؉6 to neutral net charge. The effects of charge variation were analyzed in combination with three Streptomyces promoters and two transcriptional terminators. Introduction of additional positive charges significantly decreased the amount of secreted tendamistat. On the contrary, a charge reduction to ؉2 resulted in the doubling of inhibitor production. After exclusion of transcriptional effects, the observed alterations of inhibitor secretion by the mutants with a charge of ؉6 to ؉2 were attributed to a modulation of precursor synthesis. Furthermore, a tight coupling of synthesis and export was stated. Charge reduction to ؉1 or neutral charge generally reduced the yield of secreted tendamistat, yet remarkable differences were found for mutants with identical net charge. Elimination of the positive charge at a defined position resulted in the release of tendamistat precursor protein, which suggested a specific uncoupling of synthesis and translocation.
The majority of secreted proteins are synthesized in the cytoplasm in a precursor form with an N-terminal extension termed the signal peptide. Although little homology in the overall amino acid sequence is found, signal peptides show a common tripartite structure irrespective of their origin. A central core of hydrophobic residues is flanked by an amino-terminal region, also termed the N-domain, 1 which is characterized by its content of basic residues, and a C-terminal part with an appropriate recognition site for a signal peptide peptidase. Excellent reviews on signal peptide structure and its function in bacterial protein secretion have been submitted by several authors (1)(2)(3)(4)(5).
The impact of charge variation in the N-domain was studied essentially in signal peptides of Escherichia coli secretory proteins; the results are reviewed by Gennity et al. (3). Concurring experimental data (6 -8) indicate that signal peptides with neutral or even negative N-terminal charge are still functionable, but reduced in their ability to promote the translocation of a given protein. For maximal efficiency of translocation, a positive net charge is necessary, although this dependence may be influenced by length and overall hydrophobicity of the core region (9). From the observation that signal peptides with reduced but still positive net charge function as efficiently as the wild-type leader, it was concluded that multiple charges represent a functional redundancy (10). One assignment of the charged region is the definition of the signal peptide orientation during insertion into the membrane in accordance with the "positive-inside" rule proposed by von Heijne (11), which has been amply documented during the last years (12)(13)(14)(15), and the "loop model" of secretion (16,17). On the other hand, involvement of the charged domain in the interaction of the signal peptide with components of the secretory machinery like the E. coli SecA protein (18) is proposed.
In contrast to the variety of information on signal peptide mutations in E. coli, few publications are available concerning Gram-positive bacteria with emphasis on members of the Bacillus genus (reviewed in Ref. 19). A systematic study of the influence of signal peptide variations on the secretion of Bacillus amyloliquefaciens levansucrase in Bacillus subtilis (20,21) is in general agreement with the results obtained for E. coli. However, the dependence on positive charges in the N-terminal region was more pronounced; signal peptides with neutral net charge significantly impaired the export of the precusor protein.
Comparison of signal peptide sequences from different organisms indicated that variations in length and amino acid composition of the three parts are species-specific (22). N-domains from Gram-positive bacteria are longer and higher charged than the respective parts of signal peptides from Gram-negative bacteria, human or plant origin. Streptomycete signal peptides occupy a special position because of their extraordinary long and highly charged N-domains as well as the strong preference of arginine for lysine in this segment.
The obvious capacity of the Streptomyces secretory machinery, demonstrated by the variety of proteins that are secreted at high levels into the culture medium, raised growing interest among bioscientists and biotechnologists (23)(24)(25). A valuable model of Streptomyces secretory protein is tendamistat, a polypeptide of 74 amino acids that specifically and almost irreversibly inhibits ␣-amylases of the mammalian type (26). The tendamistat gene was cloned from an amplified genomic sequence of the original producer Streptomyces tendae 4158 and was successfully expressed in Streptomyces lividans 66 (27).
Here we describe the influence of N-terminal charge modifications of the tendamistat signal peptide on tendamistat secretion in combination with the original tendamistat promoter and terminator as well as with two other Streptomyces promoters, the mel promoter of Streptomyces antibioticus and the ermEup promoter of Saccharopolyspora erythrea, and an additional terminator derived from the aph gene of Streptomyces fradiae. Charge variations have a significant effect on tendamistat expression and secretion in S. lividans, thus indicating that the N-terminal region of a signal peptide plays an important role in the synthesis of the precursor. Our results suggest that secretion of tendamistat occurs essentially cotranslationally. The coupling of translation and translocation, however, can be partially relieved by specific signal peptide mutations. Moreover, a hypothesis is developed in order to understand the modulation of precursor synthesis due to signal peptide variation.

MATERIALS AND METHODS
Bacterial Strains-For site-specific mutagenesis, E. coli strains BMH71-18 and MK30-3 as well as the M13 phagemids M13mp18 and M13mp18rev (28) were used. Further cloning was done in E. coli XL1-Blue (Stratagene, La Jolla, CA). The host for expression plasmids with the tendamistat and the mel promoter was S. lividans 66 TK24, which was kindly provided by K.-P. Koller (Hoechst AG, Frankfurt, Germany). Shuttle vector plasmids were expressed in S. lividans 66 TK23, contributed by the group of W. Piepersberg (Wuppertal, Germany). TK23 equals TK24 with respect to expression of tendamistat.
