Production of FimH in mammalian cells

The low yields of WT FimH_LD or FimCH produced in E. coli (3-10 mg/L of bacterial culture) are well documented [18,19]. Given these challenges, we explored production of FimH using an Expi293 mammalian cell expression system (Fig 1A), wherein the protein is secreted into the cell culture medium using a eukaryotic signal peptide. Fusion of FimH to the mouse IgGκ signal peptide yielded protein that was correctly processed at the N-terminus (confirmed by mass spectrometry (Fig A in S1 Figs), also by X-ray crystallography described below). In contrast, use of the native FimH signal peptide resulted in a protein that was not correctly processed (Fig A in S1 Figs). Non-native N-linked glycosylation commonly occurs during heterologous expression of proteins in mammalian cells and can interfere with biological function as well as antigenicity of recombinant proteins. Non-native N-linked glycosylation on FimH_LD at residues N7 and N70 was removed via mutation of target Asparagine residues to Serine, which was selected as a conservative substitution. WT and conformationally locked FimH_LD (V27C L34C, described previously [19]), harboring N7S and N70S mutations to remove glycosylation, were expressed in 20 ml Expi293 cultures, which yielded 17 mg and 11 mg protein respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. FimH_LD produced in mammalian cells has comparable biophysical properties to material produced in E. coli.

WT or conformation stabilized FimH_LD were expressed in Expi293 cells along with mutants designed to remove non-native glycosylation on residues N7 and N70. (A) Batch purified proteins run on SDS-PAGE gel stained with Coomassie blue. C1 and C2 designate two clones of FimH_LD N7Q V27C L34C N70S and FimH_LD V27C L34C C1 that were evaluated for expression. (B). Near-UV CD spectra of FimH_LD produced in E. coli and mammalian expression systems. Spectra of FimH_LD WT (blue) and V27C L34C (red) are shown. (C) Affinity of WT and V27C L34C FimH_LD produced in E. coli or mammalian cells for alpha-D-mannopyranoside by FP. (D) Thermal stability of WT and V27C L34C FimH_LD produced in E. coli or mammalian cells. (E) Mouse study design: CD-1 mice were immunized 3 times with 10 µg FimH proteins with QS21 adjuvant. (F) Sera were analyzed for the ability to block FimH-expressing E. coli binding to yeast mannan; bars represent geometric mean IC₅₀ and 95% confidence intervals. Statistical significance (p-value) of differences in responses between groups was determined using an unpaired t-test with Welch’s correction applied to log-transformed data; the bars and asterisk illustrate the significance of the difference in response for comparisons. Tabulated IC₅₀ values are shown in Table A in S1 Text.

https://doi.org/10.1371/journal.ppat.1012325.g001

FimH _LD produced in mammalian cells has comparable biophysical properties to material produced in E. coli

Mammalian and E. coli derived FimH_LD WT and disulfide lock mutant V27C L34C were further characterized using biophysical and biochemical assays. The E. coli and mammalian derived FimH_LD WT and mutant proteins had essentially identical near-UV circular dichroism (CD) spectra (Fig 1B), indicating that the respective tertiary structure of these conformationally distinct proteins is very similar. Ligand binding affinity was determined using a direct binding fluorescence polarization (FP) assay with a fluorescein-conjugated octylbiphenylmannopyranoside (BPMP) ligand as described previously [19,43]. E. coli and mammalian produced material bound BPMP ligand with similar affinity (Fig 1C); thermal stability, as determined by differential scanning calorimetry, was also comparable (Fig 1D).

Design of a full-length donor strand complemented FimH (FimH-DSG)

To mimic the presentation of FimH on the assembled fimbrial tip [20], and to understand the contribution of FimH_PD in ability to elicit anti-FimH antibodies with the ability to prevent bacterial binding to mannose, we considered production of a full length FimH antigen. Full length FimH cannot be produced in E. coli without the chaperone FimC or as part of a pilus [16,19,44,45]. In previous studies in humans and non-human primates, full length FimH in complex with FimC was used as an immunogen [23–25,33,34]. Whether the FimH_PD is important for optimal immunogenicity is unknown, therefore the design of a full length FimH was explored to compare immunogenicity with that of FimH_LD alone. Sauer et al previously produced a stable, isolated FimH molecule by displacing FimC in FimCH with a FimG donor strand peptide (FimG residues 1-14) [46]. This approach requires coexpression of FimH with the chaperone, FimC. In contrast, Barnhardt et al produced a stable, single chain FimG donor strand complemented FimH in E. coli using a 4-residue amino acid linker (DNKQ), enabling production of full length FimH without a chaperone [17]. The single chain concept was explored further in the current study. Donor-strand complemented, full-length FimH proteins (FimH-DSG) were produced by attaching a 14-mer donor strand peptide from FimG to the C-terminus of the FimH_PD via a Glycine-Serine linker of various lengths, as well as the previously described DNKQ linker [17]. Flexible polypeptide linkers consisting of Glycine and Serine are often used in construction of multidomain proteins as these flexible and hydrophilic spacer sequences prevent formation of secondary structure between protein domains, reducing the possibility that linkers will interfere with the folding or function of the target protein [47]. Constructs were screened in mammalian cells and expression levels were similar, as assessed by SDS-PAGE (Fig B in S1 Figs). FimH-DSG with a 7-residue Glycine-Serine linker (GGSSGGG) was selected as optimal, based on structural analysis. Non-native glycosylation in FimH_LD was prevented by incorporation of N7S and N70S mutations as described above and an additional mutation to Glutamine was also introduced at N228 in the FimH_PD. Analysis by mass spectrometry confirmed that FimH-DSG retained a single glycosylation site at N235 in the exploratory antigen initially characterized in this study (Fig A in S1 Figs). Note, fully aglycosyl variants of FimH-DSG were subsequently generated following introduction of an additional mutation at N235 and expression in ExpiCHO cells (described in brief below).

