Indispensable functions of ABL and PDGF receptor kinases in epithelial adherence of attaching/effacing pathogens under physiological conditions

Abstract

Enteropathogenic Escherichia coli (EPEC) and Citrobacter rodentium are attaching-and-effacing (A/E) pathogens that cause intestinal inflammation and diarrhea. The bacteria adhere to the intestinal epithelium, destroy microvilli, and induce actin-filled membranous pedestals but do not invade the mucosa. Adherence leads to activation of several host cell kinases, including FYN, n-SRC, YES, ABL, and ARG, phosphorylation of the bacterial translocated intimin receptor, and actin polymerization and pedestal formation in cultured cells. However, marked functional redundancy appears to exist between kinases, and their physiological importance in A/E pathogen infections has remained unclear. To address this question, we employed a novel dynamic in vitro infection model that mimics transient and short-term interactions in the intestinal tract. Screening of a kinase inhibitor library and RNA interference experiments in vitro revealed that ABL and platelet-derived growth factor (PDGF) receptor (PDGFR) kinases, as well as p38 MAP kinase, have unique, indispensable roles in early attachment of EPEC to epithelial cells under dynamic infection conditions. Studies with mutant EPEC showed that the attachment functions of ABL and PDGFR were independent of the intimin receptor but required bacterial bundle-forming pili. Furthermore, inhibition of ABL and PDGFR with imatinib protected against infection of mice with modest loads of C. rodentium, whereas the kinases were dispensable for high inocula or late after infection. These results indicate that ABL and PDGFR have indispensable roles in early A/E pathogen attachment to intestinal epithelial cells and for in vivo infection with limiting inocula but are not required for late intimate bacterial attachment or high inoculum infections.

enteropathogenic Escherichia coli (EPEC) can cause serious diarrheal illness with potentially high mortality in infants, especially in developing countries (5). EPEC is classified as an attaching-and-effacing (A/E) pathogen because of its ability to adhere to the intestinal epithelium, efface microvilli, and induce characteristic, actin-filled membranous pedestals (25). Adherence to the epithelium is critical in pathogenesis, since the bacteria are only minimally invasive into the mucosa. Several bacterial factors that mediate attachment, including type IV bundle-forming pili (BFP) (17), EspA filaments (25), and the adhesin intimin (12), have been identified. Initial attachment of EPEC can be observed within 10 min (8) and is partly mediated by the plasmid-encoded BFP (13). After 30 min, morphologically distinct bacterial microcolonies form on epithelial cells, a process that is also aided by BFP (17, 41).

These initial stages of attachment are followed by translocation of several bacterial effector proteins into host epithelial cells. Translocation is mediated by a type III secretion system that acts as a molecular syringe. EPEC and the murine A/E pathogen Citrobacter rodentium harbor a 30- to 40-kb chromosomal pathogenicity island, the locus of enterocyte effacement (6), which carries genes encoding the proteins that form the type III secretion system (1), as well as translocated effector proteins (including EspB, EspF, EspG, EspH, EspZ, and Map), the translocated intimin receptor (Tir), and intimin. Tir, which is inserted into the epithelial plasma membrane, possesses a cytoplasmic tail and an extracellular domain that binds to bacterial intimin (15). Upon bacterial binding of EPEC, a signaling cascade that involves activation of several tyrosine kinases, phosphorylation of Tir (23), and actin polymerization and pedestal formation (22), the final stage of intimate EPEC attachment to the epithelium, is initiated over several hours in the host cell.

A number of host cell kinases, including the SRC kinase family members FYN, n-SRC, and YES, the ABL family kinases ABL (ABL1) and ARG (ABL2), and other factors, have been shown to be involved in tyrosine phosphorylation (4, 9, 18, 32, 40). These nonreceptor tyrosine kinases are located predominantly in the cytoplasm and play central roles in signal transduction and regulation of cytoskeletal functions. ABL and ARG are activated in pedestals and are sufficient for pedestal formation in standard cell culture models, but they are not necessary, because adherence proceeds normally even when they are blocked (40). These findings suggest that host cell kinases are functionally redundant for bacterial attachment and that EPEC may exploit different kinases in promiscuous ways to ensure phosphorylation of Tir. It has been postulated that a cascade of tyrosine kinases is involved in pedestal formation upon initial attachment of EPEC, since not only is tyrosine phosphorylation required, but also the polyproline region within Tir and several SRC homology 3 domains in host cell kinases (4). Furthermore, activation of ABL by host-derived factors, such as platelet-derived growth factor (PDGF) acting on its receptor tyrosine kinase PDGF receptor (PDGFR) (33), may indirectly contribute to bacterial attachment by promoting Tir phosphorylation or activation of other attachment pathways, thereby allowing EPEC to passively benefit from host kinase activation. Furthermore, it is possible that host kinases could have other functions that promote bacterial attachment independent of pedestal formation.

Despite extensive experimentation in cell culture models, the physiological relevance of the different kinases and signaling pathways in EPEC attachment is not clear. The goal of the present studies was to determine the physiological functions of host cell kinases in the pathogenesis of intestinal infections with A/E pathogens. Using a novel dynamic in vitro infection model, we demonstrate here that ABL and PDGFR kinases have indispensable functions in early bacterial attachment under conditions of limited bacterial loads in vitro and in vivo. Our data also suggest that ABL and PDGFR control host cell processes that are necessary for the early, BFP-mediated EPEC attachment to epithelial cells.

