Crystal structure of the sugar acid‐binding protein CxaP from a TRAP transporter in Advenella mimigardefordensis strain DPN7T

Recently, CxaP, a sugar acid substrate binding protein (SBP) from Advenella mimigardefordensis strain DPN7T, was identified as part of a novel sugar uptake strategy. In the present study, the protein was successfully crystallized. Although several SBP structures of tripartite ATP‐independent periplasmic transporters have already been solved, this is the first structure of an SBP accepting multiple sugar acid ligands. Protein crystals were obtained with bound d‐xylonic acid, d‐fuconic acid d‐galactonic and d‐gluconic acid, respectively. The protein shows the typical structure of an SBP of a tripartite ATP‐independent periplasmic transporter consisting of two domains linked by a hinge and spanned by a long α‐helix. By analysis of the structure, the substrate binding site of the protein was identified. The carboxylic group of the sugar acids interacts with Arg175, whereas the coordination of the hydroxylic groups at positions C2 and C3 is most probably realized by Arg154 and Asn151. Furthermore, it was observed that 2‐keto‐3‐deoxy‐d‐gluconic acid is bound in protein crystals that were crystallized without the addition of any ligand, indicating that this molecule is prebound to the protein and is displaced by the other ligands if they are available.


Introduction
Nutrient uptake is essential for growth of all living organisms. Versatile strategies have evolved using different types and mechanisms for the uptake of nutrients. The well-studied phosphotransferase system is often used for carbohydrate uptake in bacteria [1]. Other strategies include permeases of the major facilitator superfamily or ATP-binding cassette transporters [2] for a broad range of substrates. Furthermore, a tripartite ATP-independent transport (TRAP) system mediates the uptake of D-gluconate in Sinorhizobium meliloti [3]. Subsequent to their discovery in the 1990s [4], several TRAP transport systems have been described accepting various ligands. The only limitation appears to be the need of a carboxylic acid in the structure of the ligand [5][6][7][8]. TRAP transport systems consist of three components: the two transmembrane components, DctQ and DctM, and the substrate binding protein (SBP), DctP [9]. The transport of the ligands starts with an interaction of the ligand with the SBP, which subsequently interacts with the complex of the transmembrane proteins DctQ and DctM. The translocation of the ligand is then mediated by DctM, the larger of both membrane proteins, consisting of twelve transmembrane-spanning helices that form a barrel structure [6].
Besides the TRAP transporter from S. meliloti, Bacillus halodurans also possesses an SPB of a TRAP transporter, which was also shown to accept D-gluconate, D-xylonate and D-galactonate as ligands, although the corresponding transport system has not been characterized [8]. Recently, a TRAP transporter involved in the uptake of sugar acids was described in Advenella mimigardefordensis strain DPN7 T [10] as part of a novel strategy of carbohydrate uptake in this Gram-negative bacterium. The deletion of the individual components of the TRAP transporter led to restricted growth of the bacterium with the monosaccharides D-glucose, D-galactose, D-fucose and L-arabinose, as well as to poor growth with D-xylose. Additionally, intracellular accumulation of 14 C-labeled D-glucose was also shown to be dependent on the identified TRAP transporter, demonstrating that the same single transporter is crucial for the uptake of carbohydrates [10]. This was unexpected because the substrate range of so far described TRAP transporters is limited to organic acids [5,8].
Further investigations of the SBP in A. mimigardefordensis DPN7 T revealed that not the sugar molecules, but the corresponding acids are the ligands of this protein [10]. In silico analyses revealed that the strain possesses putative membrane-bound quinone glucose dehydrogenases, and the oxidation of the respective sugars was observed in the membrane fraction of cells of A. mimigardefordensis. This indicated that, prior to transport, the carbohydrates are oxidized to their corresponding sugar acids in the periplasmic space of the bacterial cell, which explains the restricted growth of the strain with D-glucose and the lack of 14 C-labeled D-glucose accumulation in the deletion mutants.
Because the protein is not a binding protein for dicarboxylates, the name DctP Am could lead to confusion regarding its function. To reflect the substrates of this particular SBP and thus the function with respect to the transport of sugar acids in A. mimigardefordensis DPN7 T , it is renamed CxaP. This new name is used throughout the present study.
During the screening of ligands of CxaP, it became obvious that the terminal carboxylic acid group and the hydroxylic groups at position C2 and C3 are important for ligand binding. In general, the overall sequence similarity of TRAP type SBPs is relatively low. Structurally, however, they appear to be conserved. DctP proteins, similar to CxaP, usually consist of two domains linked by a hinge and spanned by a long α-helix, as is typical for SBPs of TRAP transporters [5,11,12]. Generally, the ligand binding in TRAP type SBPs involves a salt bridge formed by a highly conserved arginine and the carboxylic acid of the ligand [5,13]. This was also identified in CxaP at position Arg175 [10]. Because a large variety of ligands are to be bound by CxaP [5,14], it would be of interest to investigate the binding site of the protein in great detail by X-ray crystallography in combination with mutagenesis analysis. The present study shows highresolution crystal structures of CxaP in complex with different ligands and thereby provides information about the architecture of the binding pocket and coordination of the ligands in the protein. This is the first crystal structure with sugar acid ligands reported for a TRAP type SBP.

