Microbial Colonization Coordinates the Pathogenesis of a Klebsiella pneumoniae Infant Isolate

Enterobacteriaceae are among the first colonizers of neonate intestine. Members of this family, such as Escherichia and Klebsiella, are considered pathobionts and as such are capable of inducing local and systemic disease under specific colonization circumstances. Interplay between developing microbiota and pathogenic function of pathobionts are poorly understood. In this study, we investigate the functional interaction between various colonization patterns on an early colonizer, K. pneumoniae. K. pneumoniae 51-5 was isolated from stool of a healthy, premature infant, and found to contain the genotoxin island pks associated with development of colorectal cancer. Using intestinal epithelial cells, macrophages, and primary splenocytes, we demonstrate K. pneumoniae 51-5 upregulates expression of proinflammatory genes in vitro. Gnotobiotic experiments in Il10−/− mice demonstrate the neonate isolate induces intestinal inflammation in vivo, with increased expression of proinflammatory genes. Regulation of microbiota assembly revealed K. pneumoniae 51-5 accelerates onset of inflammation in Il10−/− mice, most significantly when microbiota is naturally acquired. Furthermore, K. pneumoniae 51-5 induces DNA damage and cell cycle arrest. Interestingly, K. pneumoniae 51-5 induced tumors in ApcMin/+; Il10−/− mice was not significantly affected by absence of colibactin activating enzyme, ClbP. These findings demonstrate pathogenicity of infant K. pneumoniae isolate is sensitive to microbial colonization status.

Similar to E. coli, K. pneumoniae is a rod shaped, facultative anaerobe, and a member of the Enterobactericiae family known to colonize the lung, oral cavity and intestine. Indeed, K. pneumoniae has been identified as a source of varied systemic and local infections 20 . Although most commonly associated with infection and inflammation of the lung 21,22 , there are reports that highlight an association between Klebsiella and intestinal maladies 23,24 . Mouse models of colitis have demonstrated inflammation can enhance the intestinal colonization of K. pneumoniae 25,26 , although K. pneumoniae colonization does not initiate disease. Additionally, there are few studies that report cooperation of Klebsiella with other microbes 25 or chemical agents 27 to enhance inflammatory reponse in preclinical models. To date, no study has shown an intestinal Klebsiella isolate could singly promote inflammation or carcinogenesis and whether colonization pattern influence this response.
With the use of gnotobiotic mice, we were able to simulate various colonization methods to investigate the pathogenic potential for the gut of a neonate isolate, K. pneumoniae . We demonstrate that a nonexistent and a slowly acquired microbiota, provide optimal conditions for K. pneumoniae 51-5 to induce intestinal inflammation and influence CRC.

K. pneumoniae infant isolate induces inflammatory gene expression in vitro.
A gram-negative bacterium (K. pneumoniae  was isolated from the stool of a premature infant. Using 16S sequencing and multi-locus sequence typing (MLST) analysis (Supplementary Table S1), we confirmed that the isolate belonged to the species Klebsiella pneumoniae, belonging to sequence type 1243 (ST1243).
To assess the inflammatory potential of our K. pneumoniae 51-5 infant isolate, we infected intestinal epithelial cells (IEC) and immune cells and examined cytokine mRNA expression using real-time PCR. We observed a significant upregulation of Cxcl1 (p = 0.0077, p = 0.0005) and Cxcl2 (p = 0.0019, p = 0.0002) mRNA in K. pneumoniae infected IEC-6 cells (Fig. 1a) and in Mode K cells ( Supplementary Fig. S1), respectively, with no significant changes in Tnfα or Il6 mRNA levels. In addition, a significant upregulation of Tnfα (p = 0.0407), Il-1β (p = 0.0388), and marked increase in Il-6 (p = 0.0529) mRNA transcripts, compared to uninfected controls, was observed in immune RAW cells (Fig. 1b). Finally, primary murine splenocytes infected with K. pneumoniae 51-5 showed increased expression of Tnfα (p = 0.0069), Il-6 (p = 0.0002), and Il-1β (p = 0.0018) mRNA (Fig. 1c). These data demonstrate the potential of the infant K. pneumoniae isolate to trigger an inflammatory response in both IECs and immune cells.
