Architecture and function of human uromodulin filaments in urinary tract infections

See allHide authors and affiliations

Science  21 Aug 2020:
Vol. 369, Issue 6506, pp. 1005-1010
DOI: 10.1126/science.aaz9866

How uromodulin helps flush out bacteria

Urinary tract infections (UTIs) are one of the most frequent bacterial infections in humans. The glycoprotein uromodulin is the most abundant urinary protein and can provide some protection from UTIs, but the precise mechanism has been unclear. Weiss et al. found that uromodulin forms stacked, fishbone-like filaments that act as a multivalent decoy for bacterial pathogens with adhesive pili that attach to the uromodulin glycans (see the Perspective by Kukulski). The resulting uromodulin-pathogen aggregates prevent bacterial adhesion to glycoproteins of the urinary epithelium and promote pathogen clearance as urine is excreted. This innate protection against UTIs is likely to be particularly important in infants and children.

Science, this issue p. 1005; see also p. 917


Uromodulin is the most abundant protein in human urine, and it forms filaments that antagonize the adhesion of uropathogens; however, the filament structure and mechanism of protection remain poorly understood. We used cryo–electron tomography to show that the uromodulin filament consists of a zigzag-shaped backbone with laterally protruding arms. N-glycosylation mapping and biophysical assays revealed that uromodulin acts as a multivalent ligand for the bacterial type 1 pilus adhesin, presenting specific epitopes on the regularly spaced arms. Imaging of uromodulin-uropathogen interactions in vitro and in patient urine showed that uromodulin filaments associate with uropathogens and mediate bacterial aggregation, which likely prevents adhesion and allows clearance by micturition. These results provide a framework for understanding uromodulin in urinary tract infections and in its more enigmatic roles in physiology and disease.

The glycoprotein uromodulin (UMOD) is secreted in the kidney and is the most abundant urinary protein (1). A UMOD promoter variant present in ~80% of the human population drives a twofold increase in urinary UMOD concentration (2), which results in reduced susceptibility to bacterial urinary tract infections (UTIs) (3). Uropathogenic Escherichia coli (UPEC) utilize adhesive type 1 pili to attach to high-mannose–type N-glycans displayed on the uroepithelial surface (4). It has been suggested that UMOD acts as a soluble adhesion antagonist for UPEC (511).

Mature UMOD consists of three epidermal growth factor (EGF)–like domains (EGF I to III), a cysteine-rich domain (D8C), a fourth EGF domain (EGF IV), and the bipartite zona pellucida (ZP) module (subdomains ZP-N and ZP-C) (Fig. 1A) (1). Produced as a glycosylphosphatidylinositol-anchored precursor, UMOD is then cleaved by the protease hepsin and assembles into homopolymeric filaments with an average length of ~2.5 μm (12, 13). Despite its multiple roles in human health and disease (1), the molecular architecture and interactions of UMOD in urine are poorly understood.

Fig. 1 UMOD site-specific glycosylation pattern and filament architecture.

(A) Domain organization of mature UMOD with four EGF-like domains (I to III, light blue; IV, green), the cysteine-rich D8C domain (light blue), and the bipartite ZP module (ZP-N and ZP-C, green) (1). The most abundant N-linked glycan forms identified at each N-glycosylation site are shown schematically (amino acid numbering according to UMOD with N-terminal signal peptide). Except for Asn275 and Asn513, all N-glycans were confirmed to be di-, tri-, or tetra-antennary complex type that could be sialylated and/or fucosylated. We observed mixed N-glycan structures at Asn513, composed of Man6 and, as previously reported, complex-type N-glycans (8). The sum of the masses of all identified glycans corresponded well to the average mass difference (20.1 kDa) between glycosylated and deglycosylated UMOD. The elastase-digested form of UMOD (eUMOD = Ser292 through C terminus Phe587) is marked in green, and scissors indicate the elastase cleavage site after Ser291. Neu5Ac, N-acetylneuraminic acid; Gal, galactose; Fuc, fucose; Man, mannose; GlcNAc, N-acetylglucosamine. (B to D) Cryo-ET of purified UMOD filaments revealed irregular bending and helical parameters. (C) and (D) show magnified views of the two most common orientations, revealing the zigzag core (C) and the lateral arms causing a fishbone-like appearance (D). 13.8-nm slices through cryotomograms are shown. Scale bars, 100 nm in (B) and 50 nm in (C) and (D). (E) Different orientations of a UMOD subtomogram average (surface renderings). The subtomogram average was low-pass filtered to 27 Å to demonstrate the complete 3D architecture. (F) Different orientations (surface renderings, green) of the eUMOD subtomogram average. The superposition of eUMOD with native UMOD (transparent, light blue), low-pass filtered to the same resolution of 27 Å, demonstrates the absence of the lateral arms in eUMOD. The unfiltered averages can be seen in fig. S9. (G) Proposed alternating ZP-N– and ZP-C–stacking model of the UMOD filament architecture [same orientation as in (F), right]. The two high-mannose N-glycan structures in each UMOD monomer are indicated by green glycan trees, and complex-type N-glycans are shown as gray triangles.

