doi: 10.1111/age.12131
Identifying genetic loci controlling neonatal passive transfer of immunity using a hybrid genotyping strategy G. A. Rohrer, L. A. Rempel, J. R. Miles, J. W. Keele, R. T. Wiedmann and J. L. Vallet USDA, Agricultural Research Service, U.S. Meat Animal Research Center, PO Box 166, Clay Center, NE 68933, USA.
Summary
Colostrum intake is critical to a piglet’s survival and can be measured by precipitating out the c-immunoglobulins from serum with ammonium sulfate (immunocrit). Genetic analysis of immunocrits on 5312 piglets indicated that the heritabilities (se) for direct and maternal effects were 0.13 (0.06) and 0.53 (0.08) respectively. To identify QTL for direct genetic effects, piglets with the highest and lowest immunocrits from 470 litters were selected. Six sets of DNA pools were created based on sire of the litter. These 12 DNA pools were applied to Illumina Porcine SNP60 BeadChips. Normalized X and Y values were analyzed. Three different SNP selection methods were used: deviation of the mean from high vs. low pools, the deviation adjusted for variance based on binomial theory and ANOVA. The 25 highest ranking SNPs were selected from each evaluation for further study along with 12 regions selected based on a five-SNP window approach. Selected SNPs were individually genotyped in the 988 piglets included in pools as well as in 524 piglets that had intermediate immunocrits. Association analyses were conducted fitting an animal model using the estimated genetic parameters. Nineteen SNPs were nominally associated (P < 0.01) with immunocrit values, of which nine remained significant (P < 0.05) after Bonferroni correction, located in 16 genomic regions on 13 chromosomes. In conclusion, the pooling strategy reduced the cost to scan the genome by more than 80% and identified genomic regions associated with a piglet’s ability to acquire c-immunoglobulin from colostrum. Each method to rank SNPs from the pooled analyses contributed unique validated markers, suggesting that multiple analyses will reveal more QTL than a single analysis. Keywords GWAS, immunoglobulin, pig, pooled DNA
Introduction Litter size in commercial pig production has increased along with increased piglet mortality (Johnson et al. 1999; Sorensen et al. 2000). Inadequate consumption of colostrum can lead to neonatal mortality due to infection, hypothermia, starvation or dehydration (Sangild 2003; Farmer & Quesnel 2009). There are numerous factors that can contribute to effective passive transfer of immunity. The most important factor is the quantity and quality of colostrum that is produced by the dam (Farmer & Quesnel 2009). However, the piglet also must acquire the colostrum via competitive nursing with its littermates and then
Address for correspondence G. A. Rohrer, USDA, Agricultural Research Service, U.S. Meat Animal Research Center, PO Box 166, Clay Center, NE 68933, USA. E-mail: [email protected] Accepted for publication 17 December 2013
340
absorb IgG through the lining of the small intestine (Sangild 2003). Without genetic markers for factors affecting passive transfer, improvement in piglet survival with direct selection for this important component trait will be difficult. Development of a high-density SNP genotyping system for pigs (Ramos et al. 2009) has enabled genome-wide association studies to progress rapidly. Although the number of markers per animal is quite high, the cost to collect data on 1000 or more animals also is expensive. Therefore, researchers have worked on methodologies to enable QTL discovery by genotyping pools of DNA from affected and normal/control individuals. This approach has proved successful to identify cancer susceptibility loci in humans (Huang et al. 2010b; Gaj et al. 2012) as well as a variety of other phenotypes with simple and complex inheritance (for example Pearson et al. 2007). In livestock, evaluation of pooled DNA data has successfully identified loci associated with fertility in cattle (Huang et al. 2010a; McDaneld et al. 2012). However, for all of
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity these studies, researchers selected a set of SNPs from the pooled DNA results to individually genotype in phenotyped animals to confirm the associations. Typically, a low percentage of selected markers were validated with individual genotypes, despite what statistical procedures were used to assess the data from the pooled DNA reactions. The objectives of this study were to develop a pooling strategy to identify quantitative trait loci (QTL) associated with the piglet’s ability to acquire and absorb IgG from colostrum and to determine which analysis of the pooled DNA data yielded the most validated SNP markers upon individually genotyping phenotyped piglets.
Materials and methods Data set The experimental procedures were approved and performed in accordance with the U.S. Meat Animal Research Center’s Animal Care Guidelines and the Guide for Care and Use of Agricultural Animals in Research and Teaching (FASS 2010). Piglet’s measured were from two generations (generations F7 and F8) of a ½ Landrace ¼ Duroc ¼ Yorkshire (LDY) composite population described by Holl et al. (2010). Data were available for 5312 piglets from 592 litters of 407 first and 185 second parity sows. Litters were farrowed in six discrete batches or groups (November 2009, January 2010, May 2010, July 2010, September 2010 and November 2010). The procedures used to collect immunocrit values for piglets were as described by Vallet et al. (2013). Briefly, all live piglets in each litter were bled when approximately 24 h old. Immunoglobulin was precipitated by adding 50 ll of serum to 50 ll of 40% (NH4)2SO4 and then centrifuged in a hematocrit microcapillary tube. The immunocrit value was defined as length of the immunoglobulin precipitate divided by the length of the total solution present and multiplied by 100 to create a percentage.
DNA pool construction and genotyping A sire was required to have at least 75 phenotyped progeny to be included, and litters with at least eight live
piglets at 24 h of age were sampled. A total of 480 litters sired by 30 different boars met the minimum criteria (13 sires of F7 piglets and 17 sires of F8 piglets). For pool set 1, F7 generation piglets with the highest immunocrit value from litters sired by the five boars with the most progeny were included in pool 1H, whereas the piglets with the lowest immunocrit value from these litters were put into pool 1L. Pool set 2 was constructed in a similar manner, using F8 generation piglets sired by the five boars with the most progeny. Pool set 3 included F7 generation piglets with the highest (3H) and lowest (3L) immunocrit values for the six boars with the next highest number of phenotyped progeny as well as the second highest and second lowest piglets from the two largest litters for each boar. The addition of the second highest and lowest piglets attempted to equalize the number of unique DNAs included within a pool. Pool set 4 was constructed using the same protocol as that for pool set 3, except F8 generation piglets were sampled. Pool set 5 included the piglets with the highest (5H) and lowest (5L) immunocrit values from the remaining litters, and a total of eight boars were represented (two boars sired F7 and six boars sired F8 generation piglets). For pool set 6, the piglets with the highest (6H) and lowest (6L) immunocrit values from litters sired by three boars in each generation that had a large number of progeny and a high standard deviation among his progeny were selected. No trends were evident in the sex or birth weight of piglets included in each set of pools. Descriptive statistics of immunocrit values for animals included in each DNA pool are presented in Table 1. Genomic DNA was extracted from tail tissue of each piglet using Promega Wizard SV96 Genomic prep kits based on manufacturer’s protocols. The purified DNA was quantified using a Nanodrop spectrophotometer. An aliquot of 100 ng of genomic DNA was put into the tubes from each animal assigned to the pool. Pools of DNA were mixed and stored at 4 °C until genotyped. Each pool contained DNA from 91 to 113 individual piglets. In total, 988 individual piglets were represented in at least one pool from 480 litters and 30 sires. Pooled DNAs were sent to a commercial genotyping laboratory to be applied to the Illumina Porcine SNP60 BeadChip per manufacturer’s protocols. Bead-level data were acquired.