The proposed mechanism of action of FimH-containing vaccines to prevent UTIs is via inhibition of bacterial binding to the urinary tract epithelium [25]. We developed a live, whole E. coli binding inhibition assay to quantify the ability of anti-FimH antibodies to prevent binding of fimbriated bacteria to a mannosylated yeast mannan surface (mimicking the mannosylated FimH receptor on the surface of bladder epithelial cells), based on previous work [48]. Using this assay, the relative ability of E. coli and mammalian derived FimH proteins to elicit inhibitory serum antibodies in mice was evaluated (Fig 1E and 1F). Inhibitory titers elicited by E. coli and mammalian derived WT FimH_LD proteins in mice were equivalent, although the potency of the stabilized lock mutant FimH_LD V27C L34C produced in mammalian cells was slightly lower than that of the E. coli produced FimH_LD V27C L34C. Disulfide bond formation may differ in the periplasm of E. coli compared to the mammalian cytoplasm [49]. Inhibitory titers elicited by full length FimH proteins (FimH-DSG, only produced in mammalian cells) including the previously described FimCH protein [50] were markedly higher than FimH_LD proteins in terms of percentage of responders and geometric mean IC₅₀ values (Fig 1F and Table A in S1 Text).

Rational design of FimH variants stabilizing the low-affinity, open conformation

We employed in silico analysis to identify novel variants predicted to stabilize FimH_LD in a low affinity conformation, which is unable to bind its cognate mannose receptor. Crystal structures of FimH in complex with fimbrial structural proteins in the absence of ligand or presence of D-mannose are shown in Fig 2A, illustrating the differences in FimH_LD conformation. The crystal structure of full length FimH in complex with fimbrial structural proteins (PDB ID 3JWN (20)) was used as a model. Suggested amino acid substitutions identified from this analysis were introduced into either FimH_LD or FimH-DSG (Fig 2B), harboring the same mutations to prevent glycosylation described above (N7S, N70S in FimH_LD and N7S, N70S and N228Q in FimH-DSG). To stabilize FimH_LD in an open conformation, the following design strategies were applied: 1) nonpolar residues exposed to solvent in the pre-bound state but buried within the protein interior in the bound structure were mutated to polar or charged residues, disfavoring the high-affinity, closed conformation of FimH_LD; 2) disulfide linkages were introduced between residue pairs in close proximity in the pre-bound state, but not in the bound conformation; 3) mutation of glycine residues having a backbone ϕ-angle < 0° in the pre-bound state but > 0° in the bound structure to prevent closure of the ligand binding site; 4) a full length, single chain FimH-DSG was designed (described above); and 5) cavity filling mutations designed to stabilize the interface between FimH_PD and FimH_LD were introduced separately into full length FimH-DSG. An alignment of all FimH mutants that were evaluated, in comparison to wild type FimH sequence from J96, is provided in S1 File.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Rational design of FimH mutations stabilizing the native state.

(A) Structures of FimH in unbound form (PDB structure 3JWN) and bound to mannoside ligand (PDB structure 1KLF). In pili, unbound FimH (left), complexed with the donor strand of FimG, adopts a compact conformation that binds the FimH cognate receptor, the terminal mannose moiety of glycosylated proteins, with low affinity. Upon binding a mannose moiety (right), the FimH_LD and FimH_PD separate, sidechains (colored in green) flip from protein interior to surface, and backbones of Gly residues (colored in blue) exhibit large conformational changes. Residues shown in blue and green were targeted for mutagenesis. (B) Strategies employed to stabilize FimH conformation. Unbound (grey) and bound (green) full length FimH structures with residues targeted by mutagenesis are highlighted.

https://doi.org/10.1371/journal.ppat.1012325.g002

Screening of FimH _LD and FimH-DSG mutants in in vitro assays

Sixty-four mutant and WT versions of FimH_LD and FimH-DSG proteins were expressed in Expi293 cells as described above. Purified proteins were evaluated in a series of in vitro and in vivo studies (Fig 3A). Mutants that were expressed at low levels were excluded from further evaluation. Binding affinities (K_d) of FimH mutants for mannoside ligand were determined by FP assay using BPMP as described above. K_d values for a subset of FimH_LD and FimH-DSG mutants relative to WT are shown in Fig 3B (data for additional mutants can be found in Table B in S1 Table). The sequence of FimH_LD WT is derived from E. coli UTI isolate J96 [51]. V27A is a natural variant that is associated with virulent UTI isolates and those associated with Crohn’s Disease [11,52]. Introduction of the single point mutant V27A did not alter mannoside ligand binding affinity relative to FimH_LD WT, while combining single or double glycine loop mutations at G15 or G16 positions with V27A significantly impaired ligand binding (Table B in S1 Table). Ligand binding was completely lost in the triple mutant, FimH_LD G15A G16A V27A (FimH_LD TM, K_d > 2000 nM). Ligand binding affinity of FimH-DSG WT was reduced more than 100-fold compared to FimH_LD WT, in agreement with the allosteric role of the pilin domain in regulating ligand binding by the lectin-binding domain. FimH-DSG V27A also had a two-fold lower ligand-binding affinity, compared to FimH-DSG WT (S2 Table B in S1 Table). This is consistent with previous data showing that a FimH mutant containing A27 has the propensity to adopt a less-active state that binds mannose with low affinity [11]. Like FimH_LD, FimH-DSG mutants containing V27A and Glycine loop mutations have substantially reduced ligand binding activity. Altogether, our data suggest that the flexible Glycine loop plays a stabilizing role in ligand binding. Mutations in this loop result in FimH adhesins with poor affinity for the synthetic BPMP ligand used as representative of cognate mannosylated host cell glycoprotein receptors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Identification of FimH mutants with improved thermal stability and reduced mannoside ligand affinity.

