Species, sex and geographic variation in chlamydial prevalence in abundant wild Australian parrots

Chlamydia psittaci (order: Chlamydiales) is a globally distributed zoonotic bacterium that can cause potentially fatal disease in birds and humans. Parrots are a major host, yet prevalence and risk factors for infection in wild parrots are largely unknown. Additionally, recent research suggests there is a diverse range of novel Chlamydiales circulating in wildlife. We therefore sampled seven abundant parrot species in south-eastern Australia, taking cloacal swabs and serum from n = 132 wild adults. We determined C. psittaci and Chlamydiales prevalence and seroprevalence, and tested for host species, sex, geographical and seasonal differences, and temporal changes in individual infection status. Across all species, Chlamydiales prevalence was 39.8% (95% CI 31.6, 48.7), C. psittaci prevalence was 9.8% (95% CI 5.7, 16.3) and C. gallinacea prevalence was 0.8% (95% CI 0.1, 4.5). Other Chlamydiales species were not identified to species level. We identified two C. psittaci strains within the 6BC clade, which is highly virulent in humans. Seroprevalence was 37.0% (95% CI 28.5, 46.4). Host species (including crimson rosellas, galahs, sulphur-crested cockatoos and blue-winged parrots) differed in seroprevalence and Chlamydiales prevalence. Galahs had both highest Chlamydiales prevalence (54.8%) and seroprevalence (74.1%). Seroprevalence differed between sites, with a larger difference in males (range 20–63%) than females (29–44%). We reveal a higher chlamydial prevalence than previously reported in many wild parrots, with implications for potential reservoirs, and transmission risks to humans and other avian hosts.


Scientific Reports
| (2020) 10:20478 | https://doi.org/10.1038/s41598-020-77500-5 www.nature.com/scientificreports/ relationship between infection and host body condition, and (4) how chlamydial infection status changes over time in recaptured individuals. Our goal was to improve knowledge about chlamydial prevalence in a wild parrot community, including risk factors for infection, and to identify whether these species may be a reservoir of chlamydial infection for other wild and domestic bird species, humans and other mammals.

Results
Identification and prevalence of Chlamydiales. Across all host individuals tested, and taking the first capture of every individual (Supplementary Tables S1-S2), mean total Chlamydiales PCR prevalence was 39.8% (95% CI 31.6, 48.7; 49/123 positive), C. psittaci PCR prevalence was 9.8% (95% CI 5.7, 16.3; 12/123 positive), and C. gallinacea PCR prevalence was 0.8% (95% CI 0.1, 4.5; 1/123 positive). Including all recaptures (n = 179), there was a total of n = 67 Chlamydiales positive sampling events. Of 14/67 samples positive for C. psittaci, n = 4 were identified by species-specific PCR, and n = 10 identified by sequencing, identified from BLASTn analysis using both the non-redundant nucleotide (nr/nt) database and 16S ribosomal RNA database (Table 1; Supplementary  Table S3). One positive sample was identified as C. gallinacea, by using the gidA and CTU/CTL primers (individual reported in a previous study 41 , where ompA sequencing was carried out as confirmation). We sequenced 19 of the 52 remaining unknown Chlamydiales positive samples, of which n = 11 were successfully sequenced: n = 2 samples were identified as Parachlamydiaceae, n = 2 samples represent potentially novel species within the Chlamydiales, and n = 7 samples were mixed chlamydial infections, confirmed as Chlamydiales from BLASTn analysis using the nr/nt and 16S databases. Partial sequences of the ompA gene from two C. psittaci samples indicated that the C. psittaci genotypes grouped most closely with genotype A and the 6BC clade (Supplementary Figure S1). All 16S and ompA accession numbers obtained are provided in Table 1. Seroprevalence (according to the ImmunoComb assay) was 37.0% (95% CI 28.5, 46.4; 40/108 individuals were positive).
