Properties of a Cyclodextrin-specific, Unusual Porin from Klebsiella oxytoca *

The function of CymA, 1 of the 10 gene products involved in cyclodextrin uptake and metabolism by Klebsiella oxytoca, was characterized. CymA is essential for growth on cyclodextrins, but it can also complement the deficiency of alamB (maltoporin) mutant of Escherichia colifor growth on linear maltodextrins, indicating that both cyclic and linear oligosaccharides are accepted as substrates. CymA was overproduced in E. coli and purified to apparent homogeneity. CymA is a component of the outer membrane, is processed from a signal peptide-containing precursor, and possesses a high content of antiparallel β-sheet. Incorporation of CymA into lipid bilayers and conductance measurements revealed that it forms ion-permeable channels, which exhibit a substantial current noise. CymA-induced membrane conductance decreased considerably upon addition of α-cyclodextrin. Titration experiments allowed the calculation of a half-saturation constant,KS , of 28 μm for its binding to CymA. CymA assembled in vitro to two-dimensionally crystalline tubular membranes, which, on electron microscopy, are characterized by a p1-related two-sided plane group. The crystallographic unit cell contains four monomeric CymA molecules showing a central pore. The lattice parameters are a = 16.1 nm, b = 3.8 nm, γ = 93°. CymA does not form trimeric complexes in lipid membranes and shows no tendency to trimerize in solution. CymA thus is an atypical porin with novel properties specialized to transfer cyclodextrins across the outer membrane.

lized by the respective organisms as carbon and energy source. In the case of Klebsiella oxytoca, the utilization pathway has been studied in detail and it could be shown that the products of at least 10 genes, designated cymA to cymJ, are involved (3). Four of them are constituents of a periplasmic binding proteindependent uptake system (3), from which CymE is the binding protein (4), CymF and CymG are the integral membrane components, and CymD the ATPase. There is convincing evidence that ␣and ␤-CDs are taken up as intact entities (3,4) and that they are linearized in the cytoplasm by the product of the cymH gene that is a cyclodextrinase (5). CymE, CymF, CymG, and CymD are homologs of the components of the paradigmatic maltose uptake system, MalE, MalF, MalG, and MalK, respectively (3,6). The metabolism of ␥-CDs is dependent on the activity of the cyclodextrin-glucanotransferase (3), so ␥-CDs must be converted into ␣or ␤-CDs or linearized (7,8) 2 to be taken up.
How the bulky molecules of ␣-CD and ␤-CD traverse the outer membrane was still unknown. Because of their outer diameter of 1.37 (␣-CD) and 1.53 (␤-CD) nm and their physical properties, it was a priori improbable that they use the LamB channel for entry into the periplasm. In this report, we show that ␣-CDs cross the outer membrane of K. oxytoca and also of recombinant Escherichia coli cells via a specific porin that has been identified as the product of the cymA gene. We demonstrate that this porin also accepts linear maltodextrins as passenger molecules. The purification of CymA and its intriguing properties are reported.

EXPERIMENTAL PROCEDURES
Strains and Plasmids, Media, and Growth Conditions-Bacterial strains and plasmids are listed in Table I. Media and growth conditions were as described previously (3,4).
Standard Genetic Procedures-Standard genetic procedures were adopted from Miller (10) and Sambrook et al. (11). Enzymes for recombinant DNA techniques were used according to the recommendations of the manufacturers (Roche Molecular Biochemicals GmbH, Mannheim, Germany; Amersham Pharmacia Biotech, Freiburg, Germany; New England Biolabs, Schwalbach, Germany). The oligonucleotide primers used were synthesized by MWG (Ebersberg, Germany).
Construction of Plasmid pCYMA2-Plasmid pCYMA2 is a pSU2719 derivative (12) which carries a K. oxytoca DNA fragment starting 216 base pairs 5Ј of cymA and extending to base pair 175 of cymB. For construction, pSU2719 was restricted with SacI and XbaI, dephosphorylated, and ligated with the insert, which was prepared by analogous restriction of plasmid pCYMA (3).
