Public Veterinary Medicine: Food Safety - PubAg - USDA

7 downloads 0 Views 128KB Size Report
TX 77845 (Bauer); and the USDA, Food Safety and Inspection Service, Washington, DC 20250 ..... Animal and Plant Health Inspection Service is critical for.
Public Veterinary Medicine: Food Safety Use of bovine single nucleotide polymorphism markers to verify sample tracking in beef processing Michael P. Heaton, PhD; James E. Keen, DVM, PhD; Michael L. Clawson, PhD; Gregory P. Harhay, PhD; Nathan Bauer, DVM, MS; Craig Shultz, DVM; Benedict T. Green, PhD; Lisa Durso, PhD; Carol G. Chitko-McKown, PhD; William W. Laegreid, DVM, PhD

Objective—To determine whether a selected set of 20 single nucleotide polymorphism (SNP) markers derived from beef cattle populations can be used to verify sample tracking in a commercial slaughter facility that processes primarily market (ie, culled) dairy cows. Design—Prospective, blinded validation study. Animals—165 cows and 3 bulls from 18 states (82% Holstein, 8% other dairy breeds, and 10% beef breeds). Procedure—Blood was collected by venipuncture from randomly chosen animals just prior to slaughter. The purported corresponding liver samples were collected during beef processing, and genotype profiles were obtained for each sample. Results—On the basis of SNP allele frequencies in these cattle, the mean probability that 2 randomly selected individuals would possess identical genotypes at all 20 loci was 4.3 X 10–8. Thus, the chance of a coincidental genotype match between 2 animals was 1 in 23 million. Genotype profiles confirmed appropriate matching for 152 of the 168 (90.5%) purported bloodliver sample pairs and revealed mismatching for 16 (9.5%) pairs. For the 16 mismatched sample pairs, 33% to 76% of the 20 SNP genotypes did not match (mean, 52%). Discordance that could be attributed to genotyping error was estimated to be < 1% on the basis of results for split samples. Conclusions and Clinical Relevance—Results suggest that this selected set of 20 bovine SNP markers is sufficiently informative to verify accuracy of sample tracking in slaughter plants that process beef or dairy cattle. These or similar SNP markers may facilitate high-throughput, DNA-based, traceback programs designed to detect drug residues in tissues, control of animal diseases, and enhance food safety. (J Am Vet Med Assoc 2005;226:1311–1314)

A

ccurate food animal identification is essential for improving disease control and enhancing food safety. The US Code of Federal Regulations (CFR, Title 9 Animals and Animal Products) presently requires individual animal identification and tracing of cattle with tuberculosis (§77.17) or brucellosis (§78.1), for traceback purposes. The Food Safety and Inspection Service (FSIS) also requires collection of animal identification for all cattle tested for bovine spongiform encephalopathy (BSE, FSIS notice 28-04). In beef packing plants, animals with antemortem or postmortem signs of disease are screened for antimicrobial, chemical, and drug residues that may violate food safety standards. Tissues from high-risk animals are tracked with a system of physical labels, the accuracy of which is critical for correctly identifying violations and removing condemned carcasses from the human food chain. Current research1-4 suggests that DNA-based technology represents a promising means for verifying the accuracy of physical labels for identification of cattle. After ear tags and other physical identification devices have been removed, an animal’s DNA remains as a stable, accessible, integral, and identifiable component of its beef products. Thus, DNA-based identification technologies provide a gold standard for auditing the fidelity of physical labels. Historically, the feasibility of using DNA markers for identification of individual cattle has been limited by cost, availability, and demand. However, these constraints have been diminished by 3 recent events. First, the human genome project5 has resulted in a variety of efficient genotyping technologies that appear to be capable of producing single nucleotide polymorphism (SNP) genotypes for less than $0.01 each. Single nucleotide polymorphisms are common DNA sequence variations among individuals. Second, a set of well-char-