We first developed a protocol to purify stable UMOD filaments from urine (table S1 and fig. S1A). Mass spectrometry (MS) of dissociated UMOD monomers from different donors revealed highly similar, broad mass-distribution profiles (fig. S1, B to D), which showed that UMOD glycosylation was gender- and genotype-independent. Next, we established a site-specific N-glycosylation map of UMOD using liquid chromatography with tandem mass spectrometry (LC-MS/MS) of tryptic glycopeptides (Fig. 1A, figs. S2 to S6, and tables S2 and S3). We identified individual N-glycans attached to asparagine (Asn) residues 38, 76, 80, 232, 275, 322, 396, and 513 (figs. S2 to S6 and tables S2 and S3). High-mannose N-glycans were found at Asn275 and Asn513, whereas other N-glycans were confirmed to be di-, tri-, or tetra-antennary complex types. Because UPEC adhere to uroepithelial cells by means of type 1 pili that spec>ifically recognize high-mannose N-glycans (4, 14), both glycans at Asn275 and Asn513 were candidates for mediating UMOD’s antiadhesive activity.

To contextualize the glycan arrangement within UMOD polymers, we imaged UMOD filaments using cryo–electron tomography (cryo-ET) (15) (Fig. 1B and movie S1). One orientation (Fig. 1C) was consistent with the previously observed zigzag shape (12, 13, 16, 17), though filaments showed different degrees of curvature and irregular rotations around the long axis. A second, prominent orientation had a fishbone-like appearance, with a central core filament and regularly protruding arms (Fig. 1D). We calculated a structure by subtomogram averaging, which resolved repeating-filament subunits arranged in a helix with a 180° twist and 6.5-nm rise. The filament core consisted of 8.5-nm-long modules zigzagging at 95° angles. The 12.5-nm-long arm segments were flexible and protruded at 45° angles (Fig. 1E, fig. S7, and movie S1).

To assign UMOD protein domains to densities in the subtomogram average, we studied an elastase-digested form of UMOD (eUMOD), wherein UMOD is cleaved between the D8C and EGF IV domains (18). We observed that eUMOD retained the zigzag core structure, but densities for the arms were absent (Fig. 1F; fig. S7; fig. S8, B to D; fig. S9; and movie S2). On the basis of the fitting of the previously published crystal structure of truncated UMOD (EGF IV–ZP module) (16) into our subtomogram averages, we propose a model in which the ZP module polymerizes into the filament backbone and the EGF I to III and D8C domains constitute the protruding arms (Fig. 1G and fig. S9, C and D). This alternating ZP-stacking model differs from previously suggested architectures (16, 19) of the UMOD filament (fig. S10).

Next, we investigated the interaction between UMOD filaments and the adhesin, FimH, of type 1–piliated UPEC strains. FimH at the pilus tip recognizes terminal mannosides in high-mannose N-glycans of the uroepithelial receptor uroplakin 1a (4, 20). Binding of type 1–piliated cells to UMOD filaments had been demonstrated, but the exact FimH binding site in UMOD remained unknown (6). We recorded the affinity of the isolated FimH lectin domain (FimHL) to UMOD filaments and obtained a single apparent dissociation constant (Kd) of 2.2 × 10−8 M (Fig. 2A). Consistent with the high affinity of FimHL for UMOD, spontaneous dissociation of FimHL from UMOD filaments was very slow, with a half-life of 2.1 hours (Fig. 2B). Because FimHL binds mannosides with a 2000-fold higher affinity compared with full-length FimH—as a consequence of the ability of FimH to form catch bonds under tensile mechanical force (21, 22)—we calculated a Kd of ~4 × 10−5 M for the UMOD-FimH complex in the absence of shear stress. Using analytical gel filtration and gel-band densitometry, we determined the stoichiometry of the UMOD-FimHL complex to be 1:2 (Fig. 2, C and D, and fig. S11).

