Study design

We conducted a cross-sectional study nested in our primary study entitled “Eco-epidemiological study of ectoparasite-borne diseases in Peru.” All methods were performed in accordance with relevant guidelines and regulations, as approved by the Gerencia Regional de Salud (GERESA) Loreto and by the Universidad Peruana Cayetano Heredia (UPCH) IRB and IACUC (project number: 201148).

Sample collection for this study was conducted in Iquitos (Peru) in April 2023.

Iquitos (3.7437° S, 73.2516° W) has a total population of 479,866 individuals (2017 national census) and comprises four districts: Iquitos, San Juan Bautista, Punchana, and Belen. April is in the rainy season, with daily temperatures ranging from 22 to 32 °C. During the rainy season, Iquitos experiences widespread flooding in various areas: Nuevo Versalles and Masusa in Punchana district; riverbank area of the historic center and Moronococha in Iquitos district; Zona baja in Belen district; and Pampachica, Cabo López, and Nueve de Octubre in San Juan district. Zona baja in Belen is the largest and most severely flooded sector. April was chosen to conduct field activities as the 2023 rainy season was ending, thus allowing us to access most areas of the Belen district without the need for a boat, thereby reducing costs, and minimizing the likelihood of fieldwork disruptions due to extreme heat and heavy rainfall.

In the present study, we included dogs living in houses in Iquitos City, which is formed by the urban areas of the four aforementioned districts. Exact district border information was obtained from the Peruvian National Institute of Statistics and Informatics (Fig. 1). Accordingly, the following district borders were used: the start of the Iquitos-Nauta highway (southwest), the Nanay river (northwest), the Amazonas river (northeast), and the Itaya river (southeast), as shown in (Fig. 1).

Map of Iquitos city showing study area boundaries and houses enrolled in the study.

Sample size calculation and house selection

A minimum sample size of 286 dogs, equivalent to 286 houses (considering a conservative scenario of one dog per house), was calculated to evaluate factors potentially associated with E. canis infection. We chose 80% power and 0.95 confidence level, and assumed a 0.2 intraclass correlation coefficient. We based our sample size calculation on a previous study conducted in Panama, focusing on dogs’ age and veterinary clinic visits as key risk factors¹⁶. The calculation was done in Stata 15.0 (Stata Corp., College Station, TX).

We randomly selected 286 buildings in QGIS v.3.28.0-Firenze (https://www.qgis.org) using a base map of all the buildings present in Iquitos city provided by Dr. Amy Morrison (University of California, Davis). According to the 2017 national census, the number of houses in Iquitos (28,381) and San Juan Bautista (30,037) districts is approximately twice the number of houses in Belen (13,467) and Punchana (15,698). Therefore, we randomly selected 96 buildings in Iquitos and San Juan each, and 47 in Belen and Punchana each to ensure adequate representation of each district. The selection was done using the random selection algorithm in QGIS v.3.28.0-Firenze. Then, we used the centroids algorithm to obtain the geographic coordinates (GC) of each house. The centroids GC were downloaded as a kml file that was uploaded to the maps.me application (My.com, VK holding). This application was used in the field to locate the selected houses.

In the field, the research team visited each randomly selected building. If the selected building was a business, an abandoned house, or nobody was present, we walked to the right until we found a house with someone present.

House visits, participant recruitment, and questionnaire

Two field teams, each comprising a veterinarian, a veterinary technician, and two biologists, were deployed. Houses with someone older than 18 years old present and at least one dog owned were selected for enrollment, and informed consent was secured by explaining the study procedure, benefits, and risks. Ineligible houses were noted as such, and the research team walked to the next house to the right.

After receiving informed consent from participants, the number of people and dogs in each house were recorded prior to verbally administering a questionnaire to the dog owner or any person from the house involved in the dog’s care (referred all as “owner” hereon). To limit biases, each field team designated a member in charge of consistently administering the questionnaire until the end of the study. Our questionnaire (Sup. material 1) included 5 sections: owner’s sociodemographic information (age, gender, education level, occupation, place of work, contact with animals at work, and income level); house characteristics (electricity availability; water source; type of sewage system and presence of open sewage nearby; presence of nearby dumps/abandoned houses/food markets, and presence of stray animals around the house); knowledge, experiences, and risk perception regarding ticks and tick-borne diseases (if owner was able to identify ticks in different developmental stages, if owner knew ticks feed on various animals and humans, that people and dogs can get tick-borne diseases, including CME, how concerned owners were about tick-borne diseases in dogs and humans, and if they had ever found a tick attached to their bodies); dog’s epidemiological data (pet signalment, frequency of veterinary visits, lifestyle, type and frequency of acaricide use, travel and tick infestation during the last year); and dog clinical data (presence of clinical signs compatible with CME during the last month, mucous membranes color, body condition, and skin turgor test).

Additionally, we visually inspected the house to determine the main wall and floor material(s) inside the house and in the patio (if the house had one), recording presence or absence of each type of wall or floor material.

The exhaustive list of assessed variables is available as Supplementary Data.

Sample collection

A veterinary technician (accompanied by a veterinarian) collected blood samples from a maximum of two dogs in the house. If more than two dogs lived in the house, we randomly selected two of them (using SocialGest, https://www.socialgest.net/es/sorteos-gratis for sample randomization). We only sampled dogs if their owners indicated that they were not aggressive. Dogs were safely muzzled and manually restrained. A maximum of 6 ml of blood was collected from dogs in vacuum blood collection tubes with EDTA.

