Gut microbiome variation in juvenile blue tits in a European urban mosaic

Abstract

Urbanisation transforms natural environments, impacting not only wild animals living in cities but also the microorganisms they are hosting. To better understand urban-driven variation in microbiological composition and diversity in the gut of birds developing in urban areas, we collected faecal samples from blue tit Cyanistes caeruleus nestlings using nestboxes distributed across the capital city of Warsaw, Poland. Sampling included a variety of urban habitats, a suburban village and a natural forest area. Microbiome analysis revealed a pattern of reduced alpha diversity and significant shifts in beta diversity in urbanised settings, driven by impervious surface coverage. Additionally, we observed that this effect was year-dependent, therefore highlighting the importance of temporal replication in ecological research. Furthermore, comparing two cavity types (natural and human-made), we demonstrated that artificial nestboxes, a tool widely used in field ecology, can impact the microbiome assembly in nestlings.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-23005-y.

Introduction

Urbanisation creates a novel ecosystem with a set of specific conditions that pose an adaptive challenge for the wild animals inhabiting it—this includes, among many, habitat fragmentation, warmer temperatures, altered resource availability, higher population densities, and exposure to human disturbance^1–3. While these environmental changes may benefit certain species, others face significant challenges, leading to reduced survival, reproductive success, and overall fitness⁴.

For example, urban birds tend to lay their eggs earlier in the season and have smaller clutch sizes compared to their rural counterparts^5,6. Both juvenile and adult birds from urban environments also tend to be smaller and have a lower body mass^5,7,8, and while the stable availability of anthropogenic food positively influences adult well-being, the paucity of natural food sources poses a threat to the nourishment of nestlings⁵. Findings stemming from study sites located in Warsaw, Poland, where our study took place, are consistent with these general patterns: the percentage of impervious surface area around nestboxes was shown to have a negative impact on offspring development, body mass, and survival in both blue tits and great tits^7,9.

The gut microbiome is an important fitness component as it influences several aspects of host biology¹⁰, and the bird microbial assemblage patterns may be shaped by both intrinsic and extrinsic factors, including social relationships, diet, and environmental conditions¹¹. Consequently, urbanisation can impact the gut microbiome through various mechanisms, making it a focal point of interest for urban ecologists. For example, in American white ibises lower microbial diversity was correlated with the quantity of digested human-provisioned food¹². In a study conducted in Kenya on 57 avian species, it was found that increasing human density, coupled with higher livestock density, may be linked to the spillover of antimicrobial resistance genes into birds, potentially facilitating the spread of resistant pathogens and impacting both wildlife and public health¹³. A similar investigation involving samples from 30 different bird species collected in eight countries revealed that the prevalence of antimicrobial resistance genes in gut microbiomes was six times higher in urban environments¹⁴. In a study on house sparrows urbanisation was associated with a lower taxonomic diversity of microbiome and changes in community structure¹⁵. Interestingly, a study on the same study system reported a substantial impact of the availability of anthropogenic food sources on glucose levels and decreased uric acid levels in the blood—both of which could potentially influence microbiome composition¹⁶. Contrary to the notion of lower diversity being a prevailing characteristic of urban microbiomes, urban birds in a study on white-crowned sparrows exhibited higher microbial diversity. This outcome, however, was associated with more variable land cover types in the urban areas under investigation compared to non-urban areas, and the elevated microbial diversity was correlated with a higher tree cover density¹⁷. This result underscores the significance of the local environment, challenging the binary urban and rural dichotomy that inaccurately lumps diverse environments such as woodlands and office areas based on exposure to disturbances¹⁸.

Overall, studies on wild microbiomes in urban environments reveal patterns that are often inconsistent and highly variable, mainly due to the lack of spatial and temporal replication. As the sensitivity to disturbances may be species-dependent, it is also crucial to examine the differences between various species sharing the same habitat. We additionally remark on a visible need for exact quantification of land use that would allow between-location comparisons. Moreover, microbiome studies have predominantly focused on adult birds, and little attention has been given to juveniles. It is thus essential, considering the current scale of anthropogenic pressures, to better understand how resilience to disruptions may be shaped in the earliest stages of development.

