Structural and Population Characterization of MrkD, the Adhesive Subunit of Type 3 Fimbriae

Abstract

Type 3 fimbriae are adhesive organelles found in enterobacterial pathogens. The fimbriae promote biofilm formation on biotic and abiotic surfaces; however, the exact identity of the receptor for the type 3 fimbriae adhesin, MrkD, remains elusive. We analyzed naturally occurring structural and functional variabilities of the MrkD adhesin from Klebsiella pneumoniae and Escherichia coli isolates of diverse origins. We identified a total of 33 allelic variants of mrkD among 90 K. pneumoniae isolates and 10 allelic variants among 608 E. coli isolates, encoding 11 and 9 protein variants, respectively. Based on the level of accumulated silent variability between the alleles, mrkD was acquired a relatively long time ago in K. pneumoniae but recently in E. coli. However, unlike K. pneumoniae, mrkD in E. coli is actively evolving under a strong positive selection by accumulation of mutations, often targeting the same positions in the protein. Several naturally occurring MrkD protein variants from E. coli were found to be significantly less adherent when tested in a mannan-binding assay and showed reduced biofilm-forming capacity. Functional examination of the MrkD adhesin in flow chamber experiments determined that it interacts with Saccharomyces cerevisiae cells in a shear-dependent manner, i.e., the binding is catch-bond-like and enhanced under increasing shear conditions. Homology modeling strongly suggested that MrkD has a two-domain structure, comprising a pilin domain anchoring the adhesin to the fimbrial shaft and a lectin domain containing the binding pocket; this is similar to structures found in other catch-bond-forming fimbrial adhesins in enterobacteria.

INTRODUCTION

Type 3 fimbriae are adhesive organelles found in a variety of enterobacterial pathogens, like Klebsiella, Enterobacter, Serratia, Proteus, Providencia, and Citrobacter species (1–8). Historically, type 3 fimbriae have not been associated with E. coli; however, most recently two independent studies reported type 3 fimbriae expression in E. coli strains, and in both strains the fimbrial genes were carried on a conjugative plasmid (4, 9). Type 3 fimbriae belong to the chaperone/usher class of fimbriae and are encoded by the mrkABCDF gene cluster, in which mrkA encodes the major structural component and mrkD encodes the fimbrial adhesin (10–12). The genes mrkB and mrkC encode the chaperone and usher proteins, respectively (10). The mrkF gene encodes a protein suggested to be involved in the stabilization or assembly of the fimbriae (10, 13). The fimbriae have been shown in vitro to play a role in the binding to several different epithelial cell lines and extracellular matrix proteins (14–17). Furthermore, type 3 fimbriae have been determined to be responsible for attachment to and formation of biofilms on biotic and abiotic surfaces, including different human cell types (12, 18–20). However, the exact identity of the MrkD receptor remains elusive.

Type 3 fimbriae are expressed by virtually all clinical isolates of Klebsiella pneumoniae. K. pneumoniae is recognized as an important opportunistic pathogen that frequently causes urinary tract infections (UTIs), septicemia, or pneumonia in immunocompromised individuals (21). It is responsible for up to 10% of all nosocomial bacterial infections (22, 23) and is a frequent cause of catheter-associated urinary tract infections (CAUTIs) (8, 24, 25). In recent years, a high incidence of community-acquired K. pneumoniae pyogenic liver abscess has been reported, especially from Taiwan, but also from other Asian countries, as well as from Europe and North America (26–32).

Type 3 fimbriae are known to be present in K. pneumoniae in at least two variants, a chromosomally encoded variant and a plasmid-carried variant with 60% to 96% homology, depending on the genes compared, with the mrkD gene being the most diverse (15, 16, 33, 34). The plasmid-borne variant is found only in a few isolates, while the chromosomally encoded variant is found in the vast majority of K. pneumoniae isolates (8, 34).

Recently, a novel mode of so-called “catch bonds” (the opposite of slip bonds) or shear-enhanced adhesion was described in bacteria, where adhesive interactions become stronger rather than weaker under increasing hydrodynamic shear conditions (35, 36). Initially, the principal mechanism of catch-bond formation was discovered and extensively studied by using type 1 fimbriae of E. coli as a model. FimH, the adhesive subunit of type 1 fimbriae, was demonstrated in both E. coli and K. pneumoniae to be comprised of two domains, the mannose-binding lectin domain and the fimbria-incorporating pilin domain, and these are connected via a linker chain (37, 38). When the two domains interact under no or low shear conditions, the distal mannose-binding pocket assumes a low-affinity conformation (39, 40). The domains separate under increasing shear, and in doing so they convert the lectin domain into a high-affinity conformation. Catch bonds, or shear-enhanced binding, are beneficial when bacteria live in environments where shear stress is part of everyday life as, for example, on catheter surfaces, the urinary tract, or the colon of mammals (41, 42).

In this study, we analyzed and described the extent and pattern of structural and functional variabilities for the MrkD proteins from K. pneumoniae and E. coli. Due to high diversification between the chromosomally encoded and plasmid-borne variants of type 3 fimbriae, and in order to identify important binding residues in the native MrkD protein, we focused on the chromosomally encoded variant of mrkD, which is the most frequent variant found in E. coli and K. pneumoniae (8, 34).

MATERIALS AND METHODS

Bacterial strains and plasmids.

Bacterial isolates and plasmids used are listed in Table 1.

Table 1.