Fig. 2 Two FimHL molecules bind to the high-mannose glycan on the UMOD arm.

(A and B) Thermodynamics and kinetics of UMOD binding and release by FimHL at pH 7.4 and 25°C. (A) Competitive equilibrium displacement of the fluorescent mannoside GN-FP (21, 22) from FimHL with increasing UMOD concentrations, recorded through the decrease in GN-FP fluorescence anisotropy. (B) Kinetics of spontaneous dissociation of FimHL from UMOD, recorded through the binding of excess GN-FP to released FimHL. The obtained first-order kinetics (solid lines) were independent of GN-FP concentration and thus directly monitored dissociation of FimHL from UMOD (koff = 9.1 × 10−5 s−1). Direct determination of the affinity of full-length FimH for UMOD filaments proved to be impossible because of the limited solubility of UMOD filaments (~100-μM monomers). (C and D) Titration of FimHL with increasing amounts of UMOD filaments to determine the stoichiometry of complex formation at pH 7.4 and different temperatures (15°, 25°, and 37°C). (C) Size exclusion chromatography (SEC) elution profiles of 10 μM FimHL incubated with different amounts of UMOD filaments (UMOD monomer concentrations between 0 and 20 μM), detected through protein absorbance at 280 nm. (D) Peak area of free FimHL, plotted against the UMOD monomer-to-FimHL ratio, revealing that, on average, two FimHL molecules can bind to a single UMOD monomer. a.u., arbitrary units. (E) Different orientations of the subtomogram average (shown as a surface rendering) of UMOD filaments that were incubated with a fourfold excess of FimHL over UMOD monomers. Notable additional densities are detectable on the filament arms, whereas no extra densities could be identified on the filament core. (F) Overlay of surface renderings of subtomogram averages of native UMOD (blue) and UMOD incubated with a fourfold excess of FimHL (transparent orange), both low-pass filtered to a resolution of 27 Å. The additional densities could accommodate two copies of FimHL (fig. S12, D to G). Together, the data indicate that up to two FimHL molecules bind to the high-mannose glycosylation site at Asn275 in the D8C domain of the UMOD arm.

To test which of the two high-mannose N-glycans of UMOD was recognized by FimH, we analyzed the UMOD·(FimHL)2 complex using cryo-ET. Differences between UMOD∙(FimHL)2 and native UMOD were already visible in individual tomograms (fig. S12, A and B, and movie S3). The UMOD·(FimHL)2 subtomogram average revealed a prominent additional density on the UMOD arms, which may be sufficient to accommodate two FimHL (Fig. 2, E and F; fig. S12, C to F; and movie S3). No notable additional density was seen at the core of UMOD∙(FimHL)2 filaments or on eUMOD (containing only the high-mannose glycan at Asn513) incubated with FimHL (fig. S12, G to L). Together, our data demonstrate that the high-mannose N-glycan at Asn275 of the UMOD arm is the only accessible FimHL-recognition site.

We next analyzed binding of UMOD filaments to a mixture of type 1–piliated and nonpiliated E. coli cells by cryo-ET imaging. Each type 1–piliated cell exhibited a substantial increase in the local UMOD filament concentration around the cells (n = 6 tomograms) (Fig. 3, A and B; fig. S13, A and B; and movie S4). Contact sites between pili and UMOD were mainly at the FimH-containing pilus tips (Fig. 3A). In contrast, no UMOD filaments accumulated around nonpiliated cells (n = 7 tomograms; fig. S13, C and D) or piliated cells incubated with eUMOD (n = 3 tomograms; fig. S14). E. coli-UMOD association therefore requires both the presence of type 1 pili and the glycosylated UMOD arm.

Fig. 3 UMOD associates with piliated bacteria and can lead to cell aggregation.