Table 1 Descriptive statistics of immunocrit values for animals included in each DNA pool. Set
Sires (n)
Litters (n)
Piglets (n)
Low mean (SD)
Low range
High mean (SD)
High range
1 2 3 4 5 65
5 5 6 6 8 6
91 125 67 67 96 113
91 125 79 79 96 113
8.8 6.9 6.5 10.4 7.2 8.1
1.4–16.2 0.0–14.7 0.7–13.8 1.4–17.6 0.0–12.9 0.0–16.2
17.3 13.8 13.8 17.6 14.1 15.9
11.8–23.1 8.7–17.9 6.9–18.8 11.6–22.2 8.6–19.4 6.9–23.1
(4.5) (3.2) (3.3) (3.7) (3.3) (4.2)
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
(2.5) (2.2) (2.5) (2.5) (2.0) (2.9)
341
342
Rohrer et al.
Individual genotyping of SNPs for validation phase A validation panel for individual genotyping was created based on markers identified from the various ranking procedures described in statistical analyses. SNP markers selected from the pooled DNA results were individually genotyped across the 988 piglets represented in the DNA pools. In addition, 524 additional littermates with intermediate immunocrit values in litters of 10 or more piglets were genotyped along with all sires and dams of litters. The 524 additional littermates were selected with two approaches. First, approximately 11 intermediate piglets were selected for each of the 30 sires. Then, 26 litters from the entire population were selected to have all piglets genotyped to ensure that the entire phenotypic range was represented in the association analyses. Marker sequences were collated, and assay groups created using Sequenom’s ASSAY DESIGN 3.0 software (Sequenom). Multiplex groups of 25–40 SNPs were targeted. Assays were performed on the individual DNA samples and genotyped using Sequenom’s MassArray platform per manufacturer’s protocols. Genotypic data were manually evaluated, and adjustments were made when appropriate.
Average magnitude of difference adjusted for mean intensity (BINOM) Based on bionomial theory, the inherent variance is expected to be greater when the ratio of X to total intensity is 0.5 than when the ratio is closer to one of the extremes (0.0 or 1.0). Therefore, an approximate standard error was computed for each SNP based on binomial theory assuming approximately 100 beads contributed to each mean estimated. The approximate standard error was equal to the square root of the product of the average X ratio, the average Y ratio and 0.02. The coefficient 0.02 was derived by the estimate of 100 beads per mean and the equation (1/n1 + 1/n2). Each magnitude of difference in X (DELTA X) was then divided by its approximate standard error, and all 62 163 SNP markers were ranked based on their average difference adjusted for the approximate standard error. The 25 highest ranking SNP markers were included in the validation phase of the project.
Analysis of variance (ANOVA)
Three different approaches were implemented to individually rank SNP markers in order of importance, and then, a fourth method was used that incorporated all three approaches as well as averaged values for five adjacent markers. Each method relied on utilizing the normalized bead intensity values calculated in the Illumina BEADSTUDIO software (Illumina) frequently referred to as X and Y values (Staaf et al. 2008). These values were converted to a ratio of X to total intensity (X + Y) for each SNP marker and each pooled DNA sample. The Euclidian distance matrix among pools based on all SNPs was computed, and a neighborjoining analysis (Saitou & Nei 1987) was conducted to determine whether selection of individuals for pools introduced unexpected stratification among high and low pools.
The normalized data were analyzed under a general linear model, weighted by the inverse of the non-diagonal variance–covariance matrix among pools. A constant 12 9 12 symmetric matrix proportional to the variance– covariance matrix was estimated across all SNPs using the cov() function of R with SNPs treated as rows and pools treated as columns. All SNPs were used collectively to estimate similarities and dissimilarities among pools. Individual SNPs had minimal impact on this matrix, so no attempt was made to remove the dependency of the matrix on a SNP when that SNP was being tested. A scaling coefficient specific to each SNP was estimated using restricted maximum likelihood (Patterson & Thompson 1971). SNPs were tested individually using an F test with one degree of freedom for the numerator and 10 (12 pools minus two phenotypes) degrees of freedom for the denominator. All 62 163 SNP markers were then ranked based on their P-value, and the 25 most significant SNPs were included in the validation phase of the project.
Average magnitude of difference (DELTA X)
Five-SNP window based on combined rankings
The relative intensity of X for each SNP marker was averaged for the six high pools as well as for the six low pools. The difference in these two means was computed, and then, the SNP markers were ranked based on the magnitude (absolute value) of the average difference. All 62 163 SNPs with information from BEADSTUDIO were included in the ranking, regardless of whether BEADSTUDIO called genotypes for the SNPs or if the SNPs segregated in the USMARC population. The 25 highest ranking SNP markers were included in the validation phase of the project.
For all 62 163 SNP markers, the sum of all three rankings was computed. Markers were ordered relative to their chromosomal position in build 10.2 of the swine genome (Groenen et al. 2012), and the combined rankings for five adjacent SNPs were summed. Markers that were not mapped to build 10.2 were not included in this analysis. These summations were computed continuously such that a given SNP was present in five consecutive rankings and the ‘index SNP’ for each five-SNP window was the first SNP. The markers were then sorted by this
Evaluation of results from Illumina BeadChips with pooled DNA
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity composite five-SNP ranking, and 12 regions (less than 1 Mb in size) with multiple five-SNP windows present were selected for further study. The top two regions already had been discovered by all three of the previous methods (DELTA X, BINOM, and ANOVA), so markers for these two regions were the ones selected by the other methods. For the third and fourth highest ranking regions, four SNPs were selected and, for the remaining regions (rank 5 through 12), only three SNP markers were selected for the validation phase of the project. Markers in these regions were selected based on their individual rankings.
Variance component estimation and model selection The WOMBAT program (Meyer 2007) was used to conduct all mixed model analyses. An animal model was applied to the data, and pedigree information was included for parents and phenotyped animals back to the founders of this closed composite population. This analysis used 5312 immunocrit records. The initial model fit included the fixed effects of sex of piglet, farrowing group and parity of dam with covariates for number born alive, gestation length and birth weight. Sex was not a significant source of variation and was eliminated from the model. Variance components were estimated for direct additive genetic, maternal additive genetic, litter and error. A maternal permanent environmental effect could not be fit due to data structure. The estimated variance due to litter effects was nil; thus, this random effect was removed from the final model.
SNP association analyses Individual genotypic data were evaluated, and markers with call rates less than 80%, minor allele frequencies 30%) of the top 25 ranking markers either did not map or were not scored when individually genotyping pigs with the SNP60 BeadChip. From this information, we decided to focus our efforts on results across all six sets of pools. Another criterion that could be applied is if the marker was selected by more than one method. Six of the 14
markers that were on multiple lists were significant at the experiment-wise level. As four sequences did not get included into assay groups, 60% of the markers genotyped were significant. Rather than determining if a marker was important with more than one analysis, Gaj et al. (2012) averaged test statistics of two approaches (similar to the ANOVA and BINOM methods used in the present study) before they selected markers to test in the validation phase. They believed a simple comparison of means would result in too many false-positive and false-negative results, but the more sophisticated analysis may over-correct data if technical variation was low. The most robust method was the five-SNP window. However, this method used the ranking of the three singlemarker methods and required markers to be mapped to build 10.2 of the swine genome. So for a region to be included, markers had to pass a couple of screening
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
345
346
Rohrer et al. Table 3 Results from validation phase SNP association analyses for immunocrit data.
