Evaluation of SARS-CoV-2 identification methods through surveillance of companion animals in SARS-CoV-2-positive homes in North Carolina, March to December 2020

Ethics statement

The authors confirm that the ethical policies of the journal, as noted on the journal’s author guidelines page, have been adhered to and the appropriate ethical review committee approval has been received.

Study design and setting

This was a prospective, observational descriptive study. Households were recruited from an approximately 100-mile radius surrounding Durham, NC, from March 2020–March 2021. A written consent form was obtained for each study participant and pet enrolled in the study.

Sample collection

This study was a collaboration between Duke University and North Carolina State University, and was part of an ongoing study known as the Molecular and Epidemiological Surveillance in Suspected Infection (MESSI) study. Human sample collection was approved by the Duke Institutional Review Board (IRB) under the MESSI protocol number Pro00100241. Animal collection protocols were reviewed and approved by the Duke Institutional Animal Care and Use Committee (IACUC) before approaching an owner for permission to collect samples under the Molecular and Epidemiological Surveillance in Suspected Infection in House Pets (MESSI-HP) protocol number A079-20-03. The MESSI study team consisted of lab members with expertise in epidemiological field research and human sample collection. The MESSI-HP study team included one member of the MESSI field team with animal handling experience, one veterinary technician, and one veterinarian, who was present for house pet serum sample and oral and/or rectal swab collection. Inclusion and exclusion criteria are listed in Table 1 for humans and house pets.

The MESSI study team collected serial biological samples (nasopharyngeal swabs, serum samples) of humans throughout the course of their illness (days 0 (day of enrollment in MESSI study), 1, 3, 7, 14, 21, 28, 45, 60, 120).

Biological samples (oral swabs, rectal swabs, serum samples) were collected from house pets (dogs and cats) on visit days 0, 1, 3, 7, 14, 21, 28, 45, 60, 120, and/or 180. Some sample collection timepoints were omitted due to challenges with pet compliance or insufficient team members present who were comfortable assisting in animal sample collection. Oral and/or rectal swabs from exposed house pets were collected from days 0–60. Serum samples from exposed house pets were collected on day 28 or later to allow for seroconversion and to limit the number of people exposed to an actively infected or clinically ill human.

Oral swabs were collected by placing a sterile polyester-tipped applicator (manufacturer Puritan 25-806-1P) into the house pet’s mouth and rubbing the cheeks, gums, and tongue. Rectal swabs were collected by placing an applicator into the rectum and swabbing gently in a circular motion. Polyester-tipped applicators were immediately placed in viral transport media (VSM01) from Dasky (components included Hanks’ balanced salts, sucrose, penicillin, gentamicin, streptomycin sulfate, amphotericin B, nonessential amino acid, and phenol red) in a disposable tube for transport and storage. Serum samples were collected using standard venipuncture and processing for transport and storage.

Sample types and labels

Polymerase chain reaction for SARS-CoV-2

To assess for the presence of viral SARS-CoV-2 RNA in exposed PCR samples and unexposed PCR controls, standard methodology was used as per the recommendation of the World Health Organization at the time (testing performed November 2021) (Diagnostic testing for SARS-CoV-2). Automated QIAsymphony (Qiagen LLC, Germantown, MD) two-step RT-PCR and World Health Organization E_Sabeco primer-probe sets (Charite, Berlin) were used (Corman et al., 2020). To evaluate the number of viral copies detected within a sample, quantitative PCR was carried out on a QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA). Samples with LOQ ≥ 62 RNA copies/mL (1.79 log₁₀) from 800 µL of sample were considered positive based on a threshold determined by the manufacturer to indicate a positive test. To improve the accuracy in reporting positive tests, samples with LOQ ≥ 62 RNA copies/mL were subsequently run on the COBAS 6800 system, a qPCR test approved for emergency use by the FDA in 2021 to detect SARS-CoV-2 RNA in swab samples (Roche, Basel, Switzerland) (Roche Diagnostics, 2021). This test is typically reported qualitatively to users. The COBAS 6800 is a dual target assay where target 1 is a sequence-specific to SARS-CoV-2 and target 2 is more general to the Sarbeco virus subgenus. Detection of target 1 is considered indicative of the presence of SARS-CoV-2, whereas detection of target 2 typically indicates the presence of SARS-CoV-2 in low concentrations. Samples were considered positive for SARS-CoV-2 with a cycle threshold (Ct) value < 38 for target 1 or target 2 (Pujadas et al., 2020). Samples were only reported as positive if they were positive on both the QuantStudio 3 RT-PCR and COBAS-6800 systems. Additional PCR methods, including a description of primers, for the QuantStudio 3 RT PCR system are provided in supplementary materials (Supplemental Information 1).

