Adherent-Invasive Escherichia coli Production of Cellulose Influences Iron-Induced Bacterial Aggregation, Phagocytosis, and Induction of Colitis

Abstract

Adherent-invasive Escherichia coli (AIEC), a functionally distinct subset of resident intestinal E. coli associated with Crohn's disease, is characterized by enhanced epithelial adhesion and invasion, survival within macrophages, and biofilm formation. Environmental factors, such as iron, modulate E. coli production of extracellular structures, which in turn influence the formation of multicellular communities, such as biofilms, and bacterial interactions with host cells. However, the physiological and functional responses of AIEC to variable iron availability have not been thoroughly investigated. We therefore characterized the impact of iron on the physiology of AIEC strain NC101 and subsequent interactions with macrophages. Iron promoted the cellulose-dependent aggregation of NC101. Bacterial cells recovered from the aggregates were more susceptible to phagocytosis than planktonic cells, which corresponded with the decreased macrophage production of the proinflammatory cytokine interleukin-12 (IL-12) p40. Prevention of aggregate formation through the disruption of cellulose production reduced the phagocytosis of iron-exposed NC101. In contrast, under iron-limiting conditions, where NC101 aggregation is not induced, the disruption of cellulose production enhanced NC101 phagocytosis and decreased macrophage secretion of IL-12 p40. Finally, abrogation of cellulose production reduced NC101 induction of colitis when NC101 was monoassociated in inflammation-prone Il10^−/− mice. Taken together, our results introduce cellulose as a novel physiological factor that impacts host-microbe-environment interactions and alters the proinflammatory potential of AIEC.

INTRODUCTION

Inflammatory bowel diseases (IBD), including Crohn's disease (CD) and ulcerative colitis (UC), comprise a heterogeneous collection of chronic, relapsing immune-mediated disorders. Although the precise etiologies are incompletely understood, accumulating evidence suggests that IBD are the result of inappropriate immune responses toward a subset of resident intestinal microbes and their products in genetically susceptible individuals (1, 2).

The gastrointestinal (GI) tract is home to a complex community of microbes referred to as the intestinal microbiota. The development of intestinal inflammation is associated with community-wide changes to the intestinal microbiota, including an expansion in the relative abundance of endogenous Escherichia coli in IBD patients (3) and in rodent models of experimental colitis (4–6). A functionally distinct group of resident enteric E. coli bacteria known as adherent-invasive E. coli (AIEC) are recovered more frequently and in larger quantities from ileal tissue biopsy specimens from CD patients than from specimens from non-CD controls (7, 8). In the absence of common identifying genetic determinants (9), AIEC strains are characterized by their ability to adhere to and invade intestinal epithelial cells (10) and to survive and replicate within macrophages (11). AIEC strains are also moderate to strong in vitro biofilm producers (12). In addition, AIEC strains are capable of inducing and perpetuating intestinal inflammation in various rodent models of experimental colitis, including streptomycin-treated mice (13), dextran sodium sulfate (DSS)-treated mice (14), Toll-like receptor 5-deficient mice (15), transgenic CEABAC10 mice (16), and gnotobiotic interleukin-10 (IL-10)-deficient (Il10^−/−) mice (17). These functional attributes of AIEC, in conjunction with host factors, such as genetic polymorphisms linked to aberrant microbial sensing and clearance (18, 19), potentially enable enhanced mucosal association by AIEC strains (20, 21). Together, these characteristics provide the physical opportunity for AIEC strains to continuously stimulate the mucosal immune system, thus propagating a state of chronic intestinal inflammation.

Macrophages are a key component of host innate immune defense in the intestines, limiting systemic microbial dissemination by destroying potential invaders through phagocytosis, while also sensing and responding to microbial stimuli and informing consequent host immune responses (22). Through pattern recognition receptors (PRRs), macrophages recognize the conserved microbial molecular patterns synthesized by resident and pathogenic intestinal bacteria, including extracellular structures, such as fimbriae, flagella, lipopolysaccharides, and peptidoglycan. Environmental factors, such as iron availability, influence the microbial production of some of these extracellular structures, including curli fibrils in Salmonella enterica serovar Typhimurium (23, 24) and type I fimbriae in E. coli (25), providing the opportunity for environmental modulation of microbial interactions with macrophages. Indeed, iron impacts E. coli interactions with host cells, albeit in contrasting ways. Iron promotes the increased internalization of pathogenic E. coli by neutrophils (26) and intestinal epithelial cells (27, 28). In contrast, iron limitation promotes the phagocytosis of a nonpathogenic E. coli K-12 strain by macrophages through the decreased expression of the outer membrane protein OmpW (29). Extracellular microbial structures that impact interactions with macrophages are also produced within multicellular microbial communities, including biofilms and bacterial aggregates. Curli fibrils and the exopolysaccharide cellulose are common matrix components present within multicellular structures produced by S. Typhimurium and E. coli (30–32). Cellulose and/or curli production has also been implicated in modulating intestinal E. coli interactions with epithelial cells (33, 34) and influencing in vivo host immune responses to uropathogenic E. coli (UPEC) strains in the urinary tract (35).

Iron is a cofactor necessary for various microbial enzymes and therefore serves as an important micronutrient for most bacteria. In E. coli, changes in cytosolic iron concentrations are directly sensed by Fur (36). When bound to Fe²⁺, Fur acts as a transcription factor, regulating genes involved in diverse cellular processes, such as metabolism, metal acquisition, stress responses, motility, and biofilm formation (37, 38). Changes in extracellular iron concentrations are also sensed by the membrane-associated kinase BasS, a member of the BasRS two-component system (39). In response to Fe³⁺, BasR regulates genes involved in altering the outer membrane landscape of E. coli (40, 41). Given the importance of iron to microbial growth and function, an integral component of the innate immune response is the secretion of iron-scavenging proteins at mucosal surfaces to limit microbial iron availability, a response that is potentiated by inflammation (42, 43).

