INTRODUCTION

The terminal five reactions of the biosynthetic pathway yielding L-tryptophan are catalyzed by anthranilate-5-phosphoribosyl-1-pyrophosphate phosphoribosyltransferase (PRT¹; EC 2.4.2.18), phosphoribosylanthranilate isomerase (PRAI; EC 5.3.1.24), indole-3-glycerol-phosphate synthase (IGPS; EC 4.1.1.48), and tryptophan synthase (- and -chains, EC 4.2.1.20, see Scheme I).

In most procaryotes, tryptophan synthase is organized as a tetrameric bienzyme complex consisting of a central ₂-dimer (TRPS) combined with two peripheral -subunits (TRPS). The Escherichia coli and Salmonella typhimurium enzymes in particular have been intensively investigated in the past (1, 2). In fungi such as Saccharomyces cerevisiae and Neurospora crassa, tryptophan synthase is a dimer of identical bifunctional polypeptides (3, 4, 5). In contrast, the enzyme isolated from Euglena gracilis, here designated as multifunctional tryptophan-synthesizing enzyme (MTSE), is capable of catalyzing all reactions shown in Scheme I. Although this discovery was made 2 decades ago (6, 7), no thorough further characterization of this multifunctional enzyme has been published so far, probably due to the excessive expenditure to cultivate sufficient amounts of E. gracilis and the difficulties of purifying MTSE with outdated technology. Using state of the art equipment and materials, we established a reliable and simple procedure to obtain MTSE from E. gracilis (8). Here we present both a reinvestigation of earlier initial findings and new data with respect to structural and functional properties of the enzyme.

EXPERIMENTAL PROCEDURES

If not otherwise stated, all chemicals (reagent or ultrapure grade) were obtained from Sigma (Deisenhofen, Germany) or Merck (Darmstadt, Germany), and all enzymes were from Boehringer (Mannheim, Germany). Pyridoxal 5-phosphate was from Serva (Heidelberg, Germany), and phosphoribosyl pyrophosphate was from Sigma (Deisenhofen, Germany). Indoleglycerol 3-phosphate (IGP), phosphoribosylanthranilate (PRA), and o-carboxyphenylaminodeoxyribulose 5-phosphate (CDRP) were synthesized as described earlier (9-11). E. gracilis MTSE was prepared from algal biomass as described previously (8). Quartz-bidistilled water was used throughout. E. coli strains W3110 trp C 9830 (F) and W3110 trp D 9923 (D) were a generous gift from Dr. Charles Yanofsky.

PRT activity at substrate concentrations below 75 µM was determined fluorimetrically (12), whereas at higher substrate concentrations, a spectrophotometric assay at 278 nm was performed (11). Assays were performed both in the presence and in the absence of 0.1 mg/ml E. coli crude cell extract complementing consecutive "helper" activities but lacking PRT (W3110 trp D 9923 (D)).

PRAI activity was determined spectrophotometrically at 278 nm using a 0.1 mg/ml concentration of a PRAI-deficient E. coli helper cell extract (W3110 trp C 9830 (F)) (11).

To determine IGPS activity both a spectroscopic (13) and a fluorimetric assay system (14) were used.

The  activity of tryptophan synthase was measured using an NADH-coupled assay system (11), and the formation of L-tryptophan from L-serine and indole (-replacement, -reaction) was determined as described earlier (15).

The overall activity of MTSE was measured in 20 mM Tris/HCl, pH 7.8, containing 0.05 mM anthranilate, 1 mM phosphoribosyl pyrophosphate, 50 mM L-serine, 2 mM MgSO₄, 0.04 mM pyridoxal 5-phosphate, 0.01 mM dithioerythritol, 0.25 mM EDTA, 0.025 mg/ml BSA. The change in OD was monitored at 290 nm.

Deamination of L-serine (-elimination, -side reaction) was measured using a lactate dehydrogenase/NADH-coupled assay system as described previously (16).

Calculations to determine K_m and k_cat were carried out applying a computer program that iterates 20 times an algorithm described previously (17). All measurements were performed in 20 mM Tris/HCl, pH 7.8, at 37 °C, and the reactions were started by adding 0.04 µg of purified E. gracilis MTSE. Fluorimetric assays were carried out using an Aminco Bowman series 2 Luminescence Spectrometer (Sopra, Büttelborn, Germany), whereas spectrophotometric measurements were performed on a UV-2100 spectrophotometer (Shimadzu, München, Germany).

