Local habitat, not phylogenetic relatedness, predicts gut microbiota better within folivorous than frugivorous lemur lineages

Abstract

Both host phylogenetic placement and feeding strategy influence the structure of the gut microbiome (GMB); however, parsing their relative contributions presents a challenge. To meet this challenge, we compared GMB structure in two genera of lemurs characterized by different dietary specializations, the frugivorous brown lemurs (Eulemur spp.) and the folivorous sifakas (Propithecus spp.). These genera sympatrically occupy similar habitats (dry forests and rainforests) and diverged over similar evolutionary timescales. We collected fresh faeces from 12 species (six per host genus), at seven sites across Madagascar, and sequenced the 16S rRNA gene to determine GMB membership, diversity and variability. The lemurs' GMBs clustered predominantly by host genus; nevertheless, within genera, host relatedness did not predict GMB distance between species. The GMBs of brown lemurs had greater evenness and diversity, but were more homogeneous across species, whereas the GMBs of sifakas were differentiated between habitats. Thus, over relatively shallow timescales, environmental factors can override the influence of host phylogenetic placement on GMB phylogenetic composition. Moreover, feeding strategy can underlie the relative strength of host–microbiome coadaptation, with Madagascar's folivores perhaps requiring locally adapted GMBs to facilitate their highly specialized diets.

1. Introduction

A current difficulty within gut microbiome (GMB) research is determining the respective influence of host phylogenetic placement versus feeding ecology on GMB structure across species [1–4]. On the one hand, by scaling analyses across mammalian taxa, some researchers have shown that host phylogeny strongly predicts GMB composition [2], whereas convergent feeding strategies may produce strongly [3] or weakly [4] convergent microbial consortia. On the other hand, by scaling analyses within host lineages, others have shown host–GMB coadaptation relative to evolutionary time [5] and habitat occupation [6]. Rarely, researchers have combined these approaches, comparing multiple, speciose lineages to examine how different feeding strategies or habitats may differentially shape GMBs over shallower or deeper timescales, e.g. 1 versus 100 Myr [1]. Using wild subjects, we apply this integrated approach across speciose lineages of frugivores and folivores to relate GMB variation to host relatedness and habitat occupation.

Of the feeding strategies available to animals, folivory and herbivory specifically depend on microbial metabolism to convert complex plant fibres into nutrients [7]. Generally, folivores have GMBs that are rich in fermentative taxa [4,8] and are sensitive to seasonal diets [8]. Although GMB function may be similar across folivores, distantly related folivorous primates show little similarity in GMB structure [4]. Relative to other feeding strategies that are less reliant on microbial action, it remains unclear if and how folivory drives host–GMB dynamics within host lineages.

Madagascar's lemurs represent unique potential for comparative research to investigate GMB structure relative to host evolution and ecology. The diversity of habitats in Madagascar has led to unparalleled species diversity and endemism [9]. We focus on brown lemurs (Eulemur spp.) from Lemuridae and on sifakas (Propithecus spp.) from Indriidae, whose extant species diverged over similar evolutionary timescales [10] and radiated across habitats. Brown lemurs are frugivores, exhibit minimal feeding flexibility across seasons [11] and have short gastrointestinal tracts [12] that promote the absorption of sugars. Sifakas are frugo-folivores, rely on seasonal resources [11] and have elongated intestines and sacculated caeca [13] that facilitate microbial fermentation. Lemurs in both genera show more feeding flexibility in dry forests than rainforests [11]. Different species of brown lemurs are sometimes sympatric, but all sifaka species are allopatric: most primary forests in Madagascar contain one sifaka species and minimally one brown lemur species. Studies of these animals have shown their GMBs to be compositionally distinct from one another [4,14], but these comparisons relied on a single host species per genus. Here, examining the GMBs from multiple, sympatric and allopatric species allows greater resolution for parsing the influence of evolutionary versus ecological factors.

