Background
Toxoplasma gondii is a ubiquitous protozoan parasite that infects felines as definitive hosts and a wide range of other warm-blooded animals, including humans, as intermediate hosts. Infected cats excrete copious amounts of
T. gondii oocysts containing infectious sporozoites in their feces. The oocysts can contaminate soils and remain viable for months to years [
1,
2]; subsequent ingestion of oocyst-contaminated soil causes infection in intermediate hosts.
T. gondii infects many tissues of intermediate hosts, including muscles and the central nervous system, where it forms infectious tissue cysts. Predation of intermediate hosts by felines completes the life cycle of the parasite. In humans, life-long infections usually result from ingestion of raw or undercooked meat of infected intermediate hosts, such as pigs, as well as ingestion of environmental oocysts [
3]. Other infection routes in humans include vertical transmission from infected mother to her fetus and transmission via blood transfusion or organ transplant [
4,
5]. A new infection typically involves a transient acute phase caused by the rapidly replicating tachyzoites of the parasite followed by a latent, life-long stage with bradyzoites persisting inside tissue cysts.
Serum immunoglobulin (Ig) M response to the parasite is characteristic of the acute phase of infection. Serum IgG response, which reaches a maximum level within 2 to 3 months of initial infection and then slowly declines to a residual elevated level, is characteristic of the latent phase of infection [
4]. A combination of serum IgG and IgM responses to the parasite is used to differentiate acute and latent infection phases in diagnostic settings [
6]. Serum IgG immunoassays are a standard test used in population surveillance of
T. gondii seroprevalence [
7,
8].
According to the nationally representative data from the National Health and Nutrition Examination Survey (NHANES), IgG seroprevalence of
T. gondii in the US in individuals older than 5 years of age was 13.2% in 2009–2010. There had been a substantial decline in seroprevalence during the previous 20-year interval [
8]. Low socioeconomic status has been linked with increased odds of seropositivity in the US [
9].
There is strong epidemiologic evidence of
T. gondii transmission via consumption of raw or undercooked meat [
10,
11] suggesting that ingestion of tissue cysts is a dominant infection pathway in the US and other developed countries. Research using antibody responses to sporozoite-specific antigen demonstrated that ingestion of environmental oocysts containing sporozoites is also a common infection pathway in North America [
12,
13]. Some studies also provided evidence of an association between contacts with cats and
T. gondii infection in developed countries [
10,
11,
14], while other studies failed to confirm this association [
15‐
17]. Associations between gardening and other soil contacts with risk of
T. gondii infection have also been demonstrated [
15,
16,
18]. In addition, waterborne outbreaks of toxoplasmosis have been reported in Canada [
19] and Brazil [
20].
Only 10–30% of new
T. gondii infections in humans cause clinical symptoms [
21], but when symptoms are present, the clinical manifestations of the disease can be severe. Symptoms of acute toxoplasmosis include ocular disease, encephalitis, chorioretinitis, lymphadenitis or lymphadenopathy, and myocarditis [
22].
T. gondii infection during pregnancy and vertical transmission of the parasite to the fetus can cause mental disabilities, seizures, blindness and spontaneous abortion [
3]. In the US, 400 to 4000 infants are born annually with congenital toxoplasmosis [
3,
23].
While the latent phase of infection may appear asymptomatic, there is a growing body of evidence of behavioral modifications in intermediate hosts enhancing the probability of predation by felines. Examples from animal studies include fatal attraction to and sexual arousal by the smell of cat urine in infected rats [
24,
25], and similar fatal attraction to the smell of leopard urine in infected chimpanzees [
26]. Epidemiological studies have also linked latent infections in humans with adverse neuropsychological outcomes, including elevated risk of suicide [
27], impaired reaction time and increased risk of traffic accidents [
28], and mental health disorders including schizophrenia, depression, and obsessive compulsive disorder [
22], as well as greater odds of developing a metabolic disorder type 2 diabetes [
29]. There is also evidence of widespread immune activation and subclinical neurophysiological changes induced by
T. gondii infection in humans [
30,
31]. However, the knowledge of sub-clinical health effects of latent infections and biological pathways leading to these effects remains rather limited.