(A) 64 FimH variants were screened in vitro for ability to bind mannose, thermal stability, and conformation. A subset of constructs was screened for immunogenicity in mice. (B-C) Biochemical characterization of purified FimH_LD and FimH-DSG mutants. (B) relative average binding affinities of FimH mutants to mannoside ligand. Note, the assay limit of detection was ~2000 nM. (C) filled circles display average melting temperatures of each mutant. Open circles denote melting temperature of FimH protein in the presence of mannoside ligand. Tabulated K_d and T_m values for all mutants are in Tables B and C in S1 Text s respectively.

https://doi.org/10.1371/journal.ppat.1012325.g003

To further evaluate ligand binding abilities of FimH variants, a SYPRO orange-based differential scanning fluorimetry assay was used [19]. The temperature at which 50% of the protein is unfolded (T_m) was determined for all mutants in apo state as well as in the presence of methyl alpha-D-mannopyranoside, a derivative of alpha-D-mannose that binds to FimH with micromolar affinity [53]. The T_m of apo protein and its corresponding T_m shift (∆T_m) in ligand bound condition are summarized in Table C in S1 Table. FimH-DSG WT proteins exhibited significantly higher T_m (71.3 ºC) compared to FimH_LD WT (61.5 ºC). T_m shifts of the FimH mutant proteins from apo to mannoside ligand bound conditions are in concordance with the changes seen in mannoside ligand K_d measurement (Fig 3B and Table B in S1 Table). Of note, consistent with published data the V27C L34C disulfide lock mutant [19] exhibited significantly lower thermostability (T_m) compared to FimH_LD and FimH-DSG WT proteins, while retaining mannoside ligand binding activity (Kd of 30-60 nM). In contrast, combinatory mutations containing V27A and Glycine loop mutations at G15 and G16 positions resulted in FimH mutants that retained thermostability while losing their ability to bind mannoside ligand (Kd >2000 nM). Together with our binding affinity data, these data suggest that combining V27A and Glycine loop mutations stabilizes FimH proteins in a low affinity state that is largely incapable of ligand binding.

In previous work, WT and conformationally locked FimH_LD mutants were found to have distinct tertiary structures [19]. The secondary and tertiary structures of selected FimH_LD and FimH-DSG proteins were therefore examined by CD (Fig C in S1 Figs). The far-UV CD spectrum (secondary structure) of FimH_LD WT expressed in mammalian cells is consistent with previously published data. Overall, the secondary structures of the FimH_LD or FimH-DSG mutants are highly similar to WT proteins (Fig C in S1 Figs), suggesting that the overall secondary structure is not altered in these mutants. The near-UV CD spectra (tertiary structure) of most FimH_LD mutants except for V27A (Fig C in S1 Figs), however, were quite different from that of FimH_LD WT and are more similar to the previously reported FimH_LD V27C L34C and V27A R60P mutants which assume an open conformational state [19]. On the other hand, all FimH-DSG proteins including WT and the V27A, G15A, G16A triple mutant (or TM) had highly similar secondary and tertiary structures, resembling the open state of FimH_LD (Fig C in S1 Figs). Overall, the CD characterization is consistent with mannoside ligand binding data described above, which suggests these mutants are stabilized in an open conformation that has low mannose binding affinity.

Together, these data suggest that the Glycine loop mutations constrain FimH_LD in an open conformation, while the FimH-DSG constructs including the corresponding reference WT variant, remain in a conformation that remains unchanged upon introduction of stabilizing mutations that eliminate ligand binding.

FimH-DSG mutants induce antibodies with superior ability to inhibit bacterial binding than FimH _LD mutants in mice

To evaluate the relative ability of selected FimH mutants to elicit inhibitory antibodies, mice were vaccinated with each mutant protein and the collected sera were used to assess the ability to prevent bacterial binding in vitro. Mice were immunized three times with 10 µg FimH protein combined with QS21 adjuvant (Fig 4A) and sera from post dose 2 and 3 time points were evaluated in the yeast mannan E. coli binding inhibition assay described above (Fig 4B and Table D in S1 Text).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. FimH-DSG mutants induce antibodies with superior ability to inhibit bacterial binding compared to FimH_LD mutants in mice.