SDS-PAGE was performed on 7.5% acrylamide slab gels using 0.05 µg of MTSE per slot and myosin, -galactosidase, phosphorylase b, bovine serum albumin, and ovalbumin as molecular weight markers (18). Runs under nonreducing conditions were performed in the absence of -mercaptoethanol in the sample buffer. Gels were stained with Coomassie Blue G-250 and scanned with an MPS 940.800 densitometer (Vitatron, Rösrath, Germany).

To determine the native molecular weight of MTSE from E. gracilis, 0.2 mg of the enzyme was cross-linked for 10 h at 4 °C in 1 ml of 20 mM potassium phosphate, 2% glutaraldehyde (19). The reaction was stopped by adding 4-fold concentrated SDS-PAGE sample buffer, and the molecular weight of the cross-linked protein was estimated by SDS-PAGE under reducing conditions.

Analytical gel filtration was performed on a Superose 6/HR 10/30 column (Pharmacia Biotech, Uppsala, Sweden), calibrated with thyroglobulin, apoferritin, -amylase, alcohol dehydrogenase, albumin, carbonic anhydrase, and cytochrome c (20). 0.5 mg of MTSE was applied to the column, and elution was performed at 400 µl/min. To avoid unspecific hydrophobic or ionic interactions between the stationary phase and the proteins, 150 mM NaCl and 10% ethylene glycol were added to the elution buffer. To follow subunit dissociation, additional runs were performed using the same calibrated column at increasing concentrations of guanidinium hydrochloride.

0.1 mg of MTSE from E. gracilis was applied to a glycoprotein assay (21) using the glycoprotein detection kit from Boehringer.

Protein concentrations were determined according to Bradford (22) using bovine serum albumin as the protein standard.

To determine parts of the amino acid sequence, complete tryptic digestion of E. gracilis MTSE was performed. For that purpose, 100 µl of 5 M guanidinium hydrochloride, 30 mM dithiothreitol, 0.5 M Tris/HCl, pH 8.6, were added to 150 µg of the purified enzyme and stored at 4 °C under argon overnight. After unfolding, the sample was dialyzed against 2 liters of 50 mM NH₄HCO₃, 1 mM CaCl₂, pH 7.8, for 4 h. After dialysis, 6 µg of trypsin were added, the solution was made 5% (v/v) CH₃CN, and digestion was carried out at 37 °C overnight. After cleavage, tryptic peptides were separated by reversed phase HPLC using a 140 B inert pump (Applied Biosystems, Weiterstadt, Germany), an SPD 10 A detector (Shimadzu, München, Germany), and an MBO-25-05 column (250 × 0.8 mm, C18, 5 µ, 300 A, LC-Packkings, Amsterdam, The Netherlands). A linear gradient from 0.1 to 0.09% trifluoroacetic acid and 0-84% CH₃CN was run for 200 min at a flow rate of 10 µl/min, and peaks were detected at 215 nm as well as 295 nm. Depending on the elution profile, peaks were collected manually. Symmetrical peaks without visible contaminations as judged by the HPLC elution profile were further analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-MS) as described (23) using a Lasermat 2000 (Finnigan MAT). For that purpose, an aliquot of 0.5 µl from 20 selected peak fractions of the HPLC separation was mixed with 1 µl of -cyanohydroxycinnamonic acid, and 1 µl thereof was subjected to MALDI-MS. Seven homogenous peaks, as judged by MALDI-MS, were subjected to automated Edman degradation (24, 25), and their N-terminal amino acid sequences were determined using a 473A or 476A protein sequenator (Applied Biosystems Inc., Foster City, CA).

Limited proteolysis was carried out using 0.04 mg/ml of E. gracilis MTSE dialyzed against 50 mM Tris/HCl, pH 7.8, at 37 °C and 7 µg/ml endoproteinase Glu-C from S. aureus. After different time intervals, samples were taken, assayed for enzymatic activities, and (after boiling for 5 min in SDS-PAGE sample buffer) subjected to SDS-PAGE analysis.

RESULTS

To elucidate the flux of the respective metabolites in MTSE, the K_m and V_max values of the involved active centers for their respective substrates were measured. A summary of the kinetic data is given in Table I.

