Inflammation-associated alterations to the intestinal microbiota reduce colonization resistance against non-typhoidal Salmonella during concurrent malaria parasite infection

Childhood malaria is a risk factor for disseminated infections with non-typhoidal Salmonella (NTS) in sub-Saharan Africa. While hemolytic anemia and an altered cytokine environment have been implicated in increased susceptibility to NTS, it is not known whether malaria affects resistance to intestinal colonization with NTS. To address this question, we utilized a murine model of co-infection. Infection of mice with Plasmodium yoelii elicited infiltration of inflammatory macrophages and T cells into the intestinal mucosa and increased expression of inflammatory cytokines. These mucosal responses were also observed in germ-free mice, showing that they are independent of the resident microbiota. Remarkably, P. yoelii infection reduced colonization resistance of mice against S. enterica serotype Typhimurium. Further, 16S rRNA sequence analysis of the intestinal microbiota revealed marked changes in the community structure. Shifts in the microbiota increased susceptibility to intestinal colonization by S. Typhimurium, as demonstrated by microbiota reconstitution of germ-free mice. These results show that P. yoelii infection, via alterations to the microbial community in the intestine, decreases resistance to intestinal colonization with NTS. Further they raise the possibility that decreased colonization resistance may synergize with effects of malaria on systemic immunity to increase susceptibility to disseminated NTS infections.

malaria parasite infection on the intestine, C57BL/6 mice were inoculated with P. yoelii. Mice developed parasitemia, measured as the percentage of red blood cells harboring parasite, that peaked at maximal levels between days 10 and 15 after inoculation. We selected this phase of maximal parasitemia to interrogate effects of malaria on the intestine (Fig. 1A). During this time, P. yoelii-infected blood cells could be observed in the intestinal microvasculature, with evidence of sequestration on the vascular endothelium (Fig. 1B). Blinded histopathology analysis of the large intestinal wall at the cecum revealed mild but significant changes, including edema of the lamina propria, focal loss of goblet cells, hyperplasia of undifferentiated enterocytes, and focal infiltration of mononuclear cells into the lamina propria, but no evidence of inflammatory cell exudation into the intestinal lumen (Fig. 1C).
Analysis of the cellular infiltrates by flow cytometry revealed that they consisted primarily of T cells (CD3+ ), as well as CD11b+ and CD11c+ myeloid cells ( Fig. 1D and Fig. S1). Approximately 30% of the infiltrating CD11b+ cells exhibited an inflammatory phenotype, as evidenced by expression of Ly6C (Fig. 1E). Together, these results show that acute malaria parasite infection is associated with inflammatory changes in the wall of the intestine.
Inflammatory changes result from parasite infection, and do not require the endogenous microbiota. We next interrogated whether the mucosal inflammation observed in the P. yoelii-infected mice resulted from the parasite infection itself, or rather from resulting penetration of the endogenous intestinal microbiota into the tissue. To address this question, we compared intestinal responses of conventionally reared or germ-free C57BL/6 mice during maximal parasitemia with P. yoelii ( Fig. 2A). Both groups of mice exhibited similar levels of parasitemia at d15 after inoculation, so this time point was used for comparison ( Fig. 2A). Histopathology scoring revealed a comparable severity of histologic changes in P. yoelii-infected germ-free mice compared to conventional mice ( Fig. 2B and Fig. 1C). Based on our observation ( Fig. 1) that inflammatory infiltrates in the intestine of conventional mice contained T cells and macrophages, we analyzed mucosal expression of S100a8 and S100a9, produced by inflammatory macrophages during malaria 20 , as well of IL-10 and interferon gamma, produced by T cells 21 (Fig. 2B). In conventional mice, expression of S100a8, S100a9, Il10 and Ifng was elevated at d10 and d15 after P. yoelii infection. A similar pattern of induction was observed in germ-free mice evaluated at d15. There was a nonsignificant trend (P > 0.05) for lower induction of S100a8 and S100a9 in the germ-free mice. These results suggest that while an intact intestinal microbiota may contribute to inflammatory changes in the gut mucosa during P. yoelii infection, it is not required for this effect, rather it is the malaria parasite infection driving this inflammatory response.

The composition of the intestinal microbiota is altered during malaria parasite infection.
To determine whether malaria parasite infection impacts the resident microbiota, fecal pellets were collected from two groups of co-housed C57BL/6 mice before inoculation with P. yoelii and at days 10, 15 and 30 days post infection. Illumina MiSeq analysis of amplicons from the 16S rRNA locus in fecal DNA extracts revealed significant alterations in the colonic microbiota (Fig. 3). At the phylum level, a decreased abundance of Firmicutes and a relative increase in the abundance of Bacteroidetes were observed at d10 (Fig. 3A). These changes were not simply the result of fluctuation in the resident microbiota over time or of husbandry-related effects, since a group of mock-infected mice from the same colony exhibited a stable microbiota composition over time (Fig. S2). At the genus level, acute malaria parasite infection at d10 was associated with an increase in the relative abundance of unclassified members of the Rikenellaceae (P = 0.009), Ruminococcaceae (P = 0.007), and Bacteroidales (P = 0.024), as well as of Turicibacter (P = 0.001). A decrease in Ruminococcus was also noted (P = 0.041; Fig. 3B and Table 1). Since the P. yoelii-infected mice were co-housed, we cannot formally exclude a contribution of coprophagy to the altered fecal microbiota. However one co-housed mouse in this group (not shown) was not infected with P. yoelii and did not exhibit these alterations in the fecal microbiota, suggesting that coprophagy alone is not sufficient to alter the microbiota in a conventionally-reared mouse. Overall, as P. yoelii infection progressed, the diversity of the fecal microbiota decreased by d10, with a gradual recovery by d30 after infection (Fig. 3C). At day 30, after resolution of infection, the composition of the microbiota most closely resembled the composition prior to infection, as shown by principal component analysis (Fig. 3D), suggesting that the effect of malaria parasite infection on microbial communities in the large intestine is transient.
Malaria parasite infection lowers the implantation dose for S. Typhimurium in mice. The finding that P. yoelii infection altered the intestinal microbiota raised the possibility that these changes  Fig. S1. Significance for differences between experimental groups was determined using Student's t test on logarithmically transformed data. Mice were housed in groups of 4-5 per cage. could affect susceptibility of mice to infection with NTS. To address this question, we determined the dose at which 50% of mice would become infected with S. Typhimurium (implantation dose 50 or ID 50 ) at 1 day after infection, according to the method of Reed and Muench 22 . In control mice, the ID 50 for S. Typhimurium IR715 was 1.1 × 10 4 CFU, 34-fold higher than at the peak of P. yoelii infection, where this value was reduced to 3.2 × 10 2 CFU (Table 2 and Fig. S3A). Further, P. yoelii-infected mice inoculated with S. Typhimurium at varying doses were colonized at significantly higher levels, as assessed by determining CFU in the feces (Fig. 4A). By 4 days after S. Typhimurium infection, colonization levels were similar in both groups (data not shown), likely because S. Typhimurium infection elicits intestinal inflammation in the control mice, a factor that promotes its outgrowth in the intestinal lumen [23][24][25][26] . However, elevated intestinal colonization of S. Typhimurium at 1 day post inoculation in the P. yoelii-infected mice was independent of the ability of S. Typhimurium to elicit a mucosal inflammatory response, since an invA spiB mutant (SPN487), defective in the SPI-1 and SPI-2 encoded type III secretion systems required for mucosal invasion and inflammation 27,28 , was also recovered in higher numbers from P. yoelii-infected  Table S1. Each bar represents an individual mouse. Images were acquired with 10× (left panels) and 40× objectives (right panels). Arrow indicates mononuclear infiltration. (C) Expression analysis of inflammatory markers by qRT-PCR. Transcript levels of calprotectin (subunits S100a8 and S100a9), interferon gamma (Ifng) and interleukin-10 (Il10) were determined in cecal tissue from Conv or GF mice sacrificed at 10, 12 or 15 d after P. yoelii inoculation. Data shown as fold-change over mock-treated Conv mice (indicated with dashed line at 1) with mean + SEM for (Conv, n = 5 − 11; GF, n = 3). Asterisk (*) indicates significance (P < 0.05) when compared to mock-treated mice as determined by Student's t test on logarithmically transformed data, (ns) indicates no significance (P > 0.05). Groups of mice were co-housed.
Scientific RepoRts | 5:14603 | DOi: 10.1038/srep14603 mice ( Fig. 4B and Fig. S3B). Further, the human commensal strain Escherichia coli HS 29 , which does not cause intestinal inflammation, also colonized the intestine of P. yoelii-infected mice at higher levels than in control mice (Fig. 4C), and this elevated colonization was maintained for several days after E. coli inoculation (data not shown). Of note, we did not observe an effect of P. yoelii infection on colonization of E. coli in our 16S microbiota analysis (Fig. 3), most likely because our mice (C57BL/6J) were not consistently colonized with detectable levels of E. coli. However, mice inoculated concurrently with E. coli and P. yoelii exhibited higher colonization of E. coli compared to control mice 14 days later (data not shown), implying that if E. coli is present at the outset of infection, its outgrowth is promoted during malaria. Together, these results suggest that changes to the intestinal milieu caused by P. yoelii infection promote colonization of the intestine with both S. Typhimurium and E. coli.
Alterations to the microbiota induced by malaria parasite infection promote colonization with S. Typhimurium. To determine the significance of the altered intestinal microbiota for increased S. Typhimurium colonization during malaria, we performed a microbiota transfer experiment. Cecal contents were isolated under anaerobic conditions from three control mice and three mice acutely infected with P. yoelii at d10 post infection (Fig. S3C), and the contents from each single mouse were transferred to an individual germ-free Swiss-Webster recipient via oral gavage. After allowing 6 days for the microbiota to become established, mice were inoculated via gavage with S. Typhimurium. One day later, S. Typhimurium colonization was measured via fecal shedding. Figure 4D shows that recipients of the microbiota transplant from P. yoelii-infected mice were colonized at a tenfold higher level with S. Typhimurium than recipients of the microbiota from control mice. These results suggest that dysbiosis induced by malaria parasite infection lowers colonization resistance of mice against S. Typhimurium.

Discussion
Studies in murine models have shown that multiple responses to malaria parasite infection conspire to increase susceptibility to disseminated infection. Malaria-induced hemolysis impacts maturation of neutrophils, which play a critical role in containing spread of extracellular bacteria 10 . Further, malaria-induced IL-10, that is beneficial in the context of dampening parasite-induced inflammation, has a detrimental effect on control of intracellular S. Typhimurium replication within hepatic macrophages 11 . As a consequence, once bacteria have disseminated from the gut, control of systemic infection is compromised.  Further, disruption of intestinal barrier function and suppression of NTS-induced neutrophil recruitment to the mucosa may facilitate disseminated infection 12,14 . This study identifies a new factor that suppresses resistance to initial colonization of the intestine by S. Typhimurium, namely alterations to the community structure of the intestinal microbiota, which outnumbers the body's own cells by an order of magnitude (Ref 30). As a result, malaria reduces colonization resistance against S. Typhimurium-in our model of concurrent infection, the effective dose of bacteria needed to establish intestinal infection was decreased by 97%. Loss of colonization resistance during malaria did not involve epithelial invasion or induction of inflammation by S. Typhimurium, as it was independent of the SPI-1 and SPI-2 Type III secretion systems that are needed for invasion and induction of intestinal inflammation (Fig. 4) 31 . Further, a non-invasive commensal E. coli strain also exhibited enhanced colonization in our model, suggesting that perturbation of the microbial community by malaria opens an ecologic niche that can be occupied by either S. Typhimurium or E. coli. This altered environment was associated with mononuclear infiltration of the intestinal mucosa ( Fig. 1 and Fig. 2), suggesting the possibility that inflammatory changes may drive these changes to the microbiota. However, malaria-induced inflammation did not appear to be necessary for loss of colonization resistance, because reduced colonization resistance could be transferred to