The objectives of this study were: (i) to assess behavioral and environmental risk factors for T. gondii infections in the Durham-Chapel Hill, NC area, and (ii) to explore potential associations between latent T. gondii infections and a composite biomarker-based measure of physiological dysregulation known as allostatic load (AL), and individual biomarkers of immune, neuroendocrine and metabolic functions.
Methods
Study design and data collection
The protocol of this cross-sectional population-based observational study was approved by the Institutional Review Board of the University of North Carolina at Chapel Hill. The target population included adult (at least 18 years of age) residents of the Durham-Chapel Hill metropolitan area in North Carolina. The study was advertised on the US EPA’s website for recruitment of volunteers in epidemiological research, and by displaying study posters at various venues. Veterinary practices and animal shelters were targeted for recruitment to over-sample individuals with increased contacts with cats, dogs and other animals. Participants reported to the US EPA Human Studies Facility (HSF) in Chapel Hill, NC. A venous blood sample was drawn in a BD Vacutainer SST tube (Becton, Dickinson and Company, Franklin Lakes, NJ), and height and weight were measured by a registered nurse. Serum was separated following manufacturer instructions on the day of collection, and stored at − 80° C until analysis. Participants also completed a questionnaire addressing their demographic and socioeconomic characteristics, and behavioral and environmental factors which may be associated with exposure to T. gondii, such as contacts with cats, handling soil and consumption of undercooked meat of various types. Data collection was conducted in May – September 2013.
Serological tests
Serum samples were tested for IgG response to T. gondii using VIR-ELISA Anti-Toxo IgG assays (VIRO-IMMUN Labor-Diagnostika GmbH, Oberursel, Germany) in accordance with manufacturer’s instructions. Geometric mean values from duplicate tests were used in data analysis. Samples from two individuals with indeterminate results (average optical density values within plus/minus 10% interval around the plate-specific cut-off) were re-analyzed. If a new test again produced an indeterminate result, the infection status was classified as negative if the average ratio of optical density value for the sample to the corresponding plate-specific cut-off was less than one and as positive otherwise.
As part of a previously conducted study of environmental predictors of AL [
32], serum samples were also analyzed for 15 stress-related biomarkers, including nine biomarkers of immune function: C-reactive protein (CRP), vascular cell adhesion molecule 1 (VCAM-1), intercellular adhesion molecule 1 (ICAM-1), interleukin (IL)-1β, IL-6, IL-8, tumor necrosis factor (TNF)-α, fibrinogen, and myeloperoxidase (MPO); four biomarkers of neuroendocrine function: dehydroepiandrosterone (DHEA), epinephrine, norepinephrine, and dopamine; and two biomarkers of metabolic function: uric acid and serum amyloid A (SAA). All biomarker tests were conducted using commercially available assay kits as described previously [
32].
Analysis of allostatic load
AL was calculated as a sum of dichotomized biomarker values, which is the most commonly used approach in AL studies [
33]. Biomarker data were dichotomized at the 10th percentile of the sample distribution (DHEA and dopamine), at the 90th percentile (IL-1β, IL-6, IL-8, TNF-α, fibrinogen, uric acid, MPO, CRP, SAA, VCAM-1, and ICAM-1) or at both 10th and 90th percentiles (norepinephrine and epinephrine, two binary variables for each biomarker), depending on which tail of the biomarker distribution is known to be associated with an elevated risk of disease or death. Thus, AL measures were based on a total of 17 binary variables representing 15 biomarkers as described previously [
32].
Analysis of residential exposure to living environment
Proportions of total vegetated land cover within a 500 m radius of each residence were estimated using high resolution land cover data for the Durham-Chapel Hill, NC metropolitan area from the US EPA’s mapping application EnviroAtlas (
https://www.epa.gov/enviroatlas) as described previously [
32]. Vegetated land cover was defined as the sum of two land cover categories: Trees & forest, and Grass & other herbaceous. Exposure measures were based on average proportion of vegetated land cover within 50 m, 150 m, and 500 m radii, and distance-weighted average proportion of vegetated land cover within 500 m radius around the residence. The latter measure was calculated as an arithmetic mean of vegetated land cover proportions for ten concentric 50 m annuli from 0-50 m to 450–500 m. This weighting scheme implicitly used a constant weight of 0.1 for each annulus. As a result, a square meter of vegetated land cover within the 0–50 m annulus (7854 m
2 area) had 19 times greater impact on the weighted estimate than a square meter of vegetated land cover within the 450–500 m annulus with 19 times larger area (149,226 m
2).
