Sensitivity of Polymerase Chain Reaction for Detection of Bovine Viral Diarrhea Virus in Pooled Serum Samples and Use of Pooled Polymerase Chain Reaction to Determine Prevalence of Bovine Viral Diarrhea Virus in Auction Market Cattle

Abstract

Two reverse transcription-nested polymerase chain reaction tests, 1 quantitative (qRT-nPCR) and 1 standard (RT-nPCR), were evaluated to assess sensitivity for detection of bovine viral diarrhea virus (BVDV) of a single positive serum sample in a pool of 30. The RT-nPCR and qRT-nPCR each detected 95 of 100 known positives. The RT-nPCR was used to estimate the prevalence of BVDV in adult beef cows. Serum samples were obtained from the US Department of Agriculture brucellosis testing laboratories in 3 Midwestern states. Samples originated from auction markets and private treaty sales throughout the 3 states. A total of 2,990 serum samples were collected and randomly pooled into 100 pools for testing. Two of the 100 pools of field samples were positive, and each positive pool had a single positive individual sample upon confirmation. The estimate of BVDV prevalence in adult cows in this study was 0.07%. This study estimates the diagnostic sensitivity of RT-nPCR for BVDV and confirms that it is a useful diagnostic tool for pools of 30 serum samples and that prevalence of BVDV in adult cattle from auction markets is low.

Bovine viral diarrhea virus (BVDV) is a common disease of U.S. cattle herds. 16 BVDV causes a range of effects, including increased disease incidence 10 and reproductive disorders. 6 The disease economically impacts producers and the livestock industry through increased treatment expenses, calf and pregnancy loss, and decreased weight gain. 11 BVDV is spread primarily through persistently infected (PI) animals. 7 Therefore, these animals are the most important target of control and biosecurity programs. Control programs based on herd testing protocols have been predicted to be cost-effective based on disease models. 17 Testing costs may be decreased by pooling samples for calves entering feed yards. Limited data indicates that pooling serum samples for polymerase chain reaction (PCR) followed by confirmation of individuals in positive pools may be more cost-effective than individual testing. 11

The advent of PCR tests for BVDV has enabled sample pooling owing to the high analytic sensitivity of PCR. While the diagnostic sensitivity of PCR on pooled sera has not been determined, studies have detected 1 positive serum sample in pools of 50 and 100 samples. 19 Using pooled supernatants from ear notch samples, 1 positive animal among 99 negatives was detected with good sensitivity and specificity. 9 A stochastic model accounting for dilution effects of pooling calculated that sensitivity would decrease by 6% to 7% for pools of 20 compared with individual testing. 13 To the authors' knowledge, no laboratory-based estimation of pooled serum PCR diagnostic sensitivity has been performed.

Knowledge of PI prevalence is also vital to determining the risk posed by importation of cattle and determining the cost-effectiveness of pooled testing programs. Prevalence of PI animals in the U.S. beef herd is known to some extent for calves, 12 yearling heifers, 1 and yearling bulls. 5 Little data are available on the prevalence in adult cows: however, 1 study found no PI animals in 1,000 non-pregnant adult cattle purchased from multiple operations. 20 The purpose of this study is to estimate the sensitivity of PCR for a single BVDV-positive serum sample in a pool of 30 serum samples and to use pooled PCR to determine the prevalence of BVDV infection in adult beef cattle sold through Midwestern auction markets and private treaty sales.

One hundred BVDV-positive serum samples were obtained from diagnostic and research samples from veterinary laboratories in New York, Kansas, and Alabama. The BVDV-negative samples were obtained from 782 beef calves that were BVDV-negative by skin biopsy immunohistochemistry, buffy coat virus isolation, and serum virus isolation. The sample size of 100 was determined to establish a 95% confidence interval (CI) for the pooled test diagnostic sensitivity with a confidence width of ±5%.

