Increased Urinary Trimethylamine N-Oxide Following Cryptosporidium Infection and Protein Malnutrition Independent of Microbiome Effects.

Cryptosporidium infections have been associated with growth stunting, even in the absence of diarrhea. Having previously detailed the effects of protein deficiency on both microbiome and metabolome in this model, we now describe the specific gut microbial and biochemical effects of Cryptosporidium infection. Protein-deficient mice were infected with Cryptosporidium parvum oocysts for 6-13 days and compared with uninfected controls. Following infection, there was an increase in the urinary excretion of choline- and amino-acid-derived metabolites. Conversely, infection reduced the excretion of the microbial-host cometabolite (3-hydroxyphenyl)propionate-sulfate and disrupted metabolites involved in the tricarboxylic acid (TCA) cycle. Correlation analysis of microbial and biochemical profiles resulted in associations between various microbiota members and TCA cycle metabolites, as well as some microbial-specific degradation products. However, no correlation was observed between the majority of the infection-associated metabolites and the fecal bacteria, suggesting that these biochemical perturbations are independent of concurrent changes in the relative abundance of members of the microbiota. We conclude that cryptosporidial infection in protein-deficient mice can mimic some metabolic changes seen in malnourished children and may help elucidate our understanding of long-term metabolic consequences of early childhood enteric infections.

Malnutrition is a global health epidemic. Undernutrition results in growth and cognitive impairment, a host of inflammatory markers and an increased susceptibility to enteric infections [1]. The consequences of these infections, such as diarrhea and malabsorption from intestinal damage, further deprive the host of nutrients and lead to a vicious reciprocal cycle of undernutrition and enteric infections. The cumulative effects of this chronic cycle result in developmental shortfalls and impaired responses to vaccines [2][3][4][5][6]. Understanding how nutritional deficiencies prolong enteric infections and combine to the detriment of the host is essential to break this cycle.
Cryptosporidiosis, a protozoan infection, is prevalent in malnourished populations, areas with limited access to clean water and sanitation, as well as immunocompromised individuals [2,[7][8][9][10][11][12][13]. Cryptosporidium infections with or without overt diarrhea have been associated with significant growth shortfalls, and we have previously demonstrated that, as seen in children, malnutrition substantially worsens Cryptosporidium infection in a murine model [14][15][16]. Following 3 days of Cryptosporidium infection, significant growth impairment was observed in protein-malnourished mice, accompanied by increased intestinal injury and inflammation. Unlike nourished equivalents, protein-malnourished mice were unable to rapidly clear Cryptosporidium and continue to have detectable organisms in the stool for more than 7 days [17,18].
A previous study found strong effects of C. parvum on the murine microbiota, although the authors acknowledge that small treatment group size across multiple experiments may have amplified differences between uninfected and infected groups [19]. Studies in children and rodents have indicated that undernutrition results in both compositional and functional modulations in the resident gut microbiota [19]. In a study examining the effect of various diets on the mouse microbiota and metabolome, we observed that mice fed a low-protein diet retained a fecal microbiota more similar in composition to newly weaned mice than those on any other diet [20][21][22]. In the current study, we examined the effects of cryptosporidiosis on the fecal microbiome and urinary metabolic phenotypes of mice on the same protein-deficient diets. We disentangled the infection-specific metabolic alterations from those related to the diet. This approach reveals that several metabolic shifts after infection occur independent of changes in specific microbiota.

Animal Husbandry
Mice used in this study were male, 22 days old, C67BL/6 strain, and ordered from Jackson Laboratories (Bar Harbor, ME). Mice weighed approximately 11 g on arrival and were cohoused in groups up to 5 animals per cage. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Committee on the Ethics of Animal Experiments of the University of Virginia (protocol no. 3315).

Cryptosporidium Challenge
Mice were infected with a single inoculum of 5 × 10 6 purified C. parvum oocysts (Iowa Strain, Waterborne, Inc) by orogastric gavage as previously described [17]. Following challenge with C. parvum oocysts, mice were weighed daily. Feces and urine samples were collected every other day.

Lipocalin-2 and Myeloperoxidase Measurements
Lipocalin-2 (Lcn2) and myeloperoxidase (MPO) were measured in fecal samples collected 5 days after Cryptosporidium infection. Samples were homogenized in a radioimmunoprecipitation assay buffer with protease inhibitors, centrifuged at 8000 rpm for 10 minutes, and supernatant collected. Stool supernatant was assayed for total protein (BCA), Lcn2 (R&D Systems), and MPO (R&D Systems), according to manufacturer's instructions. Data were reported as pg of Lcn2 or MPO per µg total protein.

DNA Isolation
DNA was isolated from fecal pellets using the QIAamp DNA stool mini kit as previously described [24]. Stool DNA was analyzed for the C. parvum-specific 18S ribosomal ribonucleic acid (rRNA) gene to determine shedding of organism in the stool. Polymerase chain reaction conditions and primer sequences were as previously published [18]. The V3-V4 hypervariable regions of the 16S rRNA gene from fecal DNA samples were amplified using Illumina specific primers: (Illumina; forward: 5'-TCGTCGGCAGCGTCAGATGTGTATA-AGAG ACAGCCTACGGGNGGCWGCAG-3', reverse: 5'-GT-CTCGTGGGCTCGGAGATGTGTATAAGAGACAGG ACTAC-HVGGGTATCTAATCC-3'). For detailed methods of 16S rRNA gene sequencing and 1 H NMR spectroscopy-based metabolic profiling, please see Supplementary Methods.

RESULTS
We have previously shown significant weight loss, colonization, fecal shedding of the parasite, and inflammation in mice maintained on a protein-deficient diet infected with C. parvum [15,17,18,24]. A protein-malnourished state was necessary for the infection to establish, as C. parvum infections could not be sustained in nourished mice. Peak shedding of oocysts usually occurs approximately 5 days postinfection, which coincides with up to 20% weight loss in infected mice. Using this same model of protein malnutrition, we infected mice with 5 × 10 6 C. parvum oocysts. As in previous studies, infected mice had significant weight loss and shedding of oocysts in stool (Supplementary Figure 1).

Lcn2 Is Significantly Upregulated by Cryptosporidiosis
We have previously shown that protein deficiency alone is capable of disrupting intestinal epithelial tight junctions [18] and elevating fecal biomarkers of inflammation [23]. As shown in Figure 1, C. parvum infection with protein deficiency further increases myeloperoxidase and significantly increased Lcn2 compared with protein deficiency alone.

Urinary Metabolic Alterations Induced by C. parvum Infection
The urinary metabolic profiles of control and infected mice were compared at 6, 7, and 13 days postinfection using an orthogonal partial least squares-discriminant analysis (OPLS-DA) approach to assess the biochemical response to C. parvum exposure. Significant OPLS-DA models with good predictive abilities were obtained for this comparison at all time points. increased compared to the uninfected controls ( Figure 2). At 13 days of infection (OPLS-DA model: Q 2 Y = 0.71; P = .001) DMA and TMAO excretion remained higher in the infected mice compared with the control mice, but the excretion of the other choline-related metabolites was comparable. Products of the combinatorial metabolism between the gut microbiota and host were also increased 7 and 13 days postinfection, respectively. This included metabolites of the amino acids phenylalanine (phenylacetylglycine [PAG]), tryptophan (3-indoxyl-sulfate [3-IS]), and tyrosine (4-cresyl glucuronide [4-CG] and 4-cresyl-sulfate ). However, the excretion of 3-hydroxy-phenyl propionic acid sulfate (3-HPPA sulfate) and cinnamate and a cinnamate derivate, all gut microbial-host cometabolites of polyphenols and/or phenylalanine, was reduced 6 and/or 7 and 13 days postinfection, respectively. In addition, following 6 days of C. parvum infection (OPLS-DA model: Q 2 Y = 0.82; P = .001), several TCA cycle metabolites (succinate, citrate, isocitrate, malate, 2-oxoglutarate, and fumarate) were increased in urine compared with uninfected controls. However, some of these changes (succinate, citrate, and fumarate) were not observed at 7 days, and by 13 days postinfection most alterations were observed in the opposite direction (decreased), with the exception of cis-aconitate, which was excreted in greater amounts compared with uninfected controls. Similar trends were observed in the excretion of metabolites related to inflammation and oxidative stress (allantoin, pseudouridine, N-acetyl glycoproteins [NAG], and fucose). These metabolites were excreted in lower amounts 6 days postinfection but were  excreted in higher amounts 13 days postinfection compared with uninfected controls. No difference was observed in their excretion 7 days postinfection. Other notable changes in the urinary metabolic profiles included a decrease in sucrose, acetate, and alanine at 6 and/or 7 days postinfection, which was not observed after 13 days of infection.

Microbiota
We have recently shown that protein deficiency retards the aging-induced diversification of the fecal microbiome relative to a defined nourished diet [23]. Although cryptosporidiosis caused major changes in urinary metabolites, there was surprisingly little effect of C. parvum infection on the bacterial composition of the feces. Only 26 microbial sequence variants (SVs) were found to be differentially abundant between infected and uninfected groups at any time point, and these 26 SVs stratified infected and uninfected groups independent of time (Supplementary Figure 2). At all time points, infection slightly reduced α-diversity, consistent across a variety of α-diversity metrics ( Figure 3, Supplementary Figure 2). Considering all SVs that met filtering criteria (see Methods), β-diversity calculations and nonmetric multidimensional scaling led to delineation of infected and uninfected samples ( Figure 4).

Microbial-Metabolic Interactions in Response to Malnutrition
Unsupervised hierarchical clustering analysis (HCA) of the metabolites identified to contribute to the C. parvum infection at days 6, 7, and 13 is shown in Figure 5 (upper-right quadrant).
The interpretation that 7 days postinfection represents a transition in the metabolic shift from 6 to 13 days postinfection is more evident from the heatmap. C. parvum-infected samples at 6 days appear on the left, whereas all 13-day postinfected samples are located on the right, with some 7-day infected samples clustering with either 6-or 13-day infected samples. The clustergram, constructed using relative abundances of the bacteria identified to be significantly different between any age-matched infected and uninfected mice, resulted in distinct clusters of uninfected and infected mice ( Figure 5, lower-left quadrant). In addition, there seems to be a transition in the SVs from 6 to 13 days, with some 7-day samples clustering with either 6-or 13-day infected samples. The correlation heatmap ( Figure 5, upper-left quadrant) identified significant correlations between SVs and metabolites. The most notable metabolite-bacterial correlations included those between TCA cycle metabolites and 4 SVs identified as members of the genera Enterococcus, Blautia, Anaeroplasma, and Anaerotruncus. Additionally, an unknown bacterial group (25_NA) was strongly correlated with 3 products of microbial-specific metabolism (cinnamate, a cinnamate derivative, and 3-HPPA sulfate). The bacterial group 25_NA was not detected in any infected samples, and cinnamate and the cinnamate derivative are present at very low amounts in infected samples.

DISCUSSION
These findings of metabolites and biomarkers seen in experimentally controlled dietary (protein)-deficient mice with and without cryptosporidial infection with additional microbiome studies extend studies of cryptosporidial infection and their potential impact in field studies. In particular, we identify that modulations in choline, protein, and tryptophan metabolism seen with protein deficiency alone are magnified by cryptosporidial infection. Furthermore, these choline pathway metabolites, such as TMAO, may well have long-term risk     consequences, as they have been strongly associated with cardiovascular disease risk in humans in later life [25]. Indeed, we and others have shown that early-life malnutrition, enteric infections, diarrhea, or enteropathy may have such long-term consequences as metabolic syndrome and cardiovascular disease, as well as stunting and cognitive effects [26][27][28]. Thus, these are the first specific metabolic findings with a common enteric infection, cryptosporidiosis, in a protein-deficient state that provide experimental evidence to support such potential consequences of the common dietary and enteric infections seen in young children in resource-limited settings.
Regarding fecal biomarkers that are altered in malnourished children and in our murine cryptosporidial infections, we find significant increases in the epithelial and inflammatory marker, Lcn2, as well as the inflammatory marker MPO. Consistently, cryptosporidiosis in malnourished children living in endemic areas (Haiti and Northeast Brazil) has also been associated with evidence of mild to moderate intestinal or systemic inflammation, including elevated fecal lactoferrin, and fecal or systemic interleukin-8 and tumor necrosis factor-α, more than that seen in healthy adult volunteers [29][30][31]. Similarly, we see effects on intestinal and inflammatory biomarkers in our protein-deficient murine model of experimental cryptosporidial infection that exceed those seen in protein deficiency alone, again suggesting a compounding of modest diet-induced changes [23] with added cryptosporidial infection, thus potentially modeling (and dissecting) both diet and infection components of environmental enteropathy. Also of interest is the greater increase in Lcn2 relative to MPO in our model. Because Lcn2 is seen in epithelial cells as well as in inflammatory cells, while MPO is more specifically associated with inflammation, these effects likely reflect a more direct disruption of epithelial-cell integrity by cryptosporidial infection, opposite to the greater relative increases in MPO we see in our murine models of enteroaggregative Escherichia coli or Campylobacter infections (unpublished data, submitted).
As expected after the introduction of an enteric pathogen, cryptosporidiosis impacted the composition of the microbiota. Infected mice were most notably enriched with SVs from Clostridium XVIII and had reduced levels of several SVs that could not be assigned identity at the genus level. Given the limited taxonomic resolution of 16S rRNA gene sequencing analysis, as well as the small region we used (180 base pairs), the cause of this shift requires further investigation. Overall, it is unclear whether modulation of the microbiota is a result of interaction between C. parvum and the microbiota, or an indirect effect in which cryptosporidiosis alters the intestinal environment, thereby altering the microbiota. Consistent with our observations the study of murine C. parvum infection by Ras et al. [19] found clear differences in the fecal microbial profiles between infected and uninfected mice using beta diversity analyses. In their study, 4 Bacteroidetes operational taxonomic units (OTUs) were identified as enriched during C. parvum oocyst shedding relative to uninfected controls, while 4 other Bacteroidetes OTUs were depleted. These results contrast with the shifts we identified, which were almost exclusively within the Firmicutes phylum. However, Ras and colleagues performed 3 other mouse experiments with slightly altered infection protocols, and a meta-analysis of 16S rRNA sequencing from all 4 experiments indicated strong clustering by experiment in beta diversity analysis. The inconsistencies in specific taxonomic shifts across these experiments imply that C. parvum infection does not induce a specific alteration of the microbiota, or that the effect cannot be identified using taxonomic analyses.
A previous murine study found that cryptosporidiosis induced by C. parvum was enhanced in germ-free immunodeficient mice relative to immunodeficient mice harboring the altered Schaedler flora, a model microbiota composed of 8 bacterial species [32]. This finding indicates that C. parvum colonization is affected by interactions between the intestinal microbiota and the parasite or host, even with a reduced microbiota. C. parvum growth is also inhibited to a greater extent in mature mice compared to freshly weaned mice, a finding originally thought to be due to increasing diversity of the gut microbiota with age [33]. The protein-deficient model in our study results in lower alpha diversity than age-matched mice on other diets, reinforcing that the result shown in Harp et al. [33] may be due to age-related differences in microbiota composition. However, colonization resistance conferred by the altered Schaedler flora rules out diversity as a determinant of colonization resistance; instead, indicating that specific functionalities carried out by 8 or fewer bacterial species are sufficient. The potential for individual bacterial species to influence colonization of C. parvum suggests that direct interaction between C. parvum and the intestinal microbiota is likely during cryptosporidiosis. Studies using mouse models with varying degrees of colonization resistance, coupled with metagenomic and/or metatranscriptomic analyses, are required to further characterize interactions between C. parvum and the microbiota.
At 6 days postinfection, we observed an increase in the urinary excretion of TCA cycle intermediates, with the exception of cis-aconitate. This initial stimulation of the TCA cycle may reflect the high energy demands required to fight the infectious process, resulting from the recruitment of inflammatory cells and elevated proliferation rates among lymphocyte populations [34]. However, C. parvum lacks both the TCA cycle and cytochrome-based respiratory chain, thus relying mainly on glycolysis for adenosine triphosphate production. This could lead to a depletion of the TCA cycle intermediates as the infection progresses and explains the lower levels of TCA cycle intermediates in the urine of C. parvum-infected mice at 13 days compared with uninfected controls. However, we did not observe any difference in glycolysis metabolites, such as pyruvate, lactate, or acetate.
Thus, the suppression of TCA cycle may not be due to a stimulation of glycolysis. Alternatively, it could be related to the higher excretion of cis-aconitate, a metabolite produced from the dehydration of citrate by aconitase, an iron-sulfur protein. Successful iron acquisition from host cells is almost entirely necessary for any protozoan pathogen to survive, grow, and replicate. C. parvum lacks a sulfur assimilation pathway, which is expected to be substituted from the host cells. This can lead to an aconitase deficiency and depletion of TCA cycle intermediates. In fact, aconitase knockdown flies contained high levels of cis-aconitate, while most other TCA cycle metabolites, including 2-oxoglutarate, succinate, fumarate, and malate, as well as metabolites of glycolysis, were significantly decreased [35]. In addition, several studies have shown that aconitases are inhibited by iron deficiency or oxidative stress [36].
Decreased aconitase activity may also lead to decreased production of nicotinamide adenine dinucleotide phosphate, an important defense against oxidative stress [36]. Thus, oxidative stress seems to play an important role in the development of C. parvum infection in mice [37].
Cryptosporidiosis is also associated with a persistent systemic inflammatory response. Consistently, the unsupervised HCA showed a cluster comprising NAG, fucose, pseudouridine, and allantoin. These metabolites are related to inflammation and oxidative stress. Thus, allantoin appears to be a useful biomarker of gut inflammation, and both allantoin and pseudouridine have been associated with inflammation and oxidative stress. Elevated levels of NAG signals from acute-phase reactive proteins have been associated with inflammatory conditions [38]. Finally, fucose is a sugar that can be found in the glycosylated part of the glycoproteins and is also found abundantly in the mammalian gut. Systemic invasion of pathogens or microbial products in the circulation has been shown to induce fucosylation of the epithelium in the small intestine, where it seems to have a protective role in both gut-centered and systemic infection and inflammation [39].
Interestingly, choline and phosphocholine (PC) were excreted in higher amounts in infected mice. PC was identified as the predominant lipid in C. parvum-infected bovine kidney cells and, interestingly, C. parvum has been suggested to be capable of sequestering phospholipids from the host to incorporate them into the membranes [40]. Thus, the greater abundance of choline and PC can be related to the degradation of host membranes by the parasite. In addition, choline-derived metabolites (DMA, TMA, and TMAO) were also excreted in higher amounts by infected mice.
Our results also suggested a change in microbial activity, in some cases coupled to the relative abundance of specific SVs. Gut microbial-derived metabolites such as PAG, 4-CG, 4-CS, 3-IS, DMA, TMA, and TMAO were excreted in higher amounts, whereas 3-HPPA sulfate, cinnamate, and a cinnamate derivate were excreted in lower amounts by infected mice. This was reinforced by the fact that TMA, TMAO, choline, 4-CS, 4-CG, and 3-IS cluster together, as did 3-HPPA sulfate, cinnamate, and the cinnamate derivate. PAG, 4-CG and 4-CS, and 3-IS are derived from phenylalanine, tyrosine, and tryptophan, respectively. On the other hand, DMA, TMA, and TMAO are choline-derived metabolites, while 3-HPPA sulfate and cinnamate result from polyphenol metabolism. Similar to our findings, elevations in TMA, PAG, 4-CG and reductions in hippurate and/or 3-HPPA sulfate have been previously identified in hosts infected with helminths (such as Schistosoma mansoni [41,42] and S. japonicum [43][44][45][46]) and intestinal nematodes [47]. Similar results have also been observed in protozoan parasites such as Plasmodium berghei [48] and Trypanosoma brucei brucei [49].
The correlation between these metabolites and the SVs confirmed their microbial origin. For example, 4-CG is derived from microbial degradation of tyrosine by species such as Clostridium difficile, C. scatalogens, and certain Lactobacillus species. In accordance, 4-CG correlated with 59_NA and 123_ NA, both clostridia SVs. However, choline-derived metabolites showed no significant correlations with the microbiome. Alternatively, these changes could be due to renal damage. For example, DMA and TMAO have been associated with renal cortex damage in methanol intoxication patients [50]. An increase in DMA and TMAO was observed at 13 days, when TCA cycle intermediates were depleted. These changes could be the direct products of parasite metabolism. Future studies should also address whether altered renal function might contribute to these findings as well.
In conclusion, these metabolomic, biomarker, and microbiome studies of cryptosporidial infections in protein-deficient mice help dissect common major dietary and pathogen elements in enteropathy commonly seen in children in developing areas. They also elucidate metabolic and inflammatory pathways that have direct relevance to growth failure and even potential long-term metabolic disease consequences. They also provide a powerful experimental model to help elucidate key pathways involved, and also to assess innovative interventions to ameliorate environmental enteropathy and its triple burden of potentially lasting consequences in impoverished children.

Supplementary Data
Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes
Financial support. This work was supported by the Bill and Melinda Gates Foundation Biomarker Grant no. OPP1066140 entitled, "Novel metabonomic biomarkers of gut function and health: Modeling enteropathy (EE) and field validation" (to R. L. G.), and by the following National Institutes of Health (NIH) grants: National Institute of Allergy and Infectious Diseases (NIAID) Centers for Excellence for Translational Research U19 AI109776 (to R.L.G.; Myron Levine University of Maryland, Principal Investigator), NIH Grant RO1 GM108501 (to J.A.P.), and NIH grant T32LM012416. Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.