Serology for SARS-CoV-2

Enzyme-linked immunosorbent assay (ELISA) was used for the detection of house pet IgG and IgM to SARS-CoV-2 in the pre-2019 serological controls and exposed serum samples as previously described by (Bienzle et al., 2022). Adsorption immunoassay plates (96-well, ThermoFisher, Mississauga, ON) were coated at 4 °C with 2 µg/mL of His-tagged SARS-CoV-2 S1 (GenScript, Piscataway, NJ) and incubated overnight. The following day, wells were washed 3x, blocked with 3% skim milk in Tris buffer for 60 min, washed 3x, and then 60 µL of five 3-fold dilutions (1:100, 1:300, 1:900, 1:2,700 and 1:8,100) of each serum sample was added. Plates were incubated for 120 min, washed 3x, and secondary antibodies conjugated to horseradish peroxidase (HRP) and diluted 1:5,000 were added for 60 min. Wells were washed 3x, and HRP activity was confirmed visually by adding trimethyl benzidine (TMB) substrate. Reactions were ceased with sulfuric acid, and optical density (O.D.) at 450 nm was read. Secondary antibodies were derived from goats and consisted ofanti-dog IgG, anti-dog IgM, anti-cat IgG, and anti-cat IgM (all from Abcam, Waltham, MA). Control samples for ELISA validation differed from the pre-2019 controls. For cats, the positive control ELISA validation sample consisted of serum from an experimentally infected SARS-CoV-2 cat (kindly provided by Y. Kawaoka, Madison, WI; positive feline control, used at 1:5,000 in ELISA), and negative controls included three different batches of pooled cat serum from 2016 or 2017, two serum samples from cats with feline infectious peritonitis (due to mutated feline enteric coronavirus), and one serum sample from a cat with osteomyelitis and hyperglobulinemia. For dogs, negative controls included three different batches of pooled dog serum collected in 2017, 2018, and 2019. Since serum from experimentally infected dogs was not available, the positive control for ELISA validation consisted of a serum sample from one study dog (exposed serum sample) with a high O.D. Each ELISA plate included 16 wells that were not coated with recombinant protein (blank), five replicate 1:100 dilutions of species-specific negative control samples, and five replicates of each of 3 dilutions of the positive control and test samples (1:100, 1:200, 1:400).

From the MESSI-HP serum samples and the pre-2019 serological controls, serum with a volume of at least 0.5 mL was divided into two aliquots, and the ELISA was performed on each aliquot. Staff performing the ELISA were masked to the identity of samples and group (pre-2019 serological control or exposed serum sample) for the second round of serology. These samples were used to calculate a false positive rate from initial reporting of “positive” and “negative”.