Cloning of the lamB Gene of K. oxytoca M5a1-The lamB gene of K. oxytoca was cloned via PCR using phosphorylated oligonucleotides derived from the nucleotide sequence of the homologous gene from Klebsiella pneumoniae (13). Oligonucleotide LAMB3 is a 24-mer (5Ј-ATGAT-GATTACTCTGCGCAAACTT-3Ј) and starts with the ATG of the reading frame, whereas the 27-mer oligonucleotide LAMB4 (5Ј-TTAC-CACCACACTTCCATCTGGGCACC-3Ј) starts with the codon complementary to the stop codon of lamB. The resulting PCR fragment was inserted into SmaI-restricted pSU2719 vector. The resulting plasmids were pLAMB carrying the insert fused to the lac promoter of the vector and pILAMB carrying the insert in the opposite direction.
Construction of a K. oxytoca cymA Deletion Mutant-The K. oxytoca cymA deletion mutant was constructed as described in Pajatsch et al. (4) for the K. oxytoca cymE deletion strain. The plasmid used for introducing the deletion was pCYM⌬A (3).
Overproduction and Purification of CymA-For overproduction and purification of CymA, strain JM109 (14) was transformed with plasmid pCYMA. The transformants were grown aerobically at 37°C overnight to an A 600 of 1.2 to 1.7 in 6 ϫ 2.5 liters of LB medium (10), harvested by centrifugation, and stored at Ϫ80°C.
During purification, the protein concentration was assayed according to Whitaker and Granum (15), and that of the purified CymA according to Bradford (16). The CymA content of the fractions was analyzed by SDS-polyacrylamide gel electrophoresis (17) following the distribution of the 39-kDa band of the mature form of CymA.
Cells were suspended in 200 ml of buffer A (50 mM potassium phosphate, pH 7.5) containing 1 mM phenylmethylsulfonyl fluoride and broken by three passages through a French pressure cell at 16,000 p.s.i. The extract was clarified by centrifugation at 6,000 ϫ g for 45 min, resulting in the S 6 fraction. The membranes were pelleted by centrifugation at 100,000 ϫ g for 90 min, yielding the P 100 fraction.
For the solubilization of the cytoplasmic membrane components, the pellet was resuspended in 40 ml of buffer A containing 0.5% sarcosyl (Sigma, Deisenhofen, Germany). After stirring for 45 min at room temperature, the solubilized material was removed by centrifugation at 100,000 ϫ g for 90 min. The pellet was washed in buffer A, and the solubilization procedure was repeated resulting in fraction P CM . Remainders of sarcosyl were removed by washing the pellet twice in buffer A; the pellet was resuspended in buffer A containing 100 g/ml lysozyme and 3 mM NaN 3 and stirred overnight at 37°C.
After centrifugation at 100,000 ϫ g for 90 min, the outer membrane components were solubilized by resuspending the pellet in 40 ml of buffer B (50 mM potassium phosphate, pH 7.5, 5 mM EDTA, and 5% of octyl-polyoxyethylene (OPOE; Bachem Biochemica GmbH, Heidelberg, Germany); protein-detergent ratio 1:5). The solution was stirred for 1 h at room temperature and clarified by centrifugation at 100,000 ϫ g for 90 min.
The supernatant (fraction S OM ) was dialyzed against buffer C (10 mM potassium phosphate, pH 7.5, 0.6% OPOE) and loaded on a Q-Sepharose Fast Flow column (21-ml bed volume) that had been equilibrated in buffer C. The column was washed with 80 ml of buffer C, developed with a 200-ml linear gradient (0 -0.6 M NaCl), and finally washed with 40 ml of buffer C and buffer C containing 1 M NaCl. The flow rate was 1 ml/min. Fractions containing CymA were concentrated by ultrafiltration (Centricon 50 concentrators, Amicon) and applied to a Superdex 200 HiLoad gel filtration column (60 cm ϫ 16 mm) that had been equlibrated in buffer D (10 mM potassium phosphate, pH 7.5, 100 mM NaCl, 0.6% OPOE). The column was eluted with buffer D at a flow rate of 1 ml/min.
To remove most of the remaining impurities (mainly OmpC, OmpF, and OmpA), an ion-exchange chromatography through a Mono Q-HR 5/5 column (1-ml bed volume) was carried out. The column was equilibrated with buffer C, loaded with the sample, and rinsed with 25 ml of buffer C. It was developed with a 20-ml linear gradient (0 -0.6 M NaCl) and consecutively washed with 10 ml of buffer C and buffer C containing 1 M NaCl. The fractions containing highly enriched CymA were concentrated by ultrafiltration, and the gel filtration chromatography with the Superdex 200 HiLoad gel filtration column was repeated. The fractions containing the purified CymA were concentrated by ultrafiltration to a concentration of 1 mg/ml, supplied with 3 mM NaN 3 , and stored at 4°C.
