Functional Genomic Studies of the Intestinal Response to a Foodborne Enteropathogen in a Humanized Gnotobiotic Mouse Model*

Members of the genus Listeria provide a model for defining host responses to invasive foodborne enteropathogens. Active translocation of Listeria monocytogenes across the gut epithelial barrier is mediated by interaction of bacterial internalin (InlA) and its species-specific host receptor, E-cadherin, whereas translocation across Peyer's patches through M-cells is InlA-independent. To define microbial determinants and molecular correlates of the host response to translocation via these two routes, we colonized germ-free transgenic mice expressing the human enterocyte-associated E-cadherin receptor with wild-type (WT) or mutant L. monocytogenes strains, or its nonpathogenic noninvasive relative Listeria innocua, or with Bacteroides thetaiotaomicron, a prominent gut symbiont. Mouse Gene-Chips, combined with Ingenuity Pathway software, were used to identify canonical signaling pathways that comprise the response to WT L. monocytogenes versus the other species. Gain- and loss-of-function experiments with L. innocua and L. monocytogenes, respectively, demonstrated that the 773-member transcriptional signature of the response to WT L. monocytogenes is largely conserved in the ΔinlA mutant. Internalin-dependent responses include down-regulation of gene networks involved in various aspects of lipid, amino acid, and energy metabolism and up-regulation of immunoinflammatory responses. The host response is markedly attenuated in a listeriolysin-deficient (Δhly) mutant despite its ability to be translocated to the lamina propria. Together, these studies establish that hly, rather than bacterial invasion of the lamina propria mediated by InlA, is a dominant determinant of the intensity of the host response to L. monocytogenes infection via the oral route.

The human gut is inhabited by a complex community of trillions of microorganisms representing all three known domains of life: Bacteria, Archaea, and Eukarya (1)(2)(3). Our microbiota is dominated by members of Bacteria, with components of two divisions, the Firmicutes and the Bacteroidetes, comprising Ͼ90% of all phylogenetic types in those few individuals where comprehensive 16S rRNA gene sequence-based enumerations have been performed (1,2). Although most of the estimated 500 -1000 bacterial species in the gut microbiota appear to enjoy a mutually beneficial relationship with their host, potential pathogens are also present or may be introduced through the consumption of food or water.
Members of the genus Listeria provide a model for comparing host responses to invasive versus noninvasive foodborne bacteria. Fully sequenced genomes are available from two species: Listeria monocytogenes, an enteroinvasive human pathogen that can cross the intestinal as well as blood-brain and placental barriers; and Listeria innocua, a nonpathogenic and noninvasive relative that shares 84% of its genes with L. monocytogenes (4). L. monocytogenes is estimated to be present in the small intestines of up to 5% of individuals yet only a few, typically those who are immunocompromised, develop invasive symptomatic disease (5).
L. monocytogenes and L. innocua can cross the follicle-associated epithelium (FAE) 5 that overlies the lymphoid follicles of Peyer's patches with equal efficiency (6). L. monocytogenes, unlike L. innocua, expresses internalin (InlA), a surface protein that is sufficient to promote bacterial internalization into enterocytes that express its receptor, human E-cadherin (hEcad). Epidemiological (7) and histopathological data (8,9), as well as experiments using human primary cells and tissue explants (8), indicate that InlA is an important virulence factor in humans, mediating targeting and crossing of both intestinal and placental barriers. A single amino acid difference (Pro 16 in human versus Glu 16 in mouse) enables human but not mouse E-cadherin to function as a receptor for InlA (10). In conventionally raised, adult transgenic mice expressing hEcad under the control of an enterocyte-specific promoter (Fabpi-hEcad), L. monocytogenes is able to invade enterocytes that cover small intestinal villi and enter the underlying lamina propria. This receptor-mediated invasion results in a rate of translocation across the intestinal barrier that far exceeds that of InlA-independent translocation across the FAE (6,9).
Once internalized, L. monocytogenes is able to infect macrophages and other cells that neighbor the epithelium, and ultimately, the bacteria disseminate to organs such as the spleen and liver. The ability of L. monocytogenes to survive and multiply in professional phagocytes situated in the lamina propria, or beneath the FAE, is dependent upon another virulence factor that is not present in the genome of L. innocua: listeriolysin (LLO). LLO, encoded by the hly gene, is a pore-forming, cholesterol-dependent cytolysin that mediates bacterial escape from phagosomes to the cytosol, where the bacterium is able to replicate and spread to neighboring cells (11).