Strains and plasmids used in this study

Strain or plasmid	Description	Reference
Strains
C3091	K. pneumoniae UTI isolate	62
HB101	Nonfimbriated, noncapsulated E. coli K-12 lab strain	63
AAEC191A	Nonfimbriated E. coli strain	49
Plasmids
pUC18_mrkABCDF	High-copy-no. vector containing mrkABCDF	34
pGB17	Low-copy-no. vector used to clone mrkD variants	49
pBR322_mrkABCF	Low-copy-no. vector containing mrkABCF	This study

We screened 90 K. pneumoniae isolates for the presence of mrkD. The isolates were collected from surface waters or from patients suffering from primary liver abscesses, urinary tract infections, bacteremia, or catheter-associated urinary tract infections. For comparison, the reference strains MGH78578 (ATCC 700721; GenBank accession number NC009648), K. pneumoniae NTUH-K2044 (accession number NC012731), K. pneumoniae IA565 (accession number M24536), and K. pneumoniae strain 342 (accession number NC017540) were used (10, 43, 44).

Isolates were identified as K. pneumoniae by standard biochemical and Minibact E assays (developed by B. Nissen, Statens Serum Institut; catalog numbers SSI-905 and SSI-906).

The E. coli collection under study consisted of 608 isolates screened for the presence of mrkD. The strains were isolated from patients with bacteremia, urinary tract infection, or meningitis or from dogs with a urinary tract infection or pneumonia. mrkD sequences from E. coli MS2027 and plasmid pOLA52 as well as 2 strains from the ECOR strain collection, ECOR15 and ECOR23, were included in the study (4, 9).

PCR amplification and sequencing of mrkD.

Sequences of the type 3 fimbrial adhesin (mrkD) were obtained from 100 isolates (E. coli and K. pneumoniae strains). The primers used for amplification and sequencing of mrkD were the following: mrkDF forward (GGGGGTGCACAATAGCAGCCACGCGATAGT) and mrkDR reverse (GGGGAAGCTTTGTTTATCAGCGATGCGAAC). After sequencing, electropherograms were visually inspected for consistency between strands, and any ambiguous nucleotides were resolved by resequencing. Nucleotide sequences were aligned using ClustalW (45) (http://www.ebi.ac.uk/Tools/clustalw/index.html).

Evolutionary analysis.

Zonal phylogeny analysis and associated statistics were carried out using Zonal Phylogeny software (ZPS), in which the corresponding protein tree is depicted from the DNA phylogram and the structural variants are classified as members of two zones: the primary zone, with multiallelic structural variants (the alleles of each variant being synonymously differentiated), representing the evolutionarily long-term category, and the external zone, with single-allele structural variants (where synonymous mutations have not yet accumulated within the variants), representing the evolutionarily short-term or recent category (46). Average pairwise nucleotide diversity (π) and the rates of synonymous (dS) and nonsynonymous (dN) mutations between and within the clades were calculated using MEGA4 program (47). The analysis of statistical significance was performed using a Z-test with the π and dN/dS values (48).

Construction of pmrkABCF and isogenic recombinant mrkD genes.

The type 3 fimbriae gene cluster (mrkABCDF) had previously been cloned from the K. pneumoniae urinary tract infection isolate C3091 into pUC18 with a restriction enzyme EcoRI site located adjacent to the cluster, creating an 8,130-bp large plasmid, pUC18_mrkABCDF (34). To construct pUC18_mrkABCF, containing the entire mrk cluster except for the mrkD gene, forward primer khfF (5′GATACAGCGGCCGCGATTAATACGGGAGGGGGAATGAAGGG) and reverse primer khfR (5′GATACAGCGGCCGCTCAGCGACATACGCTATCCTTTGTTG) were used, with NotI restriction sites added to the 5′ ends. The forward primer amplified downstream of mrkD, while the reverse primer amplified upstream of mrkD but away from the mrkD gene, i.e., an amplicon of the entire pUC18_mrkABCDF except that the mrkD gene was obtained. Via the NotI restriction sites, the amplicon was ligated, creating pUC18_mrkABCF (7,134 bp). The mrkABCF region was then cut from the construct by using the EcoRI restriction sites and subsequently ligated into pBR322, creating pmrkABCF. Finally, pmrkABCF was transformed into the nonfimbriated E. coli strain HB101.

Different mrkD isogenic recombinant variants were constructed as described elsewhere (49). Briefly, mrkD genes were cloned from appropriate K. pneumoniae and E. coli isolates with primers mrkDF and mrkDR (see above) by PCR and subcloned into the pACYC184-based plasmid pGB17 downstream of the bla promoter. The mrkD-expressing plasmids were introduced into nonfimbriated E. coli strain HB101 already carrying pmrkABCF.

Mannan-binding assay.

The receptor compound, Saccharomyces cerevisiae mannan, was obtained from Sigma Chemical Co. (St. Louis, MO) and dissolved at 20 μg/ml in 0.02 M bicarbonate buffer; 100-μl aliquots were incubated in 96-well microtiter plates for 1 h at 37°C under static conditions. The wells were then washed three times with phosphate-buffered saline (PBS; pH 7.4) and quenched with 0.1% bovine serum albumin in the same buffer for 15 min at 37°C. Overnight cultures of HB101(pmrkABCF) carrying different variants of mrkD were adjusted to an optical density at 540 nm (OD₅₄₀) of 2.0, added to the wells, and incubated for 45 min at 37°C without shaking to achieve sedimentation. Afterwards, the wells were washed to remove unbound bacteria, and 150 μl LB medium was added, followed by a 4-hour incubation with shaking at 37°C. Finally, the density of bound bacteria in each well was determined by measuring the OD at 405 nm.

