Detection of Feline Leukemia Virus RNA in Saliva from Naturally Infected Cats and Correlation of PCR Results with Those of Current Diagnostic Methods

Abstract

A novel diagnostic test for feline leukemia virus (FeLV) RNA in saliva from naturally infected cats is described in this study. We evaluated different diagnostic tests and compared them with the widely used enzyme-linked immunosorbent assay (ELISA) for the detection of p27 in the diagnosis of FeLV. Blood samples from 445 cats were tested for the presence of provirus by real-time PCR and plasma and saliva specimens from those cats were tested for the presence of viral RNA by real-time reverse transcription (RT)-PCR and for the presence of p27 by ELISA. In comparison to conventional ELISA, the diagnostic sensitivity and specificity of the detection of salivary FeLV RNA by real-time RT-PCR were found to be 98.1 and 99.2%, respectively. Detection of viral RNA in saliva had a positive predictive value of 94.6% and a negative predictive value of 99.7%. The kappa value was 0.96, demonstrating an almost perfect agreement between both tests. Furthermore, we confirmed previous results showing that a number of cats which tested negative for the presence of p27 in plasma were in fact positive for the presence of DNA provirus in blood specimens (5.4%). However, 96.4% of these latently infected cats did not shed viral RNA in saliva; therefore, we assume that these cats are of relatively low clinical importance at the time of testing. This study shows considerable diagnostic value of the detection of saliva FeLV RNA in naturally infected cats. This new diagnostic method has advantages over the conventional ELISA, such as less invasive sample collection and no requirement for trained personnel.

Feline leukemia virus (FeLV) was first described by Jarrett et al. in 1964 (13) and is one of the most common fatal pathogens affecting cats worldwide. FeLV is an enveloped, positive-sense, single-stranded RNA virus that, once released in the environment, is not able to survive long periods on dry surfaces (3). Infection results mostly from oronasal exposure to saliva and nasal secretions containing high levels of the virus, especially through mutual grooming and sharing of food dishes and water bowls (10). Vertical transmission occurs occasionally (10) but it is of secondary importance. Commercially available vaccines against FeLV are generally safe, but are not 100% efficacious; therefore, preventing exposure of susceptible cats to viral sources remains an important measure to prevent infection. Collectively, the relatively high prevalence of FeLV in domestic cats in Switzerland (17.1%, determined by real-time PCR) (9), the significant number of asymptomatic infections, and the fatal consequences of having a positive individual in a multicat household emphasize the importance of screening as an essential intervention to eliminate sources of infectious virus in a given population.

There are several methods for the diagnosis of FeLV infection, including the detection of p27 by enzyme-linked immunosorbent assay (ELISA) (16), the detection of FeLV structural antigens in the cytoplasm of infected leukocytes and platelets by immunofluorescence antibody test (6), and virus isolation (12). Recently, FeLV has also been diagnosed by PCR, which detects proviral specific sequences in blood samples. However, there is no standardized protocol available for PCR so far and each laboratory uses its own specific sets of primers, making direct comparisons between studies problematic (8, 9, 11, 17).

Although virus isolation is considered the gold standard by some research groups, we have observed that some cats testing negative for virus isolation in plasma still can harbor DNA provirus in their genome. Moreover, levels of cell-associated FeLV mRNA, plasma RNA, and plasma p27 in these cats remain below the analytical sensitivity of our tests (M. A. Gomes-Keller and H. Lutz, unpublished data). We consider these cats to have a true latent infection which persists for long periods of time (many months), suggesting that the viral genome is replicated and segregated to the daughter cells. Whether these cats can eventually completely clear the virus is unknown. Nevertheless, it was observed that a cat could suppress FeLV replication to undetectable levels for 8.5 years postinfection, showing that the virus can persist for many years in an inactive form (1).

The use of virus isolation, immunofluorescence antibody, and ELISA as diagnostic methods has the drawback of needing blood or plasma as specimens to test. Blood collection is invasive and may be difficult in certain patients, such as very young kittens and aggressive animals, or in the case of owners reluctant to have blood collected from their asymptomatic cats. In addition, blood and plasma require special storage and processing, particularly if samples are sent by regular mail. Alternative specimen sources, specifically saliva, have shown encouraging results in clinical evaluation of experimentally infected cats (4). Using molecular techniques, such as real-time reverse transcription (RT)-PCR, we were able to show that viremic cats consistently shed viral RNA in saliva. A factor that favors the use of real-time RT-PCR for the diagnosis of FeLV in naturally infected cats using saliva as the substrate is the fact that FeLV RNA present in buccal swabs was stable for more than 2 months at room temperature.