A confounding problem with studies of oral listeriosis in conventionally raised, humanized Fabpi-hEcad mice is the extent to which other members of the microbiota contribute to observed host responses. Therefore, in this report, we have generated germ-free (GF) Fabpi-hEcad mice to directly assess the host response (i) as a function of the route of translocation of L. monocytogenes across the epithelial barrier (enterocyte versus nonenterocyte) and (ii) as a function of residency in the underlying lamina propria in the presence or absence of listeriolysin. To do so, we colonized the animals with (i) WT L. monocytogenes and isogenic mutant strains with inlA, inlB, or hly deletions; (ii) WT L. innocua and isogenic strains engineered to express InlA or LLO; or (iii) Bacteroides thetaiotaomicron, a sequenced, well characterized, human gut symbiont that is an adept, adaptive forager of dietary polysaccharides (12,13). Our results provide a direct view of the significance of the route of bacterial entry into and residency within the lamina propria on the host response.
Animals-All experiments involving mice were conducted using protocols approved by the Washington University Animal Studies Committee. GF C57BL/6J transgenic mice were housed in plastic gnotobiotic isolators (14) under a strict 12-h light cycle and fed a standard autoclaved chow diet (B&K Universal) ad libitum. 10 9 colony-forming units (CFU) of bacteria, cultured to mid-log phase, were taken up in 0.5 ml of PBS containing 50 mg of CaCO 3 and inoculated by gavage into 12-15week-old male mice that had been deprived of food, but not water, for 12 h. Animals were sacrificed at 72 h after inoculation; the small intestine was removed and cut into 16 equalsized segments (numbered 1-16; proximal-to-distal). Spleens were homogenized in PBS (2 ml/spleen), and CFU counts in this material and in luminal contents harvested from intestinal segment 15 were determined by streaking 10-fold serial dilu-tions onto brain-heart-infusion agar plates and incubating the plates for 2 days at 37°C under aerobic or anaerobic conditions. Immunohistochemical Studies-Optimal Cutting Temperature TissueTek compound (VWR Scientific)-embedded blocks of small intestinal segment 14 (15) were cryosectioned (7-mthick sections) and fixed in ethanol (Ϫ20°C for 5 min). Blocking steps, antibody dilutions, and washes were performed in 1% bovine serum albumin in PBS. Slides were stained with a mouse anti-human E-cadherin IgG (1:100) (9) followed by a goat antimouse IgG-fluorescein isothiocyanate conjugate (1:100; Molecular Probes). Nuclei were counterstained with 4,6-diamidino-2-phenylindole (40 ng/ml), and sections were visualized on an Axiovert 200M (Zeiss, Gottingen, Germany) fitted with an Axiocam MRm (Zeiss).
Five-m-thick sections of formalin-fixed, paraffin-embedded blocks prepared from intestinal segment 13 of mono-associated mice were stained with hematoxylin and eosin. Adjacent sections were stained with L. monocytogenes R11 or L. innocua R6 primary antibodies (16) [final dilution was 1:500 in blocking buffer (1% bovine serum albumin, 0.3% Triton X-100 in PBS)] followed by horseradish peroxidase-conjugated monkey antirabbit Ig (1:400). Sections were counterstained with hematoxylin. Ileal sections were also stained with rat anti-mouse CD45 and CD3 and F4/80 rat monoclonal antibodies (all from Pharmingen) followed by horseradish peroxidase-conjugated rabbit anti-rat Ig (1:400; Invitrogen). Antigen-antibody complexes were visualized by using reagents supplied in the Envision kit (Dakocytomation, Carpinteria, CA). Samples were viewed with an Axioskop 2 (Zeiss) microscope and images captured with an Evolution PM color camera (Mediacybernetics, Silver Spring, MD).
