Introduction
Infection with
Escherichia coli is the most frequent cause of septicaemia in humans and commonly originates from the urinary tract [
1]. Uropathogenic
E. coli (UPEC) adhere to bladder epithelium in a process mediated by type 1 fimbriae via FimH engaging uroplakin 1a on urothelium, leading to urinary tract infection [
2,
3]. Subsequently, FimH promotes invasion and is critical for blood-borne dissemination to other tissues [
4]. Thus, in neonatal meningitis, FimH is essential for the localisation of UPEC to brain microvascular endothelium and invasion of the meninges [
5,
6]. This establishes the pathogenic importance of FimH-mediated adhesion beyond the urinary tract.
FimH is located at the tip of type 1 fimbriae expressed by Gram-negative pathogens, including
E. coli,
Salmonella enterica and
Klebsiella pneumoniae [
7,
8]. It has two domains: an
N-terminal lectin domain of FimH containing the mannose binding pocket (MBP) responsible for bacterial adhesion to cellular ligands and a C-terminal pilin domain that connects FimH to the fimbrial rod [
9]. Introduction of shear stress after initial binding induces allosteric interactions between the lectin and pilin domains that increase the affinity of mannose for the MBP through a catch bond mechanism [
10,
11].
Over the past decade, mutated FimH have been used extensively to probe the molecular basis for its binding to mannose, most commonly in studies performed under static conditions using yeast agglutination [
12], FimH binding to pure mannose substrates [
7,
10] or bacterial adhesion to bladder carcinoma cell lines [
13] as end points. These have provided considerable insights into the molecular basis for the MBP binding with mannose, but necessarily poorly reflect physiological conditions in which it normally takes place. Specifically, there is a lack of data on FimH-dependent bacterial adhesion to microvascular endothelium that is thought to underlie blood-borne dissemination of
E. coli [
4]. We addressed this issue by generating and validating a panel of multiply disabled
E. coli strains that uniquely express type 1 fimbriae and normal or mutated FimH [
14], and systematically analysing the ability of the mutant strains to adhere to microvascular endothelium and bladder epithelium, under both static conditions and physiological shear stress. We show that FimH-dependent adhesion to endothelium occurs much more efficiently than to bladder epithelium and identify MBP residues that are critical for adhesion under shear stress but without detectable effects in static assays. Our results characterise important differential effects of FimH-mediated adhesion to different cellular substrates that reflect the different physiological conditions they are exposed to in vivo.
Materials and methods
Chemicals
Alpha-D-mannopyranoside (mannoside), RNase B (tri-mannosylated-3 M) and bovine serum albumin (BSA) were from Sigma-Aldrich (St. Louis, MO, USA). D-Mannose-BSA (mono-mannose-1 M) (14 atom spacer) was from Dextra Laboratories (Reading, UK), and 0.05 % Trypsin-EDTA and HEPES were from Life Technologies (Carlsbad, CA, USA).
Antibodies
The following polyclonal antibodies were used for Western blot: uroplakin 1a (ABIN955479, 1:100, antibodies-online, Atlanta, GA, USA) and beta-actin (A2066, 1:500, Sigma-Aldrich). Secondary antibodies conjugated with alkaline phosphatase were from Promega (1:5000, Madison, WI, USA).
Cell lines
Human dermal microvascular endothelial cells (G1S1) and conditionally immortalised glomerular endothelial cells (GEnC) were cultured using standard validated methods [
15,
16]. GEnC were propagated at 33 °C (proliferation phase) and differentiated at 37 °C for 5 days prior to each experiment [
15]. The human transitional cell carcinoma cell lines 5637 (ATCC HTB-9) and HT-1376 [
17] and SV40-transformed urothelial cell line SV-HUC [
18] were all kind gifts from Michael Wirth (University of Vienna).
Bacterial strains and GFP labelling of bacteria
Escherichia coli were labelled with green fluorescence protein (GFP) using phage 1 transduction of gfp:bla from
E. coli strain OS56 into the multiply disabled
E. coli MS528 (
E. coli MG1655 Δfim Δflu) [
19], resulting in strain
E. coli MSC95. The gfp gene was inserted into the chromosome of MS528 using the Lambda Red System with the lambda red proteins encoded on the plasmid pTP223, which includes a gene for tetracycline resistance [
20,
21] (kindly provided by Antony Poteete). As the source of the drug resistance cassette, pKD4 carrying a kanamycin cassette was used [
22].
