Host sex and genotype modify the gut microbiome response to helminth infection

The microbial community can be altered by direct/indirect interactions with parasites infecting host. Direct interactions can arise from physical/chemical contact with the parasite. Indirect interactions can involve parasite-induced changes in host immunity. If so, this would represent a case of genetic polymorphism in one species controlling an ecological interaction between other species. Here, we report a test of this expectation: we experimentally exposed Gasterosteus aculeatus to their naturally co-evolved parasite, Schistocephalus solidus. The host microbiome differed in response to parasite exposure, and between infected and uninfected fish. The microbial response to infection differed between host sexes, and also varied between variants at autosomal quantitative trait loci (QTL). These results indicate that host genotype regulates the indirect effect of infection on a vertebrate gut microbiome. Our results also raise the possibility that this sex-bias may be related to sex-specific microbial responses to the presence (or, absence) of helminthes. Therefore, helminth-based therapeutics as possible treatments for inflammatory bowel diseases might need to take account of these interactions, potentially requiring therapies tailored to host sex or genotype.


5 1
PCoA axis 5 exhibited three QTLs that narrowly exceeded our stringent threshold for 2 5 2 significance (Additional file: Figure S7a-c). We had a stronger signal when we fused the many 2 5 3 PCoA axes into a single metric using linear discriminant analysis (trained to distinguish the two 2 5 4 backcrosses then applied to all samples). Using this first LDA axis we detected a single well-2 5 5 supported QTL on Chr9 (Additional file: Figure S7d-e). Note that because QTL mapping was 2 5 6 done within each cross, using LDA to define an axis that distinguishes between crosses is not 2 5 7 tautological. Chromosome 9 does not have any noteworthy effect on cestode infection success 2 5 8 (infection QTL described in Weber et al, in preparation).

5 9
The lack of strong QTL for microbial ordination metrics led us to hypothesize that host 2 6 0 control of the whole microbial community is highly polygenic. If host genetic variation acts on 2 6 1 particular microbe taxa, it might act only weakly on PCoA scores and be correspondingly hard to 2 6 2 map. So we next mapped microbial Orders separately, revealing numerous taxon-specific QTL.

6 3
To illustrate, Fusobacterales exhibit two strong autosomal QTL, plus an association with the X genome are presented in Fig. 6. This summary reveals genomic 'hotspots' for QTL affecting the 2 6 7 microbiota, on Chr1, Chr2, and Chr3. Many of these microbial Orders also mapped to the sex 2 6 8 9 chromosome (Chr19), consistent with the common main effect of sex. Similar results were 2 6 9 obtained for Family level QTL mapping.

7 0
We next tested whether these QTL are contingent on cestode presence or absence, as 2 7 1 expected from the interactive effect of infection and cross type, described above. We mapped 2 7 2 QTL separately for infected fish and for uninfected fish, and then looked for loci with QTL in 2 7 3 only one of these groups. We found numerous host loci that are only associated with microbial 2 7 4 variation in uninfected fish, whereas the same locus is unrelated to the microbiota among 2 7 5 infected fish (e.g., Additional file: Figure S9a-f). In fewer instances, we identified host QTL for 2 7 6 microbiota in infected fish only (Additional file: Figure S9g-  The interaction between hosts, parasites, and gut microbes represents a rich opportunity for depend on both the sex of the host, and the host's autosomal genotype. By examining F2 hybrids 2 9 0 between two recently-diverged host populations, we were able to identify autosomal loci (QTL) 2 9 1 that contribute to variation in microbiota composition, as well as variation in microbiota response 2 9 2 to infection. The implication is that genetic variation within one species (the host) alters the 2 9 3 ecological effect of another species (the cestode parasite) on a third party (the microbial 2 9 4 community).

9 5
The inconvenient implication of this finding is that we may not readily generalize helminth-2 9 6 microbe effects beyond the genotypes (and sex) we study. This unwelcome news is tempered by 2 9 7 the opportunity it presents: host genetic variation can therefore help us identify the mechanisms 2 9 8 by which helminths alter the gut microbiota (or, vice versa). Doing so may allow us to develop a

1 3
Host autosomal genotype also affected gut microbial composition, and modified microbes' [30, 31]. What makes the above results novel is that this host genetic variation also alters the 3 1 7 microbiome's response to a third (parasitic) species. We coarsely localize these autosomal 3 1 8 effects to a modest number of chromosomal loci (QTL). These QTL do not contain MHCIIb, 3 1 9 which has previously been associated with natural variation in stickleback gut microbiota [29].

2 0
At present we lack the resolution fine-map down to specific candidate genes, but some intriguing

2 7
The QTL on Chr3 contains cadherins, which play a key role in intestinal homeostasis and barrier 3 2 8 function [37]. Substantial future work is needed to fine-map candidate genes and experimentally 3 2 9 validate their suspected effects on the gut microbiota, and on the cestode-microbiome interaction.