K. pneumoniae monoassociation induces severe colitis. To assess the inflammatory potential of K. pneumoniae 51-5 in vivo, we performed mono-association studies in germ-free (GF) Il10 −/− mice (129SvEv), a colitis model susceptible to bacteria colonization. Mice were orally gavaged with K. pneumoniae 51-5 (10 9 CFU/ mice) and maintained in gnotobiotic isolators for 8 and 12 weeks (Fig. 2). Successful colonization was confirmed measuring viable bacteria in reconstituted feces (Fig. 2d). Histological assessment revealed a significant increase of inflammation in mono-colonized mice compared to GF Il10 −/− controls (p = 0.0421) at 8 weeks, mostly localized to proximal region of the colon. Histological presentation of the small intestine of K. pneumoniae 51-5 infected mice was similar to those of GF controls, suggesting a colon-specific inflammatory effect of the bacterium (data not shown). After 12 weeks, mono-colonized mice showed development of pan-colitis and significant increase in inflammation (p = 0.0022 vs GF control), with marked hyperproliferation present throughout the colon, particularly in the distal regions (Fig. 2b).
Microbiota assembly regulates K. pneumoniae-induced intestinal inflammation. We next tested whether K. pneumoniae 51-5 can induce colitis in the presence of a complex and diverse microbiota using a co-colonization model (Fig. 3a).
In this simultaneous colonization model, GF Il10 −/− mice were colonized for 12 weeks by single oral gavage with specific pathogen free (SPF) microbiota obtained from the ceca of wild type mice alone or in combination with K. pneumoniae 51-5 (SPF + Kp). We measured fecal lipocalin 2 (Lcn-2), a sensitive marker of inflammation, through the course of the experiment. K. pneumoniae-infected mice produced a significantly higher amount of Lcn-2 (p = 0.0136) at an early timepoint (4 weeks) (Fig. 3b). Interestingly, the addition of K. pneumoniae 51-5 did not have a synergistic effect on intestinal inflammation as assessed by H&E (Fig. 3c) and mean inflammation score (1.98 vs 1.89) (Fig. 3d) after 12 weeks. To determine whether the infant isolate was excluded due to additional bacteria of the SPF inoculum, we plated reconstituted stool to determine bacterial colonization. The presence of the complex biota did not impact K. pneumoniae 51-5 colonization level over the course of the experiment (Fig. 3e). No significant difference was observed in immune infiltration by CD3 immunostaining nor cytokine (Tnfα, Il-6, Cxcl2) mRNA accumulation between cohorts, although Il1β was significantly increased in K. pneumoniae colonized mice ( Supplementary Fig. S3).
To determine whether K. pneumoniae 51-5 can initiate inflammation in mice with a pre-established microbiota, we used a delayed colonization model where SPF biota was introduced by oral gavage 2 weeks prior to K. pneumoniae infection (Fig. 3f). In this model, a significant increase in Lcn-2 (p = 0.0087) production was also observed with the addition of K. pneumoniae 51-5 after 4 weeks compared to controls (Fig. 3g). Similar to the previous model, we observed a loss of this separation over time, with no significant difference in histology (Fig. 3h) or inflammation score after 12 weeks (Fig. 3i), suggesting K. pneumoniae plays an important role in the onset of inflammation.
To examine the influence of microbiota assembly on host susceptibility to K. pneumoniae infection, we used a primary colonization model where GF Il10 −/− mice were transferred to SPF housing and immediately colonized with Klebsialla 51-5 by oral gavage (Figs 4a and S4). In this model, control (no gavage) and K. pneumoniae associated mice were allowed to naturally acquire an intestinal microbiota through husbandry environmental exposure for 20 weeks. Mice were sacrified at 8 and 20 weeks following transfer to assess intestinal inflammation at early and late time points respectively. After 8 weeks, we observed a significant increase in inflammation (p = 0.00376), particularly in the distal colon (p = 0.0263) with a significant increase in number of CD3-positive immunostained cells (p = 0.0130) ( Supplementary Fig. S4). K. pneumoniae-infected mice demonstrated significantly higher Lcn-2 production as early as 4 weeks post-infection (p = 0.0051) that was sustained to 8 weeks (p = 0.0059) ( Supplementary Fig. S4g). A significant increase of proinflammatory cytokines Tnfα (p = 0.0159), Il-6 (p = 0.0159), Cxcl2 (p = 0.0357) and Il1-β (0.0159) mRNA was also detected in colonic tissues  (Kp). Results are representative of at least three independent experiments performed in triplicates. Graphs represent means ± SD. **P < 0.01; ***P < 0.001 (Unpaired student t-test, Welch's correction).