Testing for host species differences in prevalence. Chlamydiales, including C. psittaci, were identified in all four focal host species except for blue-winged parrots (Fig. 1a). When analysing birds caught in walk-in traps only, we did not identify C. psittaci or Chlamydiales in the three species for which we had smaller sample size, namely eastern rosellas (Platycercus eximius) (n = 3), rainbow lorikeets (n = 2) and red-rumped parrots (n = 1). However, when including recapture data, we did find one breeding eastern rosella testing positive for Chlamydiales. Seropositive individuals were found for all four focal species (Fig. 1b). Of the remaining three species, one eastern rosella (33.3%; 95% CI 6.1, 79.2 [1/3 positive] assayed seropositive. Host species differed in Chlamydiales prevalence (p = 0.005) and seroprevalence (p < 0.001), but not in C. psittaci prevalence (Fig. 1, Table 2). Post-hoc pairwise comparisons revealed that galahs had significantly higher seroprevalence ( . 2c). There was a significant interaction between sex and field site on seroprevalence, with a greater difference in male seroprevalence compared to female seroprevalence (p = 0.041, Table 2, Fig. 2d).
There was a trend for higher Chlamydiales prevalence in summer (83% [95% CI 43.6, 97.0; 5/6 individuals positive) and higher seroprevalence in autumn (67% [95% CI 46.7, 82.0; 16/24 positive). However, the effect of season on Chlamydiales prevalence, C. psittaci prevalence and seroprevalence was not significant ( Table 2). Time of day had no effect on Chlamydiales or C. psittaci prevalence when all host species were tested. However, in crimson rosellas, birds were marginally more likely to test positive for C. psittaci in the morning compared to later in the day (p = 0.049; Supplementary Table S5).

Relationship between host infection status and body condition.
There was no effect of Chlamydiales or C. psittaci PCR prevalence, or seroprevalence, on body mass, packed cell volume (PCV; haematocrit) or residual body mass, although seropositive birds tended to have lower body mass (p = 0.051) and there was a tendency for Chlamydiales positive (p = 0.076) and seropositive (p = 0.071) birds to have lower PCV (Sup-  Recaptures. There were n = 39 birds which were caught more than once (30 crimson rosellas, six galahs, and three eastern rosellas). Birds were caught a maximum of three times, and the mean interval between capture events was 214.  www.nature.com/scientificreports/ assayed positive first then negative at recapture(s), and 7% (1/14) changed status twice (although only n = 5 birds were caught three times). Birds recaptured in a different season were more likely to change in Chlamydiales status compared to birds recaptured in the same season (χ 2 = 3.904, df = 1, p = 0.048). In crimson rosellas, Chlamydiales status at first capture did not predict infection status at recapture (Supplementary Table S11). Serostatus did not change in any birds tested for seroprevalence on multiple captures ( Table 3). The mean interval between seropositive capture events was 78 days (± 46.8 SD, range 13-126) and the mean interval between seronegative capture events was 242.6 days (± 103.8 SD, range 14-436). Of four birds which assayed positive for C. psittaci, 75% (3/4) were seronegative, and did not assay positive for C. psittaci upon recapture (Table 3).

Discussion
Prevalence data and risk factors for chlamydial infection in wild parrot populations are severely lacking, despite the disease risk these bacteria can pose to several avian hosts, and the potential risk of zoonotic transmission to humans. Furthermore, while Chlamydiales species outside the Chlamydia genus have been found in birds 12,43 , the overall prevalence of this bacterial order has rarely been investigated in a wild avian population or community. We identified 40% Chlamydiales prevalence in wild parrots, which is lower than that previously reported in wild mammals, namely marsupials (48%) 13 and deer (72%) 44 . Our results provide further evidence that chlamydial infections are widespread in wildlife 6 . Chlamydiales were detected by PCR in crimson rosellas, galahs, sulphur-crested cockatoos and eastern rosellas, with the former three host species also testing positive for C. psittaci. At least two birds (a crimson rosella and a galah) were shedding C. psittaci strains from the 6BC clade, which is highly virulent for humans 19,45 . We also identified birds testing positive for Parachlamydiaceae, another Chlamydiales family identified as potentially pathogenic in humans 46 . Five host species were seropositive, namely www.nature.com/scientificreports/ crimson rosellas, galahs, sulphur-crested cockatoos, blue-winged parrots and eastern rosellas. It is plausible that the other host species we tested may also be infected with C. psittaci or other Chlamydiales, but our smaller sample size prevented us from detecting positive individuals. We found that host species differed in Chlamydiales prevalence, and that there were host species and geographical differences in seroprevalence. Host species differences in C. psittaci prevalence have been reported in an early study of wild Australian parrots 23 , and in a more recent study of captive parrots 14 . Species differences in seroprevalence have been reported in captive parrots 15 . However, host species differences are rarely investigated in wild birds. We found that host species predicted