Marker 1
ALGA0001781 H3GA00553292 SIRI00006191 ALGA01071153 M1GA00060821,2,3 INRA00155361,2,5 ALGA00268692 ASGA00266412,3 ALGA00334042 ALGA01120161 ALGA01215581 ASGA00313422 MARC00533113 DIAS00005571 ASGA00331161 INRA00251941 DRGA00077945 MARC00028713 ASGA00381011 ASGA00983801 MARC00544471 MARC00279723 ASGA00406631 H3GA00271303 DIAS00009053 ASGA00474732,5 ALGA00648001,3,5 MARC00713051 ALGA00661055 ALGA00706781 ASGA00627942,3,5 ALGA01224183 ALGA00863761 ASGA00710872 ASGA00771893 ALGA00956223 H3GA00550593
Chromosome
Position
Number genotyped
Minor allele frequency
1 1 1 3 4 4 4 5 5 6 6 7 7 7 7 7 7 7 8 8 8 8 9 9 9 10 12 12 12 13 14 15 15 15 17 17
27524645 161993748 172238435 1277000004 98371966 98959246 102313213 91158735 92144681 147476183 149305769 13235493 31661556 43488854 43505936 43810375 75726206 105847910 19918384 20462147 130390886 148158336 1457532 47833978 77752795 34644896 9585899 9788471 34594122 71775121 40264909 35438261 100618897 141647107 51417426 56572871 Unknown6
1436 1464 1385 1395 1439 1466 1494 1463 1473 1420 1450 1470 1486 1462 1450 1432 1444 1471 1462 1415 1353 1477 1424 1404 1398 1492 1483 1487 1483 1480 1459 1445 1459 1510 1424 1448 1206
0.30 0.19 0.39 0.16 0.25 0.19 0.18 0.41 0.12 0.31 0.15 0.12 0.14 0.30 0.28 0.45 0.05 0.07 0.16 0.20 0.37 0.07 0.29 0.20 0.41 0.11 0.23 0.29 0.05 0.47 0.08 0.44 0.22 0.19 0.35 0.27 0.33
Regression coefficient 0.174 0.316 0.261 0.488 0.584 0.535 0.567 0.498 0.431 0.267 0.452 0.300 0.400 0.262 0.124 0.297 0.798 0.464 0.474 0.349 0.220 0.745 0.015 0.443 0.451 0.759 0.646 0.475 1.166 0.612 0.671 0.271 0.332 0.480 0.180 0.411 0.074
Standard error
P-value
0.137 0.159 0.136 0.173 0.141 0.156 0.156 0.129 0.182 0.136 0.164 0.197 0.175 0.139 0.138 0.122 0.308 0.223 0.170 0.155 0.124 0.234 0.140 0.161 0.131 0.209 0.142 0.150 0.292 0.123 0.224 0.119 0.154 0.163 0.134 0.142 0.137
0.2043 0.0471 0.0552 0.0048* 3.65E-05** 0.0006** 0.0003** 0.0001** 0.0180 0.0507 0.0059* 0.1280 0.0224 0.0596 0.3690 0.0150 0.0097* 0.0376 0.0054* 0.0245 0.0763 0.0015* 0.9147 0.0060* 0.0006** 0.0003** 5.82E-06** 0.0016* 7.04E-05** 7.42E-07** 0.0028* 0.0229 0.0313 0.0033* 0.1794 0.0039* 0.5892
1
Marker was in the 5-SNP window of the top 12 regions. Marker ranked in the top 25 of the DELTA X method. 3 Marker ranked in the top 25 of the ANOVA. 4 Estimated location in build 10.2 based on location in build 9.2. 5 Marker ranked in the top 25 of the BINOM analysis. 6 The marker sequence does not uniquely map to builds 9.2 or 10.2. *Significant association at the nominal P < 0.01 level. **Significant at the experiment-wise P < 0.05 after applying the Bonferroni correction factor. 2
procedures as well as be relatively highly ranked by the three single marker methods. Of the 12 regions tested, three were experiment-wise significant and two were nominally significant (P < 0.01). The most significant marker of the study came from the seventh ranked region. The relative ranking of the other regions discussed were 1, 2, 3, and 9. Markers from the highest two ranking regions were also identified by single marker methods. The use of an average test statistic across several adjacent markers has been used in genome-wide association studies and is commonly implemented in GENSEL (Kizilkaya et al. 2010). Pearson et al. (2007) proposed using the SNP window for pooled DNA
experiments to reduce noise as well as to minimize the number of test statistics to review when higher density (100 000 or more) genotyping platforms were used, and it was subsequently used by Janicki et al. (2011). This approach was quite useful in the current study, as three of the putative QTL would not have been discovered if this approach was not implemented. Many approaches have been initiated in attempting to identify the analysis that is most effective at discovering QTL using pooled DNA samples and high-throughput genotyping. Therefore, researchers have focused efforts on developing analyses that provide more precise estimates of
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity allele frequencies from pools of DNAs (Craig et al. 2005; Macgregor et al. 2006; Earp et al. 2011; Peiris et al. 2011). This approach seems logical but may not be necessary. If the primary focus is on the difference in allele frequency between pools, any bias in estimation of allele frequency for an assay would be expected to be similar in high and low pools and cancel each other out when evaluating differences. With the individual genotypic data, we were able to compute the allele frequencies for each pool as well as an average allele frequency for the high and low pools. The difference in allele frequencies for the 37 markers analyzed among the average of the high and low pools ranged from 0.1 to 7.1%. Surprisingly, when the association analyses of those data along with individual data from 524 intermediate phenotype piglets were completed, the correlation of differences in allele frequencies between extreme animals with significant associations was weaker than expected. Although the four markers with the greatest divergence (ranging from 5.9 to 7.1%) were significant at the experiment-wise level, none of the next four markers with allele frequency differences of 4.7 to 5.1% reached a reportable level of significance. At the other end of the spectrum, the marker with the least divergence between high and low pools (difference of 0.1%) was significant at a nominal P < 0.01. Most studies applying the pooled DNA approach utilize much higher density marker coverage. This study was conducted in a closed composite population and blocked by sires, so it leveraged the higher level of linkage disequilibrium permitting a lower density of markers to be successful. However, the amount of linkage disequilibrium in this population makes it more difficult to narrow the QTL window. Genotyping additional SNP markers across a larger group of phenotyped piglets may be necessary before markers that are predictive of immunocrit values can be developed. An evaluation of the swine genome (build 10.2) revealed an interesting candidate gene located within the SSC4 region that was detected by all four methods. The most significant SNP from the validation phase was located at 98.37 Mb, and coatomer protein complex, subunit alpha (COPA) is located at 98.08 Mb. The N-terminal 25 amino acids of the transcribed protein actually codes for a protein called Xenin, which is a gastrointestinal hormone that has been shown to regulate appetite (Alexiou et al. 1998). The third most significant region from the five-SNP window approach was located over the cholecystokinin A receptor (CCKAR) gene, which has been shown to regulate satiety (Asin & Bednarz 1992). The gene is located on SSC8 at 20.55 Mb, and a marker located at 19.92 Mb was nominally significant in the validation phase of this experiment. Variants of CCKAR previously have been associated with differences in growth rate of pigs (Houston et al. 2006). The most significant SNP of this study, located at SSC13:71.78 Mb is 1.7 Mb from the gene that codes for
ghrelin/obestatin (located at 73.48 Mb), which also regulates appetite (Wang et al. 2002); however, the associated SNP may be too far from the gene to be a viable candidate. Further refinement in position of association and annotation of the porcine genome is needed before additional candidate genes can be identified. Surprisingly, no associations were detected near two genes that have been related to passive transfer of immunity in cattle. Beta 2 microglobulin (B2M), located at SSC1:141.5 Mb in pigs, was associated with failure of passive transfer in beef cattle by Clawson et al. (2004), and the gene encoding the a-chain of FcRn (FCGRT), located at SSC6:50.3 Mb, was associated with this trait by Laegreid et al. (2002). Either these genes are not as important for passive transfer of immunity in pigs, there are no functional mutations in these genes segregating in the studied population or the SNP markers analyzed were not in linkage disequilibrium with functional variants. In summary, the pooling strategy employed was an effective approach to identifying genomic positions affecting an important novel phenotype (immunocrit value). Use of multiple analytical methods resulted in the discovery of more loci. Screening methods to improve the efficiency of the validation process should rely on marker parameters in the studied population (call rates, minor allele frequency and unique match to the genome) as well as how the marker ranked in all analyses. This study revealed that genes controlling appetite may be important factors in improving the immunocrit value in piglets and reduce pre-weaning mortality.
