Comparison of serological and molecular panels for diagnosis of vector-borne diseases in dogs

Background Canine vector-borne diseases (CVBD) are caused by a diverse array of pathogens with varying biological behaviors that result in a wide spectrum of clinical presentations and laboratory abnormalities. For many reasons, the diagnosis of canine vector-borne infectious diseases can be challenging for clinicians. The aim of the present study was to compare CVBD serological and molecular testing as the two most common methodologies used for screening healthy dogs or diagnosing sick dogs in which a vector-borne disease is suspected. Methods We used serological (Anaplasma species, Babesia canis, Bartonella henselae, Bartonella vinsonii subspecies berkhoffii, Borrelia burgdorferi, Ehrlichia canis, and SFG Rickettsia) and molecular assays to assess for exposure to, or infection with, 10 genera of organisms that cause CVBDs (Anaplasma, Babesia, Bartonella, Borrelia, Ehrlichia, Francisella, hemotropic Mycoplasma, Neorickettsia, Rickettsia, and Dirofilaria). Paired serum and EDTA blood samples from 30 clinically healthy dogs (Group I) and from 69 sick dogs suspected of having one or more canine vector-borne diseases (Groups II-IV), were tested in parallel to establish exposure to or infection with the specific CVBDs targeted in this study. Results Among all dogs tested (Groups I-IV), the molecular prevalences for individual CVBD pathogens ranged between 23.3 and 39.1%. Similarly, pathogen-specific seroprevalences ranged from 43.3% to 59.4% among healthy and sick dogs (Groups I-IV). Among these representative sample groupings, a panel combining serological and molecular assays run in parallel resulted in a 4-58% increase in the recognition of exposure to or infection with CVBD. Conclusions We conclude that serological and PCR assays should be used in parallel to maximize CVBD diagnosis.


