The microbiome of African penguins (Spheniscus demersus) under managed care resembles that of wild marine mammals and birds

Animals under managed care in zoos and aquariums are ideal surrogate study subjects for endangered species that are difficult to obtain in the wild. We compared the fecal and oral microbiomes of healthy, managed African penguins (Spheniscus demersus) to those of other domestic and wild vertebrate hosts to determine how host identity, diet, and environment shape the penguin microbiome. The African penguin oral microbiome was more similar to that of piscivorous marine mammals, suggesting that diet and a marine environment together play a strong role in shaping the oral microbiome. Conversely, the penguin cloaca/fecal microbiome was more similar to that of other birds, suggesting that host phylogeny plays a significant role in shaping the gut microbiome. Although the penguins were born under managed care, they had a gut microbiome more similar to that of wild bird species compared to domesticated (factory-farmed) birds, suggesting that the managed care environment and diet resemble those experienced by wild birds. Finally, the microbiome composition at external body sites was broadly similar to that of the habitat, suggesting sharing of microbes between animals and their environment. Future studies should link these results to microbial functional capacity and host health, which will help inform conservation efforts.


Sample collection
Samples were obtained between November 2018 and March 2019 from various body sites (described below) of 36 African penguins (Spheniscus demersus) under managed care at Georgia Aquarium (Atlanta, GA, USA).All sampled penguins were born under managed care either at Georgia Aquarium or another zoological institution before being transferred to Georgia Aquarium (Supplementary Table 1).At the time of sampling, Georgia Aquarium housed approximately 55 penguins, with the number of individuals occupying the exhibit fluctuating as chicks were born and animals were moved on and off exhibit.Penguins were maintained in an indoor, 16,500-gallon exhibit that contained an artificial rocky environment and 3 basins filled with artificial seawater 0.5-1.5 m deep.All basins were connected and supplied from the same aquarium life support system.The water from the entire exhibit was turned over every 37 min using a filtration system composed of two vertical sand filters, a protein fractionator, an ozone contactor, and a deaeration tower.The water temperature was maintained at 17 °C using a titanium plate heat exchanger, and penguins were fed a mixed piscivorous diet (capelin (Mallotus villosus), Pacific herring (Clupea pallasii), squid (Loligo sp.), night smelt (Spirinchus starksi), and silversides (Menidia menidia)) ad libitum twice a day.
Microbiome samples were collected opportunistically while the birds were removed from the exhibit for routine annual health exams or as needed to address minor medical concerns.No birds were sampled more than once.All body site samples were collected by gently rubbing sterile swabs along the cloaca, oral cavity (near the choana), brood pouch, uropygial (preening) gland, leg skin, or back skin, accumulating material over the entire surface of the swab.For skin samples, feathers were parted to access the skin.Fecal samples and feather samples were also collected opportunistically when they fell to the floor during examination-these were considered "environmental" samples as they were no longer a part of the penguin body.Samples were immediately preserved in an RNA/DNA stabilizing buffer (25 mM sodium citrate, 10 mM EDTA and 70 g ammonium sulfate per 100 ml solution, pH 5.2) and frozen at -80 °C until further processing.Not all body sites were sampled from all penguins (Supplementary Table 1).Sample collection was approved by the IACUC ethics committee at Georgia Institute of Technology, under protocol A100174.All penguin samples were collected by veterinary staff at the aquarium in accordance with the relevant guidelines and regulations, in accordance with ARRIVE guidelines.
Swabs were also collected from surfaces within the exhibit, including dry rocks, wet rocks, the waterline of the exhibit, fresh and dry guano, and three different nesting niches.Water microbiome samples from the exhibit basin were collected by filtering basin water through a 0.2 µm Sterivex filter, as described in our prior work (e.g., 25 ).Penguin food microbiome samples were also collected by rubbing a sterile swab on the external surface and into the cloaca of fish and other mollusks contained in a bucket of food to be delivered to the penguins.All environmental samples (water, surface and food swabs) were collected at a single time point in July 2018.Samples were immediately preserved in RNA/DNA stabilizing buffer, and frozen at -80 °C until further processing.A summary of all samples is included in Tables 1 and S1.

DNA extraction and sequencing
Total DNA was extracted from swabs using the PowerSoil DNA extraction kit (QIAGEN, Location, USA).Swabs were placed directly into PowerBead tubes and extracted according to the manufacturer's instructions.For each kit, an extraction blank (sterile swabs without biomass) was processed following the same procedures, resulting in four extraction blanks total.DNA from Sterivex filters was extracted following the protocol described in 25 .For each sample, the V3-V4 region of the 16S rRNA gene was amplified by PCR using primers F515 and R806

Meta-analysis data processing
To assess the relative influence of host identity, diet, and environment in shaping the oral and cloaca penguin microbiomes, we compared the African penguin oral and cloacal datasets to cloacal or fecal, and oral microbiomes of other domestic and wild birds, reptiles and mammals (including marine mammals).Datasets included in the meta-analyses were selected on the criteria that they provide informative comparisons to the African penguin, targeted the V3-V4 region of the 16S rRNA gene, and were generated using either 454 or Illumina sequencing technologies.A summary of these datasets is in Table 2.All datasets were trimmed to retain the same V3-V4 region amplified in this study.Datasets sequenced using Illumina technology were then pooled and processed independently through DADA2 using the same methods as above.Datasets sequenced using 454 technology were processed separately, with parameters adjusted in the DADA2 algorithm to deal with such data by modifying denoising parameters (HOMOPOLYMER_GAP_PENALTY = − 1, BAND_SIZE = 32) and filtering sequences by maximum length 30 .All ASV tables issued from Illumina and 454 sequencing technologies were merged and taxonomy was assigned following the same method as described above.

Statistical analysis
Statistical analyses were conducted through QIIME2 31 to test for differences between penguin body sites and environmental samples, and between different animal classes (reptiles, birds, and mammals) and penguins when

Ethical approval
This study is approved by the IACUC ethics committee at Georgia Institute of Technology, under protocol A100174.This study is in accordance with relevant guidelines and regulations.

Results
16S rRNA gene sequence data were generated for 176 penguin samples from various body sites including the cloaca, mouth, skin, brood pouch, and uropygial (preening) gland, as well as environmental samples including the exhibit water, surfaces in the exhibit, and the food (Supplementary Table S1).After quality filtering, trimming, and rarefaction, 161 samples constituting 3232 ASVs composed the final penguin dataset (Table 1).All negative controls and extraction blanks did not pass quality control, validating our quality control method.Meta-analyses were conducted comparing the penguin oral and cloaca/fecal microbiomes to those from other domestic and wild vertebrate species (Table 2).After quality filtering, trimming, and rarefaction, a total of 162 samples constituting 6522 ASVs were used for the oral meta-analysis, and 226 samples constituting 8795 ASVs were used for the fecal/cloaca meta-analysis (Table 2).Datasets included in the oral and fecal meta-analyses were selected because they allowed informative biological comparisons to the African penguin data and they were generated using methods and primer sets consistent with those used for our data.However, not all studies selected for these analyses included both oral and fecal microbiomes from all animals.These analyses revealed that the taxonomic composition and diversity indices for hen and broiler poultry samples were notably distinct compared to those of other vertebrates (Fig. 1; Fig. 2A).These datasets did not share several phyla present in all other bird hosts and were instead contained high representation (25-50%) by Lactobacillus, which represented < 3% of sequences in all other vertebrates (Supplementary Figures S1, S2).Poultry samples were therefore removed from statistical analyses.Table 2. Sample metadata and number of associated amplicon sequence variants (ASVs) for the oral and fecal microbiome meta-analyses.The datasets included in the meta-analyses were selected for their informative comparisons to the African penguin.Few studies have compared the oral microbiome across animals, potentially due to a lack of oral datasets from non-model taxa.We therefore characterized the oral microbiome of the African penguin relative to that of phylogenetically and ecologically divergent vertebrates for which oral microbiome data were publicly available.It should be noted that not all studies selected for the oral and fecal meta-analyses included both oral and fecal microbiomes from all animals.S3).In contrast, microbiome taxonomic composition-beta diversity measured by weighted UniFrac distances-differed among sample types, with cloaca and oral samples differing significantly and clustering to the exclusion of samples from more external body sites (Fig. 3, Supplementary Table S3).While composition in some external body site samples differed significantly from that of environmental samples, external body site microbiomes were generally more similar to those of the surrounding habitat than to those of internal body sites (Fig. 3, Supplementary Table S3).
All external penguin and environmental samples were primarily dominated by Proteobacteria, Firmicutes, Bacteroidetes, and Actinobacteria (Supplementary Figure S4).Actinomyces, Petrimonas, Campylobacter, and Fastidiosipila were more abundant in the cloaca microbiome, while Coenonia and Suttonella were prominent in the oral cavity (Supplementary Figure S5).To assess possible routes of microbial exchange, we examined those taxa shared among penguin and environmental samples.Over 60% of ASVs in external body site samples were also detected in environmental samples, with ASVs of the genera Psychrobacter and Oceanisphaera being particularly  S4 and S6. common in both external body site and environmental samples (Supplementary Figure S3).In contrast, less than 45% of cloaca and oral ASVs were shared with environmental samples, and the fraction of unique ASVs within the oral and cloaca microbiomes was twice that of other body sites (Supplementary Table S4).Microbiome composition of African penguin body sites was not significantly influenced by metadata variables age, sex, time since molt, reproductive status, use of a nest, use of specific medications and supplements, or antiseptic treatment (Kruskal-Wallis test for alpha diversity and PERMANOVA for beta diversity-p > 0.05).However, these tests had low statistical power due to small sample size; further investigation is needed to evaluate the influence of these variables.

The penguin oral microbiome is similar to that of marine mammals
Few studies have compared the oral microbiome across animals, potentially due to a lack of oral datasets from non-model taxa.We therefore characterized the oral microbiome of the African penguin relative to that of phylogenetically and ecologically divergent vertebrates for which oral microbiome data (generated using the  S8.S3. same primers as in our study) were publicly available.Based on the Shannon diversity index, alpha diversity in the African penguin oral microbiome was greater than that of the Cooper's hawk (Accipiter cooperii), prairie rattlesnake (Crotalus viridis), sea lion (Zalophus californiansus), and human (Homo sapiens) oral microbiomes, but lower than that in the bottlenose dolphin (Tursiops truncatus) (Fig. 1; Supplementary Table S5).Based on weighted Unifrac distances, the taxonomic composition of the African penguin oral microbiome differed significantly from that of all other oral microbiomes (Supplementary Table S6).However, principal coordinate analysis showed oral microbiomes of penguins clustering with those of sea lions and dolphins, to the exclusion of oral microbiomes of non-marine vertebrates (Fig. 4); consistent with this clustering pattern, the average weighted UniFrac distance to penguin microbiomes was lowest for sea lion and dolphin datasets (Fig. 5).These trends suggest microbiome similarity in the oral cavities of marine animals.
At the phylum level, marine vertebrate oral microbiomes were again more similar to each other than to those of non-marine vertebrates and harbored the lowest relative abundances of Actinobacteria (< 2%) and Firmicutes (< 9%), but the highest relative abundance of Bacteroidetes (> 34%; Supplementary Figures S6, S7).Interestingly, the Gram-negative bacterial genus Proteiniphilum was found in the oral microbiome of all bird species but not in non-birds, though in low abundance (< 1%).

The penguin fecal/cloaca microbiome is most similar to that of other birds
Alpha diversity in fecal/cloacal microbiomes-including richness and diversity estimated using the number of observed ASVs and the Shannon diversity index-varied among vertebrate species (Fig. 1).Based on the Shannon diversity index, the African penguin cloacal microbiome was significantly more diverse than that of the bottlenose dolphin and Komodo dragon (Fig. 1; Supplementary Table S7).Beta diversity analysis indicated that taxonomic composition in the penguin fecal/cloacal microbiome (based on weighted UniFrac distances) varied significantly from all other species (Supplementary Table S8).In contrast to the oral microbiome, the penguin fecal/cloacal microbiome was most similar to that of other birds, based on average weighted Unifrac distances (Fig. 2B; Fig. 5).The representation of Actinobacteria likely contributed to these differences, as Actinobacteria represented < 0.1% of sequences in mammal datasets, but > 13% of sequences in African penguins and wild birds (barn swallow, and great tit; Supplementary Figure S1).Other microbial taxa enriched in bird microbiomes include Patescibacteria, which was present almost exclusively in wild birds and the genera Corynebactericeae and Catellicococcus, which were uniquely shared by all bird fecal microbiomes, although these bacterial taxa represented a small fraction of the microbiome (< 1.5%) and were composed of different ASVs for each host species.

Discussion
This study compared the microbiome of African penguins under managed care to the microbiome of their exhibit.This analysis showed the microbiome of external body sites to be similar to that of the exhibit, suggesting high connectivity between the birds and their managed, artificial habitat.We also conducted meta-analyses comparing penguin oral and fecal/cloacal microbiomes to that of other vertebrates both under managed care and free-living.
-0.The meta-analyses showed that the oral microbiome of African penguins was more similar to that of marine mammals, while the penguin fecal/cloaca microbiome was more similar to that of other wild birds (Fig. 2).The African penguin microbiome is strongly tied to the environment Beta diversity analyses showed that the microbiomes of all external body sites were similar to those of the surrounding environment, and there were fewer statistically significant differences between the environmental samples and external body sites compared to internal body sites.Similarly, both environmental and external penguin samples were primarily dominated by the same phyla, including the bacterial genera Oceanisphaera and Psychrobacter, which have been associated with both planktonic 34,35 and host-associated marine microbiomes 36,37 .Together, these trends suggest that the external African penguin microbiome is strongly influenced, and likely influences, the microbiome of immediate surrounding environment, with transfer of microbes likely routinely occurring between these body sites and the water and various surfaces of the managed care exhibit.This potential high connectivity suggests a need to keep the habitats of African penguins clean and free of pathogens and that routine monitoring of water and other habitat microbiomes may be useful in identifying microbes that are also in direct association with the animal surfaces.However, further work is needed to assess the extent to which microbial taxa are shared between host-associated and environmental niches under natural settings outside managed care.

The penguin oral microbiome is similar to marine mammals
Although the African penguin oral microbiome differed significantly in composition from the oral microbiomes of all other animals tested, this microbiome was statistically most similar to that of other marine animals (Fig. 5).Moreover, all of these marine animals (the African penguin, sea lion, and bottlenose dolphin) are piscivores.Previous studies highlight the major role that diet plays in shaping the oral microbiota [38][39][40] , and suggest that diet is shaping the oral microbiome of these animals as well.However, it is difficult to disentangle the effects of diet versus other aspects specific to the marine environment (e.g., water column microbiome composition) that may also be shaping microbiome structure.It is likely that the marine lifestyle in general, combined with a piscivorous diet, are driving the similarity between the African penguin and marine mammal oral microbiomes, despite major differences in host anatomy and physiology.Multiple interacting factors-including host physiology and oral anatomy, diet, and contact with the surrounding air/water-likely influence the diversity of the oral microbiome, notably as this body site niche is continually exposed to external components including air/water and food (in contrast to most sections of the gastrointestinal tract).Indeed, alpha diversity in the penguin oral microbiome, like other oral microbiomes analyzed here, is comparable to that of the cloaca/feces, suggesting a complex community potentially associated with diverse functional or spatial niches.The oral cavity and its associated community is the first component of the gastrointestinal tract, with its microbes potentially serving as inocula for other sections of the gastrointestinal system, underlining the importance of the oral system as part of the digestive process but also as a potential reservoir and vector of pathogens and disease [41][42][43][44] .For the African penguins in Georgia Aquarium, the most abundant ASVs (> 1%) unique to both the penguin oral and cloacal microbiome were not pathogenic and corresponded to taxonomic groups that are often found associated with the respiratory system or gastrointestinal tract of birds.These included microbes from the families Cardiobacteriaceae 45 , Flavobacteriaceae 9 , Weeksellaceae, Moraxellaceae 46 , and Mycoplasmataceae 47 .However, evaluating the function of these taxa in relation to host health is difficult, as none of these taxa could be classified to the genus level.Finally, while the penguin microbiome contained relatively high abundances (up to ~ 15%) of microbial species known to harbor some pathogenic strains (e.g., Coenonia in the oral microbiome), penguins from Georgia Aquarium were healthy and under constant veterinary supervision, which suggests that the ASVs recovered from these animals are likely nonpathogenic.These patterns establish a baseline for comparing oral microbiome signatures relative to changes in host health.

The penguin fecal/cloaca microbiome is most similar to that of other birds
Based on previous research, phylogeny and diet are main factors shaping the microbiome of vertebrates [48][49][50] , with variation in the relative strength of these factors among host clades.In our fecal/cloacal meta-analysis, microbial taxonomic composition was more similar among wild birds compared to other mammals and reptiles (Fig. 5), suggesting that phylogeny plays a strong role in shaping the bird gut microbiome 12,13,51 .All birds, including the African penguin, had a fecal/cloacal microbiome composed of 4 major phyla including Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria, which occurred at variable relative abundances among species and exhibited divergence in microbial richness and abundance when evaluated at finer taxonomic levels, as also seen in 9 .Two genera identified as Corynebacteriaceae and Catellicococcus were unique to bird fecal microbiomes and have been previously reported as common in the bird gastrointestinal tract 52,53 .Unlike the oral cavity, the African penguin cloaca microbiome was more similar to that of other wild birds (barn swallow and great tit), highlighting the effect of phylogeny as a main driver of microbiome structure in bird guts 12,13,51 .
Interestingly, the fecal microbiome of factory-farmed poultry was very different from that of any other vertebrates, including all birds, and was highly dominated by Lactobacillus.This suggests that while phylogeny plays a major role in structuring the bird gut microbiome, highly controlled environments such as those used in commercial farming over multiple generations and with high antibiotic and probiotic use may cause drastic changes in microbiome composition.The similarity of the managed African penguin cloaca microbiome to that of wild birds, and its dissimilarity to that of commercial poultry, implies that the exhibit and diet conditions at Georgia Aquarium resemble the wild rather than a domesticated environment.However, samples from wild penguins and other domesticated birds are necessary to validate this hypothesis.

Conclusion
Our work identifies a baseline for evaluating the African penguin microbiome.While the cloaca and oral cavity of African penguins harbored distinct microbial communities that resembled those of other birds and marine vertebrates, respectively, all other body sites were similar to environmental samples, showing a tight connection with the managed care habitat.However, the cloaca microbiome of the African penguin did not resemble that of the domesticated chicken, possibly suggesting that the environment and diet in the managed care setting of Georgia Aquarium mimic wild conditions more closely than those applied in some domesticated bird facilities.Future studies would benefit from including wild animals to assess the influence of a managed care environment on the microbiomes of penguins and other species of conservation interest, and thereby potentially identify biomarkers of host and ecosystem health.Ultimately, this information could be used to inform management strategies in wild populations as well as those under managed care.In particular, the high percentage of taxa shared between environmental niches and the penguin external microbiome identifies exposure to surface and water microbiomes as a potentially significant vector for both commensal and pathogenic microbes.This connectivity highlights the necessity of maintaining clean exhibit conditions, as well as minimizing disturbance to microbial ecosystems influencing wild penguins.Future investigations of penguin microbiomes may inform and improve conservation efforts focused on mitigating the spread of disease.

Figure 1 .
Figure 1.Standard box plots representing differences in observed ASVs (left column) and Shannon diversity index (right column) for the oral (top row) and fecal/cloaca (bottom row) microbiomes of various vertebrate hosts including birds, reptiles, and mammals.Significant differences in the Shannon diversity index are given in Supplementary TablesS4 and S6.

Figure 2 .
Figure 2. Principal coordinate analysis (PCoA) using weighted Unifrac distances between the fecal/cloaca microbiomes of various vertebrate hosts including wild and managed birds, reptiles, and mammals.(A) Includes poultry (hen and broiler) samples, (B) does not include poultry samples.Pairwise PERMANOVA tests are given in Supplementary TableS8.

Figure 3 .
Figure 3. Principal coordinate analysis (PCoA) using weighted Unifrac distances between the microbiome of various African penguin (Spheniscus demersus) body sites and environmental samples of the penguin exhibit at Georgia Aquarium.Pairwise PERMANOVA tests are given in Supplementary TableS3.

Figure 5 .
Figure 5. Average weighted Unifrac distances from the African penguin oral (panel A) and fecal/cloaca (panel B) microbiome.From left to right, average distances are arranged from smallest (more similar to the African penguin) to largest (least similar to the African Penguin).Individual samples and distances are plotted in Figs. 2 and 4. Error bars indicate standard deviation.

Table 1 .
Summary of samples from the African penguins (Spheniscus demersus) of Georgia Aquarium.The number of samples per body site or location and the total number of amplicon sequence variants (ASVs) associated with each sample type are given.*Opportunistically collected from the floor during routine veterinary examinations.**Dry fecal matter collected from rocks in the exhibit.