Until recently, the only study exploring this aspect in the urban space was conducted on great tits, using the same gradient of urbanisation and methodology as described here. The results revealed that nestlings growing up in more highly urbanised environments exhibited lower gut microbial diversity¹⁸. Building upon these observations and aiming to fill these knowledge gaps, we focused on the blue tit (Cyanistes caeruleus), an avian species not previously examined in terms of its gut microbial diversity in an urban setting, specifically at a life stage when the microbiome is still under development¹⁹. We used 16S rRNA gene sequencing to examine the gut microbiome of blue tits sampled in 2018 and 2019 across a range of urban microhabitats within a European capital city. We tested whether (i) urbanisation impacts individual microbial diversity and composition of blue tit nestlings; (ii) these patterns are repeatable across years; and (iii) nestboxes, as examples of human-made structures, alter the microbial diversity when compared to natural cavities. Additionally, we aimed to assess how generalisable or different these patterns are between two species, great tits and blue tits, by comparing the findings of the present study with those produced for great tits (both species were sampled in the same study locations in the same year)¹⁸.

Results

Faecal samples from 15-day-old blue tit nestlings were collected in 2018 and 2019 as part of a long-term monitoring effort of the breeding activity of blue tits and great tits in a gradient of urbanisation in Warsaw, Poland. We analysed a total of 107 samples originating from 91 nestboxes (Schwegler, type 1b) distributed across 9 study sites located within and outside of the city (urbanisation gradient) and from 16 natural cavities located in an urban forest.

The rarefied dataset comprised a total of 5277 ASVs, with an average of 85 individual ASVs per sample (SD = 76). Across all samples, the microbial community was dominated by three phyla: Firmicutes (59.32%), Proteobacteria (28.5%), and Actinobacteriota (5.77%), which together accounted for the majority of the relative abundance.

Given that environmental conditions, including weather patterns and food availability, can vary between years and influence the avian microbiome, we examined whether observed patterns were consistent across both breeding seasons. To focus on the impact of urbanisation in nestlings reared in the same cavity type, the basic diversity and community structure analyses solely used data derived from nestboxes in the urbanisation gradient. To compare the impact of cavity type on microbiome, we used samples collected across one, highly homogenous location: the urban Bielany forest, where nests in natural cavities (i.e. holes in trees) and artificial cavities (i.e. nestboxes) were present (Supplementary Table 3).

Microbial community diversity and composition in relation to urbanisation and year of sampling

We examined alpha diversity metrics in relation to urbanisation with linear mixed models to account for temporal variation (two years of sampling and differences in individual hatching dates), spatial structure (percentage of Impervious Surface Area as a measure of urbanisation, and sampling site assignment), and potential survival-related health differences (fledging—successful or not). We used three key diversity metrics as response variables in separate models: Shannon’s index, Chao1, and Faith’s phylogenetic diversity (indices described in Methods section). The model with Shannon diversity as the response variable was run without the ISA*year interaction as it was not statistically significant.

Increasing impervious surface area (ISA) around nestboxes was associated with lower Chao1 and lower Faith’s PD (Table 1), indicating fewer detectable ASVs (reduced taxonomic richness) and narrower evolutionary breadth (reduced phylogenetic richness) in more urbanised locations. Shannon’s diversity index, the estimate of richness and evenness, did not change significantly with increasing ISA. These relationships were year-dependent, being evident in 2018 but not detected in 2019 (Table 1, Fig. 1).

Table 1.

Variation in gut microbiome alpha diversity as a function of Impervious Surface Area, year and their interaction, sample collection day, and whether or not the nestling survived until fledging. Significant results are indicated in bold.

Fixed effect	Response variable	Est	CI	p
ISA * year	Chao1	0.03	0.00 to 0.06	0.041
	Faith’s PD	0.02	0.00 to 0.04	0.042
ISA	Chao1	− 0.02	− 0.04 to − 0.00	0.044
	Shannon	0	− 0.01 to 0.01	0.795
	Faith’s PD	− 0.02	− 0.03 to − 0.01	0.008
Year [2019]	Chao1	− 0.7	− 1.2 to − 0.2	0.006
	Shannon	− 0.21	− 0.32 to 0.05	0.162
	Faith’s PD	− 0.43	− 0.78 to − 0.08	0.016
Sample collection day	Chao1	− 0.07	− 0.35 to 0.21	0.619
	Shannon	− 0.09	− 0.19 to 0.05	0.259
	Faith’s PD	− 0.12	− 0.32 to 0.08	0.242
Survival	Chao1	0.44	− 0.22 to 1.09	0.186
	Shannon	0.09	− 0.16 to 0.39	0.399
	Faith’s PD	0.38	− 0.08 to 0.84	0.103

Fig. 1.