Studies investigating the impact of iron on E. coli physiology and interactions with host cells have been limited to nonpathogenic K-12 or pathogenic E. coli strains. Consequently, little is known about the impact of iron on the functional attributes of AIEC strains. Therefore, the goal of this study was to characterize how iron impacts the physiology of the AIEC strain NC101 (9, 44) and subsequent interactions with macrophages. Here we show that iron promotes the cellulose-dependent aggregation of NC101. Bacterial cells recovered from the aggregates are more susceptible to phagocytosis, as prevention of aggregation through the disruption of cellulose production reduces macrophage uptake of NC101. Conversely, under iron-limiting conditions where aggregation is not induced, disruption of cellulose production enhanced NC101 phagocytosis and decreased macrophage proinflammatory responses. Abrogation of bacterial cellulose production also delayed the onset of colitis in inflammation-prone Il10^−/− mice monoassociated with NC101. Taken together, our results introduce cellulose as a novel physiological factor that dynamically impacts AIEC-host interactions in the face of changing environmental conditions.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The bacterial strains and plasmids used in this study are listed in Table S1 in the supplemental material. E. coli NC101 was isolated from the feces of a wild-type (WT) mouse as previously described (17). Unless otherwise indicated, bacteria from an overnight culture were washed prior to inoculation into M9 minimal medium with the concentrations of iron as ferrous sulfate (catalog number I146; Fisher Scientific) indicated below. Bacteria were grown at 250 rpm and 37°C for all experiments. The medium was supplemented with 50 μg/ml kanamycin or 100 μg/ml carbenicillin as appropriate.

Construction of isogenic mutant, chromosomally complemented, and GFP-labeled strains.

All deletion mutants were created using the bacteriophage λ Red recombinase system as previously described (45). Deletion mutants were chromosomally complemented using the pMCL2868 plasmid (a kind gift from M. Chelsea Lane), a mini-Tn7 vector, as previously described (46). For green fluorescent protein (GFP)-labeled strains, the pEGFP plasmid was transformed into each strain by electroporation. Transformed strains were grown with 100 μg/ml carbenicillin to maintain the plasmid.

Sedimentation assays.

Sedimentation assays were performed as previously described (47) with the following modifications. Briefly, bacteria were grown in M9 minimal medium with the concentrations of iron or the iron chelator 2,2-bipyridyl (BPD; catalog number 366-18-7; Alfa Aesar) indicated below. At the specified time points, cultures were removed from the incubator and the aggregates were allowed to settle to the bottom of 50-ml conical tubes. An aliquot from the top of the culture (planktonic phase) was taken for optical density (OD) quantification by spectrophotometry or quantitative culture. The cultures were then vortexed thoroughly, and the cells were pelleted, washed with phosphate-buffered saline (PBS), and passed through a 30-gauge needle to obtain homogeneous bacterial suspensions. An aliquot was then taken for quantitative culture or spectrophotometry as a measure of the bacterial concentration of the whole culture. Percent aggregation was calculated using the formula [(whole-culture OD₆₀₀ − planktonic culture OD₆₀₀)/whole-culture OD₆₀₀] × 100, where OD₆₀₀ is the optical density at 600 nm. Where indicated, cellulase (from Aspergillus niger [catalog number C1184; Sigma]) was added to the cultures after 2 h of growth. The cultures were then vortexed vigorously and incubated for 10 min at 37°C.

Assessment of aggregation and CF binding by fluorescence microscopy.

GFP-labeled bacteria were grown in M9 minimal medium with the concentrations of iron indicated below. For assessing the dispersion of the bacterial aggregates prior to the gentamicin protection assays, bacterial aggregates were physically disrupted as described above. When staining with calcofluor (CF), the bacteria were cultured with 1% calcofluor white staining solution (catalog number 18909; Sigma). The cultures were normalized to equivalent optical densities (600 nm). A 5-μl aliquot was spotted onto glass microscope slides in duplicate. The slides were visualized using an Olympus IX71 fluorescence microscope. ImageJ software was utilized to quantify the aggregates, where aggregates were defined as objects in the field that had a pixel intensity threshold of 5 to 215 and a pixel size of 201 to infinity. ImageJ software was also utilized to quantify CF binding per cell using the following formula: (mean intensity of CF × area of CF)/(mean intensity of GFP × area of GFP). In each experiment, at least 15 high-power fields (magnification, ×200) per slide were analyzed. Representative images were taken to demonstrate the colocalization of CF with GFP-expressing bacteria at a ×400 magnification.

Gentamicin protection assays.

Bacteria were grown for 2 h in M9 minimal medium with the concentrations of iron indicated below. The cultures were vortexed and washed with PBS to disrupt the aggregates as described above and normalized to equivalent optical densities (600 nm). For experiments using planktonic or aggregate bacterial cells, the aggregates were allowed to settle to the bottom of the 50-ml conical tubes for 5 min. The broth phase (top) of the culture was then collected, and cells recovered from this phase were defined as planktonic. The remaining broth phase was removed by aspiration. Aggregates were recovered by resuspension in PBS and were physically disrupted as described above. Gentamicin protection assays were then performed as previously described (7). Briefly, bone marrow-derived macrophages were seeded in 24-well plates, and bacteria were added at a multiplicity of infection (MOI) of 10. After centrifugation for 10 min at 228 × g, the cocultures were incubated for 30 min in RPMI 1640 medium. Gentamicin-laden medium (RPMI 1640 with 10% fetal bovine serum and 100 μg/ml gentamicin) was then added to eliminate any remaining extracellular bacteria. The macrophages were lysed with PBS containing 1% Triton X-100 at the time points indicated below for enumeration of intracellular bacteria. Intracellular bacteria were quantified by serial dilution plating.