Results and Discussion
The uptake of carbohydrates in A. mimigardefordensis strain DPN7 T by a novel strategy has been described previously [10]. It was shown that the monosaccharides D-glucose, D-fucose, D-galactose, D-xylose and L-arabinose are oxidized in the periplasm and the resulting corresponding sugar acids are subsequently transported into the cell by the characterized TRAP transport system [10]. The SBP displayed wide specificity with similar affinity toward five different ligands, which is unusual for TRAP type SBPs. Usually, these transporters exhibit high specificities toward one or two structurally very similar ligands, as well as modest affinities, with this being reduced by an approximate factor of ten toward a few other structurally similar ligands [13][14][15].

Heterologous expression and crystallization of CxaP
For crystallization, the protein was heterologously expressed. After the exchange of the elution buffer to sodium phosphate buffer (20 mM, pH 7.2, 100 mM NaCl), crystallization experiments were performed as described in the Materials and methods. Crystals were obtained with a pure CxaP protein solution and CxaP, which was supplemented with 10 mM D-galactonic, D-gluconic, D-fuconic or D-xylonic acid, respectively, prior to crystallization at similar conditions (see Materials and methods).
Crystals of the CxaP protein diffracted to a maximum resolution of 1.3Å. This high-resolution dataset was phased using the automated AUTORICKSHAW pipeline employing the automated molecular replacement mode (http://www.embl-hamburg.de/Auto-Rickshaw) and the CxaP protein sequence as input. After several rounds of manually model building using COOT [16] and subsequent refinement using REFMAC5 (https:// www.ccp4.ac.uk)/PHENIX (https://www.phenix-online.org), the structure of the full-length CxaP protein was modeled in terms of the electron density. The CxaP structure was refined to 1.3Å resolution and the resulting R work and R free values were 13.9% and 16.1%, respectively. Interestingly, within the protein, a ligand was found to be bound later identified as 2keto-3-deoxy-D-gluconic acid (see below). This ligand was not added to the protein and apparently was copurified with the protein during the whole purification and crystallization procedure. Therefore, we termed this structure CxaP:2-keto-3-deoxy-D-gluconic acid. Once the refinement of the 1.3Å structure of the CxaP:2-keto-3-deoxy-D-gluconic acid protein was finished, it was used to phase the other datasets obtained from crystals of the CxaP protein co-crystalized with D-galactonic acid, D-xylonic acid, D-fuconic acid or D-gluconic acid. Adding the ligands to the CxaP protein also yielded high-resolution crystal structures that were subsequently refined. The CxaP:D-galactonic crystal was solved at 2.0Å (R work and R free values were 20.3% and 25.7%, respectively); CxaP:D-xylonic acid crystal was solved at 1.7Å (R work and R free values were 18.2% and 20.2%, respectively); CxaP:D-fuconic crystal was solved at 2.0Å (R work and R free values were 20.8% and 26.7%, respectively); and CxaP:D-gluconate crystal was solved at 2.0Å (R work and R free values were 18% and 23.8%, respectively). A summary of the data collection statistics, refinement details and model content of these different CxaP structures is provided in Table 1.
Interestingly, the crystal parameters, especially the unit cell dimensions and space group, differ between the crystals (Table 1), resulting in a different symmetry of the CxaP D-fuconic acid co-crystals. Here, the symmetry was found to be P2 1 , whereas all other datasets display a P2 1 2 1 2 1 symmetry. All crystals contained a monomer of CxaP in the asymmetric unit except the CxaP:D-fuconic acid, where two monomers were found.

