Human brain endothelial ATP synthase β-subunit is mannose-insensitive binding target of FimH

Abstract

Binding of meningitis-causing E. coli K1 to human brain microvascular endothelial cells (HBMEC) contributes to traversal of the blood-brain barrier, which occurs in part by the mannose-sensitive binding of FimH. In this study, we showed that FimH also binds to HBMEC, independent of mannose, and identified ATP synthase β-subunit and actin proteins from the surface biotinylated HBMEC as the mannose-insensitive binding targets for FimH. Co-immunoprecipitation experiments in the presence of α-methyl mannose verified the binding of FimH to ATP synthase β-subunit of HBMEC. ATP synthase β-subunit antibody decreased E. coli K1 binding to HBMEC in the presence of α-methyl mannose. Taken together, these findings demonstrate that FimH of E. coli K1 binds to HBMEC in both mannose-sensitive and -insensitive manner.

INTRODUCTION

Most cases of E. coli meningitis develop as a result of hematogenous spread (Kim, 2008), but it is incompletely understood how circulating E. coli traverses the blood-brain barrier. We have shown that successful traversal of the blood-brain barrier by circulating E. coli K1 requires E. coli binding to and invasion of human brain microvascular endothelial cells (HBMEC), which constitute the blood-brain barrier (Kim, 2008).

We have identified several E. coli K1 structures contributing to binding to and invasion of HBMEC, such as type 1 fimbriae and outer membrane protein A for binding, and Ibe proteins and cytotoxic necrotizing factor 1 for invasion (Kim, 2008), but it remains incompletely understood how those E. coli structures contribute to binding to and invasion of HBMEC.

Type 1 fimbriae are encoded by a fim gene cluster, including at least nine genes required for their biosynthesis (Orndorff & Falkow, 1984). The lectin-like adhesin, FimH, located at the tip of the fimbrial shaft (Hanson & Brinton, 1988) is responsible for the mannose-sensitive binding to host cells, including HBMEC (Teng et al, 2005). We have previously identified CD48 on the surface of HBMEC as the mannose-sensitive binding receptor for FimH (Khan et al, 2007).

The expression of type 1 fimbriae is phase variable (Abraham et al, 1985), and a wild type E. coli strain is a heterogeneous mixture of two subpopulations, i.e., phase-on subpopulation, which expresses type 1 fimbriae, and phase-off subpopulation, which does not express type 1 fimbriae (Teng et al, 2005). To examine the role of type 1 fimbriae in E. coli K1 binding to HBMEC, we constructed isogenic phase-locked mutants of strain RS 218 whose fim promoter-containing invertible elements are fixed in either on or off orientation (Teng et al, 2005), representing type 1 fimbriated (fim⁺) and non-fimbriated strains (fim⁻), respectively. We showed that excessive amount of α-methyl mannose decreased the HBMEC binding of fim⁺ strain, but not to the level of fim⁻ strain, while FimH reduced the binding of fim⁺ strain to the level of fim⁻ strain, suggesting that FimH binding to HBMEC may occur independent of mannose. In the present study, we showed that FimH exhibited the mannose-independent binding to HBMEC, and identified for the first time the HBMEC surface-localized ATP synthase β-subunit as a mannose-insensitive binding target of FimH protein.

MATERIALS AND METHODS

Antibodies and Reagents

Monoclonal anti-ATP synthase β subunit antibodies were purchased from BD Biosciences (clone # 10, San Jose, CA) and Affinity BioReagents (clone # 4.3E8.D10, Golden, CO), monoclonal anti-β-actin antibody (clone # ACTN05 [C4]) from Abcam (Cambridge, MA), goat anti-biotin serum for co-immunoprecipitation and HRP-conjugated goat anti-biotin antibody for Western blotting from Fitzgerald Industrial International, Inc. (Concord, MA) and Cell Signaling Technology (Beverly MA), respectively, and FITC-conjugated and unconjugated donkey anti-mouse IgG antibodies from Jackson ImmunoResearch Laboratories, Inc. (Baltimore, MD). EZ-Link sulfo-NHS Biotin for surface biotinylation, AminoLink plus immobilization kit for making affinity columns, and M-PER mammalian protein extraction reagent were purchased from Pierce (Rockford, IL), mammalian protease inhibitor cocktail and α-methyl mannose (methyl α-D mannopyranoside) from Sigma (St. Louis, MO), and protein A agarose fast flow bead from Upstate (Lake Placid, NY). Precision Plus Protein All Blue Standards from BioRad (Hercules, CA) was used for molecular weight standard.