Assays of binding to mannan and antibodies for strains subjected to site-directed mutagenesis were carried out in a similar manner, with a small modification. After washing to remove unbound bacteria, microtiter plates were dried and then stained with 0.1% crystal violet for 10 min. After several washes with water, 100 μl of 50% ethanol was added to each well, and absorbance was measured at 600 nm. Experiments were carried out in triplicate, and reported results represent mean values with standard errors of the means.

Biofilm formation.

Biofilm formation was assayed by adding overnight cultures of HB101(pmrkABCF) carrying different variants of mrkD to the wells of microtiter polystyrene plates (Nunc) containing 150 μl of M9 minimal medium with 0.02 M glucose, followed by incubation for 24 h at 37°C with agitation (100 rpm). After removal of medium and two washes with phosphate-buffered saline (pH 7.4), the surface-attached cells were stained with 0.1% crystal violet (Sigma-Aldrich) for 15 min. After two washes with phosphate-buffered saline, crystal violet was dissolved by the addition of ethanol, and the absorbance was measured at 595 nm. Each strain was tested in quadruplicate.

Purification of type 3 fimbriae.

Type 3 fimbriae were purified as described previously (39) with a few modifications. Briefly, E. coli strain AAEC191A carrying pUC18_mrkABCDF and thus expressing type 3 fimbriae was grown overnight in LB with shaking at 37°C. The cells were harvested, resuspended in buffer (5 mM Tris, 15 mM NaCl; pH 7.4), and agitated in an osterizing blender. Cell debris was removed by centrifugation. Fimbriae were repeatedly precipitated by using ammonium sulfate [(NH₄)₂SO₄] to a final concentration of 40% (NH₄)₂SO₄ and dialyzed against 5 mM Tris, 15 mM NaCl (pH 7.4) buffer. Finally, fimbriae were resuspended in 50 mM Tris, 150 mM NaCl (pH 7.4), tested for protein sizes on sodium dodecyl sulfate-polyacrylamide gels, and the protein concentrations were measured with a bicinchoninic acid protein assay kit (Pierce). The identity of the purified type 3 fimbriae was verified by mass spectrometry.

Flow chamber experiments.

Polystyrene tissue culture plates were coated with purified type 3 fimbriae (4.1 mg/ml) in 0.02 M NaHCO₃ (bicarbonate) buffer for 1 h at 37°C or with bacterial suspensions for 5 min to 2 h, creating a bacterial carpet (36). Additionally, if the plates were coated with purified type 3 fimbriae, they were also quenched with 0.1 mg/ml of purified E. coli type 1 fimbriae possessing a nonbinding variant of FimH (with a Q133D mutation in the mannose-binding pocket), to prevent nonspecific binding of the yeast (baker's dry yeast [Saccharomyces cerevisiae]) to the polystyrene surfaces (50). Then, a 2.5-cm by 0.25-cm by 250-μm parallel plate flow chamber (PPFC; GlycoTech) was assembled over the plates according to the manufacturer's instructions. The assembly was mounted on a Nikon TE200 inverted microscope with a 10× phase-contrast objective. Experiments were recorded by using a Roper Scientific high-resolution charged-coupled-device camera and MetaMorph or MetaView video acquisition software (Universal Imaging Corporation). All further experiments were performed in 0.1% gelatin–PBS buffer with 0.5% α-methyl-d-mannopyrannoside (Sigma). Yeast cells were washed at least twice and resuspended in 0.1% gelatin–PBS buffer with 0.5% α-methyl-d-mannopyrannoside to a final yeast concentration of 1% to prevent any residual mannose-specific binding.

Yeast suspensions were manually loaded into an assembled PPFC and allowed to settle for 4 min. Recording was started, and buffer was pushed into the chamber by using a Harvard syringe pump at various flow rates to produce the desired wall shear rates. For experiments with purified fimbriae, yeast surface rolling was recorded at 2 frames per second, with a 100-ms shutter time to distinguish cells that were moving with velocities slower than the hydrodynamic speed. For experiments using bacteria-coated surfaces, the yeast movement was recorded at 1 frame per second with a shutter speed of 100 or 200 ms. The time of recording varied depending on the period of delay for each shear rate. Recorded movies were analyzed as follows: settled yeast cells were counted before the flow was turned on. Then, after flow in the chamber had been established, yeast cells that were in focus (moving slower than the ones that were moving with hydrodynamic speed) were counted in several sequential frames. These counts were used to calculate the average percentage of yeast cells that stayed near the surface (i.e., bound yeast). Each experiment was repeated at least two times with the different proteins and bacteria or yeast batches.

As a negative control, the same set of experiments was performed on plates quenched with bacterial cell debris obtained by osterization of the nonfimbriated AAEC191A host strain.

Homology modeling of MrkD lectin and pilin domains.

The full-length amino acid sequence and the lectin domain of MrkD from protein variant C3091 was submitted to RaptorX, a protein structure prediction program (51), and the structure was retrieved.

Site-directed mutagenesis of the 3148 mrkD variant.

Mutations were introduced into mrkD by PCR by using the QuikChange mutagenesis kit (Stratagene). Plasmid pGB17 carrying the 3148 mrkD variant was used as a template, and the mutagenic primer pairs were as follows: F, GACGTCTTTTCGTCCCGGGTATATAACACCACC, and R, GTGGTGTTATATACCCGGGACGAAAAGACGTC, for Q124R and F, CTGGGAGAGCGGCGGTAACCCGACCCTC, and R, GAGGGTCGGGTACCGCCGCTCTCCCAG, for D168G substitutions, respectively.