(A) CD-1 mice were immunized 3 times with 10 µg FimH with QS21 adjuvant. Sera were analyzed for the ability to block FimH-expressing E. coli binding to yeast mannan. (B) Inhibitory titers were determined from serial dilution of sera from vaccinated mice and represent the reciprocal of the dilution of serum at which 50% of bacteria remain bound to the plate and are shown for post dose 2 and post dose 3 timepoints. Statistical significance (p-value) of differences in responses between groups was determined using an unpaired t-test with Welch’s correction applied to log-transformed data; the bars and asterisk illustrate the significance of the difference in response between groups. Tabulated IC₅₀ values are shown in Table D in S1 Text.

https://doi.org/10.1371/journal.ppat.1012325.g004

At post dose 3, mice immunized with novel stabilized FimH_LD mutants G15A G16A V27A, G16A V27A, and the previously described mutant V27A R60P [19] had higher responder rates and increased binding inhibition (p <0.05) compared to FimH_LD WT (Fig 4B and Table D in S1 Text). Other FimH_LD mutants (G15A V27A, G16P V27A, V28C N33C) did not significantly enhance functional immunogenicity of FimH_LD. Thus, several mutants designed to enhance functional immunogenicity of FimH_LD by constraining FimH_LD in an open conformation led to improved responses relative to FimH_LD WT, confirming previous findings [54]. Following vaccination with 2 doses of FimH_LD and FimH-DSG proteins, significantly more animals yielded inhibitory titers in the groups vaccinated with FimH-DSG compared to FimH_LD (Fig 4B and Table D in S1 Text). This trend was sustained at PD3, where 95%-100% of mice responded in groups vaccinated with FimH-DSG V27A, FimH-DSG G15A V27A or FimH-DSG G15A G16A V27A (henceforth FimH-DSG TM) and tended to have higher GMC IC₅₀ values relative to FimH_LD mutants (p<0.05) (Fig 4B and Table D in S1 Text). In conclusion, FimH-DSG mutants elicit higher inhibitory antibody responses compared to FimH_LD mutants.

The ability of FimH-DSG V27A, G15A V27A double mutant and FimH-DSG TM to elicit inhibitory antibodies were similar. However, during large scale purification of FimH-DSG WT and TM from ExpiCHO cells for further characterization, a tendency for aggregation was observed for FimH-DSG WT and characterization by analytical size exclusion chromatography revealed the presence of high molecular weight complexes (Fig E in S1 Figs). We hypothesized that during CHO fermentation, and upon secretion into the culture media, the FimH-DSG WT binds glycan molecule(s) released from the surface of the host CHO cells. To evaluate this further, samples were analyzed by High pH Anion-Exchange Chromatography with Pulsed Amperometric (electrochemical) Detection (HPAEC-PAD). FimH-DSG WT preparations contained numerous monosaccharides (Fig E in S1 Figs). In contrast, FimH-DSG TM was bound to comparatively fewer glycan moieties (Fig E in S1 Figs). Furthermore, it is entirely possible that the low monosaccharide content that was detected represents sugar moieties of the N-glycan present on N235. FimH-DSG TM did not tend to aggregate or form high molecular weight complexes. Therefore, as it can be purified to homogeneity, at a high yield, we used this mutant for our subsequent studies.

Production of aglycosylated FimH-DSG TM

Removal of the N-glycan present on residue N235 of FimH-DSG TM led to additional glycosylation on residue N228. To avoid these issues, we explored removal of glycosylation by introducing the following combinations of mutations: N228S N235S, T230A T237A, N228G N235G, and N228Q N235Q. Proteins were expressed transiently in 2L ExpiCHO cell cultures and yields were between 69 mg (N228G N235G) and 311 mg (N228S N235S). Aglycosylation mutations did not impact the ability of inhibitory Mabs to bind relative to glycosylated FimH-DSG TM (S5 Table), and the capability of aglycosylated variants elicit adhesin blocking antibodies does not appear to differ from the glycosylated parent (Fig D in S1 Figs). An alignment of FimH-DSG TM and aglycosylation mutants, along with other FimH sequences used in this study, is shown in S1 File.

Structural characterization of FimH-DSG TM by X-ray crystallography

Among all conformationally stabilized FimH-DSG mutants, FimH-DSG TM was of most interest as it harbors the two Glycine loop mutations G15A and G16A designed to conformationally lock the mannose site in the ‘open’ state. To verify the conformation of FimH-DSG TM, the crystal structure of a FimH-DSG TM protein containing three glycosylation site mutations (N7S, N70S, and N228Q) was solved by X-ray crystallography at 1.9 Å resolution (Fig 5A and Table F in S1 Text). There are four molecules of FimH-DSG TM in one asymmetric unit which share high similarity, therefore, structural analysis was done with protomer A. Introduction of the glycosylation site mutations did not appear to alter either the local or global conformations of the FimH_LD and FimH_PD. Comparison of the structure to previously determined structures of FimH_LD confirmed that FimH-DSG TM lectin domain adopts a compressed structure with an open mannose-binding pocket (Fig 5C). Clear electron density was obtained for the Glycine loop, wherein the G15A G16A mutations were introduced which induced a local change in the loop conformation (Fig 5B and 5C). A15 projects towards the inside region of the clamp loop, providing rigidity, which causes the loop to widen at the end by ~1.2 Å (I13 Cα-S17 Cα) compared to the WT residue (G15), thus stabilizing the low affinity state. Additional rigidity is introduced by the G16A substitution, as it projects towards the bulk solvent. By comparison, these residues adopt very different conformations in the closed, high-affinity state [55]. The overall structure of FimH-DSG TM resembles the low affinity conformation of FimH_LD (Fig 5D) [16,18,]. In comparison to previously published structures, the donor strand complemented FimH_PD did not exhibit any conformational changes. The backbone of the exogeneous 7-residue Glycine-Serine linker that connects the FimH and FimG peptide could largely be resolved, and does not impact the conformation of the lectin or pilin domains in comparison to previously published structures (Fig F in S1 Figs) [18].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The lectin domain of FimH-DSG TM adopts an open conformation.