Two PCR protocols were used throughout the sensitivity estimation phase of the project, a reverse transcription nested PCR (RT-nPCR) and a quantitative RT-nPCR (qRT-nPCR). Initially all individual positive samples underwent confirmation testing by each of the 2 PCR tests in duplicate. The qRT-nPCR established the genotype (BVDV 1 or BVDV 2) of virus in the sample. For evaluation of test sensitivity in pools, pools were formed with 100 μl of a single known BVDV-positive sample and 100 μl each of 29 known BVDV-negative serum samples. Pools were tested in duplicate with both the RT-nPCR and the qRT-nPCR.

For both PCR protocols, RNA was isolated from samples using the QIAamp viral RNA mini kit a according to the manufacturer's instructions. The RT-nPCR has been previously described in detail. 3 The qRT-nPCR allowed simultaneous detection and differentiation of BVDV 1 and BVDV 2. A detailed description of quantitative PCR has previously been published. 2 In brief, for implementation in this study, in the first round of the nested reaction, a LightCycler RNA Amplification Kit HybProbe a was used. The outer primers BVD 180 and HCV 368 amplified a 213 base pair PCR product from 5 μl of RNA sample. For the second round, a LightCycler FastStart DNA Master HybProbe Kit b was used. The PCR primers for this reaction were BVDVL1 (TGCCATGTACAGCAGAGATTT) and BVDVU3 (CATGCCCAAAGCACATCTTA). The hybridization probes used in this reaction were a BVDVs1dg probe (AYGRAYACAGCCTGATAGGGTGY; Y = degenerate primers of C+T) labeled on the 3′ end with 6-carboxyfluorescein; a BVDVdg2 probe (CAGAGACCTGCTATTCCGCTAGTAAA) labeled on the 5′ end with Cy5.5 and containing a 3′ phosphate to prevent elongation; and a BVDs2 probe (CAGAGGCCCACTGTATTGCTACTAAA) labeled on the 5′ end with an Alexa Fluor 647 c and containing a 3′ phosphate to prevent elongation. 18 After completion of both rounds in the LightCycler Instrument, b data were analyzed and displayed as 670:530 nm fluorescence ratios for BVDV 1 and 705:530 fluorescence ratios for BVDV 2.

The U.S. Department of Agriculture (USDA) Brucellosis testing laboratories in 3 Midwestern states that required all test-eligible cattle presented to the auction market to be tested for Brucella abortus provided serum samples collected from adult cows at auction markets and private treaty sales. Personnel at Brucellosis testing laboratories selected and shipped the frozen serum samples to Kansas State University, where they were cataloged and stored at −80°C. Sample size was calculated to detect a BVDV prevalence of 0.1% with 95% confidence.

The RT-nPCR test was used in the field prevalence portion of the study based on performance in previous studies in the investigating laboratory. 3 Individual samples were pooled in groups of 27 to 30 and tested for BVDV using RT-nPCR, as described above. Individual samples from positive pools were tested by serum virus isolation with immunoperoxidase staining. 4 If all individual samples within a positive pool tested negative by this method, individual samples were repooled in pools of 6 or 8 (3 pools of 8 and 1 pool of 6 to account for the 30 individual samples) and retested using RT-nPCR. The 8 individual samples constituting the positive pool were then individually tested using RT-nPCR. Sensitivity and prevalence estimates, with CIs, were calculated using exact methods for categorical data with PROC FREQ in SAS; d measures of association for continuous data were calculated by a generalized linear model with PROC GENMOD in SAS. d

Of the individual diagnostic laboratory samples, 60 were BVDV 1 and 36 were BVDV 2. For the 4 remaining isolates, 2 were negative by qRT-nPCR assay and therefore not typed, and the genotype could not be determined for the other 2 because the qRT-nPCR was positive for both type 1 and type 2. The RT-nPCR detected 97/100 individual samples (sensitivity: 97%, 95% CI: 91.6–99.5%). One of the 2 negatives was BVDV type 1 and one was not typed. The qRT-nPCR detected 98 of 100 individual samples (sensitivity: 98%, 95% CI: 93.0–99.8%). Only 1 sample was not detected by either qRT-nPCR or RT-nPCR. There was no association observed between BVDV genotype and sensitivity of RT-nPCR (P = 0.06) on individual samples.