Plasmid Construction-Oligodeoxyribonucleotides were synthesized on an Applied Biosystems Model 380B DNA synthesizer and purified by reversed-phase chromatography. Restriction endonucleases and DNAmodifying enzymes were from New England Biolabs Inc. (Beverly, MA). DNA sequencing according to Sanger et al. (68) was carried out with the T7 sequencing kit (Pharmacia Biotech Inc.).
Shuttle Vector Constructions with the ermEup Promoter and the aph Terminator-The basic E. coli-S. lividans shuttle vector pAZ5, which in general represents a fusion of the E. coli vector pUC19 and the S. lividans vector pIJ702, was a generous gift of D. Pocta. The plasmid carries the ColE1 replicon and the ampicillin resistance gene for propagation and selection in E. coli as well as the pIJ101 replicon and the thiostrepton resistance gene for cloning in S. lividans. pAX5a (8067 base pairs), a derivative of pAZ5 with the tendamistat gene (wild-type signal sequence) under the control of the ermEup promoter of S. erythrea and with the aph transcriptional terminator from S. fradiae downstream from the coding region, was used in all cloning experiments. The above-described signal peptide variants R5R9, R5, A2, and Xtend were cloned into pAX5a as SpeI/EcoRI fragments from pUC19 derivatives. SpeI/EcoRI fragments with the promoterless and terminatorless tendamistat gene and the signal sequence mutations A2 and R5R9 were also cloned into the E. coli vector pBluescript II KS ϩ (Stratagene), resulting in plasmid pIVT2 in the case of the A2 mutant and plasmid pIVT4 in the case of the R5R9 mutant. The A2 mutation generates a singular NcoI site; the R9 mutation introduces a singular BssHII site. The new mutant signal sequences were introduced by synthetic oligodeoxynucleotide linkers covering the sequence from the NcoI to BssHII sites. The following linkers were ligated with the NcoI/EcoRI vector fragment of pIVT2 and with the BssHII/EcoRI insert fragment of pIVT4: A2A4, 5Ј-CATGGCCGTAGCCGCACTTCGACTTGC-3Ј ϩ 5Ј-CGCGGCAAGTCGAAGTGCGGATACGGC-3Ј; A2A7, 5Ј-CATGGCCG-TACGGGCACTTGGCTGC-3Ј ϩ 5Ј-CGCGGCAAGCGCAAGTGCCCGT-ACGGC-3Ј; A2A4A7, 5Ј-CATGGCGTGCGCACTTGCGCTTGC-3Ј ϩ 5Ј-CGCGGCAAGCGCAAGTGCGGCTACGGC-3Ј; A2D4, 5Ј-CATGCGTA-CGGGCACTTGACCTTGC-3Ј ϩ 5Ј-CGCGGCAAGGTCAAGTGCCCGT-ACGGC-3Ј; and A2D7, 5Ј-CTAGGCCGTAGACGCACTTCGACTTGC-3Ј ϩ 5Ј-CGCGGCAAGTCGAAGTGCGTCTCGC-3Ј. The SpeI/EcoRI fragments from pBluescript derivatives with verified signal sequence mutations were ligated into pAX5a to obtain the respective expression plasmids for S. lividans.
DNA Manipulation and Expression Cultures of Streptomyces Strains-Cultivation of S. lividans TK23 and TK24 and preparation of protoplasts and transformation were performed according to Hopwood et al. (31). Expression cultures were inoculated with 10 4 spores/ml and grown in 50 ml of modified minimal medium (5% glucose, 2% casamino acids, 0.2% (NH 4 ) 2 SO 4 , 15 mM phosphate buffer, pH 6.9, 3 mM MgSO 4 , and trace elements) supplemented with 12.5 g/ml thiostrepton (except for TK23 and TK24) at 28°C and 300 rpm for 9 days (10 days for strains harboring a shuttle expression vector). The supernatant was prepared by centrifugation.
Assays for Inhibitor Production-Selection of inhibitor-secreting transformants was carried out on agar plates as described (32). Secreted inhibitor in liquid cultures was determined by the modified dinitrosalicylic acid assay (33) and by a modified ␣-amylase assay purchased from WAK Chemie GmbH (Bad Soden, Germany). Reference ␣-amylase from porcine pancreas was purchased from Boehringer Mannheim (Mannheim, Germany). Purified tendamistat for calibration was a generous gift of Hoechst AG.