In this study, we evaluated a rapid diagnostic assay for FeLV infection using real-time RT-PCR for the detection of viral RNA in saliva. The field utility of this test was assessed by comparing the respective results with those of p27 detection by ELISA using plasma and saliva specimens, DNA provirus measurement in whole blood by real-time PCR, and viral RNA assessment in plasma by real-time RT-PCR. Although the cost of RNA detection in saliva by RT-PCR assay is higher than that of other tests such as ELISA and virus isolation, the assay has the advantage of using samples collected noninvasively. In addition, there is no requirement for trained personnel for collecting the clinical sample, which minimizes costs. An alternative for reducing costs is the pooling of saliva samples, as only those pools containing a positive sample will require reflex testing (test is repeated on all specimens within the positive pools). Extraction of viral nucleic acids from pooled saliva samples and subsequent testing may be of importance to detect FeLV infected cats in multicat households and animal shelters. Therefore, we also evaluated the analysis of pooled samples that contained saliva from an infected cat.

MATERIALS AND METHODS

Study population.

This was a cross-sectional study done over 9 months from April 2004 to January 2005 in Switzerland. EDTA-blood and buccal swabs were obtained from each of 445 privately owned cats. Veterinarians were requested to take one EDTA-blood sample (1.2 ml) and three to four buccal swabs from cats regardless of the clinical status of the animals. In addition, cats were clinically examined by the veterinarian in charge, and a questionnaire about risk behavior, clinical status, and demographic data was completed. Moreover, blood and saliva samples collected from 30 specific-pathogen-free (SPF) cats (males, 20 to 21 weeks old), acquired from Liberty Research, Inc. (Waverly, NY) and kept in groups under barrier conditions, were used as negative controls.

Buccal swab and blood collection.

The material for blood and buccal swab collection was sent to the veterinarians accompanied by an instruction sheet that described the proper swabbing procedure. Blood was collected using the closed collection system K₃EDTA S-Monovette (Sarstedt, Nümbrecht, Germany) to avoid contamination in the veterinary office. Saliva specimens were collected without stimulation with the aid of commercially available cotton wool swabs (Primella, Migros Genossenschafts-Bund, Switzerland), similar to Q-Tips. Swabs were inserted into the cheek pouches and under the tongue of the cats. Swabs were immediately placed into 1.5-ml microcentrifuge tubes to avoid contamination, the external tip was removed, and the microcentrifuge tubes were closed. We requested at least three buccal swabs from each cat to ensure enough material in case a test would have to be repeated. Blood and buccal swab samples were submitted without refrigeration to our laboratory by mail with simple packaging. Samples were processed upon receipt.

Buccal swab and blood processing.

Buccal swabs were processed essentially as described (4). Briefly, 200 μl of Hanks' balanced salt solution (without calcium chloride, magnesium chloride, or magnesium sulfate; GIBCO, Paisley, Scotland, United Kingdom) were pipetted into the tubes and briefly vortexed. Saliva samples were incubated at 42°C for 10 min. After incubation, the microcentrifuge tubes were centrifuged at 8,000 × g for 1 min to remove drops from the inside of the lid. Swabs were inverted using a pair of tweezers, and microcentrifuge tubes were recentrifuged at 8,000 × g for 1 min. Swabs were then discarded, and the whole eluate was used for extraction of total nucleic acids. In a previous study, we could show that the signal obtained in the real-time RT-PCR for the detection of salivary RNA was much higher than that obtained by PCR (4). The mean difference between C_t values in samples extracted from buccal swabs obtained by PCR (DNA measurement) and RT-PCR (RNA detection) was of 10.72 (standard deviation = 1.44), i.e., FeLV DNA accounts for only 0.06% of the total signal obtained. Consequently, although total nucleic acids were extracted from buccal swabs, the signal obtained in the real-time RT-PCR originates mainly from viral RNA and not DNA.