Nucleotide sequence accession numbers.

All sequences obtained in this study were deposited in GenBank under accession numbers KF777696 through KF777787.

RESULTS

Allelic diversity of chromosomal mrkD in K. pneumoniae and E. coli.

The “chromosomal” variant of mrkD could be amplified for 86 of 90 (94.4%) of the K. pneumoniae isolates and 14 of 608 E. coli isolates (2.3%). As expected, according to the maximum likelihood phylogenetic tree of the nucleotide sequences DNA, all chromosomal mrkD variants were phylogenetically distant from the plasmid-borne mrkD variant from strain IA565 (Fig. 1). The K. pneumoniae chromosomal mrkD formed three groups. We consider the groups as distinct clades (1, 2, and 3), where the phylogenetic distance of the ancestral sequence of each clade (marked by the blue circle in Fig. 1) from any of its descendants is significantly lower (P < 0.05) than its distance from any other sequence outside the clade. The pairwise diversity between the clades ranged from 6% to 8%, while the within-clade diversity was 10-fold lower, at 0.6% for all clades (Table 2, split between clades). The largest clade, clade 1, comprised 42 alleles, with clade 2 (4 alleles) and clade 3 (7 alleles) being much smaller. In E. coli, a total of 10 mrkD alleles were identified, with all of them being phylogenetically linked to and tightly intermingled with clade 1 of the K. pneumoniae mrkD, and 4 alleles being completely identical between the two species.

Fig 1.

Maximum likelihood phylogenetic tree of all 100 mrkD sequences, grouped into three distinctive clades (red boxes). The ancestral node of each clade is denoted by a blue circle. IA565 represents the plasmid variant of mrkD. For each allelic variant, the information includes the name of a representative isolate followed by the total number of isolates carrying the allele (preceded by “n”). The stars indicate that the allelic variant is represented by (at least) one E. coli isolate.

Table 2.

Overall nucleotide diversity and rates of nonsynonymous and synonymous mutations in mrkD from K. pneumoniae and E. coli isolates

Species and/or clade	No. of isolates	No. of hot spot residues	No. of alleles in:		π/nucleotide	dN	dS	dN/dS	Z-test P value
			Primary zone	External zone
All isolates	100	7	30	12	0.037	0.009	0.142	0.06	<0.0001
K. pneumoniae	86	4	30	6	0.040	0.008	0.159	0.05	<0.0001
E. coli	14	6	3	7	0.008	0.009	0.007	1.29	0.579
Within clade 1	84	7	20	11	0.006	0.004	0.010	0.40	0.007
Within clade 2	5	1	4	0	0.006	0	0.023	0
Within clade 3	11	1	6	1	0.006	0.001	0.021	0.05	0.001
Clade 1
E. coli	14	6	3	7	0.008	0.009	0.007	1.29	0.579
K. pneumoniae	70	3	20	5	0.005	0.002	0.011	0.16	0.004

Interestingly, both the between- and within-clade allelic diversities of K. pneumoniae mrkD were primarily due to the silent (synonymous) rather than the amino acid (nonsynonymous) variabilities of the alleles. The rate of synonymous mutations (dS) was significantly higher (P < 0.05) than the rate of nonsynonymous mutations (dN), by 17- to 26-fold between the clades (data not shown) and 2.5- to 20-fold within the clades (Table 2). This indicated a primary action of purifying selection on the K. pneumoniae mrkD against structural changes in the protein. In contrast, the vast majority of allelic variations in E. coli mrkD were due to nonsynonymous mutations, as the dS value was slightly lower than the dN value (Table 2). This indicated that positive selection for adaptive structural changes rather than purifying selection against them could be the primary force behind the mrkD microevolution in E. coli.

Thus, mrkD in E. coli (only located in clade 1) is significantly less diverse than is K. pneumoniae mrkD, but some alleles are identical between the species. Furthermore, unlike E. coli mrkD, the K. pneumoniae genes are under a strong purifying selection against structural changes in the protein, while mutations in the E. coli genes could primarily be of an adaptive nature.

Structural variability of MrkD.

While the dN/dS analysis could not show with certainty whether amino acid changes in E. coli MrkD are adaptive in nature, this microevolutionary test is highly conservative in the detection of positive selection actions. Therefore, the structural variabilities of MrkD proteins from K. pneumoniae and E. coli isolates were analyzed with a more sensitive approach: construction of a zonal phylogenetic tree of the MrkD protein variants, in which adaptive structural changes appear as evolutionarily recent and, often, in the form of “hot spot” mutations, i.e., independently acquired polymorphisms in the same amino acid position (52). A protein tree of nascent MrkD proteins, containing 18 protein variant nodes formed by 42 independently occurring mutations, is shown in Fig. 2. The five nodes inside the inner, dashed circle on the tree denote structural variants that are evolutionarily long-term (with the corresponding alleles accumulating synonymous diversity), while evolutionarily recent protein variants are positioned outside (with the corresponding alleles being either singletons or without silent variability).

Fig 2.

