INTRODUCTION

The observation of an accelerated cis/trans isomerization of the oligopeptide succinyl-Ala-Ala-Pro-Phe-4-nitroanilide in biological material led to the discovery of peptidyl-prolyl cis/trans-isomerases (PPIases, E.C. 5.2.1.8)¹ in 1984 (1). Recently, it was shown that they can also act on polypeptides as folding helper enzymes during the refolding of proteins in vitro (2) and in vivo (3, 4). By comparison of their amino acid sequences PPIases can be subdivided into three families: the cyclophilins (5), the FK506-binding proteins (FKBPs) (6), and the parvulins (7, 8). The families consist of many different members even in the same cell type or organism (9, 50).

Little data exist describing the occurrence of Mip-like proteins or PPIases of the FKBP family in Enterobacteriaceae such as Escherichia coli. In addition to two members of the cyclophilin family of PPIases, E. coli also contains the 10.1-kDa PPIase parvulin (7), the 48-kDa trigger factor (21), and three FKBP-like genes. The predicted protein product of one open reading frame, orf149, shows similarities to the FKBPs (22). The slyD gene (23) and its protein product (24) is 47.3% similar to the FKBP family but the protein does not exhibit PPIase activity in the standard enzyme assay. The amino acid sequence deduced from the recently discovered fkpA gene is much more related to the Mip-like FKBPs (25), showing 83% identity to the consensus sequence of the catalytic core as derived by Trandinh et al. (26). In all these cases, it remains unclear whether or not the deduced proteins have PPIase activity and contribute to the PPIase pattern of E. coli.

In this study we describe the purification and characterization of a new periplasmic, Mip-like PPIase from E. coli, and show that similar genes are present in other Gram-negative bacteria.

Eukaryotic PPIases interact with many cellular proteins, some of which may include in vivo substrates. These interactions have been detected by the yeast two-hybrid transcription assay, affinity chromatography, and chemical cross-linking experiments (for reviews, see Refs. 9, 10, and 50). This array of natural ligands may imply that PPIases control several cellular activities either by their catalytic activity or by sequence-specific recognition of proline residues.

Much less information is available concerning prokaryotic PPIases and their specific binding proteins. Most interesting is the superficially localized Mip protein, a virulence factor of the Gram-negative bacterium Legionella pneumophila (11), because it belongs to the family of FKBPs (FKBP25mem)² (12). These bacteria cause a severe pneumonia in men (Legionnaires' disease) by growing intracellularly in human blood monocytes, human alveolar macrophages, and also macrophage-like tissue-cultured cells (13). FKBP25mem may be involved in interactions with other components on the bacterial surface or with proteins of the host cell (14). The enzyme is a dimer in its active state and the COOH-terminal part of 106 amino acids shows homology to FKBPs. The same type of FKBP is produced by various pathogenic as well as non-pathogenic Legionella strains (15, 16).

In addition, mip-like genes have also been detected outside Legionella species, in Coxiella burnetii and Rochalimaea quintana, using Southern hybridization (17). The sexually transmitted pathogen Chlamydia trachomatis possesses a protein that is a virulence-related Mip-like PPIase (C. trachomatis FKBP27) (18). Finally, the infectious stage of the intracellular parasite Tryphanosoma cruzi secretes a Mip-like PPIase that enhances invasion of mammalian host cells (19). Enhanced invasion also occurs when externally produced Mip-like protein is added to the medium during assay of infection of a mutant strain of T. cruzi that cannot make the protein itself. The tight binding inhibitor of FKBPs, the peptidomacrolide FK506, and the non-immunosuppressive C¹⁸-hydroxy, C²¹-ethyl derivative of FK506 protected HeLa-cells from penetration by trypomastigotes. Therefore, enzyme inhibition alone, not an indirect interaction with calcineurin (20), was believed to mediate the effect of peptidomacrolides on infectivity. Since Mip-like PPIases occur in many intracellular pathogens, their involvement in entry or replication may be inferred.

MATERIALS AND METHODS

Bacterial strains used in this work include E. coli K12 HB101, E. coli XL1-Blue (recA1 lac thi endA1 gyrA96 hsdR17 supE44 relA1 [FproAB lacLy lacZM15 Tn10]) (Stratagene) and those described in the legend to Fig. 4. The cloning vector was pBR322. Bacteria were grown in LB broth and plasmids were maintained in and prepared from E. coli XL1-Blue, using the following antibiotics, as appropriate: ampicillin (50 µg/ml) and tetracycline (12.5 µg/ml) (Sigma).