(A and B) The type 1 pilus–deficient strain E. coli AAEC189 was transformed with the type 1 pilus expression plasmid pSH2, coincubated with UMOD filaments (1 μM UMOD monomers and ~1.6 × 109 E. coli cells/ml) and imaged with cryo-ET. A slice (8.6-nm thick) through a cryotomogram (A) and a segmentation of the same tomogram (B) are shown. Piliated bacteria (cell, brown; pili, orange) were always seen in close association with a loose mesh of UMOD filaments (blue lines or arrows). Note the contact site of the pilus tip with the UMOD filament. A partly lysed cell that resulted in higher image quality is shown. See fig. S13, A and B, for more examples. Scale bars, 100 nm. (C) Quantification of UMOD-mediated formation of cell aggregates with light microscopy. The bar graph shows the normalized clump area observed upon incubation (2 hours) at 37°C of a constant amount of type 1–piliated E. coli HB101 (FimAra) (~108 cells/mL), with different concentrations of UMOD filaments (0.08 to 5.0 μM; concentrations refer to the UMOD monomer). The presence of 10 mM d-mannose inhibited aggregation across all UMOD concentrations tested. Representative light microscopy images are shown on the right (without and with 10 mM Man). Bars represent the means of biological triplicates ± SEM. *P < 0.05 difference between groups [two-way analysis of variance (ANOVA), post hoc Tukey test]. Scale bars, 10 μm. (D) Cell aggregates colocalized with UMOD filaments. Bright-field and TIRF microscopy images of type 1–piliated E. coli HB101 (FimAra) cells (excitation at 405 nm, green) that were incubated with fluorescent UMOD-Alexa647 filaments (red) are shown. Scale bars, 10 μm. (E) UMOD-induced E. coli aggregates were plunge-frozen, thinned by cryo-FIB milling, and imaged with cryo-ET (an 8.6-nm tomographic slice is shown). E. coli cells (brown) were always seen embedded in a highly dense meshwork of UMOD filaments (blue arrows). For further examples, see fig. S19. Scale bar, 100 nm.

Using light microscopy, we investigated bacteria-UMOD interactions on a larger scale. Incubation of type 1–piliated E. coli with UMOD filaments resulted in the aggregation of bacteria and the formation of clumps that consisted of tens to hundreds of bacteria (Fig. 3C). Cell clumping occurred across a wide range of UMOD concentrations and was inhibited by an excess of d-mannose (Fig. 3C), which indicates that clumping was caused by FimH binding to UMOD glycans. However, preformed clumps proved to be resistant against dissociation by d-mannose (fig. S15). Total internal reflection fluorescence (TIRF) microscopy showed that cell aggregates colocalized with UMOD (Fig. 3D). UMOD-dependent cell aggregation also occurred with a UPEC strain (fig. S16) but was not detected with nonpiliated E. coli or with piliated E. coli incubated with eUMOD (figs. S17 and S18). To visualize the UMOD-induced cell aggregates by cryo-ET, we thinned the sample using cryo–focused ion beam (FIB) milling (23). The cryotomograms consistently revealed bacteria that were tightly surrounded by a dense mesh of UMOD filaments (n = 17 tomograms; Fig. 3E, fig. S19, and movie S5).

In addition to type 1 pili, many UPEC strains express multiple pili with diverse glycan specificities. The UPEC strain CFT073 encodes type 1 pili (mannoside-specific lectin), F9 or Fml pili (lectin specific for galactosides), and S-pili (lectin specific for sialic acid) (2426). We therefore tested the inhibition of UMOD-mediated cell clumping of CFT073 by the addition of d-mannose, d-galactose, and sialic acid. Although d-mannose alone only slightly decreased cell aggregation (fig. S20, A and B), the addition of a cocktail of all three monosaccharides proved to be most effective in reducing UMOD-mediated cell clumping (fig. S20B). This indicates that the complex-type UMOD glycans might interact with different types of pilus adhesins presented by different uropathogens.

We verified these findings by analyzing unprocessed urine from patients with clinically diagnosed UTIs using light microscopy and cryo-ET. Urine from a patient with an E. coli UTI revealed bacterial clumps that were embedded into fibrous structures (Fig. 4A). Individual bacteria from the same sample were heavily piliated and always surrounded by numerous UMOD filaments (Fig. 4B and movie S6; n = 27 tomograms); several contact sites between pilus tips and UMOD were resolved (Fig. 4C and movie S6). Furthermore, we analyzed urine from UTI patients with other pathogens, including Klebsiella pneumoniae, Pseudomonas aeruginosa, and Streptococcus mitis. In all cases, we observed cell aggregates associated with fibrous structures using light microscopy, and 67 to 71% of cells imaged by cryo-ET confirmed the presence of UMOD filaments (fig. S21 and movie S7).

Fig. 4 Urine from UTI patients shows cell aggregation and pilus-mediated UMOD association with bacterial cells.

(A) Differential interference contrast (DIC) light microscopy imaging of urine from a patient with acute E. coli UTI revealed clustered bacteria. Scale bar, 10 μm. (B) Cryo-ET imaging of urine from the same patient showed piliated (orange arrowheads) bacterial cells (labeled E. coli). All analyzed cells (n = 27 tomograms or cells; two representative examples are shown) were surrounded by filaments with the typical UMOD appearance (blue arrows). Scale bars, 100 nm. (C) The cryotomograms revealed multiple contact sites (black arrowheads) between pili (orange arrowheads) and UMOD (blue arrows). Scale bars, 100 nm. Magnified views are provided in the insets. Scale bars, 50 nm.