SDS-PAGE of Proteins and Western
Blotting-For SDS-PAGE according to Schä gger and von Jagow (34), gels with 16.5% T, 6% C in the separating gel, 10% T, 3% C in the spacer, and 4% T, 3% C in the stacking gel were cast. Samples were resuspended in sample buffer (4 M urea, 0.1 M Tris, 0.1 M 1,4-dithio-DL-threitol, 1% SDS, 0.05% bromphenol blue); 2 l of ␤-mercaptoethanol were added, and the sample was boiled for 2 min prior to loading. Extended boiling is not recommended, for it will cause the appearance of partially or totally oxidized tendamistat. Gels were either stained with Coomassie Brilliant Blue R-250 according to Ref. 34 or electroblotted on Immobilon P membranes (Millipore Corp., Bedford, MA) following the protocol of Plough et al. (35). The molecular mass marker was visualized by Coomassie staining of the respective part of the membrane. Tendamistat-specific polyclonal antiserum preparations developed in goat and an anti-goat IgG-alkaline phosphatase conjugate were used for immunodetection of tendamistat. The tendamistat antisera were generous a gift of K.-P. Koller. The coloring reaction was carried out with 5-bromo-4-chloro-3-indolyl phosphate substrate and p-nitro blue tetrazolium according to standard procedures.
Cell Disruption and Western Blotting-Cells from 1 ml of culture were washed three times in phosphate-buffered saline and resuspended in 200 l of ice-cold phosphate-buffered saline before they were lysed by sonification under constant ice cooling. After pelleting of the cell debris by centrifugation, the supernatant comprising the soluble protein fraction was obtained. Determination of protein concentration was performed according to Lowry et al. (69). Lyophilized aliquots of the soluble protein fraction were resuspended in sample buffer and subjected to electrophoresis and Western blotting as described above.
Preparation of Total RNA from S. lividans-Isolations were performed from fresh cells on days 3-5 of a standard shaking flask culture. Pelleted cells were washed once in phosphate-buffered saline and resuspended in 200 l of isolation buffer (160 mM Tris-HCl, pH 7.6, 20 mM MgCl 2 , 20 mM ␤-mercaptoethanol) before the addition of 200 l of 2 ϫ SEPH buffer (0.5% SDS, 10 mM EDTA, 2 mM 1,10-phenanthroline, 0.2 mg/ml heparin, pH 6.0, supplemented with 2 mg/ml proteinase K prior to use), and the suspension was incubated for 20 min at 37°C. Disruption of cells during incubation was promoted by shearing through an injection syringe. After repeated phenol extractions, the nucleic acids were precipitated with ethanol/sodium acetate and lyophilized. DNA was removed from the sample by digestion with 15 units of RNase-free DNase I (Boehringer Mannheim) in 50 l of DNase buffer (50 mM sodium acetate, 25 mM MgCl 2 , 4 mM CaCl 2 , pH 6.0) within 20 min at 37°C. 0.25 units of human placental RNase inhibitor (Stratagene) may be added to the reaction mixture. Following phenol/chloroform extraction, the RNA was precipitated with ethanol/sodium acetate from the aqueous phase. The dried preparation was resuspended in distilled water, and concentration was determined via absorbance at 260 nm. Additionally, the preparations were analyzed by electrophoresis in MOPS (50 mM MOPS, 10 mM sodium acetate, 1 mM EDTA, pH 7.0)buffered agarose gels containing 1.8% deionized Formalin and 0.5 g/ml ethidium bromide. 5-l samples of RNA solution were mixed with 13 l of denaturation buffer (mixture of 500 l of 5 ϫ MOPS buffer, 500 l of deionized formamide, and 175 l of deionized Formalin), incubated for 15 min at 55°C, and supplemented with 2 l of 10 ϫ loading buffer (50% glycerin, 10 mM sodium dihydrogen phosphate, pH 7.0, 0.4% bromphenol blue) prior to loading.
In Vitro Transcription, Northern Blotting, and Northern Slot Blotting-An SpeI/EcoRI fragment with the tendamistat gene was cloned into pBluescript II KS ϩ . Plasmid DNA purified by a cesium chloride gradient was restricted with HindIII and freed of contaminating RNases by digestion with proteinase K to produce template DNA. Purified template DNA was transcribed using the in vitro transcription kit purchased from Stratagene. The addition of any RNase inhibitors proved to be unnecessary. Template DNA was removed by digestion with RNase-free DNase I prior to phenol extraction and ethanol precipitation. Transcript RNA was resuspended in distilled water.
For Northern blots, RNA samples separated on agarose gels were transferred to nylon membranes (Hybond N, Amersham International, Buckinghamshire, United Kingdom) by standard procedures. Hybridization was performed according to the protocol of Church and Gilbert (36) at 60°C with an ␣-32 P-labeled 52-mer oligodeoxynucleotide (5Ј-GCACAGCCGTTGTCGGCCTGTGAGTACCGCCAGCTCTGGTA-GAGCGTCACGC-3Ј) complementary to the tendamistat coding strand. Membranes were washed at least three times with 0.1 ϫ SSC, 1% SDS at 60°C prior to exposure to Kodak XAR films. For Northern slot blots of RNA to Nytran-N nylon membranes (Schleicher & Schü ll, Dassel, Germany), the Minifold II filtration unit (Schleicher & Schü ll) was used. The transfer was carried out according to the manufacturer; hybridization was as described above.