Mice.

Germfree (GF) Il10^−/− and WT mice of the 129S6/SvEV background were originally derived under sterile conditions by hysterectomy at the Gnotobiotic Laboratory (University of Wisconsin, Madison). Mice were maintained under GF conditions at the National Gnotobiotic Rodent Resource Center at the University of North Carolina (UNC), Chapel Hill. An overnight bacterial culture of E. coli NC101 or the bcsA mutant was utilized to monoassociate mice via oral and rectal swabs as previously described (48). The absence of isolator contamination was confirmed by Gram staining and fecal culture. Once the mice were monoassociated, fecal and cecal E. coli loads were quantified by dilution plating on LB plates as previously described (44). All animal protocols were approved by the UNC, Chapel Hill, Institutional Animal Care and Use Committee.

Histological scoring.

At necropsy, proximal and distal colonic segments were Swiss rolled and fixed in 10% neutral buffered formalin. The histological inflammation scores (range, 0 to 4) of the proximal and distal colonic sections were blindly assessed as previously described (17). Briefly, the scoring system is as follows: 0 indicates no inflammation, 1 indicates the presence of cells infiltrating the lamina propria (LP); 2 indicates epithelial hyperplasia, a mild loss of goblet cells, and more extensive cellular infiltration within the LP; 3 indicates marked epithelial hyperplasia, a loss of goblet cells, and pronounced cellular infiltration within the LP and submucosa; and 4 indicates ulceration and transmural inflammation. Data are expressed as the composite histology score (range, 0 to 8), which was calculated by adding the proximal and distal colonic histology scores.

Statistical analysis.

P values were calculated using Student's t test when two experimental groups were compared, one-way analysis of variance (ANOVA) with Tukey's multiple-comparison posttest when three or more experimental groups were compared, or two-way ANOVA with the Bonferroni multiple-comparison posttest when more than two variables were compared. All data from the enumeration of bacteria by serial dilution and plating were log transformed to normalize the data. For quantification of NC101 aggregates by microscopy, P values were calculated using a nonparametric Kruskal-Wallis test with Dunn's posttest. For all animal experiments, P values were determined using a nonparametric Mann-Whitney test.

Detailed methods for the Congo red and calcofluor colony morphotype assays, macrophage isolation, mesenteric lymph node (MLN) cultures, enzyme-linked immunosorbent assays, RNA isolation, and reverse transcriptase PCR are described in the Materials and Methods in the supplemental material.

RESULTS

Iron promotes aggregation of E. coli NC101.

Bacterial iron availability varies within the GI tract, likely decreasing toward the mucosal interface with the secretion of host iron-binding proteins, a phenomenon that is potentiated with inflammation (49). However, the physiological responses of AIEC to changes in iron availability have not been well characterized. We therefore investigated the impact of iron on the physiology of E. coli NC101, a murine intestinal isolate that exhibits various AIEC characteristics, including increased epithelial translocation, enhanced persistence within macrophages, and the ability to induce colitis in selectively colonized, inflammation-prone Il10^−/− mice (17, 50). When cultured in minimal medium with 5, 10, or 50 μM iron at 37°C, NC101 formed macroscopic aggregates that sediment in culture (Fig. 1A). The proportion of bacterial cells associated with the aggregates significantly increased as the iron concentration increased, as assessed by quantitative plating (see Fig. S1 in the supplemental material) and determination of the optical density (Fig. 1B). Addition of an iron chelator did not impact NC101 aggregation (Fig. 1C), suggesting that further iron deprivation had no effect on this phenotype. In contrast, the E. coli K-12 substrain MG1655 does not form aggregates in response to iron (see Fig. S2 in the supplemental material), indicating that this phenomenon is not universal to all E. coli strains.

FIG 1.

Iron promotes aggregation of E. coli NC101. (A) Representative images of NC101 aggregates after 2 h of growth in minimal medium with increasing concentrations of iron. (B and C) NC101 sedimentation assays after 2 h of growth in minimal medium with various concentrations of iron (B) or the iron chelator BPD (C). Data represent the percent optical density associated with the aggregates relative to that for the whole culture. All data represent the mean ± SEM from at least three independent experiments. P values were determined by pairwise comparisons by one-way ANOVA. *, P < 0.05 compared to the condition with 0 μM iron; **, P < 0.01 compared to the condition with 0 μM iron; ***, P < 0.001 compared to the condition with 0 μM iron.

Cellulose is required for iron-induced aggregation of NC101.

The extracellular matrix (ECM) of multicellular structures produced by other E. coli strains is composed of proteinaceous components, such as fimbriae and curli; exopolysaccharides, such as cellulose; and/or extracellular DNA (51–53). The combination of matrix components varies depending on the environmental conditions and the E. coli strain (54). Therefore, to determine the extracellular composition of the NC101 aggregates, cellulase, DNase, or proteinase was added to the cultures following 2 h of growth. Addition of cellulase resulted in dispersal of the aggregates (see Fig. S3 in the supplemental material), while DNase and proteinase had no obvious impact (data not shown). This suggests that cellulose is a major extracellular component that contributes to iron-induced aggregation.