Zonal phylogenetic tree of nascent MrkD protein. The five nodes inside the dashed circle denote primary zone structural (protein) variants, i.e., evolutionarily long-term alleles, with each represented by multiple alleles synonymously differentiated from one another (the number of synonymous [syn] changes within each such variant is shown in square brackets, next to the corresponding node). For each primary zone node, the number of alleles (pie slices) and number of representatives/strains (number inside each pie slide) for each allele are shown as a pie diagram inside the node. The red nodes outside the dashed circle denote external zone structural (protein) variants, i.e., evolutionarily recent alleles, each represented by a single allele without any accumulation of synonymous variation within the variant. The node size indicates the number of strains representing the variant (in the external zone; the lowest and highest being 1 and 4, respectively). Amino acid mutation information is shown along each branch, with the number of synonymous (syn) changes in parentheses, if any. The branch length depicts the number of amino acid changes. The name of the protein variant (node) is denoted in gray next to the node. The mutational hot spot positions are underlined. Stars indicate that an E. coli sequence is present.

Only a minority of the K. pneumoniae alleles (7 out of 34) coded for evolutionarily recent MrkD variants, in contrast to the E. coli MrkD, for which 7 out of 10 were of recent origin (P = 0.02) (Fig. 2; Table 2). This affirmed that the K. pneumoniae chromosomal mrkD is primarily under purifying selection, while reenforcing the hypothesis that E. coli MrkD might accumulate amino acid changes under positive selection. Furthermore, there were seven positions found to be hot spots for structural mutations in MrkD: amino acid residues 46, 111, 121, 124, 127, 130, and 168, with the E. coli MrkD having mutations in six of them, supporting further the presence of adaptive evolution. Interestingly, some of the mutations in the K. pneumoniae MrkD also occurred in the hot spot positions, indicating that positive selection could also be, in part, a factor in the microevolution of the K. pneumoniae protein.

Altogether, these results suggested that the microevolution of MrkD in E. coli is primarily due to positive selection forces for the structural changes, thus indicating that the mutations are functionally adaptive in nature.

Functional effects of point mutations in MrkD.

In order to test whether the amino acid polymorphisms found in mrkD affected the binding function of the protein, 16 of the 18 detected MrkD variants were cloned and expressed in E. coli HB101 carrying plasmid pmrkABCF (we did not have access to strain MS2027 [9], and no clone could be obtained for stl300). By using antibodies against type 3 fimbriae, we found that that all recombinant strains expressed type 3 fimbriae at the same level, based on colony dot blot analysis (data not shown). Recently, we showed that type 3 fimbriae bind to the surface of yeast (53). We therefore utilized MrkD-expressing yeast strains to investigate the ability of the individual MrkD protein variant to bind yeast mannan (Fig. 3).

Fig 3.

(A) Mannan binding by different MrkD variants in a growth assay. White bars denote MrkD protein variants located within the primary zone. Gray bars denote MrkD protein variants located in the external zone. The statistical differences labeled in the figure are based on a comparison with the evolutionarily oldest protein variant, C3091. The mean values of three experiments are shown along with standard errors of the means. P values were calculated by using Student's t test. (B) Biofilm formation by different MrkD variants. White bars denote MrkD protein variants located within the primary zone. Gray bars denote MrkD protein variants located in the external zone. The statistical differences labeled in the figure are based on comparisons with the evolutionarily oldest protein variant, C3091. The mean values of three experiments are shown, along with standard errors of the means. P values were calculated by using Student's t test.

The strongest mannan binding was demonstrated by the most common MrkD protein sequence in the collection (denoted C3091 in Fig. 2), which is found in both K. pneumoniae and E. coli and belongs to the category of evolutionarily long-term variants, with the highest silent diversity of the coding alleles (Fig. 2). While all other variants were significantly (P ≤ 0.01) less adherent than C3091 in binding to mannan, binding was generally most reduced in the evolutionarily recent MrkD variants.

Notably, recombinant strains expressing the E. coli MrkD protein variants (3052, 3148, 3285, FHM19, pOLA, and ECOR23) were particularly reduced in mannan-binding activity compared to the C3091 variant (Fig. 3A).

Type 3 fimbriae are strong biofilm promoters (18, 19, 34). We therefore tested the 16 recombinant strains expressing different MrkD protein variants for biofilm formation capability (Fig. 3B). All evolutionarily long-term protein variants displayed biofilm formation at the same level as the most common variant, C3091. In contrast, E. coli MrkD protein variants 3052, 3148, 3285, ECOR23, and FHM19, of a evolutionarily recent origin, showed a significant reduction (P > 0.05) in biofilm formation.

Thus, the mutational changes that were predicted to emerge under positive selection had a significant functional effect on MrkD that was manifested by the decrease in mannan binding and biofilm formation.

MrkD forms catch bonds.

It has been shown that the binding level of fimbrial tip-associated bacterial adhesins can be increased under shear force, demonstrating so-called catch bond adhesive properties (35, 38, 54, 55). In order to investigate binding patterns of different MrkD protein variants under shear flow, we compared the most abundant and best binding (in static assays) MrkD variant, C3091, to two of the lowest binders, the FHM19 and 3148 MrkD variants, in a PPFC. The binding under shear conditions was tested for either purified fimbriae (for the C3091 variant) or whole bacteria (for all three variants). The fimbriae or bacterial cells were adsorbed to a polystyrene plate and overlaid with a suspension of yeast cells, which were then subjected to different hydrodynamic conditions.