www.nature.com/scientificreports www.nature.com/scientificreports/ of K. pneumoniae-infected mice after 8 weeks colonization ( Supplementary Fig. S4h). Following 20 weeks of colonization, K. pneumoniae-associated mice demonstrated increased colonic hyperplasia (Fig. 4b) and a significant increase in immune infiltration as assessed by CD3 immunostaining (p = 0.0286) (Fig. 4g), which was confirmed by increased inflammation score (p = 0.0002) (Fig. 4c). Much of this inflammation was localized to the distal colon ( Fig. 4d) of infected mice (p = 0.0021). A significant increase in fecal lipocalin (p = 0.0184),  www.nature.com/scientificreports www.nature.com/scientificreports/ Genotoxic K. pneumoniae contributes to colitis-associated tumorigenesis. As previously mentioned, the pks island is a conserved sequence found among select bacteria and identified as the source of cellular DNA damage. Using primers designed for the flanking regions of the pks in E. coli, we detected the presence of the pks island in the genome of our infant K. pneumoniae isolate, suggesting that this strain could exert a genotoxic effect ( Supplementary Fig. S6). Indeed, functional assays showed K. pneumoniae 51-5 infection increased DNA damage in IEC-6 cells as determined by increased comet tails and γH2AX immunofluorescence (Fig. 5a-c). K. pneumoniae-infected cells also demonstrated significantly increased (p = 0.0156) cell numbers in the G2 phase (Fig. 5d). To determine whether the pks was necessary for DNA damage induced by our isolate, we deleted the clbP gene which encodes for ClbP enzyme, a peptidase responsible for the release of mature toxin 28,29 ( Supplementary  Fig. S6). Infection with isogenic clbP mutant (Kp ΔclbP) was sufficient to abrogate DNA damage as observed by reduction of comet tails, decreased γH2AX immunofluorescence and cell cycle arrest (p = 0.0364) (Fig. 5a-d). We www.nature.com/scientificreports www.nature.com/scientificreports/ then tested the in vivo genotoxic potential of K. pneumoniae 51-5 by colonizing ex-GF Il10 −/− mice for seven days and examining γH2AX production (Fig. 5e). The colon of K. pneumoniae 51-5 infected mice showed increased staining for γH2AX compared to uninfected (p = 0.0202) or K. pneumoniae 51-5 ΔclbP infected mice (Fig. 5e). www.nature.com/scientificreports www.nature.com/scientificreports/ Bacterial colonization of Kp ΔclbP was confirmed by stool plating to enumerate CFU, with no significant change compared to K. pneumoniae-infected mice (Fig. 5f). These results confirm the genotoxic capability of the K. pneumoniae neonate isolate.
We recently published a novel model of spontaneous colitis-associated tumorigenesis where GF Apc Min/+ ; Il10 −/− mice developed colonic tumors when colonized with specific microorganisms 30 . To determine the pro-tumorigenic capacity of Klebsiella 51-5, we transferred GF Apc Min/+ ; Il10 −/− mice to SPF housing and colonized them with K. pneumoniae 51-5 via oral gavage. After 20 weeks of colonization, K. pneumoniae-infected mice demonstrated a significant increase (p = 0.0009) in the number of colonic tumors (Fig. 6a,b), with no significant increase in inflammation (Fig. 6c,d).