Chlamydiales PCR prevalence, although the host species in which we did find infections did not differ significantly from each other in prevalence. We found no significant association between host species and C. psittaci prevalence, although this may be due to low statistical power, given the low C. psittaci prevalence in our sample. We found that host species predicted seroprevalence, with galahs having significantly higher seroprevalence compared to crimson rosellas and sulphur-crested cockatoos, suggesting that galahs may have a higher level of exposure to Chlamydia compared with the other host species. We also identified a higher C. psittaci prevalence in galahs (10%) than previously reported in this host species, with previous estimates between 0 and 2% 23,27,28 . Ecological or behavioural differences may result in increased chlamydial exposure in certain hosts. For instance, galah foraging behaviour may facilitate higher rates of infection: they typically forage on the ground 47,48 which may cause them to become infected more frequently, as C. psittaci is transmissible through infected fomites 38 and other Chlamydiales are also hypothesised to be transmitted through environmental contamination 49 . Interestingly, we also identified one galah infected with C. gallinacea 41 , a chlamydial species more frequently associated with poultry 10 . This could suggest a potential route of chlamydial transmission between wild parrots and free-range poultry (which may be bidirectional), and may warrant increased biosecurity measures on farmland. While seroprevalence was lower in crimson rosellas and sulphur-crested cockatoos, we also identified C. psittaci in these species, at 14% and 10% prevalence respectively. C. psittaci has previously been reported in wild crimson rosellas 19,28 , but in the few studies we found of wild sulphur-crested cockatoos, no birds assayed positive for C. psittaci 27,37 , except for one individual concurrently infected with beak and feather disease virus (BFDV) 28 . We found no blue-winged parrots shedding either C. psittaci or Chlamydiales, and only one individual which tested seropositive. This may indicate that they are less susceptible to infection, or that infected birds suffer severe disease or fatality prior to detection. The latter may be more likely given that parrots in the Neophema genus are reported as hard to treat for C. psittaci infection 7 and C. psittaci has previously caused fatality in captive populations of the closely-related orange-bellied parrot 39 . To our knowledge, no other studies have tested wild blue-winged parrots for chlamydial infection.
We found geographic variation in seroprevalence, with a significant difference between field sites. However, we found no geographic variation in PCR prevalence. Consequently, our findings are more likely to indicate a previous high infection rate or outbreak in Meredith than a current high infection rate, and suggest that chlamydial exposure varies by location, over a relatively local scale. It is possible that this site variation in seroprevalence may partly be due to seasonal variation. However, we think this is unlikely, as IgG antibodies can persist in the host for several months following infection 50 . Moreover, our data shows that seroprevalence status did not change in birds recaptured several months later, suggesting that sampling date is unlikely to bias our seroprevalence data. Geographic variation in seroprevalence could arise from differences in site environmental characteristics, or bird community composition. Food availability, altitude and other habitat characteristics can predict malarial parasitaemia in wild birds 51 , and variation in bird community composition is suggested to cause geographic variation in Mycoplasma gallisepticum prevalence in house finches (Carpodacus mexicanus) 52 . Indeed, C. psittaci prevalence in pigeons in Europe has been shown to range between 16 and 28% across comparable distances to those separating our field sites (< 100 km), as well as differing between lofts in the same city 53 . The geographic variation we observed may mean that in certain locations there is a greater risk of chlamydial disease outbreak in birds, and consequently a greater risk of transmission to humans and livestock. Serological surveillance could be carried out in wild birds found in close proximity to livestock or human communities, to investigate these risks. Interestingly, while we did not find significant sex differences in overall PCR prevalence or seroprevalence, we found that the site differences we observed in seroprevalence were greater in males than in females. It is possible that a previous outbreak occurred in Meredith, and males were more susceptible to chlamydial infection than females, since males are more susceptible to infection in most vertebrate species 29 . Alternatively, as inherent sex differences in immune response and antibody persistence may occur 29,54 , it is possible that males have a longerlasting antibody response. Therefore, if there was an outbreak at this site, male seroprevalence would remain elevated for a longer time. To our knowledge, no studies have tested for sex differences in chlamydial prevalence or seroprevalence in any wild bird species. Previous studies from captive parrots have shown either no significant sex differences in prevalence 55 , or in contrast to our findings, a higher seroprevalence in females 15 .