An aliquot of 200 μl of whole blood was used for DNA extraction using the QIAamp DNA blood minikit (QIAGEN GmbH, Hilden, Germany), according to the manufacturer's recommendations. The elution volume was 100 μl. Approximately 1.0 ml of blood was centrifuged at room temperature at 2,300 × g for 10 min and an aliquot of 200 μl of plasma was removed for total nucleic acid extraction. The rest of the plasma was stored at −20°C and used for p27 ELISA testing. Total nucleic acids were extracted from 200 μl of buccal swab eluate and 200 μl plasma using MagNA Pure LC total nucleic acid isolation kit and a MagNA Pure LC instrument (Roche Diagnostics, Mannheim, Germany), according to the manufacturer's recommendations. The elution volume was 100 μl. Extracted samples were stored at −80°C until analysis.

Pooling of buccal swab eluates.

The pooling method consisted of combining aliquots of buccal swab eluates to make up the pooled sample. Although a dilution effect cannot be avoided in this method, the processing of the samples remained straightforward. The buccal swab eluate from a cat known to be shedding RNA in saliva was pooled with 1, 4, 9, 14, 19, 24, and 29 buccal swab eluates from FeLV-negative SPF cats to make up pool sizes of 2, 5, 10, 15, 20, 25, and 30, respectively. An equal volume of each eluate was used to make up a pool suspension of 200 μl, which were used for total nucleic acid isolation using the MagNA Pure LC total nucleic acid isolation kit and MagNA Pure LC instrument (Roche Diagnostics, Mannheim, Germany), as recommended by the manufacturer.

Detection of FeLV-specific nucleic acids by PCR and RT-PCR.

FeLV-specific DNA and RNA were detected by real-time PCR and RT-PCR (ABI 7700, Applied Biosystems, Foster City), respectively. Primers, probes, and assay conditions were identical to those previously described (18). Due to the high analytical sensitivity of the assay (detection of 1 copy/PCR) and high analytical specificity (detection of all three FeLV subtypes, and no false-positive obtained in 106 SPF cats), the detection of provirus in whole blood by real-time PCR was considered the gold standard.

ELISA for the detection of FeLV p27 in plasma and saliva.

FeLV p27 antigen was detected by a sandwich ELISA, as previously described (16). Results are represented as percentages of a positive control (FL74 feline lymphoblastoid cell culture supernatant), which was considered 100%. In this study, p27 values above 5% were considered positive. Cats positive for p27 were regarded as antigenemic.

Statistical analysis.

To determine whether a relationship between two diagnostic methods existed we performed contingency table analysis by using the software StatView (Version 5; SAS Institute Inc., Cary, NC). P values of <0.05 were considered as significant. Sensitivity, specificity, positive predictive value, negative predictive value, observed agreement (accuracy, level of agreement between two diagnostic tests), expected agreement (agreement expected by chance if the results of two different diagnostic tests are independent), prevalence, and the Cohen's kappa coefficient were calculated as described (5). Perfect agreement results in a kappa value of 1.0, and a kappa value of 0 indicates the level of agreement expected based on chance alone. The interpretation of the kappa values used in this study was based on the interpretation suggested by Landis and Koch (14). Kappa values of 0.2 or less indicate slight agreement, 0.21 to 0.40 fair, 0.41 to 0.60 moderate, 0.61 to 0.80 substantial, ≥0.81 almost perfect, and 1.00 indicates perfect agreement between tests.

RESULTS

Characteristics of the study population.

The detailed collected data about the study population are given in Table 1. Data were recorded by the veterinarian in charge by filling in a detailed questionnaire at the time of sample collection. The prevalence of FeLV was shown to be 12.1% when plasma samples were analyzed for the presence of p27. However, when the detection of provirus in whole blood was considered, we obtained an increased prevalence of 17.5%. Of note, the 78 provirus positive samples were collected from cats from 62 households. Therefore, this prevalence does not reflect the FeLV prevalence in the population. A number of cats (24 out of 445, 5.4%), which were in fact FeLV latently infected as shown by a positive result obtained for provirus detection, could not be identified by ELISA.

TABLE 1.