Year-dependent alpha diversity patterns in a gradient of urbanisation. The regression lines illustrate the covariation between the percentage of Impervious Surface Area and Faith’s Phylogenetic Diversity index (logarithmic scale).

Across distance metrics (described in greater detail in Methods section), community composition differed between years (PERMANOVA; Table 2; Fig. 2). ISA and site were associated with composition for Jaccard, Bray–Curtis, and unweighted UniFrac, but not for weighted UniFrac (Table 2), indicating that urbanisation alters the taxa present, their relative abundances, and the phylogenetic relationships among them; however, the most abundant taxa remain phylogenetically similar across communities.

Table 2.

Compositional differences in response to different Impervious Surface Area percentages, years, sampling sites, and sample collection dates were tested using PERMANOVA models. Significant results are indicated in bold.

Fixed effect	Response variable	Pseudo-F	p
ISA	Jaccard	1.353	0.001
	Bray–Curtis	1.734	0.0004
	UniFrac	1.424	0.017
	Weighted UniFrac	1.728	0.088
Year	Jaccard	1.493	0.0001
	Bray–Curtis	1.921	0.0001
	UniFrac	1.753	0.001
	Weighted UniFrac	2.163	0.037
Site	Jaccard	1.056	0.038
	Bray–Curtis	1.110	0.013
	UniFrac	1.115	0.030
	Weighted UniFrac	0.911	0.633
Sample collection day	Jaccard	1.149	0.075
	Bray–Curtis	1.283	0.030
	UniFrac	1.444	0.014
	Weighted UniFrac	1.119	0.298

Fig. 2.

UniFrac Beta matrix illustrating dissimilarities in the compositional structure of samples collected in the years 2018 and 2019.

To examine whether this result was driven by sampling year, we ran separate models for each year and, in accordance with the results for alpha diversity (Fig. 1), the impact of urbanisation on beta diversity was visible in 2018, whilst there was no such effect in 2019 (Table 3). The analysis also uncovered differences driven by the timing of sampling during the breeding season in Bray–Curtis and UniFrac matrices (Table 2), which might have influenced microbiome composition due to seasonal environmental variation (e.g. food availability, temperature).

Table 3.

Results of PERMANOVA analyses for 2018 and 2019 datasets. The significant results are indicated in bold, revealing year-specific effects: Impervious Surface Area and sampling sites did influence the compositional similarity of microbiomes, but only in 2018.

Year	Fixed effect	Response variable	Pseudo-F	p
2018	ISA	Jaccard	1.3861	0.0002
		Bray–Curtis	1.7601	0.0001
		UniFrac	1.6495	0.0058
		Weighted UniFrac	1.8145	0.0538
	Site	Jaccard	1.0652	0.0073
		Bray–Curtis	1.1262	0.0053
		UniFrac	1.1025	0.0508
		Weighted UniFrac	0.86878	0.7598
2019	ISA	Jaccard	1.0667	0.3691
		Bray–Curtis	1.1442	0.2036
		UniFrac	0.965	0.5778
		Weighted UniFrac	0.91963	0.4267
	Site	Jaccard	1.042	0.1288
		Bray–Curtis	1.0978	0.0618
		UniFrac	1.0459	0.237
		Weighted UniFrac	1.2515	0.2284

The homogeneity of multivariate dispersions was additionally tested between year groups with PERMDISP. The median distances for UniFrac beta diversity, but for this matrix only, were not even (p = 0.0123), suggesting that the PERMANOVA results, in this case, can be impacted by different levels of internal variability of beta diversity within year groups. In order to account for the significant variation between years in beta diversity, differential abundance analysis was applied, revealing higher abundances of Streptococcaceae and unclassified bacteria from the order Chlamydiales in 2018 (Supplementary Table 1).

Impact of cavity type on nestlings microbiome

To analyse the impact of cavity type on nestling microbiome, we randomly selected a subset of samples collected from natural cavities and nestboxes located in the same urban forest across two years (in total, 16 samples from natural cavities and 17 samples from nestboxes). We did not include nestboxes from locations other than the urban forest to eliminate the possible influence of any varying environmental conditions, especially to exclude differences in urbanisation pressure.

