En route to photoaffinity labeling of the bacterial lectin FimH

Summary Mannose-specific adhesion of Escherichia coli bacteria to cell surfaces, the cause of various infections, is mediated by a fimbrial lectin, called FimH. X-ray studies have revealed a carbohydrate recognition domain (CRD) on FimH that can complex α-D-mannosides. However, as the precise nature of the ligand–receptor interactions in mannose-specific adhesion is not yet fully understood, it is of interest to identify carbohydrate recognition domains on the fimbrial lectin also in solution. Photoaffinity labeling serves as an appropriate methodology in this endeavour and hence biotin-labeled photoactive mannosides were designed and synthesized for photoaffinity labeling of FimH. So far, the photo-crosslinking properties of the new photoactive mannosides could be detailed with the peptide angiotensin II and labeling of FimH was shown both by MS/MS studies and by affino dot–blot analysis.


Introduction
Photoaffinity labeling is a technique by which ligand binding sites of a receptor protein can be identified in solution. It requires a photoactive ligand derivative, which can form a reactive species upon photo-excitation. Thus, incubation of the photoprobe with a protein followed by irradiation can result in a photo-crosslinked product, that provides structural information on the binding site of the protein (Figure 1) [1].
It has become our goal to utilize this methodology for the investigation of carbohydrate binding of the bacterial lectin FimH. The protein FimH is found on the tips of type 1 fimbriae, long adhesive filamentous appendages on the surface of many bacteria, such as Escherichia coli [2][3][4][5]. In X-ray studies a carbohydrate recognition domain (CRD), which can complex one α-D-mannosyl residue, has been clearly identified [6][7][8].  However, other binding experiments performed with a multitude of synthetic as well as natural mannosides and oligomannosides are not in complete agreement with just one monovalent carbohydrate binding site on the FimH protein [9][10][11][12][13]. Thus, we decided to employ photoactive ligands to probe carbohydrate binding to the known CRD in solution and to identify possibly auxiliary, so far unknown, binding sites on bacterial lectin FimH [2].
The known FimH CRD was taken as the lead structure for the design of an appropriate photoactive ligand. Inspection of the available X-ray data clearly shows that α-D-mannosides are complexed in the CRD with the aglycone moiety sticking out of the binding pocket. The entrance of the CRD is flanked by the tyrosine residues Tyr48 and Tyr137 ('tyrosine gate') that can form favorable interactions with appropriate mannosidic aglycone moieties, such as π-π-stacking with the phenyl group of benzyl mannosides [14]. In a preceding work, we synthesized the three corresponding photoactive α-D-mannosides, 1, 2, and 3 ( Figure 2a) for photolabeling of the bacterial lectin FimH [15]. They differ in their photoactive functional groups, which are part of the aglycone moiety. Upon irradiation, the aryl azide 1 and the diazirine 2 expel nitrogen to yield a reactive nitrene and a carbene intermediate, respectively. Irradiation of the benzophenone 3, on the other hand, delivers a reactive triplet diradical. In addition, in order to combine a photoactive functional group with an affinity label within the same mannoside, the orthogonally protected glycoamino acid scaffold 4 was synthesized and used for the preparation of the biotin-labeled photophore 5, which is well suited for the streptavidine-based photoaffinity labeling, relying on the extraordinary high affinity of the protein streptavidine for biotin ( Figure 2b) [16].