Cellulose biosynthesis and structural proteins are encoded by bcsQABZC and bcsEFG and are required for cellulose production (31, 32, 55). However, the presence of bcs genes does not necessarily correspond with an ability to produce cellulose. We first assessed whether NC101 is capable of producing cellulose at 37°C utilizing the well-established calcofluor (CF) binding and red, dry, and rough (RDAR) colony morphotype assays. In contrast to E. coli MG1655, which served as an established negative control (56), NC101 bound CF, indicating the presence of cellulose production (Fig. 2A). Similarly, on Congo red agar, NC101 produced colonies that had the RDAR colony morphotype (Fig. 2B), a colony morphotype that is dependent on cellulose and curli production (23, 31). To confirm that NC101 is a cellulose producer, an isogenic mutant lacking the bcsA gene, which encodes the catalytic subunit of cellulose synthase (57), was created. Deletion of bcsA resulted in the loss of CF binding and the production of smooth colonies on Congo red agar (Fig. 2A and B). Moreover, the bcsA mutant did not form macroscopic aggregates (Fig. 2C and D), confirming that cellulose production is required for iron-induced aggregation of NC101.

FIG 2.

Cellulose is required for iron-induced aggregation of NC101. (A and B) Representative colony morphologies of NC101 or the ΔbcsA, ΔbcsA+bcsA, or csgA::kan mutant on calcofluor (A) or Congo red agar (B). E. coli MG1655 served as a negative control. (C) Representative images of NC101 and the ΔbcsA mutant after 2 h of growth in minimal medium with 10 μM iron. (D) Sedimentation assays of NC101 and the ΔbcsA, ΔbcsA+bcsA, or csgA::kan mutant after 2 h of growth in minimal medium with increasing concentrations of iron. Data represent the percent OD₆₀₀ of the aggregates relative to that for the whole culture. Data represent the mean ± SEM from three independent experiments. P values were determined by pairwise comparisons by one-way ANOVA. *, P < 0.05; ***, P < 0.001.

Curli are commonly coexpressed with cellulose within RDAR colonies and biofilms (31, 51, 58). To determine whether curli also contribute to RDAR colony formation in NC101, the csgA gene encoding the major subunit of the curli fibrils was disrupted (59). The csgA mutant produced pink instead of brown-red textured colonies (Fig. 2B), a phenotype that has been observed in other csgA-deficient intestinal E. coli strains (58). We also determined whether curli contribute to NC101 aggregation in response to iron. No differences in aggregation were observed between the cgsA mutant and NC101, suggesting that curli expression does not significantly contribute to iron-induced aggregate formation (Fig. 2D).

Deletion of fur decreases NC101 aggregation.

Given the impact of iron on NC101, we next sought to determine how iron-induced aggregation occurs. In E. coli, fluctuations in iron availability can be sensed intracellularly by the transcription factor Fur (36) or extracellularly by the two-component system BasRS (60). To determine whether NC101 aggregation occurs in response to intracellular or extracellular iron, we tested the aggregative abilities of fur and basRS deletion mutants. In contrast to the findings for the parental strain, the fur mutant did not form visible aggregates after 2 h of growth (Fig. 3A and B). This was not the result of a corresponding growth defect with the fur mutant at the same time point (Fig. 3D). The aggregates formed by the fur mutant after 8 h were macroscopically smaller than the aggregates formed by NC101 (Fig. 3A), with a reduced proportion of bacterial cells relative to the whole culture associated with the aggregates (Fig. 3C). In contrast, deletion of basRS did not impact NC101 aggregate formation (Fig. 3B). Taken together, deletion of fur limited NC101 aggregation, suggesting that factors promoting iron-induced aggregate formation may be under the control of the Fur modulon.

FIG 3.

Deletion of fur in NC101 limits iron-induced aggregation. (A) Representative images of NC101 or the Δfur mutant after 2 or 8 h of growth in minimal medium with 10 μM iron. (B) Sedimentation assays of NC101 and the Δfur, Δfur+fur, or basRS::kan mutant after 2 h of growth in minimal medium with 0 or 10 μM iron. (C and D) Time course sedimentation assays (C) and growth curves (D) of NC101 or the Δfur or Δfur+fur mutant in minimal medium with 10 μM iron. Data for all sedimentation assays represent the percent OD₆₀₀ of the aggregates relative to that for the whole culture. All data are represented as the mean ± SEM from three independent experiments. P values were determined by pairwise comparisons by one-way ANOVA (B) and two-way ANOVA (C and D). *, P < 0.05; ***, P < 0.001.

Decreased aggregation by the fur mutant is not the result of an inability to produce cellulose.

Given that iron-induced aggregation of NC101 requires the capacity to produce cellulose, we determined whether cellulose biosynthesis was disrupted in the fur mutant as an explanation for its reduced ability to aggregate. We first tested whether cellulose production was unperturbed in the fur mutant. Deletion of fur did not eliminate CF binding, although the distribution of CF within the inoculum was altered compared to that in NC101 (Fig. 4A). As growth conditions impact the stimulation of E. coli cellulose production, we next determined whether NC101 and the fur mutant bind CF when exposed to iron in minimal medium. NC101 microscopic aggregates colocalized with CF (Fig. 4C), confirming that cellulose is a component of the ECM. Although binding was not uniformly observed, single NC101 cells not associated with the aggregates bound CF, suggesting that some planktonic cells produce cellulose. The fur mutant also colocalized with CF in the presence of iron, further demonstrating that cellulose production was not abrogated in the mutant strain. Formation of microscopic aggregates by the fur mutant was also observed. However, consistent with its decreased ability to form macroscopic aggregates, the microscopic aggregates produced by the fur mutant were less frequent and smaller in size than those produced by NC101 (see Fig. S4 in the supplemental material). Finally, to determine whether cellulose production is decreased in the fur mutant, the extent of CF binding was assessed as a proxy for cellulose production. CF binding in the presence of iron did not differ between NC101 and the fur mutant (Fig. 4B). CF binding was also evident for NC101 and the fur mutant when they were grown in minimal medium without iron (Fig. 4B; see also Fig. S5 in the supplemental material), indicating that cellulose production occurs under iron-limiting conditions. Taken together, these data demonstrate that decreased aggregate production by the fur mutant is not due to an inability to produce cellulose, suggesting the involvement of other factors, in addition to cellulose, that promote maximal NC101 aggregation.