Taking into account the possible impact of year, we ran models that included the interaction between year and cavity type. Shannon index, being one of the three indices that we used, revealed significantly lower microbial diversity in samples collected from natural cavities and a strong effect of the interaction between year and cavity type (Table 4; the remaining models in Supplementary Table 2a). To better understand how cavity type and year interact, we ran post hoc tests, which revealed different year-to-year changes in nestboxes and natural cavities: in nestboxes, in 2019 the microbial diversity was lower than in 2018 (Shannon: p = 0.0118; Fig. 3) while in samples from natural cavities, gut microbiome alpha diversity remained stable between years (Shannon: p = 0.979; Fig. 3; Supplementary Table 2b).

Table 4.

Variation in gut microbiome alpha diversity (Shannon index) as a function of year, cavity type, and their interaction. Reference categories are 2018 for the year and natural cavity for cavity type. Diversity was significantly lower in 2019 and in natural cavities. Significant results are indicated in bold.

Predictors	Est	CI	p
(Intercept) (2018, natural cavity)	3.24	2.6–3.89	<0.001
Year (2019 vs. 2018)	1.94	1.13–2.76	0.699
Cavity type (nestbox vs. natural)	1.3	0.26–2.34	0.016
Year × cavity type	2.38	− 4.13 to − 0.62	0.01

Fig. 3.

Shannon diversity index boxplots (with medians, 25th and 75th percentiles, respectively) for nestboxes and natural cavities across years. Differences between the years are shown in colour. The significant differences revealed by the post hoc test were determined based on the linear model at p values 0.05 (*), p 0.01 (**), and p 0.001 (***).

None of the PERMANOVA analyses used to test for beta diversity dissimilarities in composition between years and cavity type revealed any differences—the exception was the effect of year in the Bray–Curtis matrix (p = 0.0377; Supplementary Table 2c).

Discussion

In our study, we quantified how variation in urbanisation is associated with the gut microbiome of blue tit nestlings across two breeding seasons and between cavity types (natural vs. human-made). Increasing impervious surface area was associated with declines in both taxonomic richness (Chao1) and phylogenetic diversity (Faith’s PD), suggesting that urbanisation filters some taxa. However, Shannon diversity, which integrates richness and evenness, remained unchanged, likely because the relative abundances of the persisting taxa became more evenly distributed, offsetting the reduction in richness. By using Faith’s PD, we intended to check whether urban conditions shrink the spread of lineages in the microbiome—something richness or evenness alone can miss. We treat it simply as a description of evolutionary breadth and avoid health or fitness claims unless supported by functional data.

These uncovered relationships were year-dependent, being evident in 2018 and not detected in 2019, therefore highlighting the importance of temporal replication of ecological studies and remarking that the pressure the urban environment exerts on avian microbiome can vary over time—in our case, this variation was clearly visible when samples were collected in just two contrasting years.

At the beta level, community structure also varied with ISA, but the signal depended on what the metric measures. ISA and site were associated with differences in presence/absence composition (Jaccard), abundance-weighted composition (Bray–Curtis), and with phylogenetic distance (unweighted UniFrac), whereas phylogenetically structured abundance shifts (weighted UniFrac) were not supported. Collectively, these results suggest that the community composition differs primarily in the presence/absence of rare taxa rather than in the dominant phylogenetic lineages. As with alpha-level patterns, year effects were pervasive, pointing to temporal variability in how urban filters act on community membership and relative abundances.

Lower Chao1 and Faith’s PD indices with high ISA, paired with results derived from beta diversity analyses, suggest that urban filters act at colonisation and early microbiome assembly. Likely contributors could include diet shifts in provisioning, contaminants and antimicrobials, microclimate differences, or environmental microbial pools that are more homogenous in built-up areas. The year effect implies these filters strengthen or weaken with annual context (e.g., food availability and weather).

Moreover, the comparison between natural and artificial cavities suggests that these two cavity types provide immediate environments that may differently support microbial assembly in the first days of nestling life, in line with earlier findings in great tits in the same study system¹⁸.

Consequently, a narrower phylogenetic breadth together with fewer ASVs could reduce the range of metabolic and immune functions available during a critical developmental window (e.g., digestion, production of microbial metabolites, immunity, resistance to pathogens). As we cannot claim fitness effects, we recommend further testing of functional consequences to see whether lineages missing at high-ISA sites map onto functions that matter for nestling performance.