Overall structure of CxaP
We compared the individual structures of CxaP with its substrates and determined rmsd values that range from 0.15 to 0.3Å over 264 Cα atoms, indicating that there were no large conformational changes present. Because the monomers of the different DctP crystal structures are almost identical in overall shape, we describe the general monomeric fold for the highest resolution structure, namely CxaP:2-keto-3-deoxy-D-gluconic acid, which was solved at 1.3Å in detail as outlined below.
CxaP comprises 335 amino acids and, as expected for a TRAP transporter binding protein, it is composed of two domains, which are colored blue (domain I, residues 1-143 and 247-260) and light brown (domain II, residues 159-236 and 292-335), respectively ( Fig. 1). Each of these domains is composed of a central antiparallel five-stranded β-sheet. Furthermore, domain I contains six α-helices, whereas domain II harbors eight α-helices and one additional β-strand ( Fig. 1), which flank the central sheet (green in Fig. 1) on both sites. This strand order is found in many other SBPs of TRAP transporters, as well as in class II ATP-binding cassette-dependent SBPs [11,12]. The domains are located non-colinearly on the gene sequence and helix α13 (amino acids 256-291) and βstrands β4 (amino acids 143-158) and β9 (amino acids 237-247) connecting both domains together (Fig. 1).
To identify the structurally closest homologs of the CxaP protein, we performed a DALI search [17,18], which revealed a variety SBP of TRAP transporters, of which the top 10 are listed in Table 2. Interestingly, the Z-scores are quite high, ranging from 29.9 to 38.4, despite the relatively low sequence identity, which ranges from 15% to 26%. This highlights that SBP from TRAP transporters adopt highly similar three-dimensional structures, suggesting that the venus fly trap mechanism for capturing the ligand remains the same. The overall fold of the CxaP protein places this protein in class IIIa of SBPs, which is a subgroup containing mainly SBPs from TRAP transporters [11,12]. Among the 10 homologous structures, five are bound with a rather linear ligands, similar to the CxaP protein homolog from Pseudomonas aeruginosa with bound L-malate [Protein Data Bank (PDB) code: 4NF0] [8], from Chromohalobacter salexigens with bound D-alanine-D-alanine (PDB code 4N8G) [8], from Desulfovibrio salexigens with bound diglycerolphosphate (PDB code 4N6K) [8], from Rhodoferax ferrireducens with bound malonate (PDB code 4MCO) and from C. salexigens with bound (S)-(+)-2-amino-1-propanol (PDB code 4XF5). The others contain a ring-like ligand structure similar to UehA from Silicibacter pomeroyi DSS-3 [19] with bound ectoine (PDB code 3XFB), as well as the DctP homologs from Ruegeria pomeroyi with bound 3,4dihydroxybenzoate (PDB code 4PAF), from R. ferrireducens with bound benzoate (PDB code 5IM2) and from Bordetella bronchiseptica with bound picolinic acid (PDB code 4YIC) [8,19]. This suggests that although the overall structure is closely related, the binding site of the ligand discriminates between a linear and a ring-like ligand.