The overall C. psittaci prevalence we found (10%) was higher than that reported in other recent wild parrot studies in Australia and worldwide 27,28,56 , which could be due to a number of reasons. Firstly, as discussed, there is likely to be geographical variation. Secondly, sampling time of year may also cause variation in prevalence. For instance, we found a high Chlamydiales prevalence during summer (83%) which could be due to increased shedding due to the stress of breeding or moulting 34,38,57 . Additionally, estimated prevalence may vary between studies because different chlamydial PCR assays vary in sensitivity 30 . The 16SIG PCR assay and detection method we used appears to be very sensitive, because (a) we identified some 16S positive samples which could not be further characterised by sequencing, and (b) some C. psittaci positive samples (identified by sequencing) tested positive using the 16SIG PCR, but not the C. psittaci specific PCR. However, pan-Chlamydiales PCR primers such as those we used can also have lower specificity than other nested PCR or qPCR methods 58  www.nature.com/scientificreports/ another gene) may help to confirm sequencing results for C. psittaci identification. Other studies have also shown that samples with low C. psittaci loads may not be amplified by all PCR protocols 59 , or may only be identified by sequencing 27 . It is plausible these low-level infections have little effect on the host, and may not be of zoonotic risk: whether this is the case remains to be determined. A limitation of our study is that we did not characterise the genotype of all C. psittaci positive samples. In future, it would be useful to characterise the genotype of all C. psittaci strains found, to facilitate comparison of strains between and across host species, to help identify potential transmission pathways, and to quantify zoonotic transmission risks. Nonetheless, because all C. psittaci strains are considered transmissible to humans 10,60 and we identified at least two individuals shedding C. psittaci strains in a clade highly virulent in humans, we consider our findings are of zoonotic relevance. Another limitation is that some of the samples we sequenced were unsuccessful, most likely due to mixed infections, or low DNA concentration. Additionally, we did not sequence all PCR positive samples, so it is possible we may have detected bacteria outside the order Chlamydiales; however we consider this unlikely, since results from this study and our previous work 42 confirmed that all successfully sequenced samples were within the Chlamydiales. The short fragment (298 bp) also prevented us from identifying all sequenced Chlamydiales to species level; future studies could use primers targeting a larger fragment, to facilitate identification to a genus or species.
Our PCR data from recaptured individuals showed that 41% of recaptured birds assayed differently between captures. Similar results were found for BFDV infection in crimson rosellas, where 77% of individuals which tested BFDV positive at least once tested differently upon recapture 61 . We found birds which assayed seropositive on both capture events up to four months apart, and we also identified seropositive birds which always assayed C. psittaci and Chlamydiales PCR negative. These could be chronically infected individuals which shed Chlamydia intermittently, as commonly found in captive parrots 7,38 . If this is occurring, it increases the risk of transmission to conspecifics, and to other species, in cases where different species share habitat or nesting hollows 62 . Our results could also be explained by sporadic infection and re-infection, infection relapses, or recovered birds assaying seropositive due to antibody persistence 50 . Re-exposure could cause longer-lasting antibodies and a boosted immune response, as hypothesised for avian influenza antibodies in recaptured waterbirds 63 . It is also possible that chlamydial shedding may follow a circadian rhythm. Indeed, in crimson rosellas at least, we found more birds testing positive for C. psittaci in the morning. To our knowledge, circadian variation in chlamydial shedding has not received prior investigation. Consequently, this may warrant further investigation, particularly as such effects may influence detectability and repeatability of chlamydial testing. To investigate whether multiple infections, chronic latent infection, or intermittent shedding is occurring, future studies could test whether recaptured birds are always infected with the same or different chlamydial strains, and test known chlamydial positive birds in captivity periodically throughout the day. Our data also suggest that chlamydial exposure is not ubiquitous in wild populations, as we identified individuals which consistently assayed seronegative at capture events more than a year apart, suggesting that birds may not be