MATERIALS AND METHODS

Reagents and bacterial strains.

A library of pharmacological inhibitors of different kinases (Kinase Screening Library), including threonine and tyrosine kinases, MAP kinases, and phosphoinositide 3-kinase, was obtained from Cayman Chemical (Ann Arbor, MI). The compounds and target kinases are listed in Table 1. Additional kinase inhibitors, including imatinib, staurosporine, bosutinib, H-89, sorafenib, nilotinib, sunitinib, lapatinib, K252a, genistein, and gefitinib, were purchased from LC Laboratories (Woburn, MA). PD169316 was obtained from Axon Medchem (Groningen, Netherlands). All kinase inhibitors were used in the screening experiments at their approximate IC₅₀ for their respective target kinases (which were obtained by comprehensive literature searches) or at the highest possible concentration as limited by solubility (Table 1). Mechanistic studies beyond the screens were done at inhibitor concentrations commonly employed in vitro (31). Cycloheximide was obtained from Sigma Aldrich (St. Louis, MO). The following bacteria were used: wild-type EPEC strain E2348/69 (serotype O127:H6) (35), EPEC ΔbfpA (34), EPEC Δtir (24), and C. rodentium (28). Fluorescently labeled bacteria were constructed by transformation of an enhanced green fluorescent protein expression plasmid (Clontech, Mountain View, CA) into EPEC E2348/69.

Table 1.

Kinase inhibitors

No.	Inhibitor	Targeted Kinases	Reported IC50	Concentration in Screens
1	CAY 10578	Casein kinase 2	0.3 μM	0.3 μM
2	Lauric acid leelamide	Pyruvate dehydrogenase kinase	9.5 μM	10 μM
3	N,N-dimethylsphingosine	Sphingosine kinase	3.1 μM	3 μM
4	PD98059	MAPKK1, Raf, MEK	2–7 μM	2 μM
5	Piceatannol	IKBα kinase, other kinases	15 μM	15 μM
6	SB216763	Glycogen synthase kinase 3	34 nM	50 nM
7	JNJ10198409	PDGFR tyrosine kinase	4.2 nM	5 nM
8	Sphingosine kinase inhibitor 2	Sphingosine kinase	0.5 μM	0.5 μM
9	NSC 210902	Casein kinase 2	1 μM	1 μM
10	Arachidonic acid leelamide	Pyruvate dehydrogenase kinase	9.5 μM	10 μM
11	CAY 10561	ERK2	0.54 μM	1 μM
12	PD169316	p38 MAPK, ERK, PKA, PKCα	89 nM	100 nM
13	Triciribine	Akt-1, -2, -3	5–10 μM	5 μM
14	d-Erythro-sphingosine C-18	PKC, multiple kinases	5 μM	5 μM
15	O-1918	Non-CB1,CB2 receptor antagonist	<30 μM	10 μM
16	AG-17	EGFR, PDGFR	460 μM	10 μM
17	RG-13022	EGFR	1 μM	1 μM
18	AG-213	EGFR	2.4 μM	2.5 μM
19	Sunitinib	VEGFR, PDGFR	10 nM	10 nM
20	Wortmannin	PI3K, polo-like kinases, p38, MLCK	1–10 nM	1 nM
21	LFM-A13	Bruton's tyrosine kinase	2.5–10 μM	2.5 μM
22	Imatinib	ABL, ARG, PDGFR	280 nM	300 nM
23	Leelamine	Pyruvate dehydrogenase kinase	9.5 μM	10 μM
24	AS-604850	PI3K	0.25–4.5 μM	0.25 μM
25	AG-1296	PDGFR	0.4 μM	0.5 μM
26	CAY 10575	Cdc7	15.8 μM	16 μM
27	NH125	Eukaryotic elongation factor-2 kinase	60 nM	60 nM
28	(R)-roscovitine	CDK2/cyclin E	0.1 μM	0.1 μM
29	AG-1478	EGFR	3 nM	3 nM
30	AG-494	EGFR	1 μM	1 μM
31	AG-82	EGFR	3 μM	3 μM
32	CAY 10571	p38 MAPK	0.03 μM	0.03 μM
33	CCT018159	Hsp90	3.2–6.6 μM	3 μM
34	Olomoucine	CDK2/cyclin A, B, E, ERK1	7 μM	7 μM
35	AG-183	EGFR	0.8 μM	1 μM
36	AS-605240	PI3K, PKB	8–300 nM	10 nM
37	AG-99	EGFR	10 μM	10 μM
38	PD0325901	MEK	0.33 nM	0.5 nM
39	ZM336372	Raf-1	70 nM	70 nM
40	PI-103	PI3K, mTOR	2–150 nM	2 nM
41	PI3Kα inhibitor 2	PI3K	2–220 nM	2 nM
42	RG-14620	EGFR	3 μM	3 μM
43	CAY 10577	Casein kinase 2	0.8 μM	1 μM
44	Erbstatin analog	EGFR	2.5 μM	10 μM
45	CAY 10576	IKKϵ	40 nM	40 nM
46	U-0126	Mer	1.1 nM	1 nM
47	ML-9	MLCK, PKB/Akt, PKA, MAPK	10–100 μM	10 μM
48	LY294002	PI3K	1.4 μM	2 μM
49	SB203580	p38 MAPK, PDK1	0.6 μM	0.5 μM
50	AG-490	JAK2	0.1 μM	0.1 μM
51	Lavendustin C	EGFR	0.012 μM	0.01 μM
52	Y-27632	Rho	800 nM	800 nM
53	HA-1077	Rho-associated kinase II, PKC	1.9–15 μM	2 μM
54	H-8	PKA, PKG, PKC, MLCK	1.2 μM	1 μM
55	CAY 10575	IKKϵ	15.8 μM	16 μM
56	AS-25242s4	PI3K	30 nM	30 nM
57	Myrecetin	JAK, STAT3	6 μM	6 μM
58	AG-825	HER-2/neu, EGFR/PDGFR	50 μM	50 μM
59	AG-370	PDGFR	20 μM	20 μM
60	AG-18	EGFR	35 μM	35 μM
61	Janex1	JAK3	78 μM	30 μM
62	Staurosporine	PKC, multiple other kinases	2.7 nM	3 nM
63	Bosutinib	c-Src, c-Abl	1–1.2 nM	1 nM
64	H-89	PKA (PKG, PKCμ)	50 nM	50 nM
65	Sorafenib	VEGFR, PDGFR, Flt3, c-KIT	20–90 nM	20 nM
66	Nilotinib	BCR/ABL	15 nM	15 nM
67	Sunitinib	VEGFR, PDGFR	10 nM	10 nM
68	Lapatinib	ErbB2,4, EGFR	367 nM	400 nM
69	K252a	NGF receptor	3 nM	3 nM
70	Genistein	EGFR, other tyrosine kinases	2.6 μM	3 μM
71	Gefitinib	EGFR	0.2–0.4 μM	0.2 μM