The ligand binding site of CxaP
In the SBP of TRAP transporters, the ligands are bound in the cleft flanked by the two domains of the protein. In the CxaP protein, which was crystallized in the apo-form, a ligand was bound that could be identified as 2-keto-3-deoxy-D-gluconate (KDG) as a result of the high resolution. This was unexpected and suggests that KDP was copurified after expression in Escherichia coli and tightly bound to the CxaP protein ( Fig. 2).
KDG is bound in the binding cleft in between and via interaction with the two domains. Interaction with the domain I is mediated via the side chains of Arg175, Arg154, Asn215, and Asn151 (Fig. 2, shown in light brown). Other interactions arise from domain II and are mediated by the side chains of Ser96 and Asp78, as well as via the backbone interaction of Tyr38, Gly39 and Leu40. These amino acids also play a role in the binding of the ligands (Fig. 3). The presence of the KDG intermediate of the Entner-Doudoroff pathway KDG in the binding pocket was unexpected because the ligand structure differs from the known sugar acid ligands of CxaP. The Entner-Doudoroff pathway is used for gluconate metabolism in E. coli and KDG is most probably formed during cultivation for protein expression [20]. This is not the first time that an unexpected molecule was identified in SBP protein crystals. Binding of glycerol to an α-keto acid-binding SBP of Rhodobacter sphaeroides has been reported prevously [15].  By comparison of the interactions of KDG with the amino acid residues in the binding pocket of CxaP and the interactions observed for the other ligands, it is obvious that KDG shows interactions with amino acid residues similar to those of sugar acids. Approximate dissociation constants for the interaction of CxaP and the sugar acid ligands have been described previously, with the values appearing to be approximately 10-fold higher compared to other TRAP transporters [6,[10][11][12]. Because the protein used for determination of the dissociation constant (K d ) was synthesized by E. coli in complex medium and prebound ligands  potentially have not been removed [10], the observation of higher K d values is most probably a result of prebound KDG in the protein samples. The prebound KDG is thereby most likely displaced by the sugar acid ligands. To our knowledge, this is the first report of an interaction of KDG with an SBP of a TRAP transporter. By contrast to the known ligands of the CxaP, KDG possesses a keto group at position C2 and no functional group at position C3. This indicates that the substrate range of CxaP could include further ligands, which probably have lower affinities but still might be accepted by the protein.
It was reported earlier that the selectivity of CxaP is based on a carboxylic group and a hydroxyl group at C2 in R-configuration and a hydroxylic group positioned at C3 in S-configuration [10]. In general, the architecture of the binding site of CxaP confirmed this observation (Fig. 2). However, it was observed that KDG was also accepted as a ligand in the binding pocket, as discussed further below.
We set out to crystallize CxaP with different bound ligands and succeeded to crystallize the CxaP in complex with D-gluconic, D-xylonic, D-fuconic and D-galactonic acid. Crystals were grown under similar conditions as the wild-type crystals. Datasets were obtained at high resolution, and the already determined structure of CxaP in complex with KDG was used as a template to solve the structures with the different ligands ( Table 1).
The carboxylic group of the different sugar acids interacts with a highly conserved Arg175 (Fig. 3), which is typical for TRAP SBPs and is present in all known structures. Furthermore, the Arg residue has been shown to be crucial for the high selectivity toward carboxylic acid ligands [21]. The carboxylic group also interacts with another arginine residue (Arg154) and an asparagine (Asn215). These interactions are not exclusively present in CxaP. Similar interactions including two Arg and an Asn have also been reported for SiaP, a sialic acid-binding protein from Haemophilus influenzae [13]. Arg154 and Asn215 additionally interact with the hydroxylic group at position C2. In addition, in the case of the C6-ligands, another interaction of Asn151 at position C2 could be observed.
The ligands are stabilized at position C3 by an interaction with either Ser96 (Fig. 3). Additionally, all sugar acid ligands are stabilized by hydrogen bonds between the backbone of the ligand in position C4 and Asp78.