E. coli binding assay in HBMEC

HBMEC were isolated and cultivated as described previously (Stins et al, 1997). The ability of E. coli strains to bind to HBMEC was examined as previously described (Shin et al, 2005).

FimCH and FimC protein purification and affinity chromatography

To purify functionally active FimH, the co-purification method with FimC, a periplasmic chaperon of type 1 pilus subunit proteins was used as described previously (Lee et al, 2005). FimC protein also was purified and used as a negative control. To make affinity column, 1.5 mg FimCH or FimC proteins were covalently immobilized in a 1 ml bed-volume of AminoLink plus coupling beads in 0.1 M sodium citrate and 0.05M sodium carbonate, pH 10. Surface biotinylation of HBMEC was performed on HBMEC monolayers grown on the plate as described in the manufacturer's manual. HBMEC monolayers were washed with ice-cold PBS and lysed with M-PER mammalian protein extraction reagent with mammalian protease inhibitor cocktail, and the insoluble debris was removed by centrifugation (20,000g at 4 °C). 100 mM α-methyl mannose was added to the lysate (10 mg), and the mixture was loaded onto the FimC (negative control) immobilized column which was equilibrated with M-PER reagent containing 100 mM α-methyl mannose (binding buffer). The FimC affinity column eliminates the nonspecific interacting proteins with column beads and FimC protein as well as to minimize any effect of any mannose-binding proteins. The pass-through fractions were reloaded to the FimCH immobilized column, and the column was washed with 10 bed-volume of the binding buffer. The FimH binding proteins were eluted with 0.2 N glycine buffer, pH 2.5, and the elution fractions were neutralized with one tenth volume of 1 M Tris, pH 9.5. Proteins were concentrated with Centricon (m.w. 5 kDa, Millipore, Billerica, MA), desalted and equilibrated with 20 mM MOPS (pH 7.0) using Quick Spin protein column (Roche, Indianapolis, IN). The protein samples were separated on SDS-PAGE (Novex TG and tris-acetate NuPAGE gels, Invitrogen) and two-dimensional gel electrophoresis with ReadyStrip IPG Strips and Criterion pre-cast gel (BioRad).

Mass spectrometry for protein identification

Protein treatment, obtaining peptide mass fingerprints, and identifying peptides were performed by the Mass Spectrometry/Proteomics Facility at Johns Hopkins School of Medicine (www.hopkinsmedicine.org/msf/). The Coomassie-stained protein bands were excised from the gel and in-gel digested by trypsin. After desalting process, a mass list of peptides was obtained for each protein digest digested peptides by MALDI-TOF mass spectrometer (Voyager DE-STR). MS-FIT (http://prospector.ucsf.edu/uscfftml3.2/msfit.htm) and MASCOT (http://www.matrixscience.com) software were used to identify the proteins.