A culture of E. coli K12 HB101 (4 liters) was grown overnight at 37 °C in LB-broth (10 g/liter bactotryptone, 5 g/liter NaCl, 5 g/liter yeast extract). Cells were harvested by centrifugation at 4 °C for 15 min at 6000 × g in a Beckmann J2-HC centrifuge and sedimented cells were resuspended in 2 mM Tris buffer, pH 8.0. Cell rupture was performed by a SLM Aminco French pressure cell. The cell lysate was stirred with 0.5% (v/v) Triton X-100 and 0.1% (v/v) Benzonase for 15 min at 22 °C and ultracentrifuged in a Beckmann L8 60M centrifuge at 20,000 × g for 30 min at 4 °C. The supernatant was applied to a Fractogel EMD DEAE-650(M) column (2.5 × 20 cm), equilibrated with 2 mM Tris buffer, pH 8.0. The column was washed with 300 ml of equilibration buffer to remove unbound protein. Bound protein was eluted by running a linear gradient from 0 to 2 M NaCl in 500 ml of 2 mM Tris buffer, pH 8.0. The PPIase active fractions were detected as described under ``Enzyme Assay and Inhibition Studies with FK506 and Cyclosporin A,'' and pooled. Protein was dialyzed overnight against 3 liters of 5 mM cacodylate buffer, pH 5.3, at 6 °C. Denatured protein was removed by centrifugation and the soluble fraction was adjusted to pH 7.8. An IMAC chromatography was performed for further protein purification using a chelating Sepharose Fast Flow column (1.6 × 16 cm) (Pharmacia, Uppsala). The column was loaded with NiSO₄ and equilibrated with 20 mM HEPES buffer, pH 7.5, containing 300 mM NaCl and 0.5 mM MgCl₂. Protein was applied to the column at a flow rate of 1.5 ml/min, after which the column was washed with 100 ml of equilibration buffer. E. coli FKBP22 was eluted with 500 mM imidazole in the equilibration buffer. PPIase-active fractions were pooled and concentrated with a Filtron OMEGACELL, exclusion size 10,000 Da. Samples (1 ml) were applied to a Superdex 75 gel filtration column (1.6 × 60 cm) (Pharmacia), equilibrated with 10 mM Hepes buffer, pH 7.8, containing 150 mM KCl, 1.5 mM MgCl₂, and 0.5 mM dithioerythritol. The flow rate was 0.8 ml/min. Fractions containing E. coli FKBP22 were pooled and ammonium sulfate was added to yield a concentration of 20%. The solution was stirred for 1 h at 6 °C. A HIC column was prepared using phenyl-Sepharose CL-4B (1 × 8 cm) and was equilibrated with gel filtration buffer containing 20% ammonium sulfate. The protein mixture was applied to the column at a flow rate of 1 ml/min. PPIase-active fractions were obtained by running a linear gradient from 0 to 0.5% CHAPS and 0 to 20% glycerine in 100 ml of 35 mM Hepes buffer, pH 7.8. The protein derived from the HIC column was dialyzed overnight against 5 liters of 5 mM Hepes buffer, pH 6.0, containing 0.5 mM MgCl₂, at 6 °C. The protein solution was successively passed through a Fractogel TSK AF-Blue column (1 × 6 cm) and a Fractogel EMD DEAE-650(M) column (1 × 2.5 cm), both equilibrated with dialysis buffer. PPIase-active protein passed through the Fractogel TSK AF-Blue column and was eluted from the Fractogel EMD DEAE-650(M) column by a linear gradient of 0-1 M KCl in 100 ml of buffer. Fractions containing E. coli FKBP22 were analyzed by SDS-PAGE, revealing the purification of a homogenous protein.

Samples for SDS-PAGE and Western blot analysis were prepared as described elsewhere (27).

Slab gels (15%) were generated on a Minigel G41 vertical unit (Biometra, Göttingen). Transfer to nitrocellulose NC45 (SERVA, Heidelberg) was conducted in 25 mM Tris buffer, pH 8.3, containing 150 mM glycine and 10% methanol, using the semi-dry electroblotting system Fast Blot B32 (Biometra, Göttingen). Visualization of the transferred proteins was performed by staining with an Amido Black solution in 1% acetic acid. Blots were destained by shaking in 50 mM Tris buffer, pH 7.5, containing 500 mM NaCl and 1% Tween 20. A polyclonal antiserum, specific to the Mip-protein (FKBP25mem from L. pneumophila Philadelphia I), was obtained from rabbit (28).