Table I. Summary of kinetic data obtained from steady state experiments Reaction Substrate Km (or K0.5) kcat kcat/Km µM s1 s1 µM1 PRTa Anthranilate 20 1.9 0.095 PRAIa PRA 10 3.6 0.36 IGPS CDRP 400 3.3 0.008 TRPS IGP 670 2.0 0.003 TRPS Indole 25 1.9 0.076 TRPS L-Serine 500 1.9 0.004 Overall TRPS Anthranilate 20 1.8 0.09 a Both PRT and PRAI assays were started with the respective enzymes after a 10-min preincubation with complementing E. coli helper extracts. No changes in absorbance or fluorescence were observed during this period.

Using anthranilate as the substrate, PRT activity was measured both in the presence and in the absence of complementing helper activities, and no difference was observed. Steady state kinetics reveal significant substrate inhibition at anthranilate concentrations higher than 50 µM. The data indicate a slight cooperativity. Nevertheless, for better comparability a K_0.5 value was calculated according to the Michaelis-Menten model.

PRAI activity, if determined without an E. coli helper extract, was about 30% lower than the value obtained in the presence of the cell extract. An influence on K_m could not be detected. PRAI activity showed Michaelis-Menten behavior, and the enzyme was inhibited competitively by anthranilate with a K_i value in the millimolar range (data not shown).

Kinetic measurements with IGPS resulted in good data fits to the Michaelis-Menten model. The K_m of MTSE for CDRP is much higher than the K_m for anthranilate and PRA.

Michaelis-Menten behavior was observed for TRPS activities with IGP as substrate. No inhibition at higher anthranilate concentrations could be detected.

Depending on the substrate, measurements of TRPS activities revealed a different picture. Whereas titration with indole resulted in significant cooperativity, conversion of L-serine obeyed the Michaelis-Menten model.

In addition to the described single enzymatic activities, the overall reaction (anthranilate  L-tryptophan) was investigated using a spectrophotometric assay system at 290 nm in the presence of all necessary cosubstrates, phosphoribosyl pyrophosphate, and L-serine. The overall reaction was inhibited by anthranilate. In contrast to the K_m values, k_cat values of all activity measurements were in the same range.

Measurements of the overall reaction at different concentrations of potassium phosphate, sodium chloride, and Tris/HCl clearly showed an activity optimum at a concentration of 40 mM for these salts. In contrast, measurements at increasing concentrations of ammonium sulfate resulted in slightly increasing activities, but no distinct maximum was detectable (Fig. 1). The pH optimum was 7.8, and the temperature optimum was at 52 °C (data not shown).

The relative molecular weight of the purified enzyme and the product of cross-linking by glutaraldehyde, respectively, was determined by both reducing and nonreducing SDS-PAGE. The relative subunit molecular weight is 136,000, whereas the cross-linked chains migrated to a position corresponding to a value of 272,000 (data not shown). Calibrated gel filtration of the native enzyme using a Superose 6/HR column yielded a relative molecular weight of 245,000 (data not shown).

To follow subunit dissociation, analytical gel filtration was performed using buffers containing different concentrations of guanidinium hydrochloride. As shown in Fig. 2 (inset), two peaks in the elution profile can be detected at 1.2 M guanidinium hydrochloride, indicating the presence of both dimeric and dissociated protein species. Since it is well known that guanidinium HCl-induced subunit dissociation is followed by further unfolding of the resulting monomers at higher concentrations of the chaotropic agent, the elution volume in gel filtration decreases as a consequence.

Using the Boehringer glycosylation detection kit, no glycosylation of MTSE from E. gracilis could be detected (data not shown). Several approaches of N-terminal sequence analysis of the native enzyme failed, indicating that the protomer is blocked N-terminally. To obtain information about the primary sequence, complete tryptic digestion followed by reverse phase HPLC was performed, and seven homogenous HPLC peaks as determined by MALDI-MS were subjected to automated N-terminal sequence analysis making use of Edman degradation. In addition to the resulting seven main sequences, impurities within four HPLC peaks that were not detected by MALDI-MS yielded additional sequences. The resulting sequences are summarized in Fig. 3.

A similarity search was performed with all entries of the SwissProt protein sequence data base release 32.0 using the BLASTP program (26) and the BLOSUM62 scoring matrix (27). Whereas peptides 2a, 3a, 4b, 6a, and 7b exhibited no sequence homology to known Trp sequences, peptides 1a, 4a, 5a, and 7a showed homologies to PRT (TrpD), and peptides 5b and 6b showed homologies to IGPS (TrpC) and TRPS (TrpB), respectively. Alignments of the sequenced peptides to Trp sequences as obtained by BLASTP are given in Fig. 4.

