CHLAMYDIA PSITTACI IN FERAL ROSY-FACED LOVEBIRDS (AGAPORNIS ROSEICOLLIS) AND OTHER BACKYARD BIRDS IN MARICOPA COUNTY, ARIZONA, USA

Chlamydiosis is a bacterial disease of free-living and domestic birds, mammals, and humans caused by obligate intracellular bacteria (Chlamydia spp.). Chlamydiosis, caused by Chlamydia psittaci, is most-commonly reported in species of parrots (Order: Psittaciformes) but occurs in other avian species worldwide (Brand 1989). Chlamydia psittaci, found in both free-living and captive birds, has as many as 15 different genotypes (Sachse et al. 2008). Holzinger-Umlauf et al. (1997) reported that 54% (215/399) of apparently healthy tits (Parus spp.) tested positive for Chlamydia spp. and, of those tits found dead, 67% (4/6) were positive for Chlamydia spp.; in two it was determined to be the cause of death. However, the prevalence of C. psittaci in other bird populations was much lower: 7.9% (26/331) for Rock Doves (Columba livia) and 0.9% (3/349) for Mallards (Anas platyrhynchos; Heddema et al. 2006; Blomqvist et al. 2012). Franson and Pearson (1995) primarily attributed the cause of death of more than 400 gulls (Order: Charadriiformes) in North America to chlamydiosis.

Chlamydia psittaci is a public health concern because it can cause a disease in humans (psittacosis) which is usually associated with mild respiratory disease but can cause severe pneumonia or other serious health issues (Smith et al. 2011). While most cases of psittacosis have been linked to companion birds, some studies have linked outbreaks to free-living birds (Smith et al. 2011). Investigations of psittacosis outbreaks in Australia revealed that direct contact with free-living birds, gardening, and mowing lawns increased the risk of human infection (Williams et al. 1998; Telfer et al. 2005). A cluster of psittacosis cases in Sweden may have been associated with free-living bird contact, but other sources of the chlamydia bacterium were not ruled out (Rehn et al. 2013).

Nonnative, feral species in the avian order Psittaciformes are potential C. psittaci reservoirs in North America where eight species are known to have established free-living populations (Butler 2005; Chesser et al. 2013). The only feral population of Rosy-faced Lovebirds (Agapornis roseicollis) in North America is in Maricopa County, Arizona where they are frequent visitors to residential bird feeders (Chesser et al. 2013). In 2011, Radamaker and Corman (2011) estimated the lovebird population at 2,500 individuals.

A mortality event of 18 Rosy-faced Lovebirds occurred in August and September of 2013 around bird feeders at a residence in Scottsdale, Maricopa County, Arizona and was attributed to infection with C. psittaci (US Geological Survey 2017). Deaths of Rosy-faced Lovebirds associated with residential backyard feeders reoccurred in Maricopa County in June–September 2014, killing an estimated 75 lovebirds (US Geological Survey 2017). Diagnostic testing again detected the presence of C. psittaci in dead birds and avian chlamydiosis was attributed as the cause of mortality (US Geological Survey 2017). We then documented the prevalence of C. psittaci in feral lovebirds, in sympatric, feeder-associated bird species, and in environmental samples. Also, we tested for the presence of psittacine circovirus pathotypes 1 and 2 (PCV-1 and PCV-2; Circoviridae) because circovirus infection may cause host immunosuppression, increasing susceptibility to other diseases (Todd 2000; Pare and Robert 2007).

We collected samples from 20–28 August 2014 at six sites in Maricopa County, Arizona; four in Scottsdale, one in Phoenix, and one in Mesa (Table 1). Sites selected were residential properties with bird feeders where lovebirds were regularly observed. Property owners at four sampling sites (sites 1–4) had previously reported lovebird mortalities in 2014, and two sites (sites 5–6) had no reported mortality. We obtained the voluntary cooperation of each property owner prior to sampling. We sampled each location once or twice between 0700–1100 hours depending on ambient temperature. We captured birds at each location using one or two mist nets (9×2.1 m, 32-mm mesh) depending on the size of the property. If doves were present, we also used a single walk-in trap. We placed the nets and the trap to maximize bird captures. When possible, we flushed birds into mist nets and one lovebird was hand captured. Sampling personnel used personal protective equipment including N95 respirators, eye protection, gloves, shoe covers or rubber boots, and gowns or coveralls. All equipment or supplies were discarded, washed, or disinfected between sites (daily at the conclusion of trapping) using either a commercially available laundry detergent for nets, holding bags, and clothing or Maxima 256 (Brulin & Company, Inc., Indianapolis, Indiana, USA) for boots, traps, net poles, and all other sampling materials. Maxima 256 was also used to disinfect surfaces or equipment while processing birds.