Results and Discussion
In order to learn more about the ligand properties of the previously prepared photoactive mannosides 1, 2, 3, and 5, we have performed computer-aided docking studies using FlexX [17][18][19] to get an idea about their binding to the bacterial lectin FimH, as reported earlier [20]. FlexX produces so-called scoring values for each docked ligand, which can be regarded as a rough estimate of its free binding energy. Low (more negative) scores correlate with high affinities, whilst higher scores reflect diminished binding potency (Table 1). Docking was based on two different X-ray structures. In the first case, crystals with an open tyrosine gate were taken as the basis [7], whilst in the second case, a closed-gate conformation of FimH was used for the modeling [8]. This led to somewhat different predictions, as expected, however, the same trends were revealed for ligands 1-3 and 5 in comparison with the wellknown standard ligands methyl α-D-mannoside (MeMan) and p-nitrophenyl α-D-mannoside (pNPMan), respectively (Table 1).
When these mannosides were tested as inhibitors of type 1 fimbriae-mediated bacterial adhesion to a mannan-coated surface in an ELISA [21,22], IC 50 -values were obtained, which reflect the concentration of the derivative employed, that leads to 50% inhibition of bacterial adhesion. Three independent measurements resulted in high standard deviations, a typical characteristic of this assay, whereas the relative trend in a series of tested ligands can be verified with high reproducibility. Hence, the inhibitory potencies of the tested ligands were referenced to an internal standard inhibitor, MeMan (inhibitory potency ≡ 1), to deduce relative inhibitory potencies, so-called RIP-values (Table 1). This uniform referencing shows, that all photoactive mannosides surpass the inhibitory potency of MeMan, and that mannosides 1 and 2 are more powerful inhibitors than 3 and 5. Consequently, the synthetic photoactive mannosides are suited as ligands for the bacterial lectin FimH. Comparison of the measured IC 50 -values with the theoretical docking results, discloses a somewhat limited value of the computer-aided predictions in this case. Docking suggested that mannosides 2, 3 and 5 are the most potent ligands of FimH in the tested series, which was not confirmed experimentally. On the other hand, it has to be kept in mind that the employed ELISA is not based on pure FimH, but rather on type 1-fimbriated bacteria, a dissimilar more complex system.
The binding studies showed that mannosides with a typical photolabel in the aglycone moiety serve as ligands for FimH, and that even the more complex glycoamino acid derivative 5 is a suitable ligand. Encouraged by these results, we set out to improve the photochemical properties of 5 in order to facilitate later photoaffinity-labeling of FimH. Our earlier work suggested that diazirines are more useful photoactive groups than aryl azides and benzophenones [15]. Therefore, the synthesis of a biotin-labeled daizirine-functionalized mannoside was our next target. In this synthesis, aspartic acid was utilized as the scaffold molecule in two different ways in order to allow fine-tuning of the spacing between the mannoside ligand and the photoactive group. Thus, the amino acid derivatives Fmoc-Asp-OtBu and Fmoc-Asp(OtBu)-OH were employed in two analogous synthetic pathways, starting with peptide coupling to the known 2-aminoethyl mannoside 6 (Scheme 1) [23]. This led to the orthogonally protected mannoside amino acid tert-butyl esters 7 and 8. The tert-butyl ester groups were then cleaved under acidic conditions and the resulting acids ligated with the biotin derivative biotinylamidopropylammonium trifluoroacetate. These two steps gave the Fmoc-protected biotin-labeled glycoamino acids 9 and 10, respectively. Fmoc-cleavage and peptide coupling with the diazirine 11 led to the target molecules 12 and 13 in good yield.
To test the prepared photoactive mannosides in crosslinking reactions, the model peptide angiotensin II (DRVYIHPF) was first employed. It was irradiated with the three different diazirine-functionalized mannosides 2, 12, and 13. From our earlier work [15] it was known that irradiation of the diazirinefunctionalized mannoside 2 with angiotensin II led to a photocrosslinked product with m/z = 698.82. This double positively charged ion correlates with a 1:1-photoaddition product of peptide and photolabel with the molecular formula C 65 H 90 F 3 N 13 O 18 2+ and a monoisotopic mass of 1397.47 Da.
We have now carried out MS/MS experiments to analyze the structure of this photo-crosslinked adduct and unequivocally shown that insertion of the carbene, which results after irradi- The photo-crosslinking experiments with angiotensin II demonstrated, that the synthesized diazirines are well suited as photoprobes for the labeling of this peptide, with a preference for the tyrosine side chain. Thus, the photoactive mannosides were next investigated with the bacterial lectin FimH. For the irradiation experiments a FimH truncate, FimH tr , which resembles the adhesin domain of the complete FimH, was used [24]. FimH tr comprises of amino acids 1-160 of FimH and is terminated by a histidine tag His 6 . FimH tr has the same carbohydrate binding properties as FimH. Mass spectrometric analysis of FimH tr revealed m/z = 17839 (calcd. m/z = 17845). Solutions of FimH tr and the photophores 2, 12, and 13 were applied in a ratio of 5:1.
Samples were incubated at 37 °C to allow formation of the lectin-ligand complex and then irradiated at λ ≥ 320 nm. This led to 1:1-photo-crosslinked products with the corresponding masses (Table 2).
In addition to the mass spectrometric analysis, dot-blots were performed with FimH tr and the photoactive mannosides.
Affinity staining was carried out with a streptavidine-HRP conjugate and the chromogene 3,3'-diaminobenzidine (DAB). The biotin-labeled mannosides 12 and 13 gave violet spots on the nitrocellulose membrane when tested. Affino dot-blot with mannoside 2, that does not contain a biotin moiety, was negative, as predicted. In addition, control experiments were carried out, leading to the expected results in all cases. Interestingly, photoaffinity probes 12 and 13 seem to exhibit unequal affinity to FimH tr as suggested by the different intensity of color of the respective precipitates ( Figure 3). Affinity staining using western blots led to analogous results (not shown).

Conclusion
In conclusion, we have demonstrated the synthesis of new biotin-labeled photoactive mannosides for photoaffinitylabeling of the bacterial lectin FimH. The target molecules 12 and 13 were selected after docking studies based on the structure of FimH and according to binding studies employing type 1-fimbriated E. coli. Photo-crosslinking was tested with the model peptide angiotensin II and the regiochemistry of the insertion reaction could be solved by MS/MS studies. Furthermore, photoaffinity-labeling of FimH tr was successful and could be demonstrated by mass spectrometric studies as well as dot-blot analysis. The overall goal of this study is to identify mannose binding sites on the bacterial lectin FimH in solution. After photocrosslinking of the lectin with photoactive, biotin-labeled mannosidic ligands, proteolytic digestion of the products of photo-crosslinking, followed by affinity chromatography and mass-spectrometric analysis of the fragments, should allow the identification of the critical amino acid residues on FimH, according to the photoaffinity methodology. However, so far we have not been successful in an unequivocal mass-spectrometric analysis of any proteolytic digest, we have obtained so far. Thus, this study is currently continued in our laboratory.

Experimental Docking studies
Computer-aided modeling to predict binding of the various FimH ligands was carried out using FlexX flexible docking and consensus scoring as implemented in Sybyl 6.8 as described earlier [20]. Docking was based on published X-ray structures of the FimH CRD. This CRD was held fixed during the minimization, whereas the sugar ligand was allowed to change its conformation freely under the influence of the force field.

ELISA
ELISAs to determine IC 50 -values of the various FimH ligands were carried out with E. coli bacteria of strain HB101pPKL4 and mannan-coated microtiter plates as described earlier [21,22].