Oligomerization, membrane anchoring, and cellulose-binding characteristics of AbpS, a receptor-like Streptomyces protein.

Streptomyces reticuli produces a 34.6-kDa surface-anchored protein (AbpS) whose surface-exposed N terminus binds strongly to Avicel, a dominantly crystalline type of cellulose. The generation of a large set of mutated abpS-genes and the subsequent analysis of the corresponding proteins in vitro as well as in vivo in a Streptomyces host allow the assignment of the following characteristics for AbpS. (i) Amino acid residues participating directly in the cellulose-interaction are located at the N terminus. (ii) As ascertained by cross-linking experiments, AbpS forms homotetramers in its soluble as well as cellulose-bound form. (iii) The intermolecular assembly of four AbpS molecules is governed by two domains (including amino acids 60-110 and 161-212). Both domains possess large portions of alpha-helical regions in which hydrophobic amino acids are located on one side as known from coiled-coil proteins. (iv) The C-terminal part of AbpS comprising 35 amino acids contains a transmembrane domain. Due to the surface-exposed N terminus of AbpS and the presence of transmembrane helix the C terminus has to be situated in the cytoplasm of the S. reticuli hyphae. Thus AbpS connects the interior of the mycelia with the extracellular space and binds cellulose using a unique cellulose-binding module.

Streptomycetes are aerobic Gram-positive soil bacteria, which are adapted optimally to their natural environment (1). They undergo a complex process of morphological differentiation and growth, characterized by the formation of substrate mycelia, followed by a phase of aerial growth. Within the hyphae, chains of spores are formed (2)(3)(4). Their resistance to heat, dryness, or cold ensures survival of the strains in unfavorable periods (5). The production of anti-microbial substances may inhibit growth of competing organisms (6). The utilization of biopolymers (i.e. cellulose, chitin, or xylan) is a typical characteristic of streptomycetes, as they possess a wide variety of the corresponding catabolic enzymes (7). Additionally Streptomyces hyphae are surrounded by cell walls, which protect them from osmotic or mechanical disruption and extracellular attacks. Murein (peptidoglycan) is the main cell wall polymer, consisting of glycan strands (␤ 1-4-linked N-acetylglucosamine and N-acetylmuramic acid) and oligopeptides. These stem peptides are linked to the carboxyl group of Nacetylmuramic acid and interconnect glycan molecules via an interpeptide bridge. Peptidoglycan acts as permeability barrier, which should also be flexible, permitting morphological changes and allowing transport out of and into the mycelia (for reviews see Refs. 8 and 9). In addition, teichoic acids are covalently linked to muramic acid residues in peptidoglycan (cell wall teichoic acids) and may constitute 60% of the cell-wall material. The physiological role of the cell wall teichoic acids is thought to be involved in ion exchange, keeping the peptidoglycan sacculus in an expanded state by charge repulsion and control of the activity of autolytic enzymes. Other functions are related to phage binding and immunogenicity (10,11).
In addition to the teichoic acids, proteins have been found to be associated with the cell wall. In Streptomyces species only few surface-exposed proteins have been described up to now. These include a 23-kDa protein from Streptomyces lividans (12), a cell-bound esterase synthesized by the cyclophilin A-producing strain Streptomyces chrysomallus X2 (13), the myceliaassociated cellulase (14), catalase-peroxidase from Streptomyces reticuli (15), and the surface-active proteins (i.e. SapB from Streptomyces coelicolor and its homologue from Streptomyces tendae), which are involved in erecting aerial hyphae (16).
Recently we identified a 35-kDa protein from S. reticuli, which is very likely covalently anchored to the cell wall (17). Its N-terminal part protrudes from the surface of the hyphae, as demonstrated by immunolabelled ultrathin sections and investigations by electron microscopy (18). The protein has no enzymatic activity, but it interacts strongly with crystalline forms of cellulose (Avicel). AbpS 1 (for Avicel-binding protein from S. reticuli) recognized other biopolymers merely weakly (chitin and Valonia cellulose) or not at all (xylan, starch, and agar). By comparing the deduced AbpS sequence, no homology was found to any discovered cellulose-binding domain, which were often present within cellulases (19). AbpS possesses an up to now unique cellulose-binding module. By analysis of the secondary structure of the deduced AbpS sequence, a large centrally located ␣-helical structure showing a weak homology to the tropomyosin protein family and the streptococcal M-proteins (20) was identified. As AbpS has also been found to be associated to protoplasts, it is predicted that a C-terminally located stretch of 18 hydrophobic amino acids anchors the protein to the cytoplasmic membrane.