Statistical data analysis
Statistical analysis was conducted using SAS version 9.4 software (SAS Institute, Cary, NC). It involved two phases: the first phase focused on environmental predictors of T. gondii infections with T. gondii serostatus being an outcome variable while the second phase focused on subclinical health outcomes of latent T. gondii infections. In the second phase, T. gondii serostatus was a predictor variable while AL and individual biomarkers were modeled as outcome variables, one variable at a time.
At the first phase, univariate analysis of associations between demographic, behavioral and environmental factors and T. gondii infections was conducted using the Chi-square Wald test for binary and nominal variables and the Cochran-Armitage test for trend for ordinal variables. Subsequent multivariate regression analysis involved developing predictive models of T. gondii seropositivity. An initial predictive logistic regression model included a set of socio-demographic and behavioral covariates. The next step involved adding cat-related variables to the initial model, one variable at a time, and selecting the cat variable that produced the best model fit. Akaike Information Criterion Corrected (AICc) values in the output of SAS procedure genmod were used as a measure of model fit.
The next step involved adding vegetated land cover measures to the previously developed model, one variable at a time. To account for spatial autocorrelation, all regression models for vegetated land cover included a two-dimensional spline smoothing function of geographic coordinates (“thin-plate smoother”), as described previously [
32]. The models involving a combination of linear and non-linear predictors also known as generalized additive models were fitted using the SAS procedure
gam. Using generalized additive models is a common approach in analysis of geographic distributions of health outcomes [
34‐
36]. The spline function was fitted using the option “method = GCV” (generalized cross-validation function) for automatically selecting degrees of freedom which define the flexibility of the “thin plate” smoother. Model fit was assessed using the deviance of the final estimate criterion from the output of SAS procedure
gam.
In the second phase, associations of T. gondii seropositivity with AL (a Poisson-distributed count variable), as well as individual dichotomized biomarkers and continuous log-transformed biomarkers were analyzed in univariate and in multivariate regression models adjusting for demographic and socioeconomic covariates. The univariate analysis of association between T. gondii and AL was followed by developing a multivariate predictive Poisson regression model including demographic covariates and body mass index (BMI). A final predictive model for AL also included vegetated land cover as a covariate and a spline function of geographic coordinates to account for spatial autocorrelation; it was fitted using the SAS procedure gam as described above. Associations between T. gondii seropositivity and individual biomarkers were analyzed using logistic regression models for binary biomarkers adjusting for socio-demographic covariates or linear multivariate models for log-transformed biomarker data.
Discussion
In this study of urban and suburban adult residents of central North Carolina, greater residential vegetated land cover was significantly associated with
T. gondii seropositivity. In previous studies, exposures to urban green spaces and other natural living environments have been linked to reduced morbidity and mortality; however, evidence of potential detrimental effects of green spaces including zoonotic infections remained rather sparse [
37]. To our knowledge, this is the first study linking residential greenness to
T. gondii infections.
The present study also found that handling soil with bare hands was a risk factor for
T. gondii seropositivity and produced some evidence suggesting that individuals living in a greener residential environment were more likely to acquire infection through contacts with local soil contaminated with environmental oocysts of this parasite. The association between contacts with soil and
T. gondii infections is corroborated by previous research [
15,
16,
18]. A study in France also demonstrated that urban and rural residents acquired
T. gondii infections through different pathways: parasites isolated from urban residents lacked geographic genetic structure, suggesting foodborne infections via products transported over long distances; in contrast, parasites isolated from rural residents exhibited a spatial genetic structure, suggesting a greater importance of local sources of infection [
38].