Limited proteolysis using Endoproteinase Glu-C revealed an intermediate fragment with a relative molecular weight of 119,000 that is slowly degraded further during the time course of proteolysis (data not shown). The corresponding M_r 17,000 fragment could not be detected, indicating its rapid further degradation, possibly due to a loss in conformational protection against proteolytic attack. N-terminal sequence analysis of the M_r 119,000 fragment by Edman degradation failed.

To shed some light on the effect of proteolytic cleavage by Endoproteinase Glu-C, we determined both the activity and the reaction specificity of E. gracilis MTSE. During the time course of proteolysis, a loss in tryptophan synthase activity accompanied by an increase in serine deaminase activity was observed (data not shown).

DISCUSSION

The turnover numbers of both the overall activity and all five single activities of E. gracilis MTSE are in the same range, indicating that none of the single activities comprises a rate-limiting step in the overall reaction. Although slightly different, the k_cat values reported here roughly correspond to the specific activities as determined by Hankins and Mills (7). Table II gives an overview of turnover rate constants of tryptophan pathway enzymes from other sources. However, care should be taken in directly comparing those values, since various parameters such as buffer composition or reaction temperatures were different in the respective investigations. Nevertheless, compared with the respective enzymes from other sources, it can be seen that k_cat of PRT, TRPS, and TRPS of E. gracilis MTSE are the lowest values, whereas k_cat of PRAI and IGPS lie in a medium range. Thus, with respect to the single turnover rates, E. gracilis MTSE has no catalytic advantage compared with the individual enzymes derived from other microorganisms. Regarding the overall pathway from anthranilate to L-tryptophan within the living cell, we conclude that due to a decreased diffusion barrier even the low turnover rates of PRT and tryptophan synthase within E. gracilis MTSE are sufficient to result in a comparable overall efficiency to synthesize L-tryptophan.

Table II. Comparison of turnover rates, kcat of PRT, PRAI, IGPS, TRPS, and TRPS from different sources In most cases, kcat was calculated on the basis of the published specific activities of the purified enzyme species and their respective relative molecular weights. Source PRT PRAI IGPS TRPS TRPS Reference s1 s1 s1 s1 s1 E. gracilis 1.9 3.6 3.3 2.0 1.9 This work E. gracilis 1.9 1.9 1.9 1.9 1.9 This work E. gracilis 12.0 7.3 4.1 1.7 7 Erwinia carotovora 16.3 61 S. typhimurium 12.0 62 S. typhimurium 5.7 63 E. coli 4.4 64 Serratia marcescens 3.1 65 S. cerevisiae 2.9 66 S. typhimurium 2.7 67 S. marcescens 27.7 4.4 68 N. crassa 6.0 2.3 69 Bacillus subtilis 1.9 1.4 70 Aerobacter aerogenes 2.6 2.6 71 S. typhimurium 2.0 2.0 71 E. coli 46.0 72 E. coli (Tris acetate) 7.7 73 S. cerevisiae 5.5 74 E. coli (potassium phosphate) 2.0 73 E. coli 1.2 75 E. coli 9.7 76 E. coli 12.1 77 S. cerevisiae 7.6 7.6 5

Comparing the K_m values for all reaction intermediates from anthranilate to L-tryptophan gives an interesting pattern: the K_m values for the first and the last intermediates of the reaction pathway catalyzed by E. gracilis MTSE (anthranilate, PRA, and indole) are smaller by more than 1 order of magnitude as compared with those in the middle of the pathway (CDRP and IGP). This is exactly what one would expect if substrate channeling via a tunnel is realized in E. gracilis MTSE. Reaction intermediates, binding to active centers located in the core of the enzyme need to overcome a greater diffusion barrier. Substrate channeling is known for tryptophan synthase in S. typhimurium and E. coli (78-81) but not for PRAI/IGPS in E. coli (82). However, further detailed kinetic data are necessary to allow conclusions regarding substrate channeling in E. gracilis MTSE.

Determination of the subunit molecular weight of E. gracilis MTSE by MALDI-MS failed, since no suitable desorption conditions could be established. Nevertheless, SDS-PAGE analysis of the denatured and of the cross-linked protomer revealed a relative molecular weight of 136,000 and 272,000, respectively. The fact that densitometry of the respective gel yields only one highly symmetric band provides strong evidence that the enzyme exists as a homodimer of 272,000. However, the possibility of the existence of a heterodimer consisting of chemically different but physically similar chains cannot be totally excluded.