GeneChip Analysis-RNA was purified from small intestinal segment 12 of each mouse (Midi RNeasy kit (Qiagen) with oncolumn DNase digestion). Equivalent amounts of RNA from mice in each treatment group (n ϭ 4/experiment) were pooled, and two biotinylated cRNA targets (40 g/replicate) were independently prepared from each sample (17). cRNAs were hybridized to Affymetrix Mouse Genome U74Av2 GeneChips, and the resulting data sets were analyzed using DNA-Chip and significance analysis of microarrays (SAM) as follows. CEL files 6 were read into dChip, and Present/Absent calls were read in from the accompanying GeneChip operating software (Affymetrix) TXT files. The GeneChip with the median overall intensity served as the base line to which all other GeneChips within that treatment group were normalized. Signals were assigned to each probe set by using model-based expression (perfect match-mismatch model). Unsupervised filtering was performed using the following criteria: (i) variation across samples (standard deviation/mean) Ͼ0.40 and Ͻ10.00; (ii) called Present in Ն20% of arrays; and (iii) variation between replicates Ͻ0.2, as assessed by median value of standard deviation/mean. For two-class SAM analysis, expression values were exported from dChip only for probe sets called Present in at least one GeneChip within the group being analyzed. Significance was defined by maintaining a false discovery rate 6 Gene Expression Omnibus (GEO) accession number GSE7013.
(q-value) below 1%. Ingenuity pathway analysis (IPA) focus genes were identified as having significantly increased expression in the specified group; 6,351 probe sets were called "Present" in at least one of the GeneChips by GeneChip operating software (Affymetrix).

Generation of Germ-free Fabpi-hEcad Transgenic Mice-
Fabpi-hEcad C57BL/6J ϫ SJL/J transgenic mice that express hEcad in small intestinal enterocytes under the control of nucleotides Ϫ1178 to ϩ28 of the rat Fabpi gene (9) were backcrossed to C57BL/6J animals to generation N7 and then rederived as GF (14). The growth rate, adult weight, and fertility of Fabpi-hEcad transgenic mice were indistinguishable from that of their GF nontransgenic littermates (n ϭ 8 mice/group). Histologic assessment of serial sections, prepared along the cephalocaudal axis of their small intestines, revealed no discernible differences between transgenic and normal littermates. Immunohistochemical studies of 12-15-week-old animals confirmed that the GF transgenic mice expressed hEcad in villus enterocytes (Fig. 1A). The FAE in the distal small intestine, but not the colonic epithelium, contained detectable levels of immunoreactive protein (data not shown).
L. monocytogenes but Not L. innocua Behaves as an Enteroinvasive Microbe in Gnotobiotic Fabpi-hEcad Mice-Twelve-tofifteen-week-old GF C57BL/6J Fabpi-hEcad mice were inoculated with 10 9 CFU of WT L. monocytogenes, WT L. innocua, or B. thetaiotaomicron. Animals were sacrificed 72 h after inoculation, a time that corresponds to peak levels of L. monocytogenes in intestinal tissue and mesenteric lymph nodes of conventionally raised animals (9,18). No mortality occurred after 3 days in any group (n ϭ 8 animals/group), although L. monocytogenes-infected mice exhibited clinical signs of infection (diarrhea, roughcast fur, 5-15% weight loss, and tremor). After 72 h, levels of bacterial colonization in the jejunum (middle third of the small intestine) and ileum (distal third) were not significantly different among mice in each treatment groups (10 8 -10 9 CFU/ml luminal contents; see supplemental Fig. S1). Examination of ileal sections stained with anti-L. monocytogenes polyclonal antibodies (16) revealed bacteria invading the tips of intestinal villi and in the lamina propria (Fig. 1B), similar to our previously reported observations in infected conventionally raised C57BL/6J ϫ SJL/J Fabpi-hEcad animals (9). In contrast, invasion was not detected in the intestines of L. innocua-colonized transgenic mice (Fig. 1B). Consistent with systemic dissemination, quantitative CFU assays of splenic homogenates disclosed viable bacteria in L. monocytogenes-infected animals at levels equivalent to those observed in conventionally raised Fabpi-hEcad hosts. In contrast, spleens did not harbor appreciable numbers of viable organisms in L. innocua-or B. thetaiotaomicron-colonized controls (Fig. 1C).
Intestinal Response to an Enteropathogenic Invasive Bacterial Species-Histologic studies revealed prominent mononuclear cell infiltrates in the lamina propria of the intestines of L. monocytogenes-infected mice (Fig. 2A). These mononuclear cells reacted with a pan-leukocyte antibody (anti-CD45) and were predominantly macrophages (F4/80 positive) and T-lymphocytes (CD3 positive) (Fig. 2B). In contrast, no lamina propria infiltrates were observed in any of our L. innocua-or B. thetaiotaomicron-colonized mice (Fig. 2, A and B, and data not shown).