FIG 4.

Deletion of fur does not disrupt NC101 cellulose production. (A) Colony morphologies of NC101 or the Δfur or ΔbcsA mutant on calcofluor plates. (B) GFP-expressing NC101 or the Δfur or ΔbcsA mutant were grown in minimal medium with 0 or 10 μM iron for 1 h and stained with calcofluor. ImageJ software was utilized to calculate the mean calcofluor binding per cell per high-power field (magnification, ×200). At least 15 fields were analyzed per sample. Data represent the mean ± SEM from three independent experiments relative to the condition with NC101 and 0 μM iron. P values were determined by pairwise comparisons by the Kruskal-Wallis test. (C) Representative images of GFP-expressing NC101 or the Δfur or ΔbcsA mutant stained with calcofluor. Bacteria were grown in minimal medium with 10 μM iron for 1 h. (Insets) Magnified views of the boxed areas on the right. Magnification = ×400 (or ×2,000 for insets); bars = 100 μm.

NC101 aggregate cells are more susceptible to phagocytosis by macrophages.

ECM components produced within microbial multicellular structures potentially alter E. coli interactions with macrophages. Given that AIEC is characterized in part by its distinct interactions with macrophages (11), we investigated whether the physiology associated with an aggregative state alters NC101 susceptibility to phagocytosis by macrophages and subsequent intracellular survival. To test this, aggregation of NC101 was induced by growth with iron, and cultures containing aggregates were physically dispersed into single-cell suspensions and cocultured with macrophages. NC101 was exposed to iron prior to coculture with macrophages, as iron availability can also impact macrophage function (61). Under aggregate-inducing conditions, NC101 phagocytosis was significantly enhanced (Fig. 5A). Although the quantity of intracellular NC101 was also increased after 4 and 8 h, the percent intracellular survival of NC101 was not substantially altered (Fig. 5B). These data demonstrate an association between iron-induced aggregation and increased bacterial uptake by macrophages.

FIG 5.

NC101 aggregates are more susceptible to phagocytosis. NC101 was grown in minimal medium with 0 or 10 μM iron prior to coculture with bone marrow-derived macrophages. (A) Following the addition of gentamicin for 30 min, intracellular NC101 was quantified after 1 h as a measure of bacterial uptake and 4 or 8 h as a measure of intracellular survival. (B) Ratios of the amount of intracellular NC101 at 4 h to that at 1 h and of the amount of intracellular NC101 at 8 h to that at 1 h as measures of percent intracellular survival. (C) The planktonic and aggregate fractions of each culture were separated prior to coculture with macrophages. Intracellular NC101 was quantified after 1 h. Data are shown as the mean ± SEM from a representative experiment of at least three independent experiments with at least four technical replicates. P values were determined by pairwise comparisons by t test (A and B) and one-way ANOVA (C). ***, P < 0.001.

To further explore this association, we physically separated the planktonic and aggregate phases from the same culture and tested whether NC101 cells recovered from the aggregates were more susceptible to phagocytosis. Irrespective of the presence of iron, the extent of internalization of planktonic NC101 remained constant (Fig. 5C). In contrast, a significantly larger amount of NC101 aggregate cells than planktonic cells was phagocytosed by macrophages, suggesting that NC101 aggregate cells are more susceptible to phagocytosis. One explanation for the increased phagocytosis of aggregate cells is the incomplete disruption of the aggregates, which could result in more cells entering the macrophage at once. To address this possibility, we investigated whether aggregates of GFP-labeled NC101 were fully dispersed utilizing microscopy. Before physical disruption, the number of aggregates per high-power field was significantly higher when NC101 was cultured with iron (see Table S3 in the supplemental material). After physical disruption, NC101 aggregates were rarely visible microscopically, and importantly, there was no significant difference in the number of NC101 aggregates per field with or without iron. These findings indicate that the increased susceptibility of aggregate cells to phagocytosis is not likely due to more bacteria entering the macrophage at once. Instead, these data suggest that the physiology of the individual cells in the aggregates promotes their phagocytosis.

Cellulose modulates NC101 susceptibility to phagocytosis.

We next sought to identify the physiological factors that mediate enhanced internalization of individual NC101 aggregate cells. Given that cellulose is required for aggregation and deletion of fur limits aggregate formation in a cellulose-independent manner, we predicted that deletion of bcsA or fur would reduce macrophage uptake of iron-exposed NC101. As hypothesized, under aggregate-inducing conditions, macrophage uptake of the bcsA and fur mutants was significantly decreased in comparison to that of NC101 (Fig. 6B; see also Fig. S6 in the supplemental material). However, deletion of bcsA or fur did not reduce the amount of NC101 internalization to levels comparable to those for NC101 nonaggregate cells, suggesting the involvement of other bacterial factors that contribute to the susceptibility of aggregate cells to phagocytosis. Conversely, under iron-limiting conditions where NC101 aggregation was not induced, the bcsA mutant was more susceptible to phagocytosis than NC101 (Fig. 6A; see also Fig. S6 in the supplemental material). This suggests that cellulose may act as an antiphagocytic factor for NC101 nonaggregate cells. Similar results were also observed with the non-cellulose-producing strain MG1655 (see Fig. S6 in the supplemental material). Importantly, the bcsA mutant and MG1655 did not form microscopic aggregates (see Table S3 in the supplemental material). Moreover, deletion of bcsA did not impact the percent survival within macrophages (see Fig. S7 in the supplemental material). Taken together, cellulose disparately modulates NC101 susceptibility to macrophage phagocytosis by enabling the formation of an aggregative physiological state under iron-replete conditions and by potentially acting as an antiphagocytic factor under iron-limiting conditions.