Co-immunoprecipitation and Western blotting

To verify protein-protein interaction (i.e., FimH-ATP synthase β subunit), purified FimCH (5 μg) was mixed with 200 μg HBMEC lysates at 4 °C for 3 h to allow the binding complex to form between FimH and ATP synthase β-subunit of the HBMEC lysates. For a negative control 2.5 μg FimC protein was used to adjust for molar ratio with FimCH. To pull-down the FimH-ATP synthase β-subunit complex, 10 μg of affinity-purified anti-FimH rabbit serum or 5 μg of anti-ATP synthase β-subunit antibody (BD Biosciences) was added and incubated overnight at 4 °C. Protein A agarose beads were incubated with the protein-antibody mixture at 4 °C for 3h, and then precipitated by centrifugation (5,000 g, for 1 min). In the case of the pull-down with anti-biotin antibody, 10 μg of anti-biotin serum was used. Protein complexes were separated by SDS-PAGE using Novex TG gel and the separated proteins were transferred to PVDF membranes. The membranes were blocked with TBST (20 mM Tris [pH 7.5], 150 mM NaCl, 0.1% Tween-20) containing 5 % BSA for 1 hr at room temperature and incubated with anti-ATP synthase β-subunit and FimH antibodies overnight at 4°C. The blots were washed with TBST and incubated with a horseradish peroxidase-conjugated anti-mouse or rabbit IgG antibody (1: 5000 dilution, Cell Signaling) in 5 % skim milk-TBST for 1 hr at room temperature. For probing biotinylated proteins, membranes were blocked with 5 % skim milk-TBST and incubated with HRP-conjugated anti-biotin antibody (Cell Signaling) at room temperature for 1h. The blots were washed with TBST and developed with ECL Western detection reagent (Amersham Biosciences).

RESULTS AND DISCUSSION

Mannose-insensitive binding of FimH to HBMEC

We have previously shown that type 1 fimbriae contribute to the binding of meningitis-causing E. coli K1 strain RS 218 to HBMEC and the binding was significantly reduced by α-methyl mannose, but α-methyl mannose did not decrease the HBMEC binding of E. coli K1 to the level of fim⁻ mutant (Teng et al, 2005). This raised a question on whether FimH interaction with mannose-containing molecules is wholly responsible for FimH-mediated binding of E. coli K1 to HBMEC. To address this question, we first examined the effect of α-methyl mannose on fim⁺ and fim⁻ E. coli K1 binding to HBMEC. The binding to HBMEC was approximately 10-fold greater with fim⁺ E. coli K1 than with its isogenic fim⁻ strain (Table 1), which is consistent with our previous finding (Teng et al, 2005). The addition of α-methyl mannose (10 mM), as expected, decreased the binding of fim⁺ E. coli K1 to HBMEC, but failed to affect the HBMEC binding of fim⁻ strain. The same concentration of other carbohydrate (e.g., galactose) did not affect the binding of both E. coli strains. The addition of higher concentrations of α-methyl mannose did not further decrease the binding of E. coli K1 to HBMEC (data not shown), suggesting that 10 mM concentration of α-methyl mannose may be close to its saturated concentration. Of interest, the binding of fim⁺ E. coli to HBMEC in the presence of α-methyl mannose 10 mM was 3-fold higher than that of the fim⁻ E. coli (Table 1). These findings suggest that type 1 fimbriated E. coli binding to HBMEC may not be entirely due to its interaction with mannose-containing molecules on HBMEC.

Table 1.

The effect of α-methyl mannose on the binding of E. coli RS218 derivative strains to HBMEC. HBMEC were infected with type 1 fimbria locked-on (fim⁺) and –off (fim⁻) E. coli strains for binding assay. To test the blocking effect of α-methyl mannose, sugars were added to HBMEC before E. coli infection. The HBMEC-bound E. coli were enumerated by culturing on trypticase soy agar with 5 % sheep blood. Results are expressed as percent relative binding compared to percent HBMEC binding of fim⁺ strain without adding sugar.

Strain	sugara	% bindingb
fim +	none	100
	galactose	107.9 ± 8.0
	α-methyl mannose	34.6 ± 3.0c
fim −	none	8.7 ± 2.1
	galactose	11.5 ± 1.6
	α-methyl mannose	12.2 ± 2.0

^a10 mM sugar was added

^bMean ± SD of three independent experiments, each performed in triplicate.

^csignificantly less than no sugar or galactose (p<0.01)

We next examined whether FimH mediates the mannose-insensitive binding of type 1 fimbriae to HBMEC. FimH protein complexed with FimC periplasmic chaperon represents functionally active FimH protein (Choudhury, 1999; Vetsch, 2002). As shown in Table 2, 50 μg of FimCH reduced the HBMEC binding of fim⁺ E. coli to the level of fim⁻ strain in the presence of α-methyl mannose. These findings suggest that FimH can interact with HBMEC surface, independent of mannose. We, therefore, hypothesize that there may be the mannose-insensitive receptor(s) for FimH on the HBMEC surface, which interacts with type 1 fimbriated E. coli. Here we presented the identification of the mannose-insensitive FimH receptors on the HBMEC surface.

Table 2.

The effect of purified FimH protein on the binding of fim⁺ and fim⁻ E. coli strains to HBMEC in the presence of 10 mM α-methyl mannose. HBMEC were preincubated with indicated amounts of purified proteins for 15 min at room temperature, and then infected with E. coli. The HBMEC-bound E. coli were enumerated by plating on trypticase soy agar with 5 % sheep blood. Results are expressed as percent relative binding compared to percent HBMEC binding of fim⁺ strain without protein.

Strain	Protein	% bindinga
fim +	none	100
	FimC 10 μg/well	105.2 ± 7.3
	FimC 50 μg/well	96.8 ± 5.9
	FimCH 10 μg/well	77.6 ± 9.3b
	FimCH 50 μg/well	35.3 ± 3.1b
fim −	none	30.6 ± 2.7c

^aMean ± SD of three independent experiments, each performed in triplicate.

^bsignificantly less than no protein (p<0.05).

^csignificantly less than fim⁺ with no protein (p<0.05).

HBMEC surface proteins interacting with FimH in a mannose-insensitive manner.

To identify mannose-insensitive FimH interacting proteins from the HBMEC surface, FimH-affinity chromatography was performed using surface-biotinylated HBMEC lysates in a mannose-oversaturated condition (i.e., 100 mM α-methyl mannose). For constructing affinity column, FimC protein alone or FimCH complex was immobilized to the agarose beads as described in Methods. The lysates of surface biotinylated HBMEC flowed through the FimC immobilization column were subjected to the FimCH column in order to identify FimH-specific HBMEC surface protein(s), and proteins interacted with FimH were eluted by acid pH (0.2 N glycine, pH 2.5).

Figure 1A showed the elution fraction of HBMEC surface proteins probed with anti-biotin antibody from FimCH affinity column. Fraction 3 contained major biotin signals. Concentrated proteins from the fraction 3 were separated and probed with anti-biotin antibody (right panel of Figure 1B). The same membrane was stained with Coomassie blue to demonstrate the total FimH interacting proteins from the HBMEC lysates (left panel of Figure 1B). Biotinylated signals ranging from 75 to 200 kDa were insufficient to determine protein identities using mass spectrometry. We were unable to recover each single spot from silver stained two-dimensional gel (data not shown). Four proteins of approximate 35, 40, 45 and 50 kDa were identified by Coomassie blue stained membrane, but only 50 and 40 kDa bands (band 1 and 2 in Figure 1B) had biotin signals, indicating their location on the HBMEC surface. The protein identities of band 1 and 2 were determined as ATP synthase β-subunit (NCBI accession number P06576, 32% coverage, MOWSE score of 3.2E+04) and cytoplasmic actin (β-[P02570] and γ-[P02571] isoforms share the same peptides, 63% coverage, MOWSE score of 5.75E+15), respectively. The 45 kDa protein which did not have biotin signal was identified as Elongation Factor 1-α1 (EF-1-α-1 [P04720], 19 % coverage, MOWSE score of 1.77E+05). ATP synthase β-subunit is the major protein interacting with FimH, as shown in Coomassie stained band 1 (Figure 1B), but its biotin signal was weaker than that of actins. Cell surface-localized surface ATP synthase β-subunit (biotinylated) as well as the mitochondrial ATP synthase β-subunit (non-biotinylated) are likely to bind to FimH during the affinity chromatography, resulting in overall weaker biotin signal. In contrast, cytoplasmic actins, which are in the state of cytoskeletal filaments, were likely to be eliminated by removing insoluble fraction of the HBMEC lysates during cell lysate preparation, resulting in surface-localized actins interacting mainly with FimH in the process of affinity chromatography.