To further distinguish the host response to L. monocytogenes infection versus colonization with L. innocua or B. thetaiotaomicron, we performed GeneChip profiling of ileal RNAs prepared from mice that had been colonized (mono-associated) with each bacterial species, as well as GF transgenic controls (n ϭ 4/group). Using the stringent filtering criteria described under "Experimental Procedures," we identified 304 genes with significantly different levels of expression between these four colonization states (Fig. 2C). Unsupervised hierarchical clustering of these 304 genes established that WT L. monocytogenes elicits a highly reproducible host response, as evidenced by the high degree of similarity in profiles from two independently colonized groups of infected animals (n ϭ 4 mice/group) (Fig.  2C). The response to L. monocytogenes is strikingly different from that elicited by L. innocua, despite the high degree of relatedness between these two Listeria species (2523 of 2853 L. monocytogenes genes have orthologs in L. innocua; the 270 genes specific to L. monocytogenes include all of its known virulence factors, such as inlA, inlB, hly, and actA) (4). The clustering also disclosed that the ileal transcriptome of L. innocua- colonized animals was more similar to GF controls than the transcriptome expressed in B. thetaiotaomicron-colonized mice.
Based on the results obtained with this unsupervised clustering, we used a two-class SAM analysis to compare GeneChip data sets from the ileums of WT L. monocytogenes-colonized mice with ileal GeneChip data sets generated from GF, L. innocua-and B. thetaiotaomicron-mono-associated animals; our goal was to define a set of genes that distinguishes the intestinal response to WT L. monocytogenes infection. The comparison yielded 614 genes that exhibited significantly higher expression in L. monocytogenes-infected mice and 159 genes with significantly lower expression (q-value Ͻ1%; total of 773 genes; Fig. 3A).
We subsequently used IPA software to identify gene networks that are significantly over-represented among the 614 L. monocytogenes-induced genes. This software utilizes a knowledge base of more than 1,000,000 functional and physical interactions for Ͼ23,900 mammalian genes (including 8,200 for the mouse). 255 of the 614 genes are functionally annotated in the IPA knowledge base; 83 of these lie within the IPA "immune response" category (p Ͻ 10 Ϫ6 by Fisher's exact test; see supplemental Fig. S2). We then utilized 134 genes from immune response and nine other high level IPA functional categories that are over-represented (p Ͻ 10 Ϫ4 ; supplemental Fig. S2) to construct an unsupervised master gene interaction network. Elimination of 51 orphan genes, which lack established interactions with other induced genes, yielded an 83-member "master" network with 237 edges (Fig. 3D) that characterizes the response to enteroinvasive WT L. monocytogenes.
By utilizing "focus genes" from the master network and adding the 6,351 probe sets called Present in the ileal RNAs, we were able to show that components of the Jak/Stat, B-cell receptor, interleukin-10, interleukin-6, Toll-like receptor, antigen presentation, NF-kB, granulocyte-macrophage colony-stimulating factor (GM-CSF), ERK/MAPK, PI3K/Akt, G-proteincoupled receptor, integrin, Vegf, Pten, apoptotic, and ephrin receptor signaling pathways are all part of the response of the ileum to L. monocytogenes infection (see supplemental  Using the stringent selection criteria described under "Experimental Procedures," 304 genes were identified as displaying significantly changed expression across the four sample groups. Unsupervised hierarchical clustering, based on this list of 304 genes, established that the host response to L. monocytogenes infection was highly distinct from the responses to noninvasive L. innocua, or B. thetaiotaomicron, both of which cluster with GF controls. See supplemental Table S1 for a list of the 304 genes and their corresponding expression values. Scale bars in A and B are 10 m. InlA Is Not the Major Bacterial Determinant of L. monocytogenes-specific Intestinal Response-We used this derived 773member molecular signature of the host response to WT L. monocytogenes (Fig. 3) to identify some of its bacterial genetic determinants. Two experiments, one a gain-of-function and the other a loss-of-function, were designed to assess the contributions of the InlA-hEcad interaction. GF Fabpi-hEcad mice were colonized with a strain of L. innocua engineered for heterologous expression of InlA (Li(inlA)) or with mutant strains of L. monocytogenes that lack either inlA (Lm(⌬inlA)) or both inlA and inlB (Lm(⌬inlAB)). Unlike its isogenic WT L. innocua parent, which lacks an InlA homolog, the engineered internalin-expressing L. innocua strain is able to invade cultured Caco-2 cells, transfected fibroblasts expressing E-cadherin, guinea pig ileal enterocytes, and ex vivo infected Fabpi-hEcad mouse intestinal explants (9,22).