IC₅₀ values are based on data from the published literature and are given as ranges or representative values for the major targets. Specific inhibitor concentrations used in the screens in Fig. 1 are listed in the last column.

CDK, cyclin-dependent kinase; EGFR, EGF receptor; MLCK, myosin light chain kinase; mTOR, mammalian target of rapamycin; NGF, nerve growth factor; PDGFR, platelet-derived growth factor receptor; PDK, phosphoinositide-dependent kinase; PI3K, phosphoinositide 3-kinase; VEGFR, VEGF receptor.

Cell culture and attachment assays.

Bacteria were grown overnight at 37°C in Luria-Bertani Broth (LB) with shaking, diluted 1:25 into LB, and grown with shaking for an additional 2 h. After the cultures were washed with PBS, bacterial density was estimated at 600-nm OD, and bacteria were resuspended in DMEM at the desired numbers. The human cervical epithelial cell line HeLa [American Type Culture Collection (ATCC) CCL-2] and the murine rectal epithelial cell line CMT-93 (ATCC CCL-223) were cultured in DMEM supplemented with 10% FCS and 1% penicillin-streptomycin. For infection experiments, cells were seeded into six-well plates and grown in antibiotic-free medium to 80–90% confluence, at which point the cultures contained ∼10⁶ adherent cells per well. The human colon epithelial cell line T84 (ATCC CCL-248) was seeded onto 42-mm² permeable filter supports (EMD Millipore, Billerica, MA) in six-well plates and grown to confluence over 10–14 days to reach a minimum transepithelial resistance of 1,000 Ω·cm². At 1 day before infection, the medium was changed to serum-free DMEM. Kinase inhibitors were added to the cells ≥30 min prior to infection and were present throughout the experiments. In some experiments, cells were fixed with 1% formalin in PBS for 30 min and washed extensively before infection. In other experiments, cycloheximide (5 μg/ml) was added 15 min prior to infection.

Bacteria were added at multiplicity of infection (MOI) of 100:1–1:1, and plates were placed onto a rocking shaker (Boekel Scientific, Feasterville, PA; 15° angle, 18 rpm) or left stationary for up to 3 h at 37°C. For assay of attached bacteria, cells were washed four times with warm PBS to remove nonadherent bacteria and incubated with 0.1% Triton X-100 in PBS for 5 min. Cells were scraped off the plates, and lysates were transferred to 1.5-ml microtubes. After vigorous vortexing, serial dilutions of the lysates were plated onto tryptic soy agar plates. Colony-forming units (CFUs) were counted after overnight incubation. For detachment assays, cells were infected with EPEC for 1 h and washed with PBS to remove nonadherent bacteria, and fresh medium containing the various inhibitors or solvent controls was added. After further incubation for 1 h, cells were washed, and numbers of attached bacteria were determined by CFU assay.

For testing attachment of EPEC ΔbfpA compared with the wild-type EPEC E2348/69, bacteria were subcultured in DMEM for 2 h prior to infection to facilitate BFP induction. Ten microliters of the bacterial culture were added to HeLa cells and incubated for 1 h with shaking, and bacterial attachment was assessed as described above. For testing bacterial viability, a 1:25 dilution of overnight cultures of EPEC was grown in DMEM for 2 h with and without kinase inhibitors, and viable bacteria were assessed by CFU assay.

siRNA transfections.