Background
When dogs develop clinical and hematological abnormalities, such as fever, cytopenias, hypoalbuminemia, hyperglobulinemia, polyarthritis or protein-losing nephropathy, veterinarians will often include vector-borne diseases among the differential diagnoses. In North America the number of recognized CVBDs has increased and now includes anaplasmosis (Anaplasma phagocytophilum or Anaplasma platys), babesiosis (Babesia canis and Babesia gibsoni), bartonellosis (numerous Bartonella sp.), heartworm disease (Dirofilaria immitis), ehrlichiosis (Ehrlichia canis, Ehrlichia chaffeensis, Ehrlichia ewingii, or Ehrlichia muris), hepatozoonosis (Hepatozoon canis and Hepatozoon americanum), Lyme disease (Borrelia burgdorferi), tularemia (Francisella tularensis), hemotropic Mycoplasma (multiple Mycoplasma sp.), Potomac Horse Fever (Neorickettsia risticii), and Rocky Mountain spotted fever (Rickettsia rickettsii). In addition, CVBDs such as E. ewingii and E. chaffeensis appear to have increased in incidence and geographic distribution due to the spread of Ambylomma americanum ticks northward throughout much of the United States. Also, B. gibsoni, which is frequently found in American Pitbull Terriers, can be transmitted directly to other dogs via fighting and bites from an infected animal [1] thus, this historically vector-borne pathogen can be directly transmitted among dogs. Although the mode(s) of transmission are incompletely understood, several Bartonella species are thought to be transmitted to dogs by fleas and ticks. These and other factors have contributed to an evolving appreciation of the role of CVBDs as a cause of disease in dogs. As the optimal treatment modalities differ for these diseases, it behooves veterinarians to choose CVBD diagnostic tests wisely, so as to economically and accurately evaluate exposure to and/or infection with a spectrum of vector-borne pathogens.
All diagnostic tests have inherent advantages and limitations. Serology relies on an immunologically appropriate and diagnostically detectable host immune response against one or more CVBD pathogens. As antibodies can persist for variable intervals after a pathogen is immunologically or therapeutically eliminated, serology does not confirm active or persistent infection in the patient, which is a diagnostic disadvantage. However, serology can be used to retrospectively confirm recent infection, by demonstrating seroconversion (i.e. a fourfold change in the patient's antibody titer between acute and convalescent serum samples) [1]. The persistence of circulating antibodies can also be an advantage of serology in that antibodies may be detectable during chronic intravascular CVBD infections, when a pathogen may be circulating below the limit of PCR detection or sequestered in tissues that are not routinely submitted for PCR testing. Another potential limitation of serology includes diminished specificity, due to antibody crossreactivity within or between CVBD genera [2][3][4][5][6][7][8]. Nevertheless, cross-reactivity between Anaplasma, Bartonella, Ehrlichia, and Rickettsia genera in dogs seems to be very unlikely, as dogs experimentally-infected with these pathogens develop very specific antibodies that do not cross react among genera [9][10][11][12][13][14][15].
Conversely, serological antigens chosen or available for assays may be too specific or mismatched to the etiological pathogens resulting in false negative results. These factors can result in the inability to accurately identify the infecting species or strain, which can have therapeutic implications for the patient. A technical limitation for the development of some CVBD serological assays is the inability to produce antigens in sufficient quantities to be used in indirect fluorescent antibody (IFA) or enzymelinked immunoabsorbent assays (ELISA) (examples include A. platys, E. ewingii, Babesia and Hepatozoon spp.), although the use of synthetic peptides may help overcome this limitation. With the advent of PCR testing, it has also become obvious that some dogs do not mount a detectable antibody response, despite persistent intravascular infection with CVBD pathogens [16][17][18][19][20].
Similar to serology, PCR has advantages and disadvantages for the diagnosis of CVBDs. PCR has the distinct advantage over serology of detecting "active" infection in a single sample from a single time-point. Additionally, PCR can be used to specifically target a pathogen at the species or strain level by using different PCR primer sets or by sequencing PCR products. While some a priori knowledge or assumptions about the DNA sequence of a pathogen are needed to design PCR-based assays, PCR does not require definitive knowledge of the pathogen DNA sequence. Additionally, PCR does not require that the pathogen(s) be isolated or their antigens produced to achieve the development and validation of an assay. Multiple pathogens or species can be detected using multiplex PCR assays [21][22][23] but these assays can be more challenging in the context of achieving optimal sensitivity for all pathogens targeted in the panel. Coinfecting pathogens may cause competition in the PCR reaction process [24,25]. Substantially higher concentrations of one pathogen compared to the other(s) can result in detection of only one organism despite the presence of a co-infection. The main limitation to PCR testing is the requirement for adequate template (nucleic acid of the target organism) in the patient sample to achieve amplification of the target DNA sequence. For vector-borne pathogens such as Anaplasma spp., B. burgdorferi, and Bartonella spp., it is well documented that the number of intravascular organisms fluctuates over time following transmission [10,11,13,14]; therefore, PCR testing at a single time point may produce a false negative result for an infected patient. Some other technical disadvantages of PCR-based testing include potential false negative results due to the presence of PCR inhibitors that were not removed during the nucleic acid purification process and the potential for laboratory contamination resulting in false positive reactions in patients that are not infected. The latter disadvantages can be minimized by the use of appropriate techniques, reagents and the incorporation of appropriate controls. Unless mechanisms are developed to overcome all of the limitations of PCR-based testing, PCR is unlikely to serve as a stand-alone assay for the diagnosis of many vector-borne infections.
In the current study, a panel of serological assays was compared to multiple PCR assays using samples from healthy dogs and dogs suspected of being infected with one or more CVBDs.

Sample group selection
Paired serum and ethylenediaminetetraacetic acid (EDTA)anti-coagulated blood samples from several defined groups of sick dogs were selected from a convenience sample of diagnostic accessions of the Vector-Borne Disease Diagnostic Laboratory (VBDDL), at the College of Veterinary Medicine, North Carolina State University. To facilitate a comparative analysis between serology and PCR, all patient samples were tested in parallel and in an identical manner, regardless of the test(s) originally requested by the attending clinician. In addition, paired serum and EDTA samples retrieved from storage at -80°C from 30 clinically healthy dogs examined during routine wellness appointments (Group I) were tested to establish background exposure to or infection with the defined population of CVBDs targeted in this study. Samples from 69 sick dogs (Groups II-IV) were selected on the basis of prior serology or PCR results. Group II consisted of 20 sick dog sample sets submitted by attending veterinarians from which no antibodies were detected using the serology panel defined below. The attending veterinarian for these cases only requested serologic testing. PCR was performed retrospectively (as described below).
Group III consisted of 25 sample sets from sick dogs that were seroreactive to at least one vector-borne pathogen. The attending veterinarian for these cases only requested serologic testing and PCR was performed retrospectively (as described below).
Group IV consisted of sample sets from 24 sick dogs that were previously PCR positive for Babesia, Bartonella, Ehrlichia, Anaplasma or Rickettsia species. The attending veterinarian for these cases only requested PCR testing (for one or more organisms). Both serologic testing and PCR testing, including testing for additional organisms (as described below) were performed retrospectively.