A 1:2,000 dilution of the anti-Mip serum was used in the primary incubation step. Peroxidase-conjugated goat anti-rabbit IgG (Sigma, Deisenhofen) was used for detection. The blot was developed using a 0.018% 4-chloro-1-naphthol solution in 50 mM Tris buffer, pH 6.0, containing 150 mM NaCl and 0.024% hydrogen peroxide. The total protein content was determined using the Bradford assay (29) and bovine serum albumin as standard.

A sample of purified protein was concentrated, desalted, and transferred into a volatile solvent by means of reversed-phase HPLC as described below (``Protein Sequencing''). The molecular mass was determined by electrospray MS on a VG BIO-Q (Fisons Instruments) consisting of an electrospray ion source and a triple quadrupole mass analyzer with a mass range of 4000 as described elsewhere (7).

The molecular mass of the native E. coli FKBP22 was determined by size exclusion chromatography on a Superdex 75 column (1.6 × 60 cm) (Pharmacia LKB). The column was equilibrated with 10 mM Hepes buffer, pH 7.8, containing 150 mM KCl, 1.5 mM MgCl₂, and 0.5 mM dithiothreitol at a flow rate of 0.8 ml/min. Purified protein (60 µg) was applied to the column and detected at 280 nm. Retention time was estimated by using the Liquid Chromatography Controller LCC-500 (Pharmacia LKB). The column was calibrated with a set of calibration proteins for gel chromatography (combithek, Boehringer, Mannheim) under the conditions described above.

Measurements were performed in 35 mM Hepes buffer, pH 7.8, at 10 °C in the protease-coupled assay (1, 5). At measurement conditions an equilibrium exists of about 80-95% trans and 5-20% cis conformer with respect of the prolyl bond of the substrates Suc-Ala-Xaa-Pro-Phe-4-NA (Xaa stands for a variable aminoacyl residue in the P₁ position of various substrates). Chymotrypsin was used to cleave the anilide bond of the trans (but not the cis) conformer in a concentration of 0.5 mg/ml in the reaction mixture. The assay was performed by adding an 2-µl aliquot of the peptide dissolved in dimethyl sulfoxide (10 mg/ml). A Hewlett-Packard 8452 diode array UV/VIS spectrophotometer was used for monitoring the time course of the reaction. The reaction product of chymotryptic cleavage, 4-nitroanilide, was detected by the absorbance at 390 nm. First-order rate constants were calculated from 500 data points. The substrate Suc-Ala-Phe-Pro-Phe-4-NA was used to measure the PPIase activity in protein purification and inhibition studies.

FK506 was a gift from Fujisawa Pharmaceutical Co., Osaka. Stock solutions of the inhibitors were prepared in 50% ethanol. The incubation time was 5 min for FK506 and 15 min for CsA. Three independent experiments were performed. The data were fitted to the equation for competitive tight-binding inhibition (30), Equation 1. Values of k_enz were obtained from the relation k_obs = k_enz + k₁ (k₁ stands for the uncatalyzed and k_enz for the PPIase-catalyzed cis to trans interconversion of the peptidyl-prolyl bond of the substrate).

(Eq. 1)

Experimental conditions were the same as described above for measurements of PPIase activity. Reported data are given as the mean of three to five measurements. Substrates were purchased from Bachem (Heidelberg). Stock solutions of various substrates were made in dimethyl sulfoxide.

Genetic manipulations were performed according to established procedures. Plasmids were prepared by alkaline lysis or boiling lysis, and linear DNA fragments for cloning were isolated from low melting agarose. DNA from Kohora  phage 655 (31) was isolated by standard procedures and a 6.05-kilobase HindIII-SpeI DNA fragment containing the fklB gene was excised and cloned into pBR322 previously digested with HindIII-NheI, creating a 10.2-kilobase plasmid, pSH219. Cloning was confirmed by comparing the restriction pattern of the subcloned DNA to the pattern predicted from the previously sequenced region (designated as ``o259,'' GenBank accession numbers U14003[GenBank] and J01749[GenBank]) (32). In addition, the 5 and 3 ends of the cloned fragment were sequenced to confirm that the fklB gene was present. DNA was sequenced by the dideoxy chain termination procedure using [³⁵S]dATP (Amersham Corp.). Two primers, 5-CCGGCGCGTGCCAGAC-3 and 5-CGGGGTCTTTGCAAC-3 (Midland Certified Reagent Co., Midland, TX), were used to prime sequencing reactions for the 5 and 3 ends of the fklB gene using the Sequenase II sequencing kit (U. S. Biochemical Corp.). Restriction endonucleases and other DNA modifying enzymes were from New England Biolabs (Beverly, MA). Alignment, restriction analysis, and comparison of DNA sequences were performed with the DNA Inspector II software (Texaco, Hanover, NH) and with PC Gene (IntelliGenetics, Mountain View, CA).