Protein Sequencing of CymA-The N-terminal sequence of purified CymA was determined in an Applied Biosystems pulsed liquid-phase sequencer 477A equipped with an on-line phenylthiohydantoin-amino acid analyzer 120 A (Applied Biosystems).
Fourier Transform Infrared Spectroscopy-The secondary structure composition of CymA was estimated with spectra obtained by Fourier transform infrared spectroscopy using the attenuated total reflection technique. Spectra were measured in a Nicolet FTIR 740 spectrometer at 2 cm Ϫ1 resolution. Samples containing 30 -70 g of protein were dried onto Germanium crystals. Spectra were collected by adding 1024 scans for each determination and transformed without apodization for further analysis. Hydrogen-deuterium (H-D) exchange was performed on the Ge crystals in a particularly constructed chamber by flushing the sample with N 2 /D 2 O at 25°C. The exchange was monitored online by taking spectra (8 scans added) every 15 s to 5 min. After 30 min the exchange was apparently completed, and a final spectrum (1024 scans) was measured. The secondary composition was judged by comparing the CymA spectrum with spectra of porins and other reference proteins whose structures are known and by quantitative analysis of the amide I band as described in Ref. 18.
Two-dimensional Crystallization of CymA-Two-dimensional crystallization was performed using the dialysis method and apparatus described by Paul et al. (19). CymA was dissolved in 10 mM potassium phosphate buffer solution, pH 7.5, plus 250 mM NaCl, 0.6% OPOE, and 3 mM NaN 3 . The solution was supplemented with dimyristoylphosphatidylcholine (DMPC) solubilized in the same detergent such as to adjust protein-to-lipid ratios from 0.23 to 8.9 (w/w). The final volume was 200 l, the final protein concentration 0.67-0.95 mg/ml, and the final OPOE content between 0.48% for the highest and 1.92% for the lowest protein-to-lipid ratio. A sample without protein served as a control for membrane formation, a sample without DMPC as a test for the capability of CymA to form crystals without additional lipid. The dialysis buffer consisted of 20 mM HEPES, pH 7.5, plus 3 mM NaN 3 ; the dialysis membrane had an exclusion cutoff of 15-20 kDa. Dialysis was performed at 33 Ϯ 1°C for 72 h at a buffer flow rate of 0.12 ml/min per dialysis chamber (200 l). Membrane formation was monitored online by light diffraction (19), and the experiment was ended when the diffraction curve had approached its plateau for at least 24 h. The samples were inspected in the light microscope and investigated by electron microscopy afterward. All samples were stored at 4°C.
Electron Microscopy and Image Processing-Aliquots of 5 l were applied to carbon-coated copper grids made hydrophilic by means of glow discharge. The grids were washed with pure water to remove buffer salts prior to negative staining with 2% unbuffered uranyl acetate solution. Electron micrographs were taken in a Philips EM420 at nominal magnification ϫ36,000. Micrographs were digitized with an Eikonix 1412 camera applying a pixel size of 0.43 nm at the specimen level. Image processing was performed with a Silicon Graphics workstation using the SEMPER 6.4 system (20, 21) for analyzing and averaging two-dimensional crystalline membranous structures via the correlation averaging approach (22). Reference images for averaging were prepared by window filtering the Fourier transforms. Only the periodic information contained in the reflections systematically distributed on a lattice in Fourier space were selected. Subregions of the back-transformed images were used as references for correlation averaging (23). The resolution was assessed by the radial correlation function criterion (22). Lipid Bilayer Experiments-Black lipid bilayer membranes were formed as has been described previously (24) from a 1% solution of diphytanoyl phosphatidylcholine (Avanti Polar Lipids, Alabaster, AL) in n-decane. The instrumentation consisted of a Teflon chamber with two aqueous compartments connected by a small circular hole with a surface area of 0.3-0.5 mm 2 across which the membranes were formed. The aqueous salt solutions (Merck, Darmstadt, Germany) were used unbuffered and had a pH around 6, if not indicated otherwise. CymA was added from the concentrated stock solution to the aqueous phase bathing a membrane in the black state. The temperature was kept at 20°C throughout. The membrane current was measured with a pair of Ag/AgCl electrodes with salt bridges switched in series with a voltage source and a current amplifier. The amplified signal was monitored with a storage oscilloscope, and the reconstitution of channels in the black lipid membrane was recorded with a strip chart recorder.