FIG 6.

Deletion of bcsA alters NC101 interactions with macrophages. (A and B) Intracellular NC101 (NC) or the ΔbcsA or Δfur mutant after 1 h of coculture with bone marrow-derived macrophages. (C and D) IL-12 p40 cytokine production by bone marrow-derived macrophages infected with NC101 or the ΔbcsA or Δfur mutant for 8 h. All bacteria were grown in minimal medium with 0 μM (A and C) or 10 μM (B and D) iron prior to coculture with macrophages. Data are shown as the mean ± SEM from a representative experiment of three independent experiments with at least four technical replicates. P values were determined by pairwise comparisons by one-way ANOVA. *, P < 0.05; ***, P < 0.001.

Cellulose alters the proinflammatory potential of NC101.

It is unclear how the differential phagocytosis of AIEC impacts macrophage proinflammatory responses. As deletion of bcsA alters macrophage uptake of NC101, we investigated whether uptake of the bcsA mutant also alters macrophage production of p40, the common subunit of the proinflammatory cytokines IL-12 and IL-23. Under nonaggregating conditions (i.e., low-iron conditions), production of IL-12 p40 was decreased in macrophages that phagocytosed the bcsA mutant (Fig. 6C). Conversely, with increased amounts of iron at levels where NC101 aggregation is induced, macrophages that phagocytosed the bcsA mutant secreted larger amounts of IL-12 p40 than the NC101-exposed macrophages (Fig. 6D). Deletion of fur did not alter IL-12 p40 secretion by the mutant-exposed macrophages, further demonstrating the importance of cellulose in mediating NC101-macrophage interactions. Taken together, these results demonstrate that, in addition to impacting NC101 phagocytosis susceptibility, cellulose modulates macrophage production of the proinflammatory cytokine IL-12 p40.

Given the contrasting effects of cellulose on macrophage proinflammatory responses, we next sought to determine how the ability to produce cellulose contributes to the in vivo proinflammatory potential of NC101. To investigate this, GF Il10^−/− mice were monoassociated with NC101 or the bcsA mutant. AIEC strains such as NC101 uniquely induce colitis when monoassociated in inflammation-susceptible Il10^−/− mice (17), whereas the monoassociation of non-AIEC strains, including MG1655 (62) and Nissle (63), does not induce chronic colitis. The severity of inflammation was assessed by histological score and proinflammatory cytokine expression. Monoassociated WT mice served as inflammation-resistant controls. After 10 days of colonization, no differences in histological inflammation were observed (Fig. 7B). After 21 days, mice colonized with the parental strain exhibited significantly worse proximal and distal colonic inflammation, characterized by increased crypt hyperplasia and leukocytic infiltration into the lamina propria (Fig. 7A and B). By 35 days, pathohistological differences were no longer apparent because the severity of inflammation increased in mice colonized with the bcsA mutant. Importantly, WT animals colonized with either strain did not exhibit any pathology (Fig. 7C).

FIG 7.

Deletion of bcsA in NC101 delays the onset of colitis in Il10^−/− mice. (A) Representative histology by staining with hematoxylin and eosin of the proximal colons of Il10^−/− or WT mice monoassociated with NC101 or the ΔbcsA mutant for 21 days. Magnification = ×200; bars = 50 μm. (B and C) Composite proximal and distal colon histology scores (range, 0 to 8) of Il10^−/− mice (B) or WT mice (C) monoassociated with NC101 or the ΔbcsA mutant. (D and E) Proximal colon transcript levels of Il12b (D) or Il17a (E) relative to that of Actb in Il10^−/− mice monoassociated with NC101 or the ΔbcsA mutant. Data are expressed as the fold change relative to that for NC101-colonized mice. (F) IL-17 production by unfractionated MLN cells restimulated with the respective bacterial lysates ex vivo. MLNs were isolated from Il10^−/− mice monoassociated with NC101 or the ΔbcsA mutant for 21 or 35 days. Each symbol represents an individual mouse, and the data are for 5 to 8 mice per group. Open circles, mice monoassociated with NC101; closed squares, mice monoassociated with the ΔbcsA mutant; lines at medians, P values determined by pairwise comparisons by a Mann-Whitney test. *, P < 0.05; **, P < 0.01.