The degree of urbanisation can be reflected by the similarity of microbial composition and its diversity

Our results build upon an earlier study on great tits¹⁸, which took place in one of the same years (2018) as the current study, and in the exact same locations. Importantly, Maraci et al.¹⁸ also reported a negative, significant correlation between ISA and microbial alpha diversity. This suggests that these two closely related species that inhabit the same niches might respond similarly to environmental stimuli. Studies on other species investigating how microbial alpha diversity varies in urban spaces have revealed mixed results. Similarly to the results reported in Warsaw, lower diversity in urban areas was observed in adult house sparrows^15,20 and herring gulls²¹. Conversely, gut microbial diversity was higher in urban populations of white-crowned sparrows than in their rural counterparts¹⁷, and no significant results were found in American white ibises¹².

The discrepancies between studies may result from differently defined urbanisation measures (urban/rural, percentage of built-up area, human population, or tree cover density) or simply originate from taxonomy or life-stage history. To capture possible patterns, we would need to use a standardised measuring system that does not raise any doubts regarding the definition of urbanisation of the area¹⁸ and replicate the studies on the same species to observe whether emerging trends are related to the taxonomy or rather the particularity of the studied environment³. Additionally, a meta-analytical quantification of effect sizes across studies could be performed to assess whether trends are emerging across species.

Although some studies showed associations between higher microbial diversity and good health (reviewed in Ottinger et al.²²), it is known that microbiome tends to fluctuate in early life^19,23. The fitness consequences of these variations in gut microbiome should be further investigated. To gain better insight into the possible implications of differences in microbial diversity, it would be necessary to assess not only species diversity but also their functional diversity²⁴.

Year-specific effects of urbanisation on microbial diversity

The analysis of samples collected over two years revealed highly significant temporal differences in gut microbiome variation, suggesting that the impact of urbanisation on developing nestlings can be highly year-specific. Since sampling site characteristics remained the same in 2018 and 2019, the most likely drivers for observed microbiome alterations are differences in climatic conditions regarding temperature, rainfall, humidity, and the downstream differences related to food availability. Indeed, as described by Sudyka et al.²⁵ in a study conducted on blue tits in the same years and in one of the exact locations as the data collected here (urban forest), weather conditions differed markedly between these years²⁵. Weather data from a local weather station indicated that in 2019, April and May were significantly colder in terms of temperature average, minimum and maximum. Additionally, in 2019, April and June tended to be drier, May was more humid with higher precipitation, and the entire season was windier. All these conditions suggest that 2019 was a less favourable year. Importantly, these climatic alterations can directly influence the availability of primary food sources such as caterpillars and other invertebrates²⁶. As the type and amount of food influence the gut microbiome^27,28, the observed differences between years can be attributable to the direct effect of food availability on the gut microbiome. Furthermore, these months are crucial for nestling growth, and changes in food availability might lead to differential growth rates⁷, ultimately influencing microbial assembly. Future research should aim to incorporate quantitative measures of environmental conditions—such as food availability, weather variables, or habitat phenology—to better understand the ecological mechanisms underlying year-to-year variation in gut microbiome composition. The increasing unpredictability of weather patterns, as a result of the human-induced global climate crisis, may drastically interfere with ecosystem stability and resilience²⁹. In environments subjected to frequent and rapid weather fluctuations, the unpredictability of foraging and feeding activities generates irreparable consequences on the endocrine and metabolic systems of both nestlings and their parents³⁰. This could also be reflected in the microbiome.

The two families detected in higher abundances in 2018, Streptococcaceae and unclassified Chlamydiales (Supplementary Table 1), are known to encompass some pathogenic bacterial species, and therefore their increased abundance might pose a health risk. Given that different weather conditions led to varying food availability between the years, these may have also contributed to differences in the bacterial abundances - certain microbial families may, in fact, thrive in response to specific dietary components¹⁵. Although some genera included in these families are commonly known as human pathogens, beneficial interactions have also been observed in wild animals. For instance, higher levels of Streptococcaceae were detected in healthy wild mallards compared to mallards infected by influenza A viruses³¹. Chlamydiales, on the other hand, are more frequently associated with bacterial infections in wild birds and can spill over to domestic animals, posing risks to public health³². Evidence linking specific microbial taxa to direct health outcomes in wild passerines remains extremely limited. One exception comes from a study on ostriches³³, where certain bacterial taxa (not detected in higher abundances in our dataset) were associated with higher mortality rates. These rare examples highlight a pressing need for integrative studies that combine strain-level microbiome data with physiological or immunological indicators to better understand the health implications of microbiome variation in free-living birds.

