Gut microbiota in a host–brood parasite system: insights from common cuckoos raised by two warbler species

ABSTRACT

An animal's gut microbiota (GM) is shaped by a range of environmental factors affecting the bacterial sources invading the host. At the same time, animal hosts are equipped with intrinsic mechanisms enabling regulation of GM. However, there is limited knowledge on the relative importance of these forces. To assess the significance of host-intrinsic vs environmental factors, we studied GM in nestlings of an obligate brood parasite, the common cuckoo (Cuculus canorus), raised by two foster species, great reed warblers (Acrocephalus arundinaceus) and Eurasian reed warblers (A. scirpaceus), and compared these with GM of the fosterers’ own nestlings. We show that fecal GM varied between cuckoo and warbler nestlings when accounting for the effect of foster/parent species, highlighting the importance of host-intrinsic regulatory mechanisms. In addition to feces, cuckoos also expel a deterrent secretion, which provides protection against olfactory predators. We observed an increased abundance of bacterial genera capable of producing repulsive volatile molecules in the deterrent secretion. Consequently, our results support the hypothesis that microbiota play a role in this antipredator mechanism. Interestingly, fosterer/parent identity affected only cuckoo deterrent secretion and warbler feces microbiota, but not that of cuckoo feces, suggesting a strong selection of bacterial strains in the GM by cuckoo nestlings.

metabarcoding, bacteria, feces, deterrent secretion, great reed warbler, Eurasian reed warbler

INTRODUCTION

Vertebrate gastrointestinal tracts are colonized by taxonomically and functionally diverse bacterial communities that interact with a broad range of host physiological processes and provide benefits to the host via increased digestion and vitamin synthesis efficiency, protection against pathogens and stimulation of the immune and nervous systems (Jumpertz et al. 2011; Koch and Schmid-Hempel 2011; Cryan and Dinan 2012). Gut microbiota (GM) exhibits pronounced variation at the interindividual and interspecific levels (Baxter et al. 2015; Yuan et al. 2015; Lewis, Moore and Wang 2016; Kreisinger et al. 2017; Kropáčková et al. 2017), which can modulate host phenotype (Claus et al. 2008; Han et al. 2016). Understanding the factors behind GM variation is considered a crucial endeavor in current ecological research as it improves our understanding of the mechanisms involved in three-way interactions between GM, host and its environment.

There are two basic sources of variation affecting GM communities, i.e. host-intrinsic and environmental factors. Host-intrinsic factors may be triggered by dozens of genes, being related, for example, to immune functions and facilitating selection of appropriate microbes from the environment or regulation of microbial populations already present in the host's gut (Benson et al. 2010; McKnite et al. 2012; Bolnick et al. 2014; Kropáčková et al. 2017). Environmental factors include a range of biotic and abiotic agents that affect the composition and spatiotemporal variation of environmental bacterial sources colonizing the gut. Physical contact with conspecifics and associated intraspecific microbiota transfer are other important sources of environmental variation shaping symbiotic bacterial communities (Lucas and Heeb 2005; Kreisinger et al. 2017; Ambrosini et al. 2019). There are also a plethora of factors linked to the external environment, including diet composition (Bodawatta et al. 2018; Loo et al. 2019b; Teyssier et al. 2020) and eukaryotic gut parasites that can directly affect the within-gut environment, thereby shaping associated microbial communities (Kreisinger et al. 2015; Newbold et al. 2017; Aivelo and Norberg 2018). Last but not the least, effects induced by environmental factors may vary depending on host-intrinsic regulatory mechanisms. For example, pronounced variation in GM changes has been observed following the introduction of wild-living species into captivity (McKenzie et al. 2017). Host-specific GM changes have also been recorded following infection by intestinal helminths, which can be explained by variation in host genetic factors interacting with both GM and the parasite (Reynolds et al. 2014). While the relative effect of environmental vs host-intrinsic factors on GM variation has been intensively studied in mammals (Benson et al. 2010; Campbell et al. 2012; Nelson et al. 2013; Menke et al. 2017), it has been somewhat neglected in non-mammalian vertebrate taxa, which can harbor strikingly distinct GM (Lucas and Heeb 2005; Ruiz-Rodríguez et al. 2009a; Hird et al. 2014; Kreisinger et al. 2017; Loo et al. 2019b).