We next determined whether differences in histological inflammation at 21 days corresponded with the differential expression of proinflammatory cytokines. Il10^−/− mice monoassociated with NC101 develop colitis that is driven by T-helper 1 (Th-1) and Th-17 responses, where the onset and exacerbation of inflammation are associated with the increased production of IL-17, gamma interferon (IFN-γ), and IL-12 (17, 48). We therefore determined whether the earlier onset of colitis in NC101-colonized mice corresponded with the differential expression of Il17a, Ifng, and Il12b. Prior to the onset of histological inflammation, the expression of Il12b encoding the p40 subunit was increased in mice colonized with NC101 (Fig. 7D). This was consistent with the differences in macrophage production of IL-12 p40 observed in vitro in response to NC101 or the bcsA mutant under iron-limiting conditions. The colonic expression of Il17a or Ifng did not differ at 10 days (data not shown). At 21 days, coinciding with a more severe histopathology, Il17a expression was increased in the proximal colon in mice colonized with NC101 (Fig. 7E). In contrast, no significant differences in Ifng transcript levels were observed (see Fig. S8 in the supplemental material). Because colitis is driven by antigen-specific responses in Il10^−/− mice monoassociated with NC101, we also quantified IFN-γ and IL-17 production by unfractionated MLN cells restimulated with the respective bacterial lysates. MLN cells recovered from mice colonized with NC101 produced larger quantities of IL-17 than MLN cells isolated from mice colonized with the bcsA mutant (Fig. 7F). The levels of IFN-γ production by restimulated MLN cells from mice colonized with NC101 or the ΔbcsA mutant were not significantly different (see Fig. S8 in the supplemental material). Finally, differences in the severity of colitis observed at 21 days also corresponded with a 2.4-fold decrease in the fecal loads of the bcsA mutant (Fig. 8). The cecal luminal densities of the bcsA mutant were also consistently decreased relative to that of NC101, although this was not uniformly observed in the feces. Decreased fecal and cecal concentrations of the bcsA mutant relative to the parental strain were likewise observed in WT mice. Taken together, Il10^−/− mice monoassociated with a cellulose-deficient NC101 mutant exhibited a delayed onset of colitis, suggesting that the disruption of cellulose production in NC101 reduced its proinflammatory potential in an experimental model of chronic immune-mediated colitis.

FIG 8.

Luminal densities of NC101 or the bcsA mutant in WT or Il10^−/− mice. (A and B) Quantitative bacterial culture of feces collected from Il10^−/− mice (A) or WT mice (B) monoassociated with NC101 or the ΔbcsA mutant. (C and D) Quantitative bacterial culture of cecal contents collected from Il10^−/− mice (C) or WT mice (D) monoassociated with NC101 or the ΔbcsA mutant. Each symbol represents an individual mouse, and the data are for 5 to 8 mice per group. Open circles, mice monoassociated with NC101; closed squares, mice monoassociated with the ΔbcsA mutant; lines at medians, P values determined by pairwise comparisons by a Mann-Whitney test. *, P < 0.05; **, P < 0.01.

DISCUSSION

Environmental factors, such as iron availability, that may promote proinflammatory interactions between AIEC, microbes clinically relevant to IBD, and the host have not been well investigated. Therefore, the purpose of this study was to characterize how iron impacts the physiology and functional attributes of the AIEC strain NC101. Our findings demonstrate that iron promotes the cellulose-dependent aggregation of NC101. Moreover, NC101 aggregate cells are more susceptible to phagocytosis by macrophages. The contribution of cellulose to NC101 phagocytosis susceptibility and consequent macrophage proinflammatory responses changes as iron availability and the physiological state of NC101 are altered, demonstrating a dynamic role for cellulose in modulating host-microbe interactions. Finally, abrogation of cellulose production in NC101 reduced its ability to induce colitis in inflammation-prone Il10^−/− mice. Taken together, our results demonstrate that cellulose production alters the proinflammatory potential of NC101.

Various environmental factors, including temperature and nutrient availability, impact multicellular behaviors such as aggregation in E. coli and related enteric bacteria (64, 65). Interestingly, iron has divergent effects on the multicellular behaviors of other E. coli functional subtypes, including enteroaggregative (27), uropathogenic (47, 66, 67), and K-12 (25) strains. Therefore, given the various responses of different E. coli strains to alterations in iron availability, it is likely that multiple strain-specific mechanisms regulate these responses. Here we show that fur-deficient, but not basRS-deficient, NC101 exhibited a reduced ability to form aggregates. This suggests that intracellular, rather than extracellular, iron sensing by NC101 contributes to the induction of this multicellular behavior. This was not the result of an inability of the fur mutant to produce cellulose, suggesting that additional factors under the control of the Fur modulon promote maximal NC101 aggregation. Cellulose production by the fur mutant may enable it to form smaller microscopic aggregates that, compared to NC101 aggregates, are not macroscopically visible until later in growth. Additionally, the fur mutant exhibited a growth defect when grown in minimal medium with iron, which could contribute to decreased aggregate formation. However, this growth defect was evident only during later stages of growth, after aggregate formation had already occurred in the parental strain. Fur has been linked to the regulation of additional multicellular behaviors in UPEC, E. coli K-12, and other Gram-negative bacteria (38, 66–69), demonstrating the importance of iron as an environmental signal in modulating the formation of microbial communities across many bacterial species.

The translocation of microbes and their products across the intestinal epithelial barrier is detected by immune cells, including macrophages, where engagement of pattern recognition receptors by microbial products activates signal transduction pathways that promote phagocytosis (70), microbial killing, and production of inflammatory mediators. The assumption of an aggregate physiological state promoted NC101 phagocytosis by macrophages, where NC101 cells recovered from the aggregates were more susceptible to phagocytosis than planktonic cells. Abrogation of cellulose production prevented aggregation and reduced NC101 susceptibility to phagocytosis under aggregate-inducing conditions (i.e., iron exposure). Coinciding with a reduced ability to aggregate, the fur mutant was also phagocytosed to a lesser extent, even though it produced cellulose. Interestingly, increased phagocytosis of iron-exposed (71, 72) or biofilm-associated (73, 74) bacteria has been observed in other bacterial species and with iron-exposed extraintestinal E. coli pathogens (26). These results suggest that although cellulose is required for aggregation and, presumably, the assumption of an aggregate physiological state, other factors contribute to the phagocytosis of NC101 aggregate cells.