For Southern blotting, chromosomal DNA was isolated from Gram-negative bacteria by standard procedures, digested with restriction enzymes, electrophoresed on a 0.8% or 1.0% agarose gel, and subsequently transferred to a nylon membrane. A 1.2-kilobase BglII-MluI DNA fragment containing the complete fklB gene was purified from pSH219 and was used to create randomly labeled DNA probes by labeling with [³²P]dCTP (Amersham Corp.), using the method of Feinberg and Vogelstein (33). In addition, probes were derived from the fkpA gene as described previously (25). Each set of probes was hybridized under conditions of low stringency to chromosomal preparations from various bacteria.

An overnight culture of E. coli K12 HB101 (1 liter) was harvested by centrifugation at 3,500 × g. The cells were washed three times with PBS buffer (10 mM phosphate buffer, pH 7.5, containing 150 mM NaCl) and suspended carefully in 50 ml of 20% (w/v) sucrose in 20 mM Tris buffer, pH 8.0. The cells were then pelleted and resuspended in 25 ml of 10 mM MgCl₂ to induce osmotic shock. After centrifugation at 3,000 × g, the periplasmic components were located in the supernatant. All steps were performed at 6 °C. The pH of the supernatant was adjusted to 8.0 by adding 100 mM Tris buffer, pH 8.0 (1 ml), and protein was concentrated in a Filtron OMEGACELL, 3,000 Da. The sediment derived from the osmotically shocked cells were suspended in 20 ml of 35 mM Hepes buffer, pH 7.5, to obtain cytoplasmic compounds. These cells were disrupted by four passages through a French pressure cell at 20,000 p.s.i. Debris was removed by ultracentrifugation in a Beckmann L8 60M centrifuge at 4 °C for 30 min at 20,000 × g.

Cells of a 100-ml overnight culture of E. coli HB101 were harvested and washed as described above. The resulting sediment was resuspended in 10 ml of PBS buffer and aliquots of 2 ml were centrifuged. Sedimented cells were suspended in 200 µl of 0, 0.5, 1, 2, or 5% (v/v) Triton X-100 containing PBS. The mixtures were shaken for 30 min at 25 °C. Cells were removed by centrifugation and supernatants were used for further investigations.

A sample of purified E. coli FKBP22 was dissolved in 75% formic acid, and cyanogen bromide (a small crystal) was added for cleavage. The mixture was incubated at room temperature under nitrogen in the dark for 24 h. A sample of the purified protein was digested with trypsin (sequencing grade, Boehringer, Mannheim) at an enzyme to substrate ratio of 5:100 in 0.1 M ammonium hydrogen carbonate, pH 8.5, for 14 h at room temperature. The digestion was stopped by acidification with 2 M acetic acid. Another sample of the purified protein was digested with endoproteinase Asp-N (sequencing grade, Boehringer, Mannheim) in 0.2 M ammonium hydrogen carbonate, pH 8.5, for 6 h at 32 °C. The enzyme to substrate ratio was 1:100.

Before sequence analysis and mass spectrometry protein was desalted by reversed-phase HPLC on a C3 column (125 × 4-mm Nucleosil 500-5 C₃-PPN, Macherey-Nagel, Düren). The protein was eluted with an acetonitrile (0.08% trifluoroacetic acid) gradient of 30-60% in 0.1% trifluoroacetic acid for 20 min at 40 °C (flow rate 1 ml/min). Peptides were separated using the same column and solvent system but with an acetonitrile gradient of 1-60% for 60 min for tryptic peptides and for 40 min for CNBr- and Asp-N-peptides (flow rate 1 ml/min, column temperature 40 °C). HPLC separations were performed on a Shimadzu LC-10A modular HPLC unit.

Amino acid sequences were determined by using an Applied Biosystems 476A sequencer model, according to the manufacturer's instructions. The COOH-terminal tryptic peptide was sequenced after covalent attachment to a Sequelon AA membrane using the attachment kit supplied by Millipore.

The E. coli FKBP22 sequence was compared to data base entries using programs from the UWGCG (University of Wisconsin Genetics Computer Group) and PIR (Protein Identification Resource) software packages. Similar sequences were searched using the FASTA and TFASTA algorithms (34). Pairwise comparisons were performed using BESTFIT and multiple alignments were produced with the CLUSTAL program (35).