Questionnaires

Three questionnaires were provided to the primary owner. Each questionnaire was in paper copy and filled out either by a primary owner (an owner who assumed primary responsibility for general caretaking of the pet and lived with the pet full time), or by one of the study team members while verbally asking each question to the owner. When possible, the same primary owner provided answers at each visit. Questionnaire 1 (Supplemental Information 2 collected demographic information about each pet and was given to owners at the enrollment visit. Questionnaire 2 (Supplemental Information 3) collected clinical signs for each pet at every visit and contained 13 specific questions regarding the presence or absence of specific symptoms. Only questions with at least one “yes” and one “no” for each column were included for analysis. An additional column was added for the “presence” or “absence” of any symptom as dictated by the responses to the other symptom questions. Questionnaire 3 (Supplemental Information 4) contained 17 specific questions regarding human-animal interactions and animal behavior and was provided at the time of exposed serum sample collection, which occurred between days 28-180. Only questions free from errors in owner reporting that might have led to a spurious outcome were analyzed. Errors included owners circling multiple answers (as opposed to one answer) for multiple choice questions, failing to answer any part of the question, and/or answering the main question but failing to fill out related subquestions.

Statistical analysis

Statistical tests were conducted using R 4.1.1 (R Core Team, 2021). Descriptive statistics were calculated for the primary outcome of the percent of exposed dogs and cats positive on PCR and serology, as well as median and range of animal age. Lin’s concordance correlation coefficient for agreement on continuous measures was used to determine agreement between round 1 and round 2 O.D.s. Fisher’s exact test was used for univariable analysis to calculate p-values and odds ratios evaluating associations between 14 behavioral factors and seropositivity. Adjusted p-values were calculated using Bonferroni’s correction to account for multiple comparisons. The full reproducible code is available at https://github.com/t-gin/SARS-CoV-2_in_housepets.

Animal demographic data and overview of sample types

Table 2 provides an overview of the number of each sample type (exposed PCR and serum samples, unexposed PCR controls, and pre-2019 serological controls). Sixty-four house pets from 32 households with SARS-CoV-2-PCR-positive humans were enrolled (exposed PCR and serum samples), and six house pets from households with SARS-CoV-2-PCR-negative humans were enrolled only in the PCR portion of the study as negative controls (unexposed PCR controls). The 43 pre-2019 serological controls came from 24 dogs and 19 cats. Figure 1 shows the types of samples collected from SARS-CoV-2-positive households by species.

Exposed PCR samples were collected from day 0 through day 60, with most swabs taken at day 0 (41 oral, 19 rectal) and day 28 (34 oral, 22 rectal). Thirty-five house pets (9 cats and 26 dogs) had oral or rectal swabs taken at more than one time point.

Thirty-nine exposed serum samples were submitted for antibody testing. One sample was removed due to suspected erroneous results, as indicated by an outlier during visualization (Fig. 2), resulting in 38 samples for analysis. This included exposed serum samples from 24 dogs and 14 cats collected on days 28 (n = 13 samples), 60 (n = 2 samples), 120 (n = 22 samples), and 180 (n = 2 samples).

Evaluating the ELISA to generate high confidence seropositivity calls

The anti-SARS-CoV-2 IgG and IgM O.D.s from replicate testing for the exposed serum samples were evaluated alongside the O.D.s for the pre-2019 serological controls. Four of the 38 exposed serum samples (one from dogs and three from cats) did not have adequate volume for two aliquots; these were excluded from the analysis to establish an O.D. cut-off. To visually assess for quantitative consistency, O.D.s for round 1 and round 2 serology from the exposed serum samples were plotted in Fig. 2, as well as O.D.s of the pre-2019 serological control samples. Additionally, Lin’s concordance correlation coefficient was calculated to measure agreement between the round 1 and round 2 O.D.s for dog and cat IgG and IgM. The correlation coefficient for dog IgG was 0.86, cat IgG was 0.96, dog IgM was 0.64, and cat IgM was 0.85, indicating relatively consistent measurements between both runs for dog and cat IgG and cat IgM, but not dog IgM. On initial visual assessment of the 34 round 1 and round 2 samples, two distinct clusters were identified for dog and cat IgG. A line was drawn as a proposed cutoff point to distinguish samples with O.D.s that were visibly higher or lower. The pre-2019 serological controls were then plotted over the exposed serum samples, and the previously drawn cutoff line was adjusted such that a large majority of the samples above the line were exposed serum samples. The chosen cutoffs represent one option for delineating between the distinct clusters of exposed serum sample O.D.s and minimizing the number of pre-2019 serological controls above the cut-off line (false positives). Dog and cat IgM samples did not cluster such that a distinct cutoff point could be established even before adding the control samples. Figure 2 graphically represents this, where positive samples cluster distant from negative pre-2019 serological controls and negative exposed house pet samples.