To assess the impact of colibactin deficiency on tumorigenesis we infected Apc Min/+ ; Il10 −/− mice with K. pneumoniae 51-5 ΔclbP or K. pneumoniae 51-5. While K. pneumoniae ΔclbP mice presented with fewer colon tumors compared to K. pneumoniae-infected mice, these differences were not statistically significant (Fig. 6e,f). Additionally, there was no significant difference detected in the level of inflammation between cohorts (Fig. 6g,h). While ClbP is essential for genotoxicity, it does not seem to significantly affect the pro-carcinogenic potential of K. pneumoniae 51-5. Overall, our data reveal that K. pneumoniae 51-5 mediates intestinal pathologies which is sensitive to microbiota status.

Discussion
Enterobacteriaceae are among the early bacterial colonizers of the lower intestinal tract, and the relative abundance of some genera from this family such as Escherichia and Klebsiella are often increased in intestinal pathologies including IBD, NEC and CRC 23,24,31,32 . Although the colitogenic potential of some strains of E. coli has been demonstrated in various gnotobiotic conditions 33 , the relationship between K. pneumoniae and colitis is still unclear. Here, we demonstrated that a K. pneumoniae infant isolate, K. pneumoniae 51-5, has the ability to promote intestinal inflammation in a host lacking the immunosuppressive cytokine Il10. This colitogenic property is observed in mono-associated and in naturally colonized Il10 −/− mice, but diminished in mice colonized with a full complement of microbiota. Since defective Il10 signaling has been associated with very early onset IBD [34][35][36] , interactions between environment, such as early bacterial colonization and host genetics, is likely a critical component of intestinal pathology.
To determine interaction between microbiota and colitogenic potential of K. pneumoniae 51-5, we utilized a varied set of microbial colonization models employing a gnotobiotic approach with Il10 −/− mice. We first mono-associated Il10 −/− mice with K. pneumoniae 51-5. We observed significant hyperplasia and immune cell invasion response throughout the colon after 12 weeks of colonization, suggesting this K. pneumoniae isolate is an alpha bacterium in this model. Interestingly, in the presence of a complex microbiota, K. pneumoniae colonization had no synergistic effect on inflammation, despite maintaining a high colonization level. Temporal analysis of Lcn-2 production allowed us to conclude that K. pneumoniae infection may play a role in the early stages of inflammation, regardless of microbial status. This supports previously published reports that utilize the addition of K. pneumoniae to accelerate inflammation in mouse models of intestinal disease 27,37 . In contrast, when GF mice acquired microbiota through environmental exposure (natural acquisition), K. pneumoniae colonization resulted in a strong inflammatory response compared to control mice. These experimental groups had comparable levels of K. pneumoniae colonization suggesting the attenuated inflammation observed in the SPF + K. pneumoniae 51-5 and the naturally colonized mice + K. pneumoniae 51-5 cohorts is not due to exclusion of the bacterium from the intestine. The extent of inflammation observed between these groups could be explained by the early interactions between K. pneumoniae 51-5 and resident microbes at the time of infection. We previously reported that natural acquisition of microbiota from the environment results in an initial varied composition of bacterial communities compared to SPF gavaged mice, that becomes more comparable between the 4 and 8 week timepoint after colonization 38 . The effects of K. pneumoniae on inflammation has been most successfully studied within the context of lung infections, an environment initially thought to be sterile. Recent identification of an oral isolate of K. pneumoniae inducing colonic inflammation was performed under germ free or antibiotic conditions 39 . Our data corroborates these findings in that mice devoid of bacteria are highly susceptible to K.pneumoniae infection, but also introduces another concept where a shift in the relative amounts of certain bacterial communities, through modulation of microbiota assembly, can regulate its colitogenic effects. K. pneumoniae belongs to the Proteobacteria phyla which is known to predominate in the immature gut 40 and a decrease in this community is associated with maturation of the host biome. Indeed we have previously demonstrated a decrease in Proteobacteria abundance with time after conventionalization of germ free mice by gavage or natural acquisition 38 . It is possible that gavage with K. pnemoniae 51-5 can influence the microbiota development by sustaining a high level of Proteobacteria which could influence pathogenesis of this isolate. Further investigation using taxonomic analysis (16S rDNA sequencing) will be needed to determine impact of early gavage of K. pneumoniae on host microbiota composition.