Acknowledgements The authors would like to acknowledge the expert technical assistance of Mr. Mike Judy for immunocrit data, Ms. Kris Simmerman for genotypic data collection and Ms. Linda Parnell for manuscript preparation. This work was funded by CRIS #5438-31000-083-00D from the Agricultural Research Service, U.S. Department of Agriculture (USDA). Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. The USDA prohibits discrimination in all its programs and activities on the basis of race, color, national origin, age, disability, and where applicable, sex, marital status, familial status, parental status, religion, sexual orientation, genetic information, political beliefs, reprisal, or because all or part of an individual’s income is derived from any public assistance program (Not all prohibited bases apply to all programs). Persons with disabilities who require alternative means for communication of program information (Braille, large print, audiotape, etc.) should contact USDA’s TARGET Center at (202) 7202600 (voice and TDD). To file a complaint of discrimination, write to USDA, Director, Office of Civil Rights, 1400
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
347
348
Rohrer et al. Independence Avenue, S.W., Washington, D.C. 202509410, or call (800) 795-3272 (voice) or (202) 720-6382 (TDD). USDA is an equal opportunity provider and employer.
Conflict of interest The authors claim no conflict of interest in the publication of this information.
References Alexiou C.H., Zimmermann J.P., Schick R.R. & Schusdziarra V. (1998) Xenin – a novel suppressor of food intake in rats. Brain Research 800, 294–9. Asin K.E. & Bednarz L. (1992) Differential effects of CCK-JMV-180 on food intake in rats and mice. Pharmacology Biochemistry and Behavior 42, 291–5. Bostrom M.A., Lu L., Chou J., Hicks P.J., Xu J., Langefeld C.D., Bowden D.W. & Freedman B.I. (2010) Candidate genes for non-diabetic ESRD in African Americans: a genome-wide association study using pooled DNA. Human Genetics 128, 195–204. Clawson M.L., Heaton M.P., Chitko-McKown C.G., Fox J.M., Smith T.P.L., Snelling W.M., Keele J.W. & Laegreid W.W. (2004) Beta2-microglobulin haplotypes in U.S. beef cattle and association with failure of passive transfer in newborn calves. Mammalian Genome 15, 227–36. Craig D.W., Huentelman M.J., Hu-Lince D., Zismann V.L., Kruer M.C., Lee A.M., Puffenberger E.G., Pearson J.M. & Stephan D.A. (2005) Identification of disease causing loci using an array-based genotyping approach on pooled DNA. BMC Genomics 6, 138. Earp M.A., Rahmani M., Chew K. & Brooks-Wilson A. (2011) Estimates of array and pool-construction variance for planning efficient DNA-pooling genome wide association studies. BMC Medical Genomics 4, 81. Farmer C. & Quesnel H. (2009) Nutritional, hormonal, and environmental effects on colostrum in sows. Journal of Animal Science 87, 56–65. FASS (2010) Guide for the Care and Use of Agricultural Animals in Research and Teaching, 3rd edn. Federation of Animal Science Societies, Champaign, IL. Gaj P., Maryan N., Hennig E.E. et al. (2012) Pooled sample-based GWAS: a cost-effective alternative for identifying colorectal and prostate cancer risk variants in the Polish population. PLoS ONE 7, e35307. Groenen M.A.M., Archibald A.L., Uenishi H. et al. (2012) Analyses of pig genomes provide insight into porcine demography and evolution. Nature 491, 393–8. Holl J.W., Rohrer G.A. & Brown-Brandl T.M. (2010) Estimates of genetic parameters among scale activity scores, growth, and fatness in pigs. Journal of Animal Science 88, 455–9. Houston R.D., Haley C.S., Archibald A.L., Cameron N.D., Plastow G.S. & Rance K.A. (2006) A polymorphism in the 5′-untranslated region of the porcine cholecystokinin type A receptor gene affects feed intake and growth. Genetics 174, 1555–63. Huang W., Kirkpatrick B.W., Rosa G.J.M. & Khatib H. (2010a) A genome-wide association study using selective DNA pooling
identifies candidate markers for fertility in Holstein cattle. Animal Genetics 41, 570–8. Huang Y., Hinds D.A., Qi L. & Prentice R.L. (2010b) Pooled versus individual genotyping in a breast cancer genome-wide association study. Genetic Epidemiology 34, 603–12. Janicki P.K., Vealey R., Liu J., Escajeda J., Postula M. & Welker K. (2011) Genome-wide association study using pooled DNA to identify candidate markers mediating susceptibility to postoperative nausea and vomiting. Anesthesiology 115, 54–64. Johnson R.K., Nielsen M.K. & Casey D.S. (1999) Responses in ovulation rate, embryonal survival, and litter traits in swine to 14 generations of selection to increase litter size. Journal of Animal Science 77, 541–57. Kizilkaya K., Fernando R.L. & Garrick D.J. (2010) Genomic prediction of simulated multibreed and purebred performance using observed fifty thousand single nucleotide polymorphism genotypes. Journal of Animal Science 88, 544–51. Laegreid W.W., Heaton M.P., Keen J.E., Grosse W.M., ChitkoMcKown C.G., Smith T.P.L., Keele J.W., Bennett G.L. & Besser T.E. (2002) Association of bovine neonatal Fc receptor a-chain gene (FCGRT) haplotypes with serum IgG concentration in newborn calves. Mammalian Genome 13, 704–10. Macgregor S., Visscher P.M. & Montgomery G. (2006) Analysis of pooled DNA samples on high density arrays without prior knowledge of differential hybridization rates. Nucleic Acids Research 34, e55. McDaneld T.G., Kuehn L.A., Thomas M.G., Snelling W.M., Sonstegard T.S., Matukumalli L.K., Smith T.P.L., Pollak E.J. & Keele J.W. (2012) Y are you not pregnant: identification of Y chromosome segments in female cattle with decreased reproductive efficiency. Journal of Animal Science 90, 2142–51. Meyer K. (2007) WOMBAT–A tool for mixed model analyses in quantitative genetics by restricted maximum likelihood (REML). Journal of Zhejiang University - Science B 8, 815–21. Patterson H.D. & Thompson R. (1971) Recovery of inter-block information when block sizes are unequal. Biometrika 58, 545–54. Pearson J.V., Huentelman M.J., Halperin R.F. et al. (2007) Identification of the genetic basis for complex disorders by use of pooling-based genomewide single-nucleotide-polymorphism association studies. The American Journal of Human Genetics 80, 126–39. Peiris B.L., Ralph J., Lamont S.J. & Dekkers J.C.M. (2011) Predicting allele frequencies in DNA pools using high density SNP genotyping data. Animal Genetics 42, 113–16. Ramos A.M., Crooijmans R.P.M.A., Affara N.A. et al. (2009) Design of a high density SNP genotyping assay in the pig using SNPs identified and characterized by next generation sequencing technology. PLoS ONE 4, e6524. Saitou N. & Nei M. (1987) The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4, 406–25. Sangild P.T. (2003) Uptake of colostral immunoglobulins by the compromised newborn farm animal. Acta Veterinaria Scandinavica 44, S105–22. Sorensen D., Vernersen A. & Andersen S. (2000) Bayesian analysis of response to selection: a case study using litter size in Danish Yorkshire pigs. Genetics 156, 283–95. Staaf J., Vallon-Christersson J., Lindgren D., Juliusson G., Rosenquist R., H€ oglund M., Borg A. & Ringner M. (2008) Normaliza-
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity tion of Illumina Infinium whole-genome SNP data improves copy number estimates and allelic intensity ratios. BMC Bioinformatics 9, 409. Vallet J.L., Miles J.R. & Rempel L.A. (2013) A simple novel measure of passive transfer of maternal immunoglobulin is predictive of preweaning mortality in piglets. The Veterinary Journal 195, 91–7. Wang G., Lee H.-M., Englander E. & Greeley G.H. Jr (2002) Ghrelin —not just another stomach hormone. Regulatory Peptides 105, 75–81.