HeLa cells were grown in reduced serum medium (Opti-MEM, GIBCO, Life Technologies, Carlsbad, CA) to 80% confluence and transfected with Silencer siRNA (Ambion Life Technologies, Grand Island, NY) targeting ABL (ID no. 1431), ARG (ID no. 1478), p38 MAPK (ID no. s3585), or PDGFR (ID no. s10242) following the manufacturer's instructions with Lipofectamine RNAiMAX reagent. Controls were transfected with negative control siRNA, either Silencer negative control #1 siRNA for ABL and ARG or Silencer Select negative control siRNA for PDGFR and p38 (Ambion). After 48 h, transfected cells were either processed for protein isolation or infected for 1 h with EPEC at MOI of 3:1 under shaking conditions, and bacterial attachment was assessed.

Immunoblotting.

HeLa cells were lysed for 10 min on ice in a buffer containing 150 mM NaCl, 5 mM KCl, 10 mM HEPES, 0.5 mM EDTA, 0.2 mM EGTA, 1 mM sodium fluoride, 1 mM vanadate, 0.05% Nonidet P-40, 1 mM DTT, and a protease inhibitor cocktail (Roche Applied Science). After centrifugation to remove debris, equal protein amounts (20–40 μg) were boiled for 10 min in loading buffer (50 mM Tris, pH 6.8, 100 mM DTT, 2% SDS, 40% glycerol, and 0.2% bromophenol blue) and size-separated by electrophoresis on a 10% Tris-glycine polyacrylamide gel (Bio-Rad Laboratories). Proteins were electrotransferred to a polyvinylidene difluoride membrane, which was blocked and stained overnight at 4°C with antibodies against ABL (BD Biosciences), PDGFR (Cell Signaling), p38 MAPK (Santa Cruz Biotechnology), or β-actin (Cell Signaling) in a buffer containing 10 mM Tris, pH 7.6, 5% BSA, and 0.05% Tween 20, and then incubated with the appropriate horseradish peroxidase-conjugated secondary antibodies and detected by enhanced chemiluminescence (GE Healthcare).

Mice and infection protocols.

Wild-type C57BL/6J mice were obtained from Jackson Laboratory (Bar Harbor, ME). Mice were infected by oral gavage with 5 × 10⁸ CFU/mouse (or 1 × 10⁷ or 5 × 10⁹ CFU/mouse in selected experiments) of C. rodentium in PBS. To determine bacterial numbers in the stool, fecal pellets were collected from individual mice, weighed, and homogenized in 5 ml of PBS. Homogenates were plated onto MacConkey agar and CFU were counted after overnight incubation. The detection limit of the CFU assay was 10³ CFU/g feces. To confirm identity of single colonies, PCR analysis was done for espB of C. rodentium (28). For inhibitor treatments, mice were given the inhibitor daily by oral gavage (50 mg/kg imatinib) or intraperitoneal injection (5 mg/kg PD169316) starting 1 day before or 3 or 6 days after infection. All animal studies were reviewed and approved by the University of California, San Diego Institutional Animal Care and Use Committee.

Histological analysis.

For histological analysis, the colon was removed, opened longitudinally, cleaned, processed as a “Swiss roll,” and fixed for 24 h in 10% phosphate-buffered formalin. Fixed tissues were embedded in paraffin, and 5-μm sections were prepared and stained with hematoxylin and eosin.

Statistical analysis.

CFU counts from mice were log₁₀-transformed, and means ± SE were calculated from the log values. Mice without detectable bacteria in the stool were assigned a log₁₀ value equivalent to half of the detection limit of the CFU assay. Results from male and female mice were combined, as no significant differences were observed between the sexes in bacterial colonization or mucosal responses after infection. Data from bacterial attachment assays are expressed as means ± SE. Differences between groups were evaluated by t-test or Wilcoxon's rank-sum test, as appropriate; P < 0.05 was considered to be significant.

RESULTS

Functional screening of kinase inhibitor library for inhibition of EPEC attachment to epithelial cells.

Attachment of EPEC to cultured epithelial cells involves activation of a number of host cell kinases; yet marked functional redundancy appears to exist, and the physiological importance of these processes is not well understood (44). Most prior studies relied on high bacterial inocula to maximize bacterial-epithelial interactions (40), but these infection conditions may not adequately reflect the in vivo situation, where intestinal motility and luminal bulk flow presumably limit opportunities for bacterial attachment to the epithelium. In fact, after infection of mice with the A/E pathogen C. rodentium, >99% of the inoculum passes through the intestinal tract without persistent adherence to the mucosa (43). Therefore, we reasoned that more “dynamic” in vitro infection conditions may better model the physiological in vivo situation and, thereby, better reveal any unique functions of specific host cell kinase in EPEC attachment. We used 80–90% confluent HeLa cells as a model epithelium, a modest inoculum with a MOI of 3:1, and a 1-h incubation time with concurrent rocking of the culture plates at 18 cycles/min. Using this modified in vitro infection model, we screened a library of 71 different inhibitors of a wide range of host cell kinases (Table 1) for inhibition of EPEC attachment to epithelial cells. Significant reduction of attachment was observed for seven compounds (Fig. 1A), one or several of which inhibit ABL, ARG, PDGFR, and p38 MAPK (10, 11, 16, 26). Several additional compounds also appeared to reduce bacterial attachment, but not in a reliable and statistically significant manner. A few others appeared to promote attachment (Fig. 1A), although we did not pursue these observations in the present study, because inhibition of attachment was more promising from a potential therapeutic perspective.