Tendamistat Expression in pIJ702
Derivatives with the Tendamistat or the mel Promoter and with the Tendamistat or the aph Terminator-On the basis of the tendamistat gene, a cassette for transport expression in Streptomyces was developed in our group (Fig. 1). The construction of the signal peptide variants and the Streptomyces expression plasmids is described under "Materials and Methods"; the amino acid sequences of wild-type and mutant tendamistat signal peptides are shown in Fig. 2. The chosen codons for arginine (5Ј-CGC-3Ј), alanine (5Ј-GCC-3Ј and 5Ј-GCG-3Ј), and aspartate (5Ј-GAC-3Ј) substitutions obey the preferences of the Streptomyces codon usage (37).
First, mutant signal peptides with an N-terminal charge ranging from ϩ2 to ϩ6 (Fig. 2) were tested for tendamistat expression in combination with the native tendamistat promoter and terminator as well as with another Streptomyces promoter (mel) and terminator (aph). The tendamistat promoter is located on a 194-base pair HindIII/SpeI fragment of the expression vector. Downstream from the coding region, a sequence was identified that represents a preferred transcriptional termination site as proven by Northern blots (data not shown) and is therefore termed the tendamistat terminator. This terminator is not very efficient because a remarkable portion of extended transcripts are found.
According to a report of Pulido and Jiménez (38), an increase of secretory protein expression in Streptomyces is possible by the introduction of the strong transcriptional terminator derived from the neomycin resistance (aph) gene from S. fradiae. Introduction of the aph terminator resulted in the occurrence of uniform tendamistat transcripts with an approximate length of 520 bases on Northern blots (data not shown).
The promoter of the melC gene of S. antibioticus is present on the well known Streptomyces plasmid pIJ702 (39,40). The construction of an expression vector with the tendamistat gene under the control of the mel promoter resulted in enhanced expression of the inhibitor as described by Schmitt-John and Engels (30). Transcripts of the tendamistat gene that initiate at the mel promoter and terminate at the aph terminator have an approximate length of 550 bases (data not shown).
Monitoring tendamistat secretion by activity assays yields a sigmoid production curve (Fig. 3A); saturation of inhibitor concentration is obtained on late days of culture. Variation of the signal peptide N-terminal charge from ϩ6 to ϩ2, however, severely affects the mean saturation concentration of inhibitor as demonstrated in Fig. 3 and Table I. The R5R9 signal peptide with an N-terminal charge of ϩ5 reduces the amount of secreted inhibitor to ϳ25%. The R5 signal peptide with a charge of ϩ4 results in a decrease to ϳ65% of the wild-type level. On the contrary, the yield of active secreted tendamistat is doubled by the use of the A2 signal peptide with an N-terminal charge reduced to ϩ2.
Substitution of the tendamistat terminator for the aph terminator has no significant influence on inhibitor secretion as shown in Fig. 3B and Table I. On the other hand, the combination of the mel promoter and the described signal peptide variants causes an increase of tendamistat secretion by a factor of ϳ4 (Fig. 3B and Table I). However, the effect of the signal peptide variation is not significantly influenced by promoter substitution.
The results of the activity assays were confirmed by SDS-PAGE and Western blotting of supernatant samples. Secreted tendamistat is totally active and migrates as a homogeneous protein band with an apparent molecular mass of 9 kDa on denaturing SDS gels (26).
The only mutant signal peptide apparently unable to promote the secretion of tendamistat in each construction (Table I) is the chimeric Xtend signal peptide, which was built by the fusion of the N-terminal domain of the XP55 signal peptide of S. lividans (Fig. 2) (41) with the hydrophobic and C-terminal part of the tendamistat leader mutant R5R9 at the DNA level as described under "Materials and Methods." In the case of this ϩ6 leader, the absence of tendamistat in culture supernatants was confirmed by SDS-PAGE and Western blotting. Furthermore, no inhibitor was found in total protein from cell disrup-tions. Hence, the introduction of the Xtend leader totally abolished the expression of tendamistat.
To test the assumption that signal peptide variations do not interfere with transcription, we attempted to quantify the amount of tendamistat mRNA in all strains. RNA isolation from many culture samples was achieved by development of a minipreparation method (see "Materials and Methods") that provides total RNA in a quality suitable for qualitative and quantitative analysis of mRNA. Successful RNA isolations were possible only from logarithmic growing Streptomyces cultures on days 3-5, with a distinct optimum in yield and quality of the prepared RNA. The results of the Northern slot blot experiments are shown in Fig. 4. Signals obtained after stringent hybridization and washing conditions were solely caused by hybridization of a tendamistat-specific oligonucleotide with tendamistat mRNA.