Escherichia coli MSC95 completely devoid of all fimbriae was used as the FimH-negative control strain.
Escherichia coli MSC95 expressing FimH (MSC95-FimH) was derived from
E. coli PC31
fimH [
23] located on the pMAS4 plasmid, together with pPKL115 carrying the entire fim gene cluster with a knock-out mutation in
fimH [
24] and used as standard FimH-bearing control strain.
Site-directed mutagenesis of fimH
Mutations were introduced into the
fimH gene from
E. coli PC31 carried by the pMAS4 plasmid [
23] by site-directed mutagenesis using the Phusion Site-Directed Mutagenesis Kit (F-541, Thermo Scientific, Waltham, MA, USA), following the manufacturer’s instructions. Specific primers were designed for each desired point mutation. For the expression of type 1 fimbriae carrying the mutated FimH protein, plasmids carrying the mutated
fimH gene were individually transformed into MSC95 containing the plasmid pPKL115 carrying the
fim gene cluster with a knock-out mutation in
fimH [
24]. Recombinant strains were grown in LB medium supplemented with 10 μg/ml chloramphenicol and 100 μg/ml ampicillin.
Yeast agglutination assay
The ability of the recombinant
fimH mutant strains to express a D-mannose binding phenotype was examined by yeast agglutination using an established method [
25]. Briefly, 20 μl of 1 % (v/w) yeast in PBS were mixed with 20 μl of serial dilutions (non-diluted up to 1:16) of bacterial suspensions in PBS (normalised to OD
600 = 0.3) of either MSC95-FimH or mutant strains on a microscopy slide, and the dilution at which agglutination occurred was recorded.
Bacterial adhesion under static conditions
GFP-expressing E. coli MSC95-FimH (4 × 106 CFU) were incubated with confluent mammalian cell lines in 12-well cell culture plates for 30 min on ice to prevent internalisation. The cultures were then washed three times with PBS to remove non-adherent bacteria before the mammalian cells were detached with trypsin-EDTA and the resulting single-cell suspensions were analysed by flow cytometry (LSRFortessa SORP, Becton Dickinson, San José, CA, USA). Bacterial adherence was quantified from the intensity of the GFP signal from single endothelial or urothelial cells identified by forward/sideward scatter, thus excluding GFP signals associated with cell clusters and/or from free bacteria. The results were analysed with FlowJo (Tree Star, Inc., Ashland, OR, USA) and expressed as the adhesion index, defined as the percentage of GFP-positive E. coli bound to mammalian cells. The adhesion of mutant E. coli strains to the cell lines was expressed in the results as a percentage of adhesion of the MSC95-FimH parent strain. The mannose dependence of adhesion was assessed by suspending the E. coli in media containing 4 % mannoside for 10 min on ice before the experiment.
Bacterial adhesion under flow conditions
Vena8 Fluoro+ biochips (Cellix, Dublin, Ireland) were coated overnight with either 200 μg/ml D-mannose-BSA (1 M), 100 μg/ml RNAse B with high 3-mannose (3 M) residues or 2 % BSA alone at 4 °C and blocked prior to use with PBS + 0.2 % BSA. A total of 1 × 10
6
E. coli prepared as described above were added to the substrates. The biochips were set up and washed according to the manufacturer’s instructions using the VenaFlux Assay Software (Cellix). The specificity of binding was assessed by pre-incubating
E. coli MSC95-FimH with 2 % mannoside in PBS for 10 min prior to the experiment. Adhesion of bacteria under a shear stress of 1 dyne/cm
2 was recorded every second in phase contrast and the settings were equal for both 1 M and 3 M (exposure time 344 ms, magnification 20×) for 5 min using an Axiovert 200M microscope (Zeiss, Oberkochen, Germany) with AxioVision 4.5 software. The total number of adherent bacteria per high-power field (HPF) was counted manually using ImageJ [
26].