Beside streptomycetes, surface proteins possessing diverse functions were also discovered in several other bacteria (21). In pathogenic bacteria streptococci and staphylococci (22)(23)(24) these proteins are often called microbial surface components recognizing adhesive matrix molecules (MSCRAMM), such as fibronectin, collagen, or immunoglobulins (25,26). Because of their involvement in adherence of the bacteria and conceal-ment of the bacterial surface from the host's defense system, MSCRAMMs were studied intensively. Surface-associated heparin-binding proteins are frequent among pathogenic mycobacteria (27,28). Various types of cell-wall-anchored proteinases are encountered in different genera of Gram-positive bacteria (29 -31).
In this study we elucidate the detailed characteristics of the up to date unique Avicel-binding protein from S. reticuli. The investigations as to the membrane anchoring and complex formation support the conclusions that AbpS connects the Streptomyces cytoplasm with the extracellular environment and functions as a cellulose-receptor.

MATERIALS AND METHODS
Bacterial Strains, Plasmids, Transformations, and Cultivation-The wild-type strain S. reticuli Tü 45 described by Wachinger et al. (32) was obtained from H. Zä hner, Tü bingen, Germany. It was cultivated in pH-stable medium (MM3) supplemented with a carbon source (1% w/v), as outlined previously (33). After protoplasting, S. lividans 66 mycelia (provided by D.A. Hopwood, John Innes Institute, Norwich, UK) was transformed with bifunctional Streptomyces-Escherichia coli vector pWHM3 (34) or its derivatives containing truncated abpS genes by standard protocols (35). pUS1, a pUC18 derivative containing a 3.2-kb genomic SalI DNA fragment from S. reticuli, on which the complete abpS gene is located, was described earlier (17,18). The DNA sequence of abpS is available in the EMBL data bank under the accession number Z97071. The E. coli vector pET21a (Novagen, Madison, WI) was used as cloning vector for truncated forms of the abpS gene. The derivatives of pET21a containing the complete abpS gene or truncated genes with 5Ј end deletions (resulting in N-terminally shortened AbpS-proteins with molecular weights of 32.3 kDa, 29.1 kDa, and 23.5 kDa) were described earlier (17). E. coli DH5␣ or the chloramphenicol resistant E. coli BL21 (pLysS) (Novagen) was transformed with plasmid DNA using the CaCl 2 method (36).
Polymerase Chain Reactions and Construction of Plasmids-The PCR reaction mixture contained different primer combinations: 10 ng of pUS1, 0.2 nM of each of the dNTPs (dATP, dCTP, dGTP, and dTTP), 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 20 mM Tris-HCl (pH 8.8), 2 mM MgSO 4 , and 0.1% Triton X-100, in a final volume of 30 l. To reduce misreading, the Vent ® DNA polymerase, which has a 3Ј35Ј proof reading exonuclease activity, was used instead of the Taq polymerase.
The primer P 1 CAGGAACCATATGAGCGACAC was utilized to replace the ATG start codon of abpS with an NdeI-site. The primers P 2 (GAGCATCTCGTCGACGTTCGTCAG) and P 3 (TGCGCGGTGTCGAC-CGTC TGGCGG) were employed to introduce SalI sites at different locations at the 3Ј end of the abpS gene. The following cycling conditions were found to be optimal for the primer combinations P 1 and P 2 or P 1 and P 3 : 95°C, 90 s; 55°C, 60 s; and 72°C, 60 s. After thirty cycles, the PCR products were purified (Qia quick spin PCR purification kit, Qiagen, Hilden, Germany) according to the supplier's instructions. The NdeI-and SalI-digested PCR products were ligated in-frame into the polycloning site of pET21a (linearized with NdeI and XhoI), and transformed in E. coli BL21 (pLysS).
Internal deletions of the abpS gene were performed by "inverse" PCR. The pET21a derivative comprising the complete abpS gene served as template. The primers were designed in such a way that the generated PCR products consisted of the vector-sequence and flanking abpS fragments. The space between the primers represented the desired deletion. For self-ligation of the PCR products, in-frame SstI sites were incorporated in the primer sequences. For amplification of DNA with the primers P intA (GGCCTTGAGCTCGCCCGCGTA) and P intB (GGCGGACGA-GCTCTTCGAG GAGAG) or P intC (GCCGAGGAGCTCCGCCTGGAGG-C), the following cycle conditions were used: 95°C, 90 s; 50°C, 90 s; and 72°C, 450 s. After ten cycles the products were purified (see above) and digested with SstI, self-ligated, and transformed in E. coli BL21 (pLysS).