Obligate avian brood parasitism represents a reproductive strategy whereby species lay eggs in the nests of other species that then foster the parasitic offspring (Davies 2000). Brood parasites utilizing multiple foster species represent a unique type of natural experiment allowing the role of host-intrinsic vs environmental factors to be disentangled, with parasitic progeny from the same genetic background being exposed to different environmental conditions, i.e. foster species. Surprisingly, GM variation in brood parasites has received relatively little attention. To our knowledge, there have been only four studies focused on GM variation in avian brood parasites and/or their fosterers (Ruiz-Rodríguez et al. 2009a,b, 2018; Hird et al. 2014), three of which compared GM between parasitic chicks and the fosterers’ own chicks (Ruiz-Rodríguez et al. 2009a,b, 2018). As these studies always comprised only one foster species of brood parasite system, they say little about the relative effects of host-intrinsic vs environmental factors on GM. In our study, we address for the first time the role of environmental vs host-intrinsic factors on GM of brood parasites raised by two different foster species and compare it with the GM of the fosterers’ genetic offspring.

We focus on GM variation in an obligate brood parasite, the common cuckoo (Cuculus canorus), which exploits a range of passerine bird species as potential fosterers (Moksnes and Røskaft 1995; Stokke et al. 2018). Approximately 40 h after hatching, a common cuckoo chick evicts its nest mates, thereby reducing all potential bacterial interactions with the fosterer's offspring (Honza, Vošlajerová and Moskát 2007). We analyzed the GM of cuckoo chicks raised by two foster species, the great reed warbler (Acrocephalus arundinaceus) and the Eurasian reed warbler (A. scirpaceus), and contrasted it with the GM of the genetic progeny of both foster parent species. This allows us to test whether host-intrinsic mechanisms shape GM structure. Great reed warblers are around three times larger than Eurasian reed warblers (Cramp 1992) and, consequently, cuckoos raised by the former species grow faster and have higher body mass at fledging than cuckoos raised by the latter species (Kleven et al. 1999). While the breeding habitats of the two fosterer species are comparable, with minor differences in microhabitat structure (Dyrcz 1981; Leisler 1981; Saino 1989), the two species differ in the diet provided to the parasitic and own chicks (Dyrcz 1979; Grim and Honza 1996, 1997), which ensures sufficient variation in environmental component of so-called nidobiome, i.e. a complex of processes shaping microbial colonization in neonates (Campos-Cerda and Bohannan 2020). Consequently, we asked if such a sort of environmental variation affects GM of warbler's own and adoptive offspring. As in many previous studies (Hird et al. 2015; Lewis, Moore and Wang 2016; Kropáčková et al. 2017; Grond et al. 2019; Loo et al. 2019a), we used fecal microbiota analyzed by high-throughput sequencing of 16S rRNA amplicons as a GM proxy for both cuckoo and warbler offspring. Furthermore, by adopting the metabarcoding approach, we were able to replicate previous research on warbler and cuckoo nestling diet (Grim and Honza 1997), i.e. one of the main sources of vertebrate GM variation (Ley et al. 2008; Zhu et al. 2017; Youngblut et al. 2019). The main aim of diet profiling was to confirm that (i) nestling diet differs according to foster/parent species and (ii) individual foster/parent species provisions their genetic progeny and cuckoo nestlings with a comparable diet. In addition to fecal samples, we also analyzed the microbial content of a dark secretion of putative cecal origin (Röder et al. 2014) produced by cuckoos but not warbler nestlings. This secretion repels nest predators owing to its high concentration of volatile compounds, many of which are putative by-products of bacterial metabolism (e.g. butyric acid, acetic acid and indoles; Röder et al. 2016). It has been suggested that symbiotic microbes also contribute to the secretion's repulsive properties (Röder et al. 2016). Consequently, the secretion may represent a further example of microbiota facilitating chemical communication in vertebrates (Theis et al. 2013; Lam et al. 2018), thus extending the host's phenotype beyond a capacity inherent to its genome. Surprisingly, there are no studies aimed at profiling the secretion microbial community.