Host genetics also play a crucial role in microbial pathogenesis. For example, K. pneumoniae was one of the bacterial species enriched in the model of T-bet −/− ;Rag2 −/− ulerative colitis (TRUC) 25 . However the bacterium was unable to induce disease on its own when introduced to GF TRUC mice 25 . When introduced to a WT or Rag2 −/− mouse, K. pneumoniae, in concert with TRUC resident microbes, was able to elicit inflammation 25 . While this response may be specific to the isolate, a possible mechanism could be due to unnoticed defects in Th1 signalling that may be intrinsic to TRUC. Recently, an oral isolate was demonstrated to elicit colonic inflammation in GF Il10 −/− model via Th1 immune signaling, particularly Ifnγ + T cells 39 . While we observed significant upregulation of Ifnγ via cytokine array ( Supplementary Fig. S2 and S5), further analysis is needed to confirm regulation through Th1 signalling in our models. www.nature.com/scientificreports www.nature.com/scientificreports/ Sequencing analysis identified a high relative abundance of K. pneumoniae within the mucosal sites of CRC patients 23 however the functional role of this bacterium in CRC is unknown. The presence of the pathogenic pks island, responsible for the production of the genotoxin colibactin, has been identified in Western CRC patients and is critical for E. coli-induced tumor formation in mice 14,41 . We demonstrated that an active colibactin is generated from the infant isolate K. pneumoniae 51-5 since genetic ablation of the clbP gene reduces DNA damage www.nature.com/scientificreports www.nature.com/scientificreports/ and cell cycle arrest in cell lines. Surprisingly, K. pneumoniae-induced CRC in Apc Min/+ ; Il10 −/− mice was not significantly attenuated when colibactin is functionally disrupted, suggesting that other mechanisms contribute to carcinogenesis. It is important to note the full length pks gene cluster was not deleted in our experiments, and as such, there remains several Clb genes whose functions extend beyond colibactin production. For example, clbS encodes for an anti-toxin 42,43 , clbM produces a xenobiotic transporter 44,45 while clbA participates in the production of iron scavenger siderophores by Enterobactin (Ent) and yersiniabactin (Ybt) found in the High-Pathogenicity Island (HPI) loci 46 . Dual loss of Ent and Ybt is sufficient to reduce cytokine production and Klebsiella colonization in lung homogenates 22 . In addition, deletion of the clbA gene from K. pneumoniae sufficiently reduced cytokine production and bacterial colonization in a murine model of meningitis 47 . It is also noteworthy that K. pneumoniae contains several virulence factors demonstrated to affect colonization, immune cell recruitment, and inflammation [47][48][49] . One example is the fimbrial adhesion gene, fimK, which functions in host cell attachment, and shown to be critical for lung colonization and proinflammatory response 48 . Further studies would be needed to address the function of pks island and other virulent genes in K. pneumoniae 51-5. In summary, we have demonstrated early acquisition of an infant isolate of K. pneumoniae, is sufficient to induce inflammation in a microbe-free environment or immature nascent microbiota, while the presence of a complex biota attenuates this response.

Methods ethics statement.
All experiments were performed in accordance with relevant guidelines and regulations.
Written informed consent was obtained from the parent(s) of the infant and in accordance with the principles expressed in the Declaration of Helsinki. The study, including consent procedure was approved by the UF Health Institutional Review Board 01.