Study population distribution by FeLV infectious state

Parameter and value type or category	Value for group		χ2d	P valuee
	FeLV infected	FeLV uninfected
Total no. (%) of catsa	78 (17.5)	367 (82.5)
Ageb
Mean ± SD	3.17 ± 3.34	2.97 ± 3.95		0.0165f
Range	0.17-13.00	0.12-16.00
Clinical statusc
Healthy	48.7	66.8	9.042	0.0026
Sick	51.3	33.2
Sexc
Male	19.4	26.3	3.633	0.3040
Female	25.8	31.6
Neutered male	25.8	21.2
Neutered female	29.0	20.9
Outdoor accessc	88.2	70.7	8.990	0.0027
Breedc
Purebred	2.9	17.3	9.628	0.0019
Not defined	97.1	82.7
Multicat householdc
Single cat	17.4	23.5	1.215	0.2704
Two or more cats	82.6	76.5
Keepingc
Private	71.4	79.1	ND	ND
Animal shelter	11.7	7.5
Breeding facility	0.0	7.8
Farm	6.5	2.5
Foundling	7.8	2.5
Feral cat	2.6	0.0
Other	0.0	0.6
Originc
Private	62.7	45.0	ND	ND
Animal shelter	6.0	2.9
Breeding facility	0.0	11.8
Farm	13.4	8.2
Unknown	17.9	30.3
Other	0.0	1.8

^an = 445.

^bAges are in years. For all cats in the study population, the mean age was 2.99 years (standard deviation, 3.87 years) and the age range was 0.12 to 16.00 years.

^cValues are percentages of the total of the subpopulations (FeLV infected and FeLV uninfected).

^dND, not done. Statistical test could not be computed due to the insufficient number of subjects in some of the categories. The χ² value was calculated for the given parameter, though in most cases it is listed along with the first category or value type.

^eP values that were <0.05 were considered statistically significant. The P value was calculated for the given parameter, though in most cases it is listed along with the first category or value type.

^fP value was obtained by Mann-Whitney U test analysis.

The proportion of FeLV provirus-positive cats was significantly higher in the sick population (0.51) than in the healthy population (0.33). The difference in proportions was significant when chi-square analysis was performed, χ²(1, n = 445) = 9.042 and P = 0.0026 (Table 1). Similar results were obtained when ELISA data were considered (χ²[1, n = 445] = 7.944, P = 0.0048).

The cats included in this study were on average 2.99 (standard deviation = 3.87) years old (range, 0.12 to 16.0 years). FeLV-infected cats (provirus positive) were on average older (3.17 ± 3.34 years old, range, 0.17 to 13.0 years) than noninfected (provirus negative) cats (2.97 ± 3.95 years old, range, 0.12 to 16.0 years), and this difference was statistically significant (Mann-Whitney U test, P = 0.0165). The proportion of subjects falling into the four categories for sex (male, female, neutered male, and neutered female) did not differ significantly between the subpopulations of FeLV-infected and noninfected cats, χ²(3, n = 420) = 3.633 and P = 0.3040. The proportion of infected subjects which had access to outdoors was 0.88, whereas the proportion of noninfected subjects was only 0.71. The difference in proportions was significant, χ²(1, n = 402) = 8.990 and P = 0.0027. In addition, the proportion of infected subjects which were purebred was only 0.03 whereas the proportion of noninfected purebred subjects was 0.17; this difference in proportions was also significant, χ²(1, n = 434) = 9.628 and P = 0.0019. The number of subjects positive or negative for the presence of provirus as a function of the number of animals kept at the same household (animal kept alone or in a multicat household) did not differ significantly, χ²(1, n = 410) = 1.215 and P = 0.2704. Statistical analysis regarding the living conditions of the animals and their origin could not be performed due to the fact that the number of individuals in some of the categories was insufficient.

Comparison of different parameters characterizing an FeLV infection.

This study evaluated the usefulness of salivary RNA detection by RT-PCR in comparison to the detection of different parameters, characterizing an FeLV infection. The detection of provirus in whole blood by real-time PCR was considered the gold standard due to the high analytical and diagnostic sensitivity and specificity.