Serology
For this study, all serum samples were tested using a commercial ELISA based kit a and by IFA assays using a panel of VBDDL antigens. All antigens were grown in vitro or, in the case of B. canis, in vivo by personnel in the VBDDL using strains of canine or feline origin. Briefly, antibody responses to B. canis, Bartonella henselae, Bartonella vinsonii subspecies berkhoffii, E. canis and spotted fever group Rickettsia were tested by traditional IFA practices with fluorescein conjugated goat anti-dog IgG (Thermo Fisher Scientifics, Waltham MA 02452) [14].
Serum samples were diluted in phosphate buffered saline (PBS) solution containing 1% normal goat serum, 0.05% Tween-20 and 0.5% powdered nonfat dry milk (BioRad, Hercules, CA) to block non-specific antigen binding sites [15]. Seropositive samples were defined as having endpoint titers ≥ 1:64 using a twofold scale of 1:16 -1:8192. Dirofilaria immitis antigen, as well as antibodies to Anaplasma spp., Ehrlichia spp. and the C6 peptide of B. burgdorferi, were detected using a commercial in-house ELISA-based kit according to manufacturer's instructions. Serum from the healthy dog population had been previously tested using a version of the ELISA kit b which did not contain Anaplasma spp. antigens. Due to limited serum, results for these dogs were not determined for Anaplasma spp.

PCR testing
Following selection of samples for the various groups (Groups I-IV), all EDTA-anti-coagulated blood samples were randomized and tested in an operator blinded fashion. Ten microbial genera (Anaplasma, Babesia, Bartonella, Borrelia, Ehrlichia, Francisella, hemotropic Mycoplasma, Neorickettsia, Rickettsia, and Dirofilaria) were targeted using specific PCR assays as previously reported [16][17][18][19][20][21][22]. To further assess potential infection with Dilofilaria spp. PCR amplification of Wolbachia spp. was used as additional indirect evidence to support a diagnosis of canine heartworm disease [21][22][23]. In all cases in which blood samples were PCR positive, direct DNA sequencing was performed. Reference sequences for this study included the following GenBank accession numbers: AY055469 (Anaplasma phagocytophilum), haematoparvum) and CP000848 (Rickettsia rickettsii). Sequences were compared to the GenBank database using the Basic Local Alignment Search Tool.

Statistical analysis
Agreement between PCR-based assays and serological assays was assessed by calculating an un-weighted kappa statistic using a statistical program. c For agreement between PCR and serological assays for Anaplasma, the detection of any Anaplasma sp. by PCR was considered a positive result and the detection of a colorimetric change in the Anaplasma ELISA was considered a positive result. For agreement between PCR and serological assays for Babesia, the detection of any Babesia sp. by PCR was considered a positive result and the detection of an antibody titer ≥ 1:64 against B. canis was considered a positive result. For agreement between PCR and serological assays for Bartonella, the detection of any Bartonella sp. by PCR was considered a positive result and the detection of an antibody titer ≥ 1:64 against any Bartonella sp. was considered a positive result. For agreement between PCR and serological assays for Ehrlichia, the detection of any Ehrlichia sp. by PCR was considered a positive result and the detection of an antibody titer ≥ 1:64 against E. canis or a colorimetric change in the Ehrlichia spp. ELISA was considered a positive result. For agreement between PCR and serological assays for Rickettsia, the detection of any Rickettsia sp. by PCR was considered a positive result and the detection of an antibody titer ≥ 1:64 against R. rickettsii was considered a positive result. Agreement between PCRbased assays and serological assays could not be calculated for Borrelia spp., Dirofilaria sp. or Mycoplasma spp.