(A) Overall structure of FimH-DSG TM from x-ray crystallography data (PDB code 8V3J). (B) Superimposition of ligand binding sites of FimH-DSG TM (orange) and WT FimH from a previously published structure of native pili (PDB code 3JWN, light grey) shows remodeling of the Glycine loop due to G15A G16A mutations. The widening of the loop between Ile13 Ca-Ser17 Ca is shown in a dotted line. (C) Electron density and atomic model of Glycine loop in FimH-DSG TM. (D) Superimposition of FimH-DSG TM (orange), WT FimH in apo (PDB code 3JWN, light grey) and trimannose-bound (PDB code 6GTV, slate) forms shows that the ligand binding site of FimH-DSG TM adopts an open conformation resembling the apo state but not the ligand-bound, closed conformation.

https://doi.org/10.1371/journal.ppat.1012325.g005

Inhibitory antibodies elicited by FimH-DSG map to novel epitopes

As described above, FimH-DSG WT and derived mutants induced superior inhibitory antibody responses in mice compared to equivalent FimH_LD variants. To further characterize inhibitory antibodies induced by FimH-DSG, monoclonal anti-FimH antibodies were developed and screened by the yeast mannan E. coli binding inhibition assay, which resulted in the identification of 10 unique clones (Table G in S1 Text).

A series of competition experiments using Bio-layer interferometry (BLI) was undertaken to facilitate Mab classification and epitope mapping. As comparators, two previously described inhibitory Mabs, 475 and 926, which bind to overlapping epitopes on the ligand binding interface, were included [48,54]. An octyl amino mannopyranoside compound was selected as the synthetic ligand for these experiments as it is expected to have a higher affinity for FimH than shorter chain alkyl α-D-mannosides [53,56], and a crystal structure of the related octyl carboxy mannopyranoside in complex with FimH in complex is available [57]. BLI experiments were performed with and without octylmannopyranoside ligand, to assess whether there is competition between the ligand and Mab binding to FimH_LD (Fig G in S1 Figs and Table G in S1 Text), and ability of antibodies to bind FimH_LD and FimH-DSG mutants (Table H and Table I in S1 Text). These experiments identified four distinct inhibitory epitopes. Site 1 was recognized by 6 novel Mabs, as well as reference Mabs 926 and 475, which effectively block ligand binding. Three antibodies were unable to block ligand binding and bound to two distinct epitopes (site 2 and site 3, Mabs 327-3 and 329-2, and 445-3 respectively). Site 4 was recognized by a single Mab (440-2) that bound exclusively to conformation-stabilized FimH_LD. Note, this Mab was derived from mice immunized with the conformationally locked construct, FimH-DSG V27C L34C. Individual Mab binding kinetics are shown in Table G in S1 Text alongside their inhibitory concentrations in the E. coli binding inhibition assay.

Fab fragments derived from representative antibodies from each new site (329-2 (site 2), 445-3 (site 3) and 440-2 (site 4)) were selected as representatives for subsequent epitope mapping by cryo-EM. The small size of FimH was an impediment to high resolution structural analysis of FimH-Fab binary complexes, which yielded only low-resolution structures that could not be modeled. To circumvent this limitation, a combinatorial approach was taken in which pairs of Fabs with non-overlapping epitopes were combined with FimH to form ternary complexes, increasing the overall particle mass to 125 kDa. Using this ternary complex strategy, a cryo-EM structure of FimH-DSG TM with Fabs 329-2 and 445-3 was determined at 3.11 Å (Figs 6A and Fig H in S1, S8 Figs, and Table J in S1 Text), with the final map comprising FimH_LD and the antigen-binding portions of each Fab, allowing molecular modeling of each component and identification of the molecular details of the respective epitopes. The two Fabs were distinguished in part by comparison of their sequences to the corresponding side chain density in the map, but to build additional confidence, the structure of an overlapping FimH-DSG TM-Fab ternary complex, in which the 329-2 Fab was replaced with 440-2, was determined by cryo-EM (Figs 6B and Fig I in S1 Figs, and Table J in S1 Text). The new map from a 4.2 Å reconstruction enabled building of Fabs 440-2 and 445-3, providing confidence in the Fab assignments and epitope definitions, and allowed mapping of the remaining 440-2 epitope. Surprisingly, Mab 440-2 binds to the tip of FimH_LD, which covers the mannose binding site and largely overlaps with the known site 1. Binding of Mab 440-2 (site 4) therefore directly competes with mannose binding, conveying an orthosteric inhibition mechanism for blocking of bacterial adhesion, like that observed for the competitive Mab 475(48). However, being derived from a conformationally locked FimH-DSG mutant antigen distinguishes Mab 440-2 from the other inhibitory Mabs. Mab 440-2 preferentially binds to the low affinity open state of FimH_LD, but not WT FimH_LD (Table H in S1 Text), suggesting that it uses a novel inhibitory mechanism. Indeed, the structure of the ligand binding site of 440-2-bound FimH_LD matches the open conformation, largely distinct from the observed conformation in the ligand bound high-affinity ‘closed’ state (Fig 7A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Identification of novel inhibitory epitopes on FimH-DSG TM.