METHODS

Field sampling

Samples were obtained from a fishpond system situated between Hodonín (48° 51′ N, 17° 07′ E) and Mutěnice (48° 54′ N, 17° 02′ E) in South Moravia, Czech Republic. All samples were collected during the breeding season between 30 May and 10 July 2015. Around 46% of great reed warbler nests were parasitized by common cuckoos in the population, while the cuckoo parasitism rate in Eurasian reed warbler nests was around 10% (Jelínek et al. 2016). Warbler nestlings were sampled 6–10 days after hatching and cuckoos 7–17 days after hatching. We collected fecal samples from 20 common cuckoo, 16 great reed warbler and 9 Eurasian reed warbler nestlings, along with 15 samples of common cuckoo nestlings’ deterrent secretions (Table S1, Supporting Information). Only one nestling was sampled from each foster brood. Day of year of sampling did not differ between samples collected from great reed warbler and Eurasian reed warbler nests (t-test: t = 0.351, P = 0.727) or between cuckoo and warbler samples (t-test: t = 0.265, P = 0.792).

Nestlings were temporarily removed from the nest and both fecal (in warblers and cuckoos) and deterrent secretion (in cuckoos) samples were collected directly into sterile DNA/RNA free cryotubes (Simport, Beloeil, Canada) filled with a self-made DNA/RNA-stabilizing buffer based on RNAlater (protocol available upon request) and kept at −20°C until the end of the field work. Then, they were transferred into the laboratory and kept at −80°C.

All field procedures were approved by the ethical committee of the Czech Academy of Sciences (Animal Care Protocol numbers 173/2008 and 128/2010) and by the relevant conservation authorities (permits JMK20189/2010 and MUHO 2680/2014 OŽP).

Microbiota profiling and bioinformatic processing of 16S rRNA data

We isolated metagenomic DNA using PowerSoil kits (Mo Bio Laboratories Inc., Carlsbad, USA) and subsequently amplified the V3–V4 region of 16S rRNA using S-D-Bact-0341-b-S-17 (CCTACGGGNGGCWGCAG) and S-D-Bact-0785-a-A-21 (GACTACHVGGGTATCTAATCC) primers (Klindworth et al. 2013) and tagged both of them with 10 bp oligonucleotides for multiplexing. Technical PCR duplicates were prepared for all samples in order to check the consistency of microbial profiles. PCR yields were low in the case of two fecal samples. Thus, we did not use them for preparation of sequencing libraries. Sequencing libraries were prepared using TruSeq nano kits and sequenced on Illumina MiSeq using v3 chemistry (300 bp paired-end reads). Further details on laboratory procedures associated with microbiota profiling and subsequent bioinformatic processing of sequencing data are provided by Kreisinger et al. (2017).

Diet metabarcoding

To gain an insight into the diet of cuckoo and warbler nestlings, metabarcoding analysis was applied using the same fecal metagenomic DNA samples as for bacterial 16S rRNA profiling. Previously published universal cytochrome c oxidase subunit I (COI) primers targeting a broad range of invertebrate taxa (Elbrecht and Leese 2017) were employed for this purpose. Details on laboratory procedures and data processing associated with the metabarcoding experiment are available in Supporting Information A1.

Bioinformatic and statistical analysis

Fastq files were demultiplexed and primers trimmed using skewer (Jiang et al. 2014). Next, we trimmed low-quality 3′ ends (250 base pairs from forward and 220 base pairs from reverse reads being retained), eliminated low-quality sequences (maximum expected error per sequence < 1) and denoised quality-filtered files using DADA2 (Callahan et al. 2016). After the denoising, forward and reverse files were merged using DADA2 and chimeric 16S rRNA variants were identified and eliminated using UCHIME (Edgar et al. 2011) and the GOLD database (Mukherjee et al. 2017).

The taxonomy of the resulting unclustered 16S rRNA variants (hereafter operational taxonomic units, i.e. OTUs) was assigned using RDP classifier and the Greengenes reference database (v. 13.8.; DeSantis et al. 2006). PyNAST (Caporaso et al. 2010) was employed for sequence alignment and an OTU phylogenic tree was subsequently constructed using FastTree (Price, Dehal and Arkin 2010). The resulting OTU table, sample metadata and phylogenetic tree were merged into a phyloseq database (McMurdie and Holmes 2013) for the purposes of further statistical analysis.