Fig 1. Affinity purification of FimH interacting protein on HBMEC surface.

(A) 35 μl of neutralized eluent from each fraction were separated, blotted into PVDF membrane, and probed by anti-biotin antibody. (B) Concentrated proteins from fraction number 3 were subjected to Western blotting with anti-biotin antibody (right panel). The same membrane was stained with Coomassie blue (left panel).

ATP synthase β-subunit is one component of F₁-F₀ ATP synthase complex (hereafter referred to as ATP synthase) localized in mitochondrial cristae. The β-subunit has been found on the tumor cell surface and found to have a role in the lymphocyte-mediated cytotoxicity (Das et al, 1994; Di Virgilio et al, 1989). Moreover, differentiated endothelial cells including human umbilical vein endothelial cells and dermal microvascular endothelial cells express the whole F₁-F₀ ATP synthase complex on their surface, and this endothelial cell surface ATP synthase α- and β-subunit functions as a receptor for angiostatin (Moser et al, 1999). Cytoplasmic actin β- and γ- isoforms are present in the cytoplasm of non-muscle cells (Sheterlin et al, 1988). In hematopietic cells including dendritic cells and activated platelets, cytoplasmic actins are secreted to extracellular environment along with other cytoplamic proteins (Coppinger et al, 2004; Thery et al, 2001; Thery et al, 1999). Dudani et al., have isolated cytoplasmic actin that is localized on the surface of venous endothelial cells, and functions as a receptor for plasminogen, tissue plasminogen activator and lipoprotein (a) (Dudani & Ganz, 1996). The biotin labeling of ATP synthase β-subunit and actin in Figure 1B suggests that both ATP synthase β-subunit and cytoplasmic actin are localized on the surface of HBMEC. To support this finding, the surface location of ATP synthase β-subunit and β-actin on HBMEC was demonstrated by immunofluorescence microscopy (Supplementary Figure S1). These findings suggest that these proteins function as mannose-insensitive surface targets for FimH. To support this concept, we further characterized the interaction between ATP synthase β-subunit and FimH.

Interaction between ATP synthase β-subunit and FimH

To verify the mannose-insensitive FimH binding to ATP synthase β-subunit of HBMEC, co-immunoprecipitation experiments of HBMEC lysates were performed in the presence of α-methyl mannose (100 mM). To minimize the non-specific interaction with protein-A agarose beads, the mixture of FimCH and HBMEC lysates were pre-incubated with protein-A agarose beads, and the non-specific complex was removed by centrifugation. The FimH-ATP synthase β-subunit complex was precipitated using anti-FimH antibody from HBMEC lysates pre-incubated with FimCH complex, as shown by Western blotting with anti-ATP synthase β-subunit antibody (Figure 3A). Controls for the non-specific reaction of anti-FimH serum with ATP synthase β-subunit protein and rabbit serum (second and third lane of Figure 3A, respectively) revealed no ATP synthase β-subunit co-immunoprecipitated from HBMEC lysates.

We used FimCH complex as a functionally active FimH, and next examined whether FimC portion of FimCH complex interacted with ATP synthase β-subunit by immunoprecipitating the mixture of biotinylated FimC and FimCH proteins and HBMEC lysate with anti-biotin antibody (Figure 3B). Only ATP synthase β-subunit interacted with biotinylated FimCH (first lane), whereas biotinylated FimC (second lane) and anti-biotin antibody itself (third lane) did not reveal ATP synthase β-subunit from HBMEC lysates.

For additional validation of the FimH interaction with ATP synthase β-subunit, we performed co-immunoprecipitation of HBMEC lysates and FimCH mixture with anti-ATP synthase β-subunit antibody, which was probed with anti-FimH antibody (Figure 3C). FimH was detected only when anti-ATP synthase β-subunit antibody was used along with HBMEC lysates and FimCH (first lane of Figure 3C). These lines of evidence indicate that ATP synthase β-subunit is the mannose-insensitive interacting target for FimH.