Calculation of the Stability Constant for Carbohydrate Binding to CymA-The mechanism of solute transport through specific porins can be explained by a simple occupancy of a binding site in a two-barrier, one-site channel similar to the situation in certain single-file ion channels (25,26). We used a similar approach here for the explanation of the binding of carbohydrates to the CymA channel. For this, it is assumed that CymA is a carbohydrate-specific single file channel similar to those formed by LamB of enteric bacteria (specific for maltose and maltooligosaccharides; Refs. 27 and 28), ScrY (specific for sucrose; Refs. 29 and 30), and Tsx (specific for nucleosides; Ref. 31). This means that it is open when no carbohydrate is bound, and closed when it is occupied by a carbohydrate. The probability, p, for the occupancy of the binding site by a carbohydrate is given by the Langmuir adsorption isotherm. The channel does not conduct ions with the probability, p.
K is the stability constant of the binding between substrate and the binding site (half-saturation constant K S ), and c is its concentration in the aqueous phase on both sides of the channel. The probability that the binding site is free and conducts ions is given by 1 Ϫ p.
This means that the membrane conductance, G(c), as a function of the substrate concentration, c, in the aqueous phase is given by the following equation.
G max is the maximum membrane conductance without the substrate, i.e. when the binding site is free. According to the assumptions, only one carbohydrate can bind to the binding site at a given time and no ion can pass the channel when the binding site is occupied. This means that a carbohydrate can enter the channel only when the binding site is free. The model also assumes a symmetric channel in view of the location of the binding site inside the channel and its access from both sides.

Identification of CymA as the Putative Cyclodextrin Porin-
The sequence analysis of the products of the cym genes had indicated that three of them (CymA, CymC, and CymE) contain signal peptides (3). CymE has been identified as the cyclodextrin-binding protein in the periplasm (4), and CymC was shown to be dispensable for the utilization of cyclodextrins as sole carbon source (3). The only candidate left for the porin function, therefore, was CymA. Previous evidence for such a role was that an E. coli strain transformed with the 10 cym genes from K. oxytoca was able to grow at the expense of CDs but not when the cymA gene was deleted. In addition, the lamB gene from E. coli could not complement this deficiency (3).
We also introduced an in-frame deletion into the chromosomal cymA gene from K. oxytoca (yielding strain CYMA). The mutant was not able to grow on CDs as sole carbon and energy source, but it was still able to grow on maltohexaose, indicating that lamB was functionally expressed. This was confirmed by cloning the lamB gene from K. oxytoca via PCR and by demonstrating that it could complement the lamB deficiency of an E. coli mutant to grow on maltohexaose (data not shown).
It was further analyzed whether the presence of CymA is required solely for growth on CDs or whether it can also support utilization of maltodextrins. A lamB mutant from E. coli (strain GM7) was, therefore, transformed with plasmid pCYMA2 that expresses cymA. Table II shows that the presence of the cymA gene confers the capability to the lamB mutant to grow on maltodextrins with four or more glucose units. Growth on glucose, maltose, or maltotriose also occurred in the absence of the cymA gene since these substrates are taken up via alternative porins (32,33).
Overproduction and Purification of CymA-CymA was overproduced in E. coli JM109 transformed with plasmid pCYMA, and the protein was purified as detailed under "Experimental Procedures." Fig. 1 displays the course of purification as analyzed by SDS-polyacrylamide electrophoresis of the pooled fractions after each purification step. A symmetrical peak of highly enriched CymA was eluted from the last column; the yield was 5.5 mg of protein out of 15 liters of culture.
Secondary Structure Properties of CymA-The N-terminal amino acid sequence of the purified protein was determined to be ASDQR, which confirms that the signal sequence has been cleaved off ( Fig. 2A). An alignment of the sequence with those of porins with established structure and function did not reveal any conspicuous similarities. Inspection of the primary sequence using a previously proposed formalism (36) suggested that it contained possibilities for amphipathic ␤-strands, which is typical for outer membrane porins (37) (data not shown). A clear ␤-sheet structure is also visible at the C terminus (Fig.  2B), which contains the ultimate phenylalanine residue, which has been shown to be important for insertion of porins into the outer membrane (38). The consensus of the sequence -X-h-X-h-X-h-X-Y-X-F matches that of CymA.