exposed to Chlamydia for several months at a time. We also identified seronegative birds which assayed PCR positive for C. psittaci. These birds may be in an early stage of infection, and not yet producing a detectable immune response 50 , or alternatively, they may have had low-level infections, which did not induce an immune response, since the infectious dose of a pathogen can affect host antibody response 64 . Two birds tested C. psittaci positive on initial capture, but seronegative upon their recaptures several months later; it is possible these birds seroconverted after initial capture, then stopped producing IgG antibodies following recovery 65 . A limitation is that we do not know the precise sensitivity or specificity of the ImmunoComb. It is plausible that this assay (designed to detect C. psittaci) also detects antibodies against other chlamydial species 41 , and cross reactivity to other bacteria may also occur 66 . Future studies could develop species-specific peptide based ELISAs to test for exposure to each chlamydial species; such an approach has been found to have increased specificity for C. abortus detection in livestock 66 , and could similarly increase the reliability of seroprevalence estimates for wild birds. Interestingly, we found no relationship between seroprevalence status and any of the combinations of PCR status tested. This lack of relationship (and our recapture findings) indicate that neither PCR nor serology alone can confirm the presence or absence of chlamydial infection in a population. Similar results were found for feline foamy virus infection in wild pumas (Puma concolor) where ELISA and qPCR did not have strong diagnostic agreement 33 . We suggest that using both PCR and serology is desirable for accurate estimation of chlamydial prevalence, and epidemiological inference.
We found no effect of chlamydial infection on host body condition, which accords with our recent study of crimson rosellas 42 . This could be indicative of endemic infection, whereby wild parrots have a stable host-parasite relationship with Chlamydiales, as similarly hypothesised by de Freitas Raso et al. for hyacinth macaws in Brazil 24 . The infections we observed may have a low bacterial load, or may be of low virulence. Indeed, C. psittaci genotype A (which we identified in two individuals) is endemic among captive psittacine birds 10 , so under natural conditions, they may suffer few adverse consequences of chlamydial infection. However, we did find that seropositive birds tended to have lower body mass and PCV, suggesting that there may be a link between infection and host body condition. This would be a useful area for further study in both captive and wild individuals.
In conclusion, we show that wild individuals of common parrot species in south-eastern Australia are both exposed to, and shedding, C. psittaci and other Chlamydiales, at a higher prevalence than previously reported in most wild parrot populations. For the first time in wild parrots, we demonstrate that host species within a community differ in Chlamydiales prevalence, and that seroprevalence differs between host species, and for males at least, geographical location. We also reveal that some individuals show evidence of antibody persistence and potentially chronic infection, which has implications for direct and environmental transmission. Highlighting the wide range and abundance of potentially zoonotic chlamydial bacteria in wild birds, our findings suggest that conservation managers should investigate the presence of these bacteria when managing threatened species, and investigate the potential spill-over risks in locations where humans and livestock are in contact with wild birds. . Parrots were captured in two study areas in south Victoria, Australia: either within 10 km of Bellbrae (S38°19′ E144°10′) or within 12 km of Meredith (S37°51′ E144°06′). These areas are located approximately 75 km apart, and we therefore considered these two different parrot communities, as previous data indicate that most recaptured crimson rosellas were caught or resighted within 10 km of their banding site 67 , and galahs and sulphur-crested cockatoos within 20 km of their banding site 34,35,68 . Birds were caught using walk-in traps and mist nets. Upon capture, each bird was placed in a bag and weighed. Following this, each bird was banded, where possible the wing, head-bill, tail and tarsus length was measured, and blood and cloacal swab samples taken. Blood was collected from the brachial vein 69 , stored at 4 °C immediately after collection, then centrifuged for 9 min at 16,000g within 3 h of collection, after which serum was separated using a Hamilton syringe and stored at − 80 °C. Cloacal swabs were stored at 4 °C immediately after collection, then stored at − 80 °C within 12 h of collection. PCV was measured as described by Ots, Murumägi & Hõrak 70 .