From the USDA, Agricultural Research Service, US Meat Animal Research Center, State Spur 18D, PO Box 166, Clay Center, NE 68933-0166 (Heaton, Keen, Clawson, Harhay, Green, Durso, Chitko-McKown, Laegreid); the USDA, Food Safety and Inspection Service, College Station, TX 77845 (Bauer); and the USDA, Food Safety and Inspection Service, Washington, DC 20250 (Schultz). The use of product and company names is necessary to accurately report the methods and results; however, the USDA neither guarantees nor warrants the standard of the products, and the use of names by the USDA implies no approval of the product to the exclusion of others that may also be suitable. The authors thank Jacky Carnahan for technical assistance. Address correspondence to Dr. Heaton. JAVMA, Vol 226, No. 8, April 15, 2005

Vet Med Today: Public Veterinary Medicine

1311

acterized, highly polymorphic SNPs in popular US cattle breeds is now available for use.6 Third, on December 23, 2003, the USDA announced the identification of a case of BSE in the United States, highlighting the immediate need for a national identification system for cattle. When combined with physical, electronic, and database tools, DNA markers provide a means for verifying animal identification. The purpose of the study reported here was to determine whether a selected set of 20 SNP markers could be used to verify sample tracking in a commercial slaughter facility that processes primarily market dairy cows.

DNA extraction and genotype analysis—For blood samples, DNA was extracted by use of a solid-phase system incorporating 96-well microtitration plates, according to the manufacturer’s instructions.a For liver samples, DNA was extracted as described.7,8 The DNA samples were randomly arrayed in 96-well microtitration plates for genotype analysis. In addition, DNA from each of 48 randomly selected blood and liver samples was split and submitted in duplicate to estimate the discordance rate attributable to genotyping error. A set of 20 di-allelic bovine SNP markers spanning 12 chromosomes was used. Markers were selected on the basis of availability for scoring with a commercial systemb by means of matrix-assisted, laser desorption-ionization, time-of-flight mass spectroscopy at a commercial contract laboratoryc that used assays and procedures similar to those described previously.6 Oligonucleotides for polymerase chain reaction amplification and mass spectroscopy extension assays were derived from sequences available in GenBank, as described.6 The DNA from a multibreed beef panel was used to compare SNP allele frequencies between beef and dairy cattle. This well-characterized multibreed panel (MARC beef cattle diversity panel, version 2.1)9 consisted of 92 sires from 16 popular beef breeds and 4 sires from the Holstein dairy breed. To maximize the total number of unshared haploid genomes (187), sires within each breed were selected for pedigrees with minimal relationships between ancestors. On the basis of the number of registered progeny for each breed, the beef breeds in this panel comprised > 99% of the germplasm used in the US beef cattle industry. Mean probability of identity (PI) for an individual SNP marker was defined as the estimated probability that 2 unrelated individuals selected from the group at random would possess identical genotypes.10 Briefly, for an SNP marker (eg, locus A) with alleles A1 and A2, mean probability that 2 particular

Materials and Methods Animals—The study was conducted at a single large federally inspected commercial slaughter facility. Each week during a 4-week period in the summer of 2003, 42 animals were randomly chosen for inclusion in the study (168 total). Blood samples (9 mL) were collected from the tail vein of the selected cattle as they were driven to an enclosure for stunning. Matched liver samples (15 g) were collected during beef processing by conventional methods under which the position and sequence of each carcass was synchronized with its various beef-derived products as they moved down the processing line. Consistent with normal operational procedures of the plant, no additional physical labeling mechanisms were used in this study. The paired samples (ie, purported blood and liver samples from each animal) were labeled and shipped and stored at –20oC until use. This reference set of blood samples from 168 animals was labeled MARC-FSIS random market cattle panel version 1.1 (purified DNA samples are available upon request).