Differential impact of natural cavities and nestboxes on nestling microbial diversity

By using samples collected only in a single location (urban forest), we eliminated the possible impact of varying levels of urbanisation and other external factors such as food availability. This allowed us to specifically test the effect of cavity type. Nestboxes, most frequently used in studies on cavity-nesting birds, differ on many grounds from the natural nesting environment - most notably, providing poorer insulation, leading to lower humidity and higher fluctuations in daily ambient temperatures^34,35. These differences can impact various aspects of early development. For example, the study by Sudyka et al.²⁵, conducted on blue tits in the same years and location as the data collected for this study, focused on reproductive success dependent on cavity type. Blue tits breeding in nestboxes produced fewer nestlings and had lower hatching and fledging success compared to those nesting in natural cavities.

Results reported here show that alpha diversity was dependent on cavity type, with the microbiome being generally more diverse in nestboxes. However, it is essential to remark that these results were again year-dependent: while in 2018, the median alpha diversity was higher in nestboxes (importantly, the same effect was observed in our previous study on great tits where samples were collected only in 2018¹⁸), this trend reversed in 2019. Strikingly, the year-specific patterns differed between the cavity types: in nestboxes, microbial diversity dropped significantly in 2019, while in natural cavities, diversity remained at comparable levels between years (Fig. 3). This result could possibly imply higher susceptibility of nestlings reared in nestboxes to external conditions. The aforementioned study on blue tit breeding success²⁵ also reported a significantly longer duration of nesting in birds raised in nestboxes. This could suggest that even though all of them were examined and sampled on their 15th day of life, their microbial communities might have been at different stages of development. Additionally, their microbiomes could have been influenced by the overall higher temperature and its fluctuations in nestboxes²⁵, which could mediate immune and metabolic functions, although the relationship between temperature and microbial community assembly remains mostly unknown in wild passerine birds. The only study up to date that has touched upon this aspect, performed on great tits Parus major, found an association between lower temperatures and higher alpha diversity³⁶.

Conclusions and outlook

Our study serves as an extended continuation of previous observations of gut microbiome variation in great tits (Fig. 4), which we have referred to multiple times throughout the text¹⁸. We report a highly equivalent pattern of decreasing alpha diversity with increasing urbanisation, as observed in the same year (2018) and study system. In terms of beta diversity in blue tits, community composition differed with changing ISA percentages. In great tits, no such relationship between ISA and composition was detected, although there were marked differences in microbial composition between urban and rural areas. This might indicate species-specific differences in the sensitivity of microbial composition to fine-scale alteration in the local environment. Additionally, in both species, the type of cavity impacted alpha diversity, with lower values observed in natural cavities. In summary, these results highlight the consistency and complementarity of our studies, providing deeper insight into urban-driven relationships between the two species that share the same ecological niche and parallel gut microbiome variation.

Fig. 4.

Comparison of results obtained in this study and in the previous study on great tits¹⁸. The microbial alpha diversity of both species in 2018 was negatively correlated with urbanisation. Alpha diversity (Shannon index) was also lower in natural cavities than in nestboxes. However, the impact of urbanisation on the structural similarity of microbiomes (beta diversity) was species-specific.

All in all, this study showed clear urban-related alterations in the gut microbiome of a small passerine. However, these results were also found to be year-dependent, highlighting the critical importance of study replication in terms of avoiding overgeneralisation. Additionally, we observed higher diversity in microbiomes of nestlings originating from nestboxes, but their microbiomes were notably more susceptible to change between years than those from natural cavities. While providing valuable insights into how the avian microbiome responds to anthropogenic alteration, our study can also benefit from further refinements: it would be valuable to investigate urban microbiome variation through taxonomic, spatial and temporal replication. A valuable follow-up includes investigating how urban diet and food availability shape microbiome composition, as well as how it can possibly be impacted by weather conditions, particularly in the context of their globally observed unpredictability. Answering these questions would allow us to better comprehend the ecological and evolutionary response of non-human animals to human-induced rapid anthropogenic change. With rapidly increasing global climate change and urbanisation and the progressing disruption of natural habitats, we emphasize the importance of examining the microscale of organismal dependencies to understand how our welfare depends on the seemingly invisible interconnectedness between us and other beings.