The average C_t values (mean ± standard deviation) obtained for the determination of RNA in saliva and plasma were 19.54 ± 4.45 and 18.18 ± 5.53, respectively (Table 2). The level (mean ± standard deviation) of p27 in plasma obtained in the ELISA test was 60.18 ± 23.55, as shown in Table 2. The ranges, 95% confidence intervals, and the 10th, 25th, 50th, 75th, and 90th percentiles for these parameters are given in Table 2. As 200 μl of whole blood was used for DNA extraction regardless of the number of cells present in the aliquot, no absolute quantitative statement can be made.

TABLE 2.

Measurement of RNA levels in plasma and saliva and p27 levels in plasma from naturally FeLV-infected cats^a

Testb	Mean	SDc	95% CId	Range	Percentilee
					10th	25th	50th	75th	90th
RNA saliva	19.54	4.45	18.33-20.74	12.02-37.67	16.07	16.66	18.45	20.11	25.45
RNA in plasma	18.18	5.53	16.69-19.68	11.58-44.00	13.95	15.16	17.13	18.50	21.83
p27 in plasma	60.18	23.55	53.75-66.61	8.03-106.81	33.87	48.69	60.68	71.79	92.97

^aResults are expressed in cycle threshold values except for those for p27 in plasma, which are expressed as a percentage of a positive control value (FL74 feline lymphoblastoid cell culture supernatant), which was considered 100%. Values below 5% were considered negative.

^bDetection of FeLV RNA sequences by real-time RT-PCR in saliva samples (n = 55); detection of FeLV RNA sequences by real-time RT-PCR in plasma samples (n = 55); and detection of p27 by ELISA in plasma samples (n = 54).

^cSD, standard deviation.

^dCI, confidence interval.

^eData represent the 10th, 25th, 50th (median), 75th, and 90th percentiles of the sample analyzed.

The diagnostic sensitivity and specificity, accuracy, positive predictive value, and negative predictive value of the different determinations on the population of 445 privately owned cats are presented in Table 3. The analysis of the results obtained with different FeLV detection methods showed that the detection of FeLV-specific RNA by RT-PCR in saliva specimens was slightly more sensitive (69.2%) than the detection of p27 in plasma specimens by ELISA (67.9%) when provirus detection by real-time PCR was used as the gold standard. The specificity, accuracy, positive predictive value, and negative predictive value did not differ between the tests at a prevalence of 17.5% (provirus detection). The detection of viral RNA in plasma by RT-PCR showed a sensitivity of 70.5% and a specificity of 100.0%. The positive and negative predictive values obtained were 100.0 and 94.1%, respectively. The detection of p27 in buccal swabs by ELISA had the lowest sensitivity (56.4%), specificity (94.4%), accuracy (87.4%), positive predictive value (69.8%), and negative predictive value (90.5%) among the tests.

TABLE 3.

Comparison of different tests for the diagnosis of FeLV, considering the detection of provirus in whole blood as the gold standard^a

Testb	Diagnostic sensitivity (%)	Diagnostic specificity (%)	Accuracy (%)	PPVc (%)	NPVd (%)
RNA saliva	69.2	99.7	94.4	98.2	93.9
p27 saliva	56.4	94.4	87.4	69.8	90.5
RNA plasma	70.5	100.0	94.8	100.0	94.1
p27 plasma	67.9	99.7	94.2	98.2	93.6

^aThese results were obtained from the analysis of 445 field cats; diagnostic specificity was evaluated in 106 SPF cats and shown to be 100%; diagnostic sensitivity was measured by testing samples from 46 experimentally FeLV infected cats, collected and processed essentially as described in the Materials and Methods section; analytical specificity was evaluated using genomic DNA prepared from FEA cells carrying molecular cloned FeLV-A, -B, or -C or uninfected FEA cells; analytical sensitivity was determined by coamplifying serial 10-fold dilutions of a plasmid containing the fragment amplified by the same primer pair of the FeLV-A/Glasgow strain genome and found to be 1 proviral copy/reaction.

^bDetection of FeLV-specific RNA in buccal swabs by real-time RT-PCR, detection of p27 in buccal swabs by ELISA (values below 5% were considered negative), detection of FeLV-specific RNA in plasma by real-time RT-PCR, detection of p27 in plasma by ELISA (values below 5% were considered negative).

^cPPV, positive predictive value.

^dNPV, negative predictive value.