Fig. 1.

Screen of kinase inhibitors for blocking of EPEC attachment to epithelial cells. A: cultures of 80–90% confluent HeLa cells in 6-well plates were treated with 1 of 71 kinase inhibitors (see Table 1) and inoculated with the wild-type enteropathogenic E. coli (EPEC) E2348/69 [3 × 10⁶ colony-forming units (CFU)/well] with continuous rocking for 1 h; then cultures were washed, and attached bacteria were enumerated by CFU assay. Data (means ± SE, n = 3) are shown as percent attachment relative to control cultures treated with the solvent DMSO. ○, Inhibitors that yielded significant (P < 0.05) inhibition. B: inhibitors were tested at 1×, 5×, and 25× concentrations in Table 1 for inhibition of EPEC attachment to HeLa cells. Data (means ± SE; n = 3 experiments) are shown as percent attachment relative to control cultures treated with the solvent DMSO.

Further testing of the seven compounds with significant inhibitory activity in the screens revealed that one, H-89, was directly toxic for EPEC when grown in LB and was not further examined, while the other six inhibitory compounds had no toxic effects on EPEC grown in LB or DMEM. Evaluation of these six kinase inhibitors showed that they suppressed bacterial attachment in a concentration-dependent manner, with approximate IC₅₀ values for attachment inhibition that were within the range of the IC₅₀ values for kinase inhibition for each of the respective inhibitors (Fig. 1B).

Dynamic infection conditions are critical for revealing effects of kinase inhibition.

Our inhibitor screens revealed significant inhibition of EPEC attachment by the ABL/ARG and PDGFR inhibitor imatinib, whereas prior studies did not find an effect of ABL inhibition on bacterial attachment (40). This apparent discrepancy suggested that the new dynamic infection conditions may have been important for our finding, because prior studies utilized static conditions (40). To test this notion, we systematically varied incubation time, inoculum, and plate movement during infection in our cell culture model with two goals: 1) to identify dynamic conditions that limit bacterial attachment but still allow for significant attachment to epithelial cells compared with parallel cultures without cells and 2) to find conditions that maximize imatinib inhibition of bacterial attachment.

Variation of the incubation time led to a marked reduction of bacterial attachment to HeLa cells from 100–200% of the initial inoculum under standard conditions (i.e., 3-h incubation at 37°C under stationary conditions with an initial inoculum of 10⁸ bacteria/well in a 6-well plate, which is equivalent to MOI of ∼100:1) to 4–6% when the incubation time was shortened to 1 h with no change in the other infection parameters. However, imatinib did not inhibit bacterial attachment after 1 or 3 h of incubation in these experiments, consistent with prior reports on the apparently redundant role of ABL and other kinases in EPEC attachment (4, 40). We next varied the size of the bacterial inoculum. Decreasing the inoculum by ∼30-fold to 3 × 10⁶ CFU/well (MOI of ∼3:1) reduced bacterial attachment to 1–2% of the initial inoculum but, more importantly, revealed a significant inhibitory effect of imatinib (Fig. 2A). Even lower inocula were associated with inconsistent recovery of attached bacteria due to low numbers. As a further step in varying the infection conditions, we explored the impact of “bulk movement” of the inoculum suspension on attachment by placing the culture plates onto a rocker platform during incubation at 37°C. Shaking the plates further decreased bacterial attachment to 0.1–1% of the initial inoculum but also led to the greatest relative inhibition of attachment by imatinib (Fig. 2B). Moreover, these infection conditions revealed significant attachment inhibition by imatinib in another epithelial model, the human colonic epithelial cell line T84 grown as polarized monolayers on filter supports (Fig. 2C), underlining the general applicability of our findings. Collectively, these results demonstrate that dynamic infection conditions with limited bacterial attachment opportunities are critical for revealing the unique functions of ABL, ARG, and PDGFR in EPEC attachment to epithelia.

Fig. 2.

Impact of infection conditions on imatinib blockade of EPEC attachment. HeLa cells in 6-well plates (A and B) or polarized T84 monolayers on filter supports (C) were treated with imatinib (20 μM, gray bars) or DMSO as a control (black bars) for 30 min and infected for 1 h with the indicated inocula (A) or 3 × 10⁶ CFU (B and C) of EPEC per well (T84 cells were infected on the apical side only). Monolayers were washed, and bacterial attachment was assessed by CFU assay. Attachment is expressed as percentage of the initial inoculum. Values are means ± SE of 3 experiments. *P < 0.05 vs. respective control not treated with imatinib.

Confirmation of kinase involvement by genetic knockdown approaches.