For the C3091 fimbriae, the proportion of yeast cells that accumulated on the surface increased from 5% at the lowest shear force (between 0.01 and 0.1 pN/μm²) to 60% at 0.097 pN/μm² (Fig. 4A). With a further stepwise increase in shear force up to 1.4 pN/μm², yeast accumulation decreased to 5.3%, and all yeast cells were washed away at 1.4 pN/μm². Thus, in this shear range, fimbria-yeast interactions exhibited a pattern consistent with shear force enhancement. A similar phenomenon, albeit somewhat less pronounced, was observed with the whole-cell bacteria that expressed C3091 fimbriae. During static binding to mannan, MrkD variants FHM19 and 3148 were less efficient than variant C3091 in binding under all shear conditions (Fig. 4B). However, both protein variants were still capable of binding in a shear-dependent manner.

Fig 4.

(A) Yeast cells binding to C3091 fimbriae in a PPFC. Fimbriae were immobilized on the surface of the flow chamber, yeast cells were loaded and allowed to settle, and then the flow was turned on at various shear stresses. The numbers of yeast cells at the surface moving slower than the free-floating cells were determined for all shear stress values and divided by the initial number of cells, to calculate the percentage. Error bars represent the standard deviations of bound yeast cells throughout the recorded video for duplicate experiments. As a negative control, the same set of experiments was performed on bacterial cell debris quenched (i.e., debris from the expressing strain from which type 3 fimbriae were purified, without type 3 fimbriae), with plates without immobilized type 3 fimbriae. Here, no slowing of yeast cells on the surface was observed; all the cells were moving with hydrodynamic velocity (data not shown). (B) Yeast cells binding to isogenic strains expressing MrkD protein variants from strains C3091, FHM19, and 3148. Bacteria were immobilized on the surface of the flow chamber, yeast cells were loaded and allowed to settle, and then the flow was turned on at various shear stresses. The numbers of yeast cells at the surface moving slower than the free-floating cells were determined for all shear stress values and divided by the initial number of cells, to calculate the percentage. Error bars represent the standard deviations of bound yeast cells throughout the recorded video for duplicated experiments.

Thus, the MrkD adhesin of type 3 fimbriae is capable of mediating catch-bond-like shear-dependent binding of the bacteria to the target cells. The decreased effect of functional mutations to the mannan-coated surface (under static conditions) could be restored only partially under the increased shear forces.

Localization of the functional mutations in a predicted tertiary structure of MrkD and crystallized lectin domain of a MrkD plasmid variant.

The two-domain (lectin and pilin domains) structure found in FimH adhesin (the adhesive subunit of the K. pneumoniae and E. coli type 1 fimbriae, with shear-dependent binding properties), containing a ligand-binding groove located in the lectin domain, has been suggested to be characteristic for tip-associated fimbrial adhesins in general (37, 56). This prompted us to predict the MrkD three-dimensional structure by submitting the full-length protein sequence of the C3091 MrkD protein variant to the RaptorX protein prediction program (51). Interestingly, although MrkD and FimH only have 12% identity in amino acid sequences when aligned, the homology modeling strongly suggested that FimH was the best-suited template (alignment score of 59; an alignment score of 100 would be a perfect model [51]). Therefore, MrkD, like FimH, is suggested to have a two-domain structure (Fig. 5A). The domains are comprised of a lectin domain from amino acids 24 to 184 of the nascent (mature) peptide and a fimbria-anchoring pilin domain from amino acids 185 to 332 (the putative leader peptide from amino acid 1 to 23 was removed for the modeling). Furthermore, a homology model of the putative lectin domain of C3091 was also predicted; not surprisingly, the best-suited template was the recently crystallized structure of MrkD (plasmid variant) lectin domain (amino acid sequence similarities of approximately 60%; structural alignment score, 76) (Fig. 5B) (57). When we superimposed both homology models (the entire MrkD and the lectin domain of C3091, respectively) with the crystallized structure of MrkD (the plasmid variant), root mean square deviation (RMSD) values of 10.9 and 0.175 were obtained, respectively.

Fig 5.

(A) Homology modeling of the chromosomal variant of MrkD, folded into a two-domain structure, similar to other fimbrial adhesins with a lectin domain and a pilin domain. (B) Homology modeling of the lectin domain of the chromosomal variant of MrkD. (C) Crystallized structure of the lectin domain from the plasmid variant of MrkD (57). For all structures, the seven hot spot residues (magenta [not all are visible]) found in this study have been mapped. Yellow residues denote mutational residues found in both E. coli and K. pneumoniae. Green denotes mutations found only in K. pneumoniae, and blue denotes mutations found only in E. coli. The location of the binding pocket is marked by an arrow.

Of the 42 structural mutations identified in MrkD variants of K. pneumoniae and E. coli, one occurred in the leader peptide, 29 occurred in the lectin domain, and 13 occurred in the pilin domain regions (Fig. 5A), with a predominant clustering of the mutations in the lectin rather than pilin domains (P = 0.028). The clustering in the lectin domain was even more pronounced among the evolutionarily recent (putatively adaptive) structural changes, with 23 in the lectin and 6 in the pilin domain (P = 0.004). Furthermore, all of the mutational hot spots were located in the lectin domain.

In a recent crystallization study of MrkD (the plasmid variant), a deep pocket was discovered near the N terminus of the lectin domain that was suggested to mediate adherence to collagen V. There was good conformational correspondence between the crystallized lectin domain of the plasmid MrkD and the predicted lectin domain of the chromosomal MrkD determined here (Fig. 5B and C). When we mapped the naturally occurring mutations onto the crystal structure and the homology model (Fig. 5B and C), eight mutations (S24V, A111E, F121Y, N127QSY, N130D, S132F, G168SD, and I171T) were found in or around the binding pocket, with five residues (111, 121, 127, 130, and 168) being mutational hot spots, indicating a positive selection for the structural diversification in this specific area (Fig. 5B and C). Interestingly, the six low-affinity mannan-binding protein variants, from strains 3052, 3148, 3285, FHM19, pOLA, and ECOR23, carried at least two of the eight mutations (strain 3052 carried five of the mutations) found in or around the binding pocket. Note that residues 124 and 126, which were found in this study to carry mutations in the “chromosomal variant” of MrkD, were not present in the sequence of the “plasmid variant” of MrkD and thus are not present on the crystal structure (Fig. 5C).