Results
The overall agreement between PCR and serology test results for dogs in Groups I to IV are summarized in Table 1. The number of serology versus PCR positive dogs within each Group for each pathogen is summarized in Table 2. Molecular and serological results for each pathogen and for each dog in each group are detailed in Tables 3, 4, 5 and 6. DNA of Neorickettsia, Wolbachia and Francisella spp. was not PCR amplified, from any dog in the study.

Group I: clinically healthy dogs
Of the 30 Group I healthy dog blood samples, 16 had no evidence of infection with or exposure to CVBDs (Table 3). Combined serological and molecular testing identified exposure to or infection with CVBDs in 14 healthy dogs. Thirteen dogs were seroreactive (antibody titer ≥ 1:64) to one (n = 12) or more (n = 1) antigens by IFA testing. No healthy dog was IFA seroreactive to B. canis, or B. vinsoni subsp berkhoffii antigens. One, six, and seven dogs were seroreactive to E. canis, B. henselae or R. rickettsii antigens, respectively. All 30 Group I dogs were negative by ELISA, including the one E. canis IFA seroreactor. Based upon PCR amplification and DNA sequencing, 6 dogs were infected with B. henselae, of which only two were B. henselae seroreactive. Three R. rickettsii seroreactors were PCR positive for B. henselae, one of which was also E. canis seroreactive. One dog was PCR positive for Mycoplasma haemocanis. This dog was seroreactive against B. henselae antigens but did not have detectable Bartonella DNA in its blood.
Group II: sick dogs in which CVBD antibodies were not detected and for which no PCR testing had been requested Nineteen of 20 Group II sick dogs had no evidence of infection with or exposure to CVBDs ( Table 4). The remaining dog was PCR positive for B. gibsoni and seronegative to all antigens tested. Babesia gibsoni IFA testing was not performed as a component of this study.
Group III: sick dogs in which CVBD antibodies against at least one pathogen were detected and for which no PCR testing had been requested Of the 25 dogs in Group III, two dogs were PCR positive ( Table 5). One dog was actively infected with R. rickettsii based upon PCR and DNA sequencing and was seroreactive against R. rickettsii with a titer of 1:128. The other dog had antibodies against B. burgdorferi and Anaplasma spp. detected by ELISA and was PCR positive for Mycoplasma haemocanis. All eight E. canis IFA seroreactive dogs were positive by ELISA; however, none were PCR positive. Six dogs were seroreactive to B. henselae antigens and one to B. vinsonii subsp. berkhoffii antigens, but were not Bartonella PCR positive.
Group IV: sick dogs from which DNA of at least one CVBD pathogen was detected by PCR testing, but serological testing had not been requested Group IV, consisted of 24 dogs that were previously PCR positive with Ehrlichia, Anaplasma, Babesia or Bartonella spp. When serological and PCR assays were performed beyond the PCR requested by the attending clinician exposure to or infection with additional CVBD of different genera were identified in 13 dogs (Table 6).

Comparative serological and molecular testing
For the PCR assays requested by the attending veterinarian there was complete agreement between VBDDL PCR results and repeat blinded testing. Overall, the agreement between PCR and serology ranged from 28.6% for the healthy dogs to 4 -50% for sick dogs depending on groups ( Table 1). The agreement between PCR and serology depended greatly on the tested species. There was strong agreement (kappa: 0.637, 95% CI 0.453-0.822) between PCR-based assays and serological assays for Babesia spp. There was moderate agreement (kappa: 0.353, 95% CI 0.117-0.589) between PCR-based assays and serological assays for Anaplasma spp. There was moderate agreement (kappa: 0.289, 95% CI 0.130-0.447) between PCR-based assays and serological assays for Ehrlichia spp. There was poor agreement (kappa: 0.111, 95% CI -0.081-0.304) between PCR-based assays and serological assays for Bartonella spp. There was poor agreement (kappa: 0.069, 95% CI -0.003-0.141) between PCR-based assays and serological assays for R. rickettsii.