The values found in this investigation do not correspond to earlier reports making use of native molecular weight estimations by gel filtration over Sephadex G-200 as well as density gradient centrifugations (234,000; Ref. 6) and are even in contradiction with refined measurements (325,000 and 155,000 for the native enzyme and the subunit respectively; Ref. 7). However, we believe that our estimation resulted in the most accurate value, since the migration behavior of E. gracilis MTSE as shown in the original publication of Hankins and Mills (7) is identical to that observed in our experiments. The discrepancy obviously is due to a wrong calibration based on a false assumption of the molecular weight of -galactosidase (150,000), one of only three molecular weight markers in their experiment. The molecular weight of -galactosidase was later determined to be 118,000 (83). Compared with SDS-PAGE of cross-linked enzyme multimers, native molecular weight estimations using gel filtration in general are less accurate. In the latter case, the dependence on molecular shape as well as possible retardations caused by ionic and/or hydrophobic interactions with the stationary phase further increases possible errors. Indeed, we observed hydrophobic matrix interactions in the absence of ethylene glycol (data not shown), and even under improved buffer conditions the molecular weight under native conditions as determined by gel filtration (245,000) is smaller than expected, indicating the presence of significant residual retardation effects. We exclude the possibility that the enzyme under those conditions exists as a monomer, because the addition of guanidinium hydrochloride clearly demonstrated a transition to monomers requiring a concentration of 1.2 M (Fig. 2).

Table III gives a summary of the molecular weights of all monofunctional Trp enzymes catalyzing reactions that are also catalyzed by E. gracilis MTSE. A polypeptide chain consisting of five domains homologous to the above mentioned subunits requires an average molecular weight of 164,771 (Table III,  means). Even if the smallest molecular weight values ever published of each single subunit are considered, the minimum sum is still 152,495 (Table III, smallest values). Evidently, this sum is considerably higher than the molecular weight of E. gracilis MTSE (136,000). Thus, simple gene fusions of all five genes as known for the gene corresponding to bacterial trpA and trpB in S. cerevisiae (59), N. crassa (103), or Corprinus cinereus (56) can be excluded. Moreover, gene fusion would require additional sequences acting as connectors, resulting in an even higher molecular weight. Thus, either a considerably different architecture is realized, or one of the active centers possibly catalyzes a combined reaction e.g. isomerization of PRA to CDRP followed by IGP synthesis. The structural scaffold of three subunits (PRAI, IGPS, TRPS) has a (/)₈ barrel topology. Although the sequence homology between PRAI and IGPS is rather low, they are structurally very similar so that they can be superimposed with a root mean square deviation of 2.03 Å for 138 C pairs (104). Thus, it would be conceivable that one domain catalyzes both PRA isomerization and IGP synthesis. This argument is corroborated by the finding that the K_m value of the respective active center for CDRP is rather high, meaning that CDRP might not be a true intermediate in Euglena MTSE. Consequently, assuming two instead of three TIM barrels in E. gracilis MTSE reduces the expected molecular weight by 27,982, the size of an average TIM barrel enzyme (Table III, mean of TIM barrel means). The resulting expected molecular weight would then be 136,789 (Table III,  means  mean of TIM barrel means), a value almost identical to the molecular weight of E. gracilis MTSE as determined in this investigation. However, since MTSE is so far the only enzyme exhibiting all activities based on the (/)₈ barrel motif within one polypeptide chain, no comparison with known species can be made.