In order to validate the notion that the binding pocket of MrkD is indeed under selective pressure for diversification, hot spot residues 124 and 168 in the low-affinity mannan-binding MrkD variant of strain 3148 were subjected to site-directed mutagenesis. Glutamine was changed to arginine in residue 124 (Q124R), in order for the residue to resemble the most common and highest-affinity mannan-binding MrkD variant, C3091. Likewise, aspartic acid was changed to glycine at residue 168 (D168G), again to resemble the high-affinity mannan-binding MrkD variant, C3091. The altered 3148 mrkD variants were then transformed into HB101(pmrkABCF) and evaluated for binding efficiency to mannan compared to that of the native 3148 mrkD variant (Fig. 6A). While the residue change in hot spot 168 partially increased binding to mannan, the residue change in hot spot 124 led to a more-than-2-fold increase in binding to mannan compared to the native 3148 MrkD variant. Furthermore, in order to rule out different expression levels of type 3 fimbriae, mannan was replaced with antibodies against type 3 fimbriae in the binding assay, and the three strains [MrkD 3148, 3148(Q124R), and 3148(D168G)] were evaluated for type 3 fimbrial expression levels (Fig. 6B). The strains showed similar expression patterns.

Fig 6.

(A) Relative binding to mannan by MrkD variants 3148, 3148(Q124R), and 3148(D168G) in a binding assay. (B) Relative binding of antibodies against type 3 fimbriae. Mean values of three experiments are shown, along with standard errors of the means. P values were calculated by using Student's t test.

The high resemblance between the predicted folding of MrkD and the complex 2 domain structure of FimH, which renders the protein capable of a conformation change due to shear flow, strongly support our finding that MrkD is shear dependent, i.e., it forms catch bonds when subjected to tensile forces, like the FimH adhesin. Furthermore, as supported by the crystal structure of the plasmid variant of MrkD, the diversification of MrkD seems to be located around the binding pocket.

DISCUSSION

Type 3 fimbriae are expressed by a range of pathogenic Gram-negative species, and it has been demonstrated that type 3 fimbriae promote biofilm formation (8, 18–20). However, the exact role of the fimbriae is not well understood, and both the receptor and the ligand remain elusive. The ability to form catch bonds, i.e., shear-enhanced binding by two-domain fimbrial adhesins, has been extensively described for the type 1 fimbrial subunit FimH. In this study, we characterized the adhesive subunit of type 3 fimbriae, MrkD, for naturally occurring structural polymorphisms and binding patterns under static and flow conditions.

Close to 700 E. coli and K. pneumoniae isolates of various origins were screened for carriage of the chromosomal variant of mrkD. The mrkD locus could be amplified from almost 95% of all K. pneumoniae isolates, indicating the ubiquitous nature of type 3 fimbriae in this species. Until recently, type 3 fimbriae had not been associated with E. coli; however, we demonstrated that more than 2% of E. coli isolates carry mrkD, regardless the origin of isolation. Interestingly, 11 of the 14 (it was not possible to evaluate the remaining 3) E. coli isolates carrying type 3 fimbriae could be assigned to E. coli phylogeny group A or D (58).

By creating a maximum likelihood phylogenic tree of the mrkD genes sequenced, we grouped the genes into three distinctive clades, separated from the plasmid variant of MrkD (376 changes). The largest clade, clade 1, included all of the E. coli sequences and all of the mutational hot spots, indicating that it is within this clade that selective pressure is highest.

Notably, evaluation of the diversity of mrkD amplified from E. coli and K. pneumoniae via construction of a zonal phylogenetic tree revealed that most K. pneumoniae alleles (27 of 34) exhibited a long-term evolutionary pattern of accumulation of multiple silent changes over time, while only 7 were of the evolutionarily recent origin (7 of 34). In contrast, E. coli had more alleles of recent origin (7 alleles) than of the evolutionarily long-term origin (4 alleles). The high number of evolutionarily recent MrkD E. coli variants (Fig. 2) suggests that type 3 fimbriae and mrkD were horizontally transferred to E. coli relatively recently. If the transfer had happened a long time ago, a higher number of nonsynonymous changes within mrkD from E. coli would be expected. Additionally, when type 3 fimbriae are expressed by E. coli, MrkD is under positive selection for structural changes. Especially in the lectin domain of E. coli MrkD, a strong selection for diversification was identified. Indeed, when the different MrkD protein variants were expressed and tested for binding to mannan or evaluated for biofilm formation, the evolutionarily recent MrkD protein variants were reduced in binding activity and biofilm formation compared to the evolutionarily long-term MrkD protein variants, thus suggesting that the externally located strains are diversifying toward a lower-affinity mannan-binding phenotype. Interestingly, the notions that type 3 fimbriae were recently acquired by E. coli and that mrkD is under selective pressure indicate that the type 3 fimbrial function in E. coli may differ from that in K. pneumoniae. We have hypothesized that this is due to different habitats of the two species and that in E. coli type 3 fimbriae are still evolving toward the optimal binding specificities for this genus.