To validate the pharmacological approaches, we next conducted RNA silencing experiments. Immunoblot analysis showed that addition of siRNAs targeting ABL, PDGFR, and p38 caused a >90% decrease in abundance of the respective proteins in HeLa cells (Fig. 3A), demonstrating that the knockdown strategies were highly effective in our system. Relative to cells transfected with control siRNA, knockdown of ABL and PDGFR significantly inhibited EPEC attachment to levels similar to those induced by the ABL/ARG/PDGF inhibitor imatimib and the PDGFR inhibitor JNJ10198409 (Fig. 3B). Attachment was also significantly inhibited by p38 knockdown, albeit less effectively than after addition of p38 inhibitors (Fig. 3B), suggesting that their inhibitory effects may have been partly due to other kinases. Combined knockdown of ABL and PDGFR, or additional knockdown of ARG, had no further effects on the inhibition of bacterial attachment. Together, these results indicate that at least two host cell kinases, ABL and PDGFR, have indispensable functions in the epithelial attachment of EPEC under dynamic conditions. PDGFR kinase was not known to be linked to EPEC attachment, while ABL was previously postulated to have a redundant role (40) but was shown in the present study to be critical for early attachment.

Fig. 3.

Effect of genetic knockdown of host cell kinases on epithelial EPEC attachment. HeLa cells were transfected for 48 h with siRNAs against the indicated kinases or the appropriate control siRNAs. A: protein extracts were prepared and analyzed by immunoblotting with antibodies against ABL, platelet-derived growth factor receptor (PDGFR), and p38, as well as β-actin as a loading control. B: attachment of EPEC to transfected cells was assessed under dynamic infection conditions as described in Fig. 1 legend. Attachment is expressed as percentage relative to cells transfected with control siRNA (dashed line). Indicated kinase inhibitors (with DMSO as a control) were used for comparison. Values are means ± SE (n = 5–6). *P < 0.05 vs. controls.

Imatinib inhibits rapid bacterial attachment but not later pedestal formation.

Since imatinib proved to be remarkably effective in decreasing bacterial attachment, we further explored the underlying mechanism of action in cell culture models. Inhibition of EPEC attachment required active kinase signaling in live epithelial cells, because prior fixation of the cells with 1% formalin abolished the effect (Fig. 4A). Fixation prevents specific cellular responses to infection, such as protein phosphorylation and actin recruitment (42), but does not generally interfere with attachment of EPEC or another cell-adhering E. coli, enteroaggregative E. coli, to epithelial cells (37) and may even nonspecifically promote attachment. Furthermore, imatinib had to be present throughout the infection, since preincubation of epithelial cells for 2 h followed by removal of the drug just before infection did not lead to inhibition of attachment. Finally, the effect of imatinib was not dependent on de novo protein synthesis, because inhibition of bacterial attachment occurred even in the presence of the protein synthesis blocker cycloheximide (Fig. 4B).

Fig. 4.

Characteristics of imatinib effect on EPEC attachment to epithelial cells. HeLa cells in 6-well plates were incubated with imatinib (20 μM, gray bars) or DMSO as a control (black bars) for 30 min before and throughout subsequent infection with EPEC (3 × 10⁶ CFU/well) for 1 h under rocking conditions, and bacterial attachment was determined after 1 h by CFU assay. A: selected cultures were fixed with formalin and washed before infection. B: some cells were treated with cycloheximide (CHX, 5 μg/ml) for 15 min before imatinib addition and infection. C: cultures were infected for 1 h without imatinib, nonattached bacteria were washed off, and cultures were continued for 1 h with or without added imatinib. Number of retained bacteria was determined by CFU assay. D: infections were done with wild-type (WT) EPEC and a Δtir mutant of EPEC, with and without imatinib. E: infections were done with WT EPEC or a ΔbfpA mutant. Values are means ± SE of ≥3 experiments. *P < 0.05 (by t-test) vs. respective controls (i.e., live HeLa cells, WT bacteria, solvent only); ns, not significant.

Our observation of inhibitory effects during the short (1 h) in vitro infection period suggests that imatinib influences early bacterial attachment processes, rather than later events such as pedestal formation (6). To evaluate this idea, we examined the effect of imatinib on detachment, i.e., the ability to induce detachment of already intimately attached EPEC. No difference between imatinib-treated and control cells could be detected, suggesting that imatinib did not affect A/E lesion formation or function (Fig. 4C). Consistent with this interpretation, imatinib inhibited early epithelial attachment of a Δtir mutant of EPEC, which lacks the translocated intimin receptor essential for pedestal formation (24), similar to wild-type bacteria (Fig. 4D).

To further explore the mechanism of early attachment inhibition by imatinib, we focused on the role of BFP, which are important for mediating initial bacterial attachment in the form of localized adherence (17). Wild-type EPEC and a ΔbfpA mutant were grown in DMEM to late log phase (to maximally induce BFP expression) and used to infect HeLa cell monolayers. EPEC ΔbfpA showed less overall attachment (Fig. 4E), consistent with prior observations (8). Importantly, imatinib did not inhibit attachment of the mutant to epithelial cells (Fig. 4E). Furthermore, treatment of wild-type EPEC with imatinib before incubation with epithelial cells without the inhibitor did not attenuate attachment, suggesting that bacterial BFP assembly is not impacted by imatinib. Together, these results show that kinase inhibition by imatinib interferes with an early bacterial attachment process that does not involve intimin-dependent pedestal formation and is no longer required after intimate attachment is established. Furthermore, host cell processes that mediate BFP-dependent attachment are apparently controlled by the imatinib targets ABL and PDGFR.