RESULTS

A typical purification procedure started with a 4-liter overnight culture of E. coli cells yielding 170 µg of homogeneous protein (Table I). Since measurement of PPIase activity during protein purification does not discriminate between various enzyme species, the values of specific activity and purification (Table I) refer to the total measurable PPIase activity in the protein solutions. A Coomassie Brilliant Blue-stained SDS-PAGE showed a single band after the last purification step indicating a homogenous protein (Fig. 1).

Table I. Purification steps resulting in homogeneous E. coli FKBP22 protein Measurements of PPIase activity were performed in 35 mM Hepes buffer, pH 7.8, at 10 °C, with Suc-Ala-Phe-Pro-Phe-4-NA as substrate. Purification step Total protein Total activity Specific activity Purification Recovery mg arbitrary units arbitrary units/mg -fold % Lysate from French pressure cell press 3,000 350,000 115 1.0 100 Fractogel-EMD DEAE, pH 8.0 960 218,000 227 1.9 62 pH denaturation, pH 5.3 550 200,000 364 3.1 57 Chelate-Sepharose, pH 7.5 210 30,000 143 1.2 9 Superdex 75 HiLoad 16/60 47 21,400 455 4.0 6 Phenyl-Sepharose, pH 7.8 3.6 4,876 1,353 11.8 1.4 Fractogel TSK AF-Blue and Fractogel-EMD DEAE, pH 6.0 0.17 765 6,024 52.4 0.2

Molecular mass of the E. coli FKBP22 as calculated from the migration behavior in SDS-PAGE was 25,100 Da. Using electrospray mass spectrometry a molecular mass of 22,084.86 ± 1.47 Da was determined. The molecular mass of the native enzyme was estimated by gel filtration on a Superdex 75 column. Under nondenaturing and mild reducing conditions, the protein eluted as a single peak at a retention time according to a molecular mass of 53,200 Da. The protein peak corresponded perfectly with the enzyme activity.

The influence of the immunosuppressants CsA and FK506 on the enzymatic activity of E. coli FKBP22 was investigated. CsA did not affect the PPIase activity of the enzyme up to a concentration of 5 µM in the reaction mixture (data not shown). In contrast, the enzymatic activity of the E. coli FKBP22 was sensitive to FK506 (Fig. 2). An inhibition constant K_i of 25.1 ± 3.4 nM was determined by fitting the data to the equation for competitive tight-binding (Equation 1).

The specificity constant k_cat/K_m of E. coli FKBP22 for the most preferred substrate Suc-Ala-Leu-Pro-Phe-4-NA is 1,332,500 s¹ M¹. The substrate specificity profile of E. coli FKBP22 shows a pattern typically for FKBPs with preference of amino acid residues with a hydrophobic side chain in the P₁ position of the substrate (36). Interestingly, the enzyme shows an unusually high affinity for the substrate with a lysyl residue at this position (Table II). The correspondence of the substrate specificities of L. pneumophila FKBP25mem and E. coli FKBP22 indicates a close relationship for both enzymes.

Table II. Comparison of substrate specificity of E. coli FKBP22 with other FKBPs The ratios of the specificity constants kcat/Km for peptide substrates of the type Suc-Ala-Xaa-Pro-Phe-4-nitroanilide are shown. E. coli FKBP22 is compared with rhFKBP12 and L. pneumophila FKBP25mem. Suc-Ala-Phe-Pro-Phe-4-nitroanilide was set to 100%. Measurements were performed in 35 mM Hepes buffer, pH 7.8, at 10 °C. Xaa E. coli FKBP22 L. pneumophila FKBP25mem rhFKBP12cy Gly 1 0 0 Ala 6 6 12 Val 7 9 48 Ile 12 32 88 Leu 205 122 145 Phe 100 100 100 Trp 21 32 19 His 19 18 4 Lys 53 9 6 Glu 0 0 0 Gln 9 7 6

The fklB gene was cloned from Kohara phage 655 as described under ``Materials and Methods.'' The restriction map and predicted open reading frames are diagrammed in Fig. 3. Twelve of 16 Enterobacteriaceae species contained fklB-like genes (Fig. 4A). When the nylon membrane was stripped and rehybridized with probes derived from the fkpA gene (Fig. 4B), several fkpA-specific DNA fragments were identified to which the fklB-derived probes cross-hybridized (designed by A in Fig. 4A). In each case, probing with fkpA resulted in a stronger hybridization signal than with fklB, indicating that the respective fragments represented fkpA-like genes. When these bands were excluded, 8 of 12 Enterobacteriaceae species contained a fklB-like gene(s) on a single HincII DNA fragment (Fig. 4A).