germ-free mice independently of malaria, by transfer of the cecal microbiota (Fig. 4D). Of note, based on the different types of inflammatory responses observed in the intestinal mucosa, the mechanism by (D) Susceptibility of germ-free mice reconstituted with colonic microbiota from P. yoelii-infected or control mice to colonization with S. Typhimurium. Each reconstituted mouse was housed individually for the duration of the experiment. Each symbol represents an individual mouse, with horizontal bars representing the geometric mean. Dashed lines indicate limit of detection. Significance of differences between experimental groups was determined using a Student's t test on logarithmically transformed data. which malaria alters the endogenous microbiota is likely to be different from the mechanism by which S. Typhimurium promotes its own outgrowth via inflammation. In our study, we observed an infiltration of T cells and mononuclear phagocytes in P. yoelii-infected mice (Fig. 1). In addition, we observed an increase in Fcε RI-positive cells, which is consistent with our previous report of an increase in mucosal mast cells in this model ( Fig. S1 and 14 ). In contrast, in the murine colitis model used to model enteric pathology of S. Typhimurium infection, a massive exudation of neutrophils into the mucosa and the intestinal lumen results in production of oxygen and nitrogen radicals that alter the environment and promote outgrowth of S. Typhimurium in the gut lumen 25 .
P. yoelii infection resulted in a decrease in the complexity of the cecal microbiota, as well as a decrease in the abundance of Firmicutes. Interestingly, members of this phylum have been shown to be decreased after treatment with antibiotics including cefoperazone 32 , shown to reduce colonization resistance against Clostridium difficile 32 and streptomycin, which enhances colonization with S. Typhimurium 33 . Further, a decrease in members of the Firmicutes has been observed in patients with inflammatory bowel disease, a condition that is associated with increased colonization of E. coli 26,[34][35][36][37][38] . However, whether shared mechanisms underlie outgrowth of S. Typhimurium and E. coli in each of these conditions is unknown, since the mechanisms linking alterations in the microbiota with reduced colonization resistance are incompletely understood.
Taken together, the results of this study suggest that malaria, via alterations to the intestinal environment, shifts the community structure of the gut microbiota to provide a benefit to colonizing S. Typhimurium and E. coli.

Methods
Plasmodium yoelii nigeriensis (P. yoelii). Parasite stocks were obtained from the Malaria Research and Reference Reagent Resource and the species and strain identities were confirmed by DNA sequencing of merozoite surface protein-1 (MSP-1) 14 . Parasite stocks were prepared by passage in CD-1 mice. For experiments, mice were inoculated intraperitoneally (i.p.) on day 0 with blood containing approximately 4 × 10 7 infected red blood cells (iRBCs). Mock-treated controls were injected with an equal volume of blood from uninfected CD-1 mice.

Animal experiments.
All experiments were performed in accordance with guidelines and regulations as outlined and approved by the UC Davis or Ohio State University Institutional Animal Care and Use Committees (IACUC). Specific pathogen free (SPF) mice: 6-8 week-old female C57BL/6J mice were purchased from the Jackson Laboratory (Bar Harbor, Maine) and maintained under SPF conditions. Germ-free (GF) mice: GF C57BL/6 and Swiss Webster mice were bred inside germ-free isolators. Experimentation in GF mice was performed in an independent GF isolator and for fecal microbiota reconstitution; mice were transferred to a biosafety cabinet for inoculation and maintained in sterile cages for the duration of the experiment. After reconstitution, GF mice were caged individually.
Microbial readouts of colonization. Parasite infection was monitored by blood collection from tail snips. Parasitemia was determined by counting the percentage of Plasmodium yoelii iRBCs on thin blood smears stained with Giemsa (Acros Organics). For quantification of S. Typhimurium or E. coli, fecal pellets, collected 1 day after intragastric inoculation, were homogenized and serial dilutions spread on LB agar plates containing appropriate selective antibiotics.