To assess FimH-dependent adhesion to mammalian cells under flow conditions, Vena8 Endothelial 8-channel biochips (Cellix), 800 nm long and 120 nm wide, were sterilised by UV-light and coated with FNC coating buffer (AthenaES, Baltimore, MD, USA) at 4 °C overnight. Cells were seeded into the biochips at 5 × 105 cells per channel and allowed to adhere for 1 h, resulting in confluent cell layers. The cells were incubated for another 24 h in the biochip connected to the Kima pump (Cellix) with the following shear stress conditions: for bladder epithelial cells, 150 μl/min for 6 min, followed by 20 min of absence of flow; for microvascular endothelial cells, 450 μl/min for 6 min, followed by 20 min of absence of flow. Both were incubated at 37 °C with 5 % CO2. Bacterial samples were prepared as described for the 1 M and 3 M assays. The flow chamber was then connected to the Mirus Evo Nanopump (Cellix) and the channels were rinsed three times with 25 μl of media prior to each experiment, and bacterial adhesion was initiated by the addition of 1 × 106 of bacterial suspension. As above, adhesion of bacteria was recorded every second under a shear stress of 1 dyne/cm2 in phase contrast and the settings were equal in all conditions (exposure time 344 ms, magnification 32×) for 5 min. In some experiments (stop/flow), 1 dyne/cm2 was exerted and paused for 5 min once bacteria were observed in the HPF, before 1 dyne/cm2 shear stress was re-applied. This, however, induced some gaps between the cells; any E. coli adhering to this were excluded, as mentioned above. The total numbers of E. coli that were adherent were counted as above. Escherichia coli were considered adherent when they were adhering for at least five frames at the end of the 5 min. Excluding criteria were E. coli that adhered to any plastic surface.
Transmission electron microscopy
Transmission electron microscopy (TEM) was performed to confirm the expression of intact fimbriae on all mutant strains. Five independent TEM micrographs (final magnification 60,000×) were analysed from each mutant strain and were used to count the number of fimbriae along three circumferential areas of 500 nm each in which individual fimbriae were clearly distinguishable. The total number of fimbriae was then calculated from the circumferential outline of the bacteria (∼4500 nm) using ImageJ.
Structural modelling
The crystal structure of FimH (PDB ID: 2VCO) [
27] was uploaded in Visual Molecular Dynamics [
28] (VMD, University of Illinois, Urbana–Champaign, IL, USA). The residues that we experimentally tested were highlighted.
Statistical analyses
All calculations were made using GraphPad Prism 5.0 (GraphPad Prism Software, La Jolla, CA, USA) and
p-values <0.05 were considered significant. Absolute values for the number of adherent bacteria were summarised as means ± standard error of the mean (SEM) and the significance of differences between them was assessed by Student’s
t-test. Adhesion index assays were expressed as medians with interquartile ranges and illustrated using box and whisker plots. The overall significance was tested by Kruskal–Wallis followed by, when appropriate, individual two-tailed Mann–Whitney tests. In both cases, individual pairwise
p-values were corrected for multiple comparisons using the Benjamini–Hochberg method, with a false discovery rate of α set to 0.05 [
29]. For the correlation data, the average adhesion of the mutant strains was analysed; MSC95-FimH and MSC95 were excluded from the correlation calculations.
Discussion
Type 1 fimbriated pathogens, like
E. coli, adhere to a variety of biological surfaces, both epithelia and endothelia. Accordingly, they have adapted by modifying the sequence and structure of adhesins like FimH to enable improved adhesion under different—static or shear stress—conditions [
10,
41]. Our study explores the methodologies used to analyse FimH-mediated adhesion of
E. coli using mutations in the MBP of FimH [
31]. We first used yeast agglutination as an established screening assay [
34] to test the functionality of mutant FimH and found that, to a varying degree, all our mutant strains were able to adhere to yeast. These results were comparable to those obtained with the more physiological binding to mannosylated substrates and mammalian cells under both static and shear stress conditions, which confirmed that three mutant strains lacking intact fimbriae were unable to adhere. All our strains were engineered to bear type 1 fimbriae only and the differences observed in these assays cannot be attributed to other pili, whose possible influence remains enigmatic. They confirm, however, the necessity for additional investigations to establish the adhesive properties of (mutant) FimH. The analysis of FimH-mediated adhesion under flow to coated substrates [
42,
43] or cells [
44] was established previously. In this study, by exposing FimH-expressing
E. coli under shear stress conditions to endothelial cells seeded onto biochips, we identified potential novel effects of FimH mutations.