Table 1—Allele and genotype frequency of selected single nucleotide polymorphism (SNP) markers in 96 beef cattle (MARC beef cattle diversity panel version 2.1; MBCDP2.1) and 168 cattle randomly selected at the time of slaughter at a facility that primarily processes culled dairy cattle. MBCDP2.1 Allele frequencies‡ Locus CHR* identifier

SNP marker

GenBank accession

Alleles [1,2]†

Market cattle

Genotype frequencies

Allele frequencies

Genotype frequencies

1

2

1,1

1,2

2,2

P I§

1

2

1,1

1,2

2,2

PI

BTA02 BTA05 BTA05 BTA07 BTA10

3250 2019 2105 4575 5277

MBS042-1 MBS007-3 MBS030-2 MBS044-1 MBS031-1

AF458963 AF458966 AF458967 AF465156 AF465158

ggtct-[G,A]-tttgc aggag-[C,T]-cagta tcctg-[G,A]-aggac caggg-[C,T]-gatga ctgtg-[C,T]-agctg

0.59 0.67 0.66 0.70 0.77

0.41 0.33 0.34 0.30 0.23

0.36 0.52 0.49 0.47 0.59

0.45 0.29 0.32 0.47 0.35

0.19 0.19 0.18 0.06 0.05

0.369 0.391 0.382 0.443 0.481

0.67 0.83 0.69 0.84 0.83

0.33 0.17 0.31 0.16 0.17

0.46 0.70 0.46 0.71 0.70

0.41 0.27 0.46 0.26 0.26

0.13 0.03 0.08 0.03 0.04

0.398 0.561 0.427 0.568 0.555

BTA11 BTA13 BTA13 BTA17 BTA18

3254 PRNP 5033 2121 3190

MBS015-1 AH25-4 MBS046-1 MBS018-1 MBS033-1

AF465160 AF465161 AF465162 AF465164 AF465166

ccatc-[T,C]-ggaag ccagg-[T,C]-tccag ttagc-[T,C]-tttct ttaaa-[G,T]-gtcta ttctt-[T,C]-ggtcc

0.60 0.65 0.66 0.53 0.56

0.40 0.35 0.34 0.47 0.44

0.40 0.47 0.47 0.34 0.36

0.42 0.36 0.38 0.38 0.40

0.19 0.17 0.16 0.28 0.24

0.365 0.381 0.385 0.338 0.347

0.50 0.58 0.65 0.33 0.61

0.50 0.41 0.35 0.68 0.39

0.22 0.32 0.44 0.14 0.41

0.55 0.51 0.42 0.36 0.41

0.23 0.16 0.14 0.49 0.19

0.406 0.392 0.390 0.396 0.366

BTA18 BTA19 BTA19 BTA19 BTA19

3144 NOS2 NOS2 2889 5067

MBS021-1 AH13-4 AH13-1 MBS034-1 MBS054-7

AF465167 AF465168 AF465168 AF465169 AF465170

tttcc-[T,G]-attcc gagaa-[A,G]-ttggt ggagt-[C,T]-gtcac ccacc-[T,C]-cctgc ggctt-[G,A]-gccgc

0.54 0.65 0.87 0.66 0.65

0.46 0.35 0.13 0.34 0.35

0.36 0.44 0.76 0.42 0.44

0.35 0.43 0.22 0.48 0.42

0.28 0.14 0.02 0.10 0.14

0.337 0.245¶ NA 0.414 0.391

0.37 0.72 0.80 0.86 0.45

0.63 0.28 0.20 0.14 0.55

0.13 0.52 0.67 0.75 0.25

0.47 0.40 0.26 0.23 0.40

0.40 0.08 0.08 0.02 0.34

0.396 0.202¶ NA 0.614 0.345

BTA23 BTA23 BTA23 BTA24 BTA25

3463 5793 5279 3635 5915

MBS025-1 MBS039-1 MBS035-1 MBS028-1 MBS051-1

AF465172 AF465173 AF465174 AF465175 AF465177

ttgta-[C,T]-tcaac ggtgg-[A,T]-tccag atggg-[A,G]-taggg ctgca-[G,A]-tgttg ttctg-[T,C]-gggac

0.67 0.59 0.58 0.52 0.65

0.33 0.41 0.42 0.48 0.35

0.50 0.36 0.33 0.29 0.46

0.33 0.45 0.49 0.46 0.39

0.17 0.19 0.18 0.25 0.16

0.389 0.369 0.382 0.358 0.383

0.60 0.47 0.65 0.55 0.75

0.40 0.53 0.35 0.45 0.25

0.43 0.24 0.42 0.29 0.61

0.33 0.47 0.45 0.53 0.28

0.24 0.29 0.13 0.18 0.11

0.351 0.361 0.397 0.399 0.462

Average allele frequency 0.64

0.36

NA

NA

NA

NA

0.67

0.33

NA

NA

NA

NA

*Bos taurus chromosome designation. †The nucleotide alleles for the sense strand are shown in the order of their prevalence in the beef cattle (ie, allele 1 corresponds to the major allele). ‡The allele and genotype frequencies are presented as the fraction of the number of animals genotyped. Frequencies that sum to 0.99 or 1.01 are the result of rounding errors. §PI is the probability of identity and is the average probability that 2 unrelated individuals selected at random would have identical SNP genotypes. ¶Both NOS2 SNPs (AH13-1 and AH13-4, 161 nucleotides apart) were informative, and the haplotype frequencies were used to estimate the PI. NA = Not applicable.

1312

Vet Med Today: Public Veterinary Medicine

JAVMA, Vol 226, No. 8, April 15, 2005

individuals would have the same genotype was expressed as follows: PI = (χ11)2 + (χ12)2 + (χ22)2, where χ11, χ12, and χ22 were the relative genotype frequencies of A1A1, A1A2, and A2A2, respectively, in the target population. The PI for a combination of multiple SNP markers is the product of the PI for each individual marker. The underlying assumption is that the marker spacing is sufficient for meiotic recombination to cause alleles to be randomly associated with one another. The SNP genotypes were compared between samples by first creating a genotype profile of concatenated SNP genotypes for each sample. In some instances, assay failure resulted in missing genotypes for 1 or more SNP markers in an animal’s profile. In these cases, the SNP genotype was identified as “O” to track missing data in the genotype profile. Concatenated genotype profiles were compared between sample pairs with a Perl script that compared the SNP genotype at each position in profile A with the same position in profile B. This process was repeated for every possible sample pairwise combination of genotype profiles. Numbers of matches (concordant genotypes), mismatches (discordant genotypes), and incomplete SNP pairs (either or both genotypes missing at a particular site) were scored separately and summed to 20. Results were parsed into a relational database for further analysis.

Results Animals—Random selection of cattle at the slaughter facility resulted in inclusion of 165 cows and 3 bulls originating from 18 states. Of these, 138 (82%) were Holsteins, 13 (8%) were other dairy breeds, and 17 (10%) were beef breeds. Probability that randomly selected cattle would have identical SNP genotypes—The mean PI for the 20 SNP markers was 7.4 X 10–9 for beef cattle and 4.3 X 10–8 for market cattle collected at the slaughter facility (Table 1). Consequently, the chances of a coincidental genotype match between 2 animals were approximately 1 in 135 million for the beef cattle and 1 in 23 million for the cattle from which samples were collected at the slaughter facility. Accuracy of matching of purported blood-liver sample pairs—Genotype profiles of all 168 blood samples were compared with profiles for all 168 liver samples in pairwise fashion (n = 28,224 comparisons [ie, 1682]). The mean number of SNP genotypes available for comparison between 2 genotype profiles was 17.6 (ie, 20 possible SNP genotypes for each sample and a 93.8% genotype scoring rate). When all 28,224 possible pairwise comparisons of blood and liver sample genotypes were examined, the number of discordant SNP genotypes per blood-liver sample pair was normally distributed with mean, mode, and SD of 10, 10, and 2.4, respectively (Figure 1). However, when only the 168 purportedly matched blood-liver sample pairs were examined, the distribution of discordant SNP genotypes appeared bimodal. There were 152 bloodliver sample pairs with ≤ 2 mismatches, and 16 bloodliver sample pairs with ≥ 6 mismatches (mean, 9.6 [52%] mismatches; range, 6 to 14 [33% to 76%] mismatches). The discordance rate that could be attributed to genotyping error, estimated from results for the 48 split samples, was estimated to be < 1% (8 discordant genotype pairs among 843 evaluated). Possible causes of mismatching—Direct examination of genotype profiles for all samples revealed 5 samJAVMA, Vol 226, No. 8, April 15, 2005