Anti-ATP synthase β-subunit antibody affects E. coli K1 binding to HBMEC

We next examined whether anti-ATP synthase β-subunit antibody blocks the E. coli K1 binding to HBMEC in the presence of 10mM α-methyl mannose. As shown in Table 3, anti-ATP synthase β-subunit antibody blocked the HBMEC binding of fim⁺ strain in a dose-dependent manner compared with anti-mouse IgG control, while it did not affect the binding of fim⁻ E. coli to HBMEC (Table 3). However, 2 μg of anti-ATP synthase antibody did not decrease the HBMEC binding of fim⁺ E. coli to the level of fim⁻ E. coli (65% vs 29 % for fim⁺ and fim⁻ E. coli, respectively). These findings suggest that 2 μg of the antibody may not be sufficient for complete blocking of FimH binding to ATP synthase β-subunit, or the binding epitope of anti-ATP synthase β-subunit antibody, clone # 4.3E8.D10 (Affinity BioReagents Co.) may not be completely overlapped with FimH binding site on the ATP synthase β-subunit. Another possibility is that ATP synthase β-subunit may be one of several mannose-insensitive binding targets on HBMEC for fim⁺E. coli K1.

Table 3.

Anti-ATP synthase β-subunit antibody blocks fim⁺ E. coli binding to HBMEC. HBMEC were incubated with indicated amounts of antibodies for 15 min at room temperature, and then infected with E. coli. The HBMEC-bound E. coli were enumerated by culturing on trypticase soy agar with 5 % sheep blood. Results are expressed as percent relative binding compared to percent HBMEC binding of fim⁺ strain preincubated with anti-mouse IgG antibody. This experiment was performed in the presence of saturated amount of α-methyl mannose (10 mM).

Strain	Antibody	% bindinga
fim +	Anti-mouse IgG (2 μg/well)	100
	Anti-ATP synthase (β-subunit (0.5 μg/well)	90.4 ± 5.7
	Anti-ATP synthase β-subunit (2 μg/well)	64.7 ± 3.7b
fim −	Anti-mouse IgG or anti-ATP synthase β-subunit (2 μg/well)	29.4 ± 1.4c

^aMean ± SD of three experiments, each performed in triplicate.

^bsignificantly less than anti-mouse IgG (p<0.05).

^csignificantly less than fim⁺ with anti-mouse IgG (p<0.05).

In summary, type 1 fimbriated E. coli K1 binds to HBMEC in both mannose-sensitive and -insensitive manner. We have identified that CD48 is the mannose-containing HBMEC surface receptor interacting with FimH (Khan et al, 2007). In the present study, the mannose-insensitive receptors for FimH on the surface of HBMEC were identified, which include ATP synthase β-subunit. The mannose-insensitive FimH binding may contribute to E. coli K1 binding to HBMEC in the mannose moiety-rich environment such as bloodstream, where meningitis-causing E. coli K1 interacts with the blood-brain barrier to penetrate into the central nervous sytem. Additional studies are needed to further elucidate the role of mannose-insensitive HBMEC binding in the pathogenesis of E. coli K1 meningitis.

Fig 2. Co-immunoprecipitation for proving FimH interaction with ATP synthase β-subunit in the presence of α-methyl mannose (100 mM).

(A) FimH- ATP synthase β-subunit complex of HBMEC lysates was pulled-down with anti-FimH antibody, and then Western blotting with ATP synthase β-subunit antibody (WB: ATPS beta) was performed. (B) To exclude the possibility of FimC interaction with ATP synthase β-subunit, biotinylated FimC- and FimCH-ATP synthase β-subunit complexes of HBMEC lysates were pulled-down using anti-biotin antibody. The pull-downed complex was subjected to Western blotting with ATP synthase β-subunit antibody. (C) FimH- ATP synthase β-subunit complex was pulled-down with anti-ATP synthase β-subunit antibody, and then probed with anti-FimH antibody.