Table III. Summary of the molecular weights of all relevant monofunctional Trp enzymes as obtained from SwissProt The respective smallest molecular weight values of all subgroups are printed in boldface type. Source PRT PRAI IGPS TRPS TRPS Mr Ref. Mr Ref. Mr Ref. Mr Ref. Mr Ref. ACICA 37,556 39 23,005 85 30,216 39 44,322 84 ANTSP 28,761 a AZOBR 36,721 60 28,123 60 BACPU 36,789 50 BACST 28,736 87 44,012 87 BACSU 35,898 88 24,005 88 27,928 88 29,450 88 43,691 88 BRELA 36,649 42 29,222 42 44,704 42 BUCAP 38,090 b 30,685 b 44,247 b CANAL 24,696 89 CAUCR 22,902 90 28,519 90 43,491 90 CORGL 36,751 46 CRYNE 33,070 91 CYACA 26,807 92 ECOLI 28,724 58 42,852 58 HAEIN 35,625 36 28,730 36 43,256 36 HALVO 26,754 93 29,720 93 45,943 93 KLEAE 28,543 94 KLULA 22,880 95 LACCA 36,490 45 21,404 45 28,943 45 28,724 45 43,675 45 LACLA 35,845 31 39,586 31 29,732 31 27,685 31 43,746 31 METTH 37,296 43 24,383 43 30,403 43 28,803 43 42,115 43 METVO 31,974 55 44,685 55 PSEAE 37,384 34 30,347 34 28,488 96 43,607 96 PSEPU 37,203 35 30,550 35 28,460 97 44,047 97 PSESY 28,439 98 44,507 98 RHOCA 27,975 99 SALTY 28,670 100 42,806 32 SULSO 28,588 57 THETH 28,924 101 43,809 101 VIBPA 35,415 33 28,567 33 43,027 33 YEAST 41,374 37 24,158 102 Mean 37,006 26,009 29,051 28,887 43,818   means 164,771   smallest values 152,495 Mean of TIM-barrel means (PRAI, IGPS, TRP) 27,982   means  mean of TIM barrel means 136,789 a GenBankTM accession number p31204. b GenBankTM accession number p42389.

Alignment of the sequences of the 11 peptides to all SwissProt entries resulted not only in similarities to known Trp sequences but interestingly also in five sequences with no homology at all. This finding indicates unknown parts of the primary structure of E. gracilis MTSE that may belong to regions connecting different domains. Homology to IGPS was observed for peptide 5b. In the three-dimensional structure, this part of the sequence corresponds to the 4-5 loop, helix 5, and strand 5 of the essential TIM barrel fold as described for IGPS from E. coli (104) and Sulfolobus solfataricus (75). Thus, we are inclined to expect a similar fold in the E. gracilis enzyme. Peptide 6b is homologous to the N-terminal region of the bacterial -subunit, which in S. typhimurium has no defined secondary structure and is located at the contact surface to the -subunit (105). Assuming that this contact depends on this highly conserved region (Fig. 4), it is conceivable that, as in bacterial tryptophan synthase, the - and -domain of the E. gracilis enzyme are in contact. Four peptides are homologous to PRT. However, three-dimensional structures of PRT from different sources are not yet known.

Limited proteolysis is an ideal tool to detect solvent-exposed loops lacking defined secondary structure as described for the hinge region in the -subunit of tryptophan synthase from E. coli (106) or for the connector region of tryptophan synthase from S. cerevisiae (86). Using endoproteinase Glu-C (and some other endoproteinases; data not shown), only one major cleavage site could be detected. This finding gives evidence that either only one connector in E. gracilis MTSE is present or that possible other connecting peptides are not susceptible to proteolytic attack under the given experimental conditions.

Limited proteolysis of E. coli TRPS, using endoproteinase Glu-C results in a shift from -replacement to -elimination specificity and can be expressed by the ratio of tryptophan synthase to serine deaminase activity (44). Consequently, we measured both tryptophan synthase and serine deaminase activities of MTSE during the time course of proteolysis. Albeit less significant, we observed the same shift in reaction specificity as described for the E. coli enzyme. Therefore, it is likely that proteolytic cleavage occurs near or within the -domain of E. gracilis MTSE. The difference between the molecular weight of native MTSE and the large fragment after proteolysis roughly corresponds to the smaller (F2) fragment after limited proteolysis of the -subunit of tryptophan synthase from E. coli. Thus, proteolytic cleavage is likely to occur in a structural element corresponding to the hinge region in E. coli. Since N-terminal sequence analysis of the larger proteolytic fragment failed, we assume that this fragment contains the blocked N terminus. Thus, the site of proteolytic cleavage, and possibly the -domain itself, may be located near the C terminus.

For three reasons, E. gracilis MTSE is probably not simply the product of gene fusions of all five trp genes corresponding to the five detectable activities: 1) The subunit molecular weight of the purified enzyme is too small, 2) there are new amino acid sequences (no homologies detected by BLASTP), and 3) only one of four postulated connecting peptides can be defined by limited proteolysis. Future work, cloning the MTSE-encoding gene from E. gracilis by using degenerate PCR primers based on the presented amino acid sequences should yield the complete sequence and help to solve this structural puzzle.