We compiled a GMB database, across habitats, based on faecal samples from 12 species of brown lemurs and sifakas. Consistent with previous research [4,14], we expect the GMBs of brown lemurs and sifakas to cluster predominantly by host genus. If host phylogenetic relationships underlie GMB variation, we expect more closely related species to share more similar GMBs. If feeding strategy and dietary diversity (i.e. local habitat) underlie GMB variation, we expect GMBs to differ between hosts living in dry forests versus rainforests [6]. Because folivores are more reliant than frugivores on GMBs to process their diets, we expect folivores to show more distinct GMB patterns relative to host phylogeny or habitat.

2. Material and methods

The faecal samples derived from 128 lemurs, representing 12 species inhabiting seven sites in Madagascar (figure 1a). We collected fresh samples from March through early November in 2016 and 2017. Based on local circumstances, the samples were immediately frozen at –20°C or placed in 96% ethanol or OMNIgene.GUT buffer. These storage conditions produce reproducible results [15] (electronic supplementary material, S1). Storage conditions remained consistent during transit to Duke University, where samples were frozen at –80°C.

Figure 1.

Features of the lemurs and their gut microbiota, colour-coded by host genus. Shown in (a) is a representative phylogenetic tree of the host species [10], their habitats, sample sizes, sampling locations and photographs (courtesy of David Haring). Shown in (b,c) are species comparisons based on alpha diversity measures, including (b) the Shannon index and (c) phylogenetic diversity. Shown in (d–g) are results based on the beta diversity measure of unweighted UniFrac distance, illustrated as (d) a principal coordinate (PC) plot, (e) pairwise comparisons averaged across species and (f,g) microbiome phylogenetic divergence relative to host phylogenetic divergence, both (f) across all species comparisons and (g) within host lineages. White circles (in f) represent comparisons between genera; dashed lines (in f,g) represent regression lines; ** denotes significance at p < 0.01; *** denotes significance at p < 0.001. (Online version in colour.)

We extracted DNA, amplified and sequenced 151 × 151 bp, paired-end amplicons of the v4 region of the 16S rRNA gene using the 515f-806r primers and Illumina's MiSeq platform [16]. We generated approximately 5.5 million sequences and implemented a bioinformatics pipeline in the Quantitative Insights Into Microbial Ecology (QIIME) software (v. 1.9.1) [16]: we retained samples represented minimally by 10 000 reads and binned sequences into operational taxonomic units (OTUs) using the de novo UClust method based on 97% sequence identity and default parameters. During quality filtering, we removed one sample from downstream analyses. Mean sample coverage was 43 307 sequences (standard deviation: 12 190). Further data normalization did not influence results (electronic supplementary material, S2). We used OTUs to determine GMB composition, via comparison to the GreenGenes database (UCLUST method: v13_8) and to calculate alpha (Shannon and phylogenetic diversity (PD) indices) and beta (unweighted (UUF) and weighted UniFrac distances) diversity. We report UUF metrics in the main text (for comparable results from weighted metrics, see electronic supplementary material). Sequences are available online [17].

We performed statistical analyses using R software (version 3.2.0) in RStudio (version 1.1.463). We tested whether alpha diversity varied by host genus and habitat using linear mixed models (LMM) (glmmADMB package: version 0.8.3.3) [18]. All indices were normally distributed. We ran one model per index: indices were entered as the dependent variable; host genus, habitat, storage condition and sequence depth were entered as explanatory variables; species was included as a random term.

We tested whether beta diversity varied by host phylogenetic relatedness, genus and habitat. We performed permutational multivariate analyses of distance with 999 permutations (vegan package: version 2.4-5 [19]). The dependent variable was UniFrac distance; the explanatory variables included host genus, habitat, storage condition, sequence depth and host species nested within the genus. To further check whether the habitat was a significant predictor of lemur GMBs, we reran analyses within lineages, retaining the same explanatory variables, except for host species, which was nested within the habitat.