To determine the secondary structure composition of CymA by experimental means, infrared spectra of the native and deuterated protein (H-D exchange by exposure to D 2 O vapor) were measured. The shape of the amide I band (1700 -1600 cm Ϫ1 ) of the CymA spectrum (Fig. 3) is very typical for outer membrane proteins, and porins in particular, which consist of 55-60% antiparallel ␤-sheet forming a ␤-barrel (36). The prominent peak at 1628 cm Ϫ1 is indicative for a high content of ␤-strands (60 Ϯ 5% of the total amino acids, analysis not shown) and the shoulder at 1693 cm Ϫ1 for ␤-turns, i.e. both together for antiparallel ␤-sheet. The loss of the underlying absorption band around 1655 cm Ϫ1 upon deuteration and its shift to wave numbers below 1650 cm Ϫ1 are indicative for loops and folds without preferred secondary structure characteristics (18). The ␣-helix content is very small and is assessed to be less than 10%. The spectra of both solubilized protein and CymA reconstituted into DMPC membranes did not differ. The H-Dexchange was extremely rapid as judged by spectra taken at 10 s, 60 s, 120 s, and higher up to 30 min after turning on the N 2 /D 2 O stream (data not shown). In fact, the final spectrum of CymA embedded in lipid membranes (Fig. 3) is almost indistinguishable from the very first one in the time series, meaning that the loops were easily accessible to D 2 O and were not shielded by membrane lipids or buried inside the protein.

CymA Does Not Form Trimers in Solution or in Lipid
Membranes-A typical property of bacterial porins is the trimeric structure of high thermal stability (39). Classical porins such as OmpF from E. coli or Omp32 from Comamonas acidovorans (40) at least partially retain their trimeric structure after incubation in SDS at 30°C and are only denatured upon heating in SDS buffer at temperatures above 80°C (41). As illustrated by Fig. 4, CymA is migrating as a monomer in SDS gels after incubation at 30°C for 10 min as well as upon heating at boiling temperature. Although the gel was overloaded to render even traces of oligomers visible, and although the porin Omp32 remained in oligomeric form under these conditions, there was no indication for the existence of corresponding CymA complexes.
To clarify whether CymA forms trimeric complexes in lipid membranes as bacterial porins used to do, we reconstituted CymA in DMPC membranes by means of the dialysis approach (19). CymA formed tubules in presence of DMPC at molar lipid-to-protein ratios between 10:1 and 285:1. If considerably less lipid was added, CymA precipitated upon dialysis. The flattened tubules had dimensions of about 3 m in length and 30 -80 nm in width (Fig. 5). They showed faint striations perpendicular to the longitudinal axis, indicating a regular arrangement of molecules. While long tubules occurred at high lipid-to-protein ratios without exception, the tubular mem-branes tended to apparently "unwind" in form of ribbons if the lipid concentration was low (Fig. 5). These membrane ribbons showed strong reflections in reciprocal space and were selected for image processing while the tubules showed superimposed sets of reflections that were more difficult to analyze. The diffraction spots refer to a lattice in real space with the parameters a ϭ 16.1 nm, b ϭ 3.8 nm, and ␥ ϭ 93° (Fig. 5). The angle ␥ was close to 90°in most of the membrane sheets analyzed. This observation and the fact that the odd-numbered reflections with the indices h0 and 0k in the power spectrum were usually very weak or completely missing, are compatible with the properties of the two-sided plane group c12 and a unit cell containing four CymA molecules or asymmetric structural units, respectively (23). The averaged unit cells show stainfilled channels or cavities that are not arranged in trimers, being typical for other bacterial porins (42,43), but occur in pairs representing molecules in upside-down orientation with respect to each other (central pair in the unit cell shown in Fig.  5). The two-dimensional crystal type and the structure of the projected molecule are consistent with the characteristics of a monomeric protein not forming symmetric complexes in the membrane.
The diameter of the CymA molecule is 3.8 -4.0 nm according to the lattice parameters, the heavily stain-filled pore crosssection about 1.2-1.4 nm as assessed from the averages. The latter value is close to the resolution limit of the average, being 1.4 nm, and should therefore be taken as an estimate. If CymA has a cylindrical shape, the pore diameter is constant along the channel, and the height of the molecule is 5.0 -5.5 nm, corresponding to the dimensions of the E. coli porins OmpF, PhoE, and maltoporin (determined with the atomic structures available in the Brookhaven Protein Data Bank), the calculated mass of one structural unit in Fig. 5 is about 42-46 kDa (23). This estimate is in reasonable accordance with the molecular mass of one CymA polypeptide, i.e. 38 kDa, and it confirms that the repetitive units in the averaged image indeed represent CymA monomers.