Supporting information Additional supporting information may be found in the online version of this article. Table S1. Information on all SNP markers selected for validation phase based on results from the Illumina BeadChip analysis using pooled DNAs.
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
349
Identifying genetic loci controlling neonatal passive transfer of immunity using a hybrid genotyping strategy G. A. Rohrer, L. A. Rempel, J. R. Miles, J. W. Keele, R. T. Wiedmann and J. L. Vallet USDA, Agricultural Research Service, U.S. Meat Animal Research Center, PO Box 166, Clay Center, NE 68933, USA.
Summary
Colostrum intake is critical to a piglet’s survival and can be measured by precipitating out the c-immunoglobulins from serum with ammonium sulfate (immunocrit). Genetic analysis of immunocrits on 5312 piglets indicated that the heritabilities (se) for direct and maternal effects were 0.13 (0.06) and 0.53 (0.08) respectively. To identify QTL for direct genetic effects, piglets with the highest and lowest immunocrits from 470 litters were selected. Six sets of DNA pools were created based on sire of the litter. These 12 DNA pools were applied to Illumina Porcine SNP60 BeadChips. Normalized X and Y values were analyzed. Three different SNP selection methods were used: deviation of the mean from high vs. low pools, the deviation adjusted for variance based on binomial theory and ANOVA. The 25 highest ranking SNPs were selected from each evaluation for further study along with 12 regions selected based on a five-SNP window approach. Selected SNPs were individually genotyped in the 988 piglets included in pools as well as in 524 piglets that had intermediate immunocrits. Association analyses were conducted fitting an animal model using the estimated genetic parameters. Nineteen SNPs were nominally associated (P < 0.01) with immunocrit values, of which nine remained significant (P < 0.05) after Bonferroni correction, located in 16 genomic regions on 13 chromosomes. In conclusion, the pooling strategy reduced the cost to scan the genome by more than 80% and identified genomic regions associated with a piglet’s ability to acquire c-immunoglobulin from colostrum. Each method to rank SNPs from the pooled analyses contributed unique validated markers, suggesting that multiple analyses will reveal more QTL than a single analysis. Keywords GWAS, immunoglobulin, pig, pooled DNA
Introduction Litter size in commercial pig production has increased along with increased piglet mortality (Johnson et al. 1999; Sorensen et al. 2000). Inadequate consumption of colostrum can lead to neonatal mortality due to infection, hypothermia, starvation or dehydration (Sangild 2003; Farmer & Quesnel 2009). There are numerous factors that can contribute to effective passive transfer of immunity. The most important factor is the quantity and quality of colostrum that is produced by the dam (Farmer & Quesnel 2009). However, the piglet also must acquire the colostrum via competitive nursing with its littermates and then
Address for correspondence G. A. Rohrer, USDA, Agricultural Research Service, U.S. Meat Animal Research Center, PO Box 166, Clay Center, NE 68933, USA. E-mail: [email protected] Accepted for publication 17 December 2013
340
absorb IgG through the lining of the small intestine (Sangild 2003). Without genetic markers for factors affecting passive transfer, improvement in piglet survival with direct selection for this important component trait will be difficult. Development of a high-density SNP genotyping system for pigs (Ramos et al. 2009) has enabled genome-wide association studies to progress rapidly. Although the number of markers per animal is quite high, the cost to collect data on 1000 or more animals also is expensive. Therefore, researchers have worked on methodologies to enable QTL discovery by genotyping pools of DNA from affected and normal/control individuals. This approach has proved successful to identify cancer susceptibility loci in humans (Huang et al. 2010b; Gaj et al. 2012) as well as a variety of other phenotypes with simple and complex inheritance (for example Pearson et al. 2007). In livestock, evaluation of pooled DNA data has successfully identified loci associated with fertility in cattle (Huang et al. 2010a; McDaneld et al. 2012). However, for all of
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity these studies, researchers selected a set of SNPs from the pooled DNA results to individually genotype in phenotyped animals to confirm the associations. Typically, a low percentage of selected markers were validated with individual genotypes, despite what statistical procedures were used to assess the data from the pooled DNA reactions. The objectives of this study were to develop a pooling strategy to identify quantitative trait loci (QTL) associated with the piglet’s ability to acquire and absorb IgG from colostrum and to determine which analysis of the pooled DNA data yielded the most validated SNP markers upon individually genotyping phenotyped piglets.
Materials and methods Data set The experimental procedures were approved and performed in accordance with the U.S. Meat Animal Research Center’s Animal Care Guidelines and the Guide for Care and Use of Agricultural Animals in Research and Teaching (FASS 2010). Piglet’s measured were from two generations (generations F7 and F8) of a ½ Landrace ¼ Duroc ¼ Yorkshire (LDY) composite population described by Holl et al. (2010). Data were available for 5312 piglets from 592 litters of 407 first and 185 second parity sows. Litters were farrowed in six discrete batches or groups (November 2009, January 2010, May 2010, July 2010, September 2010 and November 2010). The procedures used to collect immunocrit values for piglets were as described by Vallet et al. (2013). Briefly, all live piglets in each litter were bled when approximately 24 h old. Immunoglobulin was precipitated by adding 50 ll of serum to 50 ll of 40% (NH4)2SO4 and then centrifuged in a hematocrit microcapillary tube. The immunocrit value was defined as length of the immunoglobulin precipitate divided by the length of the total solution present and multiplied by 100 to create a percentage.
DNA pool construction and genotyping A sire was required to have at least 75 phenotyped progeny to be included, and litters with at least eight live
piglets at 24 h of age were sampled. A total of 480 litters sired by 30 different boars met the minimum criteria (13 sires of F7 piglets and 17 sires of F8 piglets). For pool set 1, F7 generation piglets with the highest immunocrit value from litters sired by the five boars with the most progeny were included in pool 1H, whereas the piglets with the lowest immunocrit value from these litters were put into pool 1L. Pool set 2 was constructed in a similar manner, using F8 generation piglets sired by the five boars with the most progeny. Pool set 3 included F7 generation piglets with the highest (3H) and lowest (3L) immunocrit values for the six boars with the next highest number of phenotyped progeny as well as the second highest and second lowest piglets from the two largest litters for each boar. The addition of the second highest and lowest piglets attempted to equalize the number of unique DNAs included within a pool. Pool set 4 was constructed using the same protocol as that for pool set 3, except F8 generation piglets were sampled. Pool set 5 included the piglets with the highest (5H) and lowest (5L) immunocrit values from the remaining litters, and a total of eight boars were represented (two boars sired F7 and six boars sired F8 generation piglets). For pool set 6, the piglets with the highest (6H) and lowest (6L) immunocrit values from litters sired by three boars in each generation that had a large number of progeny and a high standard deviation among his progeny were selected. No trends were evident in the sex or birth weight of piglets included in each set of pools. Descriptive statistics of immunocrit values for animals included in each DNA pool are presented in Table 1. Genomic DNA was extracted from tail tissue of each piglet using Promega Wizard SV96 Genomic prep kits based on manufacturer’s protocols. The purified DNA was quantified using a Nanodrop spectrophotometer. An aliquot of 100 ng of genomic DNA was put into the tubes from each animal assigned to the pool. Pools of DNA were mixed and stored at 4 °C until genotyped. Each pool contained DNA from 91 to 113 individual piglets. In total, 988 individual piglets were represented in at least one pool from 480 litters and 30 sires. Pooled DNAs were sent to a commercial genotyping laboratory to be applied to the Illumina Porcine SNP60 BeadChip per manufacturer’s protocols. Bead-level data were acquired.
