Investigating the association between diet and infection with Trichomonas gallinae in the European turtle dove ( Streptopelia turtur )

Trichomonas gallinae , a protozoan parasite infecting a wide range of birds, has been a cause for conservation concern since its recognition as an emerging infectious disease, having had notable impacts on several avian species, including causing wide-spread mortality in greenfinches ( Chloris chloris ), and chaffinches ( Fringilla coelebs ), and hindering the recovery of the endangered pink pigeon ( Nesoenas mayeri ). Horizontal transmission between birds congregating at feeding sites, such as those provided for conservation and species management purposes, is an important driver in the spread of T. galliane . Supplementary feeding is a key conservation intervention for the European turtle dove ( Streptopelia turtur ), which is declining across its range, driven at least partially by a loss of natural food resources. Due to the link between T. gallinae transmission and supplementary feeding, we consider the prevalence of this parasite among European turtle dove in relation to diet, in the first study to analyze these two factors in the decline of this species together. Using birds sampled from breeding and wintering grounds, the dietary composition of individuals was compared to the presence of T. gallinae , and specific T. gallinae strains. Dietary variation was summarized into two axes using detrended correspondence analysis; neither was associated with the presence of T. gallinae or any specific strains. The proportion of diet accounted


| Importance of disease as a conservation issue
In recent decades, emerging infectious diseases (EIDs) have increasingly become recognized as an important conservation issue, contributing to the decline of species across a wide range of taxa worldwide (Daszak et al., 2000;McCallum & Dobson, 1995).From the amphibian population crashes caused by chytridiomycosis to the extinctions of endemic Hawaiian birds linked to avian malaria, EIDs have the potential to spread rapidly among populations, resulting in mass mortality, local extinctions, and contributing to global species decline (Berger et al., 1998;Cunningham & Daszak, 1998;McCallum, 2012;Warner, 1968).
One of the most damaging epidemics for wild bird populations of recent years is the widespread epidemic of trichomonosis in greenfinches (Carduelis chloris) and chaffinches (Fringilla coelebs).The spread of this disease, caused by the protozoan parasite Trichomonas gallinae, began in western England and Wales in 2005, and spread across large areas of mainland Europe, with outbreaks of trichomonosis in finches occurring in Ireland and across mainland Europe, in Northern Germany, Austria, France, and Slovenia (Amin et al., 2014;Marx et al., 2017;McBurney et al., 2015).Trichomonosis has caused significant declines in the British population of greenfinches and chaffinches, resulting in the highest recorded impact of an infectious disease on a European wild bird population (Lawson et al., 2018).This parasite affects a wide range of avian orders, including Passeriformes, Columbiformes, and Falconiformes (Amin et al., 2014;Bunbury et al., 2008;Sansano-Maestre et al., 2009), with at least 23 strains of the parasite of varying virulence identified (Thomas et al., 2022).It infects one of the UK's most rapidly declining species, the European turtle dove (Streptopelia turtur), and has been linked to both adult and nestling mortality in this species (Stockdale et al., 2015).While this parasite is commonly spread via direct transmission, passing both horizontally, between adults during courtship, and vertically, from parent to offspring during feeding of crop milk (Bunbury et al., 2007), the congregation of birds at artificial feeding sites is believed to have been a key contributor to the rapid spread of T. gallinae, transmitted through contaminated food and water resources at these sites (Lennon et al., 2013;McBurney et al., 2015;Robinson et al., 2010;Thomas, 2017).

| Disease transmission routes and feeding behavior
Transmission route is an important factor when assessing the impacts of disease on a population, and understanding both inter and intraspecific transmission of disease is vital in species management (Herrera & Nunn, 2019).Numerous factors can have an impact on disease transmission and its consequences for wild populations, including the presence of a reservoir host, ecological changes, such as introduced species and population fragmentation, and host behavior (De Castro & Bolker, 2005;Herrera & Nunn, 2019).
Certain behaviors increase or decrease the risk of disease transmission, and one factor that has been attributed to an increase in the spread of disease is the congregation of animals.Wildlife often aggregates in small areas in response to resource availability, in particular, food (Sorensen et al., 2014).The provision of supplementary food for wildlife by humans has become a common activity, with many ecological and social benefits, including supplementing shortfalls in natural food resources (Weidman & Litvaitis, 2011), increasing reproductive success (Robb, McDonald, Chamberlain, & Bearhop, 2008) and eco-tourism (Orams, 2002).Supplementary feeding to account for shortfalls in the availability of wild food resources is commonplace in agri-environment schemes, with active schemes in the UK for both winter and summer supplementary feeding for farmland birds.Providing food resources can result in less severe declines being observed in target species of over-winter supplementary feeding programs (Siriwardena et al., 2007), and increased breeding success as a result of earlier laying dates and increased fledging success (Robb, McDonald, Chamberlain, Reynolds, et al., 2008).However, there are important negative implications of larger inter and intraspecific aggregations at feeding sites, such as disease transmission (Sorensen et al., 2014).
for by cultivated seeds did not affect parasite presence, despite the spread of this parasite being associated with supplementary feeding due to an increase in contact between potentially infected and uninfected birds gathering at high densities at feeding sites.Significant dietary overlap was observed between infected and uninfected individuals in all sites, with almost complete dietary overlap being observed in one breeding site.The level of dietary overlap between individuals infected with specific strains fluctuated between sampling seasons, indicating no consistent relationship between diet and infection status.Higher densities of individuals in an area increase the probability of interaction and close contact, thus elevating the risk of direct disease transmission and pathogenic spillover (Lawson et al., 2011).
Close contact in the form of aggressive behaviors has been observed with animal aggregation at feeding sites in the banded mongoose (Mungos mungo), and is believed to be associated with increased Mycobacterium tuberculosis infection (Flint et al., 2016).Increased animal-to-animal contact when herds were corralled into areas for supplementary feeding, along with higher stress and poor hygiene, has also been linked to the simultaneous outbreak of three infectious diseases in a herd of semi-domesticated reindeer (Rangifer tarandus tarandus) (Tryland et al., 2019).Trichomonas galliane outbreaks have been reported in several different species of columbiforms.It is suggested that Trichomonas may have contributed to the rapid decline and subsequent extinction of the passenger pigeon, Ectopistes migratorius, after the introduction of the domestic pigeon to North America (Bucher, 1992;Stabler, 1954).More recently, a large outbreak of T. galliane infection was recorded in a wintering population of woodpigeons (Columba palumbus), where approximately 2600 woodpigeons died in a wintering site in Spain in 2001 (Höfle et al., 2004).It is suggested that the congregation of woodpigeons at game bird feeders may have contributed to the disease outbreak, as birds have been seen gathering at such feeders when yields of natural food resources have been poor.
Increased indirect pathogen transmission is common when animals congregate at feeding sites as a result of infective stages being shed in the environment from infected hosts, via routes such as feces and saliva (Murray et al., 2016).The length of time such stages can persist in the environment varies between organisms and with environmental factors (Turner et al., 2021;Walther & Ewald, 2004), but higher densities of wildlife at food sources will increase the likelihood of these transmissible stages being deposited and accumulating, facilitating an increase in indirect disease transmission (Murray et al., 2016).Furthermore, the abundance of food at a single site reduces the need to move to forage, encouraging sedentary behavior which contributes both to indirect transmission, via a build-up of infective agents in the environment, and increased chance of direct transmission through contact between individuals as animals are not dispersing, thus increasing exposure time (Murray et al., 2016).
With widespread habitat degradation observed across the world resulting in reduced availability of natural food sources, birds are increasingly feeding opportunistically at man-made sites and garden bird feeders (Browne & Aebischer, 2003;Dunn et al., 2018;Lawson et al., 2018;Robb, McDonald, Chamberlain, & Bearhop, 2008).For example, the turtle dove, an obligate granivore feeding on naturally occurring weed seeds, has lost a great deal of its natural food resources as a result of agricultural intensification, and a dietary switch has been observed, with this species now feeding on far higher levels of cultivated seeds than previously observed (Browne & Aebischer, 2003;Dunn et al., 2018).There is evidence of turtle doves feeding at supplementary feeding sites, as Dunn et al. (2018) detected hemp (Cannabis sativa), niger (Guizotia abyssinica), and sorghum (Sorghum bicolor), common components of supplementary seed mix, in the diet of turtle doves in the UK.In addition, Browne and Aebischer (2003) recorded turtle doves feeding primarily at anthropogenic food sources, such as spilt grain, animal feed and grain stores, and there are anecdotal reports of turtle doves being seen under garden bird feeders (Dunn et al., 2018).The use of feeding sites results in much higher rates of inter-and intra-specific interactions between birds than would occur under natural foraging behaviors (Lawson et al., 2018).This increased contact facilitates the spread of parasites, such as T. gallinae, and other wildlife pathogens, with a common factor observed in all geographic areas where trichomonosis has emerged being mortality where birds gather at feeding and watering sites (McBurney et al., 2015;Robinson et al., 2010).
As well as being commonplace in residential gardens, with millions of households provisioning food in this way every year (Robb, McDonald, Chamberlain, & Bearhop, 2008), supplementary feeding is commonly used by game managers to encourage high densities of species for hunting (Gortázar et al., 2006), and is an important tool used in the management of vulnerable species, particularly when habitat degradation, and the concurrent loss of food resources, has been a driver in a species' decline (Ewen et al., 2015).It is therefore crucial to consider this route of disease transmission so as not to introduce or increase the risk of disease as a threat to already declining species (Sorensen et al., 2014).
Supplementary feeding is commonly used for turtle doves, both as a management strategy to increase turtle dove population and as a means to facilitate the capture of birds for the purpose of research and population monitoring (Dunn et al., 2015;Thomas, 2017).Baiting is also used to attract game birds for hunting (Rocha & Quillfeldt, 2015).Turtle doves are widely hunted across much of mainland Europe and utilize bait which may be laid to attract them, or other game species (Rocha et al., 2022).Given the mounting evidence of disease transmission at supplemental feeding sites, as well as the association of T. gallinae with birds congregated at supplementary feed, this is an important consideration in plans to mitigate the decline of this species.Here, we therefore bring together two key elements of research to address the interaction between the diet of European turtle doves, and infection with T. gallinae.By using data obtained through highthroughput sequencing on the diet and infection with T. gallinae in turtle doves (Young, 2022), we investigated the relationship between the proportion of the diet consisting of cultivated crop seeds and the presence of T. gallinae, assessing whether there was any association between diet consumed and strain of T. gallinae detected, in particular, a positive association between infection with T. galliane and consumption of non-natural food resources, indicating the potential use of supplementary feed sites.