(A) Structure of FimH-DSG TM in complex with 329-2 and 445-3 Fabs solved by cryo-EM. Top image shows the cryo-EM map, colored by chain, as indicated. Insets show respective epitope interfaces, with each Fab shown in transparent surface representation and FimH in cartoon representation, with residues contributing to each epitope surface shown in stick representation (alpha carbons are represented by transparent spheres). (B) CryoEM structure of FimH-DSG TM in complex with 440-2 and 445-3 Fabs, colored by chain, as indicated. Inset shows the epitope interface, with the Fab shown in transparent surface representation and FimH in cartoon representation, with residues contributing to each epitope surface shown in stick representation (alpha carbons are represented by transparent spheres). (C) Sequence of FimH (with stabilizing mutations underlined) with residues contributing to each Fab’s epitope highlighted and colored by their respective Fabs as in A and B. (D) Cartoon representation of the FimH_LD in the FimH-DSG TM crystal structure. Epitopes identified in this study and in previous work by others (Mab 926, 824, 475 and 21) are highlighted, with participating residues shown as spheres and colored as indicated. Numbered gray residues indicate coincidence between two epitopes (1 = Mab 926 and Mab 475, 2 = Mab 926 and Mab 445-3, 3 = Mab 475 and Mab 824, 4 = Mab 440-2 and Mab 926).

https://doi.org/10.1371/journal.ppat.1012325.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Structural basis of inhibitory mechanisms of different Mabs.

(A) Superimposition of FimH WT bound to trimannose (PDB code 6GTV, blue) and FimH-DSG TM (orange) cryo-EM structure resolved from the complex with Fabs of 440-2 (green surface) and 445-3 (not shown). (B) Superimposition of FimH WT bound to trimannose (PDB code 6GTV, blue) and FimH-DSG TM (orange) cryo-EM structure resolved from the complex with Fabs of 329-2 (purple) and 445-3 (dark blue). Surface representations for the two Fabs are shown. Side chains from FimH-DSG TM that are part of the Mab epitopes are shown in stick format. In (A) and (B), shift of the clamp loop and Glycine loop from FimH-DSG TM to mannose-bound FimH WT is highlighted by dashed lines. Conformations of the important FimH epitope loops for 440-2, 329-2, and 445-3 binding are shown as orange (FimH-DSG TM structure) and blue (FimH WT-mannose bound structure) lines.

https://doi.org/10.1371/journal.ppat.1012325.g007

Epitopes of Mabs 329-2 (site 2) and 445-3 (site 3) were mapped to the opposite sides of FimH_LD. Despite not overlapping with ligand binding pocket, the 329-2 and 445-3 epitopes are composed of residues in adjacent regions with conformations that change significantly during the switch between inactive and active states. In particular, Mab 329-2 recognizes a novel inhibitory epitope comprised of residues T5, S7, T9, A10, Y21 and N23. These residues are connected to the Glycine loop that harbors the G15A and G16A mutations. As a result, binding of Mab 329-2 stabilizes the Glycine loop in the open conformation that has low mannose affinity (Fig 7B). Binding to the other side of FimH_LD, Mab 445-3 interacts with residues including R132, T134, S139, D140 and D141 from the clamp loop in its open conformation, which shifts 4.4 Å away from the compact-mannose-bound conformation seen in the closed form (PDB ID 6GTV, Fig 6B). In addition, S80, S81, Y82 and P91 were also mapped to 445-3 binding site. These residues were revealed in previous work as being part of a critical epitope for Mab 824, an inhibitory antibody that allosterically prevents transitions of FimH from low-affinity to high-affinity states [58]. Taken together, Mab 445-3 likely blocks bacterial adhesion via a combinatory allosteric mechanism which modulates both the global state transition as well as the local ligand binding site. These new epitopes described herein (mapped onto the FimH-DSG TM sequence in Fig 6C), along with previously identified FimH Mab epitopes are highlighted on the structure of FimH-DSG TM in Fig 6D.

Of note, none of the binding sites identified for these three Fabs overlap with the conformation stabilizing mutations introduced into the FimH-DSG TM vaccine candidate (S2 File). In addition, the residues of the mapped epitopes exhibit little to no variation across UTI isolates (S2 File). Collectively, these results broaden the repertoire of FimH_LD functional epitopes, highlighting the diverse interference mechanisms by which antibodies contribute to FimH vaccine efficacy.

Ethics statement

Mouse immunogenicity studies were performed at Pfizer, Pearl River, NY, which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). All procedures performed on mice were in accordance with local regulations and established guidelines and were reviewed and approved by an Institutional Animal Care and Use Committee (IACUC). The work was in accordance with United States Department of Agriculture Animal Welfare Act and Regulations and the NIH Guidelines for Research Involving Recombinant DNA Molecules, and Biosafety in Microbiological and Biomedical Laboratories.

Experimental design

This work aimed to develop a E. coli vaccine candidate by applying a computational and experimental design strategy to enhance the stability of the FimH lectin binding domain (FimH_LD) and single chain full length FimH (FimH-DSG). Mutations were designed to introduce amino acid substitutions in FimH_LD to stabilize the protein conformation in a low affinity state. FimH_LD and FimH-DSG mutants were expressed in mammalian cells and purified mutant proteins were characterized by thermal stability and rate of dissociation, in comparison to wild type FimH_LD. The crystal structure of FimH-DSG G15A G16A V27A (FimH-DSG TM) was resolved to confirm the mutant’s open conformation. Monoclonal antibodies raised against FimH-DSG were evaluated in competition experiments to group into non-overlapping epitope bins. Representatives of each distinct bin were complexed with FimH-DSG TM and analyzed by Cryo-EM to determine epitopes.

Expression and purification of WT FimH _LD from E. coli

DNA encoding J96 FimH_LD sequence was cloned into a pET28 vector and E. coli BL21(DE3) cells were transformed with the resulting construct. Expression, extraction and purification of FimH_LD or FimCH from the E. coli periplasm was performed as previously described [50,72].