For cloning of abpS genes prolonged with the codons for 6 histidines in S. lividans as host PCR was performed under the above described conditions with primer P Xho (CTGCACCTCGAGACGCAG) and P His6 (AAATGATCAGTGGTGGTGGTGGTGGTGGGAGCCCCGGGACTGC-TGCGCCGGGAC) or P ⌬CHis6 (AAATGATCAGTGGTGGTGGTGGTGG-TGGGAGCCCCGG ACGTTCGTCAGCTGGGC) and as template pUS1 (containing the abpS gene from S. reticuli) was used (see above). After purification and digestion with XhoI and BclI, the resulting DNA fragment was used to substitute the abpS 3Ј end in pUS1, cleaved before with XhoI and BamHI. Consequently, the corresponding plasmids contained genes encoding a full-length AbpS or an AbpS protein without the hydrophobic segment both prolonged C-terminal with RGSH 6 . Subsequently, the genes were transferred into the bifuncional Streptomyces-E. coli vector pWHM3 and named pWA1 and pWA⌬C, respectively. Both plasmids were propagated in E. coli DH5␣ and after isolation transformed into S. lividans 66 protoplasts. The presence of the pWHM3-based constructs was guaranteed by selection with thiostrepton (25 g/ml).
General DNA Techniques-Application of restriction enzymes and the T4-ligase were carried out using the standard procedures (36). DNA sequencing was performed with the help of the T7 sequencing kit and Cy5-labeled standard primers (Amersham Biosciences).
Purification of Truncated His-tagged AbpS Proteins from E. coli or S. lividans-The E. coli BL21 (pLysS) transformants containing the designed constructs (see above) were grown at 37°C in SOC medium (20 g of Bacto-Trypton, 5 g of yeast extract, 0.5 g of NaCl, 0.18 g of KCl, and 20 ml of 1 M glucose/liter, which had been supplemented with chloramphenicol (34 g/ml) and ampicillin (100 g/ml)). Isopropyl-␤-Dthiogalactopyranose (final concentration 1 mM) was added when A 600 had reached 0. S. lividans transformants were grown in pH-stable medium (MM3) supplemented with 1% glucose for 48 h. After harvesting the mycelia by centrifugation, they were disrupted and the His 6 -tagged proteins were isolated as described above with or without urea-containing buffers.
Isolation of Murein-associated AbpS from S. reticuli-Anti-AbpS antibodies had been gained previously (18). The IgGs were purified using a 1 ml HiTrap protein G column (Amersham Biosciences) according to the supplier's manual. Subsequently, the purified IgGs were coupled to CNBr-activated Sepharose 4 Fast Flow (Amersham Biosciences) as described in the corresponding manual.
The murein layer was isolated from S. reticuli hyphae grown in 1 liter of minimal media as described earlier (18). Murein was treated with buffer (25 mM Tris-HCl (pH 8) and 20 mM EDTA) comprising 5 mg of lysozyme (Roche Applied Science) per ml and incubated at 37°C for 3 h. After centrifugation at 14,000 ϫ g for 30 min, the AbpS-containing supernatant was mixed with immobilized (see above) anti-AbpS antibodies. After incubation for 3-5 h under gentle stirring, the Sepharose was washed 3 times with 50 ml PBS (80 g of NaCl, 2 g of KCl, 2 g of KH 2 PO 4 , 11.5 g of Na 2 HPO 4 per liter (pH 7)). The proteins were released from the immobilized IgGs by applying 50 mM phosphate-citrate buffer (pH 2.7). For renaturation the pH was adjusted to 7, and purity and quantity were determined by SDS-PAGE followed by staining with Coomassie Brilliant Blue.
Separation of Cell Compartments-S. lividans mycelia were harvested by centrifugation, washed 3 times with 10.3% sucrose, and murein was hydrolyzed by incubation with 3 mg of lysozyme ml Ϫ1 mg Ϫ1 wet weight of mycelia (35). After filtration the protoplasts were resuspended in water supplemented with 5 mM EDTA to destroy them osmotically. The membranes and associated proteins were separated from soluble proteins by ultracentrifugation (250,000 ϫ g for 45 min).
SDS Gel Electrophoresis and Western Blotting-SDS-PAGE was performed with 10% polyacrylamide gels in the presence of 0.1% SDS (37). If desired, proteins were transferred onto nylon membranes, which were incubated in PBS containing the primary antiserum (1:100,000 dilution), the anti-His 5 antibodies, or the anti-RGSHis 6 antibodies (both 1:10,000 dilution) (Qiagen). After three washes, the blot was incubated with alkaline phosphatase conjugated with the AffinyPure F(abЈ) 2 fragments of goat anti-rabbit IgG or mouse anti-rabbit IgG (Dianova, Hamburg, Germany). Color development was performed as described by West et al. (38).
Urea-gradient PAGE-The lower polyacrylamide-gel (in a concentration of 6% and without SDS) was poured sidelong into the glass plate in 4 successive stages, containing 6, 3, 1.25 M, or no urea. After polymerization, the standard upper gel was overlaid so that one wide lane was formed. The proteins were loaded with a native loading buffer (36), and the electrophoresis was done at 15 mA to avoid heating above 25°C. Finally the proteins were stained with Coomassie Brilliant Blue.