At the level of transcription, no influence of the signal sequence variations is observed. The amount of tendamistat mRNA is only dependent on the given promoter. Moreover, the amount of tendamistat mRNA is not affected in the case of the nonproducing Xtend mutant. The level of tendamistat mRNA was estimated between 0.2 and 0.4% of total mRNA for constructions with the tendamistat promoter. Substitution for the mel promoter raises this portion ϳ5-fold, which is coincident with the observed increase of secreted inhibitor (Table I); the aph terminator has no influence on the amount of mRNA. The results of the Northern slot blot experiments definitely rule out that the signal peptide variations have an effect on the transcription of the mutant tendamistat genes.
Expression of the Signal Peptide Variants with Charges of ϩ6 to 0 in an E. coli-S. lividans Shuttle Vector with the ermEup Promoter Region of S. erythrea-The cloning of the entire set of signal sequence variants (Fig. 2) comprising a charge variation from ϩ6 to 0 into the novel pAX shuttle expression vector is described under "Materials and Methods." The shuttle plasmid was found to be as stable as conventional pIJ702 vectors.
The ermE promoter region of the erythromycin resistance gene of S. erythrea (formerly Streptomyces erythreus) represents one of the strongest known actinomycetes promoters and includes two transcriptional start sites, P1 and P2 (42). By deletion of 3 base pairs in the Ϫ35 region of the P1 promoter, a mutant with enhanced transcriptional activity was obtained, termed the ermEup promoter. 2 Expression of tendamistat by the aid of the ermEup promoter raises the amount of secreted inhibitor by a factor of 10 in comparison with the native tendamistat promoter according to activity assays (Table I) and SDS-PAGE.
Further reduction of the N-terminal charge below ϩ2 (Fig. 2) again reduces the yield of secreted inhibitor in comparison with the wild type (Fig. 5). However, remarkable differences are observed for variants with identical net charge. Substitution of the arginine residues at positions 2 and 4 with alanine (mutant A2A4, charge ϩ1) lowers the inhibitor yield to 54% of the wild type, whereas exchanges at positions 2 and 7 (mutant A2A7, charge ϩ1) have a stronger effect, reaching only 25% of the wild-type level. The same bias is observed for the neutrally charged mutants A2D4 and A2D7, which secrete 24 and 9% in comparison with the wild type, respectively. Thus, the yield of secreted tendamistat is significantly more impaired by substitution of the arginine residue at position 7 than by a similar substitution at position 4. In addition, the introduction of a negatively charged aspartate residue more severely reduces the inhibitor expression than the introduction of a neutral alanine. The lowest value with only 2% of the wild-type yield is 2 M. J. Bibb, personal communication.

FIG. 3. Secreted tendamistat in the culture supernatant of S. lividans strains transformed with pIJ702 derivatives.
A, comparison of typical production curves determined by the activity assay. Tendamistat was expressed by its native promoter and terminator; the N-terminal charge of the given signal peptide (SP) is indicated. The exponential increase of secreted inhibitor until the 5th day is correlated with an increase of cell mass. After the 6th day, the cell mass is reduced by autolytic processes within the mycelium. Because the extremely stable inhibitor is almost not proteolytically degraded, tendamistat reaches a saturation concentration. B, influence of promoter and terminator substitutions on tendamistat secretion. The columns represent the mean saturation concentration of three independent cultures per strain. The exact values are given in Table I. Tend/Tend, tendamistat promoter/tendamistat terminator; Tend/aph, tendamistat promoter/ aph terminator; mel/aph, mel promoter/aph terminator. found in the case of the variant with an elimination of all charged residues (A2A4A7, charge 0). The results were confirmed by SDS-PAGE of the supernatants (Fig. 6) and Western blotting.
Analysis of Intracellular Tendamistat by Western Blotting and by Activity Assay-The investigation of tendamistat expression was completed by immunological detection of tendamistat in the soluble protein fraction of cell homogenates. The results for strains transformed with a shuttle vector are shown in Fig. 7. The major immunoreactive protein has an apparent molecular mass of 9 kDa and is therefore identical to the mature inhibitor from supernatants as shown by comparison with the controls of purified secreted tendamistat. Two bands attributed to degradation of the mature form appear below. These observations were also made for the constructions with the tendamistat and the mel promoter.
To rule out that the detected mature tendamistat is just a remnant of the secreted inhibitor from the supernatant, the host strain TK23 was grown in the presence of 200 mg/liter purified tendamistat, and samples were subjected to Western blot analysis. In fact, a small amount of mature inhibitor was present in the soluble protein fraction of the control (Fig. 7,  lanes 25 and 26). But comparison with the fraction found in the case of the R5 variant (lanes 9 and 10), which secretes ϳ200 mg/liter inhibitor, reveals that the remnant from the supernatant represents Ͻ5% of the processed tendamistat detectable in a cell disruption sample from a producing strain. In addition, the proportionality of cell-associated mature inhibitor to secreted inhibitor is striking if Figs. 6 and 7 are compared.
The soluble protein fractions from disrupted cells were also subjected to the activity assays. An ␣-amylase inhibiting activity, which has to be attributed to correctly folded tendamistat, was detected in samples from each producing strain, except the A2A4A7 variant, which is characterized by an extremely low inhibitor secretion, too. In accordance with the course of inhibitor secretion (Fig. 3A), the cell-associated inhibitor activity increased with time and reached saturation after ϳ1 week. A comparison of the saturation concentrations is given in Fig. 8.
The relative values clearly resemble the saturation concentrations of secreted inhibitor compared in Fig. 5. Hence, we suggest that the active cell-associated inhibitor represents a measure for the amount of translocated tendamistat.
If the cell-associated inhibitor in fact resides in the cytoplasm or if it is attached to the cell wall or the membrane has not yet been proven without doubt. The appearance of cellassociated mature inhibitor is probably a tendamistat-specific phenomenon due to extremely fast folding characteristics be-cause tendamistat adopts its active conformation in ϳ20 ms. 3 In accordance with the results for the constructions with tendamistat and the mel promoter, no tendamistat precursor protein (104 amino acids) is observed intracellularly in the case of the variants with the wild-type, R5R9, R5, or A2 signal peptide expressed by the ermEup promoter (Fig. 7). This may indicate that tendamistat synthesis and export are normally tightly coupled. On the contrary, in the soluble protein fraction of cell disruptions from strains with a ϩ1 or neutrally charged signal peptide, precursor protein with an apparent molecular mass of 11.5-12 kDa is observed. The tendamistat precursor is subject to accelerated degradation as indicated by multiple degradation products below its major immunoreactive band. The appearance of the precursor protein provides evidence that the assumed coupling of synthesis and translocation can be impaired by specific signal peptide mutations.
The A2A4 signal peptide reduces tendamistat secretion to 54% of the wild-type level. However, a remarkable portion of precursor protein is found in the cytoplasm, thus indicating that the reason for the reduction of secreted inhibitor is an intracellular release of the precursor, which has no or negligible possibility to translocate post-translationally. Significantly less precursor protein is found in the case of the A2A7 variant, yet the amount of overall intracellular tendamistat protein is reduced. Therefore, the decrease of tendamistat secretion to 25% seems to be essentially due to a reduction of precursor synthesis.
An increased amount of free precursor protein in the case of a mutation at position 4 is also found for the signal peptide variants with a neutral net charge, if the results for mutants A2D4 and A2D7 are checked in Fig. 7. In addition, a comparison of the results for mutants A2A4 and A2D4 indicates that an acidic aspartate significantly raises the intracellular precursor fraction over a neutral alanine (Fig. 7). In accordance with the observations made for the ϩ1 variants, it is assumed that the reduction of inhibitor secretion to 24% in mutant A2D4 is in principle caused by the release of the precursor due to the amino acid substitution at position 4, whereas the major effect responsible for the decrease to 9% in mutant A2D7 is a reduction of synthesis connected with the exchange at position 7. In the case of variant A2A4A7, which combines the substitutions of the A2A4 and A2A7 variants, a combination of both effects (which means a reduction of precursor synthesis and a release of the already synthesized precursor) is obtained, which rea-

promoter, and terminator variation on tendamistat secretion in S. lividans
The mean saturation concentrations of active inhibitor calculated from supernatant samples from three independent cultures per strain are compared for each combination. The transcriptional terminator of the aph gene of S. fradiae had no influence on tendamistat secretion. Use of the mel promoter of S. antibioticus results in a 4-fold increase of inhibitor secretion; the ermEup promoter of S. erythrea increased inhibitor yield by a factor of 10 with respect to the native tendamistat promoter. sonably explains the pronounced reduction of inhibitor secreted to the supernatant (Ͻ2% of the wild type).

DISCUSSION
This study presents the first results obtained for the influence of a systematic charge variation in the amino-terminal region of a signal peptide on protein secretion in Streptomyces. The chosen model system was the secretory expression of the S. tendae ␣-amylase inhibitor tendamistat in S. lividans. Ten tendamistat signal peptide variants with an N-terminal charge ranging from ϩ6 to neutral net charge (Fig. 2) were examined. With the exception of the ϩ6 variant, all signal peptides were able to promote the transport expression of tendamistat (Figs. 3 and 5). This tolerance for charge variation is in general agreement with the results obtained for E. coli and B. subtilis (see the Introduction).
However, significant effects on the amount of secreted inhibitor were observed (Figs. 3 and 5 and Table I). An increase of the positive charge resulted in a gradual decrease of tendamistat secretion from 100% for wild type (ϩ3) to ϳ65% for mutant R5 (ϩ4), to ϳ25% for mutant R5R9 (ϩ5), and to a total inhibition of expression in the case of mutant Xtend (ϩ6). On the other hand, a significant increase of secreted inhibitor was obtained by charge reduction to ϩ2 in mutant A2. The presented signal peptide effects were found in combination with three different promoters, the tendamistat promoter, the mel promoter, and the ermEup promoter, as well as with two transcriptional terminators, the tendamistat and aph terminators ( Table I). Introduction of the strong transcriptional terminator of the aph gene of S. fradiae had no effect on the yield of secreted tendamistat (Fig. 3B and Table I) and on the amount of tendamistat mRNA (Fig. 4). A stabilization of tendamistat mRNA is therefore not supposed to be a key to optimization of inhibitor expression. Substitution of the tendamistat promoter for the mel promoter from the melC gene of S. antibioticus increased the amount of inhibitor by a factor of 4 ( Fig. 3B and Table I) as a consequence of an ϳ5-fold increase of tendamistat mRNA (Fig. 4). A 10-fold increase of inhibitor expression was found for constructions with the ermEup promoter of S. erythrea (Table I).