The RT-nPCR detected BVDV in 95 of 100 pools of diagnostic samples (sensitivity: 95%, 95% CI: 88.7–98.5%). Three of the negatives were also negative with individual RT-nPCR. The qRT-nPCR detected 95 of 100 pools (sensitivity: 95%, 95% CI: 88.7–98.5%). Two of the negatives were also negative with individual qRT-nPCR. Four pools were negative by both RT-nPCR and qRT-nPCR. There was no association observed between BVDV genotype and sensitivity of the RT-nPCR (P = 0.17) on pooled samples. No difference in sensitivity was detected between the individual tests, between the pooled tests, or between the individual and pooled test protocols (P > 0.05 for all associations).

A total of 2,990 serum samples were received from the 3 states. Reported cow ages ranged from 1 to 13 years. Over the 3 states 529 cows (17.7%) were reported to be 1 to 3 years of age, 1,378 cows (46%) were reported to be 4 to 7 years of age, and 604 (20.2%) were reported to be 8 years of age or greater. An additional 190 cows were reported as “short and solid-mouthed,” 169 as “broken-mouthed,” 114 as “adult,” and 2 as “gummers.” Four of the samples did not have a cow age recorded.

From state A, 997 samples were collected: 992 were from auction markets and 5 were from private treaty sales. Seventy-two counties were represented from sale dates between January 7 and March 9, 2006. Reported cow ages ranged from 1 to 11 years. Age was not reported on 4 cows, and for 2 others the age was reported as 1+. The mean of the reported cow ages was 5 years.

From state B, 1,000 samples were collected: 800 were from auction markets and 200 were from private treaty sales. Forty-seven counties were represented from sale dates between January 2 and January 31, 2006. Reported cow ages ranged from 2 to 13 years, with 475 listing only the categories of adult (114), short and solid-mouthed (190). broken-mouthed (169), or gummer (2) as ages. The mean of the reported cow ages was 5 years.

From state C, 1,000 samples were collected from auction market sales, but 7 sample tubes were broken during shipment, so analysis was performed on 993 samples. Thirty-three counties were represented. Sale dates were not reported but samples were shipped immediately after processing and received on April 6, 2006. Reported cow ages ranged from 2 to 10 years, with 60 cows reported as 1+, 1 cow reported as 2+, and 338 cows reported as >8 years. The mean of the reported exact cow ages was 5.5 years.

Seven pools did not consist of 30 samples because of sample loss in shipment and handling. Pools ranged in size from 27 to 30 samples. Two of the 100 pools were positive by RT-nPCR, 1 from state B and 1 from state C (each containing 30 samples). One individual sample in the positive pool from state C was positive by serum virus isolation. In the positive pool from state B, none of the individual samples were positive by serum virus isolation, but 1 of the samples was positive by RT-nPCR. Positive samples from both states were obtained from public auction market sales. The positive sample from state B was from a 6-year-old cow, while the positive sample from state C was from a 5-year-old cow. The estimated prevalence of BVDV in adult cows was 0.07%, with exact 95% CIs of 0.01–0.24%.

This study demonstrated a high sensitivity of 2 PCR assays (RT-nPCR and qRT-nPCR) for detection of BVDV in serum samples, both individually and in pools of 30. The 1 individual diagnostic laboratory sample that was negative on both results may have been due to primer or probe mismatches that failed to detect that sample. Alternately, that sample could have been a false positive from the submitting diagnostic laboratory. If the sample was a false positive, test sensitivity reported here may be an underestimation, although the effect would be small. The sensitivity of PCR to BVDV in pooled samples reported here relates to a single viremic sample pooled with 29 BVDV-negative samples.