Cellulose-binding Assay-Proteins were incubated with Avicel (15 mg/ml) in 50 mM potassium phosphate buffer (pH 7) for 30 min. Avicel recovered by centrifugation was washed 3 times with 50 mM potassium phosphate buffer containing 1 M NaCl. Avicel-bound proteins were released by heating in SDS sample buffer for 5 min at 100°C and subjected to SDS-PAGE. Staining of proteins was done with Coomassie Brilliant Blue.
Detection of Protein-Protein Interactions-Total proteins or purified AbpS isolated from S. reticuli were denatured by SDS, separated on an SDS-PAGE, and subsequently transferred onto a polyvinylidene difluoride membrane (because of the cellulose-binding capability of AbpS, nitrocellulose membrane was avoided). The membrane was treated for 1 h with PBS (see above) containing 1% bovine serum albumin. The His 6 -tagged fusion proteins (which were to be tested for the interaction of the immobilized AbpS) were incubated with 8 M urea in small volumes for 1 h. Rapid renaturation was performed by diluting this solution 1:100 in PBS, which contained a piece of the nylon membrane with immobilized AbpS from S. reticuli. After incubation at 25°C the membrane was washed 3 times with 50 ml of PBS, and the fusion proteins that were trapped by the immobilized AbpS were detected immunologically by applying antibodies specific to a stretch of 5 histidines (for further details see "SDS Gel Electrophoresis and Western Blotting").
Cross-linking of AbpS Complexes-Because of the absence of cysteine residues in AbpS the reactive bifunctional cross-linker dithiobis(succinimidyl propionate) (DSP), which is cleavable by reducing agents, was chosen to interconnect subunits of protein complexes. For this purpose the concentration of each of the His 6 fusion proteins (35.7, 27.1⌬C, 23.5⌬N, or 31.4⌬I) was adjusted to 10 mM. 3 l (corresponds to 30 nmol) were diluted in 50 l of buffer (25 mM Tris-HCl (pH 7)). The addition of 50 mM NaCl was found to inhibit ionic protein interaction, which increases the cross-linking specificity. To avoid intermolecular crosslinkages the molar ratio between DSP and denaturated proteins (0.5% SDS) was adjusted so that no multimers were found. The molar ratio (DSP/protein) of 3:1 was found to be optimal. Therefore the crosslinking reaction was started by adding 3 l of 10 mM solution of DSP (dissolved in Me 2 SO). After 10 min at 30°C, the proteins were denaturated by SDS at 95°C and analyzed by PAGE. 0.6 g of AbpS isolated from S. reticuli were incubated with 10 mg Avicel in a final volume of 100 l for 1 h. After 3 washes, 50 l of 25 mM Tris-HCl (pH 7) was added to the Avicel, and subsequently the bound proteins were cross-linked as described above. As control, the same amount of unbound AbpS was used.
Circular Dichroic Spectroscopiccal Analysis-To assess the structure of AbpS or its His 6 fusion variants (35.7, 27.1⌬C, 23.5, and 31.4⌬I) the concentrations of each of the proteins were adjusted to 0.2 mg ml Ϫ1 . The spectra (190 -260 nm) were recorded at 25°C using a Jasco 600 spectrophotometer in a 0.1-mm path length cell. Data were recorded 5 times, and an average value was determined. The observed ellipticity [] (degrees) is converted in mean residues ellipticity by: [] mrw,l ϭ MRW/10 dc, where MRW is obtained by dividing the molecular mass by nϪ1 (n ϭ number of amino acids in the protein), d is the path length (cm) and c is the protein concentration (g/ml). The calculated ellipticity was further converted to the difference of the molar absorbance ([] mrw ϭ 3298 ⌬⑀). Subsequently the data were analyzed by CD Spectroscopy Deconvolution program CDNN 2.1 (bioinformatik. biochemtech.uni-halle.de/).
Reconstitution of AbpS-Soybean Azolectins (L-␣-phosphatidylcholine solution obtained from Sigma) were washed twice with acetone in the presence of butylated hydroxytoluene and once with ether/vitamine E, dissolved in chloroform, and dried under a nitrogen stream. The preparation of liposomes was done according to Jung et al. (51). After solubilization of the liposomes by addition of 3% octyl ␤-D-glucopyranoside, the His 6 -tagged fusion proteins isolated from S. lividans were added and sonicated for 1 s and repeated 9 times. After removing the detergent by Bio-Beads SM-2 (Bio-Rad) the proteoliposomes were washed and concentrated by centrifugation at 250,000 ϫ g.

RESULTS AND DISCUSSION
Creation of Truncated AbpS Variants-A PCR-based approach was followed to produce abpS genes with deletion at the 3Ј end (31.0 ⌬C and 27.1 ⌬C) to complement a series of modified abpS genes that have shortened 5Ј ends (18) (Fig. 1). After cloning of the genes the encoded proteins were predicted to be tagged with a valine, a glutamine, and six histidines at the C terminus. Internal deletions within the abpS gene were generated by inverse PCR with the pET21a derivative containing the complete abpS gene (resulted in the deduced proteins 31.4 ⌬I or 25.6 ⌬I). Each of the plasmids was introduced into E. coli BL21 (pLysC). Sequence analysis of the plasmids revealed that the designed reading frames were preserved at the religated sites and that no base exchange had occurred.
Each E. coli BL21 (pLysS) transformant harboring one of the different constructs was grown to the logarithmic phase, induced with varying concentrations of isopropyl-␤-D-thiogalactopyranose at different temperatures. Independently of the used conditions, each type of the fusion protein was found dominantly within insoluble inclusion bodies (more than 90%). Therefore the proteins were bound to Ni 2ϩ -NTA in the presence of 8 M urea. Removing urea stepwise (see "Materials and Methods"), the proteins were subsequently released by imidazol, and each type was found to have the predicated molecular weight (see Fig. 1D).
Identification of ␣-Helical Structures-To analyze whether the renatured truncated AbpS forms (isolated from E. coli) kept structural characteristics of the S. reticuli wild-type AbpS a circular dichroism spectrum of each of the proteins should be recorded. As requirement a purification method of AbpS from S. reticuli has to be established first. As reported previously (18) AbpS is covalently linked to the peptidoglucan layer of S. reticuli. Therefore, its murein was isolated as previously described (18). Subsequently, AbpS was released by the action of lysozyme and immobilized by Sepharose-coupled anti-AbpS antibodies. After elution, AbpS was obtained in a good degree of purity (Fig. 1C); however, only ϳ5 g AbpS can be gained from 10 g mycelia (wet weight), grown in l liter culture.
Additionally the 35.7 His-tagged full-length AbpS fusion protein, the 23.5⌬N, 25.⌬I, and 31.⌬C forms were purified from E. coli as described (see "Material and Methods"). The concentration of each protein was adjusted to 0.2 mg/ml, and its CD spectrum was determined ( Fig. 2A). Between 205 and 260 nm the spectra of all proteins were close to identical. Below 205 nm AbpS, isolated from its natural host (S. reticuli), showed a slightly higher difference in the absorption (⌬⑀) value. Therewith, the deletion of different regions of the proteins did not significantly alter the overall structure of the protein. The competency for assembly of the secondary structure seems to reside in the composition of the protein sequence itself. The structural integrity is also reflected by the evaluated occurrence of 5 different protein-folding motifs (helix, anti-and parallel ␤-sheets, ␤-turns, or random coils). Their quantification as deduced from the CD spectra revealed that wild-type AbpS and its truncated forms consist predominantly of ␣-helical structures (ranging between 93 and 98.2%), and other structural motives are under-represented. This finding is in good agreement with a computer-supported prediction of the secondary structural elements within AbpS (Fig. 1B, scheme), comprising a large, centrally located ␣-helix flanked by two shorter helices, whereas the C-terminal helix is built by hydrophobic amino acids.

Analyzing the Function of the C-terminal Hydrophobic
Stretch of AbpS-In vivo AbpS was found to be associated with the membrane of the protoplasts, generated by removing the murein layer from the hyphae of S. reticuli (18). The C-terminally located hydrophobic helix was thus suspected to anchor AbpS to the membrane. To analyze this in more detail comparative in vivo and in vitro studies of AbpS with and without the hydrophobic segment were designed.
As a first requirement AbpS and the designed C-terminally truncated variant have to be synthesized in a Streptomyces host to guarantee native conditions, for example allowing the proper protein folding or membrane anchoring. Because there were no stable plasmids available for S. reticuli, the genetically best studied S. lividans strain was chosen as host. A disadvantage of S. lividans is the presence of an abpS homologue located within its chromosome (17). AbpS-negative S. reticuli or S. lividans mutants or AbpS-negative Streptomyces wild-type strains were not available.
Vector constructs (based on pWHM3) were designed having either the complete abpS gene (pWA1) or the 5Ј end deleted abpS gene (pWA⌬C) (encoding AbpS without the hydrophobic helix). They were kept under the transcriptional control of the upstream region of the S. reticuli abpS gene. In total, proteins of the transformants S. lividans (pWA1) or S. lividans (pWA⌬C) each prolonged with a His 6 tag (named AbpSHis or ⌬CAbpSHis, respectively) could be found in addition to the endogenous AbpS homologue (see S. lividans (pWHM3) as control) with the help of anti-AbpS antibodies. Due to the similar molecular weight of the endogenous S. lividans, AbpS, and the plasmid-encoded His-tagged fusion protein, they could not be separated by PAGE. Instead, the use of anti-RGSH 6 antibodies allowed the specific detection of the modified proteins (Fig. 3A). AbpSHis or AbpSHis could be isolated in high purity with the help of Ni-NTA under denaturating conditions (Fig. 3D, right  panel). After renaturation they were reconstituted into liposomes. Only AbpSHis was found to be integrated in the vesicles. Deletion of the hydrophobic part within ⌬CAbpSHis led to inhibition of its integration into proteoliposomes (Fig. 3B). To study the membrane integration within the original host in more detail, S. reticuli membranes were isolated and the release of AbpS was tested. Only in the presence of SDS (1%) or Triton X-100 (1%) but not with EDTA, urea (up to 1 M), nor water, AbpS was unhinged from the membranes (data not shown). Therefore a membrane-association of AbpS by ionic forces or a bedding of AbpS on the membrane-surface by hydrophobic interactions could be excluded.
Summarizing the data, the C-terminal helix was clearly identified as a transmembrane spanning segment. With simultaneous consideration that the N terminus of AbpS was previously found to protrude from the cell surface (18) and a transmembrane spanning helix is present, the C-terminal end of AbpS has to extend into the cytoplasm of the S. reticuli hyphae. Thus AbpS connects the interior of the mycelia with the extracellular space, where it binds insoluble cellulose. Such organization is reminiscent of the assembly of diverse classes of receptor proteins (39), often involved in signal transduction cascades.
To study the effect of the deletion of the transmembrane segment in vivo (in S. lividans mycelia), the overproduction of AbpSHis or ⌬CAbpSHis in S. lividans was prevented by the cloning strategy (using the wild-type regulatory elements). As a result, AbpS and its derivatives were found in approximately the same concentrations (Fig. 3A) within the total proteins isolated from S. lividans containing pWHM3, pWA1, or pWA⌬C, respectively.
It was expected that at least the C-terminally deleted protein ⌬CAbpSHis would be released by digestion of the murein. But surprisingly the distribution of the endogenous S. lividans AbpS and ⌬CAbpSHis in the different cell compartments (extra cellular, murein-associated, membrane-integrated, or cytoplasmic) was found to be equal (Fig. 3C). A possible cause for this in vivo effect, which contrasted the in vitro data, appeared to be the fact that the endogenous membrane-integrated S. lividans- FIG. 4. Identification of the interacting domains. A, proteinprotein interactions. Immobilized wild-type AbpS was incubated with the indicated truncated His 6 -labeled proteins. Formation of the protein complexes was tested immunologically by anti-His 5 antibodies (lanes 1-9) and quantified by densitometrical evaluation. The values are given as percentages in reference to the protein amounts shown in lane 1. Immobilized AbpS incubated with none of the soluble proteins (lanes 9 and 10) were treated with the indicated antibodies and served as controls. B, AbpS and bovine serum albumin as control were dotted onto a polyvinylidene difluoride membrane and incubated with (lanes 1 and 2 and 1Ј and 2Ј) or without (lane 3 and 3Ј) the His 6 -tagged full-length AbpS form. After elution of the proteins from the membrane they were loaded onto SDS-PAGE, blotted and identified using the AbpS-(lane 1) or His 5 -specific antibodies. C, protein complexes consisting of immobilized AbpS and the His 6 -tagged full-length AbpS form were incubated under the indicated conditions, and subsequently the retaining amount of the His 6 -tagged protein was determined.
AbpS molecules form complexes with ⌬CAbpSHis (plasmidencoded) and impede, therefore, the release of the C-terminally deleted protein. The expected complex formation could be satisfactorily shown by co-purification of the endogenous S. lividans-AbpS (unable to bind to Ni-NTA) together with the His 6tagged AbpS derivatives (AbpSHis and ⌬CAbpSHis) by Ni-NTA-based affinity chromatography (Fig. 3D).
Identification of AbpS Domains Required for Intermolecular Interaction-To study the deduced interaction among AbpS molecules in more detail, total proteins (including the S. reticuli wild-type AbpS) from S. reticuli were denatured, separated by SDS-PAGE, and then immobilized on a nylon membrane. Subsequently identical amounts of the truncated and denatured AbpS forms were added to the membrane-immobilized AbpS in low salt buffer. The degree of intermolecular interaction was determined by immunological quantification of the membrane-retaining His 6 fusion proteins (Fig. 4A). Independently of the presence (35.7 kDa form) or absence (31.0⌬C and 27.1⌬C) of the C-terminal part, each of the proteins strongly interacted with the immobilized S. reticuli-AbpS. However, continued deletions of the N-terminal part (32.3⌬N, 29.1⌬N, and 23.5⌬N) led to an increased loss of the binding ability. Therefore it can be concluded that the portion of the protein (including amino acids 60 -110) situated at the beginning of the central ␣-helix plays an essential role in the complex forming process. The remaining binding level of the 23.5⌬N protein was attributed to a second domain identified by analyzing the characteristics of additional truncated forms. They extend from the middle (31.4⌬I) to the right side (25.6⌬I) of the centrally located ␣-helix (including amino acids 161-212). The 25.6⌬I form was found to retain only 15% of the maximal binding ability, which corresponded closely to that of the 31.4⌬I form (Fig. 4A).
As outlined above on the nylon-membranes (on which the separated cellular proteins from S. reticuli were immobilized), each of the added truncated AbpS forms were found to interact only with AbpS, demonstrating a high specificity of the proteinprotein interaction. To verify this in more detail, the formed complexes were precisely cut out of the membrane, which was overlaid with the full-length His 6 -tagged fusion protein. Immunological identification clearly demonstrated the sole participation of AbpS and its His 6 -tagged form (35.7 kDa) in the intermolecular interaction (Fig. 4B). Based on these data, two domains responsible for AbpS-AbpS interaction were identified, matching regions predicted to contain only ␣-helical structures (Fig. 4A).
Conditions for Intermolecular Interaction-Mixtures of membrane-immobilized S. reticuli-AbpS and its His 6 -tagged full-length 35.7 kDa form were incubated under different conditions (20 -90°C or at pH 2.7) or treated with different amounts of chaotropic urea, leading to protein-denaturation protein (0. 25-8 M urea) or increasing the ionic strength (1 M NaCl) (Fig. 4C). The protein complex was found to be stable in the presence of up to 1 M urea or 1 M NaCl. Because these conditions are unfavorable for ionic bonding, oligomerization based on the ionic strength could be excluded. At very low pH (2.7), high concentrations of chaotropic salt (commencing at 1 M) or high temperatures (higher than 50°C) the protein-protein interaction was usually inhibited strongly. The above-described conditions for resolving the ability for complex formation exactly match those that inhibited the binding of AbpS to cellulose (17). These findings suggest that the S. reticuli AbpS interacts with cellulose only in an oligomerized form.
Characterization of Oligomerization-Using standard methods (sucrose gradients or gel filtration with various globular reference proteins) it was found that the molecular weight determined for AbpS neither correlated to that one for the AbpS monomer nor to any expected oligomer. This was attributed to the fact that the secondary structure of AbpS was found to be dominantly ␣-helical, governing an elongated shape of the protein (Figs. 1B and 2). Because correspondingly shaped reference proteins are missing, the mobilities of the various truncated AbpS forms and the S. reticuli-AbpS were comparatively studied in native PAA gels in the presence of increasing concentrations of chaotropic urea (0, 1.25, 3, and 5 M) (Fig. 5). The S. reticuli AbpS, the 35.7-kDa full-length AbpS-His 6 fusion protein, and the 31.4⌬I protein (with a short internal deletion) were found to migrate into the gel only in the presence of 6 M urea. When applying to 3 M urea, no protein or only traces could be recovered. This effect was also ascertainable at low (5%) polyacrylamide concentrations. The 31.0⌬C truncated form possesses a deletion of 46 C-terminal amino acids, including the above-characterized hydrophobic membrane-integrated helix. The removal of consecutive 35 amino acids resulted in the 27.1⌬C truncated form. Both protein types were found to migrate into the gel containing up to 3 M urea. In comparison, the relative mobility increased under denaturing conditions (6 M urea). These data suggest that the homomultimeric protein complexes were formed by native conditions, whereas the presence of the hydrophobic segments provoked protein aggregation (Fig. 5).
Interestingly, the 23.5⌬N and 25.6⌬I proteins (both lacking one domain responsible for the intermolecular protein interaction) were monomeric under all conditions (0 -6 M urea). This finding is in agreement with their low protein-protein interaction ability (Fig. 6). In contrast, the 32.3⌬N as well as the 29.1⌬N form, both possessing an increasing portion of a deleted N terminus, migrated already in the presence of 1.25 M urea as denatured protein complexes; this decreasing stability of the protein complexes correlates with their intermediate ability to interact with the wild-type AbpS (see Fig. 4A).
Again this finding indicates a protein-protein interaction mechanism, which governs the formation of a stable complex. Together with the structural information of AbpS a formation of coiled-coil multimers appears likely (Fig. 7). In coiled-coil proteins the ␣-helices provides hydrophobic amino acids to one side of the helix. This hydrophobic side is able to interact with the corresponding side of the next subunit. Ionic amino acids, proximate to the hydrophobic ones, direct the formation of dimeric, tetrameric, or higher organized complexes. Typical coiled-coil proteins are the dimeric tropomyosins (40), trimeric or tetrameric leucinezippers (41), or the pentameric Comp (cartilage oligomeric matrix protein) whose oligomerization do-mains marked similarities with proposed models of the pentameric transmembrane ion channels in phospholamban and the acetylcholine receptor (42).