All three bacterial strains colonized the intestines of GF transgenic mice at levels that were not significantly different from their WT parents (n ϭ 4 -8 mice/group; see supplemental Fig. S1). Expression of L. monocytogenes inlA in L. innocua (Li(inlA)) permits entry into villus enterocytes: however, no bacteria were observed in the lamina propria. Leukocytic infiltrates were not detected and systemic dissemination to the spleen was not observed (Fig. 1C plus data not shown). Using the molecular signature defined above as reference, GeneChip analysis of ileal RNAs from Li (inlA)-colonized Fabpi-hEcad mice revealed a host response that was most similar to that elicited by WT L. innocua (Fig. 3A).
Strains of L. monocytogenes lacking InlA or InlA and InlB (Lm(⌬inlA); Lm(⌬inlAB), respectively) were undetectable in villus enterocytes or in the underlying lamina propria (n ϭ 4 -8 mice/group; e.g. Fig. 3B). Nonetheless, leukocytic infiltrates in the lamina propria were present in each mouse surveyed (compare Fig. 3C with Fig. 2, A and B). Moreover, like its WT L. monocytogenes parent strain, the isogenic Lm(⌬inlA) mutant disseminated to the spleen (Fig. 1C), albeit at a lower level than WT. Together, these findings suggest that internalin-mediated invasion of hEcad-expressing villus enterocytes is not strictly required for systemic dissemination in gnotobiotic mice.
GeneChip analysis indicated that the intestinal response to Lm(⌬inlA) or Lm(⌬inlAB) contrasts sharply with that elicited by L. innocua or Li(inlA), and resembles that of WT L. monocytogenes (Fig. 3A). Direct comparison of ileal expression profiles from Lm(⌬inlA)-and Lm(⌬inlAB)-mono-associated mice with WT L. monocytogenes-colonized mice revealed 157 genes that are regulated via an internalin-dependent mechanism (see supplemental Fig. S5A). When the levels of expression of these 157 genes were referenced to the differences noted between WT L. monocytogenes versus the GF, L. innocua and B. thetaiotaomicron base-line groups, it was apparent that the internalin mutants produce attenuated responses for both up-and downregulated genes (supplemental Fig. S5B). Ingenuity-based functional categorization of the 77 genes that were significantly down-regulated upon L. monocytogenes infection via the hEcad-mediated route (defined as decreased expression in WT L. monocytogenes infection compared with Lm(⌬inlA) and Lm(⌬inlAB) infection), revealed two broad categories that were enriched: "lipid metabolism" and "small molecule biochemis- A, results obtained from two-class SAM that identified (q-value Ͻ1%) a response distinctive for L. monocytogenes (Lm) compared with that elicited by B. thetaiotaomicron (Bt)-colonized and L. innocua-colonized mice and GF controls. 614 genes show higher expression in the group infected with L. monocytogenes, whereas 159 genes exhibit lower expression. The expression pattern for these 773 genes is also shown in gnotobiotic mice colonized with isogenic Lm(⌬hly) or Lm(⌬inlA) strains or with mice mono-associated with L. innocua expressing InlA (Li(inlA)) or LLO (Li(hly)). For a complete list of genes, their associated q-values, and -fold differences in expression see supplemental Table S2. B, immunohistological localization of L. monocytogenes InlA mutant (Lm(⌬inlA)) or LLO mutant (Lm(⌬hly)) in the ileums of Fabpi-hEcad mice (red). Cell nuclei are counterstained with hematoxylin (blue). Bacteria are shown at higher magnification in the insets. C, histochemical (left panels) and immunohistochemical (right panels) analyses of bacteria populating the lamina propria in Lm(⌬inlA)-or Lm(⌬hly)-colonized mice. For immunohistochemical studies, intestinal sections were stained with antibodies to leukocytes (CD45), T-lymphocytes (CD3), or macrophages (F4/80) (red, positive staining; blue, cell nuclei).