When the results from the detection of viral RNA in buccal swabs by real-time RT-PCR were compared to those from p27 measurement in plasma samples by conventional ELISA, we observed an agreement of 99.1% (expected agreement = 78.3%) and a Cohen's kappa value of 0.96, indicating an almost perfect agreement between both tests (Table 4). If ELISA is considered the standard test (due to its broad use in the clinical practice), the sensitivity obtained by the detection of viral RNA by RT-PCR in saliva was 98.1%, and the specificity was 99.2%. Moreover, this diagnostic method showed a positive predictive value of 94.6% and a negative predictive value of 99.7%. In addition, viral RNA detection using saliva specimens showed an observed agreement of 99.1% (expected agreement = 78.3%) compared to its detection in plasma (kappa = 0.96, Table 4).

TABLE 4.

Comparison of the detection of salivary RNA by real-time RT-PCR with that of other tests and level of agreement

Testa	POb (%)	PEc (%)	Kappad	Interpretatione
Provirus	94.4	74.4	0.78	Substantial agreement
p27 plasma	99.1	78.3	0.96	Almost perfect agreement
RNA plasma	99.1	78.3	0.96	Almost perfect agreement
p27 saliva	92.9	75.8	0.70	Substantial agreement

^aDetection of provirus in whole blood by real-time PCR; detection of p27 in plasma by ELISA (values below 5% were considered negative); detection of FeLV-specific RNA in plasma by real-time RT-PCR; detection of p27 in buccal swabs by ELISA (values below 5% were considered negative).

^bPO, observed agreement.

^cPE, expected agreement.

^dCohen's kappa value.

^eCohen's kappa values were interpreted according to Greiner (5).

Samples from 30 control SPF cats tested negative for the presence of both RNA and DNA in blood and saliva. We also did not observe any false-positive results in ELISA using either plasma or saliva from these control cats.

Pooling of buccal swab eluates.

Despite the low invasiveness associated with the clinical specimen collection, the diagnosis of FeLV by real-time RT-PCR using saliva samples has a disadvantageous high cost in comparison to commonly used commercially available ELISA tests. This disadvantage may be set off when pooled samples are extracted and subsequently tested for the presence of FeLV RNA. Only cats of positive pools have to be retested on an individual basis in order to identify and separate them from the uninfected cats. Therefore, we evaluated the possibility of pooling buccal swab eluates before nucleic acid extraction and RT-PCR. Although a loss in assay sensitivity has to be expected due to a dilution effect, we were able to detect a single positive cat among pooled samples of 2, 5, 10, 15, 20, 25, and 30 saliva eluates (Fig. 1). In addition, pooling of buccal swab eluates for subsequent nucleic acid extraction and testing did not have any influence on the analytical specificity: testing the samples obtained by the extraction of total nucleic acids from the pool containing samples from only SPF cats (n = 30) resulted in a negative RT-PCR (results not shown).

FIG. 1.

Detection of FeLV-RNA by real-time RT-PCR assay using pooled buccal swab eluates for nucleic acid extraction. Results are expressed in cycle threshold values.

DISCUSSION

The diagnosis of FeLV is relatively straightforward, and different diagnostic methods are available, such as ELISA for the detection of p27 in blood, serum, or plasma (16), virus isolation (12), and fluorescent antibody for the detection of FeLV structural antigens in the cytoplasm of infected leukocytes and platelets (6). Recently, modern molecular methods have been described for the detection of FeLV proviral sequences using whole blood (18). Blood is required for all diagnostic methods cited. Due to the invasive nature of the specimen collection, it can become difficult to assess the FeLV state of a population. In many cases, blood sampling is not accepted by the owner or cannot be performed because of the age or clinical condition of the patient. The present study describes the possible application of a novel diagnostic assay for FeLV infection based on the detection of FeLV-specific RNA in the saliva of naturally infected cats. The use of saliva as the substrate represents an alternative for assessing the infectious state of a cat without leading to unnecessary stress to the patient. In addition, saliva can be used as test material for the diagnostics of FeLV in epidemiological studies and studies requiring repetitive sampling.

