Cat and dog serum samples collected during 2019 (pre–COVID-19 cohort, n = 45 each) were obtained from the serum bank of Utrecht University (Utrecht, the Netherlands). Paired and postinfection serum samples of feline coronavirus (FCoV) type I–infected specific pathogen-free (SPF) cats (n = 9) were obtained from SPF cats infected with FCoV strain UU2 or RM in a previous study (16). The SARS-CoV-2–exposed cohort consisted of 44 serum samples from stray cats roaming on SARS-CoV-2–positive mink farms (17) and 1 serum sample of a dog from a COVID-19–confirmed household. The 2020 cohort is composed of domestic cat and dog serum or plasma samples (n = 500 each) that were sent to the University Veterinary Diagnostic Laboratory or the Veterinary Microbiological Diagnostic Center at Utrecht University for routine diagnostics during April–May 2020. Data on SARS-CoV-2 exposure of these animals was not available. All samples were stored at −20°C until use and heat-inactivated at 56°C for 30 min before use.

We produced streptavidin–tagged SARS-CoV-2 S1 and RBD proteins in eukaryotic cells as described (18,19), and cloned and similarly produced streptavidin-tagged bovine coronavirus (BCoV) S1 and HCoV-229E S1. SARS-CoV-2 N protein was obtained from Sino Biological (https://www.sinobiological.com). We produced mouse Fc-tagged FCoV type I S1, FCoV type II S1, or BCoV S1 proteins as described (20). Vesicular stomatitis virus (VSV) pseudotyped with SARS-CoV-2 S protein (SARS2-VSV) was prepared as described (18) and titrated on Vero E6 cells.

We first screened samples from the 3 cohorts with indirect ELISAs for the different proteins as described (20). In brief, high-binding microtiter plates were coated with equal molar amounts of protein (1 pmol/ well after optimizing by using checkerboard titration), diluted in phosphate-buffered saline, and blocked with blocking buffer (phosphate-buffered saline containing 0.05% Tween-20 and 5% milk powder). A standard 1:50 dilution of serum samples or serial 2-fold dilutions of serum samples starting at a 1:50 dilution were added to the wells. After incubation for 1 h at 37°C, plates were washed and subsequently incubated with horseradish peroxidase (HRP)–conjugated secondary antibody (1:4,000 for goat anti-cat IgG/HRP; Rockland Immunochemicals, Inc., https://rockland-inc.com) and 1:6,000 for goat anti-dog IgG/HRP; Cappel, http://ziobio.com) diluted in blocking buffer for 1 h at 37°C. Peroxidase reactions were visualized by incubation with 3,3′,5,5′-tetramethylbenzidine (10 min at room temperature) and quenching with sulfuric acid. Optical densities (ODs) were measured at 450 nm. Cutoff values were determined at 6-fold SDs above the mean value of reactivity of all negative serum samples from the pre–COVID-19 cohort (19).

To verify that the 2 betacoronavirus infections in dogs (SARS-CoV-2 and canine respiratory coronavirus [CRCoV]) can be distinguished serologically, we designed an antigen S1 adsorption assay. We incubated serum samples with Strep-Tactin Sepharose Beads (IBA Lifesciences, https://www.iba-lifesciences.com) conjugated with S1 protein of SARS-CoV-2, BCoV, or HCoV-229E and titrated mock-absorbed and protein-absorbed serum samples in the ELISA. We expressed IgG titers as the reciprocal of highest serum dilution resulting in OD values above the cutoff value.

We conducted a VN assay by using luciferase-encoding VSV particles pseudotyped with S protein of SARS-CoV-2 (SARS2-VSV), which was conducted on Vero E6 cells in a 96-well plate (18). Antigenicity of SARS2-VSV was validated previously, and VN titers (VNTs) for SARS2-VSV correlated well with those for live SARS-CoV-2 (18). Samples (starting at a 1:8 dilution) were serial diluted 2-fold and mixed 1:1 with SARS-2-VSV. Mixtures were preincubated at 37°C for 1 h and used for inoculation on cells. Twenty-four hours postinfection, cells were lysed and relative luminescence units (RLU) of luciferase activity was determined as described (18). RLU reduction rates of samples were calculated by using the formula (Figure 6)

Sample neutralization titers were determined by using the reciprocal of the highest dilution that resulted in >50% reduction of luciferase activity. A VNT >16 was considered positive (21).

All statistical analyses were performed by using Prism version 7.04 for Windows (GraphPad, https://www.graphpad.com). The Pearson correlation coefficient was calculated to determine the correlation between different ELISA ODs and VNTs. The 95% CIs were determined by using the modified Wald method.

Serum samples from the pre–COVID-19 cohort were tested against SARS-CoV-2 antigens to screen for potential cross-reactive antibodies elicited by endemic coronaviruses in cats and dogs because they are natural reservoirs of several coronaviruses (i.e., FCoV [genus Alphacoronavirus] in cats, canine coronavirus [CCoV; genus Alphacoronavirus] and CRCoV [genus Betacoronavirus] in dogs) (20,22,23). We summarized sequence identities of SARS-CoV-2 antigens used and matching endemic coronavirus antigens (Table 1). FCoV type I S1 was used as an additional antigen to assess the reactivity of cat serum samples. For dog serologic analysis, FCoV type II S1 (92.1% similar to S1 of CCoV) was used as a proxy antigen for CCoV, and BCoV S1 (95.7% similar to S1 of CRCoV) was used as a proxy antigen for CRCoV. Many serum samples were positive for FCoV and BCoV S1, but all samples were negative for antibodies against SARS-CoV-2 S1 and RBD (Figure 1). Because of limited sample volumes, a selection of serum samples (n = 34 for cats and n = 24 for dogs) was tested for SARS-CoV-2 S–bearing VSV pseudovirus (SARS2-VSV) neutralization, and all showed negative results (VNT <16).