Concerning the biotechnological aspect of this study, the construction of the A2 signal peptide variant, which doubled the yield of secreted inhibitor, demonstrates that an optimal production of a given secretory protein may not be reached with its native signal peptide. Hence, signal peptide mutagenesis has to be considered as a suitable approach to the evaluation of secretory protein expression. The fact that only few comparable examples are published (for example, the 1.9-fold increase of the secretion of human lysozyme in yeast by engineering of the hydrophobic segment (43)) suggests, however, that the prospective success of signal peptide variation is limited. On the other hand, the effects of an optimized signal peptide and a strong promoter can be combined as shown by the expression of tendamistat with the A2 signal peptide and the mel or ermEup promoter. Moreover, the increase in yield of secreted protein attributed to the A2 signal peptide is not limited to the tendamistat model protein, but was also transferred to a quite different protein, the ␣-amylase of Streptomyces limosus (44), comprising 538 residues. 4 In addition to the ϩ6, ϩ5, ϩ4, ϩ3, and ϩ2 variants, two mutants with an N-terminal charge of ϩ1 and three mutants with a neutral net charge were constructed and expressed in S. lividans in combination with the ermEup promoter in a novel shuttle vector system. Charge reduction below ϩ2 again decreased the yield of secreted inhibitor (Fig. 5), indicating that a certain positive net charge is necessary for maximal efficiency of tendamistat secretion. However, net charge is not the crucial factor because the secretion of variants with identical net charge significantly differed. The results suggest that charged residues in distinct locations are of special importance for signal peptide function. In the case of tendamistat secretion, the need of a positively charged residue in the vicinity of the hydrophobic core is assumed.
Substitution of arginine at position 4 or 7 with an acidic residue more severely impaired the yield of secreted inhibitor than the introduction of a neutral amino acid. These results may hint at an involvement of the positively charged residues in ionic protein-protein interactions.
To understand how signal peptide variations give rise to such pronounced differences in inhibitor secretion, several efforts were made. The independence of the signal peptide effect from the promoter used (Table I) as well as the results of the quantitation of tendamistat mRNA (Fig. 4) strongly suggest that signal peptide mutations do not act on transcription. Con-cerning the mRNA structure of the signal sequence variants, no significant changes were found with a current program for secondary structure prediction (Fold, Version 7, Genetics Computer Group, Madison, WI). The mutations at position 7 of the signal peptide that resulted in a gradual decrease in the amount of expressed inhibitor from a basic arginine via a neutral alanine to an acidic aspartate residue also indicate that not the sequence of the mRNA, but the nature of the polypeptide is responsible for the observed effects.
A blocking of the general secretion pathway by export-deficient precursors was not taken into consideration because the pattern of exoproteins was not significantly altered in all variants (Fig. 6). That the export-deficient tendamistat precursors remain in a membrane-bound form is unlikely as well because no influence on viability and growth was observed for strains with reduced inhibitor secretion. Thus, we assume that the signal peptide variations already influence the synthesis of the mutant tendamistat precursor, which was investigated by analysis of tendamistat inside the cells.
The detection of cell-associated tendamistat revealed that mature inhibitor is present in the case of each producing strain (Fig. 7) and at least partially adopts a native conformation as proven by the determination of an ␣-amylase inhibiting activity (Fig. 8). The amount of cell-associated processed tendamistat is proportional to the amount of secreted inhibitor (Figs. 5 and 8) and thus represents a measure for the amount of translocated inhibitor.
The variants with wild-type and mutant R5R9 (ϩ5), R5 (ϩ4), and A2 (ϩ2) signal peptides lack any detectable free precursor protein (Fig. 7). Therefore, the differences in the amount of both cell-associated mature tendamistat and secreted inhibitor have to be attributed to a modulation of tendamistat synthesis. On the other hand, the absence of precursor protein suggests that synthesis and export of tendamistat are coupled in general.
However, in the case of the new mutants with a charge of ϩ1 or 0, the situation is different because the tendamistat precursor is observed (Fig. 7). Substitution of the arginine residue at position 4 resulted in accumulation of a significant fraction of precursor protein. This effect is strongly pronounced if an aspartate residue with an acidic side chain is introduced instead of a neutrally charged alanine. Thus, we suggest that the residue at position 4 interacts with a proteinaceous component of the secretory apparatus, which accounts for the coupling of translation and translocation of a secreted protein. The release of precursor protein seems to be essentially responsible for the reduction of inhibitor secretion by variants A2A4 and A2D4.