CymA Is an Outer Membrane Channel Permeable for Ions-When CymA was added in small concentration (10 -100 ng/ml) to the aqueous phase on one or both sides of a black lipid bilayer membrane, we observed an increase of the specific membrane conductance by several orders of magnitude. Control experiments in the presence of the same concentration of the detergent Genapol but without the protein demonstrated that the detergent alone did not lead to any appreciable increase of the membrane conductance. Single-channel experiments in the presence of small concentrations of CymA (5 ng/ml) demonstrated that the conductance increase was caused by the formation of small ion-permeable channels. Fig. 6 shows a single-channel recording taken from a lipid bilayer membrane in the presence of 5 ng/ml CymA. Interestingly, the CymA channels did not show the steplike appearance of general diffusion porins, which exhibit normally very little current noise in reconstitution experiments using lipid bilayer membranes (26). Instead, we observed a strong current noise, which was dependent on the number of reconstituted channels in a membrane. This means that part of the CymA channels, probably one or several loops between two successive ␤-strands, do not possess defined positions within the channel-forming unit and undergo rapid transitions between different states thus modulating the ion flux through the channel. Single-channel experiments were also performed at different pH in the aqueous phase (pH 5-9). However, also in these cases, we observed the rapid current fluctuations of the open channels, which made it rather difficult to evaluate the single-channel conduct-ance of the CymA channels. Fig. 7 shows the histogram of conductance fluctuations of the CymA channel in 1 M KCl. The channel distribution showed a maximum between 200 and 600 picosiemens. However, we observed also many other conductance steps probably caused by the rapid fluctuations of the CymA channels and the occurrence of multiple steps, which resulted in a broad histogram of single-channel conductance distributions. The dependence of the average single-channel conductance on the concentration of electrolyte in the aqueous phase was also difficult to obtain because of the considerable current noise of the open channels. Experiments at different KCl concentrations suggested that the single-channel conductance was a linear function of the electrolyte concentration. This means the CymA channel does probably not contain a binding site for potassium or chloride ions inside the channel.
Titration of CymA-induced Conductance with ␣-CD and Evaluation of the Stability Constant of Carbohydrate Binding-CymA confers to K. oxytoca the capacity to grow on cyclodextrins (3,4). To study the possibility that this porin has a binding site for the CDs similar to LamB for maltooligosaccharides (27,28), we performed titration experiments with CymA reconstituted into lipid bilayer membranes. The measurements were performed in the following manner. CymA was added to black lipid bilayer membranes in a concentration of about 100 ng/ml, and the membrane conductance started to increase after a lag time of a few minutes caused by slow aqueous diffusion of the protein. Simultaneously, the current noise increased considerably similar to the situation in single-channel experiments. At 30 min after the addition of the protein, the rate of conductance increase caused by reconstitution of CymA into the membrane had slowed down considerably. Then the experiment demonstrated in Fig. 8 started. Small amounts of concentrated ␣-CD solutions were added to the aqueous phase to both sides of the membrane, with stirring to allow equilibration. The results demonstrate that the membrane conductance decreased as a function of the ␣-CD concentration. Furthermore, the current noise of the recording on the strip chart recorder also decreased considerably. The data of Fig. 8 and of similar experiments were analyzed using the following equation derived here and previously from Equation 3. It describes the block of ion current through a one-site, two-barrier channel caused by substrate binding (27).
G max is the maximum membrane conductance before the first addition of the ␣-CD. G(c) is the membrane conductance at a given ␣-CD concentration c. Equation 4 means that the titration curve given in Fig. 8 can be analyzed using a Lineweaver-Burke plot as shown in Fig. 9. The straight line corresponds to a stability constant, K, of 35,300 M Ϫ1 (half-saturation constant K S ϭ 28 M). The mean value of the stability constant for ␣-CD-binding to the CymA channel was 29,000 Ϯ 9,000 M Ϫ1 (K S ϭ 34 M).