Imatinib attenuates in vivo infection with an A/E pathogen.

To establish the physiological relevance of our cell culture findings, we next tested the effects of the ABL/ARG/PDGFR inhibitor imatinib and the p38 inhibitor PD169316 in an in vivo infection model. Because EPEC infects mice only poorly, if at all, we employed the murine A/E pathogen C. rodentium. These bacteria, first identified as the causative agent of murine transmissible colon hyperplasia, resemble human EPEC in their ability to attach to intestinal epithelial cells and cause microvillus effacement (14). C. rodentium does not express BFP but has type IV pili (colonization factor Citrobacter) with strong similarity to BFP (30). Inoculation of mice with C. rodentium induces a robust infection that lasts 3–4 wk under normal conditions (28). We first tested whether imatinib and PD169316 attenuated epithelial attachment of C. rodentium under dynamic infection conditions in vitro. Attachment to CMT-93 murine rectal epithelial cells was significantly inhibited by both kinase inhibitors, although imatinib was more effective than PD169316 (Fig. 5A). These findings suggest that C. rodentium utilizes analogous attachment mechanisms to EPEC, even though C. rodentium lacks BFP. Subsequently, we administered imatinib and PD169316 daily to adult C57BL/6J mice and infected them orally with a standard inoculum of 5 × 10⁸ CFU of C. rodentium. Imatinib significantly reduced bacterial colonization throughout the course of infection, while PD169316 had only marginal effects (Fig. 5B). Imatinib also prevented infection-induced epithelial damage and hyperproliferation and mucosal inflammation (Fig. 5C), which typically occur 10–14 days after infection of wild-type mice (29). These results suggest that ABL and PDGFR (and perhaps ARG, although this kinase had no inhibitory function for EPEC attachment in vitro) have indispensable physiological functions, while p38 appeared to have no unique role in mediating mucosal attachment of a model A/E pathogen. The in vivo data also underline that our dynamic in vitro assays had correctly revealed a unique physiological attachment function of ABL and PDGFR, which was not apparent with other in vitro approaches (40).

Fig. 5.

Imatinib protects mice against C. rodentium infection at low inocula. A: CMT-93 mouse rectal epithelial cells, grown as adherent monolayers in 6-well plates, were incubated with imatinib (20 μM, light gray bar), PD169316 (100 nM, dark gray bar), or DMSO as a control (black bar) for 30 min and infected with C. rodentium (3 × 10⁶ CFU/well) under rocking conditions. Bacterial attachment was determined after 1 h by CFU assay. Values are means ± SE of 3 experiments. *P < 0.05 vs. DMSO control. B–E: adult C57BL/6 mice were treated with imatinib [50 mg/kg body wt po (B, left; C–E)], PD169316 [5 mg/kg ip (B, right)], or vehicle (PBS) once daily starting 1 day before (B–D) or 3 days (E, left) or 6 days (E, right) after oral infection with the indicated inocula of C. rodentium. Bacterial numbers in the stool were determined by CFU assay. Values are means ± SE of ≥6 mice/data point. *P < 0.05 (by rank-sum test) vs. vehicle-treated mice on the same day. Dashed horizontal lines represent assay sensitivity. C: paraffin sections of colon from uninfected mice or mice on day 14 after C. rodentium infection, treated or not with imatinib, were prepared and stained with hematoxylin and eosin. ∗, Inflammatory cell infiltrates and edema in mucosa and submucosa. Scale bar, 50 μm.

The in vitro findings indicate that imatinib is most effective in blocking bacterial attachment to epithelial cells under conditions that minimize attachment opportunities. On this basis, we predicted that imatinib would be most protective against low bacterial inocula, when the efficiency of attachment might be critical for establishing infection, while infection with high inocula should be less impacted, because higher inocula provide greater redundancy for bacterial attachment to the mucosa. To test this prediction, we evaluated imatinib efficacy in mice infected with different inocula. Imatinib did not attenuate infection initiated with a 10-fold-higher C. rodentium inoculum (5 × 10⁹ CFU) than our standard inoculum (Fig. 5C). In contrast, it completely blocked infection in mice given a lower inoculum of 10⁷ CFU (Fig. 5D). These results demonstrate that imatinib is most effective against A/E pathogen infection under conditions of limited bacterial exposure to the host. Consistent with this notion, a delayed start of imatinib treatment in mice already infected with C. rodentium for 3 days and, therefore, already partially colonized with bacteria could only delay infection for several days but not prevent it, while imatinib treatment started after 6 days, at which time colonization is fully established, had no impact on infection (Fig. 5E). Together, these data suggest that imatinib is most effective in the early stages of A/E pathogen infection with limited bacterial loads, while established stages with greater bacterial colonization are independent of the imatinib-targeted kinases ABL and PDGFR.