The observed 8.3% seroprevalence of
T. gondii in this study involving adults was lower than the 13.2% national estimate of unadjusted seroprevalence in children (>5 years of age) and adults in the 2009–2010 NHANES study [
8]. The present study was conducted in an urban area with a relatively high educational and socioeconomic status compared to the general US population, which may explain the relatively low observed seroprevalence rate. Local conditions affecting the spread of
T. gondii might also differ from those in other US regions. Recruitment methods for this study attempted to oversample cat or dog owners. Cat ownership was a risk factor for
T. gondii seropositivity; this finding was consistent with results of previously conducted studies [
10,
11,
14]. Having a dog was a negative predictor of cat ownership; it also tended to be inversely associated with
T. gondii seropositivity, although the effect was not statistically significant. This finding is consistent with previous research which demonstrated a significant protective effect of dog ownership on
T. gondii infections [
39]. Thus, the effects of oversampling cat and dog owners on
T. gondii seropositivity in the study population could cancel out each other. Although the seroprevalence estimate from this study is not generalizable due to non-random sampling, findings about risk factors for
T. gondii infections are likely to reflect transmission pathways of this parasite in central North Carolina.
Furthermore, the present study showed that latent
T. gondii infections were associated with a detrimental systemic effect reflected in elevated AL, a composite measure of physiologic dysregulation based on multiple biomarkers. Our previous study in the same group of adult residents of North Carolina demonstrated that greener residential environment was associated with reduced AL [
32]. Thus,
T. gondii infections acquired through more frequent contacts with contaminated soils in greener neighborhoods where cats were allowed to roam outdoors could partially offset health benefits of contacts with the natural living environment. This finding shows the importance of minimizing risks of zoonotic infections in green areas.
A limitation of this cross-sectional observational study is that it could only demonstrate statistical associations; it was not designed to establish a cause-effect relationship. While one of the hypotheses of this study was that latent T. gondii infections cause chronic inflammation resulting in greater physiological dysregulation and elevated AL measures, there may be alternative explanations for the observed effect. It is possible that individuals with higher AL were more susceptible to T. gondii infections (reverse causation). It is also possible that the observed association was due to confounding by unknown behavioral factors that affect the risk of T. gondii infection as well as AL.
The beneficial effect of residential vegetation on AL demonstrated in our previous study in the same population [
32] remained highly significant after adjusting for
T. gondii serostatus in the present study. Further stratified analysis showed that the beneficial effects of vegetated land cover were pronounced in both seropositive and seronegative individuals. In seropositive individuals, the detrimental effect of
T. gondii and the beneficial effect of an IQR increase in residential greenery were of similar magnitudes. However, due to the small sample size (only 15 seropositive individuals with AL data), these effect estimates should be interpreted with caution.
Analysis of associations with individual biomarkers, which comprised the AL index, showed that
T. gondii infection was linked with a significantly increased aOR of having serum MPO above the 90th percentile and, in analysis of continuous biomarker data, with a higher median concentration of MPO. These findings are logical as MPO is an enzyme involved in immune response to pathogens. Elevated levels of MPO have been linked with inflammation and with cardiovascular diseases [
40,
41].
T. gondii seropositivity was also linked with elevated levels of VCAM-1 and IL-6 (
p < 0.05 for each association), although these effects did not remain significant after applying Bonferroni correction for multiple comparisons. Previous research demonstrated associations between
T. gondii infection and an elevated serum level of VCAM-1 and some pro-inflammatory cytokines [
31] corroborating the results of this study. Experimental research also demonstrated that pro-inflammatory cytokines including IL-6 and TNF-α play critical roles in inhibiting the replication of this parasite in humans [
42]. Previous studies have also shown associations between
T. gondii and elevated serum ICAM-1 [
43], and reduced serum DHEA (both associations suggesting detrimental health effects) [
44]. Those findings were not replicated in the present study possibly due to the small sample size or particular characteristics of the study population.
Acknowledgements
The authors are grateful to Dr. Laura Jackson and the EnviroAtlas land cover classification team at US EPA for developing the land cover dataset and providing consultations on its use, and to Dr. Eric Villegas (US EPA) for helpful comments on a draft manuscript. This study was an intramural research project at US EPA. Jennifer N. Styles was funded by the US EPA-UNC-CH Cooperative Training Agreement CR-83591401-0, with the Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill.