Combination of Reverse Transcription Real-Time Polymerase Chain Reaction and Antigen Capture Enzyme-Linked Immunosorbent Assay for the Detection of Animals Persistently Infected with Bovine Viral Diarrhea Virus

Virus and cell culture

Bovine viral diarrhea virus Singer (type 1) and BVDV 125c (type 2) were obtained from the National Veterinary Services Laboratories (NVSL). a The Madin–Darby bovine kidney cell line free of BVDV was maintained in minimum essential medium supplemented with 5% horse serum b in 75-cm tissue culture flasks. After inoculation of Madin–Darby bovine kidney cell line with BVDV, the cultures were incubated at 37°C and observed daily for cytopathic effect. The inoculated cultures that showed 70–80% cytopathic effect were subjected to a freeze–thaw cycle and low-speed centrifugation for 10 min at 1,000 × g. The culture supernatants were collected, divided into aliquots, and then stored at −80°C before use. The virus stocks were titrated in 96-well microtiter plates, and the titers (log 50% tissue culture infective dose [TCID₅₀]) were calculated as previously described. 21

Sample collection, preparation, and pooling

Serum samples from 899 animals and skin biopsy (ear notch) samples from 27,033 animals submitted to the Mississippi Veterinary Research and Diagnostic Laboratory (Pearl, Mississippi) between 2006 and 2008 for the diagnosis of bovine viral diarrhea were used in the present study. Approximately 2 ml of nonhemolytic serum was taken from each of the original samples for further testing. The ear notches, approximately 1.25 cm in diameter (0.19 ± 0.02 g), were placed in individual sterile tubes that contained 2 ml of 0.1 M phosphate buffered saline solution (PBS; pH 7.4), supplemented with 0.02% sodium azide as a preservative. After a brief vortexing (10 sec) and 10-min incubation at room temperature, the supernatants of the ear notch–PBS suspensions were collected for further testing. The aliquots of serum and ear-notch extracts were refrigerated at 4°C before being tested by real-time RT-PCR and ACE within 1 week. All ACE-positive as well as 50 ACE-negative ear notches were cut into 2 halves. One-half was fixed in 10% neutral formalin for IHC, whereas the other half, in PBS, was stored at −80°C for later use. To screen a herd for the presence of BVDV PI animals, the samples were pooled by mixing equal volumes (200 μl) of sera or ear notch–PBS extracts from 25 to 50 individual animals of the same herd. The pool sizes (25–50) were chosen based on 2 factors, the number of samples to be tested and the lowest possible diagnostic cost per animal. The prevalence (0.5%) data for the year 2005 generated by testing individual samples were used for cost estimation.

RNA extraction

Two methods were used to extract RNA from different samples. For serum samples, RNA was extracted from either 200 μl of nonhemolytic individual or 200 μl of pooled sample by using a commercially available RNA purification system, designated as “A.” c With ear-notch extracts, cell-cultured BVDV, viral stocks, and bacterial stocks, RNA was isolated from 500 μl of supernatant by using a commercial RNA purification kit, designated as “B.” d RNA extraction was carried out by following the manufacturers' instructions with minor modifications (2 additional washes of RNA pellets or columns). RNA was either resuspended in A or eluted with B in 50 μl of distilled DNase- and RNase-free H₂O. The RNA preparations were stored at −80°C before real-time RT-PCR analysis.

Primers and probes

Two sets of primers and probes that target BVDV 5′-UTR were derived from previous publications. 3,13 Two additional sets of primers and probes that target the same region were designed by using the Primer Express e software. All primers and probes were commercially synthesized. f Two common primers for BVDV-1 and BVDV-2, and 2 distinct probes capable of differentiating the 2 genotypes were included in the real-time RT-PCR assay. The 5′ ends of the BVDV-1 and BVDV-2 probes were labeled with fluorescent reporter molecules, 6-carboxyfluorescein (FAM) and cyanine 5 (Cy5), respectively, whereas the 3′ ends were labeled with the Black Hole Quencher dye (BHQ-1). To clone the 5′-UTR, primers BVDV-LYF 5′-TAGCCATGCCCT TAGTAGGAC-3′ and BVDV-LYR 5′-CTCCATGTGCCATGTACAGCA-3′ were used.

Real-time RT-PCR procedure

One-step real-time RT-PCR was performed by using commercial reagents according to the manufacturer's instructions. g The optimal annealing temperature for a given combination of primers and probes was determined by increasing the temperature from 55°C to 59°C in increments of 1°C. The optimal concentrations of primers and probes were determined by titration between 0.1 μmol and 1 μmol. The optimal MgCl concentration was determined by titration between 1.5 mM and 2.5 mM. The final volume of real-time RT-PCR was 25 μl, which contained 8 μl of sample RNA and 17 μl of reagent mixture with 2.5 mM MgCl₂, 10 mM deoxyribonucleotide triphosphates, 0.2 μmol of forward primer, 0.2 μmol of reverse primer, 0.24 μmol of BVDV-1 probe, 0.96 μmol of BVDV-2 probe, 5 μmol of random hexamers, 10 U of RNase inhibitor, and 1 μl of enzymes. g The reactions were carried out on a programmable thermocycler h as follows: 48°C for 25 min; 95deg;C for 10 min; 50 cycles of 95deg;C for 25 sec, 56deg;C for 25 sec, 72deg;C for 30 sec; and final elongation at 72deg;C for 10 min. To avoid potential contaminations associated with BVDV-contaminated bovine serum, equine serum was used for cell culture and propagation of viruses. Procedures for sample processing, PCR reaction setup, PCR amplification, and cell culture were performed in separate laboratory sections.

Preparation of BVDV control RNA

The 5′-UTRs of BVDV-1 and BVDV-2 were amplified by PCR and cloned in the TOPO (topoisomerase I) TA (Taq-amplified) cloning vector i downstream of the T7 promoter to yield plasmids pBVDV-I-5′ and pBVDV-II-5prime;, respectively. The 2 plasmids were linearized with restriction enzyme BamHI and purified by phenol-chloroform extraction and ethanol precipitation. The control RNA was transcribed from the T7 promoters by using a T7 RNA polymerase according to the manufacturer's instructions. j The transcribed RNA was purified by using a RNA purification kit d to remove nonincorporated nucleotides and then quantified by spectrophotometry.

Determination of real-time RT-PCR sensitivity and specificity

The analytical sensitivities were determined by performing real-time RT-PCR assays by using RNA prepared from the following: 1) 10-fold serial dilutions of BVDV-1 and BVDV-2 stocks, 2) pooled negative ear-notch samples (n = 25) spiked with serial dilutions of viral stocks, 3) 10-fold serial dilutions of in vitro transcribed BVDV-1 and BVDV-2 5′-UTR RNA, and 4) pooled negative ear samples (n = 25) spiked with serial dilutions of in vitro transcribed RNA. The input amount of viruses and RNA ranged from 0.33 to 6.33 log TCID₅₀ or 1–8 log RNA copies per reaction. Standard curves were generated by plotting threshold cycle (Ct) values in the logarithmic linear phase of amplification against the log TCID₅₀ or log RNA copies. The detection limits of real-time RT-PCR were determined as the last dilution at which all replicate reactions gave positive results.

The specificity was evaluated by performing real-time RT-PCR by using RNA extracted from 10-fold serial dilutions (10⁷–10 TCID₅₀ or colony-forming units) of various bovine pathogens, including Bovine herpesvirus 1 (NVSL), Bovine respiratory syncytial virus (NVSL), Bovine parainfluenza virus 3 (NVSL), Escherichia coli (American Type Culture Collection [ATCC] 25922), Mycobacterium paratuberculosis (ATCC 19698), Pasteurella multocida (Mississippi Veterinary Research and Diagnostic Laboratory clinical isolate), Mannheimia haemolytica (ATCC 43270), Arcanobacterium pyogenes (ATCC 19411), and Histophilus somni (ATCC 700025). The viral and bacterial concentrations ranged from 1 to 8 log TCID₅₀ and 1 to 8 log colony-forming units, respectively. The fluorescent signals of real-time RT-PCR were analyzed for each organism and compared with that of BVDV-1 and BVDV-2.

To assess the effect of sample pooling on the sensitivity of real-time RT-PCR, positive (PCR and ACE) ear-notch extracts (n = 8) were used to artificially contaminate negative ear-notch pools of different sizes. The pools were created by mixing equal volumes (100 μl) of ear-notch extracts from 1 positive sample and 11, 24, or 49 negative samples, which yielded a final pool size of 12, 25, or 50, respectively. Real-time RT-PCR assays were subsequently carried out by using RNA prepared from these pools.

Antigen capture ELISA and IHC

ACE detection of BVDV in sera and ear notches was conducted by using a commercially available test kit per manufacturer's instructions. k Immunohistochemical detection of BVDV antigen was conducted as described previously. 4,28 Briefly, the ear-notch samples were fixed in 10% neutral formalin for 2–3 weeks, embedded in paraffin, and sectioned at 5 μm. Tissue sections were treated with proteinase K, l followed by incubation with a primary monoclonal antibody against BVDV m and alkaline-phosphatase reagents n per manufacturers' instructions. The IHC staining was graded by 2 pathologists by using a 4-point scale system similar to a previously described study. 28 A consensus grade of the 2 pathologists was assigned to each sample. In brief, positive results were interpreted as 1 = faint minimal staining, 2 = mild staining, 3 = moderate staining, and 4 = marked intense staining. Negative staining was scored as 0.

Testing scheme and cost structural analysis

For cost-effectiveness, the samples were pooled and screened by real-time RT-PCR, and individual samples in positive pools were then tested by ACE. The cost for screening BVDV PI animals was predicted for herd prevalence that ranged from 0.25% to 2.0% as described previously. 20 The costs of the 3 tests, estimated based on the costs of reagents and labor as well as the actual fees of several diagnostic laboratories, were as follows: ACE, $4–$8 (average $6); IHC, $3–$7 (average $5); real-time RT-PCR, $25–$45 (average $35); and VI, $20–$40 (average $30). The cost was assumed to be the same for a given type of test performed on an individual sample or a pooled sample. The prevalence of BVDV-positive animals in a herd was assumed to be 0.25%, 0.3%, 0.5%, 0.75%, 1.00%, 1.25%, 1.50%, 1.75%, and 2.00%. The costs for individual real-time RT-PCR and ACE tests were assigned as c and d, respectively. The prevalence was represented by π. The total number of animals in a herd was designated as N, the pool size as k, and the number of pools as r, where r = N/k, with assuming r is an integer. A positive pool was expected to contain at least 1 positive sample. The probability that a pool would test positive was the binomial probability, ρ = 1 − (1 − π) ^k . Thus, the cost for pools of ρ was cr + dr[1 − (1 − π) k ]^*k , and the cost for testing each animal was c/k + d[1 − (1 − π) ^k ]. The costs for testing individual animals for different pool sizes and prevalence rates were determined. The cost reductions associated with the application of the combined real-time RT-PCR and ACE testing scheme, compared with testing individual animals by ACE, IHC, and real-time RT-PCR were assessed based on the local prevalence and sample pool size used.

Genetic analysis of the field isolates

The 5′-UTR of BVDV isolates were PCR amplified and cloned into a TA cloning vector, i and sequenced by a commercial sequencing facility. o The sequences of 22 local BVDV isolates, 23 reference strains (12 BVDV-1 and 11 BVDV-2), 4 reference Border disease virus strains, and 4 reference CSFV strains were compared. The 5′-UTR sequences were aligned by using ClustalX version 2.0. 12 Maximum parsimony and neighbor-joining analyses were performed by using PAUP^* 4.0 Beta as described previously. 30 Maximum likelihood tree estimation was evaluated by using GARLI version 0.951. 30 Tree topologies were confirmed between each of these 3 methods. Bootstrapping support for tree topologies was performed by using the neighbor-joining method implemented in PAUP^* 4.0 Beta with 1,000 replicates.

Statistical analysis

The correlations between real-time RT-PCR Ct values and the input numbers of BVDV (log TCID₅₀) and between real-time RT-PCR Ct values and the pool sizes (ln dilution factors) was determined by linear regression analysis by using Excel 2000 software. p The correlations between different BVDV diagnostic tests, including real-time RT-PCR, ACE, and IHC, were determined by using the CORR procedure of SAS 9.1 software. q The Pearson correlation coefficient (r) was used as a measure of the strength of correlation, where +1 indicated a perfect linear correlation and −1 suggested a perfect inverse (negative) correlation. The criteria for interpreting the strength of correlations were 0.00–0.39, no correlation to weak correlation; 0.40–0.70, moderate correlation; and 0.7–1.0, strong correlation. A P value less than 0.05 indicated statistical significance.

OPEN IN VIEWER

Figure 1.

Real-time reverse transcription polymerase chain reaction sensitivity and linearity for in vitro transcribed 5′-untranslated region RNA (A) and RNA-spiked ear-notch samples (B). Standard curves were obtained by plotting the threshold cycle (Ct) values versus log RNA copies per reaction. Data shown are the average of 3 independent experiments. BVDV = Bovine viral diarrhea virus.

Sensitivity and specificity of real-time RT-PCR

Of the multiple primers and probes tested, a previously published set of primers and probes (BVDV-F2/BVDV-PESTR, BVDV-P-1, and BVDV-P-2) 13 demonstrated the highest sensitivity and lowest background noise (data not shown). These primers and probes were used throughout the current study to amplify the 5′-UTR of BVDV-1 and BVDV-2. Linear correlations (R > 0.99) were found between the Ct value and the log TCID⁵⁰ or log RNA copies. The detection limits of the assay defined based on virus concentrations were 1.33 log TCID₅₀ per reaction for BVDV-1 and 2 log TCID₅₀ per reaction for BVDV-2. The detection limits on the input RNA copy numbers were 1 log copies per reaction for BVDV-1 and 2 log copies per reaction for BVDV-2 (Fig. 1A). When in vitro transcribed RNA mixed with negative ear-notch samples, the detection limits were 2 log copies per reaction for BVDV-1 and 3 log copies per reaction for BVDV-2 (Fig. 1B). When the real-time RT-PCR assays were applied on 3 commonly isolated bovine viral pathogens and 6 bacterial strains, no fluorescent signals were detected, which suggested a satisfactory specificity.

OPEN IN VIEWER

Table 1.

Strength and significance of correlations between semiquantitative results of different Bovine viral diarrhea virus tests. *

	Pearson correlation coefficients (n = 34)
Test methods	RT-PCR (Ct value)	ACE (S/P ratio)	IHC staining grade
Real-time RT-PCR (Ct value)	r = 1.000	r = −0.432	r = −0.421
	P = 0.000	P = 0.004	P = 0.013
ACE (S/P ratio)	r = −0.432	r = 1.000	r = 0.338
	P = 0.004	P = 0.000	P = 0.051

*

RT-PCR = reverse transcription polymerase chain reaction; Ct = threshold cycle; ACE = antigen capture enzyme-linked immunosorbent assay; S/P ratio = sample-to-positive ratio; IHC = immunohistochemistry. Pearson correlation coefficients (r) and the P value are indicators of the strength and significance of the correlation between 2 tests. A negative number indicates an inverse correlation between 2 tests. Results from 34 individual samples were subjected to analysis.

Correlation between testing methods

To validate the testing system, the correlations between the semiquantitative results of 3 methods were determined. Thirty-four archived ACE-positive ear notches (from positive pools) with sample-to-positive (S/P) ratios that ranged from 0.4 to 1.4 were subjected to real-time RT-PCR analyses. The realtime RT-PCR Ct values were initially between 34 and 45 for pools that contained those individual positive samples. The pool sizes were between 25 and 50. After 1-year storage at −80°C and several cycles of freeze–thaw, 4 individual samples tested negative by realtime RT-PCR. For calculation convenience, a Ct value of 50 was assigned to these 4 samples. The Ct values for the 34 archived individual samples ranged from 29 to 50. For IHC confirmation, ear notches were fixed in 10% formalin within a week of their arrival to the diagnostic laboratory and then were processed for microscopic evaluations. Results of the present study indicated that the BVDV antigen was mainly present in the epithelial cells located at the epidermis and follicles. No differences in staining pattern and intensity were observed between infections with genotype 1 and 2 viruses or between infections with a single genotype and both genotypes. Consensus scores (1–4) obtained by 2 pathologists were assigned to the positive samples. Pearson correlation coefficients (r) and the probability values (P) showed that the correlations between the semi-quantitative results of real-time RT-PCR and ACE or IHC were moderate and significant (r > 0.4, P < 0.05), whereas the correlation between ACE and IHC was weak (r = 0.338, P = 0.051; Table 1). The inverse correlation between Ct values and ACE S/P ratios or IHC grades was in accordance with the fact that Ct value is inversely related with the amount of input virus RNA.

OPEN IN VIEWER

Table 2.

Real-time reverse transcription polymerase chain reaction threshold cycle values of individual positive samples before and after dilutions with negative samples. *

Sample ID\pool size	1	2	3	4	5	6	7	8
1×	30.54 ± 0.50	34.10 ± 1.40	33.35 ± 0.37	31.96 ± 1.74	33.85 ± 1.69	32.21 ± 1.55	31.27 ± 0.54	32.77 ± 0.65
12	34.58 ± 0.75	38.60 ± 2.87	37.32 ± 0.58	37.23 ± 0.98	38.62 ± 0.37	37.57 ± 0.38	36.62 ± 0.48	39.68 ± 2.62
25	36.15 ± 1.21	41.92 ± 2.89	39.39 ± 0.95	38.87 ± 1.41	41.29 ± 0.20	38.53 ± 0.72	39.60 ± 1.14	39.98 ± 2.00
50	38.15 ± 2.02	44.64 ± 4.71	40.86 ± 1.43	39.63 ± 1.76	45.13 ± 4.23	39.12 ± 1.44	40.16 ± 1.64	44.62 ± 4.81

*

Data shown are average threshold cycle values and standard deviations of 3 independent experiments with replicate reactions in each assay.

Effect of sample pooling on real-time RT-PCR detection of BVDV

Eight PCR-positive ear notches with real-time RT-PCR Ct values that ranged from 30.54 to 34.10 were diluted with equal volumes of varying numbers of negative ear-notch samples to create pools of different sizes. Threshold cycle values for individual positive samples and the corresponding pools at sizes of 12, 25, and 50 are summarized in Table 2. The changes in Ct versus the changes in sample pool size are presented in Figure 2A. Overall, a linear correlation (R ² = 0.993) was found between the changes in Ct and Ln pool sizes (Fig. 2B). At pool sizes of 12 and 25, the average (SD) increases in Ct were 5.02 ± 0.94 and 6.49 ± 1.53 cycles, respectively. At a pool size of 50, sharp increases in Ct values as well as intertest variations for 3 of the 8 samples (nos. 2, 5, and 8) were detected (Table 2, Fig. 2A).

Prevalence of BVDV

From 2006 to 2008, a total of 27,932 animals were tested for BVDV persistent infection (Table 3). A total of 22,732 samples (ear notch or sera from different animals) were subjected to real-time RT-PCR screening of pooled samples, followed by ACE testing of individual samples in positive pools. Approximately 5,300 ear-notch samples were tested directly by ACE. Immunohistochemistry was used to confirm results when it was necessary. Eighty-five of more than 1,000 pools at sizes that ranged from 25 to 50 tested positive by real-time RT-PCR, and 26 individual ear notches tested positive by direct ACE. The majority of positive pools contained 1 positive sample that was subsequently identified by ACE. Of the 85 positive pools, 3 pools (3.53%) contained 2 positive samples and 1 pool (1.18%) contained 3 positive samples. The average prevalence of BVDV in Mississippi for a 3-year period was 0.40%, with annual fluctuations of 0.24–0.72%, which was similar to previously reported prevalence of 0.3–0.5% in the central south United States. 7

OPEN IN VIEWER

Figure 2.

Effect of sample pooling on threshold cycle (Ct) values for 8 individual ear-notch samples. Bars (A) represent the changes in Ct after mixing 1 positive sample with 11, 24, and 49 negative samples to yield pools of 12, 25, and 50. The line graph (B) indicates a linear correlation between the increases in Ct value and Ln pool sizes. Data represent the average of 3 independent experiments.

Cost-effectiveness

The cost-effectiveness of the combined real-time RT-PCR and ACE testing scheme was evaluated for different pool sizes and prevalence rates (Fig. 3). At all pool sizes, the cost per animal associated with realtime RT-PCR/ACE increased as the prevalence increased. The least-cost pool size at prevalence of 0.25–0.50% was 50 and at a prevalence of 0.75–2.0% was 25. The second-best pool sizes for different prevalence were as follows: 100, 0.25%; 25, 0.30–0.50%; and 50, 0.75–2.0%. For instance, the cost per animal at a prevalence of 0.5% with sample pool sizes of 25 and 50 were predicted to be $2.11 and $2.03, respectively. For an average local prevalence of 0.40%, the predicted cost per animal for pools of 25 was $1.97 and for pools of 50 was $1.78. However, the true prevalence for a given farm or herd was unknown, and the cost per animal might have deviated slightly from the predicted dollar amount. At an average prevalence of 0.4%, application of the real-time RT-PCR and ACE screening protocol could reduce the cost per animal at a pool size of 25 by 67%, 61%, 94%, and 93%, compared with testing individual animals by ACE, IHC, real-time RT-PCR, and VI, respectively (Table 4).

OPEN IN VIEWER

Table 3.

Prevalence of Bovine viral diarrhea virus during a 3-year period. *

Year	No. of ear notches	No. of sera	No. of total samples	Prevalence (%)
2006	8,610 (36)	26 (0)	8,636 (36)	0.42
2007	13,001 (31)	357 (1)	13,358 (32)	0.24
2008	5,422 (30)	516 (13)	5,938 (43)	0.72
Total	27,033 (97)	899 (14)	27,932 (111)	0.40

*

The total number of samples represents the number of animals being tested. Ear-notch and serum samples were derived from different individual animals. The numbers in parentheses are positive samples.

Genetic analysis of BVDV field isolates

Genotyping by real-time RT-PCR showed that 92.94% of PI animals were infected with BVDV-1, 3.53% with BVDV-2, and 3.53% with both genotypes (1 and 2). Analysis of the 5′-UTR of 22 field isolates suggested that all 18 BVDV-1 isolates belonged to the BVDV-1b subgenotype, whereas the 4 BVDV-2 isolates fell into the BVDV-2a subgenotype (Fig. 4). Genetic variations among the isolates of each subgenotype were apparent. Within the BVDV-1b subgenotype, MS7 and MS14 were clustered with the KE strain; MS2 and MS9 were apart from other known strains; MS19 was closely related to the Osloss strain; MS5, MS1, and MS3 were related to the ILLNC strains; and MS6, MS10, MS8, and MS15, along with CP 7 and PAT, constituted another cluster. The BVDV-2b field isolates formed 2 distinct clusters with MS13 and MS20 more closely related to BVDV-II-AU501, and MS4 and MS12 to BVDV-II-890.