Figure 1—Distribution of SNP genotype mismatches between sample pairs. Genotype profiles of sample pairs were aligned and analyzed as described in the Materials and Methods. A—Pair-wise comparision of genotype profiles from 168 blood samples with those from 168 liver samples (ie, 28,224 sample pairings). B—Pairwise comparison of genotype profiles from 168 blood samples with those from their purported liver samples (ie, 168 sample pairings).

ples that did not match their purported sample pair but had a substantial number of genotype matches with another sample. Four of these mismatched cases were the apparent result of a position shift in the sample collection process at the slaughter plant (ie, the genotype of the blood sample from the first animal matched the genotype for the liver sample from the next animal in line). In the fifth case, the genotype profile from the liver sample (purported to be from animal No. 60) matched the genotype profiles from both the blood and the liver sample from animal No. 73, which were collected on a different day. However, when the samples from this case were regenotyped, there were 11 discordant genotypes between samples from animals 60 and 73, demonstrating that the genotype profiles were consistent with their original purported pairs. The liver sample identification numbers for these 2 animals were similar (ie, 20359590 and 20357590), suggesting that a DNA sample label was transposed at some point in the process after collection. Discussion Results of the present report suggest that this selected set of 20 bovine SNP markers was sufficiently informative to measure the accuracy of sample tracking in a slaughter plant that processes about 90% dairy catVet Med Today: Public Veterinary Medicine

1313

tle. Moreover, 12 of the 20 SNP markers met previously published criteria6 for high informativity,6 whereby the frequency of the minor homozygous genotype was > 0.10 (minor allele frequency > 0.31; PI < 0.416). Importantly, this genotype frequency cutoff lends significant power in parentage testing situations. These 20 SNP markers have been shown to be polymorphic in diverse bovine populations and thus may facilitate DNA-based traceback programs for US beef and dairy cattle. Subsequent to the present study, the discovery of BSE in a Holstein cow at a Washington state slaughter facility in December 2003 highlighted the need for accurate and rapid traceability. The SNP markers described in the present report were successfully used to verify pedigree relationships of the Washington state BSE index case and thereby validate the accuracy of physical records derived from the case. Three of the investigators (MPH, MLC, WWL) were able to use the DNA evidence to confirm that the BSE-affected cow was of Canadian origin. The USDA’s FSIS requires that carcasses and their corresponding parts be identified “as being derived from the particular [live] animal” (9 CFR §310.2). This regulation also requires that live-animal identification devices be retained “in such a way to relate them to the carcass.” However, we are not aware of any published studies by the USDA or the US FDA evaluating the accuracy of slaughter plant identification systems. Epidemiological traceback by the US FDA and the USDA’s Animal and Plant Health Inspection Service is critical for enhancing food safety and controlling animal disease. The FDA uses live-animal identification to trace animals back to their point of origin when drug residues are found in slaughter samples. The Animal and Plant Health Inspection Service is responsible for the brucellosis and tuberculosis eradication programs (9 CFR, §78 and §77, respectively) and uses live-animal identification and slaughter samples to trace infected animals to their premise of origin. Animal and Plant Health Inspection Service regulation 9 CFR, §78.1 anticipates a large percentage of traceback failure, in part, because of misidentification at slaughter. In states with a class B or C brucellosis designation, Animal and Plant Health Inspection Service regulation requires only “80 percent of all brucellosis reactors found in the course of MCI [market cattle identification] testing be traced to the farm of origin.” For states designated as class A or brucellosis free, the regulations require that at least 90% of all brucellosis reactors be traced to the farm of origin. Although the physical labeling in the present study did not demonstrate a high-fidelity system for maintaining correct animal identification at slaughter, it was within the level of expectations. The actual success of traceback is expected to be higher than the sample matching rate at slaughter because additional information is available in a traceback situation (eg, manual lot and position recordings and carcass weights). Two important technologic factors that influence the power to discern matched samples are the genotype scoring rate and the genotype scoring error rate. For example, the genotype scoring rate in the present study was 93.8%. Thus, the probability that a particular SNP was scored in both samples was 0.938 X 0.938. In other words, only 88% of potential SNP genotype comparisons were evaluated because of missing data. On the 1314