The present study provides an estimate of the diagnostic sensitivity of PCR for detection of BVDV in pools of 30 serum samples. Although BVDV has been detected in larger pools, 19 the optimal pool size depends on the prevalence of disease in the population, the sensitivity of the test at different pool sizes, and the cost of the tests used. When prevalence is low, as is the case with BVDV, using larger pools becomes more cost-effective by decreasing the total number of tests required as long as detection sensitivity is maintained. If prevalence is high, a large number of pools may test positive, and the cost of individual testing on positive pools soon becomes prohibitive. Pool size decisions should therefore be made on the basis of the test sensitivity on pooled samples and the prevalence of the disease. Currently, the lowest cost pooling method may be determined more accurately by using the calculated sensitivity in pools of 30 and Monte Carlo simulations as previously published. 14

Sensitivity of PCR on pooled supernatant from tissue samples has been recently estimated, 9 but in some instances serum samples may be more available or convenient. In the present study, the authors used serum samples available in connection with the USDA Brucella eradication program. Adult cows moving between farms are often purchased and sold through auction markets. The Midwest region contains a substantial proportion of the beef cows in the U.S. beef cow herd as well as a large number of auction markets. States included in this survey required collection of blood and testing for Brucella abortus antibodies on all test-eligible cows that passed through auction markets in the state. Private treaty sales of cows are also often tested for Brucella abortus. These samples provide a useful cross-section of the adult cows available for purchase.

No previous studies have provided an estimate of prevalence of BVDV in adult beef cows in the United States. It is often assumed to be 0.1%, 8,13 but no studies have verified that assumption. This study provides an estimate of the prevalence of BVDV in adult auction market cattle in the Midwest United States. This prevalence estimate is consistent with previous assumptions of adult prevalence and of adult cattle being a low-risk source of BVDV introduction to herds, compared with calves and yearlings. 1,5,12

Due to the low prevalence of BVDV, the present study had insufficient power to detect any associations between prevalence and age. However, the ages of the 2 positive samples (5 and 6 years) indicate that some middle-aged cows from U.S. auction markets are viremic. This is the age range that was sampled most heavily overall. Not all cows in the study had exact ages reported but the categories of “broken-mouthed” and “gummer” would most likely fit in the category of 8 years of age or greater. Cows designated as “short and solid-mouthed” and “adult” might fit into either the 4- to 7-year-old group or in 8 years of age or greater.

The results of PCR testing in this study do not differentiate between persistently infected (PI) and transiently infected animals. In order to confirm PI status, an animal must be diagnosed as viremic at more than 1 time point. As PI animals carry the greatest risk of BVDV introduction for a herd, 15 such follow-up would be useful in future studies. In a study of calves entering the feedlot. PCR was performed on pools of 5 to 43 whole blood buffy coat samples. It was found to be approximately 90% specific for detection of pools containing PI animals, as opposed to transient infections, when compared with immunohistochemistry. 11 Estimation of the diagnostic specificity for PI animals of this pooled PCR on serum samples, as opposed to transient viremia, would be valuable. However, PI animals are viremic for life, whereas transient infections are viremic for only a short period and, as a result, offer fewer detectable viremic days. Therefore, any random sample testing positive for BVDV is more likely to be from a PI animal than a transiently infected animal.

The cows sampled for the prevalence estimate do not represent a random sample of all adult beef cows in the United States. They do, however, represent the population of adult beef cows available for purchase through auction markets in the Midwest. Many of the purchased cattle on cow-calf farms in the Midwest will be sourced from the public auction market and private treaty sales sampled in this survey. Because the primary interest of this study was to identify the risk of importing an adult PI cow, this was an acceptable sampling strategy. In addition, the risk for importing adult PI cows from these sales may be higher than the actual prevalence of BVDV in the adult cattle population, as PI animals tend to have low fertility and other health issues and, thus, are more likely to be culled.

In conclusion, the present study provides an estimate of the sensitivity of RT-nPCR for detection of BVDV in pools of 30 serum samples. Based on this study, prevalence of BVDV in adult beef cattle appears to be very low and as such a low risk for importation of BVD. This information should be useful in the design and implementation of biosecurity programs in U.S. beef herds.

Acknowledgements. The authors would like to thank Dr. Edward Dubovi for providing BVD-positive samples for the project. This study was funded by the United States Department of Agriculture Risk Management Agency.