We established two assays that enabled us to compare adhesion to pure mannosylated substrates and relevant urothelial and endothelial cells under static and shear stress conditions: in the first assay, similar to a recently described method [
38], we allowed FimH-bearing
E. coli to adhere to mammalian cells under static conditions and analysed the numbers of bacteria attached by FACS. The second assay measured the adherence under shear stress by allowing bacteria to bind 1 M or 3 M substrates and cells under flow. We found that the results between static assays that measure adhesion to mannosylated substrates or mammalian cells correlated well. But there were distinct differences when we compared the results of wild-type or mutated FimH-bearing
E. coli adhering to cells under static conditions to those under shear stress conditions.
FimH mediates the adhesion of
E. coli to brain endothelial cells and is a virulence factor critical for blood-borne dissemination of
E. coli [
4]. We show that FimH also mediates efficient bacterial adhesion to human microvascular endothelium of the skin and the glomerulus of the kidney under static conditions and during physiological shear stress. In agreement with other studies, bacteria adhered better under conditions of high shear stress (1 dyne/cm
2), which reflects the physiological condition [
40], than low shear stress (0.2 dyne/cm
2) [
7,
44]. Once attached, adherent bacteria could not be dislodged. It is known that shear stress enhances the strength of FimH-mediated adhesion through allosteric coupling [
41,
45,
46], which could contribute to the failure of shear stress to dislodge adherent MSC95-FimH. The high affinity binding of terminal mannose residues on cellular receptors to MBP of FimH is the central event in
E. coli adhesion.
Our panel of
E. coli MSC95-FimH strains with unique FimH mutations enabled us to identify, in addition to those described [
10], amino acids that differentially modulated the effect of shear stress. The ability of FimH mutations to disrupt fimbriogenesis is a potential limitation to this approach [
33] that affected three of our strains. We measured the agglutination potential of these mutant strains and found that they correlated relatively well with other reports [
31]. Minor differences could lie in the glycosylation of guinea pig red blood cells compared to yeast, as reported by others [
47‐
49]. Regardless, the results with the remaining nine strains with normal fimbriae demonstrated that the mannose binding pocket is crucial for the adhesion of
E. coli to endothelial cells [
31]. Comparison of adhesion to endothelium under static and shear stress conditions revealed that E50 and T53 were essential under shear stress. This only partly correlated with binding to mannosylated substrates under identical assay conditions, demonstrating that there are important differences for the adhesion of isolated mannosylated substrates and mammalian cells. E50 has been demonstrated to not interact with the sugar ligand, but to stabilise R98, which is, thereby, oriented to interact with the ligand [
50]. Due to its location on the backbone of the MBP, T53 interactions with 3-mannose are blocked by the side chains of I52 and D54. The effect of T53 mutations can, therefore, only be indirect through changes in its backbone energetic minimum, which could lead to gentle changes in the 3-mannose binding site [
27,
50]. This supports a model in which these amino acid residues contribute to initial interaction between FimH and its cellular receptors, and that facilitates subsequent insertion of their terminal mannose residues into the MBP to establish firm adhesion. This would be consistent with the lack of reported mutations of these residues in clinical
E. coli isolates, suggesting an important evolutionary conserved contribution to virulence [
51].
The
E. coli strain MSC95-FimH adhered efficiently to microvascular endothelium both under static conditions and shear stress, whereas adhesion to all three urothelial cell lines tested in static conditions was less efficient and appeared abrogated by flow. The observation that MSC95-FimH fails to adhere to bladder epithelial cells under flow is in apparent contrast to an earlier study [
44]. However, the discrepancy probably relates to technical differences between the two studies and specifically to our use of an
E. coli strain specifically engineered to express type 1 fimbriae exclusively, which enabled us to examine the unique effects of FimH.
In our experiments, differences between FimH-mediated adhesion to urothelial cells under static and shear stress conditions could be regarded, as the experimental results correlate stop/flow conditions that occur in the urinary bladder during voiding. Allowing MSC95-FimH to adhere under static conditions before applying shear stress resulted in an increase of bacteria bound to the non-neoplastic urothelial cell line SV-HUC and, to a lesser extent, to the neoplastic cell lines 5637 and HT1376 that were not dislodged by flow. Using this setup, we confirmed the reduced adhesion capabilities of E50A and T53A, although the effect was less obvious than during continuous flow. This supports the model in which these residues contribute to mannose binding. When mutated, FimH is limited in making contact with mannose residues under shear stress conditions.
In summary, we show that, by analysing bacterial adhesion under physiologically relevant conditions, novel effects of FimH mutations are observed. Our study thus highlights the fact that the results obtained from investigating FimH adhesion are dependent on the methodological differences of the assays used and the physiological situation they represent. The results with MSC95-FimH, expressing native or mutated variants of FimH, identified the importance of the MBP that critically influences the adhesion under flow, and provide novel insights into screening methods to determine the effect of FimH mutants and potentially FimH antagonists.