Methods

Study sites and environmental metrics

To better reflect urban landscape heterogeneity, the sampling locations, especially within the city borders, varied greatly in the environmental structuring: they spanned from natural and urban forests, through peri-urban village and urban woodlands, to residential and office areas (Fig. 5). Therefore, we decided not to assign the dichotomous categories of urban and rural and instead used a more precise urbanisation measure with values specific to each individual sampling location. We used the percentage of Impervious Surface Area (referred to as ISA) around each nest location (measured in 100 m radius around each nestbox in the study system, and calculated using the imperviousness map downloaded from Copernicus Land Monitoring Services³. The 100 m radius was used based on the food foraging distance observed in this species while feeding the nestlings³⁷. ISA includes both all built-up areas and soil-sealing surfaces, and strongly covaries with various environmental parameters (such as tree cover density³) that may influence the gut microbiome by impacting food availability and, consequently, diet composition. All of the environmental parameters have been calculated for each nestbox as part of the long-term monitoring effort, and they included spatial variables (distance to the city centre, closest road, and closest path), variables collected on the ground (human presence, sound pollution, and temperature), and variables extrapolated from digital photography and satellite imagery (tree cover density and light pollution) (in detail in Supplementary Text 1).

Fig. 5.

(a) Study species: 15 day old blue tit Cyanistes caeruleus on sampling day; (b) Boxplot illustrating proportions of samples in every given location with marked median values of Impervious Surface Area percentage; (c) Map of sampling locations.

Data collection and laboratory work

Sample collection took place depending on individual hatching dates and spanned from May 16 to June 10 in 2018 and May 12 to June 23 in 2019. As food availability may change with time, and diet constitutes an important aspect of microbial community structure, the temporal variability was taken into account by including standardised sampling date. The Julian date of sampling (15th day of life, dates varied for individual nestlings) was standardised by subtracting the population sampling mean and dividing it by its standard deviation, obtaining a population-wide mean sampling date of zero and a standard deviation of 1 for every study year. Faeces were collected from one chick per nest using non-invasive methods and deposited directly in 5 ml sterile Eppendorf tubes filled with 3 ml of RNAlater (Qiagen). All samples collected across the two field seasons (2018 and 2019) were stored at C, transferred to the University of Bielefeld in Germany, and processed together within consecutive days. DNA extractions, PCR amplifications, and sequencing were performed simultaneously using the same protocols, reagents, and equipment to minimise potential batch effects and ensure comparability across years.

The laboratory work was carried out in accordance with the pipeline previously described in the earlier study on great tit nestling microbiome, which was performed on the same urbanisation gradient¹⁸. Microbial DNA was extracted using the DNeasy PowerLyzer PowerSoil Kit (Qiagen), as described in the manufacturer’s protocol. The hypervariable V3–V4 region of the 16S ribosomal RNA gene was targeted following Illumina 16S Metagenomic Library Preparation Guide, 15044223-B. The final amplicon pool, alongside 119 biological samples, contained two blank controls for DNA extraction and one blank control for PCR amplification. All of these were sequenced on the Illumina MiSeq system (Illumina, Inc., San Diego, CA, USA) in CeBiTec at the University of Bielefeld.

Data analysis

The bioinformatic processing was performed using the nf-core/ampliseq standardised pipeline³⁸ and involved subsequent steps. First, the raw sequencing data in FASTQ format was subjected to FastQC and the overall quality of reads was assessed manually. Then, the amplification primers (FW-5′-CCTACGGGNGGCWGCAG-3′ and RV-5′-GACTACHVGGGTATCTAATCC-3′) were used to trim the reads using Cutadapt³⁹. This process required a minimum primer overlap of ten bases with a maximum error rate of 10%. The remaining pool was quality filtered and denoised using DADA2⁴⁰ to correct sequencing errors, remove chimaeras, and create the output of amplicon sequence variants (ASVs). ASVs were then filtered for ribosomal RNA sequences only, using Barrnap⁴¹ as a prediction tool. Eventually, taxonomic classification of the ASVs was assigned using DADA2 and SILVA v138 as a reference database⁴². The phylogenetic tree was generated using Pplacer⁴³.