False positive rates

With the new cutoff points chosen, the false positive rate for IgG seropositivity in our pre-2019 serological controls was 0/24 (0%) in dogs and 1/19 (5.3%) in cats. A false positive rate for IgM was not calculated, as samples did not cluster such that a distinct cutoff point could be established. This contrasts with the ELISA cut-off procedure described earlier in our methods section, as well as in a previous manuscript, which resulted in a false positive rate for IgG seropositivity in the pre-2019 serological controls of 3/24 (12.5%) in dogs and 6/19 (31.6%) in cats. (Bienzle et al., 2022) Based on that same ELISA cut-off procedure, the false positive rate of IgM for the pre-2019 serological controls was 8/24 (33.3%) in dogs and 7/19 (36.8%) in cats.

Serology results

Based on the seropositive thresholds established with the pre-2019 serological controls, 9/24 (37.5%) dogs and 4/14 (28.6%) cats were IgG-positive. A breakdown of positive samples by timepoint includes 7/13 (53.8%) day 28 samples, 0/2 (0%) day 60 samples, 6/22 (27.3%) day 120 samples, and 1/2 (50%) day 180 samples. A summary of results from seropositive animals is provided in Table 3.

PCR results

Three of 64 (4.8%) animals from 2/32 (6.3%) SARS-CoV-2-positive households were deemed SARS-CoV-2 positive based on a positive result from the QuantStudio 3 RT-PCR and the COBAS 6800 analyzer. Table 4 provides testing information on all five positive samples from the three animals mentioned, as well as one QuantStudio 3 RT-PCR-positive, COBAS-negative sample, which came from a cat in the same household as one of the QuantStudio 3 RT-PCR-positive dogs. The QuantStudio 3-PCR-positive, COBAS-negative cat had no target 1 or 2 genetic material detected on the COBAS analyzer. All nine of the IgG-positive dogs and all four of the IgG-positive were found to be RT-PCR-negative by the QuantStudio 3 between days 0–28. Among the three QuantStudio 3-RT-PCR-positive animals, two underwent serological testing, both on day 60. Neither animal was positive for anti-SARS-CoV-2 IgG.

Analysis of questionnaire data related to seropositivity

Of the 38 animals with serology results, 36 (23 dogs, 13 cats) had a complete Questionnaire 2, and 38 (24 dogs, 14 cats) had a complete Questionnaire 3 from the time of sample collection. Questionnaires were analyzed with serum data but not PCR data due to the large number of missing questionnaires from animals with PCR but not serology and very few PCR-positive animals.

Fourteen exposures were tested for a relationship with IgG seropositivity in dogs and cats. Associations and calculated statistics are reported in Table 5. No statistically significant associations were found in cats. In dogs, prior to applying Bonferroni’s correction, a statistically significant positive association was found between an IgG-positive ELISA and the owner reporting that dogs were allowed on the owner’s bed (p = 0.04) or furniture (p = 0.048). A negative association was identified between dogs within IgG-positive ELISA and owners reporting that dogs were known to lick plates in the dishwasher (p = 0.04). None of these associations were found to be significant following Bonferroni’s correction, which is represented by the “adjusted p-value” in Table 5.

Conclusions regarding symptom data were limited by the statistically underpowered number of observations of each symptom in the dataset. As such, symptom data was not found to be significantly associated with a positive IgG serology (p = 0.40). Three out of nine (33.3%) IgG-positive dogs were reported to have symptoms at one or more timepoints during sample collection. A table of the testing timeline and results for each of these dogs is included in supplementary materials (Supplemental Information 5). No symptoms were reported in any house pet with a positive SARS-CoV-2 PCR. No symptoms were reported in any cats with serology performed.