Table 1 Descriptive statistics of immunocrit values for animals included in each DNA pool. Set
Sires (n)
Litters (n)
Piglets (n)
Low mean (SD)
Low range
High mean (SD)
High range
1 2 3 4 5 65
5 5 6 6 8 6
91 125 67 67 96 113
91 125 79 79 96 113
8.8 6.9 6.5 10.4 7.2 8.1
1.4–16.2 0.0–14.7 0.7–13.8 1.4–17.6 0.0–12.9 0.0–16.2
17.3 13.8 13.8 17.6 14.1 15.9
11.8–23.1 8.7–17.9 6.9–18.8 11.6–22.2 8.6–19.4 6.9–23.1
(4.5) (3.2) (3.3) (3.7) (3.3) (4.2)
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
(2.5) (2.2) (2.5) (2.5) (2.0) (2.9)
341
342
Rohrer et al.
Individual genotyping of SNPs for validation phase A validation panel for individual genotyping was created based on markers identified from the various ranking procedures described in statistical analyses. SNP markers selected from the pooled DNA results were individually genotyped across the 988 piglets represented in the DNA pools. In addition, 524 additional littermates with intermediate immunocrit values in litters of 10 or more piglets were genotyped along with all sires and dams of litters. The 524 additional littermates were selected with two approaches. First, approximately 11 intermediate piglets were selected for each of the 30 sires. Then, 26 litters from the entire population were selected to have all piglets genotyped to ensure that the entire phenotypic range was represented in the association analyses. Marker sequences were collated, and assay groups created using Sequenom’s ASSAY DESIGN 3.0 software (Sequenom). Multiplex groups of 25–40 SNPs were targeted. Assays were performed on the individual DNA samples and genotyped using Sequenom’s MassArray platform per manufacturer’s protocols. Genotypic data were manually evaluated, and adjustments were made when appropriate.
Average magnitude of difference adjusted for mean intensity (BINOM) Based on bionomial theory, the inherent variance is expected to be greater when the ratio of X to total intensity is 0.5 than when the ratio is closer to one of the extremes (0.0 or 1.0). Therefore, an approximate standard error was computed for each SNP based on binomial theory assuming approximately 100 beads contributed to each mean estimated. The approximate standard error was equal to the square root of the product of the average X ratio, the average Y ratio and 0.02. The coefficient 0.02 was derived by the estimate of 100 beads per mean and the equation (1/n1 + 1/n2). Each magnitude of difference in X (DELTA X) was then divided by its approximate standard error, and all 62 163 SNP markers were ranked based on their average difference adjusted for the approximate standard error. The 25 highest ranking SNP markers were included in the validation phase of the project.
Analysis of variance (ANOVA)
Three different approaches were implemented to individually rank SNP markers in order of importance, and then, a fourth method was used that incorporated all three approaches as well as averaged values for five adjacent markers. Each method relied on utilizing the normalized bead intensity values calculated in the Illumina BEADSTUDIO software (Illumina) frequently referred to as X and Y values (Staaf et al. 2008). These values were converted to a ratio of X to total intensity (X + Y) for each SNP marker and each pooled DNA sample. The Euclidian distance matrix among pools based on all SNPs was computed, and a neighborjoining analysis (Saitou & Nei 1987) was conducted to determine whether selection of individuals for pools introduced unexpected stratification among high and low pools.
The normalized data were analyzed under a general linear model, weighted by the inverse of the non-diagonal variance–covariance matrix among pools. A constant 12 9 12 symmetric matrix proportional to the variance– covariance matrix was estimated across all SNPs using the cov() function of R with SNPs treated as rows and pools treated as columns. All SNPs were used collectively to estimate similarities and dissimilarities among pools. Individual SNPs had minimal impact on this matrix, so no attempt was made to remove the dependency of the matrix on a SNP when that SNP was being tested. A scaling coefficient specific to each SNP was estimated using restricted maximum likelihood (Patterson & Thompson 1971). SNPs were tested individually using an F test with one degree of freedom for the numerator and 10 (12 pools minus two phenotypes) degrees of freedom for the denominator. All 62 163 SNP markers were then ranked based on their P-value, and the 25 most significant SNPs were included in the validation phase of the project.
Average magnitude of difference (DELTA X)
Five-SNP window based on combined rankings
The relative intensity of X for each SNP marker was averaged for the six high pools as well as for the six low pools. The difference in these two means was computed, and then, the SNP markers were ranked based on the magnitude (absolute value) of the average difference. All 62 163 SNPs with information from BEADSTUDIO were included in the ranking, regardless of whether BEADSTUDIO called genotypes for the SNPs or if the SNPs segregated in the USMARC population. The 25 highest ranking SNP markers were included in the validation phase of the project.
For all 62 163 SNP markers, the sum of all three rankings was computed. Markers were ordered relative to their chromosomal position in build 10.2 of the swine genome (Groenen et al. 2012), and the combined rankings for five adjacent SNPs were summed. Markers that were not mapped to build 10.2 were not included in this analysis. These summations were computed continuously such that a given SNP was present in five consecutive rankings and the ‘index SNP’ for each five-SNP window was the first SNP. The markers were then sorted by this
Evaluation of results from Illumina BeadChips with pooled DNA
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity composite five-SNP ranking, and 12 regions (less than 1 Mb in size) with multiple five-SNP windows present were selected for further study. The top two regions already had been discovered by all three of the previous methods (DELTA X, BINOM, and ANOVA), so markers for these two regions were the ones selected by the other methods. For the third and fourth highest ranking regions, four SNPs were selected and, for the remaining regions (rank 5 through 12), only three SNP markers were selected for the validation phase of the project. Markers in these regions were selected based on their individual rankings.
Variance component estimation and model selection The WOMBAT program (Meyer 2007) was used to conduct all mixed model analyses. An animal model was applied to the data, and pedigree information was included for parents and phenotyped animals back to the founders of this closed composite population. This analysis used 5312 immunocrit records. The initial model fit included the fixed effects of sex of piglet, farrowing group and parity of dam with covariates for number born alive, gestation length and birth weight. Sex was not a significant source of variation and was eliminated from the model. Variance components were estimated for direct additive genetic, maternal additive genetic, litter and error. A maternal permanent environmental effect could not be fit due to data structure. The estimated variance due to litter effects was nil; thus, this random effect was removed from the final model.