Activity and CxaP variants
To confirm the function of the amino acids in the binding pocket of CxaP, single-exchange mutants were generated via site-directed mutagenesis. All proteins were purified resulting in pure protein as confirmed by SDS/PAGE. Especially, the variant Asp78Leu appeared to be less stable and resulted in significantly less protein compared to the others. The targeted amino acids were Gly39, Asp78, Ser96, Arg154, Arg175, Phe198 and Asn215. Three different mutants were generated for Gly39. This was changed to alanine, leucine and valine to determine whether the size of the amino acid at this site has an impact on the activity of the protein toward differently sized ligands. Asp78 was changed to leucine and Ser96 was converted to alanine and valine, respectively. The Arg154 residue was mutated to alanine and Arg175 was changed to lysine. Isoleucine was introduced for Phe198 and Asn215 was changed to leucine. The mutations were chosen based on sequence alignment using sequence of different homologous proteins (Fig. 4). Here, the different CxaP homolog sequences that were aligned appear to have rather narrow substrate specificity.
K d values of wild-type CxaP and the generated variants were determined to analyze the impact of the different amino acid exchanges on protein and substrate binding via a fluorescence binding assay. The analyzed ligands were D-gluconic, D-galactonic, D-fuconic and D-xylonic acids as ligand for CxaP. All calculated dissociation constants are shown in Table 3. The wildtype protein binds D-gluconic acid with a K d value of 8.0 µM; for D-galactonic acid, the wild-type K d value was 6.6 µM. D-fuconic acid was bound to CxaP with a K d value of 1.9 µM and, with D-xylonic acid, a dissociation constant of 3.9 µM could be determined. To determine the K d value, the measured fluorescence decrease (%) is plotted against the used substrate concentration (µM). Fits for the K d value determination for all of the tested ligands were generated using ORI-GIN (OriginLab, Northampton, MA, USA). In Fig. 5, the curves for the wild-type CxaP are shown. All of the calculated K d values for the CxaP variants were determined accordingly. In summary, the K d values calculated for the wild-type protein CxaP were in the range 2-8 µM for the different ligands which is in line with previous studies [10]. In most cases, the K d value of the generated variants is higher than the K d of the wild-type CxaP, indicating a lower affinity (Table 3). Especially, the variants Arg154Ala, Arg175Lys, and Phe198Ile led to an increase of the dissociation constant with most ligands or even to a lack of interaction, which can be explain by the important interactions being observed in the wild-type structures. For Gly39Val, the K d values of the different ligands were two-to six-fold higher compared to the wild-type K d . The K d values of Arg154Ala were approximately five-fold higher compared to the wild-type with D-gluconic and D-xylonic acid and 10-fold higher using D-galactonic acid as the ligand. No interaction between the protein and the ligand was detectable either for Arg175Lys with D-gluconate and D-xylonic acid or for Phe198Ile with D-fuconic and D-xylonic acid. The K d value of Phe198Ile with D-gluconic acid was approximately three-fold higher compared to the K d of the wild-type with the same ligands (Table 3). In summary, however, the observed changes of the K d values are similar. This is likely a result of the multiple interactions present for binding a sugar acid ligand. A loss of a single interaction does not lead to a dramatic loss of affinity. This again highlights that CxaP has a wide substrate range and the binding site appears to be optimized with the ability to accommodate several different ligands, although these are of the same nature. This is striking because CxaP is unique in being able to bind a variety of ligands with a very similar affinity. Usually, other proteins of the TRAP transporter family are optimized for one specific substrate and can bind other ligands with a lower affinity.

Function of Gly39
An interesting amino acid in the binding pocket of CxaP is Gly39. This residue directly interacts with the hydroxylic groups of D-gluconic acid and KDG and also with the methyl group of D-fuconic acid, in each case at position C6. The shorter pentanoic acid D-xylonic acid does not interact with Gly39 at all. This indicates that the Gly39 determines the size of accepted ligands with regard to the chain length. This can also be deduced form the sequence alignment with protein displaying a narrow substrate spectrum (Fig. 4). Here, the Gly residue is not conserved throughout the proteins. Instead, it appears to be exchanged to other amino acids, such as Asp, Asn, Tyr, Leu and Val. By intruding any other amino acid, a side chain would point to within the binding site, emphasizing that this might alter the substrate specificity. We confirmed this hypothesis by introducing three variations at the Gly39 position. Here, glycine was mutated to alanine (Gly39Ala), leucine (Gly39Leu) or valine (Gly39Val). After expression and purification, the K d values for the substrates D-gluconic, D-galactonic, D-fuconic and D-xylonic acid were measured (Table 3). Especially, the introduction of alanine and leucine appeared to have a stabilizing effect on D-xylonate. Here, the affinity toward this substrate increased. The wild-type protein displayed a K d of 3.9 AE 0.5 µM. Whereas, in the variant Gly39Ala this was reduced to 2.4 AE 1.5 µM. The strongest change, however, was observed for the Gly39Leu mutation where this affinity was even further strengthened resulting in a further drop of the K d to 1.3 AE 0.1 µM. It appears that the side chains introduced are interacting with the D-xylonic acid substrate giving a stabilizing effect. A further increase of the side chain appears to hinder the binding capacity for this substrate as observed by the lowered K d for the Gly39Val mutation and the D-xylonic acid substrate. Here, the K d drops even below the K d measured for the wild-type protein.
A similar effect was observed when galactonic acid was used as substrate with the corresponding mutations, although to a smaller extent. Because the other used substrates were already interacting with the backbone of Gly39, the introduction of another amino acid (e.g. side chain) resulted in a sterically unfavored binding, as observed by lower affinity for the Gly39 mutation toward these substrates.