Vet Med Today: Public Veterinary Medicine

other hand, genotyping error caused ambiguity in otherwise perfectly matched samples. Thus, when the total number of SNP markers that are used is kept low to gain economy, it is imperative that the genotype scoring rate is high (eg, > 98%) and that the genotype scoring error rate is low (eg, < 1%). One way to overcome these technical barriers is to increase the number of SNP markers used. The combined power of identification for multiple unlinked loci increases geometrically with each informative unlinked locus.6 Larger, more powerful sets of DNA markers will become available as additional SNP markers are identified for use in US beef and dairy cattle and the price of genotyping decreases. A variety of factors may have contributed to the sample misidentification rate (9.5%) observed in the present study. For example, the position of some livers in the processing line appeared to have shifted with respect to their expected position. Because samples were often collected in groups of 1 or 2, a shift in line position may have resulted in collection of samples with no other genetic matches in the data set. Of the 32 blood and liver samples that did not match their expected sample partners, 27 had no other matches in the data set. Another important factor is sample handling after collection. To generate a genotype, the DNA must typically be extracted, diluted, and transferred to reaction mixtures. Depending on the genotyping method, there may be several biochemical steps required before the genotype is assayed. Each step in this process provides an opportunity for cross-contamination or mislabeling. Although DNA testing is not infallible, it provides a powerful means of verifying the fidelity of animal identification and sample tracking in US beef and dairy cattle. a. b. c.

Gentra Systems Inc, Minneapolis, Minn. MassArray system, Sequenom Inc, San Diego, Calif. GeneSeek Inc, Lincoln, Neb.

References 1. Werner FA, Durstewitz G, Habermann FA, et al. Detection and characterization of SNPs useful for identity control and parentage testing in major European dairy breeds. Anim Genet 2004;35:44–49. 2. Vignal A, Milan D, SanCristobal M, et al. A review on SNP and other types of molecular markers and their use in animal genetics. Genet Sel Evol 2002;34:275–305. 3. Vazquez JF, Perez T, Urena F, et al. Practical application of DNA fingerprinting to trace beef. J Food Prot 2004;67:972–979. 4. Cunningham EP, Meghen CM. Biological identification systems: genetic markers. Rev Sci Tech 2001;20:491–499. 5. Kwok PY, Chen X. Detection of single nucleotide polymorphisms. Curr Issues Mol Biol 2003;5:43–60. 6. Heaton MP, Harhay GP, Bennett GL, et al. Selection and use of SNP markers for animal identification and paternity analysis in US beef cattle. Mamm Genome 2002;13:272–281. 7. Miller SA, Dykes DD, Polesky HF. A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res 1988;16:1215. 8. Heaton MP, Grosse WM, Kappes SM, et al. Estimation of DNA sequence diversity in bovine cytokine genes. Mamm Genome 2001;12:32–37. 9. Heaton MP, Chitko-McKown CG, Grosse WM, et al. Interleukin-8 haplotype structure from nucleotide sequence variation in commercial populations of US beef cattle. Mamm Genome 2001;12:219–226. 10. Holt CL, Stauffer C, Wallin JM, et al. Practical applications of genotypic surveys for forensic STR testing. Forensic Sci Int 2000; 112:91–109. JAVMA, Vol 226, No. 8, April 15, 2005