DISCUSSION

Activation of multiple host cell kinases upon EPEC attachment to cultured cells is well established (4, 40), but their relative physiological importance has remained uncertain. Our data show that ABL and PDGFR kinases and, to a lesser degree, p38 kinase have indispensable functions for EPEC attachment under dynamic infection conditions in vitro. Furthermore, inhibition of ABL/ARG and PDGFR with imatinib can protect against infection with C. rodentium in vivo, further supporting the conclusion that these kinases have indispensable physiological functions in the pathogenesis of the infection. The key difference between our studies and prior in vitro work was the use of dynamic infection conditions that seemed suboptimal and counterintuitive at first, as they markedly compromised bacterial attachment opportunities but turned out to be more predictive than prior infection approaches (40) for the in vivo situation. Limiting bacterial numbers and times for attachment to the epithelium in vitro apparently mimics the intestinal environment in ways that are evidently relevant for defining the role of specific host cell kinases in attachment of EPEC and C. rodentium. This concept may also be of interest for studies of other enteric pathogens, such as enterohemorrhagic E. coli, for which attachment is also inefficient but crucial for pathogenesis (7). In contrast, we would predict that dynamic infection conditions are less relevant for pathogens such as Yersinia enterocolitica (36) with very high attachment efficiencies. Another concept that arises from our studies is that even the modest (40–60%) differences in bacterial attachment we observed after inhibitor treatment in vitro can apparently have marked consequences for infection in vivo, presumably because the relationship between effective inoculum and infection severity is not linear but, rather, determined by thresholds and nonlinear effects.

EPEC uses several host cell kinases, including ABL, ARG, FYN, n-SRC, and YES, to phosphorylate bacterial Tir in an apparently redundant manner under conditions that favor bacterial attachment, whereas our data suggest that ABL and PDGFR have unique attachment functions under limiting conditions that cannot be compensated by other kinases. However, the specific mechanism by which these kinases exert their functions in early attachment is unclear. Our initial hypothesis that ABL is critical for Tir phosphorylation and, thus, subsequent intimate attachment is probably incorrect, since ABL inhibition attenuated attachment even in the absence of Tir. In contrast, ABL/PDGFR inhibition was no longer effective in BFP-deficient bacteria, suggesting that these kinases influence cellular processes required for attachment mediated by BFP in EPEC or by other attachment factors in C. rodentium (30). The major structural subunit of BFP is bundlin, which binds to an enterocyte receptor containing the N-acetyllactosamine glycan (20, 21). A role for phosphatidylethanolamine has also been described in binding of BFP (3). Thus, ABL/PDGFR signaling may be required for normal production of these or other critical glycans or phospholipids or for their normal display and function on the epithelial surface (2, 39). The kinases could also play a role in regulating cytoskeletal functions involved in bacterial attachment that are independent of Tir phosphorylation (19, 27). Furthermore, we cannot formally exclude the possibility that imatinib can directly compete with binding of BFP or other attachment factors to cells, although the loss of inhibition in fixed cells and reduced attachment upon ABL and PDGFR knockdown would argue against this mechanism.

Pharmacological strategies that target host processes, rather than bacterial functions, have the potential to fight bacterial infections that cannot be treated by classical antibiotics. In the case of EPEC, antibiotics are generally not effective, and infections are managed with oral or intravenous fluid substitution. In addition, clinical EPEC isolates from different regions of the world show varying levels of resistance to most classes of antibiotics, which further complicates effective use of antibiotics (38). Our data raise the possibility that inhibition of certain host cell kinases may have therapeutic potential for treating A/E pathogen infections. In evaluating this possibility, it must be noted that kinase inhibition was only effective in the early stages of infection and at low inocula in our models. Most clinical situations probably require treatment of already fully established infections, because manifestation of clinical symptoms and the ensuing laboratory diagnostics are likely to delay treatment decisions. Under those conditions, kinase inhibition may be less effective, at least for the kinases identified in this study. Furthermore, we observed that genetically engineered mice double-deficient for ABL and ARG in the intestinal epithelium are highly susceptible to infection and tissue damage. Therefore, as conceptually promising as kinase inhibition may be as an alternative strategy for managing A/E infections, careful balancing of beneficial and detrimental effects is necessary to determine whether such a strategy can ultimately be employed in the clinic.

GRANTS

This work was supported by National Institutes of Health Grants DK-035108 and DK-080506 (L. Eckmann) and CA-043054 (J. Y. J. Wang) and by Canadian Institutes of Health Grants 115180 and 126051 (B. A. Vallance). C. F. Manthey was supported by a fellowship from the German Research Foundation (MA 4980/1-1). B. A. Vallance is the Children With Intestinal and Liver Disorders Foundation Chair in Pediatric Gastroenterology and the Canada Research Chair in Pediatric Gastroenterology.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.F.M., B.A.V., J.Y.W., and L.E. are responsible for conception and design of the research; C.F.M., C.B.C., A.W., and E.H. performed the experiments; C.F.M., C.B.C., E.H., A.G., M.G.M., and L.E. analyzed the data; C.F.M., A.G., M.G.M., J.Y.W., and L.E. interpreted the results of the experiments; C.F.M. and L.E. prepared the figures; C.F.M. and L.E. drafted the manuscript; C.F.M., E.H., B.A.V., A.G., and L.E. approved the final version of the manuscript; B.A.V., A.G., M.G.M., J.Y.W., and L.E. edited and revised the manuscript.