The intracellular appearance of the precursor was also obtained by substitution of the arginine residue at position 7, but to a significantly minor extent. Substitutions at position 7 are characterized by a reduction of synthesis, which is supposed to be the major reason for the decrease of inhibitor secretion observed for these mutants. Analogous to the variations at position 4, this reduction is more pronounced for an acidic amino acid than a neutral amino acid at position 7.
These results, as well as similar observations made for the influence of signal peptide charge variation on the synthesis of precursor protein in E. coli (6,7,(45)(46)(47), are hardly explainable by the post-translational export mechanism represented by the Sec-dependent or even the Sec-independent translocation pathway (reviewed in Ref. 4). However, concurring evidence (48 -52) strongly suggests the existence of a prokaryotic equivalent of the eukaryotic signal recognition particle (SRP)-dependent pathway determined for protein translocation into the endoplasmic reticulum lumen of mammalian cells (53)(54)(55). On the basis of an essentially cotranslational pathway involving the prokaryotic homologue of the SRP, a conception may be developed. The tendamistat precursor (104 amino acids) meets the basic requirement for a SRP-dependent translocation because it exceeds the lower size limit, which may be somewhat different with respect to the particular protein (56,57), but lies within the range of 64 (58,59) to 74 (60) amino acids. For the secretion of proteins into the lumen of the endoplasmic reticulum, it was shown that the translocation arrest caused by the interaction of the mammalian SRP with the signal peptide of a nascent protein is only released if the complex of ribosome, nascent polypeptide chain, and bound SRP interacts with the SRP receptor on the face of the endoplasmic reticulum membrane. This interaction and the following release of the SRP are governed by the binding and hydrolysis of GTP, which is influenced by the signal peptide (51). Recent publications (51,62) demonstrated that a similar GTPase cycle is found for the interaction of the E. coli homologue of the SRP54 subunit, the Ffh protein (63)(64)(65)(66), with the membrane-associated FtsY protein, which is the bacterial homologue of the ␣-subunit of the mammalian SRP receptor (63). The Ffh protein forms a ribonucleotide particle (RNP) with the 4.5 S RNA, which is the E. coli equivalent of the 7 S RNA participating in mammalian SRP (67). Miller et al. (51) have shown that the interaction of the RNP and the Ffh protein is modulated in vitro by synthetic signal peptides derived from the E. coli LamB leader. The GTPase activity of the RNP-FtsY protein complex was inhibited by signal peptide variants, which also promoted protein export in vivo. An export-deficient signal peptide, however, failed in preventing GTP hydrolysis. The authors suggested that the stimulation of GTP hydrolysis caused by binding of the RNP-signal peptide complex to the FtsY protein represents the key step that leads to the release of the signal peptide. Our hypothesis is that deficient signal peptides prevent binding of the RNP to the homologue of the E. coli FtsY protein by stabilization of a RNP conformation that tightly binds GTP or, more reasonably, retards dissociation of GDP. The latter model strongly resembles the GTPase cycle of the bacterial elongation factor Tu (61) as an example for GTP/GDP-binding proteins. Presuming that translation is also arrested by the prokaryotic FIG. 8. Comparison of cytoplasmic inhibitor activity. ␣-Amylase inhibiting activity in the soluble protein fractions from disrupted cells of strains transformed with the ermEup promoter-containing shuttle vector was determined by activity assay on days 5-8 of a selected culture. The calculated saturation concentrations, representing the portion of active inhibitor with respect to the total amount of protein in the fraction, are compared: Mutant R5R9 (ϩ5), 0.50 mg of inhibitor/g of total protein; mutant R5 (ϩ4), 1.20 mg/g; wild type (wt) (ϩ3), 2.57 mg/g; mutant A2 (ϩ2), 5.84 mg/g; mutant A2A4 (ϩ1), 1.63 mg/g; mutant A2A7 (ϩ1), 0.68 mg/g; mutant A2D4, 0.78 mg/g; and mutant A2D7, 0.19 mg/g. Inhibitor activity in the soluble protein fraction from samples of mutant A2A4A7 was below the detection limit.
SRP equivalent until interaction with the appropriate membrane receptor, the "inactivated" complex of a deficient signal peptide and the RNP with bound GDP may be released without completion of precursor synthesis. This hypothesis offers the possibility of explaining modulations of precursor synthesis by signal peptide mutations at the level of the polypeptide. According to this model, release of precursor protein into the cytosol is understandable by proposing an impairment of the interaction between the SRP homologue and the mutant signal peptide, which causes an uncoupling of translation and the SRP-dependent export pathway. Released precursors that are not able to translocate post-translationally will reside in the cytoplasm until they are proteolytically degraded.
Nevertheless, very little is known about the molecular mechanism of protein secretion in Streptomyces. Only cloning of the Streptomyces secretory apparatus proteins, which is recently in progress, 5 mutational analysis, and the in vitro reconstitution of the translocation pathway will give a satisfying answer to the questions evoked by the presented effects that signal peptide mutations have on the secretion of tendamistat in S. lividans.