We identified captured birds to species, inspected for abnormalities, marked (clipping <1 cm of the end of the outer rectrix), sampled, and released them. For swab sampling, we used a prepackaged sampling and transport system with swab and media provided (BD CultureSwab Liquid Amies; Becton Dickinson, Sparks, Maryland, USA). A single, appropriately sized, sterile swab was used to first swab the conjunctiva and then the choana. A second swab was used to sample the cloaca. Each swab was placed individually in tubes containing Amies transport medium and held chilled on ice packs in the field and then stored in a commercial refrigerator until shipment to the laboratory using ice packs. Blood was collected using an appropriately sized needle (28 ga or smaller) and syringe via jugular venipuncture and transferred to individual cryovials, with ≤10% of estimated blood volume obtained from each individual (i.e., a 35-g bird resulted in a blood sample ≤0.35 mL). Blood samples were stored on ice packs until centrifuged, when serum was collected and stored frozen (approximately −20 C) until shipment to the laboratory. The National Wildlife Health Center's Animal Care and Use Committee approved all procedures performed on live birds (Protocol no. EP140808).

We collected three environmental samples from each site during a single visit by swabbing bird feeders, perches, and edges of bird baths. Each swab was placed individually in tubes containing Amies transport medium and handled as previously described.

Real-time PCR-based testing for C. psittaci, PCV-1, and PCV-2 DNA from swabs and elementary body agglutination (EBA) detection of antibodies to C. psittaci from serum was performed by the Infectious Diseases Laboratory, University of Georgia (Athens, Georgia, USA). Testing was conducted on all swab samples for C. psittaci and PCV-1 DNA and a randomly selected subset of Rosy-faced Lovebird samples for PCV-2 DNA. For PCR-based testing, DNA was extracted from each swab and tested against positive and negative controls using organism-specific primers and probes in a LightCycler® 480 (Roche Diagnostics, Indianapolis, Indiana, USA). For the EBA assay, elementary bodies were prepared by centrifuging chlamydia-infected Vero cells at 46,467 × G for 1 h. The pellet was collected, washed in Opti-MEM (Thermo Fisher Scientific, Grand Island, New York, USA), and centrifuged at 1,200 × G for 5 min. The elementary bodies were inactivated, layered on renografin, and centrifuged at 60,226 × G for 90 min. The band containing elementary bodies was collected, pelleted, and resuspended in diluent to working concentration. Bird serum samples were diluted 1:10 for EBA; assays included known positive- and negative-control sera.

We used a Bayesian framework and fit a latent state logistic regression model using Markov-chain Monte Carlo (MCMC) methods implemented in JAGS software using R (R Core Team 2014; Plummer 2016) to describe detection of C. psittaci DNA. The model accounted for detection uncertainty in C. psittaci DNA, antibodies to C. psittaci, and PCV-1 DNA based on diagnostic test sensitivity and specificity (Joseph et al. 1995; McInturff et al. 2004).

The response variable for the epizootiologic analysis was the C. psittaci PCR-based result, and we made inference to the unknown true presence of C. psittaci DNA, z, in each individual, i, using a Bernoulli model,

where the latent true prevalence of C. psittaci, π, was inferred from the observed PCR-based result, y, and vague informative priors on the C. psittaci PCR-based diagnostic specificity, C, and sensitivity, S (prior specification and rationale below),

Next, we incorporated explanatory variables of capture site, species type, antibodies to C. psittaci, and PCV-1 detection using a logit-link function to estimate the influence of each explanatory variable on C. psittaci prevalence (π) by estimating the coefficients β_1–4,

Capture site was included as a factor influencing prevalence to account for unobserved variation that could contribute to C. psittaci prevalence and was modeled as separate intercepts for each site, j.

For the species explanatory factor, we grouped the species data as lovebirds and all other species because lovebirds were the only member of the order Psittaciformes and were the only presumptive natural hosts for C. psittaci (genotype A) and PCV-1. We estimated the effect that being a lovebird had on C. psittaci prevalence, where spp_i=1 if the individual was a lovebird and spp_i=0 for all other species.