Interestingly some M-proteins of the group A streptococci, which belong to the MSCRAMMs (see Introduction), have been shown to be coiled-coil dimers, appearing as fibrils on the bacterial surface in the electron microscope (23,43,44). The arrangement of the hydrophobic amino acids is also reflected in the primary sequence of a protein. Numbering the amino acids in a coiled-coil forming ␣-helix by "abcdefg" amino acids "a" and "d" were found to be hydrophobic (41). Analyzing the deduced AbpS protein sequence according to this rule, large protein sections (including amino acids 20 to 246) were identified with high potential for coiled-coil interactions (Fig. 7).
Constitution of the AbpS Complexes-To determine the constitution of the AbpS complexes, cross-linking experiments with the unspecific linking agent DSP were performed. For this FIG. 6. Multimerization and cellulose-recognition. A, the indicated proteins in their soluble form as well as Avicel-bound AbpS were incubated with a cross-linking DSP as control in the presence of 0.5% SDS (lane 1) or absence of SDS. The proteins were separated with the help of a SDS-PAGE. After Coomassie staining, the molecular masses were determined by comparison with a protein marker and divided through the molecular mass of the monomer. The wholenumbered quotients are indicated. B, the indicated proteins were incubated with Avicel. The mixture was washed in the presence of 1 M NaCl to avoid unspecific bindings. The bound proteins were released from Avicel by heating, and after PAGE and Coomassie staining their regained amounts were compared in percent to the deployed amounts of each protein.
purpose (i) the concentration of the proteins, (ii) the molar ratio of the proteins and the cross-linker, and (iii) the concentration of NaCl (for inhibition of weak ionic protein interactions) were carefully optimized to prevent unspecific linkages (for details see "Material and Methods"). Under optimized conditions, the 35.7-kDa full-length His 6 -tagged AbpS, 32.3⌬N, 31.4⌬I, and 31.0⌬C proteins were found to form up to tetramers (Fig. 6A). In contrast the N-terminally deleted 29.1⌬N and the 23.5⌬N protein shaped only trimers (ϳ87 kDa) or dimers (ϳ 47 kDa), respectively. The cross-linking of the 25.6⌬I protein (with the extended internal deletion) resulted in the formation of ϳ75 kDa and ϳ50 kDa complexes, corresponding to trimers and dimers, respectively. The composition of the intermediate complexes with the apparent molecular mass between 75 and 50 kDa was not explored.
The cross-linking experiments were repeated with purified S. reticuli AbpS, which initially was bound to the crystalline Cellulose (Avicel) or kept in solution. The soluble AbpS as well as the Avicel-bound was found to tetramerize, showing that the complex organization is not altered by the interaction of AbpS with cellulose.
The ability to dimerize, or in general to oligomerize, is widespread among other types of carbohydrate-binding proteins. For example lectins specifically recognize diverse sugar structures and mediate a variety of biological processes, such as cell-cell and host-cell interaction (45). For some bulb lectins, which share some similarities in their three-dimensional structure and the ability to specifically bind mannose, the degree of oligomerization was found to modulate carbohydrate-binding specificity. The tetrameric lectins snowdrop (46), daffodil (47), or bluebell (48) bind a surface glycoprotein on human immunodeficiency virus, whereas the dimeric garlic lectin (49) does not.
Cellulose Recognition-Comparative analysis of AbpS and each of the truncated forms revealed that a reduced oligomerization (29.1 ⌬N, 23.5 ⌬N, and 25.6⌬I) is accompanied with a decreased capability for cellulose binding (Fig. 6B). This supports the finding that the conditions favorable for the intermolecular interactions of AbpS are identical to those that allow cellulose recognition (17). Dissociation of the complexes abolishes the cellulose-binding ability of AbpS. In contrast, the 32.3 ⌬N form with a moderate deletion in the N terminus (29 amino acids) is still able to shape tetramers, whereas the cellulosebinding capability is reduced 10 times. Therefore amino acids that are involved directly in the cellulose interaction process have to be located within the first 29 residues of AbpS. This mode of cellulose-recognition via AbpS is so far unique and differs considerably from cellulose-binding domains within cellulases (50). As shown by biochemical and crystallographical analysis, the interaction of these cellulose-binding modules is mediated by aromatic amino acids exposed on one side of the proteins. The distance between the aromatic amino acids correlates with those between the glucose units within the cellulose (50). Multimerization of such cellulose-binding domains is obviously not advantageous; on the contrary, it leads to an aggregation of the domains containing the active site for cellulose hydrolysis. In the case of AbpS the amino acid residues that are directly involved in cellulose recognition seem to be distributed within 4 molecules. Only multimerization arranges these amino acids in a topology required for optimal interactions. FIG. 7. Arrangment of hydrophobic amino acids. A, part of the deduced amino acid sequence of AbpS (from amino acid 20 to 246) is given in the one letter code. The arrangement of hydrophobic residues (AVLIPMFW, marked with gray boxes) according to the abcdefg rule (given in the second line) of coiled-coil proteins is indicated by continuous boxes.