The data characterizing the present study population (n = 445) were recorded by the veterinarian in charge of the patient by filling in a detailed questionnaire at the time of blood and saliva collection. Considering the detection of provirus in whole blood by real-time PCR as the gold standard, we observed that the prevalence of FeLV was higher in the sick population compared to the healthy population. In addition, FeLV-infected cats were statistically significantly older than noninfected cats; however, the average difference was only 2.4 months and most likely not biologically significant. Furthermore, FeLV-positive cats had more frequent access to outdoors. Only a small proportion of the animals were purebred. Our results showed that 5.4% of the cats testing positive for the presence of provirus in whole blood were negative for the presence of p27 in plasma. These results confirm those of a previous study from this laboratory, showing that a number of cats test positive only for the presence of provirus in whole blood (9). It is important to mention that the 78 samples testing positive for the presence of provirus were collected from cats from 62 households. Therefore, the high prevalence observed by provirus real-time PCR analysis does not reflect the FeLV prevalence in the population of cats in Switzerland.

The present study shows that the detection of FeLV-specific RNA by RT-PCR in saliva specimens was slightly more sensitive (69.2%) than the detection of p27 in plasma specimens by ELISA (67.9%), using real-time PCR for the detection of proviral sequences as the gold standard. The specificity, accuracy, positive predictive value, and negative predictive value did not vary between both tests. The detection of viral RNA in plasma by RT-PCR showed the highest sensitivity (70.5%) and specificity (100.0%). The positive and negative predictive values obtained were 100.0 and 94.1%, respectively. In contrast, the detection of p27 in buccal swabs by ELISA had the lowest sensitivity, specificity, accuracy, positive predictive value, and negative predictive value among the tests.

However, when the detection of viral RNA in saliva by real-time RT-PCR was compared to the detection of p27 in plasma samples by conventional ELISA, we observed a high level of agreement between both tests (observed agreement = 99.1%, expected agreement = 78.3%, kappa = 0.96). Using our in-house p27-ELISA as the reference test, the diagnostic sensitivity and specificity obtained by the detection of viral RNA by RT-PCR in saliva were 98.1 and 99.2%, respectively. It has to be considered that commercially available ELISA tests are likely less sensitive and specific than the microtiter plate ELISA used in this study. In addition, this novel diagnostic method showed a positive predictive value of 94.6% and a negative predictive value of 99.7%.

Taken together, FeLV RNA detection in salivary secretions can easily replace conventional ELISA testing and can be used as an alternative tool for identifying FeLV-infected cats and fighting FeLV infection. It is important to point out that to develop a standardized test for the detection of FeLV RNA in saliva by real-time PCR, one should be aware of incorrect biosample collection. In some swabs mailed to the laboratory insufficient material could be present due to shorter or less vigorous swabbing. To avoid such failures, buccal swabs were accompanied by an instruction sheet that described the proper swabbing procedure. The inclusion of an internal control, such as 28S rRNA, proposed by Helps et al. (7) in a multiplex real-time PCR for the detection of Chlamydophila felis and feline herpesvirus would allow the evaluation of the correctness of the swabbing procedure (7). Nevertheless, due to the high sensitivity of the assay, even if only a small fraction of the expected amount of RNA is present, we would still be able to detect it, thus overcoming the problem of insufficient starting material. This was observed in our pooling experiment discussed below, which showed that the presence of only 3.35% of the expected amount of RNA present in the saliva of a positive cat would be sufficient to give a positive real-time RT-PCR result.

The detection of FeLV salivary RNA by RT-PCR had a lower diagnostic sensitivity than the detection of provirus in whole blood by real-time PCR. In total, 24 cats, despite being provirus positive, did not shed viral RNA in saliva. These animals were not antigenemic either. Such cats are considered to have a true latent infection and at the time of testing did not pose any risk to susceptible cats. Whether and when FeLV infection will be reactivated remains unknown; consequently, such cats should be treated as a potential peril.

From a clinical perspective, it is important to know if a cat is shedding FeLV RNA to develop strategies to reduce the likelihood of transmission. We showed recently that shedding of FeLV RNA in saliva does not necessarily mean shedding of infectious virus (4), but it may be taken as an indicator by the clinician that the cat is a potential source of infectious virus. From the cats harboring FeLV, not a single cat tested positive for the presence of p27 in plasma and negative for the presence of viral RNA in saliva.