Isolation and characterization of the neonate Klebsiella pneumoniae clinical isolate. The
K. pneumoniae 51-5 isolate was obtained from a frozen or glycerol preserved sample from Mac Conkey agar by screening for lac + colonies that lack motility. Identity of the isolated strain was then confirmed via Sanger sequencing of the 16S rRNA gene and qPCR for K. pneumoniae specfic genes, gltA 50 and ITS 51 . Initial characterization using standard microbiological methods resulted in positive test result of the isolate for urease test, negative for indole test, and negative for gram staining (data not shown). The presence of the genotoxic pks island was determined by PCR using primers specific for flanking regions of the pks island (R1 and R2) as described previously 14 . Finally, we performed a Multi-Locus Sequence Typing (MLST) Analysis on our K. pneumoniae 51-5 using seven house keeping genes commonly used to confirm the identity of Klebsiella pneumoniae isolates 52 . Briefly, genomic DNA was isolated from an overnight culture of K. pneumoniae 51-5. Using the primers provided by the University of Oxford PubMLST 53 (https://pubmlst.org/kpneumoniae/info/primers.shtml), genes were amplified by PCR, gel purified (Qiagen Mini-Elute Gel Extraction) and submitted for sequencing. Forward and reverse sequences were quality trimmed and joined using the Staden package 54 (http://staden.sourceforge.net) to generate a consensus sequence for each locus. The consensus sequence was then aligned to the reference sequence from PubMLST using the online NCBI BLAST align two sequences program (accessed on October 29, 2018) to generate the percent sequence identity between the our consequence and the full length PubMLST reference. The sequences were also submitted to the Institute Pasteur https://bigsdb.pasteur.fr/klebsiella/klebsiella.html to identify the sequence type.
Generation of isogenic ΔclbP mutant. All bacterial strains and plasmids generated in this study are listed in Supplementary Table S2. All strains were stored at −70 °C in peptone-glycerol. E. coli Top 10 was used as an intermediate host for cloning procedures and grown routinely at 37 °C in LB broth 55 or on LB solidified with 1.2% Bacto Agar (Difco). The K. pneumoniae strain used in this study was grown in LB medium at 37 °C. When required, chloramphenicol (30 μg/ml, Cm) was added. LB containing 8% sucrose was used for sacB gene-based counter selection in allelic exchange experiments. Primers clbP1/clbP2 and clbP3/clbP4 (Supplementary  Table S3) were used for amplifying the flanking regions of the clbP gene. The fused flanking region (ΔclbP) was ligated into the KpnI and XmaI sites of a suicide plasmid pRE112 to form the plasmid pKp-clbP1. All constructed plasmids were verified through sequencing. The suicide plasmid pKp-clbP1 (KpΔclbP) was conjugationally transferred from E. coli χ7213 56 to the infant isolate K. pneumoniae. Single-crossover insertion strains were isolated on LB agar plates containing Cm. Removal of the linear suicide vector from chromosome after the second recombination between homologous regions was selected by using the sacB-based sucrose sensitivity counter-selection system 57 . The Cm sensitive colonies were screened and verified by PCR using a primer set clbP1/clbP4. The mutant strain was confirmed by DNA sequence analysis.
In vitro infection studies. Rat small intestinal epithelial cells (IEC-6), murine duodenal epithelial (MODE-K), and mouse macrophage (RAW) cell lines were used in this study. For infections, K. pneumoniae was grown overnight in LB broth, and subcultured the following day for 2 hours. Infections were performed as previously described 14 . Cultured cells were infected (MOI: 100) for 3 hours, in triplicate, and harvested for qPCR analysis. Primary splenocytes were isolated from the spleens of specific pathogen free (SPF) WT C57Bl/6 mice as described previously 58 . Briefly, harvested spleens were freshly ground, collected with sterile D-PBS, and centrifuged (400 × g) to collect cells. To deplete red blood cells, the pellet was treated with RBC lysing buffer (0.15 M NH 4 Cl), filtered and washed with 10 mL of complete RPMI media. Splenocytes (2 × 10 6 ) were seeded in 6 well plates for infection. After addition of bacteria, the plate was centrifuged (200 × g) to facilitate adhesion and incubated for three hours. RNA was isolated using Trizol and reverse-transcribed using iScript cDNA synthesis kit (BioRad, #170-8891) to generate cDNA. www.nature.com/scientificreports www.nature.com/scientificreports/ Immunostaining and cell cycle analysis. Following bacterial infection, cells were washed several times, followed by overnight gentamicin treatment. For immunostaining, cells were formalin-fixed, and permeabilized with methanol. After a one-hour block in 5% NGS, cells were incubated overnight in Phospho-γH2AX antibody (Cell signaling # 9718S). Flow cytometric analysis of Phospho-H2AX and cell cyle was performed as previously described 14,44 . Cells were stained overnight with conjugated with H2AX antibody (Alex Fluor 647 anti-H2A.X-Phosphorylated (Ser 139), Biolegend), and stained with propidium iodide. Data was collected using BD LSR Fortessa cytometer. Data analysis was performed using FCS-6 express software (De Novo Software).