We determined UniFrac distances between species by averaging pairwise comparisons of individual lemurs. To compare matrices of host divergence times [10] against averaged UniFrac distances, across and within host genera, we performed Mantel tests with 999 permutations (ade4 package: version 1.7-4 [20]). We performed Wilcoxon tests to compare average UniFrac distances across and within host genera and habitats. We used linear discriminant analysis (LDA) effect size (LEfSe) and the Benjamini–Hocherg correction factor for multiple testing [21,22] to determine whether the relative abundances of microbes were enriched in different host genera and habitats. For additional information, see electronic supplementary material, S1–S6.

3. Results

Across Madagascar, brown lemurs and sifakas harboured distinct GMBs (electronic supplementary material, S5), with brown lemurs having greater alpha diversity (LMMs of Shannon: z = 2.97, p = 0.003; PD: z = 4.00, p < 0.001; figure 1b,c). Also, there was an overall effect of host genus on beta diversity (PERMANOVA of UUF: R² = 0.013, p = 0.024; figure 1d). In pairwise comparisons of beta diversity, averaged across species, GMB variation between host genera exceeded that within genera (Wilcoxon tests of UUF: W = 0, z = 7.194, p < 0.001, for both comparisons; figure 1e). Moreover, there was significantly more between-species variation in sifakas than brown lemurs (Wilcoxon test of UUF: W = 0, z = 5.688, p < 0.001). Reconstructing evolutionary patterns using mean pairwise comparisons revealed a significant relationship between host phylogenetic divergence and GMB divergence (Mantel test of UUF: r = 0.962, p < 0.001; figure 1f); however, this relationship was not present within lineages (Mantel tests of brown lemur UUF: r = −0.0270, p = 0.449; sifaka UUF: r = 0.223, p = 0.286; figure 1g). Lastly, the GMBs of brown lemurs and sifakas differed in their microbial memberships: the sifakas' GMBs were enriched with microbes unassigned at the kingdom level (log(LDA) = 5.22, p < 0.001). Of the assigned microbes, the relative abundances of 68 versus 25 OTUs were enriched in the consortia of brown lemurs and sifakas, respectively (table 1, electronic supplementary material, S7).

Table 1.

Known bacterial genera that were enriched in the gut microbiomes of different host genera.

host genus	microbial taxa			log (LDA)	p-value
	phylum	order	[family] genus
brown lemurs	Acidobacteria	Acidiobacteriales	Terriglobus	2.11	<0.001
	Actinobacteria	Actinomycetales	Actinomycetospora	2.38	0.015
			Pseudonocardia	2.16	<0.001
	Bacteroidetes	Bacteroidiales	[Paraprevotellaceae] Prevotella	3.37	<0.001
			Paraprevotella	3.51	<0.001
			Bacteroides	4.74	<0.001
			Parabacteroides	3.68	<0.001
			[Prevotellaceae] Prevotella	4.61	<0.001
	Firmicutes	Bacillales	Paenibacillus	2.34	0.022
		Clostridiales	Clostridium	2.25	<0.001
			Coprococcus	3.39	<0.001
			Dorea	3.37	<0.001
			Roseburia	2.90	<0.001
			Faecalibacterium	3.32	<0.001
			Oscillospira	3.41	<0.001
			Anaerovibrio	3.76	<0.001
			Megamonas	3.43	<0.001
			Megasphaera	2.60	<0.001
			Mitsuokella	2.36	0.020
		Erysipelotrichales	Eubacterium	2.70	<0.001
			Bulleidia	3.73	<0.001
	Proteobacteria	Caulobacterales	Mycoplana	2.34	0.041
		Burkholderiales	Sutterella	4.08	<0.001
		Desulfovibrionales	Desulfovibrio	3.20	<0.001
		Campylobacterales	Campylobacter	3.56	<0.001
		Aeromondales	Succinivibrio	2.04	<0.001
		Pasteurellales	Actinobacillus	2.14	<0.001
			Avibacterium	2.07	<0.001
			Haemophilus	2.00	0.008
		Pseudomonadales	Acinetobacter	2.37	<0.001
			Moraxella	2.14	0.01
			Pseudomonas	2.42	<0.001
	Spirochaetes	Sphaerochaetales	Sphaerochaeta	3.12	<0.001
		Spirochaetales	Treponema	4.19	<0.001
	Tenericutes	Anaeroplasmatales	Anaeroplasma	3.20	<0.001
		Mycoplasmatales	Mycoplasma	3.14	<0.001
	Verrucomicrobia	Verrucomicrobiales	Akkermansia	2.42	<0.001
sifakas	Actinobacteria	Coriobacteriales	Adlercreutzia	2.48	<0.001
	Fibrobacteres	Fibrobacterales	Fibrobacter	3.62	<0.001
	Firmictues	Lactobacillales	Streptococcus	2.13	0.023
		Clostridiales	Anerostipes	2.09	<0.001
			Buyrivibrio	4.03	<0.001
			Lachnobacterium	2.14	0.002
			Lachnospira	2.20	<0.001
			Phascolarctobacterium	3.72	<0.001
	Proteobacteria	Burkholderiales	Rubrivivax	2.31	<0.001
		Desulfovibrionales	Bilophila	3.24	<0.001
		Aeromondales	Anaerobiospirillum	2.91	<0.001
			Succinatimonas	2.34	0.041