The fklB gene is present and controlled by its own regulatory elements on plasmid pSH219 (Fig. 3). When pSH219 was present in E. coli, no overproduction of the E. coli FKBP22 protein occurred (data not shown). This suggests that the fklB gene is poorly expressed or that the level of E. coli FKBP22 is tightly controlled.

A Mip-specific polyclonal antiserum was used (37) to investigate a possible cross-reaction with E. coli FKBP22. Purified E. coli FKBP22 shows a weaker but still clear response at the titer of the Mip-specific antiserum used for detection of recombinant Mip protein in extracts from E. coli cells, (Fig. 5). The dilution of antiserum used in this experiment was too high for the detection of E. coli FKBP22 protein in the whole cell extract.

The subcellular distribution of E. coli FKBP22 was studied. The periplasm of cells was concentrated 10-fold, yielding a protein concentration of 0.65 mg/ml. A sample of periplasm was applied to a 15% SDS-PAGE. Mip-specific antiserum (dilution 1:200 (v/v)) was used in an immunoblot experiment. We found a visible band at the same position as for purified E. coli FKBP22 (Fig. 6). E. coli cells were investigated for a possible outer membrane localization of the FKBP22. The cells were washed with PBS buffer containing an increasing amount of Triton X-100. A protein concentration below 0.01 mg/ml was detected in the supernatant after sedimentation of the cells, when no detergent was added. A 0.5% (v/v) PBS/Triton X-100 solution yielded 0.69 mg/ml protein. Increasing the detergent concentration of the mixture to 5% (v/v) did not further increase the protein content of the supernatant. Samples of equal protein concentration from E. coli periplasm and of the Triton X-100-soluble cell surface proteins were applied to a 15% SDS-PAGE and blotted to nitrocellulose. No immunoreaction could be detected in the lanes containing the soluble surface proteins (Fig. 6), in contrast to the periplasmic fraction and the reference protein. Therefore, a periplasmic localization of E. coli FKBP22 without exposure on the cell surface is very possible.

The strategy and results of amino acid sequence analysis are shown in Fig. 7. The NH₂-terminal sequence analysis of the intact E. coli FKBP22 yielded the first 80-amino acid residues. Digestion of FKBP with endoproteinase Asp-N generated four peptides (D) due to incomplete cleavages behind Val-54, Thr-119, Ile-131, and Phe-136. These peptides were separated by reversed-phase HPLC. Sequence analysis of the complete peptide D3/4 (position 43-68) and of the NH₂-terminal part of D5/6/7 (position 69-205) confirmed the NH₂-terminal sequence and the presence of a methionine in position 79. Cleavage of E. coli FKBP22 with cyanogen bromide and separation of the peptides resulted in three major fragments (CNBr) and a minor fragment due to an incomplete cleavage at Met-79. The CNBr-peptide CNBr2 (position 80-164) could be sequenced to Asp-137, with the latter steps of this sequence being confirmed by sequences of the tryptic fragment T15 (position 124-129) and T16 (position 130-142), derived from the tryptic digest after reversed-phase HPLC separation. CNBr3 (position 165-205) could only be sequenced to Pro-190 because of its unfavorable composition. Some uncertainties were clarified by the sequences of the tryptic peptides T18 (position 170-183) and T19 (position 184-205). The complete sequence analysis of the latter COOH-terminal peptide could only be performed after covalent attachment to a Sequelon AA membrane via its carboxyl groups. The gap between Arg-142 and CNBr3 was closed by the complete sequence analysis of T17 (position 143-169). At this stage, the complete protein sequence was established and overlapped by Edman degradation with the exception of the link of T16 to T17. However, this sequence assignment was unambiguously confirmed by the results of the ES-mass spectrometry of the cyanogen bromide peptides and the tryptic peptides T16-T19. The following molecular masses of the tryptic peptides were determined by ES-MS: T16, 1378.20; T17, 2809.77; T18, 1704.75; and T19, 2303.00, giving the sum (-3H₂O) of 8141.72. This is in excellent agreement with the calculated sequence-based mass from positions 130-205: 8142.26. The molecular mass of CNBr2 was determined to be 9113.56 ± 3.3 and 9129.29 ± 2.0 which is in good agreement with the calculated values of 9111.09 for the homoserine lactone peptide and 9129.09 for its homoserine form. The molecular mass of the COOH-terminal CNBr3 was determined to be 4458.36 ± 0.85 which fits perfectly the calculated value of 4458.1. Furthermore, the molecular mass of the E. coli FKBP22 was determined by ES-MS as 22084.86 ± 1.47, the calculated value from the sequence is 22084.53. 