In vivo infection studies in mice.
All mouse experiments were performed under the guidelines and approval of the Institutional Animal Care and Use Committee of the University of Florida (Protocol numbers: 20169606 and 201608025). For monoassociation studies, Germ-free (GF) Il10 −/− mice (129SvEv; n = 8/group) were transferred to gnotobiotic isolators and one week later orally gavaged with the clinical isolate K. pneumoniae (10 9 CFU) and euthanized at 8 and 12 weeks following colonization. Stool samples from each mouse were taken every 2 weeks post-gavage and upon end point prior to euthanesia to measure bacterial colonization. For all other studies, GF Il10 −/− or Apc Min/+ ; Il10 −/− mice were transferred to SPF housing and orally gavaged with K. pneumoniae and/or SPF microbes from reconstituted cecal contents as indicated. All mice were monitored throughout the experiment for signs of distress. Stools were collected at regular intervals in liquid nitrogen and stored at −80 °C until further processing. Colonization was confirmed 1-2 weeks post-gavage via serial plating of reconstituted stool on MacConkey agar and reported as the number of colony forming units per gram (CFU/g) or via qPCR. Progression of inflammation was monitored using the Duoset Murine Lipocalin (Lcn-2) ELISA kit (R&D Systems Cat# DY1857) to measure fecal lipocalin. Upon termination of the experiment, sections of proximal and distal colon tissue were snap frozen for RNA analysis. The remaining tissue was swiss-rolled, formalin fixed and submitted for further processing. Fixed samples were paraffin-embedded and H&E stained by the UF Molecular Pathology Core. Inflammation score (0-4) was evaluated by at least one blinded independent investigator based upon histological examination of the mucosa, immune cell infiltration of the lamina propria, epithelial hyperplasia and crypt architecture as described previously 59 .
Immunostaining of formalin fixed samples for CD3 (goat anti-CD3-ε, Santa Cruz #sc-1127) was performed using Vectastain ABC (goat IgG) and M.O.M peroxidase kits (Vector labs, #NC9685284). Immune cell infiltration was assessed by quantification of CD3-positive (CD3 + ) cells per high power field (hpf, 400x). A total of 10 hpfs for each mouse, n = 4, were quantified. Results are represented as average of the mean for each mouse per indicated group.
RNA isolation and qPCR. RNA was harvested as previously described 30 . Briefly, tissues were homogenized in Trizol reagent followed by phenol-chloroform separation. Following DNAse treatment (Turbo DNA-free kit, Ambion), cDNA was generated using iScript cDNA synthesis kit (Bio-Rad). Real-time PCR was performed on BioRad CFX-384 Real-Time PCR Detection System using SsoAdvanced Universal SYBR Green (BioRad) and primers specific for indicated genes 30 . qPCR array. Transcripts from the monoassociation and 20wk infection samples were analyzed using the Qiagen RT2 Profiler PCR array Mouse Inflammatory Response & Autoimmunity PAMM077ZE. 400 ng of RNA was reverse-transcribed using the RT 2 First Strand Kit according to manufacturer protocol. Real-time PCR was performed using RT 2 SYBR Green qPCR Mastermix. Data was analyzed using the RT2 Profiler PCR Array DATA Analysis version 3.5 (Qiagen) normalized to 3 housekeeping genes. Data was presented as fold regulation using the 2^(-Delta Delta C T ) method and p values were calculated using Student's t-test of replicate 2^(-Delta C T ) values for each gene in control groups and K. pneumoniae-treated groups.
Statistical analysis. All data were analyzed using GraphPad Prism 6, version 6.0 h (GraphPad Software, Inc., La Jolla, CA). Cell culture data was evaluated using student t-test. Mouse experiments were analyzed using the Mann-Whitney nonparametric test, one-way ANOVA, and two-way ANOVA where appropriate and as indicated.