Although host phylogeny did not explain GMB differences within genera, the habitat was associated with microbiome phylogenetic distance in sifakas. Analysis of beta diversity revealed an effect of habitat on GMBs across all subjects (PERMANOVA of UUF: R² = 0.065, p = 0.001). Within lineages, there was no relationship between habitat and GMB beta diversity for brown lemurs (PERMANOVA of UUF: R² = 0.018, p = 0.330; figure 2a), but a significant relationship between habitat and GMB diversity for sifakas (PERMANOVA of UUF: R² = 0.033, p = 0.001; figure 2b). Pairwise comparisons of beta diversity, averaged across species, indicated that, compared to brown lemur congenerics, there was significantly more GMB variation for sifaka congenerics inhabiting different habitat types (Wilcoxon test of UUF: W = 0, z = 4.101, p < 0.001; figure 2c). There was also significantly more GMB variation for sifaka congenerics inhabiting different habitat types compared to congenerics inhabiting different rainforests (Wilcoxon test of UUF: W = 0, z = 2.609, p = 0.009), but no such differences were apparent within brown lemurs. The relative abundances of five and 24 microbial taxa, respectively, were enriched within species of brown lemurs and sifakas living in different habitats (table 2, electronic supplementary material, S7).

Figure 2.

Host habitat and gut microbiota, coded by host genus (colour families), species identity (shapes) and habitat (colour). Shown are results based on the beta diversity measure of unweighted UniFrac distance, including (a,b) principal coordinate (PC) plots and (c) pairwise comparisons averaged across species. ** denotes significance at p < 0.01; *** denotes significance at p < 0.001. (Online version in colour.)

Table 2.

Known bacterial genera that were enriched in the gut microbiomes of lemurs inhabiting different habitats.

host		microbial taxa			log (LDA)	p-value
genus	habitat	phylum	order	[family] genus
brown lemurs	dry forest	Firmicutes	Clostridiales	Butryivibrio	2.66	0.031
	rainforest	Firmicutes	Clostridiales	Anaerovibrio	3.78	0.020
		Proteobacteria	Burkholderiales	Oxalobacter	2.35	0.013
			Desulfovibrionales	Desulfovibrio	3.30	0.045
sifakas	dry forest	Firmicutes	Lactobacillales	Streptococcus	2.63	<0.001
			Clostridiales	Anaerovorax	2.67	0.027
				Phascolarctobacterium	3.10	<0.001
		Proteobacteria	Desulfovibrionales	Bilophila	2.94	0.017
			Campylobacterales	Campylobacter	2.81	0.002
				Flexispira	2.39	<0.001
		Spirochaetes	Spirochaetales	Treponema	3.08	<0.001
	rainforest	Bacteroidetes	Bacteroidales	[Paraprevotellaceae] Prevotella	3.93	<0.001
		Firmicutes	Clostridiales	Oscillospira	2.67	<0.001
		Proteobacteria	Aeromondales	Anaerobiospirillum	3.03	0.005