A total of 8 (17.8%) of 45 pre–COVID-19 cat serum samples and 1 (2.2%) of 45 dog serum samples showed positive results in the SARS-CoV-2 N protein ELISA (Figure 1, panels A, B). To explore this finding, we analyzed paired serum samples of SPF cats infected with FCoV (Figure 1, panel C). Serum samples from uninfected SPF cats were negative. After FCoV infection, 8 (88.9%) of 9 cats had antibodies reacting with SARS-CoV-2 N protein. When compared with S1 and RBD proteins, we found that the N protein was more conserved among CoVs (Table 1), which might explain the cross-reactivity between FCoV and SARS-CoV-2 detected in our ELISAs.

We tested the serum of a dog from a COVID-19–confirmed household, as well as serum samples from SARS-CoV-2–exposed stray cats found in the surroundings of SARS-CoV-2–positive mink farms (17). These cats had access to the stables and cages in which the minks were housed. This cohort was expected to contain a higher number of SARS-CoV-2–positive samples because of close contact between the cats and minks and the dog and its owner and was a source of suitable samples for validation of our ELISA and VNT. A total of 11 (24.4%, 95% CI 14.1%–38.8%) of 45 serum samples from 10 cats and 1 dog were positive by ELISA for SARS-CoV-2 S1 and RBD, and 10 (22.2%, 95% CI 12.4%–36.5%) of 45 samples (were reactive against SARS-CoV-2 N protein (Figure 2, panel A). All S1- and RBD-positive samples could neutralize SARS2-VSV infections, but N protein positivity and VN ability were not well associated (Figure 2, panel B).

OD values obtained for the SARS-CoV-2 S1 and RBD ELISAs showed a strong correlation with each other (R = 0.95), and both correlated well with VNT (R = 0.87) (Figure 3, panels A–C). Conversely, only a poor correlation was observed between OD values obtained for N protein ELISA and VNT (R = 0.57) (Figure 3, panel D). These data validate SARS-CoV-2 S1 and RBD and exclude N protein as antigen for serologic screening of cat and dog serum samples.

A total of 500 cat samples from the 2020 cohort were tested by using SARS-CoV-2 S1 and RBD ELISAs (Figure 4, panels A, C). FCoV type I S1 was included as an additional antigen in the ELISA, and 71% of cat samples were FCoV type I antibody positive. Six cat samples were positive for SARS-CoV-2 S1 and RBD, and an additional 6 samples were positive only for RBD (Figure 4, panel C). We have summarized results of different tests (Table 2). We tested by VN assay all samples positive for SARS-CoV-2 S1 or RBD by ELISA , together with 50 randomly chosen samples that showed negative results in the S1 and RBD ELISAs. Two samples that reacted with SARS-CoV-2 S1 and RBD were able to neutralize SARS2-VSV infection, and all ELISA-negative samples were also negative in the VN assay (Table 2; Figure 4, panel C). On the basis of results obtained for SARS-CoV-2–exposed animals, we defined a seropositive sample as any sample being ELISA positive for SARS-CoV-2 S1 and RBD, and with a VNT >16. Samples that did not consistently show diagnostic thresholds (ELISA positive for S1 and RBD, but VNT <16) were considered as being suspected (Table 2). Accordingly, 2 (0.4%, 95% CI 0.01%–1.55%) of 500 domestic cat samples with unknown SARS-CoV-2 exposure had reached the diagnostic thresholds, and henceforth were confirmed as seropositive. Four serum samples were defined as suspected.

We tested 500 dog samples by using the SARS-CoV-2 S1 and RBD ELISAs (Figure 4, panels B, D). FCoV type II S1 was included as an additional antigen, and results showed that 40.4% were positive for FCoV type II S1 antibody (indicator of CCoV exposure). Nine samples were positive for SARS-CoV-2 S1, of which only 1 was positive for RBD (Table 2; Figure 4, panel D). Only the sample that reacted with SARS-CoV-2 S1 and RBD was able to neutralize SARS2-VSV. Randomly chosen ELISA negative samples (n = 50) were negative in the VN assay (Table 2; Figure 4, panel D). Thus, 1 (0.2%, 95% CI, <0.01%–1.24%) of 500 of domestic dog samples with unknown SARS-CoV-2 exposure was considered seropositive.

The 2 seropositive dog samples also contained antibodies against CRCoV, which belongs to genus Betacoronavirus, as does SARS-CoV-2 (Figure 5). To corroborate SARS-CoV-2 seropositivity, we performed an antigen S1 adsorption assay with S1 proteins of SARS-CoV-2 or BCoV (proxy for CRCoV). HCoV-229E (genus Alphacoronavirus) S1 was used as a control. Although adsorption of 229E S1 did not change ELISA reactivity for serum samples against SARS-CoV-2 and BCoV antigens, adsorption of SARS-CoV-2 and BCoV S1 specifically removed ELISA reactivity against the corresponding protein (Figure 5). These data confirmed that ELISA reactivity against SARS-CoV-2 for these 2 dog samples is specific, in accordance with the screening of CRCoV-positive pre–COVID-19 dog samples described earlier, which did not show cross-reactivity with SARS-CoV-2 S1 in our ELISAs.