All blood samples were immediately stored on ice and a cold chain was maintained until they were processed at the Emerging Diseases and Climate Change Research Unit (Emerge) laboratory in UPCH, Lima, Peru.

Ticks were also conveniently collected from dogs to be identified and to improve the well-being of canine participants. Dogs were systematically inspected from head to tail. Special attention was paid to the ears, neck, between the front and back legs, between the toes, and around the tail. Additionally, house walls and floors were visually inspected to ascertain environmental tick infestation and collect any ticks found. Collected ticks were placed in cryovials containing 70% ethanol and kept at room temperature at UPCH. A maximum of 20 ticks were placed in the same cryovial if they were completely covered by ethanol. One cryovial was used for each dog and ticks collected from their house were stored in another, with a specific labeling code linking all cryovials from the same location.

DNA extraction and real time-PCR

DNA was extracted from canine whole blood samples using the DNeasy Blood and Tissue kit (Qiagen, Valencia, CA) following the kit’s protocol. We added 3 µl of an internal amplification control (IAC) before the addition of the AL buffer. The IAC is a linearized primer containing a sequence of Arabidopsis thaliana and was added to identify potential qPCR false negative results.

A previously standardized qPCR targeting the E. canis disulfide bond formation protein gene (dsb) was performed using primers and probes shown in (Sup. Table 1) and the following thermocycling conditions: 50 °C for 2 min, 95 °C for 10 min, and 50 cycles of 95 °C for 15 s, followed by 60 °C for 1 minute^6,12. A dsb gene gblock gene fragment was used as the positive control (IDT, IA, USA). Water was used as the negative control. Samples with a qPCR cycle threshold (Ct) below 35 were considered positive.

Conventional PCR and Sanger sequencing

Samples that had a qPCR Ct value between 35 and 40 were run by conventional PCR targeting a segment of the Tandem Repeat Protein 36 gene (trp36) to confirm positivity, using the TRP36-F2 and TRP36-R1 (Sup. Table 1)¹⁷. All conventional PCR were performed in 25 μl reaction volumes containing 3.5 μl of nuclease-free water, 2.5 μl of each primer (at a 10uM concentration), 12.5 μl of GoTaq Green Master Mix 2X (Promega, Madison, WI), and 4 μl of DNA. Amplification was carried out by denaturation at 95 °C for 5 min, 35 cycles of denaturation (30 s, 95 °C), annealing (1 min, 50 °C), and extension (1 min, 72 °C), and a final extension of 72 °C for 5 min. This protocol was optimized from a previously published protocol¹⁷. A sample with a qPCR Ct of 24.33 was used as a positive control, while a qPCR negative sample and a blank sample (product of DNA extraction without the addition of a sample) were used as the negative controls.

Conventional PCR amplicons from three samples with a Ct lower than 35 (quality control) and 12 samples with a Ct between 35 and 40 were sequenced. These samples were amplified using 6 μl of nuclease-free water, 1.25 μl of each primer (at a 10uM concentration), 12.5 μl 2X Phusion Master Mix Buffer (Thermo Scientific, Waltham, MA), and 4 μl of DNA. Amplification was carried out by denaturation at 98 °C for 1 min, 35 cycles of denaturation (10 s, 98 °C), annealing (30 s, 54.6 °C), and extension (30 s, 72 °C), and a final extension of 72 °C for 5 min. The 12 samples with a Ct between 35 and 40 that were processed for sequencing, were amplified using the GoTaq Green Master Mix 2X as previously described while the three samples with Ct 

These 15 samples were sent to Macrogen Inc., Chile for purification and Sanger sequencing on the 3730xl DNA Analyzer platform (Applied Biosystems). Sequence evaluation and alignment were performed using Molecular Evolutionary Genetics Analysis (MEGA X). Sequences were compared between them and those in NCBI GenBank using the Basic Local Alignment Search tool (BLAST).

Statistical analysis

The covariates included in the analysis, defined a priori, were organized as dog-, house-, and owner-related factors (Sup. Table 2).

We conducted two statistical analyses with different outcomes to assess whether the associated risk factors remained consistent regardless of the Ct cut-off used: i) samples with a qPCR Ct E. canis.

We evaluated the association between a set of covariates, shown in the conceptual framework (Fig. 2), and both outcomes. We used mixed effects logistic regression models (melogit command on Stata) with a random intercept to account for clustering of dogs at the house level. Bivariate analyses were conducted to evaluate the association between E. canis positivity and dog-, house-, and owner-related factors for each aforementioned outcome.

Conceptual framework of factors hypothesized to be associated with E. canis infection among dogs.

Manual forward nested models were built to determine which variables entered the final multilevel multivariable regression models of factors associated with E. canis qPCR positivity. Here, variables were sequentially added and retained only if their inclusion resulted in a decrease in the model’s Akaike Information Criterion (AIC), which balances model fit and complexity. A total of 61 observations were excluded from this model due to missing values in the covariates (the total number of observations included was 327, which belong to 243 houses). Because 39/61 (63.9%) missing values were due to owners’ unwillingness to share their house income, we compared all factors included in the analysis as covariates between individuals who did not indicate their house income and individuals who did indicate their house income (Sup. Table 3). For the dog-related factors, one dog from the house was randomly selected. Multicollinearity between variables included in the final model was assessed using the Variance Inflation Factor (VIF).

Statistical analyses were done using a confidence interval of 95% in Stata 15.0 (Stata Corp., College Station, TX). Variables with a p-value