In our study, we provide separate analyses regarding GM between (i) cuckoo and warbler fecal samples and (ii) between cuckoo deterrent secretion and cuckoo feces. Shannon diversity indices were calculated for alpha diversity analysis using rarefaction-based normalized OTU tables (random subsetting of read counts per sample corresponding to 1268 sequences, i.e. minimal sequencing depth). Shannon diversity was subsequently used as a response variable in analysis of variance (ANOVA) to test whether microbial diversity varied due to explanatory variables that included the effect of foster/parent species (Eurasian reed warbler vs great reed warbler) and the effect of sample type (cuckoo vs warbler nestling feces or cuckoo feces vs cuckoo deterrent secretion). Variation in microbial composition between samples was assessed using two types of ecological dissimilarity, each capturing different aspects of GM divergence, i.e. Bray–Curtis, which accounts for OTU abundance (Bray and Curtis 1957), and a binary version of Jaccard dissimilarity, which accounts for OTU absence or presence only (Jaccard 1901). Furthermore, to check if bacterial phylogeny affects GM variation pattern, UniFrac dissimilarities were calculated (Lozupone and Knight 2005; Supporting Information A1). In order to account for uneven sequencing depth between samples, we rarefied the OTU table to achieve the same sequence coverage per sample prior to calculation (n = 1268 sequences per sample, i.e. minimal sequencing depth).

Between-sample variation in GM composition was visualized using non-metric multidimensional scaling (NMDS) considering two axes. As stress values were relatively high (∼0.2 in some cases, specified below), we also provided NMDS solutions for three axes (Supporting Information A1). Using PERMANOVA (adonis function from the vegan package in R; Oksanen et al. 2013), we tested whether there was any divergence in microbial composition between cuckoo and warbler nestling feces and whether composition varied according to foster/parent species identity. Next, we tested using PERMANOVA whether there was any difference in diversity and composition between cuckoo feces and deterrent secretion in the microbial profiles. Finally, generalized linear models with negative binomial error distribution, obtained from the DESeq2 package (Love, Huber and Anders 2014), were employed for differential abundance analyses. Read counts for individual bacterial genera were used as response, and explanatory variables included the effect of foster/parent species and the effect of sample type (cuckoo vs warbler nestling feces or cuckoo feces vs cuckoo deterrent secretion). Significance of the explanatory variables was tested using likelihood ratio tests. Model fitting and statistical testing included all default steps implemented in the DESeq2 pipeline. All statistical analyses and their visualizations were conducted using R v.3.4.4 (R Core Team 2018) and the R-based packages mentioned above.

RESULTS

Characteristics of cuckoo and warbler microbial profiles

After all filtering steps, our final dataset comprised 475 604 sequences assigned to 1651 OTUs (mean number of sequences per sample = 8200). According to phylum-level classification, Firmicutes bacteria were the dominant component of both warbler and cuckoo microbiota (average proportion of reads 55% for warbler feces, 51% for cuckoo secretion and 56% for cuckoo feces). Substantial variations in abundance patterns were observed in other bacterial phyla. Proteobacteria, for example, tended to be more abundant in warbler feces (26% of reads) compared to cuckoo secretion samples (15% of reads) and cuckoo feces (11% of reads), and while Bacteroidetes abundance was relatively high in some cuckoo samples (25% of reads in secretion and 8% in feces), it was low in warbler feces (2% of reads). Tenericutes, Actinobacteria, Fusobacteria and Chlamydia all occurred at relatively high levels in some samples (Fig. 1); however, their uneven distribution prevented us from drawing any general conclusions as to systematic variation due to the effect of offspring or foster/parent species identity. A more detailed overview of the taxa in our dataset is provided in Figure S1 and Table S2 (Supporting Information).

Figure 1.

Proportion of dominant bacterial phyla and classes with relative abundance > 1% in feces and deterrent secretion of common cuckoo (CC feces and CC secretion) and feces of warbler nestlings (RW feces) raised by great reed warblers (GRW) or Eurasian reed warblers (ERW).

Variation in fecal microbiota with respect to nestling and foster/parent species

Foster/parent species had no effect on Shannon diversity estimates (ANOVA, F_(1,41) = 0.004, P = 0.953, Fig. 2); however, fecal microbiota exhibited higher richness in cuckoo nestlings compared with warbler nestlings (ANOVA, F_(1,41) = 5.867, P = 0.020, Fig. 2). We found significant differences between cuckoo and warbler nestling fecal microbiota composition, and a significant effect of foster/parent species for all GM dissimilarities but weighted UniFrac (Table 1; Fig. 3; Supporting Information A1). DESeq2 analysis further identified six bacterial genera that were more abundant in warbler feces (Rhodobacter, Sodalis, ‘Candidatus Arthromitus’ and Lactobacillus, along with Legionellaceae and Mycoplasmataceae, which were not classified to genus level; Table S3, Supporting Information), and five genera that were more abundant in cuckoo fecal samples (Escherichia, Methylobacterium, Ruminococcus, Cetobacterium and Clostridium; Table S3, Supporting Information).

Figure 2.

Boxplots for Shannon diversity of microbiota associated with feces and deterrent secretion of cuckoo nestlings (CC feces and CC secretion) and feces of warbler nestlings (RW feces) raised by great reed warblers (GRW) or Eurasian reed warblers (ERW). Groups A and B are significantly different (P = 0.020).

Figure 3.

NMDS considering two axes for fecal and deterrent secretion microbiota of cuckoo (CC feces and CC secretion) and fecal microbiota of warbler nestlings (RW feces). Species identity of fosterers/parents (great reed warbler (GRW) or Eurasian reed warbler (ERW)) is indicated by different plotting characters as detailed in the figure legend. Two types of dissimilarity were used for NMDS ordination: Bray–Curtis (stress = 0.22) and Jaccard (stress = 0.20).

Table 1.

Compositional differences in fecal microbiota between cuckoo and warbler nestlings (offspring) raised by two foster/parent species, Eurasian and great reed warblers. Calculations were based on PERMANOVA and two types of community dissimilarity. Significant predictors are in bold.

Dissimilarities	Variables	Df	F	R2	P
Bray–Curtis	Foster/parent species	1, 40	1.752	0.040	0.001
	Offspring	1, 40	1.679	0.039	0.001
Jaccard	Foster/parent species	1, 40	1.493	0.034	0.003
	Offspring	1, 40	2.427	0.055	0.001

Separate PERMANOVA analyses found no effect of foster species on cuckoo fecal microbiota composition (Table 2). On the other hand, separate PERMANOVA of warbler samples revealed a significant difference in composition between the two species (Table 2). Furthermore, DESeq2 identified four bacterial genera whose relative abundances varied between the two warbler nestling species (Sodalis, Marinomonas, Carnobacterium and an unclassified Mycoplasmataceae; Table S3, Supporting Information).

Table 2.

Separate PERMANOVA analyses testing for the effect of foster/parent species (Eurasian or great reed warbler) on microbiota composition in (A) warbler nestling feces, (B) cuckoo nestling feces and (C) cuckoo nestling deterrent secretion. Calculations were based on two types of community dissimilarity. Significant predictors are in bold.

		A: Warbler feces
Dissimilarities	Variables	Df	F	R2	P
Bray–Curtis	Parent species	1, 21	1.978	0.086	0.001
Jaccard	Parent species	1, 21	1.590	0.070	0.002
		B: Cuckoo feces
Dissimilarities	Variables	Df	F	R2	P
Bray–Curtis	Foster species	1, 18	1.129	0.059	0.241
Jaccard	Foster species	1, 18	0.891	0.047	0.703
		C: Cuckoo deterrent secretion
Dissimilarities	Variables	Df	F	R2	P
Bray–Curtis	Foster species	1, 13	1.607	0.110	0.023
Jaccard	Foster species	1, 13	1.924	0.129	0.003

Differences in GM between cuckoo feces and deterrent secretion

We collected both feces and deterrent secretions for 10 cuckoos (five raised by great reed warblers and five by Eurasian reed warblers). Interindividual differences between fecal and secretion microbiota profiles were not intercorrelated (Mantel test, P > 0.2, range of correlation coefficients: 0.145 to −0.168 for all dissimilarity types). Consequently, we did not account for within-individual covariation in microbiota structure between fecal and deterrent secretions in subsequent analyses.

While we observed a marginally non-significantly higher bacterial diversity in cuckoo deterrent secretion compared with cuckoo feces (ANOVA, F_(1,33) = 4.087, P = 0.051; Fig. 2), PERMANOVA revealed pronounced differences in the composition of cuckoo fecal microbiota vs deterrent secretion (PERMANOVA, p < 0.002 for all GM dissimilarity types but weighted UniFrac; Table 3; Supporting Information A1). In addition, deterrent secretion microbial profiles were also distinct from those of both warbler and cuckoo nestling fecal microbiota according to NMDS (Fig. 3). DESeq2 identified 43 bacterial genera differentiating cuckoo secretion profiles from those of cuckoo fecal microbiota (e.g. Methylobacterium, Marinomonas, Streptococcus, Sediminibacterium, Anaerofilum, Fusobacterium; Table S3, Supporting Information). As in the case of the cuckoo fecal microbiota, cuckoo deterrent secretion microbial diversity was unaffected by foster species (ANOVA, F_(1,33) = 0.713, P = 0.405; Fig. 2). Unlike fecal microbiota, however, deterrent secretion composition varied significantly between cuckoos raised by different warbler species (Table 2). Despite the significant difference, the effect size of deterrent secretion composition variation due to foster species was low, DESeq2 only identified the genus Clostridium as more abundant in the deterrent secretions of cuckoos raised by great reed warbler compared with cuckoos raised by Eurasian reed warbler (log2fold change = −8.224, adjusted P < 0.001; Table S3, Supporting Information).

Table 3.

Composition of cuckoo fecal vs cuckoo deterrent secretion microbiota (sample type) between individuals raised by two foster species, Eurasian and great reed warblers. Calculations were based on PERMANOVA and two types of community dissimilarity. Significant predictors are in bold.

Dissimilarities	Variables	Df	F	R2	P
Bray–Curtis	Foster species	1, 32	1.624	0.046	0.003
	Sample type	1, 32	1.667	0.047	0.002
Jaccard	Foster species	1, 32	1.655	0.045	0.012
	Sample type	1, 32	2.998	0.082	0.001

Diet variation

Using COI metabarcoding, we were only able to retrieve useful data on diet composition for 49% of fecal samples. This was mainly due to poor PCR amplification of the COI and a high representation of reads corresponding to taxa that could not be considered as diet components within the resulting COI profiles. As such, just 21 samples yielded diet profiles useful for quantitative analyses. Consistent with the basic assumptions of our study, statistical analysis of this subset revealed that parents of the same species provided a comparable diet to both genetic progeny and cuckoo parasites, though the diet fed by the two parent species was different. Specifically, Eurasian reed warbler brought a more diverse diet that included a higher percentage of true bugs, while beetles tended to predominate in the diet provisioned by great reed warbler parents. For more details on the diet profile analyses, see Supporting Information A1.

DISCUSSION

We asked to what extent environmental and host-intrinsic factors affect the GM structure during early stages of the post-natal development. To achieve this goal, we studied brood parasite's GM exposed to two distinct environmental contexts represented by two different foster species and compared it with GM of fosterer's genetic progeny. Despite the two environmental contexts that were characterized by distinct diet composition, one of the main forces shaping GM in vertebrates (Ley et al. 2008; Zhu et al. 2017; Youngblut et al. 2019), we showed in our study system that host-intrinsic factors dominated over environmental factors. At the same time, however, variation explained by both host-intrinsic and environmental effects was rather low. This is consistent with most studies on free-living birds, where GM exhibit pronounced interindividual variation and rapid temporal changes (Kreisinger et al. 2017; Escallon et al. 2019; Grond et al. 2019). Consequently, variables predicting systematic avian GM changes at both interpsecific and interindividual levels are typically of low effect size.

Our conclusion that host-intrinsic factors have more decisive effect on GM than environmental factors is based on systematic differences in the composition and diversity of fecal microbiota between cuckoo and warbler nestlings, irrespective of foster/parent species. GM differences between brood parasite young and genetic progeny were already reported by previous studies on the great spotted cuckoo (Clamator glandarius) fostered by magpies (Pica pica; Ruiz-Rodríguez et al. 2009a, 2018). In addition, Ruiz-Rodríguez et al. (2018) noted that the GM of great spotted cuckoo nestlings fostered by magpies was a mixture of GM from magpie nestlings and the great spotted cuckoo adults. However, these results do not allow direct comparison of environmental vs host-intrinsic factors as they were based on uniform environmental context represented by a single foster species. In addition, unlike the common cuckoo, the great spotted cuckoo does not evict its nest mates, which increases the complexity of this model system due to social GM transmission among individuals sharing the same nest (see also Kreisinger et al. 2017; Ambrosini et al. 2019). Interestingly, all dissimilarity measures, with the exception of weighted UniFrac, provided congruent support for microbiota divergence between cuckoo and warbler nestlings. We assume that lack of host-specific signal in the case of weighted UniFrac was caused by relatively low phylogenetic divergence of bacterial taxa that, according to DESeq2 analyses, varied between cuckoo and warbler nestlings.

Unfortunately, our data cannot provide direct insights into the mechanisms causing the differentiation in fecal microbiota between warbler and cuckoo nestlings. Nevertheless, variation in gut anatomy and function between cuckoos and passerines is the most likely source of the observed divergence. The most striking difference in the lower digestive tract of these two avian groups is probably the considerable reduction in passerine ceca, which are otherwise well developed in cuckoos (Clench and Mathias 1995; Ruiz-Rodríguez et al. 2009a) and typically host an abundant bacterial community involved in food decomposition (White 2005; Skadhauge 2012). Based on our data, we propose that migration of bacteria from cecal content to fecal material may have contributed to compositional differences in fecal microbiota between cuckoo and warbler nestlings. Consistent with this possibility, deterrent secretion composition (putatively produced in ceca; Röder et al. 2016) was more dissimilar to warbler fecal microbial samples than to those of cuckoos. In addition, most bacterial genera that were more abundant in cuckoo feces (e.g. Ruminococcus, Clostridium and Cetobacterium) corresponded to obligatory anaerobes predisposed to colonization of cecal content (Julliand et al. 1999; Zhu et al. 2002; Tsuchiya, Sakata and Sugita 2008; Suzuki and Nachman 2016). At the same time, however, we cannot exclude contributions of other factors, e.g. differences between passerine and cuckoo immune systems.

The deterrent secretion produced by the cuckoo chick is a rare example of a chemical anti-predation defense in birds (Canestrari et al. 2014; Röder et al. 2014; Trnka et al. 2016). The secretion is believed to be of cecal origin (Röder et al. 2016). Though this view requires direct verification, a similar type of secretion, so-called ‘cecal feces’, is known to be produced in ceca of various avian lineages (Clarke 1979; Villanúa et al. 2006). Moreover, there is also indication that the cecal feces, similarly to cuckoo's deterrent secretion, provide protection against predators (Swennen 1968). The cuckoo's secretion contains high concentrations of volatile molecules directly linked with its repulsive properties. Several of the bacterial genera over-represented in the cuckoo deterrent secretion (compared to cuckoo feces), i.e. Ruminococcus, Bacteroides and Parabacteroides, are known to produce short-chain fatty acids and other volatile molecules, previously detected in the deterrent secretion at high concentrations (Stack, Hungate and Opsahl 1983; Koh et al. 2016). Furthermore, the deterrent secretion was also enriched by the genus Fusobacterium, whose volatiles cause oral malodor in humans (Sterer and Rosenberg 2011) and form a component of scent gland microbiota, important for chemical signaling, in mammalian carnivores (Theis et al. 2013). As such, our data suggest that symbiotic microbiota contributes to the expression of antipredatory chemical signaling in the common cuckoo. Nevertheless, further research should employ functional metagenomic and proteomic tools in order to link GM changes observed at the taxonomic level with functional pathways responsible for production of deterrent volatiles.

Compared to the deterrent secretion, cuckoo fecal microbiota was enriched with several genera corresponding to lactic acid bacteria (i.e. Carnobacterium, Vagococcus, Lactobacillus, Lactococcus and Streptococcus), which are common inhabitants of avian GM (Kropáčková et al. 2017; Bodawatta et al. 2018; Grond et al. 2019). Members of this bacterial clade rely on anoxic decomposition of monosaccharides, prefer energy-rich substrates and exhibit tolerance to high acidity levels (Hijum et al. 2006; Nazef et al. 2008). Consequently, we speculate that a higher proportion of lactic acid bacteria in feces may reflect avian GM functional variation in different gut compartments and spatial variation of biotic and abiotic conditions within the gut. Fecal microbiota is also more likely to be affected by bacteria from external environmental pools than secretion microbiota. This is supported by the fact that plenty of aerobic bacteria, and/or bacteria that are regularly detected in environmental sources (e.g. Sphingomonas, Photobacterium, Marinomonas, Sediminibacterium and Phycicoccus), were more abundant in cuckoo feces than in cuckoo deterrent secretion.

Our data provide ambiguous support for the effect of foster/parent species on GM structure in nestlings. While we observed significant differences in fecal microbiota composition between great reed warbler and Eurasian reed warbler juveniles, we were unable to distinguish the extent to which this variation was affected by environmental or host-intrinsic factors, given the correlative nature of the data used. Nevertheless, consistent with the findings of the previous studies, nest environment has already been shown to be an important predictor of GM at the intraspecific level in passerines (Kreisinger et al. 2017; Teyssier et al. 2018; Ambrosini et al. 2019). Moreover, a study by Grond et al. (2017) reported that shorebird hatchling's GM exhibited comparable composition with microbiota from environmental samples, which supports the effect of environmental bacterial pools on avian GM. In the case of cuckoo GM samples, the effect of nest environment and diet was confirmed for microbiota associated with the deterrent secretion, but not for fecal samples. However, just one bacterial genus (Clostridium) at moderate abundance was associated with these changes, suggesting a limited effect of rearing conditions on cuckoo GM content.

We conclude that interspecific variation in host-intrinsic regulatory factors and gut compartment variation were the most probable sources of the observed changes in GM, the effect of environmental variation apparently being of lower importance in our model system. Moreover, our results also imply that symbiotic GM may have played an important role in the evolution of the cuckoo nestling's unique chemical antipredatory defense.

DATA ACCESSIBILITY

Sequencing data are available at European Nucleotide Archive under project accession numbers PRJEB38922 and PRJEB38897. Accession numbers for each sample are provided in Table S1 (Supporting Information).

ACKNOWLEDGEMENTS

We are grateful to Kirsten Grond, Kevin Roche and anonymous reviewers for helpful comments on an earlier version of the manuscript. We also thank Miroslav Čapek, Václav Jelínek, Jaroslav Koleček, Kateřina Sosnovcová and Michal Šulc, for their help with the fieldwork. Access to computing and storage facilities owned by parties and projects contributing to the National Grid Infrastructure MetaCentrum, provided under the programme ‘Projects of Large Research, Development, and Innovations Infrastructures’ (CESNET LM2015042), is greatly appreciated.

FUNDING

This research was supported by the Czech Science Foundation Projects (grant numbers 17-12262S and 19-19307S) and the Charles University Grant Agency Projects (grant numbers 281315 and 1438417). LS was supported by SVV project no. 260571/2020. We acknowledge the CF Genomics of CEITEC supported by the NCMG research infrastructure (LM2015091 funded by MEYS CR) for their support with obtaining scientific data presented in this paper.

AUTHOR CONTRIBUTIONS

Study design: JK, PP; field sampling: MP, PP, MH; laboratory analysis: LS, JFM; data analysis: LS, JK; funding: JK, LS, MH; manuscript drafting: LS, JK. All authors provided helpful comments and recommendations and approved the final version of the manuscript.

Conflicts of interest

None declared.

Author notes

These authors have contributed equally to this work.