Only one cat tested positive for the presence of FeLV-RNA in saliva and negative for the presence of provirus in blood. We believe that cross-contamination during sample collection might have played a role because this cat was held in an animal shelter together with FeLV-positive cats sampled on the same day. A similar result was obtained for ELISA. A single cat tested positive for the presence of p27 in plasma, but was provirus negative; this cat showed a very weak positive ELISA result of 10.2% in comparison to our positive control (considered 100%). In contrast to the first probable false-positive case, the reasons for such a result are unknown, considering that the test was repeated, and identical results were obtained. Unfortunately, no additional tissue sample could be obtained from this cat to check for the presence of a possible sequestered infection.

One single cat tested positive for the presence of RNA viral sequences in saliva, but negative in p27 ELISA using plasma. One possibility is that this cat had a high viral replication established locally, for instance in the salivary glands. This cat showed a very weak test result for the detection of FeLV RNA in plasma. Although this cat is a potential risk for susceptible animals because of the salivary shedding, it would remain undetectable if only an ELISA test had been performed.

The detection of p27 in buccal swabs by ELISA had the lowest sensitivity, specificity, accuracy, positive predictive value, and negative predictive value among the tests, confirming previous results (15). Therefore, we do not recommend it for the diagnosis of FeLV when screening individual cats. In total, 19 cats were positive in ELISA detecting p27 in saliva specimens but negative in PCR for the detection of provirus. These results indicate a too-low diagnostic specificity, resulting in a large number of false positives when saliva is used as a substrate for ELISA testing. Compared to the detection of FeLV RNA in saliva, none of those 19 cats shed viral RNA in saliva, which strongly suggests that those cats were in fact not infected. The most specific test among those compared to provirus real-time PCR was the detection of viral RNA in plasma samples, although the sensitivity was only slightly higher than that observed for the same test using saliva as the clinical sample.

Our results show that pooling of buccal swab eluates for subsequent total nucleic acid extraction did not have any effect on either the analytical or the diagnostic specificity of the RT-PCR assay. In this pooling experiment we could show that the presence of only 3.35% (one positive sample coextracted with 29 negative samples) of the expected amount of RNA present in the saliva of a positive cat would be sufficient to give a positive real-time RT-PCR result. However, a possible limitation of the pooling method is a loss of analytical and consequently diagnostic sensitivity because of the dilution effect, especially if positive subjects are shedding very small amounts of RNA in saliva. Therefore, it is important to determine the correct pool size to rule out the possibility of reducing sensitivity by diluting out individual specimens in the pool.

Nevertheless, according to our observations, in cats shedding RNA in saliva (n = 55) the mean ± standard deviation C_t value obtained in the real-time RT-PCR assay was 19.54 ± 4.45 (95% confidence interval, 18.33 to 20.74; median, 18.45; 25th and 75th percentiles, 16.66 and 20.11, respectively). Based on these values, even if larger pool sizes are chosen, a single positive cat can be detected without compromising the analytical sensitivity.

Of note, additional care must be exercised to prevent sampling or pipetting errors. In theory, false-negative results could also occur due to nonspecific inhibitors in saliva specimens. In the spiking experiment performed in this study, in which negative individual samples (1 to 29) were coextracted with a positive one, no inhibition could be observed even when a single positive buccal swab was coextracted with 29 negative ones. This method of sample pooling significantly reduces reagent costs. As a rule, the higher the prevalence, the lower the reagent cost savings, due to an increased reflex testing rate (2). Therefore, such pooling testing is ideal for the use in closed multicat households and animal shelters that have been previously screened for the presence of p27 by ELISA and where the prevalence is expected to be low.

In conclusion, the detection of FeLV-specific RNA in the saliva of infected cats represents a useful, sensitive, and specific alternative to currently used tests. The importance of using salivary secretions as a substrate for molecular diagnosis is that it circumvents problems and limitations associated with collection of blood samples. A possible drawback would be the increase in costs associated with the use of modern molecular methods for the diagnosis of the infection. Pooling of buccal swab eluates is a simple procedure that allows RT-PCR testing to be used as a screening tool in FeLV control, but also offers a method of saving costs. Only pools containing a positive specimen will require reflex testing. Pooling of saliva samples before extraction opens up an opportunity for veterinarians and breeders who are interested in controlling FeLV infections while minimizing costs. Of note, determination of the optimal pool size is of utmost importance to ensure that the sensitivity of the assay will not be affected.