4. Discussion

In these contemporaneous lineages of primates studied across Madagascar, GMBs clustered predominantly by host genus; however, within genera, host relatedness did not track GMB phylogenetic relatedness. Within the folivorous sifakas, community variability was instead predicted by habitat, but the same pattern was not evident within the frugivorous brown lemurs. Although host phylogenetic relationships relate to GMB membership at broad taxonomic scales [2,4], the same may not be always true within lineages [14]. Instead, environmental [1] or dietary [3] challenges may produce shifts in GMBs that, at the extreme, override the signal of the host's phylogeny. As the sifaka lineage radiated across Madagascar, the species’ unique microhabitats and folivorous diets were likely to be strong drivers of GMB variation over shallow timescales.

Although various mechanisms could account for species differences in GMB evenness, PD and taxonomic enrichment, we noted clear linkages to host dietary patterns. We suggest that the fruit-rich diets and shorter gastrointestinal tracts of brown lemurs selected for the relatively even distribution of generalist microbes from diverse phylogenetic groups, and for greater abundances of simple-fibre metabolizers, like Prevotella [4] and oxygen-tolerant Proteobacteria [23]. By contrast, the folivorous diets of sifakas, even when comprising seasonal fruit [11], likely selected for microbial communities enriched for complex-fibre specialists, like Fibrobacter [4], or tannin specialists.

Also notable is the finding that habitat had the strongest effect on sifakas, evidenced by their GMBs displaying a reduced number of enriched taxa shared by all individuals, a greater number of taxa differentiating rainforest and dry-forest congenerics, and greater community distances between species living in different habitats. We posit that these patterns likely stem from the nutritional challenges associated with folivory [7], but not frugivory. The finding that extant brown lemurs shared more homogeneous GMBs perhaps suggests that the consortia in early members of this clade could process fruit-based diets, even as hosts diverged across Madagascar. By contrast, the consortia of early sifakas perhaps responded to the unique dietary composition in microhabitats encountered throughout speciation. The persistence of varied diets may have driven stronger host–microbiome coadaptation in sifakas.

Future studies could address whether the greater dietary seasonality of dry-forest inhabitants [11] contributes to greater GMB variation. Notably, frugivorous brown lemurs in dry forests (versus rainforests) may fall back on foliage in leaner seasons, explaining the greater abundance of complex-fibre metabolizers, like Butyrivibrio, in their consortia. In folivorous sifakas, the enrichment for simple-fibre metabolizers, like Prevotella, in the GMBs of rainforest (versus dry-forest) dwellers may point to more consistent fruit availability. Our results indicate that feeding strategy can underlie the relative strength of host–microbiome coadaptation and that the considerable microbial requirements necessary to sustain folivory drive adaptation for hosts and microbiomes alike.

Ethics

Our procedures were approved by Madagascar's Ministry of Environment, Ecology and Forests (permit nos: 162/16; 028/17; 083/17; 136/17; 146/17; 164/17; 035/18).

Data accessibility

DNA sequences: NCBI SRA accession PRJNA495032.

Authors' contributions

L.K.G. conceived of the study. All authors contributed to the study design and sample collection. L.K.G. analysed data and, with C.M.D., drafted the manuscript. All authors contributed to manuscript preparation and revision, gave final approval and agree to be held accountable for the content.

Competing interests

We declare we have have no competing interests.

Funding

Research was funded by the Margot Marsh Biodiversity Foundation and Duke University (L.K.G. and C.M.D). Additional sampling was funded by the University of Minnesota and College of Veterinary Medicine (J.B.C.), Imperial College London (R.S.R), Virginia Tech (B.P.S.), the Rufford Foundation (M.A.S) and Emory University (T.R.G.).