Histopathology. Histological samples were collected at the time of necropsy. 5 μ m sections were cut from formalin fixed paraffin embedded tissues and stained with hematoxylin and eosin or Giemsa by the UC Davis Veterinary Pathology Laboratory. A veterinary pathologist (MXB) performed histopathology scoring in a blinded fashion, according to scoring criteria detailed in Table S1.
RNA extraction, reverse transcription-PCR (RT-PCR), and real-time PCR. Animal tissues were frozen in liquid nitrogen at necropsy and stored at − 80°C. RNA was extracted from tissue as described previously 42 using Tri-Reagent (Molecular Research Center) according to the manufacturer's instructions. RNA was treated with DNAseI (Ambion) to remove genomic DNA contamination. For a quantitative analysis of mRNA levels, 1 μ g of total RNA from each sample was reverse transcribed in a 50-μ l volume (TaqMan reverse transcription [RT] reagent; Applied Biosystems), and 4 μ l of cDNA was used for each real-time reaction. RT-PCR was performed using the primers listed in Table S2, SYBR green (Applied Biosystems) and ViiA 7 Real-Time PCR System (Applied Biosystems). Data was analyzed by using the comparative threshold cycle (C T ) method (Applied Biosystems). Target gene transcription of each sample was normalized to the respective levels of beta-Actin mRNA and represented as fold change over gene expression in control animals. Microbiota Sequencing. DNA was extracted from homogenized stool samples using the protocols and reagents specified in the PowerFecal ™ DNA Isolation Kit (MoBio Laboratories, Inc.). To facilitate efficient assemblies and longer accurate reads, paired end (PE) libraries were constructed. Bacterial DNA was amplified by PCR enrichment of 16S rRNA encoding sequences from each sample using primers 515F and 806R that flank the V3-V4 hypervariable region and were modified by adding a unique set of 8 oligonucleotide barcodes for purposes of multiplexing.
The resulting PE 16S rRNA amplicons were purified and quantified on an Invitrogen Qubit system. Libraries were normalized and quality assessed on an Agilent Bioanalyzer prior to sequencing with an Illumina MiSeq system. As quality control, sequences containing uncalled bases, incorrect primer sequence, or runs of ≥ 12 identical nucleotides were removed from the data.
Phylogenetic analysis of the 16S rRNA sequences was accomplished using customized Linux-based command scripts for trimming, demultiplexing, and quality filtering the raw PE sequence data. Using the QIIME 43 open source software package, the demultiplexed sequences were aligned, clustered, and operational taxonomic units (OTUs) were determined utilizing the Greengenes reference collection (greengenes.lbl.gov). Principal Component Analysis was performed using METAGENassist 44 . Alpha and beta diversity were evaluated using QIIME and the Megan 45 open source software package. Student's T-tests were used to identify taxa that displayed statistically significant differences between experimental groups and controls. 16S rRNA sequences are deposited in the Sequence Read Archive (Bioproject PRJNA287262) at the National Center of Biotechnology Information (NCBI).
Microbiota reconstitution of Germ-free Mice. Control or parasite-infected C57BL6/J mice at 10 days post P. yoelii inoculation were euthanized and ceca removed aseptically with cuts 2 cm above and below the cecum to minimize oxygen exposure. Ceca were then transferred to an anaerobic chamber (Bactron I Anerobic Chamber; Sheldon Manufacturing, Cornelius) for processing. The cecal contents from each donor mouse were collected and suspended in 2 ml pre-reduced PBS. Each recipient germ-free Swiss Webster mouse was orally inoculated with 0.2 ml of cecal suspension from one donor mouse and housed in an individual cage for 6 days to allow for microbiota reconstitution.
Statistical analysis. The statistical significance of differences between groups was determined by a Student's t test on data transformed to a logarithmic scale. A P value of 0.05 or less was considered to be significant. All data were analyzed using two-tailed tests.