Expression and purification of WT FimH and mutants in mammalian cells

DNA encoding J96 FimH sequence and conformation-stabilized mutants was codon optimized for mammalian cells and cloned in frame behind sequence encoding mouse IgGκ or native FimH signal peptide, along with a C-terminal 8xHis tag in pcDNA3.1(+). FimCH was expressed from a pBudCE4.1 plasmid containing a CMV promoter to drive expression of FimC with a C-terminal 8xHis tag and a second promoter, EF1α, to drive expression of untagged FimH. For biochemical assays and mouse immunogenicity studies, endotoxin-free DNA was transiently transfected into Expi293 cells (ThermoFisher Scientific) according to manufacturer’s instructions. Supernatants were filtered through a 0.22 µm filter unit (Nalgene sterile disposable filter with PES membranes, ThermoFisher Scientific). Nickel Sepharose excel (Cytiva) was incubated with supernatants overnight at 4ºC and purification was performed according to manufacturer’s instructions. The eluate was loaded onto a S200 16/600 column in 50 mM TrisHCl pH 8.0, 300 mM NaCl. Fractions containing pure protein were pooled.

For X-ray crystallography and cryoEM, FimH-DSG TM was produced in ExpiCHO cells (Thermo Fisher Scientific) according to manufacturer’s instructions. The purification procedure is described in S1 Text.

Computational design of FimH mutants stabilizing the native state

The crystal structure of native full length FimH in complex with fimbrial structural proteins (PDB ID 3JWN) was used as a model for the low-affinity, open conformation. The crystal structure of FimH_LD in complex with butyl α-D-mannoside was used as a model for the high-affinity, closed conformation (PDB 1UWF). Schrodinger BioLuminate (release 2017–2) was applied to analyze structural differences between the two states and to identify residue locations for mutations. Nonpolar residues that are exposed to solvent in the pre-bound structure and buried in the bound state were changed to polar or charged residues. Residue pairs selected for mutations to cysteine to form disulfide bonds that stabilize the native state were proximate (C_β-C_β ≤ 5 Å) in the pre-bound state and distant (C_β-C_β ³ 10 Å) in the bound conformation. Gly residues that have a negative backbone ϕ-angle in the pre-bound state and a positive backbone ϕ-angle in the bound state were mutated to Ala or Pro. An optimal Gly-Ser linker to connect the C-terminus of FimH and the N-terminus of FimG was identified using the Linker Modeler implemented in Molecular Operating Environment (MOE v2018, Chemical Computing Group).

Characterization of WT and mutant FimH proteins

WT and mutant FimH proteins were characterized by fluorescence polarization, circular dichroism and differential scanning fluorimetry modified from Rabbani et al [19] and are described in the S1 Text.

Antigenicity assay

Inhibitory Mabs 299-3, 304-1 and 440-2 (developed in-house) were used to confirm the conformational state of FimH mutants; 299-3 and 304-1 bind to similar epitopes as reference Mab 475 and 926 [37,48] while 440-2 recognizes a different epitope and appears to preferentially bind FimH_LD in an open conformational state. Mutants that maintain the same structure as WT FimH_LD are expected to bind all antibodies. Octet HTX from ForteBio was used for all the kinetic real-time biomolecular interaction experiments to measure antibody reactivity with each mutant. Experiments were carried out at 30 °C with 1000 rpm agitation in 96-well black plates containing 240 µl per well. Ni-NTA biosensors were equilibrated in buffer containing 1x PBS buffer containing 0.5% BSA and 0.05% Tween 20 (PBT) before allowing them to load his-tagged FimH mutant proteins at 5 µg/ml for 5 min. FimH loaded biosensors were allowed to reestablish baseline in PBT for 3 min before allowing them to associate with antibodies from different bins at 5 µg/ml for 5 min. Octet data analysis software was used for kinetic analysis of association step and obtain response in nm shift (tabulated).

FimH whole cell binding inhibition assays

CFT073 (ATCC) was serially passaged in 10 ml of Luria Bertani broth in static growth conditions at 37°C to enrich for FimH expression. Surface expression reaching ≥95% was confirmed via flow cytometry using anti-FimH Mab 926 [48]. Prior to the assay, 384 well white Maxisorp plates (Nunc) were coated with 20 µg/ml of yeast mannan (Sigma-Aldrich) and blocked in 1% BSA (Thermo). FimH-expressing E. coli cells (confirmed by flow cytometry using an anti-FimH antibody) were then incubated with a titration of vaccinated mouse sera and controls. Sera were diluted in PBS + 0.1% BSA and titrated 2.5-fold for 7 points, starting at 1:100. After 45 minutes at 37 ⁰C, the mixture was added to the plate and incubated for 45 minutes at 37⁰C before washing away any unbound cells. A titration of anti-FimH_LD rabbit sera was used as an internal control on every plate. Specificity of bacterial binding to mannan was established by the inclusion of Methyl α-D-mannopyranoside (Sigma) as a negative control, which reduced binding by >95% at 50 mM levels. Adherent cells were measured with a luminescent probe BacTiter Glo (Promega) and read on a Clariostar Plus plate reader. IC₅₀ inhibition values were interpolated using sigmoidal dose response variable-slope curve fitting (Graphpad Prism). Titers are the reciprocal of the serum dilution at which half-maximal inhibition is observed. Responders were defined as those with ≥50% inhibition at the starting dilution, had a defined IC₅₀, a positive hillslope, r² ≥0.80 and at least two points trending towards binding inhibition. The limit of detection (LOD) is designated as a dilution of 50, which is half of the maximum dilution. The statistical significance (p-value) of differences in responses between groups was determined using an unpaired t-test with Welch’s correction applied to log-transformed data.