Our data provide a three-dimensional (3D) structure of native UMOD filaments and support the hypothesis that the polymerization of UMOD into a multivalent filament is required for its function as a FimH antagonist to effectively compete with the high concentration of uroplakin 1a on the urinary epithelium (4, 14, 27). The intrinsic flexibility of UMOD filaments allows their adaptation to the bacterial surface, multivalent binding to pili from the same bacterium, and eventually encapsulation of entire cells. Analogous to the mechanism by which antibody–cross-linked clumps of enchained Salmonella cells facilitate the removal of pathogens from the gut (28), UMOD-mediated cell aggregation may have a role in the efficient clearance of bacteria through micturition.

Our study also sheds light on interactions between UMOD and pathogens other than type 1–piliated UPEC. UPEC genomes, for example, encode up to 16 fimbrial gene clusters, and UPEC switch expression between pilus types with distinct receptor specificities (24, 2931). The seven different complex-type N-glycans on UMOD (Fig. 1A) may therefore competitively inhibit adhesion of other pathogens to other epithelial receptors. Finally, the resolved site-specific N-glycosylation pattern and architecture of UMOD filaments will serve as a framework for studying the mechanisms that underlie UMOD’s roles in the regulation of salt transport, kidney diseases, and innate immunity (1, 19).

Supplementary Materials

Materials and Methods

Figs. S1 to S21

Tables S1 to S4

References (3263)

MDAR Reproducibility Checklist

Movies S1 to S7

References and Notes

Acknowledgments: We thank the Functional Genomics Center Zürich, specifically S. Chesnov and P. Hunziker for performing the matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) analysis and Edman sequencing, H. Debaix for the genotyping; E. Olinger and A. Yoshifuji for support in urinary UMOD analyses; P. Tittmann and M. Peterek for technical support during electron microscopy data collection; and ScopeM for instrument access at ETH Zürich. F. Eisenstein is acknowledged for help with subtomogram average processing and J. Xu for help with analyzing cryo-ET data. We thank G. Navarra and B. Ernst (University of Basel) for providing the fluorescent mannoside GN-FP. C. Giese is acknowledged for providing E. coli AAEC189 (pSH2) cells. T. Hennet is acknowledged for initial discussions about UMOD glycosylation. Funding: G.L.W. was supported by a Boehringer Ingelheim Fonds Ph.D. fellowship. M.P. was supported by the Swiss National Science Foundation (no. 31003A_179255), the European Research Council (no. 679209), and the NOMIS foundation. R.G. was supported by the Swiss National Science Foundation (nos. 310030B_176403/1 and 31003A_156304). O.D. was supported by the Swiss National Centre of Competence in Research Kidney Control of Homeostasis (NCCR Kidney.CH) program, the Swiss National Science Foundation (no. 310030_189044), and the Rare Disease Initiative Zürich (Radiz). J.T. was supported by the Swiss National Science Foundation (nos. PZ00P3_161147 and PZ00P3_183777). M.P. and R.G. were also supported by basic funding from ETH Zürich. Author contributions: G.L.W., J.J.S., M.M.S., O.D., M.P., and R.G. designed experiments. J.J.S., M.M.S., and J.E. purified and biophysically analyzed UMOD. C.-W.L. and M.A. performed and analyzed MS experiments. G.L.W. collected and processed cryo-ET data. D.S.Z. constructed the expression plasmid and performed TIRF microscopy. J.J.S. performed and analyzed light microscopy experiments. G.L.W. and J.T. collected patient urine. G.L.W., J.J.S., M.M.S., M.P., and R.G. wrote the manuscript with comments from all authors. Competing interests: The authors declare no competing interests. Data and materials availability: All subtomogram averages shown in this study were uploaded to the Electron Microscopy Databank (EMD) together with their respective half maps, masks for Fourier shell correlation (FSC) calculations, and an example tomogram. Accession numbers: EMD-11128 and EMD-11129 (native UMOD), EMD-11130 and EMD11131 (eUMOD), EMD-1133 and EMD-11134 (UMOD incubated with FimHL), and EMD-11134 and EMD-11135 (eUMOD incubated with FimHL). Example tomograms of bacterial cells with UMOD were also uploaded to the EMD with accession numbers EMD-11136 to EMD-11143. All other data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article