Under non-aggregate-inducing conditions (i.e., low-iron conditions) where NC101 aggregation does not occur but cellulose is expressed, deletion of bcsA enhanced NC101 susceptibility to phagocytosis. This suggests that in a nonaggregative state, cellulose acts as an antiphagocytic factor, potentially masking bacterial factors that interact with macrophage receptors to promote phagocytosis. Indeed, as no host receptor for cellulose has been identified, it is unlikely that microbial cellulose interacts directly with host cells. This is consistent with the contrasting effects of the disruption of cellulose production on NC101 phagocytosis, as both microbial iron exposure and the resulting physiological state of NC101 are altered. Thus, our study introduces cellulose as a novel factor that modulates interactions between AIEC and macrophages and highlights the complex interplay between bacterial and environmental factors in modulating host-microbe interactions.

Although our investigation demonstrates that cellulose is required for in vitro aggregation by NC101 and modulates interactions with macrophages, it is unclear whether NC101 cellulose production and aggregation occur in vivo. In a recent study by Arthur and colleagues, the impact of the inflamed and noninflamed colonic environments on the NC101 transcriptome was investigated (75). bcs transcripts were detected at 2, 12, and 20 weeks following the monoassociation of formerly GF Il10^−/− and inflammation-resistant Il10^−/− rag2^−/− mice, indicating that bcs genes are transcribed in vivo. However, as cellulose biosynthesis is primarily regulated through the allosteric control of cellulose synthase activity (76), the presence of bcs transcripts is not conclusive evidence of NC101 cellulose production in the colon. Moreover, current biochemical techniques for assaying the presence of bacterial cellulose in vitro are not easily adaptable to the intestinal environment, given the presence of plant cellulose and other polysaccharides consisting of glucose monomers.

The contribution of cellulose to the in vivo fitness and virulence potential of E. coli and related bacteria has been investigated only in UPEC strains in the urinary tract (35, 77) and S. Typhimurium when administered intraperitoneally (78). Therefore, to establish whether cellulose contributes to the colitogenicity of AIEC in the GI tract, the severity of colitis was assessed in Il10^−/− mice monoassociated with NC101 or the cellulose-deficient bcsA mutant. The onset of colitis was delayed in mice colonized with the cellulose-deficient mutant, which corresponded with decreased Th-17-associated immune responses, including the decreased expression of Il12b. This is consistent with our in vitro observations demonstrating decreased IL-12 p40 production by macrophages infected with the bcsA mutant following exposure to non-aggregate-inducing and iron-limiting conditions. The reduced macrophage production of IL-12 p40 also corresponded with the increased phagocytosis of the bcsA mutant. Therefore, cellulose may enhance the proinflammatory potential of NC101 by preventing the mucosal clearance of NC101 and, consequently, promoting increased proinflammatory immune responses. However, as the deletion of bcsA did not completely prevent colitis development and its effects were lost over time, other microbial factors likely contribute to the ability of NC101 to induce colitis. Finally, these results also suggest that iron may be limiting within the inflamed intestines, a finding that has been reported by others (79). However, the precise bioavailability of iron remains unclear, especially as iron concentrations likely vary throughout the GI tract and depend on other factors, including host iron status, inflammation, and diet.

Coinciding with decreased inflammation, the luminal loads of the bcsA mutant were significantly decreased. However, it is unclear whether a 2.4-fold decrease in fecal loads and a 1.8-fold decrease in cecal luminal loads significantly contribute to decreased immune activation in mice colonized with the bcsA mutant. The cecal luminal densities of the bcsA mutant were also decreased in Il10^−/− mice prior to evidence of histological inflammation and in noninflamed WT mice. However, this early difference in luminal bacterial loads was not observed in the feces. Therefore, cellulose production may modestly enhance AIEC colonic fitness, which provides a possible additional mechanism for augmenting the proinflammatory potential of NC101. Cellulose provides microbial resistance against a variety of stressors both in the environment (55, 67, 80) and within the host (77). For example, deletion of bcsA in UPEC enhanced bacterial clearance from the kidneys in a neutrophil-dependent manner (77). Cellulose-dependent multicellular behaviors can also be induced by stressors likely present at mucosal surfaces along the normal and inflamed GI tract, including fluctuations in iron availability, peroxide stress, and microbial contact with soluble IgA antibodies (47, 67, 81). Finally, as intestinal E. coli isolates demonstrate an ability to produce cellulose at 37°C more frequently than UPEC clinical isolates (58), it is tempting to speculate that cellulose may contribute to the intestinal fitness of resident intestinal E. coli strains.

In the colon, under homeostatic conditions, the mucosal surface is home to a distinct community of bacteria. The composition of the mucosa-associated microbial community is significantly altered in chronic disease states such as CD, which includes an increased abundance of mucosally associated resident E. coli isolates (3). Host inflammatory responses and intrinsic host genetic defects compromise mucosal and epithelial barrier integrity, enabling the enhanced proximity of mucosally associated bacteria to host cells. Consistent with this, enhanced intestinal tissue AIEC loads, mucosal association, and translocation (8, 15, 82) are correlated with more severe disease in CD and experimental models of colitis. Our study highlights the importance of environmental factors in altering AIEC physiology and subsequent host-microbe interactions and impact on inflammation. Given the lack of identifying genetic loci within AIEC, it would be interesting to investigate whether iron alters the physiological state of clinical AIEC isolates as a novel functional determinant. Finally, future studies confirming the biosynthesis of cellulose in vivo and, more broadly, assessing the in vivo physiological state of AIEC, especially within more defined intestinal niches, such as the normal and inflamed mucosa, are warranted. This could enable the identification of novel therapeutics that modulate the physiology of E. coli to limit adverse interactions with the underlying mucosa in CD patients and individuals genetically susceptible to CD.