| Datasets
Dietary data and parasite infection datasets used in this study were collected from turtle doves from three countries, across both wintering and breeding grounds (Appendix 1: Figure A1) as part of a wider study (Young, 2022).A wintering population of turtle doves was sampled using mist nets at the Beer Sheba Project (Beer Sheba Project, 2002), an agricultural resource center encompassing a 100 ha Acacia woodland near Sandiara (14°22′ N, 16°48′ W), Senegal between November and March over four winters (2014/2015-2017/2018).Birds were caught at two different European breeding locations, in western Europe (France) and eastern Europe (Hungary) in order to sample geographically distinct populations using different migratory flyways (Marx et al., 2016), early in the breeding season.In France, samples were collected in spring 2017 (May 15-30) using baited potter traps from two mainland sites, comprising large areas of deciduous forest: Chizé Forest (46°07′ N, 0°25′ W) and Aulnay Forest (46°02′ N, 0°14′ W), and a more open and varied habitat on Ile d'Oléron (45°93′ N, 01°28′ W), an island just off the west coast (Appendix 1: Figure A2).In Hungary, birds were sampled in spring 2018 (May 20-June 7) using mist nets from Balotaszállás-Öttömos (46°16′ N, 19°35′ W), consisting of a mix of plantations and native forest (Appendix 1).
While there was variation in the surrounding environment, all sites were located within a landscape containing a mix of agricultural land and small towns or villages, often interspersed with natural grassland.
Full details of sampling, DNA extraction, and DNA sequencing can be found in Appendix 2. In short, for dietary analysis, fecal samples were collected from sterile bird bags in which birds were held after capture, and dried to preserve DNA (Rayé et al., 2011).DNA was extracted using either a QIAmp DNA Stool Mini Kit, or QIAmp Fast Stool Mini Kit (Qiagen), following the manufacturer's protocols with minor modifications to maximize DNA yield, as detailed in Dunn et al. (2016).Two extraction negatives, containing nuclease-free water instead of sample DNA, were included in every extraction batch to detect potential contamination.To test for T. gallinae infection, samples were collected from the oral cavity, esophagus, and crop of turtle doves using a sterile swab, which was then used to inoculate individually labeled InPouch™ culture kits (BioMed) and incubated at 37°C for 7 days to allow parasite proliferation.Following incubation, culture medium was preserved with an equal volume of 100% ethanol.All culture samples were processed using the following protocol, and the absence or presence of T. gallinae was confirmed by PCR.Culture medium was processed as per Riley et al. (1992) in order to remove culture medium from sample and isolate parasites if they are present.
Following the isolation step, all samples underwent the DNA extraction protocol, using one of two methods, either modified ammonium acetate method (Nicholls et al., 2000) or using a DNeasy Blood and Tissue kit (Qiagen).All samples were tested using PCR for the presence of T. gallinae DNA.
For dietary analysis, the second internal transcribed spacer of nuclear ribosomal DNA (hereafter ITS2) region was amplified using the UniPlant primer pair (F: 5′-TGTGA ATT GCA RRA TYCMG-3′; R: 5′-CCCGH YTG AYY TGR GGTCDC-3′) (Moorhouse-Gann et al., 2018), and for sequencing of T. gallinae, the ITS1/5.8S/ITS2ribosomal region of the genome (hereafter referred to as the ITS region) was amplified from oral swab samples using the TFR1/TFR2 primer pair (F: 5′-TGCTT CAG TTC AGC GGG TCTTCC-3′; R: 5′-CGGTA GGT GAA CCT GCC GTTGG-3′) (Felleisen, 1997).PCR products from oral swab samples were electrophoresed through a 1% gel, stained using SYBR®Safe (Thermo Fisher Scientific) in 0.5× TBE buffer and visualized using UV light to determine the infection status of birds.The presence of T. gallinae infection was indicated by the occurrence of a band of the appropriate amplicon size.Where no amplicon was observed, the sample was run through PCR a second time to confirm the absence of T. gallinae.
For sequencing, all primers were labeled with MID-tags (Multiple Identifiers) to provide a unique combination of tags for each DNA sample.Each plate of MID-tagged PCR product was combined into a single pool of uniquely identifiable samples, with approximately equal concentration per sample.Pools were cleaned using SPRIselect beads (Beckman Coulter) following the manufacturer's instructions, and the library preparation for Illumina sequencing was carried out on this pool using NEXTflex Rapid DNA-seq Kit (Bioo Scientific), following the manufacturer's protocols.The library was sequenced using a V2 chip, with 2 × 250 bp paired-end reads on an Illumina MiSeq sequencer (Illumina).Bioinformatic analyses were carried out using a custom pipeline designed for analysis of diet metabarcoding data from Illumina MiSeq (Drake et al., 2021).Full description of bioinformatics analysis is given in Appendix 3.
An additional dataset, collected for another study in the UK, was analyzed in order to incorporate the strain Type A, a high-virulence strain which was detected in the previous study in the UK, but did not occur in the samples collected in this study (Dunn et al., 2018;Thomas et al., 2022).Data were collected in the UK from 12 farmland sites across Essex, Suffolk, Cambridgeshire, and Norfolk in 2011-2014.Sampling and DNA extraction were conducted following the same protocols used in sample collection from Senegal, France, and Hungary, and full details of sampling are given in Dunn et al. (2018).
Due to the application of different bioinformatic processing pipelines and data cleaning in the two datasets, UK data were analyzed separately from data collected on the breeding grounds in this study.