An alternative version of the assay was used for only Fig 1 and involved detection using a directly labelled anti-O25b-AF488 antibody and an O25b clinical E. coli isolate PFEEC0547 (collected as part of the ATLAS surveillance program).

Animal studies

Animal studies were conducted according to Pfizer local and global Institutional Animal Care and Use Committee (IACUC) guidelines at an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International-accredited facility.

FimH murine immunogenicity studies

6–8-week-old female CD-1 mice were obtained from Charles River Laboratories. For each group of mice, 20 animals were immunized three times subcutaneously with 10 µg FimH protein mixed with 20 µg Quillaja Saponaria-21 (QS-21) from a 5.1 mg/ ml QS-21 stock solution containing 5 mM Succinate, 60 mM NaCl, 0.1% PS80, pH 5.6. Mice were bled 2 weeks following immunization. Blood was withdrawn in 3.5-mL serum tubes at each time point and spun in a centrifuge at 3,000 rpm for 10 min. The serum fractions were collected and stored in cryovials.

Anti-FimH Monoclonal antibody production

Anti-FimH Mabs 926 and 475 were described previously [48,73]. DNA sequences encoding heavy and light chain sequences provided in the patent [73] were codon optimized by Blue Heron and cloned into a pTT5 human IgG1 vector. 50 ml Expi293 cells (ThermoFisher) were transfected according to manufacturer’s instructions with heavy and light chain encoding sequences. Culture supernatants were harvested by centrifugation and antibodies were purified using Magne Protein A beads (Promega) according to manufacturer’s instructions. Antibodies were dialyzed in cold PBS overnight in a 10,000 molecular weight cutoff cassette, with 3 buffer exchanges.

CD-1 mice immunized with purified WT FimH-DSG or FimH-DSG V27C L34C proteins in combination with QS21 adjuvant described above received a final intraperitoneal boost of 10 µg mixed with 20 µg QS21 FimH protein 4 days before fusions. Spleens cells from mice with high titers were harvested and fused with the myeloma P3X63-Ag8.653 cell line using polyethylene glycol (P7306, SIGMA HYBRI-Max). Fused cells were cultured in 96 well plates at 37 °C, 8% CO₂ in DMEM containing HAT supplement (21060-017 Gibco). After 10 days in culture, hybridomas were screened by enzyme-linked immunosorbent assay (ELISA) using Maxisorp high binding 96 well plates (442404, Thermo Fisher Scientific) coated with 100 ng of FimH-DSG WT protein. Positive hybridomas secreting anti-FimH antibody were subcloned and clonality was assessed by sequencing. Over 300 parent hybridoma Mabs were screened for the ability to inhibit E. coli binding to yeast mannan using a single point titration assay. Following clonal expansion and competitive binding (binning) experiments (described in S1 Text), 5 groups of antibodies with non-overlapping binding sites were identified.

FimH-DSG TM crystallization and structure determination

Purified FimH-DSG TM, with triple stabilizing mutations (G15A, G16A, V27A), aglycosylation mutations (N7S, N70S, N228Q), a 7-residue Gly-Ser linker, FimG donor strand peptide and a C-terminal 8xHis tag, was buffer exchanged with 20 mM Tris (pH 7.5) using PD-10 column and concentrated to 10 mg/ml. Crystallization was performed at 20 °C using sitting drop vapor diffusion method by mixing equal volumes of protein and reservoir solution containing 1 M Sodium Acetate (pH 4.5) and 25% (w/v) PEG 3350. Crystals grew to their maximum size in ~7 days. Crystals were cryoprotected using the reservoir solution supplemented with 15% glycerol and flash frozen in liquid nitrogen. Diffraction data were collected at APS 17-ID. The data was processed using autoPROC (Global Phasing Limited) and the structure was solved using WT FimH-DSG complex (PDB code 4XOD) as a starting model. Model building and refinement were carried out using COOT [74] and BUSTER (Global Phasing Limited).

Characterization of ability of Mab to target ligand binding site

This experiment was conducted on OCTET HTX using Ni-NTA biosensors and buffer containing 1x PBS, 1% BSA and 0.0 5% Tween 20. Two-fold dilutions (from 10 µg/ml to 0.078125 µg/ml) of Octyl-mannopyranoside ligand were prebound to 5 µg/ml of His-tagged FimH_LD WT for 10 min. Biosensors were allowed to capture FimH_LD WT prebound to ligand for 5 min. The baseline was established before letting the biosensors loaded with FimH and different concentrations of ligand allowed to bind 5 µg/ml FimH Mab for 5 min. Nanometer response of antibody binding obtained from each dilution of the ligand was plotted against ligand concentration.

Cryo-EM sample preparation, data collection, and processing

For cryo-EM studies, Fab fragments of Mabs 299-3, 304-1, 329-2, 440-2 and 445-3 were generated using one of two methods 1) using a Mouse IgG1 Fab and F(ab´)² Preparation Kit (Thermo Scientific Pierce) which uses immobilized Ficin for cleavage 2) production of recombinant Fab fragments with a C-terminal his tag in ExpiCHO cells, by LakePharma. Cryo-EM sample preparation, data collection and processing are described in full in the S1 Text.

Statistical analysis

Statistical significance (p-value) of differences in responses between mouse groups was determined using an unpaired t-test with Welch’s correction, applied to log-transformed data.