DISCUSSION

In contrast to several other prokaryotic organisms, a PPIase active member of the FKBP family has not been found in E. coli. Although, the E. coli gene fkpA encodes a Mip-like protein, its PPIase activity has not yet been demonstrated (25). We have now purified a PPIase from E. coli (FKBP22) that belongs to the FKBP family.

Identification and purification of a new PPIase from E. coli is complicated because of the existence of various other PPIases. Additional PPIases in E. coli can be detected only after identifying and removing known enzymes in the course of the purification procedure. Cyclophilins Cyp18cy and Cyp21peri (39) contribute the major PPIases activity in E. coli: about 80 and 10% of the total measurable PPIase activity, respectively. Trigger factor, a ribosome-associated PPIase in E. coli (21) with some sequence similarity to FKBPs, represents about 5% of measurable PPIase activity. A fourth enzyme, parvulin, represents a new family of PPIases with a distinct primary structure (7, 8). This protein contributes less than 1% of the total measurable PPIase activity. E. coli FKBP22 also represents approximately 1% of the total PPIase activity in E. coli, taking into account the loss of protein at each step of the preparation.

The molecular mass of E. coli FKBP22 was calculated to be 22,085 Da from the amino acid sequence and this mass was confirmed by mass spectrometry and SDS-PAGE, indicating a relationship to the Mip subfamily of FKBPs with molecular masses ranging from 20 to 30 kDa. Under non-denaturing conditions (gel filtration), a molecular mass of 53,200 Da could be estimated for E. coli FKBP22, suggesting that the protein is a dimer in solution. FKBP25mem also behaves as a dimer in solution and probably on the surface of the Legionella cells (38). Perhaps this is a characteristic property of other prokaryotic Mip-like members of the FKBP family.

A comparison of the primary structure of E. coli FKBP22 with known sequences from the data bases reveals a close relationship to the family of Mip-like FKBPs. Mip-like proteins have a FKBP-homologue region in the COOH-terminal part of the protein with an NH₂-terminal extension of about 100-150 amino acids of unknown function. The COOH-terminal part of E. coli FKBP22 starts at residue 102 (Gly) and shows considerable amino acid sequence homology with human FKBP12. Comparison with the 30 amino acid residues which are considered to form the consensus sequence of FKBPs (26) resulted in 90% similarity and 83% identity, respectively, for E. coli FKBP22. When the nucleotide and amino acid sequences of fklB and its protein product were aligned with those of fkpA and its predicted product, two different regions were apparent. The most conserved portion was the carboxyl terminus of the proteins (Fig. 8). In this region, there is 49% identity and an additional 11% similarity in the amino acid sequences. The DNA sequences that encode these carbonyl fragments were especially highly related: the fklB sequence was 60% identical to that of fkpA, and only four gaps were required to align the two genes (three 1-nucleotide gaps and one 2-nucleotide gap) (Fig. 8). These gaps form two sets of insertions and/or deletions that first interrupt, then restore the correct reading frames between the two genes (shaded areas in Fig. 8). The simplest conclusion is that the carbonyl regions of fklB and fkpA are very closely related by common descent from a single ancestral gene. The second region of comparison between fklB and fkpA and their protein products occurs in the amino termini of the two proteins. The amino acid sequences in these areas were only 21% identical to one another, with an additional 17% similarity (data not shown). The DNA sequences that encoded these amino termini were 53% identical but required 27 gaps and/or insertion (of up to 11 nucleotides) for alignment (data not shown). The implication is that the function of this portion of the protein has not been conserved between the two.

Alignment of the amino acid sequences of the Mip-like proteins from prokaryotic organisms (L. micdadei, C. trachomatis, Coxiella burnetii, E. coli fkpA gene) with human FKBP12 and comparison with its three-dimensional structure (46) revealed that they all lack an amino acid residue in the turn region of human FKBP12 between the ₁- and ₄-sheets and two amino acid residues between the ₅-sheet and the -helix (see Fig. 9).

Certain similarities in the amino acid sequences of Mip-like PPIases from different organisms exist also in their NH₂-terminal parts; the identity and homology between the sequences of E. coli FKBP22 and L. pneumophila FKBP25mem are 28 and 49%, respectively.

According to secondary structure prediction using the Chou-Fasman algorithm (48) the predominant structure of the NH₂-terminal part of E. coli FKBP22 is -helical. For all prokaryotic Mip-like proteins, the predicted structure of the NH₂-terminal part of the Mip-proteins seems to contain up to three larger -helices followed by a short -sheet or turn. No transmembrane spanning region could be found by estimation of free transfer energies (49) of the appropriate -helical structures. No amphipathic structures were present, as determined by the hydrophobic properties.