DISCUSSION
The CymA protein of K. oxytoca is required for growth of the organism on ␣-CD or ␤-CD as carbon source. Mutants with lesions in cymA loose this capacity, but they are still able to grow on linear maltodextrins that enter the periplasm via the LamB maltoporin. A defect in lamB abolishes this capacity (33); if such a lamB mutant is provided with a functional cymA gene, growth on maltodextrins is restored. Therefore, CymA has a role in the uptake of CDs and linear maltodextrins whereas LamB only accepts linear maltodextrins. K. oxytoca possesses both systems.
There is convincing evidence that CymA is a component of the outer membrane and that it functions as a porin specific for CDs. First, CymA is not solubilized from the total membrane fraction by sarcosyl, which is a property of outer membrane proteins (34). Second, CymA also requires a signal sequence for correct cellular location, indicating that it traverses the cytoplasmic membrane via the sec-dependent pathway (44). Third, the secondary structure composition is typical for outer membrane proteins and porins in particular (36). The antiparallel ␤-sheet very likely forms a ␤-barrel with a hydrophobic outer surface and a more hydrophilic inner side. copy and functionally proven by conductance measurements.
The single-channel conductance is similar to that of specific porins (45). Interestingly, the current traces following the stepwise increases showed a considerable current noise (see also Fig. 6), which may indicate that part of the protein, possibly surface exposed loops, have no defined location within the channel and modulate the current. This is presumably also the reason for the unusually broad channel distribution in the histograms.
We investigated the binding of ␣-CD to the binding site inside the CymA channel. This was done by measuring the inhibition of the ionic current through the channel with increasing concentrations of ␣-CD and assuming that a one-site two-barrier model is valid for the movement of the sugars through the CymA-channel (27). Implicit in the model is also that the affinity of the CD is vastly higher than the near-zero affinity of the metal ions for the binding site, which is justified by the single-channel experiments at different salt concentrations.
The binding of ␣-CD to CymA agrees with its function as specific outer membrane porin because CymA is responsible for the facilitated diffusion of ␣-CD across the outer membrane. Binding is a prerequisite for facilitated diffusion. We compared the binding constants derived here for ␣-CD to CymA with carbohydrate binding to LamB of E. coli (27) and to ScrY (30) using the same approach. The stability constants for the binding of maltooligosaccharides to these specific porins increase with increasing chain length and reach a stability constant approximately similar to that for maltoheptaose binding as derived here for ␣-CD. This means that the binding site inside the CymA channel probably provides an advantage for the transport of ␣-CD across the outer membrane similar to that of LamB and ScrY for maltooligosaccharides. However, the on and off constants for ␣-CD binding, which finally describe the velocity of substrate movement through CymA (46), remain to be determined.
Although the effective channel size cannot be extracted from projection images for fundamental reasons, it is obvious from the averages that the stain-filled regions revealed by electron microscopy possess a diameter of 1.2-1.4 nm, which is somewhat less than the apparent width of a funnel-shaped porin pore in electron micrographs (42,43). However, this value is close to the size expected to be required for the CDs and might, thus, represent a reasonable estimate for the channel diameter. CDs are relatively stiff molecules, but they were found to adapt their structure upon binding to proteins (47). We assume that similar effects may facilitate the passage through the apparently narrow porin channel.
Electron microscopical investigations also revealed that CymA behaves distinctly different from classical porins, concerning the formation of oligomeric complexes in lipid membranes. CymA neither occurs in trimers in reconstituted membranes nor does it trimerize in solution. Solubilized porins from other sources usually crystallize in two dimensions with p3 symmetry or on p2-related rectangular lattices but always with a trimer as the "asymmetric" unit (42), whereas CymA formed two-dimensional crystals being related to p1-type plane groups and built from monomers not forming symmetric complexes. This is an unusual and unexpected finding for CymA, functioning as a specific porin. On the other hand, it is not unusual for outer membrane proteins in general. There are at least three classes of bacterial outer membrane proteins not occurring as trimers, i.e. the large ␤8 family, consisting of eight-stranded ␤-sheet proteins including OmpA (48), the group of siderophore-specific pore proteins like FhuA (49), and the monomeric autotransporters (50). CymA is certainly not structurally related to the first two groups because of its size. However, at present we cannot clearly decide the relationship of CymA to other outer membrane proteins forming channels. It would be interesting to exactly determine the number of ␤-strands in CymA since general diffusion porins, specific porins, and the outer membrane proteins mentioned above can be distinguished by this property.