Of the 17 isolates (7 K. pneumoniae and 10 E. coli) that express evolutionarily recent variants, 14 (6 K. pneumoniae and 8 E. coli) were isolated from either urine or urinary catheters. This suggests that these isolates are under a selective pressure during infection of the urinary tract, followed by a diversification of MrkD. Thus, a strong positive selection is indicative of the fimbriae being functionally important in this habitat, one way or another. A recent study reported that in K. pneumoniae strain C3091, which harbors the most abundant MrkD protein variant, expression of type 3 fimbriae did not influence the ability to cause urinary tract infections in mice (34). Thus, we speculate that mrkD is diversifying away from the C3091 MrkD protein variant, either to diminish its binding to a somewhat-optimal level or to increase tropism toward a yet-unknown receptor that is different in structure from yeast mannan.

Type 3 fimbriae and the MrkD adhesin have been demonstrated in numerous studies to adhere to various substrata. Taken together, these observations indicate that type 3 fimbriae play a complex role in binding and, thus, are capable of adherence to a wide variety of substrata, both biotic and abiotic (14, 15, 17–20, 59). Additionally, type 3 fimbriae were identified in almost 95% of the K. pneumoniae isolates, independent of origin, suggesting a broad habitat span and thus versatile tasks for type 3 fimbriae. Underlining this notion of the ubiquitous nature of type 3 fimbriae is the fact that several MrkD protein variants less adherent in the mannan binding assay also showed reduced biofilm formation, i.e., adherence to a plastic surface, which is unlikely to be similar in structure to mannan.

The lectin and pilin, a two-domain structure found in FimH with a ligand-binding groove located in the lectin domain, was suggested to be characteristic for all mono-adhesive subunits of fimbriae (37, 56). Indeed, we identified a putative domain subdivision and folding pattern similar to those found in FimH and other fimbrial adhesive organelles. This is in agreement with previously published data (13).

The location of the mutational residues (residues 24, 111, 121, 127, 130, 132, 168, and 171) in the vicinity of the binding pocket in the crystallized structure of the plasmid variant of MrkD suggests that the two mutational residues not present in the plasmid variant, 124 (hot spot) and 126, play a role in the structure and affinity of the binding pocket in the chromosomal variant of MrkD. Indeed, in the putative binding pocket of the two-domain homology modeling of MrkD (Fig. 5A), the two residues (124 and 126) are found at the side of the pocket and in the pocket itself.

By altering a low-affinity mannan-binding MrkD variant (strain 3148) with changes in hot spot residues 124 and 168, we were able to show that alterations in these residues prompted the MrkD variant to increase its binding toward mannan and therefore that these residues indeed play an important role in the binding of type 3 fimbriae to substrata.

Since a domain-domain interaction has been hypothesized to exist in FimH and to be the basis of its ability to form allosteric catch bonds (39, 50), we investigated here whether type 3 fimbriae mediate catch-bond adhesion. This was especially interesting since the MrkD adhesin is believed to be the sole protein at the tip of the type 3 fimbriae, unlike type 1 fimbriae, which have a distinguishable tip fibrillum consisting of multiple minor adaptor subunits that anchor the adhesin to a pilus rod (FimG, FimF, and FimH) (60, 61).

We conducted an inverted PPFC adhesion assay as previously described (55), using a surface on which purified fimbriae were immobilized, and we added a flow of yeast cells over this surface. An increase in binding under higher shear conditions defined the interaction between type 3 fimbriae and yeast cells as shear-dependent binding.

The described shear-enhanced binding to yeast cells was seen to be similar to that of FimH binding to mannose. Thus, the shear-enhanced adhesion is most likely due to the same mechanisms as observed with FimH. This suggests that the MrkD receptor-binding domain can exist in two binding states, weak and strong, and an intact, two-domain form of the MrkD adhesin is capable of converting from the former to the latter under application of an external tensile force, similar to the catch-bond-forming FimH protein. However, unlike FimH, where point mutations increase the binding, the most abundant and evolutionarily original MrkD protein variant described here is the highest-affinity binder, under both static conditions and hydrodynamic flow conditions.

The two low-affinity mannan-binding protein variants (in the static assays) from E. coli strains 3148 and FHM19 also showed reduced binding to yeast cells when tensile forces were applied. However, the shear-dependent phenotype remained intact, suggesting that the mutations found in these protein variants do not influence the overall domain-domain interaction between the lectin domain and the pilin domain of the MrkD protein, but instead they influence the specificity of the binding pocket toward the surface of the yeast cells.

Furthermore, we observed that MrkD binding to yeast cells during shear stress was strongest at around 0.1 pN/μm²; this is similar to the strongest binding point observed for FimH, indicating that type 3 fimbriae may mediate binding in the urinary tract in a similar manner as has been suggested for type 1 fimbriae.

Taken together, our findings in this study reinforce an important concept in bacterial adhesion and pathogenesis, namely, that the shear-dependent catch-bond mechanism of adhesive interactions is a conserved and, thus, physiologically important phenomenon that is widespread in nature.

In conclusion, we established that MrkD has a two-domain structure comprised of a lectin domain containing a putative binding pocket and a fimbria-anchoring pilin domain; together, these domains are able to form catch bonds when under shear stress conditions. Furthermore, we suggest that type 3 fimbriae were horizontally transferred to E. coli recently, in which mrkD is under strong selective pressure for functional modification. We propose that the role of type 3 fimbriae is to be elucidated in a milieu with shear forces, e.g., the urinary tract.