9
In the specific case examined here, helminth effects on the microbiota are especially likely 3 4 0 to involve indirect effects via host immunity, rather than a direct microbe-helminth interaction.

4 7
The above results specifically apply to variation among cestode-exposed stickleback. But, 3 4 8 our Experiment 1 suggests that stickleback gut microbiota depended most strongly on parasite lasting ripples through the microbial community, then the study of the wild microbiome will be 3 5 3 still more difficult. We rarely if ever know our study animals' history of unsuccessful infections, 3 5 4 so these ripples might generate substantial and untraceable variation among wild individuals. Yet

5 5
infection success also alters the gut microbiome, as we then revealed in the follow-up 3 5 6 Experiment 2.

5 7
At present we do not know the fitness consequences of the altered microbiota that we 3 5 8 document here. Some studies have reported detrimental helminth effects on the microbiota.

5 9
Experimental T. muris infection in mice alters the microbiota composition leading to reduced 3 6 0 availability of microbial metabolomics products needed by the host (vitamin D2/D3 derivatives, suggest that these positive or negative effects of helminth-microbiota interactions will be 3 6 4 contingent on the specific sex and genotype of the hosts considered.

6 5
The absence of helminths in wealthy counties has been postulated to contribute to the   conducted two sets of experimental infections (Additional file: Figure S1).

9 5
(Experiment 1) As a pilot study, we evaluated the effects of cestode exposure, and 3 9 6 successful infection, on the host gut microbiome. We experimentally exposed adult pure-bred 3 9 7 fish from six full-sib families from Gosling Lake, to S.solidus. Within each family, five 3 9 8 individuals were controls fed uninfected copepods (sham exposure), while the rest received 3 9 9 infected copepods, only some of whom were ultimately infected. This experimental design 4 0 0 yielded three categories of fish within each family: unexposed controls, exposed-but-uninfected 4 0 1 controls, and infected fish (N=30, Additional file: Figure S1a and Table S1). For these exposures 4 0 2 we followed a standard procedure described by Weber et al. [50]. Briefly, we dissected mature interactions, we experimentally exposed adult F2 hybrid fish (intercross and backcross, N=711, 4 1 2 Additional file: Figure S1b and Austin.

3 6
The raw paired-end reads were demultiplexed, and subsequent sequence processing was

1
Chimeric sequences were detected using VSEARCH within mothur in each sample and removed.

2
The remaining sequences were classified by using Bayesian classifier with a training set ( Analyses of pure GOS fish experimentally exposed to S.solidus or a negative control 4 4 9 We sequenced the gut microbiota of GOS fish (Experiment 1) to obtain counts of microbe used ANOVAs to test for effects of sex, cross, and infection status on each LDA axis. We also 4 7 7 confirmed these statistical effects using a MANOVA applied directly to the top 50 PCoA axes to 4 7 8 test for effects of sex, cross, infection, host mass, and interactions among these variables. We mapped quantitative trait loci (QTL) for several microbiome measures: alpha diversity (using 4 9 2 2000 or 4000 read depth normalization), the top 10 weighted PCoA axes (or unweighted axes), 4 9 3 and the relative abundance of the common microbial Orders (as described above for GLMs). We 4 9 4 built linkage maps for each cross separately in R/qtl [60], and used the scanone function with 4 9 5 'hk' interval mapping (using rank-based nonparametric tests for microbial Order relative 4 9 6 abundance). To account for the complex cross design (with backcrosses and F2 intercrosses), we 4 9 7 built separate QTL maps within each of the three cross types, then summed their LOD scores.

9 8
This yields one summary statistic measuring a locus' association with the focal trait, while 4 9 9 accounting for between-cross differences in QTL effect size or marker linkage. We compared 5 0 0 this summed LOD against null expectations obtained by 1,000 permutations of the focal 5 0 1 phenotype across fish within each cross, each time redoing each cross' QTL map and summing 5 0 2 cross null LOD scores. Conservatively, we consider an observed QTL significant when its 5 0 3 summed LOD exceeded the 99.99% quantile from that marker's null values at that same marker.

0 4
We double-checked each significant QTL with a GLM (as described above) testing for fish 5 0 5 genotype effect (at the nearest genetic marker) on the focal microbiome variable.

6 5
Proteomic analysis of secretory products from the model gastrointestinal nematode  Rhodobacterales are more abundant in fish with a greater fraction of Roberts lake ancestry, but only among 8 1 4 uninfected fish; infection leads to a higher Rhodobacterales abundance that is similar across fish genotypes. Fig. S5 8 1 5 2 7 shows the same plots, but with all data points included to show the small effect size relative to high among-8 1 6 individual variation.   Each chromosome is plotted as a vertical grey bar, and to its right we plot microbial QTL located on that chromosome. A dot symbol indicates the location of maximum LOD score for each QTL, and thin vertical lines indicate the inferred width of the QTL. For sex-linked microbes, their QTL span the entire X chromosome, so we omit that linkage group. The key to the right lists the locations of the significant QTL for each focal taxon.