| Dietary variation and infection status
All statistical analyses were carried out in R, version 4.0.1 (R Core Team, 2020).Due to differences in the amplification efficiency of DNA from different plant species, the presence or absence of each taxonomic unit within each sampling unit (frequency of occurrence), rather than read count, was used in all statistical analyses (Lamb et al., 2018).
An index of bird body condition was calculated using a GLM with Gaussian family and identity link function.Bird weight was the dependent variable, and wing length and time measurements were predictor variables, the latter to account for within-day variation in weight as an approximate linear increase in weight through the daily sampling period was observed in this study.Residuals from the model represented the condition of individual birds.To assess the relationship between diet and infection status, detrended correspondence analysis (DCA) was first carried out to summarize dietary data, based on the presence or absence of dietary taxa within each bird sampled, to capture the source of dietary variation in a small number of variables, to be included in a generalize linear model (GLM).Analysis was carried out using the "decorana" function in the R package vegan (Oksanen, 2020) and visualized using functions from the ggplot2 package (Wickham, 2016).
A high proportion of dietary items occur in a very low number of samples, which may mask co-occurrence patterns of plant taxa in dietary samples; therefore, DCA was conducted on a subset of data, containing only frequently consumed taxa.Four different thresholds representing how frequently items occurred in the diet were tested (taxa occurring in the diet of at least 3%, 5%, 7%, and 10% of birds sampled) to determine the most appropriate threshold for defining "frequently occurring taxa" (Appendix 4).DCA was conducted on each subset and the scores of samples inspected to assess for outliers within the sample.The optimum threshold for DCA was determined based on the absence of outliers in the data, as defined by Bendixen (2003).
Following selection of the appropriate threshold, the number of dimensions to be included as independent variables in the GLM was determined using a scree plot, illustrating the variation explained by the DCA dimensions (Appendix 5: Figure A4).A GLM was fitted with the presence or absence of T. gallinae as the dependent variable, using the binomial family and logit link function.Independent variables included in the model were season, body condition index, and the coordinate scores for DCA1 and DCA2.This model was repeated to assess individual strains of T. gallinae, with the presence or absence of GEO, Tcl-1, Type C, and Type IIIc being the dependent variable in four respective models.A reduced dataset containing 153 samples with genetic information for strain of T. galliane present was used for GLMs on specific strains.Model simplification was performed on all GLMs using the "step" function.

| Dietary overlap
Dietary overlap between infected and uninfected birds, and birds infected with different strains of T. gallinae, was visualized using nonmetric multidimensional scaling (NMDS).For the purpose of NMDS analysis, genera occurring in a single dietary sample were removed, as this analysis is liable to outliers skewing the results.The command "metaMDS" from the vegan package (Oksanen, 2020) was used to estimate differences between the presence of genera in individual samples, using Jaccard distance."Ordispider" was used to visualize results from nMDS via plotting with ggplot2.
Pianka's measure of overlap (Pianka, 1986) was used to assess the extent to which diet overlapped for uninfected and infected birds, and in birds infected with different strains.This measure is given as a value from 0, indicating no dietary overlap, to 1, indicating complete dietary overlap.Two matrices were created, including plant genera present in the diet, against either infection status (presence or absence of T. gallinae infection) or the presence or absence of each T. gallinae strain, recording counts of the number of times each genus was present in the diet of infected/uninfected birds, or birds infected with a specific strain present.Due to the segregation of dietary items between Europe and Africa, Pianka's overlap statistic was calculated for the breeding and wintering grounds separately.Average pairwise niche overlap was calculated using the "niche_null_model" command in the R package EcoSimR (Gotelli et al., 2015)) with 9999 repeats.Pairwise overlap was also assessed between specific strains.
Pianka's niche overlap was repeated as above in individual sampling seasons to assess whether patterns were consistent across years (Senegal) and between countries (Europe).Two sampling seasons, W2 and S2, were excluded from this analysis due to small sample sizes (N = 11 and 10, respectively), as small sample size increases the likelihood of underestimating the overlap (Linton et al., 1981).

| Source of food and Trichomonas gallinae infection
The effect of different food sources was assessed by considering dietary items categorized as "wild" or "cultivated" based on the likely source of food items (Appendix 6).Cultivated seeds included crop plants, animal fodder, and seeds common in supplementary feed, and wild were those naturally occurring in the environment.The effect of the proportion of wild seeds in the diet upon the infection probability by T. gallinae was assessed using a binomial GLM.The presence of T. gallinae infection was the dependent variable, and the independent variables were season and proportion of the diet accounted for by wild seeds.This model was repeated to assess individual strains, modeling the presence/absence of GEO, Tcl-1, Type C, and Type IIIc in turn as the dependent variable.
The presence of T. gallinae was also considered in the context of the dominant food type occurring in the diet.As quantitative diet data were not available, the proportion of taxonomic units detected in the diet classified as cultivated or wild was used as a proxy to indicate whether an individual's diet was dominated by cultivated or wild seed, where the majority of taxonomic units detected fell within these categories, or "even" in instances where the number of taxonomic units detected classed as wild and cultivated was the same.