Conclusions
CxaP was described as the SBP of a TRAP transport system with a selectivity filter for sugar acids involving particular characteristics. By analysis of the crystal structure, the binding of the ligands depending on a carboxylic group at position C1, as well as the position of the hydroxylic groups at C2 and C3 in R-and Sconfiguration [10], respectively, was confirmed. However, the substrate range must be expanded by a ligand with a less similar structure, KDG, which is most probably accepted by the SBP with modest affinity. Further analysis of the binding pocket of CxaP is desirable to confirm these observations. The present study is the first description of the structure of a TRAP type SBP accepting sugar acid ligands with the described structural properties. The majority of the SBPs of TRAP transporters occur as monomers, although stable dimers have also been reported [15]. CxaP has a monomeric structure. The interaction of the ligand with a highly conserved arginine residue is essential for the coordination and binding of the solute in SBPs of TRAP transporters [21] and has been reported for all characterized TRAP transporters and confirmed in the obtained crystals of CxaP. A special feature of CxaP is the glycine residue in position 39, which allows multiple ligands to bind to the substrate binding pocket. This is not common for other proteins of the TRAP transporter family and was achieved via the function of Gly39 in the binding pocket of CxaP. Mutational analysis of this amino acids revealed changes in the affinity toward some of the ligands.
To our knowledge, the interaction of an SBP of a TRAP transporter with its sugar acid ligands has only been reported in such detail three times to date [3,8,10]. Only amino acid sequences for which experimental data are available have been chosen for the preparation of the alignment. By comparison of the amino acid sequences, it is striking that the amino acid residues, which are directly involved in the coordination of the ligand, are highly conserved in the sequences of the D-gluconic acid accepting SBPs. To our knowledge, the uptake strategy, consisting of oxidation in the periplasm and subsequent import via CxaP and its corresponding transport system, has only been described for A. mimigardefordensis DPN7 T [10,22].

Cultivation of bacterial strains
Cells of E. coli strains (Table 4) were cultivated in lysogeny-broth medium at 30°C or 37°C supplemented with addition of ampicillin (Amp, 75 µgÁmL −1 ), if necessary. Liquid cultures were incubated in baffled flasks on a rotary shaker at 130 r.p.m. E. coli Top10 was used for cloning procedures and plasmid propagation (Table 4). E. coli BL21 (DE3) pLysS was used for heterologous expression of cxaP and the respective mutants. Solid medium contained 1.8% (w/v) agar.

Expression and purification of CxaP
Heterologous expression of CxaP was accomplished in E. coli BL21 (DE3) by applying pET19b:cxaP as described before [10]. For this, a preculture of 20 mL of LB medium containing ampicillin in a concentration of 75 µgÁmL −1 was grown overnight. A main culture of 100 mL ZYP- autoinduction medium was inoculated with 1 mL of this preculture [23] and grown for 16-20 h on a rotary shaker at 130 r.p.m. and 30°C. Cells were harvested by centrifugation (10 min, 4°C, 7000 g) and washed in 20 mL of sodium phosphate buffer (20 mM, pH 7.2) containing 20 mM imidazole and 500 mM NaCl. After washing, the cell pellet was resuspended in 5 mL of sodium phosphate buffer and then cells were disrupted with three passages through a French press. Insoluble components of the lysate were removed by centrifugation for 30 min at 21 000 g and 4°C. CxaP was purified as described previously [10]. Prior to the crystallization experiments, the elution buffer from the purification was exchanged to 20 mM sodium phosphate buffer containing 100 mM NaCl using Vivaspin 500 tubes (GE Healthcare, Freiburg, Germany).
The protein concentration was determined via a Bradford assay [24] and the purity of the protein was analyzed via SDS/PAGE.