SNP association analyses Individual genotypic data were evaluated, and markers with call rates less than 80%, minor allele frequencies 30%) of the top 25 ranking markers either did not map or were not scored when individually genotyping pigs with the SNP60 BeadChip. From this information, we decided to focus our efforts on results across all six sets of pools. Another criterion that could be applied is if the marker was selected by more than one method. Six of the 14
markers that were on multiple lists were significant at the experiment-wise level. As four sequences did not get included into assay groups, 60% of the markers genotyped were significant. Rather than determining if a marker was important with more than one analysis, Gaj et al. (2012) averaged test statistics of two approaches (similar to the ANOVA and BINOM methods used in the present study) before they selected markers to test in the validation phase. They believed a simple comparison of means would result in too many false-positive and false-negative results, but the more sophisticated analysis may over-correct data if technical variation was low. The most robust method was the five-SNP window. However, this method used the ranking of the three singlemarker methods and required markers to be mapped to build 10.2 of the swine genome. So for a region to be included, markers had to pass a couple of screening
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
345
346
Rohrer et al. Table 3 Results from validation phase SNP association analyses for immunocrit data.
Marker 1
ALGA0001781 H3GA00553292 SIRI00006191 ALGA01071153 M1GA00060821,2,3 INRA00155361,2,5 ALGA00268692 ASGA00266412,3 ALGA00334042 ALGA01120161 ALGA01215581 ASGA00313422 MARC00533113 DIAS00005571 ASGA00331161 INRA00251941 DRGA00077945 MARC00028713 ASGA00381011 ASGA00983801 MARC00544471 MARC00279723 ASGA00406631 H3GA00271303 DIAS00009053 ASGA00474732,5 ALGA00648001,3,5 MARC00713051 ALGA00661055 ALGA00706781 ASGA00627942,3,5 ALGA01224183 ALGA00863761 ASGA00710872 ASGA00771893 ALGA00956223 H3GA00550593
Chromosome
Position
Number genotyped
Minor allele frequency
1 1 1 3 4 4 4 5 5 6 6 7 7 7 7 7 7 7 8 8 8 8 9 9 9 10 12 12 12 13 14 15 15 15 17 17
27524645 161993748 172238435 1277000004 98371966 98959246 102313213 91158735 92144681 147476183 149305769 13235493 31661556 43488854 43505936 43810375 75726206 105847910 19918384 20462147 130390886 148158336 1457532 47833978 77752795 34644896 9585899 9788471 34594122 71775121 40264909 35438261 100618897 141647107 51417426 56572871 Unknown6
1436 1464 1385 1395 1439 1466 1494 1463 1473 1420 1450 1470 1486 1462 1450 1432 1444 1471 1462 1415 1353 1477 1424 1404 1398 1492 1483 1487 1483 1480 1459 1445 1459 1510 1424 1448 1206
0.30 0.19 0.39 0.16 0.25 0.19 0.18 0.41 0.12 0.31 0.15 0.12 0.14 0.30 0.28 0.45 0.05 0.07 0.16 0.20 0.37 0.07 0.29 0.20 0.41 0.11 0.23 0.29 0.05 0.47 0.08 0.44 0.22 0.19 0.35 0.27 0.33
Regression coefficient 0.174 0.316 0.261 0.488 0.584 0.535 0.567 0.498 0.431 0.267 0.452 0.300 0.400 0.262 0.124 0.297 0.798 0.464 0.474 0.349 0.220 0.745 0.015 0.443 0.451 0.759 0.646 0.475 1.166 0.612 0.671 0.271 0.332 0.480 0.180 0.411 0.074
Standard error
P-value
0.137 0.159 0.136 0.173 0.141 0.156 0.156 0.129 0.182 0.136 0.164 0.197 0.175 0.139 0.138 0.122 0.308 0.223 0.170 0.155 0.124 0.234 0.140 0.161 0.131 0.209 0.142 0.150 0.292 0.123 0.224 0.119 0.154 0.163 0.134 0.142 0.137
0.2043 0.0471 0.0552 0.0048* 3.65E-05** 0.0006** 0.0003** 0.0001** 0.0180 0.0507 0.0059* 0.1280 0.0224 0.0596 0.3690 0.0150 0.0097* 0.0376 0.0054* 0.0245 0.0763 0.0015* 0.9147 0.0060* 0.0006** 0.0003** 5.82E-06** 0.0016* 7.04E-05** 7.42E-07** 0.0028* 0.0229 0.0313 0.0033* 0.1794 0.0039* 0.5892
1
Marker was in the 5-SNP window of the top 12 regions. Marker ranked in the top 25 of the DELTA X method. 3 Marker ranked in the top 25 of the ANOVA. 4 Estimated location in build 10.2 based on location in build 9.2. 5 Marker ranked in the top 25 of the BINOM analysis. 6 The marker sequence does not uniquely map to builds 9.2 or 10.2. *Significant association at the nominal P < 0.01 level. **Significant at the experiment-wise P < 0.05 after applying the Bonferroni correction factor. 2
procedures as well as be relatively highly ranked by the three single marker methods. Of the 12 regions tested, three were experiment-wise significant and two were nominally significant (P < 0.01). The most significant marker of the study came from the seventh ranked region. The relative ranking of the other regions discussed were 1, 2, 3, and 9. Markers from the highest two ranking regions were also identified by single marker methods. The use of an average test statistic across several adjacent markers has been used in genome-wide association studies and is commonly implemented in GENSEL (Kizilkaya et al. 2010). Pearson et al. (2007) proposed using the SNP window for pooled DNA
experiments to reduce noise as well as to minimize the number of test statistics to review when higher density (100 000 or more) genotyping platforms were used, and it was subsequently used by Janicki et al. (2011). This approach was quite useful in the current study, as three of the putative QTL would not have been discovered if this approach was not implemented. Many approaches have been initiated in attempting to identify the analysis that is most effective at discovering QTL using pooled DNA samples and high-throughput genotyping. Therefore, researchers have focused efforts on developing analyses that provide more precise estimates of
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity allele frequencies from pools of DNAs (Craig et al. 2005; Macgregor et al. 2006; Earp et al. 2011; Peiris et al. 2011). This approach seems logical but may not be necessary. If the primary focus is on the difference in allele frequency between pools, any bias in estimation of allele frequency for an assay would be expected to be similar in high and low pools and cancel each other out when evaluating differences. With the individual genotypic data, we were able to compute the allele frequencies for each pool as well as an average allele frequency for the high and low pools. The difference in allele frequencies for the 37 markers analyzed among the average of the high and low pools ranged from 0.1 to 7.1%. Surprisingly, when the association analyses of those data along with individual data from 524 intermediate phenotype piglets were completed, the correlation of differences in allele frequencies between extreme animals with significant associations was weaker than expected. Although the four markers with the greatest divergence (ranging from 5.9 to 7.1%) were significant at the experiment-wise level, none of the next four markers with allele frequency differences of 4.7 to 5.1% reached a reportable level of significance. At the other end of the spectrum, the marker with the least divergence between high and low pools (difference of 0.1%) was significant at a nominal P < 0.01. Most studies applying the pooled DNA approach utilize much higher density marker coverage. This study was conducted in a closed composite population and blocked by sires, so it leveraged the higher level of linkage disequilibrium permitting a lower density of markers to be successful. However, the amount of linkage disequilibrium in this population makes it more difficult to narrow the QTL window. Genotyping additional SNP markers across a larger group of phenotyped piglets may be necessary before markers that are predictive of immunocrit values can be developed. An evaluation of the swine genome (build 10.2) revealed an interesting candidate gene located within the SSC4 region that was detected by all four methods. The most significant SNP from the validation phase was located at 98.37 Mb, and coatomer protein complex, subunit alpha (COPA) is located at 98.08 Mb. The N-terminal 25 amino acids of the transcribed protein actually codes for a protein called Xenin, which is a gastrointestinal hormone that has been shown to regulate appetite (Alexiou et al. 1998). The third most significant region from the five-SNP window approach was located over the cholecystokinin A receptor (CCKAR) gene, which has been shown to regulate satiety (Asin & Bednarz 1992). The gene is located on SSC8 at 20.55 Mb, and a marker located at 19.92 Mb was nominally significant in the validation phase of this experiment. Variants of CCKAR previously have been associated with differences in growth rate of pigs (Houston et al. 2006). The most significant SNP of this study, located at SSC13:71.78 Mb is 1.7 Mb from the gene that codes for
ghrelin/obestatin (located at 73.48 Mb), which also regulates appetite (Wang et al. 2002); however, the associated SNP may be too far from the gene to be a viable candidate. Further refinement in position of association and annotation of the porcine genome is needed before additional candidate genes can be identified. Surprisingly, no associations were detected near two genes that have been related to passive transfer of immunity in cattle. Beta 2 microglobulin (B2M), located at SSC1:141.5 Mb in pigs, was associated with failure of passive transfer in beef cattle by Clawson et al. (2004), and the gene encoding the a-chain of FcRn (FCGRT), located at SSC6:50.3 Mb, was associated with this trait by Laegreid et al. (2002). Either these genes are not as important for passive transfer of immunity in pigs, there are no functional mutations in these genes segregating in the studied population or the SNP markers analyzed were not in linkage disequilibrium with functional variants. In summary, the pooling strategy employed was an effective approach to identifying genomic positions affecting an important novel phenotype (immunocrit value). Use of multiple analytical methods resulted in the discovery of more loci. Screening methods to improve the efficiency of the validation process should rely on marker parameters in the studied population (call rates, minor allele frequency and unique match to the genome) as well as how the marker ranked in all analyses. This study revealed that genes controlling appetite may be important factors in improving the immunocrit value in piglets and reduce pre-weaning mortality.