| Dietary variation and infection status
A total of 190 birds sampled from the breeding and wintering grounds, from which diet and T. gallinae infection status data were successfully obtained were assessed to investigate the relationship between diet and parasite infection (Table 1).The overall prevalence of T. galliane infection across the dataset was 70%, but varied from 49% to 100% between seasons (Table 1).Five strains of T. gallinae were very rare within this study, occurring in just one (Tcl-TD, 26374943, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/edn3.402 by Cardiff University, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License GEO-TD3, Type IIIc-TD, ESWD), two (GEO-TD2), three (Ttl-TD), or four birds (GEO-TD), reflecting 0.6%, 1.1%, 1.7%, and 2.3% of the total number of birds sampled, respectively, and, due to their low occurrence, were not included in strain-specific analysis.Data from an additional 35 birds sampled in the UK as part of another body of work were analyzed (Dunn et al., 2018;Thomas et al., 2022).
When accounting for outliers in the dataset following threshold testing to remove uncommon taxa from analysis (Appendix 4), the threshold for the total proportion of samples within which a taxon had occurred was set at 7%, as this was the lowest threshold which did not introduce outliers, potentially skewing the results.All reported results are based on this data subset.
DCA1 explains the geographical separation of samples, with the majority of samples from Senegal having positive values on dimension 1, and birds sampled from France largely having negative values (Figure 1).Following the removal of outliers, only one sample from Hungary was retained in analysis.When visualizing the distribution of taxa consumed (Figure 2), there was an evident division between Europe and Africa.All seeds consumed in Africa, apart from Sorghum and Helianthus, had positive scores, and all seeds occurring only in Europe, but absent in Africa occurring in the negative pole of DCA1.Sorghum, which was commonly detected in both French and Senegalese samples, was close to 0 for DCA1, and Helianthus, which was consumed in Senegal, but less commonly had a more negative DCA1 score.There was less of a clear pattern to be observed in DCA2, but it does illustrate less variation in the diet of bird samples from France, with points being grouped more closely together, with positive values on dimension 2. Birds sampled in Senegal were distributed more widely across the poles, with greater dissimilarity between the points indicating more varied diet.
DCA1 and DCA2 explained the majority of the variation observed in this dataset; therefore, only they were included as explanatory variables in GLM analysis (Appendix 5).
The final dataset contained only one sample from Hungary, which was removed for GLM analysis following model assumption testing, as this point had leverage equal to zero, potentially disrupting the model.When assessing the relationship between infection status, season, and diet, season was the only variable to affect the presence of T. gallinae (GLM: LRT 4 = 33.494,p < 0.001).When considering individual strains, no variables were significantly associated with the presence of Tcl-1 or Type C, but season was significantly associated with the presence of GEO (GLM: LRT 4 = 48.809,p < 0.001)

| Dietary overlap
Pianka's niche overlap suggests significant dietary overlap between infected and uninfected individuals on the wintering and breeding grounds.The lowest overlap was observed on the wintering grounds, with almost complete overlap identified on the breeding grounds (Table 2).When visualizing dietary segregation, centroids are close together, supporting this result of little difference in the diet between infected and uninfected individuals (Figure 3).
When considering sampling seasons separately, the winter of 2015/16 (W2) and spring 2018 samples from Hungary (S2) had small sample sizes, and all individuals sampled in winter 2014/15 (W1) were infected, thus Pianka's niche overlap between infected and uninfected individuals could not be performed on these sampling seasons.However, when looking at years sampled in Senegal individually, a higher level of dietary overlap was detected than when looking at the data as a whole, with a relatively consistent level of dietary overlap between infected and uninfected individuals across the winters of 2016/17 (W3) and 2017/18 (W4) (Table 2).The dietary overlap was highest in France (S1), where almost complete dietary overlap was observed between infected and uninfected individuals and the lowest dietary overlap between infected and uninfected individuals was observed in data collected from the UK (Table 2).
Dietary composition was investigated in relation to strains present, considering the four dominant strains infecting turtle doves: GEO, Type C, Type IIIc, and Tcl-1, as well as uninfected individuals.
The average pairwise overlap was greater in the breeding grounds than in the wintering grounds (Table 3), but both exhibited reasonably strong overlap of dietary items consumed by individuals infected with different strains of T. gallinae.When visualizing dietary overlap, centroids representing different strains in both the breeding and wintering grounds are grouped together, supporting the high level of dietary overlap detected by Pianka's niche overlap (Figure 4).
This being said, in the breeding grounds, the strain GEO is segregated from the other points, suggesting that there may be some dietary segregation in this strain (Figure 4b).
Again, when considering sampling seasons separately, W2 and S2 were excluded due to small sample sizes.Within individual sampling seasons, low dietary overlap was detected in W1 and W3 (Table 3), supported by the distribution of centroids in Figure 5a,b.
Much higher dietary overlap between birds infected with different strains was observed in W4 and S1 (Table 3, Figure 5c,d).In the UK, strains assessed were GEO, Type A, Type C, and Tcl-1.
Two strains, Type IIIc and WQR-Env, were only detected in one bird from this population, and thus were removed from analysis as a rare strain, detected in 2.8% of the population sampled.Results for niche overlap in the UK were quite high, and a similar level to those observed in W4 and S1 (Pianka = 0.770, p < 0.001, SES = 5.479).When visualizing these data, the overlap of centroids observed in UK data is not as strong as that of samples from W4 and S1 (Figure 6).

| Source of food and T. gallinae infection
The proportion of diet accounted for by wild and cultivated seeds was considered in relation to infection with T. gallinae (N = 190).
Season was significantly associated with the presence of T. gallinae infection (GLM: Dev 5 = 33.608,p ≤ 0.001), as well as strain-specific associations for GEO (GLM: Dev 5 = 46.65,p ≤ 0.001) and Type IIIc (GLM: Dev 5 = 22.71, p ≤ 0.001).The proportion of wild seeds consumed was not associated with the presence of T. gallinae infection or on the presence of any individual strain of T. gallinae.
Overall, a higher proportion of individuals with a diet dominated by cultivated seeds were not infected with T. gallinae, whereas for both individuals with a diet dominated by wild seeds and those with an even balance of wild and cultivated genera in the diet, a higher proportion of birds were infected with T. gallinae (Figure 7).

| The relationship between rate of infection and consumption of cultivated seed
While T. gallinae is commonly transmitted during courtship, and the feeding of crop-milk to offspring, based on the findings of numerous previous studies implicating individuals congregating at supplementary feeding sites in the increased spread of disease (Dhondt et al., 2007;Fogell et al., 2019;McBurney et al., 2015;Pennycott et al., 1998), we investigated the associations between diet and the presence of T. gallinae.We hypothesized that infection with T. gallinae would be positively associated with the proportion of cultivated seeds in the diet.This is because a higher proportion of cultivated seed was expected to indicate increased use of supplementary food resources, and thus increased chance of contact with other individuals.We did not find evidence to support this, with no association being detected between diet and overall infection status, and high levels of dietary overlap being observed when comparing the diet of infected and uninfected individuals.However, the effect of supplementary feeding on infection should not be ruled out and remains a key consideration in actions involving provisioning of food resources.A number of potential explanations for the lack of association within the present study are discussed.
Over half of uninfected birds sampled had a diet dominated by cultivated seed expected to be obtained through supplementary feeding.This finding may be related to the types of foraging and to support high densities of game species for hunting, such as partridges and pheasants (Phasianidae), ducks (Anatidae), and doves and pigeons (Columbidae) (Gortázar et al., 2006;Millán et al., 2003).
It is possible that a relationship between supplementary food and infection was not observed in this study due to the nature of supplementary food delivery, as it did not result in the congregation of animals.Bait was used to attract individuals into potter traps in France.Once they entered the trap and triggered the door, solitary individuals were shut in traps until they were removed.
This contrasts with previously reported incidences of transmission at supplementary feeding stations, where large numbers of individuals congregate, facilitating disease transmission, as birds were not seen to congregate around potter traps.There may still have been an elevated risk of disease transmission at feeding sites, as T. gallinae has been shown to persist on seed resources for over 48 hours in laboratory conditions (McBurney et al., 2017), and has been detected in swab samples from seeds at a range of supplementary food sources (Thomas, 2017) and multiple birds to attend the same seed pile within a day.However, large aggregations of individuals at a single site did not occur as a result of supplementary feeding at this site.
In Senegal, birds feeding on cultivated seed were unlikely to be congregating at feeding sites, due to the high level of agriculture around the capture site in Senegal and the abundance of cultivated seeds in the environment, such as sorghum and millet, which were commonly detected in turtle dove diet in the wintering grounds (Young, 2022).It is therefore most likely that birds were obtaining this food resource from harvested fields, rather than supplementary feed sites.While birds do feed in groups and mix with other columbids (mourning dove [Streptopelia decipiens] and Namaqua dove [Oena capensis]) in harvested fields, these are spread over a wide area of arable land, rather than dense congregations observed at single bird feeding stations.It has been previously shown that T. gallinae was less likely to be detected from low-density supplementary feeding sites, such as sown seed plots, than high-density feed sites, like seed piles (Thomas, 2017).Therefore, it is likely that while birds may come into contact at a higher rate than if they were feeding on wild resources, large inter-and intra-specific groups of birds would not congregate at feeding sites where cultivated seed is available to the extent that they do at bird feeders on breeding grounds, reducing the risk of disease transmission in this instance.In wintering sites, such as sub-Saharan Africa, water sources are suggested to pose a higher risk of disease transmission, as water is generally scarcer, leading to high densities of birds congregating at watering holes, increasing both the likelihood of parasites being deposited in pools by infected hosts, and infective agents being picked up by uninfected hosts (Amin et al., 2014).Typically, when in Africa, turtle doves go to drink at first light or sunset, when large congregations of birds are observed (Browne & Aebischer, 2005).
In European sampling sites, it is unlikely that individuals were obtaining cultivated grain from agricultural fields as they were observed to be in Senegal, due to the timing of sampling.The two most prevalent cultivated crops in our sampling site in Senegal, millet and sorghum, are harvested from September to November (GIEWS, 2020), resulting in an abundance of spilt grain early in the wintering season, when many of the birds in Senegal were sampled.In contrast, turtle doves caught in Europe were sampled early in the breeding season and pre-harvest, before cultivated crops in the surrounding agricultural landscape would have matured.Therefore, such food resources would not be as readily available in the environment in European sampling sites as in Africa; thus, they most likely are from supplementary food sources in Europe.Anecdotal evidence from yearly sampling at the French field sites indicates that post-harvest, turtle doves do forage in harvested fields, as they are very difficult to catch, not coming into even baited traps when there is an abundance of seed in agricultural fields (personal observation).
Another potential explanation for the lack of association between consumption of cultivated resources and T. galliane infection may be a result of mild exposure to parasites, as has been observed in house finches in response to the pathogen Mycoplasma gallisepticum (MG) (Moyers et al., 2018).The rapid spread of this bacterial pathogen has been linked to use of garden bird feeders, with transmission via fomites deposited at bird feeders by infected individuals (Dhondt et al., 2007;Moyers et al., 2018).It has been identified in two studies that exposure to MG at feeding sites may be linked to a level of immune protection against more severe symptoms (Dhondt et al., 2007;Moyers et al., 2018).Dhondt et al. (2007) provided the first empirical evidence of transmission of this pathogen indirectly between birds, via pathogens deposited at feeding stations.This study identified that finches infected via this route developed milder cases of the disease, from which they were quicker to recover than when birds were directly exposed to MG via infectious swabs, indicating that indirect transmission from pathogens deposited at supplementary feeding sites can result in lower levels of infection (Dhondt et al., 2007).Moyers et al. (2018) investigated the relationship between bird feeder density and infection with MG, finding that a higher density of bird feeders would result in individuals staying at feed sites for longer, increasing pathogen transmission.They also observed that treatments with a lower density of feeders available resulted in birds having higher concentrations of MG-specific antibodies than those fed at high densities, but with very low rates of detectable infection (Moyers et al., 2018).This suggests that birds in the low feeder density treatments may have been exposed to subclinical doses of MG, leading to them acquiring an immune response against this pathogen.
There is evidence that prior exposure to milder strains of T. galliane can infer protection against future infection.In an experimental study, Stabler (1948) inoculated rock doves with T. gallinae strains of varying virulence and found that no birds with prior exposure to less virulent strains succumbed to infection with the second, most severe strain, supporting the idea that previous infection acted as an immunization and protected individuals from future infection.It is therefore possible that birds feeding at potter traps in France, which are low-density feeding stations, spread throughout the forest, may have been exposed to low levels of T. gallinae when feeding here, potentially resulting in some level of immune protection.Similarly, birds feeding on cultivated seeds in Senegal do forage in groups, but in lower densities than at supplemental feeders as seeds are dropped in harvested fields, again leading to the possibility of lowlevel exposure to T. gallinae.However, this is very difficult to assess in wild birds, unless the same birds are recaptured and tested on a regular basis, providing information on rate of recovery from infection and a history of prior infection.One aspect that the majority of studies of T. gallinae on wild birds have in common, the present study included, is that they are only able to provide a snapshot of the infection status of birds carrying T. gallinae (Forzan et al., 2010;Lawson et al., 2006;Marx et al., 2017;Sansano-Maestre et al., 2009;Thomas, 2017) As well as these potential behavioral explanations, it is possible that birds in better condition as a result of utilizing abundant cultivated food resources were more capable of fighting off infection, as associations have been shown between immune system functioning and body condition (Becker & Hall, 2014;Chandra, 1999;Møller et al., 1998).Previous dietary analysis on these data (Young, 2022) detected a positive correlation between body condition and the proportion of diet accounted for by cultivated seed.Multiple other studies investigating the effects of supplemental feeding have reported improvement of health metrics of birds, including increased antioxidant levels, more rapid feather growth, reduced stress indicators, improved body condition index, improved innate immune defense, and improved fat score (Galbraith et al., 2016;Wilcoxen et al., 2015).
Furthermore, some commonly fed seeds, such as Helianthus sp., have a high calorific value (Hullar et al., 1999), which may help improve body condition by alleviating nutritional stress (Becker & Hall, 2014).
These findings support the suggestion that use of supplementary food may improve body condition, and therefore boost immune function, reducing host susceptibility to T. gallinae.However, it is important to note that, while these studies did demonstrate positive impacts of supplementary feeding on hosts, they also both showed significant increase in parasite and disease transmission associated with supplementary feeding.While this association between body condition and infection status was not supported in the current study, it is still a key consideration that should be addressed when providing supplementary food (Galbraith et al., 2016;Wilcoxen et al., 2015).

| Trichomonas gallinae strains in relation to diet
There was variation in both the T. gallinae strain composition and the dietary composition between the sites sampled in this study (Young, 2022); it was hypothesized that this may be caused by an association between certain strains and food types.Previous work identified four strains of T. gallinae (Type A, Type C, GEO and Tcl-1) detected at supplementary food sites from the UK (Thomas, 2017).
Type A was most commonly detected in food resources, followed by Tcl-1, indicating a possible link with these strains and survival on supplementary food resources.However, when considering the association of specific strains of T. gallinae with dietary items, we found little support for any association, with significant dietary overlap between infected and uninfected individuals.In addition, when investigating the dietary overlap of birds infected with different strains per sampling season, no consistent patterns were observed across season, with lower dietary overlap being observed in 2 years in Senegal, and high dietary overlap occurring in 1 year in Senegal, and in France.This lack of support for dietary items being linked to specific strains suggests that variation in the strains previously detected in supplementary food resources may reflect another unknown environmental factor or simply reflect variation in the parasite community composition, rather than an association with dietary items.
While we were unable to find evidence supporting a link between diet and infection with T. gallinae, the results of this study potentially provide support for advice on supplementary feeding.As discussed, there are potential benefits of alleviating nutritional stress through supplementary feeding, which may improve body condition, such that improved immune function can fight off disease.While it is important not to overlook the trade-off between potential disease transmission and nutritional benefits, it is possible that the mode of delivery of supplementary feed is important in reducing the impact of disease transmission.For example, distributing supplementary seed over a wide area, similar to dropped grain in harvested fields in Senegal, may reduce the risk of disease transmission at bait sites.were included as controls (Taberlet et al., 2018).

How to cite this article
Negatives were arranged within the plate such that every MIDtag had a negative control, facilitating identification of any contamination in primers.Two wells per plate were left empty to ensure that certain MID-tag combinations were not used at all during PCR amplification to allow quantification of tag jumping during the sequencing process.Replicate samples were also incorporated for quality control purposes, with a total of 16 samples to assess the consistency of sequencing.
PCR amplification was carried out using SimpliAmp™ Thermal Cycler (Thermo Fisher Scientific), using the following thermocycling protocol: 95°C (15 min); 40 cycles of 95°C (30 s), 58°C (90 s), 72°C In preparation for high-throughput sequencing, following pooling of samples, the quality of the pooled samples and the size of amplicons were checked via TapeStation2200 (Agilent).Pools were cleaned using SPRIselect beads (Beckman Coulter) following the manufacturer's protocols, at a ratio of 1:1.1, to retain fragments of the desired size range (207-400 bp).Once cleaned, the concentration of DNA was measured for each pool using a Qubit dsDNA High-sensitivity Assay (Thermo Fisher Scientific) and pools were combined into a single equimolar pool.The library preparation for Illumina sequencing was carried out using NEXTflex Rapid DNA-seq Kit (Bioo Scientific), following the manufacturer's protocols.

INFECTION
Parasite isolation from culture medium was performed using protocols adapted from Riley et al. (1992)  DNA was extracted from isolated parasites using two methods, either modified ammonium acetate method (Nicholls et al., 2000) or using a DNeasy Blood and Tissue kit (Qiagen) depending on the date of extraction.For the ammonium acetate method, the pellet was obtained from 2.5 mL of culture medium, for extraction using  and centrifuged at 13,000 rpm for 10 min.The supernatant was discarded and 500 μL 70% ethanol was added to the DNA pellet before being spun at 13,000 rpm for 5 min to wash the pellet.
The supernatant was discarded and the pellet was left to air-dry until no ethanol remained, typically taking 1-2 h.The DNA pellet was dissolved in 20-50 μL of low TE buffer (0.1 mM EDTA, 10 mM Tris), depending on the size of the pellet, and incubated at 65°C For this reason, PCR products were electrophoresed and samples grouped together based on band brightness (faint, medium and strong).The average DNA concentration, per band brightness group, was calculated from the QIAxcel results, and the pooling calculation was carried out again, using the average concentration for strong bands as C max .This gave the volumes to be added for all samples based on band brightness grouping.The average of all sample volumes added to the pool was calculated, and this was the volume added for all negative samples.
Cleaning of the pools and library preparation for Illumina sequencing was conducted as for DNA extracted from fecal samples, with one modification: in the initial cleaning of the pools using SPRIselect beads (Beckman Coulter), a ratio of 1:0.9 was used to remove fragments smaller than the desired 420 bp.

A PPE N D I X 3 BIOINFORMATIC ANALYSIS AND DATA CLEANING
Bioinformatic analyses were carried out on both diet and parasite infection datasets using a custom pipeline designed for analysis of diet metabarcoding data from Illumina MiSeq (Drake et al., 2021).
Data were first checked for truncation in the MID-tags, using the "grep" function to extract information of the number of reads with and without the full 10 base pair MID-tag before the primer.FastP (Chen et al., 2018)  mismatch was permitted; for DNA extracted from oral swabs, no primer mismatches were permitted, in order to prevent inflation in the number of OTUs generated and lead to the retention of higher quality reads, which are more likely to represent true variation (Thomas, 2017).This was necessary for analyzing parasite samples as even one base mismatch leads to the assignment of a new strain of Trichomonas gallinae.Reads were then clustered based on zOTUs, also in Unoise3.Clustering based on zOTUs uses a 100% clustering threshold, as opposed to the standard 97%, assuring high resolution and preventing variants being incorrectly collapsed into clusters (Edgar, 2016).BLASTn v2.7.1 (Camacho et al., 2009) was then used to assign sequences to their closest matching taxa from GenBank, with the minimum percent identity set to 95% for DNA from fecal samples and 97% for DNA from oral samples in order to remove poor matches.BLAST results were filtered using the dplyr package in R to retain only the top bit score for each zOTU, calculated by combining the percent ID and e-value to indicate how good the alignment is.
These BLAST outputs were analyzed using MEGAN v6.15.2 (Huson et al., 2016) to assign taxonomic names to each zOTU.In instances where there was more than one top hit, zOTUs would be assigned to the highest common taxonomic rank.
Data processing and cleaning following this initial bioinformatic processing differed for DNA extracted from fecal samples for diet analysis and DNA extracted from oral swabs for T. gallinae analysis.
For DNA from fecal samples, following taxonomic assignment, results were manually checked to identify zOTUs not assigned to species level, generally resulting from variations in nomenclature or erroneous sequences on GenBank.Where possible, sequences were assigned to species level; however, where there were multiple top hits on GenBank, they were assigned to the highest taxonomic rank in common.The geographic range of all taxa identified was checked using Kew Science "plants of the world online" (POWO, 2019).
Fifteen species assignments were identified which did not occur within the sampling range of the present study, and thus were assigned the most appropriate taxonomic level occurring within the study region.The final step was aggregation, whereby all zOTUs that had been assigned to the same taxon were aggregated in R using the "aggregate" function.
For DNA from oral swab samples, any sequences that did provide a 100% match to published T. galliane strains were assigned a new strain identifier, based on the closest published match.

Cleaning the dataset
There were two stages to the cleaning of data to remove artifacts from the final dataset, which cause false positives.The two stages of filtering complement each other to provide a more comprehensive method of data cleaning (Drake et al., 2021).
The first stage of cleaning was to remove contamination.
Contamination can occur at two stages: in the laboratory, exogenous contaminants may be introduced into the samples during extraction or PCR, or internal contamination, such as tag jumping, during library preparation or sequencing (Taberlet et al., 2018).In order to remove potential false positives occurring as a result of exogenous contamination, the highest read count in negative controls was identified for each taxon.The same was done for unused MIDtag combinations in order to account for tag-jumping.Theoretically, both negative controls and unused MID-tag combinations should contain no DNA, and thus any reads occurring within these samples are a result of contamination (Drake et al., 2021).The highest read count (out of either negative control of unused combinations) was designated as the maximum contamination per taxon, and any read count below this maximum contamination threshold was considered exogenous contamination and read count was replaced with zero.By removing read counts below the threshold, as opposed to subtracting the maximum contamination across the board, the overall read count of samples exceeding the contamination threshold was not affected.
The second stage is intended to remove internal contaminants: artifacts created during PCR and sequencing, and is carried out on a per-sample basis.Artifacts generally occur in far lower proportions than genuine sequences, and thus are removed by filtering out low-abundance OTUs.Despite every effort being made during the pooling process to make samples equimolar, thus limiting variation in read depth amongst samples, this remains an inevitable occurrence.This variation in read count between samples can have strong implications for the efficacy of data filtering if a single threshold is set based on raw read count, such as removal of samples with fewer than 10 reads, a commonly cited filtering method (Drake et al., 2021).With this method, artifacts are more likely to pass filtering thresholds in samples with higher read counts, as artifacts tend to occur in greater numbers at higher sequencing depths (De Barba et al., 2014;Drake et al., 2021).In order to overcome this, thresholds were set based on frequency of occurrence of taxa within samples.Following this proportional approach accounted for variation in read depth between samples (in this case, the read depth per sample range from 4276 to 172,171) facilitating a standardized clean-up across all samples.
In order to designate a filtering threshold for DNA extracted from fecal samples, the positive controls were used, each of which contained one of three known plant species, endemic to Mauritius, which would therefore not occur in samples from this study.Samplebased filtering thresholds were assigned by looking at how often non-target DNA occurred within the positive control.Different sampling thresholds were tested to assess the efficiency of removal of false positives from positive controls.In order to balance false positives and false negatives, a slightly more conservative threshold of 4% was selected, which removed most false positives from positive controls, without running the risk of removing too many genuine reads.This 4% threshold was applied across the dataset in a sample filtering basis, and read counts for any taxa failing to exceed 4% of total reads of a sample were replaced with zero.
These two filtering methods were combined and only read counts exceeding both the maximum contamination threshold and the 4% per sample threshold were retained for further analysis.Following data filtering, fungal and bacterial taxa were removed, as were any taxa which were not identified to at least family level.To further clean the data, a cumulative depth curve was generated for each sample using the top 10 variants occurring in each sample (Lighten et al., 2014;Thomas, 2017).Based on the assumption that artifacts occur in much lower numbers than genuine reads, the presence of a clear inflection point (as seen in Appendix 3: Figure A3a) indicated the divide between a true sequence and an artifact, and any reads occurring beyond this inflection point would be considered artifacts and not included in further analysis (Lighten et al., 2014;Thomas, 2017).This was carried out for all samples to determine the percentage of reads accounted for beyond this inflection point.In some instances, determining an inflection point was more subjective (Appendix 3: Figure A3b), where there was one stronger inflection point, followed by a less clear one.Finally, in a small number of cases, it was not possible to discern a clear inflection point (Appendix 3: -throughput sequencing, parasitology, supplementary feeding, Trichomonas galliane, turtle dove | 3 YOUNG et al.

F
I G U R E 1 Biplot illustrating the variance accounted for by DCA1 and DCA2, with the associated eigenvalues of 0.827 and 0.568, respectively.Shaded ellipses represent 95% confidence interval for countries with more than one sample.and Type IIIc (GLM: LRT 4 = 13.5121,p = 0.009).The variables explaining dietary variation (DCA1 and DCA2) were not associated with the presence of any strain of T. gallinae.

F
I G U R E 2 Biplot showing the distribution of dietary taxa across DCA1 and DCA2.TA B L E 2 Pianka's niche overlap analysis on the similarity of diet between individuals infected and uninfected with Trichomonas galliane within difference sampling season.

F
Spider plot derived from non-metric multidimensional scaling of genera consumed by turtle doves (a) in the wintering grounds, differentiated by status of infection with Trichomonas galliane.Stress = 0.07, N = 122; (b) in the mainland European breeding grounds, differentiated by status of infection with T. gallinae.Stress = 0.07, N = 71.Ordination of multidimensional data condenses information into two-dimensional axes, representing variation in the diet.Each small point is representative of the diet of a single turtle dove sampled, and the distance between the points is representative of the similarity of diet composition between individual (a shorter distance between points indicates greater similarity between diets).The larger points forming the centroid of each group (season) represent the mean coordinates per group.

F
Spider plot derived from non-metric multidimensional scaling of genera consumed by turtle doves across birds sampled (a) in the wintering grounds, differentiated by the strains of Trichomonas galliane present.Stress = 0.066, N = 110; (b) in the mainland Europe breeding grounds, differentiated by the strains of T. gallinae present.Stress = 0.086, N = 63.While there is a large body of research reporting the increased transmission of disease at supplementary food sites, this study specifically investigated the relationship between diet and infection status of individuals.Very few significant associations were found between infection with T. gallinae and diet.Dietary variation was not associated with the probability of T. gallinae infection, and significant dietary overlap was observed between infected and uninfected individuals.When considering specific strains, there was again no significant dietary segregation identified.

F
Spider plot derived from non-metric multidimensional scaling of genera consumed by turtle doves across birds sampled in the breeding grounds, differentiated by the strains of Trichomonas galliane present.(a) Winter 2014/15 (W1), (b) winter 2016/17 (W3), (c) winter 2017/18 (W4), (d) spring, France (S1).26374943,0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/edn3.402 by Cardiff University, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensemethod of supplementary food delivery.The rationale for our expectation that birds feeding on a greater proportion of cultivated seeds would be more likely to be infected with T. gallinae is based on previously published associations between this parasite and congregation of birds at food and water sources (McBurney et al., 2015; Robinson et al., 2010; Thomas, 2017).This congregation behavior is commonly observed at artificial food sites, such as garden bird feeders (McBurney et al., 2015), and supplementary feeding sites, intended F I G U R E 6 Spider plot derived from non-metric multidimensional scaling of genera consumed by turtle doves across birds sampled in the UK, differentiated by the strains of Trichomonas galliane present.Stress = 0.246, N = 34.F I G U R E 7 Proportion of total birds (N = 190) infected or uninfected with Trichomonas galliane, where the taxonomic units detected in the diet were dominated by wild or cultivated seeds, or with an even split of wild and cultivated food sources.Calculated based on the proportion of dietary items consumed by individuals accounted for by wild seeds.Infected birds, n = 134, uninfected birds, n = 56.

(
90 s); 72°C (10 min).Concentration of amplified DNA was determined by testing 5 μL of PCR product using a DNA fast analysis cartridge with QIAxcel (Qiagen).Each plate of MID-tagged PCR product was combined into a single pool of uniquely identifiable samples, with approximately equal concentration per sample.Concentrations from the QIAxcel were used to calculate the volume of each PCR product to be added to the final pool, with plates being pooled individually due to unavoidable variation between QIAxcel runs.The maximum DNA concentration, C max , was determined for each plate, and divided by the concentration of each sample to produce the volume to be added to the pool.The volume of negative samples to F I G U R E A 1 Turtle dove sampling locations in Europe and Senegal.26374943, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/edn3.402 by Cardiff University, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License add to the pools was the average of all sample volumes added to the pool.
and are as follows: a volume of culture medium was centrifuged at 3200 rpm for 5 min at 4°C and supernatant discarded.The remaining pellet was washed in 500 μL phosphate-buffered saline (PBS) by centrifugation, at 3200 rpm for 5 min at 4°C and supernatant discarded.The rinsing stage was repeated to wash the pellet a second time.The pellet was resuspended in 200 μL PBS and stored at −20°C until DNA extraction was carried out.

DNeasy
Blood and Tissue kit, 200-400 μL of culture medium was used.Samples from the UK dataset, and DNA extracted in 2018 used the modified ammonium acetate method, whereby isolated parasite samples were centrifuged at 3200 rpm for 5 min, the supernatant discarded and the pellet retained.250 μL of Digsol buffer (20 mM EDTA, 50 mM Tris, 120 mM NaCl, 1% SDS, pH 8.0) and 20 μL Proteinase K (Qiagen) were added and samples were incubated overnight at 37°C to allow digestion.Following digestion, 300 μL of 4M ammonium acetate was added, mixed by vortexing, and incubated at room temperature for 30 min, vortexing every 10 min.Samples were centrifuged at 13,000 rpm for 10 min and the supernatant transferred into a new tube, discarding the pellet.A volume of 1 mL 100% ethanol was added to the supernatant to allow precipitation of the DNA.Each sample was vortexed F I G U R E A 2 Sampling sites in France.
until the DNA was fully resuspended.Extracted DNA was then stored at −20°C.For the Blood and Tissue kit, extraction was carried out following the manufacturer's instructions for extraction from tissue samples.The final elution was in 50 μL TE buffer.Two extraction negatives were included for every extraction, whereby all reagents were added but no sample was included in order to detect contamination in extraction buffers.Polymerase chain reaction was used to determine the infection status of each sample.The TFR1/TFR2 primer pair(Felleisen, 1997) were used to amplify the 400 bp length target region of the ITS1/5.8S/ITS2ribosomal region of the genome (hereafter referred to as the ITS region).The PCR recipe used consisted of 1.8 μL Multiplex PCR Master Mix (Qiagen), 0.225 μL of 10 μM forward and reverse primers (Merck KGaA), and 2.25 μL nuclease free water, giving a reaction volume of 4.5 μL, to which 0.5 μL DNA was added.A negative control of molecular-grade water and a positive control known to contain T. gallinae DNA were included in each PCR run.The following touchdown PCR thermal cycling program was performed using SimpliAmp™ Thermal Cycler (Thermo Fisher Scientific).The program consisted of initial denaturing for 15 min at 95°C, 11 cycles of 1 min at 94°C, 30 s at 66°C (decreasing by 1°C every cycle until 56°C), 1 min at 72°C, followed by 24 cycles of 1 min at 94°C, 30 s at 55°C, and 1 min at 72°C, with a final elongation step of 10 min at 72°C.PCR products were electrophoresed though a 1% gel, stained using SYBR®Safe (Thermo Fisher Scientific) in 0.5× TBE buffer and visualized using UV light.The presence of T. gallinae infection was indicated by the occurrence of a band of the appropriate amplicon size when visualized.In instances where no amplicon was observed, the sample was run through PCR a second time to confirm the absence of T. gallinae.Samples which tested positive for T. gallinae were prepared for high-throughput sequencing by making up 25 μL reaction volume consisting of 10 μL Multiplex PCR Master Mix (Qiagen), 9 μL nuclease free water, 2.5 μL each of 3 μM TFR1[f] and TFR2[r] primers (Eurofins), and 1 μL template DNA.Both forward and reverse primers were labeled with MID-tags (Multiplex Identifiers) to create unique labels for individual DNA samples, with all samples having a unique MID-tag combination.Each PCR 96-well plate included 12 negative controls (11 extraction negatives and one PCR negative, containing 1 μL nuclease-free water), arranged as for fecal samples.Replicate samples were also incorporated for quality control purposes, with 10% of the total sample number being included as replicates.PCR amplification and confirmation of successful PCR reaction were carried out following the touchdown PCR and gel electrophoresis protocols described previously.Each plate of MID-tagged PCR product was combined using a modification of the method described for diet analysis, based on the maximum DNA concentration per plate, using known concentrations determined by QIAxcel (Qiagen), to produce one equimolar pool of uniquely identifiable samples per 96-well PCR plate.Due to the wide range of concentrations of ITS amplicons (0.18-40.5 ng/μL), the original method would have resulted in a large number of samples being excluded from the pool due to insufficient PCR product in samples with a concentration far lower than C max .

For
T. galliane DNA isolated from oral swabs, no positive controls were included; therefore, the filtering threshold was set at 0.3%, as per as per Taberlet et al. (2018).After removing maximum contamination and low proportion reads from samples, there were still reads present in very low numbers, indicating the retention of some artifacts or polymorphic gene copies.An additional filtering step was therefore required for T. gallinae DNA.This is because the genomic marker widely used in studies of Trichomonas species (ITS region) encodes the small 5.8S rRNA subunit (SSU), as well as the flanking internal transcription spacers.In all known animal genomes, rRNA occurs with multiple copies per cell, with some level of polymorphism between copies (Porazinska et al., 2010).Bearing in mind this polymorphism, slight variation in sequences produced by HTS may be produced by a different copy of an rRNA gene, as all reads generated during sequencing are returned (Porazinska et al., 2010).Porazinska et al. (2010) investigated the repeatability of using HTS to analyze nematode mock communities, also targeting SSU rRNA.By using mock communities, all expected sequences were known, and about half of the reads generated in the dataset differed from the presumed consensus sequence by at least 1 bp.Without extensive use of technical repeats, these polymorphisms cannot be distinguished from artifacts, but are likely to account in part to the high number of zOTUs differing from known sequences returned from sequencing in this study.

Figure
Figure A3c).In instances where no clear inflection point could be determined,Lighten et al. (2014) discarded the sequence as poor quality.In this study, such instances may be a result of infection with multiple strains, and thus, the maximum percentage accounted for by a clear inflection point was used as the threshold for all samples to determine co-infection.Different thresholds were tested for cleaning steps two and three.Analyses were conducted on datasets generated using different thresholds to assess whether threshold values effected the 26374943, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/edn3.402 by Cardiff University, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Pianka's niche overlap analysis on the similarity of diet between individuals infected with different strains of Trichomonas galliane (GEO, Type C, Type IIIc and Tcl-1) within difference sampling season.Note: W3 and W4 refer to the wintering seasons in Senegal of 2016/17 and 2017/18, respectively, and S1 refers to spring sampling season in France (2017).
TA B L E 3 Bunbury et al. (2008)rk carried out byBunbury et al. (2008)on the Mauritian pink pigeon, which repeatedly tested birds over a 2-year period for the presence of T. gallinae, finding evidence of an increase in infection rate with increased host age and a negative impact on survival in birds consistently testing positive for this parasite.However, many questions regarding the dynamics of this parasite infection remain unanswered.It is unclear how rapidly birds are able to clear infection of this parasite, how likely they are to become re-infected and within what time frame this may occur, and if this varies between species of host and parasite strains.Greater monitoring of disease and these host-parasite dynamics would help shed further light on disease transmission, and potential risk factors, such as congregation at supplementary feeding sites, and help inform management plans for declining species.
26374943, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/edn3.402 by Cardiff University, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License , 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/edn3.402 by Cardiff University, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PCR errors and chimeras.Dereplication was performed using the fastx_uniques command.Denoising of data, performed by means of a UNOISE algorithm specifically designed for use on Illumina sequences, recovers correct biological sequences from reads obtained, referred to as zero-radius operational taxonomic units (hereafter, zOTUs).For DNA extracted from fecal samples, one primer 26374943