Site-directed mutagenesis
Site-directed mutagenesis was used for amino acid substitutions of the respective CxaP variants. For this, a pair of phosphorylated primers was generated (Table 5) to substitute a certain amino acid DNA triplet via point mutation.
The forward primer introduced the desired mutations into the nucleotide sequence. The reverse primer was located adjacently. In a polymerase chain reaction using pET-19b:: cxaP as template, a linear fragment consisting of the complete plasmid sequence was amplified. The polymerase chain reaction was performed using Phusion polymerase (Fermentas, St Leon-Rot, Germany) and Phusion High Fidelity buffer supplemented with 3% (v/v) dimethylsulfoxide. Afterward, the parental plasmid was digested using DpnI (Thermo Scientific, Darmstadt, Germany) and the linear fragments were ligated using T4-Ligase (Thermo Scientific). Subsequently, E. coli Top10 was transformed with the mutated plasmid. All plasmids were verified by sequencing. The following protein variants were generated: G39A, G39L, G39V, D78L, S96A, S96V, R154A, R175K, F198I and N215L.

Crystallization of CxaP
Protein samples for crystallization were initially screened by applying a commercial screen (Molecular Dimensions,  Table 5.
Oligonucleotides used for generation of cxap variants.
Oligonucleotide Sequence (5 0 -to 3 0 ) Amino acid substitution  7) and 25% Jeffamine. The crystals of the initial screening obtained in this condition were shredded and used in microbatch set-ups (1 µL + 1 µL drops, 300 µL of reservoir solution) as seed stock in the optimization of the crystallization conditions in grid screens around the initial conditions, varying the concentrations. Thereby, protein crystals were obtained by incubation of CxaP with the 100 µM D-fuconic acid, D-xylonic acid and 1 mM D-gluconic acid and D-galactonic acid for 1 h on ice prior to crystallization. The protein crystals formed were fished after overlaying the drop with mineral oil and then flash-frozen in liquid nitrogen.

Data processing and structure determination
Data sets were collected from a single crystal of either CxaP or CxaP in complex with the different ligands at the ID29 beamline (European Synchrotron Radiation Facility, Grenoble, France) at 100 K. Initial images were collected and a strategy was calculated using the beamline software [25]. These data sets were processed using the XDS package and scaled with XSCALE [26,27]. Initial phases were obtained by molecular replacement using the web server autorickshaw employing solely the sequence of CxaP as template.
Model building and refinement were performed using COOT [16] and PHENIXREFINE [28] as part of the PHENIX software. Data refinement statistics and model content are summarized in Table 3

Fluorescence spectroscopy
Fluorescence spectroscopy was used to determine the K d values of CxaP and the generated mutants of CxaP with gluconate, galactonate, fuconate and xylonate. For this, 1 µM protein was used per measurement starting with a substrate concentration of 0.2 µM, which was added to the protein. To achieve the desired substrate concentrations, four different stock solutions were used for each ligand containing 0.1, 1, 5 or 50 mM of the respective substrate. In every titration step, 2 µL was used, starting with the 1 mM stock solution up to the 50 mM stock solution, to cover the whole concentration range. Fluorescence was measured at 328 nm. After volume correction and normalization, the K d values were determined using values for the decrease in fluorescence (%) plotted against the substrate concentration (µM). For that, software ORIGIN (OriginLab) was used and the equation F 0 þ F amp Â 0:5 Â ð1=PÞ Â P þ K d þ x ð Þ ð À P þ K d þ x ð Þ 2 À 4 Â P Â x 0:5 Þ was applied. F 0 is set as the lowest value for the fluorescence decrease and F amp is set as the amplitude of values for the fluorescence decrease. The used protein concentration is indicated by P (1 µM).
The K d value determination was calculated from at least biological triplicates. As a control, the same experiment was performed with buffer was added instead of one of the substrates and used for background correction.