The explanatory factor of C. psittaci exposure was based on the observed presence of antibodies to C. psittaci using the EBA test, and we included diagnostic uncertainty in the observed value by specifying the factor value (ab=1 if present, 0 if absent) with a latent Bernoulli process to include detection uncertainty and vague informative priors for EBA specificity (C_EBA) and sensitivity (S_EBA),

where observed EBA antibody detection in each individual, ⁠, was the input variable, and the true antibody status of each individual, ab_i, was inferred from a latent seroprevalence of antibodies (p^ab). This model for antibody detection also accounted for the missing data from captured individuals where blood was not collected, assuming no bias in the individuals where blood was not collected with respect to antibody status and C. psittaci DNA detection (Greenland 2009).

The explanatory factor of PCV-1 infection was also specified by a latent Bernoulli process where observed PCV-1 DNA detection, (y^PCV-1) was the input variable,

and the infection status of each individual was inferred based on prior information on the sensitivity (S_PCV-1) and specificity (C_PCV-1) of PCR-based detection of viral DNA.

We used vaguely informative priors for the regression coefficients β_1–4 and, hence, the prevalence of C. psittaci: normal distributions centered on 0 with a standard deviation of 5. With no information to inform the effect of these factors, this prior distribution contained a high amount of uncertainty (vague), but the coefficients assigned 95% of the prior probability to coefficient values between 10 and −10 (informative by limiting the probability of extreme effects). With the prior specification for the coefficients on the log scale, a coefficient of 10 results in a change from 50% to >99% probability of success on the response scale. We consider this a reasonable amount of information to put into our coefficient priors because a factor with an effect that large is unlikely, with our model evaluating four predictor variables with a modest amount of data (Gelman et al. 2008). The choice of these vaguely informative normal priors also resulted in much better convergence and shrinkage than vaguely prior distributions suggested by Gelman et al. (2008), which allowed for more probability of extreme coefficient values.

We specified uninformative priors, beta∼(1,1), for the seroprevalence of C. psittaci (p^ab) and prevalence of PCV-1 DNA detection (p^PCV-1) because there are no existing data about these values from feral lovebirds and sympatric feeder-associated birds that could be informatively applied to our study area. We used vaguely informative priors for diagnostic sensitivity and specificity for the three diagnostic tests based on reported applications in the literature using beta priors (Table S1).

We made inference using MCMC and Gibbs sampling implemented in JAGS using 50,000 interactions with 20% burn-in using five independent MCMC chains with randomly selected starting values from each parameter's prior distribution. We assessed MCMC convergence visually and by using the Gelman-Rubin statistic with the five MCMC chains to ensure that the scale reduction factor for each parameter was <1.1 (Brooks and Gelman 1988). After discarding the burn-in, we combined the five MCMC chains and sampled every 10th iteration to calculate the mean untransformed coefficient estimate and 95% high posterior density (HPD) intervals from the marginal posterior parameter distributions. We considered a factor to significantly influence the prevalence of a C. psittaci (π) if the coefficient (β) value did not overlap zero. To illustrate the influence of the significant factors we generated posterior distributions of predicted C. psittaci prevalence (π) among the combinations of significant factors and imputing the values of the nonsignificant values by imputing their values from their posterior distributions.

We captured and swab-sampled 188 live birds of 16 species including 46 Rosy-faced Lovebirds; 143 of these, from all 16 species, were also blood sampled. We found 30% (57/188) were positive for C. psittaci DNA by at least one of the two swabs. The highest detection rate was in lovebirds with 93% (43/46) positive by at least one swab (66% positive by both swabs) followed by Rock Doves (75%, 6/8; Table 2). All six sampling locations had at least one C. psittaci DNA-positive swab from a live bird (Table 2). We detected positive antibody titers for C. psittaci in 27% (38/143) of all birds sampled (Table 3). In lovebirds the prevalence of antibodies to C. psittaci was 76% (31/41) and in sympatric birds 7% (7/102) had antibodies to C. psittaci: two House Sparrows (Passer domesticus), two Inca Doves (Columbina inca), two Rock Doves, and one Curve-billed Thrasher (Toxostoma curvirostre). From 1–3 environmental samples were positive for C. psittaci DNA from each of the six sampling sites with 78% (14/18) total environmental swab samples positive.

We detected PCV-1 DNA in 40% (75/188) of individual birds with the greatest number of detections (for species with more than two samples tested) in lovebirds (80%, 37/46) followed by Rock Doves (38%, 3/8; Table 4). We also detected PCV-2 DNA in 14% (4/29) of lovebirds when both swabs were tested and none (0/10) were positive when only one swab was tested (Table 5).

We noted lesions on captured birds at just one site (site 2). At this location, 27% (6/22) of House Finches (Haemorhous mexicanus) exhibited variable crusty lesions on the head, bill, or legs, with some of these individuals missing parts of the upper or lower bill. Of these six birds, four also had positive conjunctival-choanal swabs (but not cloacal swabs) for PCV-1 DNA with the remaining two birds negative on both swabs. Five of these six House Finches were C. psittaci DNA-negative: the sixth bird, which was negative for PCV-1 DNA, was C. psittaci DNA-positive on both swabs. The etiology of the lesions was not further investigated. We captured 85% (22/26) of House Finches during this study at this site.

In our model, species had the largest significant influence on C. psittaci DNA prevalence, with the mean estimate of lovebirds being positive at 177 times greater than a non-lovebird (Table 6; odds ratio of β₂ 95% HPD=3.2–7.26). There was significant variation among sites with site 3 having higher mean C. psittaci DNA prevalence (Fig. 1). The coefficient estimates for serologic exposure and PCV-1 DNA detection did not have a significant effect on C. psittaci DNA prevalence (Table 6). The estimated seroprevalence of C. psittaci over all sites and species was 36% (95% HPD=5–42%) and the estimated prevalence of PCV-1 DNA over all sites and species was 44% (95% HPD=15–74%). We could only make these estimates over all individuals and sites because their observed values were highly correlated with species (Fig. 2). Results for PCV-2 testing were not included in the model due to only lovebird samples being tested for this virus.

Following Rosy-faced Lovebird mortality in 2013 and 2014 that was attributed to C. psittaci infection, our field investigation at the same area revealed a high incidence of C. psittaci-positive tests at all study sites among lovebirds from swab sample testing for C. psittaci DNA (94%) and from serum samples tested for antibodies (75%). In comparison, concurrently sampled sympatric species had a much lower prevalence of C. psittaci DNA (10%) and antibodies (7%). We also detected C. psittaci DNA in environmental samples at all study sites, confirming its widespread presence even at locations with no previous reports of disease.

A large proportion (80%) of lovebird samples was also positive for PCV-1 DNA. At the time of sampling, we did not observe feather abnormalities or other signs of disease associated with circovirus infection in lovebirds; however, infection of PCV-1 in lovebirds may only rarely result in clinical disease (Phalen 2006). We detected PCV-1 DNA in 28% of sympatric bird species with the highest prevalence in House Sparrows (27%, 17/62) and House Finches (38%, 10/26). Infection with PCV-1 can cause feather abnormalities in affected psittacines and circoviruses can cause immune dysfunction when they infect the bursa of Fabricius and thymus (Todd 2000; Pare and Robert 2007). There was no evidence that PCV-1 DNA detection influenced C. psittaci test positivity in our study. However, the variation in PCV-1 DNA detection was strongly correlated to C. psittaci test positivity and species type, with lovebirds much more likely to be positive than were other species for both pathogens (Fig. 2). Such high collinearity made further evaluation of any mechanistic inference between the infections unfeasible.

We detected a low prevalence of PCV-2 DNA, a pathotype that was originally described in lories (Subfamily Loriinae), in the tested lovebird samples (Raue et al. 2004). This pathotype was less virulent than PCV-1, and lories with advanced feather disease were able to recover from PCV-2–associated disease (Raue et al. 2004). While disease associated with PCV-2–only infection in lories is less severe than classic PCV-1–associated disease in other psittacine birds, it is unknown how coinfection of both pathotypes may affect the progression of infection of either virus in free-living bird populations. How infection with these viruses might change the pathogenicity of chlamydial infection is also unknown.

Although C. psittaci is an obligate intracellular bacterium, it can survive in a metabolically inactive extracellular state for over a month, which allows for transmission between vertebrate hosts (Smith et al. 2011). In our study, sampling occurred from nonsterile sites on the birds which could have detected contamination from an exogenous source (e.g., the feathers of the bird) and not necessarily through infection (Guzman et al. 2010). We considered that the PCR detections could have resulted from contamination by field personnel, but we removed birds from nets and processed and sampled them opportunistically as they were captured; no effort was made to group birds during processing that may have led to systematic contamination of a particular species. Our consistently positive findings of C. psittaci DNA in lovebirds and Rock Doves, and the general lack of positive results in all other species, suggests that this is not just environmental contamination from the work area or the capture site but represents an accumulation of C. psittaci DNA on or within the individual birds sampled. We also caution that PCR detection does not differentiate between viable or nonviable pathogens, so positive PCR results from swabs should not necessarily be equated with the presence of infectious material (Guzman et al. 2010; Smith et al. 2011).

The presence of C. psittaci antibody-negative and PCR-positive lovebirds may have indicated an exogenous source of C. psittaci by suggesting that the detected DNA was not infecting the individual and thus was not initiating an immune response (Guzman et al. 2010). Without repeat serologic testing of birds that were PCR positive for DNA but antibody negative, we could not determine when or if these seronegative birds would have been infected from ongoing exposure to C. psittaci. During sampling of live birds in late August, we observed lovebird mortality at two of the sampling locations. At one location (site 6) where lovebird mortality had not been previously detected, we recovered a lovebird carcass that was swab-positive for C. psittaci DNA. Moreover, the hand capturing of a single lovebird at this site suggests it was also sick; no other lovebirds could be captured in this way during the study. At a second location (site 2), the property owner found two dead lovebirds (that we had not previously sampled) between our sampling efforts approximately 4 d apart, and both birds were also C. psittaci DNA-positive. Even if only exogenous C. psittaci DNA was present on some or all birds, as long as the bacteria was viable, this would indicate that all or most of the C. psittaci DNA-positive birds were contributing to its maintenance in the environment and within the avian community.

Four sampling sites (sites 1–4) experienced consistent, low-level mortality of Rosy-faced Lovebirds (as reported by property owners), with onset in late June and early July and continuing through the end of August. Lovebird mortality was not previously reported at site 5 but at least one lovebird and one Inca Dove at this location were C. psittaci DNA-positive. Although site 6 had no reported mortality until late August, 76% (16/21) of the lovebirds we sampled were antibody positive and 91% (21/23) were C. psittaci DNA-positive. In an experimental study of Cockatiels (Nymphicus hollandicus), Guzman et al. (2010) reported a 33% (2/6) mortality of untreated positive control birds within 20 d of C. psittaci infection with no other comorbidity observed. There are few reports of naturally occurring avian mortality due to C. psittaci in the literature; however, a number of studies have been conducted to determine the prevalence of this organism in free-living birds (Franson and Pearson 1995; Holzinger-Umlauf et al. 1997; Heddema et al. 2006; Blomqvist et al. 2012). In our study, on-going mortality as well as the high C. psittaci DNA prevalence within lovebirds was indicative of the acute disease process. Unfortunately, little is known of this population of lovebirds prior to the first mortality event in August and September 2013. Without previous disease surveys, we were unable to determine if it was the first exposure of this lovebird population to C. psittaci (or possibly a change in its pathogenicity), the addition of PCV-1 in the population, or another factor that was responsible for the observed mortality.

Given the documented high prevalence of both C. psittaci DNA and PCV-1 DNA among Rosy-faced Lovebirds in this region, further understanding of the significance of these dual infections is warranted. Our findings of these pathogens in both free-living birds visiting backyard bird feeders and environmental samples near and on those same feeders indicate these congregation areas may be facilitating the transmission of these pathogens in free-living birds. In addition, while we found a generally low prevalence of C. psittaci DNA and PCV-1 DNA in native sympatric species, this study may not have detected impacts to the health of these birds and further study would be helpful. Feral parrot populations have become established in urban areas around the US and likely these populations will continue to expand; however, the presence of C. psittaci and PCV-1 in these populations is unknown (Butler 2005; Chesser et al. 2013). Heightened awareness of the risk for C. psittaci among lovebirds and other feral psittacines, particularly surrounding mortality events, is critical in evaluating the risk to native bird populations, companion and aviary birds, and humans.

We thank the residential property owners who provided access. We would also like to thank A. Stumpf, L. Plante, L. St. Louis, R. Levy, J. Pistole, J. Jones, C. Sefat, R. Lira, R. Hassan, K. Herrick, M. Luc, and K. Komatsu for assistance with field work and B. Wolff for C. psittaci genotyping. Funding was provided by the US Geological Survey and the University of Georgia Infectious Diseases Laboratory. The use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the US Government. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. Data are available at https://doi.org/10.5066/F76T0KKD.

Supplementary material for this article is online at http://dx.doi.org/10.7589/2017-06-145.