All of the following statistical analyses were conducted in R version 4.3.1⁴⁴ and PRIMER v7⁴⁵. The raw dataset was initially filtered for contaminants with the help of decontam R package⁴⁶—in total, 42 potential contaminant ASVs were excluded from the dataset. In order to account for the uneven distribution of the number of reads between samples (min = 1136, max = 46,033), all samples that remained after filtering were rarefied to a depth of the lowest read count observed, leaving 3488 ASVs across 107 samples (restricted from 119 as a result of quality filtering).

The impact of urbanisation was measured using three different metrics of alpha diversity: Shannon’s diversity index (phyloseq R package, 1.46.0⁴⁷), Chao1 (phyloseq R package, 1.46.0⁴⁷), and Faith’s phylogenetic diversity (picante R package, 1.8.2⁴⁸). Shannon’s index quantifies within-sample diversity as the joint effect of richness and evenness; it increases when communities contain more ASVs and when their relative abundances are more evenly distributed⁴⁹. Chao1 is a richness estimator that places greater weight on rare ASVs, so lower values indicate reduced taxonomic richness⁵⁰. Faith’s phylogenetic diversity (PD) measures phylogenetic richness as the total branch length connecting the ASVs present in a sample on the underlying phylogeny⁵¹.

These metrics of microbiome diversity were then assessed in the context of impervious surface variation estimated around each nestbox. Due to the non-normal distribution of values in each of these diversity indices, the data was normalised employing square root transformation in the case of Shannon diversity and logarithmic transformation in the remaining two. Three linear mixed models were fitted using the lme4 R package (version 1.1-34⁵²), using ISA, year and its interaction as fixed effects, and the diversity metrics as the response variables. The interaction aimed to test whether a possible ISA effect on microbiome diversity was year-specific. To account for the potential impact of differences in the sample collection dates on alpha and beta diversity, the number of days between the initiation of the study and sample collection was added as a fixed effect. Additionally, nestling survival (i.e., whether or not the chick fledged successfully) was included as a covariate in the alpha diversity models to control for potential differences in microbiome composition associated with individual health status to account for variation in microbial profiles that could arise from underlying health differences. The associations between sampling sites were also added to the model as a random variable to account for the non-independence of the samples originating from the same sampling site. The spatial autocorrelation was also tested in the simulated scaled residuals of the fitted LMMs using Moran’s I test with the use of DHARMa package (version 0.4.6⁵³).

To test whether the microbial communities were affected by cavity type, we employed a subset of data, including samples collected only from the urban forest (17 samples from nestboxes and 16 from natural cavities, one sample per nest). Chao1 and Faith’s PD values were normalised employing logarithmic transformation, and three basic linear models that included interaction between the cavity type and year were fitted.

To visualise the between-group differences in community composition, Principal Coordinate Analysis plots were generated for four types of dissimilarity matrices: Jaccard (qualitative, accounts for presence/absence of taxa⁵⁴), Bray–Curtis (quantitative, abundance-weighted composition⁵⁵), Unifrac (qualitative metric taking into account the phylogenetic relatedness among taxa⁵⁶), and Weighted Unifrac (which accounts for phylogenetic relatedness and abundance of taxa⁵⁷). Then, PERMANOVA models were fitted, taking into account the ISA values, year, site, and sampling date. The models for the subset, including natural cavities focused on cavity type and year.

Finally, to detect finer differences between years and to determine which ASVs were differently abundant, we employed corncob analysis (0.4.1⁵⁸). To detect ecologically relevant patterns, we focused on the family level.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

L.F., Ö.M., and M.S. conceptualized the research idea and planned the experiments. M.S., I.dL. and J.S. collected the biological samples and environmental data. L.F. and Ö.M. carried out the laboratory experiments. S.J. performed the bioinformatic analyses. L.F. carried out the statistical analyses with the supervision of Ö.M. and M.S., with input from J.S., L.F. wrote the manuscript in consultation with Ö.M and M.S., and all authors approved the final version of the manuscript.

Funding

Open Access funding enabled and organized by Projekt DEAL. This research was supported by the Polish National Science Centre under grant number 2021/41/B/NZ8/04472.

Data availability

Sequence data that support the findings of this study have been deposited in the European Nucleotide Archive with the primary accession code PRJEB89086. All the remaining data are available at the following URL: https://doi.org/10.5281/zenodo.17094217.

Declarations

Ethics statement

The research was carried out with a permit from the Regional Directorate for Environmental Protection (RDOŚ) in Warsaw, Poland.

Competing interests

The authors declare no competing interests.