Discussion
Development of DNA amplification techniques or other molecular approaches by which CVBD pathogens can be detected and identified to the species or strain level in diagnostic samples continues to facilitate a more rational basis for selection of treatment modalities. However, from a diagnostic perspective, it does not appear that either PCR or serological assays alone are adequate to identify infection or exposure in every case ( Table 2). The results of this study demonstrate that evidence of exposure to or infection with vector-borne pathogens in dogs can be increased by using a combination of serological and PCRbased assays in parallel. Use of such comprehensive testing not only resulted in detection of exposure or infection with CVBD pathogens in a greater number of dogs, it also resulted in the identification of co-infections that would have been missed if only one testing method was used. While the antibiotic treatment history and the rationale for the submission of each sample (other than the healthy control group) was not known, each group of dogs is likely to represent a scenario in which clinical, therapeutic or public health decisions are likely to be different had the clinician submitted for both PCR and serological assays in parallel.
Number of dogs positive for both serology and PCR (agreement) are listed between brackets. Group I (clinically healthy dogs) could represent dogs similar to those being screened as blood donors. Depending on individual hospital blood donor screening practices (Rickettsia, Borrelia or Dirofilaria exposure or infection are not typically included as exclusion criterion for canine blood donors), the use of PCR and serological assays in parallel increased the number of donors that could be excluded to 33% (10/30) compared to 23% (7/30) if either PCR or serology were used alone.
Group II (sick dogs in which CVBD antibodies were not detected and for which no PCR testing had been requested) represents a population of dogs for which only serological assays were requested by the attending veterinarian. It seems plausible to assume that after the requested testing these dogs were considered less likely to be infected with or exposed to CVBD. However, the use of PCR and serological assays in parallel would have facilitated the identification of an infection in 5% (1/20) of those cases.  Group III (sick dogs in which CVBD antibodies against at least one pathogen were detected and for which no PCR testing had been requested) represents a population of dogs for which only serological assays were requested by the attending veterinarian. The addition of PCR testing facilitated the identification of an additional organism in 4% (1/25) of these cases. It is also interesting to consider that if these clinicians had only selected PCR assays, exposure to CVBD would have been missed in 92% (23/25) of the cases.
Group IV (sick dogs from which DNA of at least one CVBD pathogen was detected by PCR testing, but serological testing had not been requested) represents a population of dogs for which only one or more PCR assay(s) were requested by the attending veterinarian. The use of more extensive PCR testing and serological assays would have identified exposure to or infection with additional CVBD (i.e. potential co-infections) in 54% (13/24) of the cases. Additionally, if the attending clinician had only requested serological assays on these samples, exposure to or infection with additional CVBD (i.e. potential co-infections) that would alter the case management would have either been missed in 58% (14/24) of cases, or CVBD would have been missed entirely in 13% (3/24) of the cases. Examples where the identification would alter the case management include a dog that had antibodies against B. canis but was actually PCR positive for B. gibsoni (i.e. Group IV #14 or #16) or a dog that had evidence of infection with A. phagocytophilum and B. burgdorferi and exposure to or infection with R. rickettsii but was also PCR positive for B. gibsoni. In some cases the additional information would not have changed case management. An example of such is the dog that was IFA seroreactive against E. canis yet PCR positive for E. ewingii. Treatment for ehrlichiois may not have been started based upon seroreactivity, but would have been indicated on the basis of the PCR result. There are several specific reasons for lack of agreement between PCR-based and serological assays, but differences in sensitivity and specificity between assays are of primary importance for interpretation. Specific examples that will result in differences include timing of sample collection, biological behavior of the pathogen and variation in host immune responses to antigens. As our understanding of CVBD and individual host responses to infection increases it is becoming clearer that our currently available diagnostic assays all have shortcomings.

Conclusions
Cumulatively, the results of this study suggest that the use of PCR and serological assays in parallel is likely to increase the detection of infection with or exposure to CVBDs. This clinical scenario is not unknown to the veterinary profession. For example, veterinarians are comfortable accepting that a multi-modal approach, including antigen and antibody testing combined with diagnostic imaging, is ideal to diagnose feline heartworm infection [26]. However, it is important to note that the increased detection of infection with or exposure to CVBD does not always confirm disease causation and the attending veterinarian is still responsible for interpretation of test results in light of the patient's clinical signs, laboratory abnormalities and response to treatment. Clinicians must consider the prevalence of each sign or laboratory abnormality with each type of infection and a failure to respond to appropriate treatment should prompt the consideration of alternative diagnoses. In conclusion, it still appears that an ultimate test for the diagnosis of CVBDs remains elusive. To optimize clinical decision making, clinicians should consider using panels that include serological and PCR assays in parallel to maximize their chances of detecting infection or exposure to CVBD pathogens.