DNA extraction and sequence analysis. DNA was extracted from cloacal swabs using an ammonium acetate extraction method 71 modified for swabs, with a no-template control sample included in each batch. To summarise, swabs were placed into 250 µl of Digsol buffer (20 mM EDTA, 120 mM NaCl, 50 mM Tris-HCl, 20% SDS) with 10 µl of Proteinase K (10 mg/ml). Samples were digested overnight (minimum 15 h) at 37 °C, and following this 300 µl of 4 M ammonium acetate was added. 100% ethanol was added to precipitate the DNA, after which each sample was washed with 70% ethanol and re-suspended in low Tris-EDTA buffer (10 mM Tris-HCl, 0.1 mM EDTA, pH 7.5-8.0). DNA quantity was verified using a NanoDrop 1000 spectrophotometer, and prior to PCR analysis, samples with a DNA concentration of > 50 ng/µl were diluted to 50 ng/µl. A multistep PCR protocol was used to determine which chlamydial species were present in samples. DNA samples were firstly assayed for the presence of Chlamydiales using the pan-Chlamydiales 16SIG PCR 72 (Table 4). Following this, positive samples were assayed using two separate species-specific PCR assays, using the C. psittaci-specific F3/B3 primers 73 and the C. gallinacea-specific gidA primers 74 to identify whether C. psittaci or C. gallinacea was present ( Table 4). The 16SIG reaction was performed in 50 µl total reaction volume, containing 2 µl of extracted DNA, 5 µl of each 10 µM primer, 5 µl each of 10× buffer and dNTPs, 3 µl of MgSO 4 and 1 µl of KOD Hot Start Polymerase (Novagen). Cycling conditions were as follows: initial denaturing period of 10 min at 95 °C, then 35 cycles of 1 min at 94 °C, 30 s at 68 °C and 1 min at 72 °C, followed by a final extension period of 7 min at 72 °C. Reaction conditions were the same for both the F3/B3 and gidA primers, with annealing temperatures of 57 °C and 59 °C respectively. Positive controls included a dilution of C. psittaci DNA for the 16SIG primers, DNA from a known C. psittaci positive bird for the F3/B3 primers, and a dilution of C. gallinacea DNA for the gidA primers. All negative controls were nuclease free water. All reactions were carried out in a GeneAmp PCR System 9700 thermocycler (Applied Biosystems, California U.S.A.). PCR product was visualised on a 1.5% Agarose gel, using 0.5× Tris-Borate EDTA buffer and SYBR Safe DNA Gel Stain (Invitrogen). Agarose gels were viewed under UV light and analysed using ImageLab 6.0.1 (Bio-Rad, California, U.S.A.). Samples with bands of intensity ≥ 5% of that of the positive control were considered positive. Where samples produced multiple bands or smears, they were re-assayed by PCR. Samples (n = 4) which did not produce a clear single band following re-analysis were considered 'inconclusive' as they were deemed of insufficient quality for further PCR or sequencing analysis. A subset of samples (n = 29) which assayed positive using the 16SIG PCR but negative for the two speciesspecific PCR protocols were sequenced using Sanger sequencing, to further interrogate the genetic identity of the amplicon. The amplified product from the 16SIG PCR was purified, then underwent a Big-Dye terminator reaction and dual-direction Sanger sequencing at the Australian Genome Research Facility (Melbourne). Sample chromatograms were analysed using MEGA X 75 , and sequences were compared against the nr/nt database using the BLASTn tool 76 . Samples were classed as C. psittaci if the top 10 BLAST hits had > 99% nucleotide identity and 99-100% query cover length with previously described C. psittaci 16S sequences, and E values of < 0.00001 (Supplementary Table S3). Similarly, samples with top BLAST hits with > 85% similarity with other published Chlamydiales sequences and above parameters in the nr/nt database were classified to family level or listed as 'other uncultured bacteria' . Chromatograms with multiple double peaks but > 90% percentage similarity with www.nature.com/scientificreports/ Chlamydiales (n = 7) were classed as mixed infections. Samples which were not successfully sequenced or had chromatograms of bad quality (n = 8) were classed as inconclusive. For two C. psittaci positive samples with high DNA concentration and band intensity (from one crimson rosella, and one galah), we used the CTU and CTL primers 77 to amplify an approximately 1070 bp fragment of the ompA gene, to investigate which genotypes are present (Table 4). Sequences were aligned with other publicly available ompA sequences in GenBank in MEGA-X 75 , using ClustalW. The ompA product of the positive C. gallinacea sample was also sequenced, the results of which are reported in a separate study 41 . All sequences are deposited in GenBank (see Table 1 for accession numbers).
Serological analysis. Serum samples (n = 108) were assayed for antibody presence using the ImmunoComb solid-phase ELISA (Biogal, Kibbutz Galed, Israel). This kit has been validated for use in rosellas and cockatoos 78 and a wide range of other psittacine and non-psittacine birds 78 . In brief, each serum sample corresponds with a colour change, with the colour intensity indicating whether or not antibodies are present 79 . Samples were allocated a colour intensity score from 0 to 5.5 in increments of 0.5, and compared for intensity against a positive control, which had a score of 3. Samples with scores of 0 were classed as negative, samples with scores of ≥ 2.5 were classed as positive, and samples with scores of 1-2 (n = 18) classed as inconclusive 79 . Inconclusive samples were excluded from analysis. A negative control was included in every reaction. The reliability of this method was confirmed by carrying out repeated scoring analysis of a subset of samples (n = 40) in a previous study 42  We also tested all two-way and three-way interactions between model terms in addition to main effects, but interactions were only retained in the final models when significant (p < 0.05). Only one twoway interaction was retained ( Table 2) as all other interactions had a p-value > 0.1. Post-hoc Tukey tests were used to estimate pairwise differences between species. We also ran GLMs for PCR prevalence with 'time caught' included as an additional fixed factor, to test whether prevalence varied between different times of day. We tested this first for all focal species, and then for crimson rosellas separately, as the host species with largest sample size and most capture time variation. For focal species with positive and negative birds (crimson rosellas, galahs and sulphur-crested cockatoos, plus blue-winged parrots for seroprevalence), we used GLMs to test for an effect of Chlamydiales prevalence, C. psittaci prevalence and seroprevalence on raw body mass and PCV. We used body mass as this is a reliable measure of body fat content 83 and we used PCV as this is a physiological measure often affected by disease or other environmental stressors 84 . We controlled for host sex and species, as these are strong predictors of body mass and can also cause variation in haematocrit 84 . For crimson rosellas and galahs (the host species with more morphometric data), we also carried out analyses using residual body mass (from regression of body mass on tarsus length) 85 as an additional measure of condition, to control for size differences 70,85 . Residual body mass was calculated separately for each species. Each condition index was modelled as a linear response. Mean PCV for each species (± SD) is reported in Supplementary Table S8.
We report the level of diagnostic agreement between our PCR analysis and seroprevalence analysis, based on the assumption that the ImmunoComb detects antibodies against both C. psittaci and C. gallinacea 41 . We tested the relationship between PCR and seroprevalence by using separate GLMs to investigate whether seroprevalence (binary logistic response) was predicted by: [1] C. psittaci PCR status, [2] Chlamydia (genus) PCR status, and [3] Chlamydiales PCR status. For these analyses, we excluded Chlamydiales positive samples where the bacterial species was unknown. For all prevalence analyses described above, we only analysed data from the first capture of each individual bird. For analysis of individual changes in infection status, we included data from recapture events, including recapture data from breeding birds (n = 16) which were caught in nest box traps as part of a separate study. We first used a GLM to test whether birds recaptured in different seasons were more likely to change in PCR status, with 'change in PCR status (Y/N)' modelled as a binary response, and 'caught in different season (Y/N)' as a predictor. For crimson rosellas, we used a binary logistic regression to investigate whether Chlamydiales infection status at first capture (positive or negative, according to PCR) predicted Chlamydiales status at recapture. We repeated this analysis excluding individuals (n = 2) which had a recapture interval of less than 4 weeks, to account for the possibility that recaptures after very short periods of time would not reveal any biologically relevant changes in infection status. We also ran these analyses including 'number of days between capture' as a covariate, to account for the likelihood that individuals recaptured at shorter intervals were more likely to assay the same upon recapture.
Ethical statement. All sampling for this study was approved by the Animal Ethics Committee of Deakin University (permit B31-2015) and carried out under ABBBS banding authority 2319. All handling and use of animals conforms to the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.