L. pneumophila FKBP25mem is located on the bacterial cell surface (38). The same observation has been described for the Mip-like protein of the protozoan parasite T. cruzi (19). In contrast, C. trachomatides FKBP27mem was not detectable on the surface of Chlamydia cells (40). Nevertheless, a signal peptidase II recognition sequence was deduced from the open reading frame and palmitic acid was incorporated into the native protein, indicating a membrane localization of the mature gene product (40). In contrast to L. pneumophila FKBP25mem (38), we were not able to detect E. coli FKBP22 within the Triton X-100 soluble cell surface proteins. However, E. coli FKBP22 was clearly localized in the periplasm (Fig. 6). In contrast to L. pneumophila FKBP25mem (11) and the deduced amino acid sequence derived from the E. coli fkpA gene (25), a recognition sequence for signal peptidase I does not exist in E. coli FKBP22.

The specific PPIase activity of E. coli FKBP22 is in the same order of magnitude as other known FKBPs such as L. pneumophila FKBP25mem and human FKBP12cy. The substrate specificity of the enzyme reflects a close relationship to the Mip subfamily of FKBPs (Table II). Particularly striking is the preference for a substrate with a lysyl residue in the P₁ position.

Besides the primary structure, affinity for the immunosuppressants CsA and FKBP is an additional criterion to distinguish between the cyclophilins and FKBPs, in contrast to parvulin which is not inhibited by CsA or FK506 at inhibitor concentrations up to 5 µM (7). The PPIase activity of E. coli FKBP22 is inhibited by FK506 in the nanomolar range (K_i = 25 nM). The enzyme is 10-fold more sensitive to the macrolide FK506 than is L. pneumophila FKBP25mem (K_i = 220 nM (37)), but only one-tenth as sensitive as human FKBP12cy (K_i = 0.4 nM (42)).

However, a comparison of the inhibitory constants or the IC₅₀ values shows that the K_i of E. coli FKBP22 is in a range typical for other FKBPs, e.g. human FKBP16mem (55 nM) (43) and Neurospora crassa FKBP13 (48 nM) (44). On the other hand, the K_i of human FKBP25mem of 160 nM (45) corresponds remarkably well to the K_i of 220 nM of L. pneumophila FKBP25mem (37).

The three-dimensional structure of human FKBP12 revealed that the ligand binding pocket is a hydrophobic cavity between the -helix and the interior wall of a -sheet that is lined with a number of highly conserved aromatic residues (46). A set of 14 amino acid residues are involved in FK506 binding. Three of these residues are not conserved in E. coli FKBP22, Phe-46 is replaced by Ala, Glu-54 by Gly, and His-87 by Ala. This may be the reason for the loss of sensitivity to the inhibitor.

fklB-like genes are present in other members of the Enterobacteriaceae. Shigella species and E. coli are so closely related that the Shigellae and E. coli are considered to be one group (47). Consequently, the DNA fragments carrying the fkpA genes and the fklB-like genes were present in Shigella (Fig. 4, A and B).

fklB probes hybridized to DNA fragments, which varied in size among the organisms: Salmonella, Enterobacter, Klebsiella pneumoniae, Citrobacter freundii and Serratia marescens (Fig. 4A). Although Yersinia enterocolitica contained a fkpA gene (Fig. 4B, lane 16) it does not appear to carry a fklB-like gene (Fig. 4A), making it the only tested organism to have one gene but not the other. Providencia stuartii and Proteus mirabilis have DNA fragments that hybridize so weakly to the two probes that the existence of similar gene in these species is questionable (Fig. 4, A and B, lane 19).

Although FKBP25mem is a virulence factor of Legionella, the protein was also found in non-pathogenic Legionella strains with almost identical amino acid sequence and identical protein chemical and enzymatic properties (37). Similarly, in the genus E. coli there exist non-pathogenic as well as pathogenic strains. Non-pathogenic E. coli belong to the normal flora of the gastrointestinal tract of humans and various animals. However, some E. coli strains cause infectious diseases within or outside the intestine. Characterization of the Mip-like E. coli FKBP22 and the discovery of a second mip-homologue gene (fkpA) strongly suggests that these PPIase proteins may also play a general role in the virulence of some Gram-negative bacteria.

The Legionella model can be used to study the function E. coli FKBP22 further. Invasion assays should be performed using Mip-negative mutants of L. pneumophila complemented with the fklB gene from E. coli.