Acknowledgements The authors would like to acknowledge the expert technical assistance of Mr. Mike Judy for immunocrit data, Ms. Kris Simmerman for genotypic data collection and Ms. Linda Parnell for manuscript preparation. This work was funded by CRIS #5438-31000-083-00D from the Agricultural Research Service, U.S. Department of Agriculture (USDA). Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. The USDA prohibits discrimination in all its programs and activities on the basis of race, color, national origin, age, disability, and where applicable, sex, marital status, familial status, parental status, religion, sexual orientation, genetic information, political beliefs, reprisal, or because all or part of an individual’s income is derived from any public assistance program (Not all prohibited bases apply to all programs). Persons with disabilities who require alternative means for communication of program information (Braille, large print, audiotape, etc.) should contact USDA’s TARGET Center at (202) 7202600 (voice and TDD). To file a complaint of discrimination, write to USDA, Director, Office of Civil Rights, 1400
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
347
348
Rohrer et al. Independence Avenue, S.W., Washington, D.C. 202509410, or call (800) 795-3272 (voice) or (202) 720-6382 (TDD). USDA is an equal opportunity provider and employer.
Conflict of interest The authors claim no conflict of interest in the publication of this information.
References Alexiou C.H., Zimmermann J.P., Schick R.R. & Schusdziarra V. (1998) Xenin – a novel suppressor of food intake in rats. Brain Research 800, 294–9. Asin K.E. & Bednarz L. (1992) Differential effects of CCK-JMV-180 on food intake in rats and mice. Pharmacology Biochemistry and Behavior 42, 291–5. Bostrom M.A., Lu L., Chou J., Hicks P.J., Xu J., Langefeld C.D., Bowden D.W. & Freedman B.I. (2010) Candidate genes for non-diabetic ESRD in African Americans: a genome-wide association study using pooled DNA. Human Genetics 128, 195–204. Clawson M.L., Heaton M.P., Chitko-McKown C.G., Fox J.M., Smith T.P.L., Snelling W.M., Keele J.W. & Laegreid W.W. (2004) Beta2-microglobulin haplotypes in U.S. beef cattle and association with failure of passive transfer in newborn calves. Mammalian Genome 15, 227–36. Craig D.W., Huentelman M.J., Hu-Lince D., Zismann V.L., Kruer M.C., Lee A.M., Puffenberger E.G., Pearson J.M. & Stephan D.A. (2005) Identification of disease causing loci using an array-based genotyping approach on pooled DNA. BMC Genomics 6, 138. Earp M.A., Rahmani M., Chew K. & Brooks-Wilson A. (2011) Estimates of array and pool-construction variance for planning efficient DNA-pooling genome wide association studies. BMC Medical Genomics 4, 81. Farmer C. & Quesnel H. (2009) Nutritional, hormonal, and environmental effects on colostrum in sows. Journal of Animal Science 87, 56–65. FASS (2010) Guide for the Care and Use of Agricultural Animals in Research and Teaching, 3rd edn. Federation of Animal Science Societies, Champaign, IL. Gaj P., Maryan N., Hennig E.E. et al. (2012) Pooled sample-based GWAS: a cost-effective alternative for identifying colorectal and prostate cancer risk variants in the Polish population. PLoS ONE 7, e35307. Groenen M.A.M., Archibald A.L., Uenishi H. et al. (2012) Analyses of pig genomes provide insight into porcine demography and evolution. Nature 491, 393–8. Holl J.W., Rohrer G.A. & Brown-Brandl T.M. (2010) Estimates of genetic parameters among scale activity scores, growth, and fatness in pigs. Journal of Animal Science 88, 455–9. Houston R.D., Haley C.S., Archibald A.L., Cameron N.D., Plastow G.S. & Rance K.A. (2006) A polymorphism in the 5′-untranslated region of the porcine cholecystokinin type A receptor gene affects feed intake and growth. Genetics 174, 1555–63. Huang W., Kirkpatrick B.W., Rosa G.J.M. & Khatib H. (2010a) A genome-wide association study using selective DNA pooling
identifies candidate markers for fertility in Holstein cattle. Animal Genetics 41, 570–8. Huang Y., Hinds D.A., Qi L. & Prentice R.L. (2010b) Pooled versus individual genotyping in a breast cancer genome-wide association study. Genetic Epidemiology 34, 603–12. Janicki P.K., Vealey R., Liu J., Escajeda J., Postula M. & Welker K. (2011) Genome-wide association study using pooled DNA to identify candidate markers mediating susceptibility to postoperative nausea and vomiting. Anesthesiology 115, 54–64. Johnson R.K., Nielsen M.K. & Casey D.S. (1999) Responses in ovulation rate, embryonal survival, and litter traits in swine to 14 generations of selection to increase litter size. Journal of Animal Science 77, 541–57. Kizilkaya K., Fernando R.L. & Garrick D.J. (2010) Genomic prediction of simulated multibreed and purebred performance using observed fifty thousand single nucleotide polymorphism genotypes. Journal of Animal Science 88, 544–51. Laegreid W.W., Heaton M.P., Keen J.E., Grosse W.M., ChitkoMcKown C.G., Smith T.P.L., Keele J.W., Bennett G.L. & Besser T.E. (2002) Association of bovine neonatal Fc receptor a-chain gene (FCGRT) haplotypes with serum IgG concentration in newborn calves. Mammalian Genome 13, 704–10. Macgregor S., Visscher P.M. & Montgomery G. (2006) Analysis of pooled DNA samples on high density arrays without prior knowledge of differential hybridization rates. Nucleic Acids Research 34, e55. McDaneld T.G., Kuehn L.A., Thomas M.G., Snelling W.M., Sonstegard T.S., Matukumalli L.K., Smith T.P.L., Pollak E.J. & Keele J.W. (2012) Y are you not pregnant: identification of Y chromosome segments in female cattle with decreased reproductive efficiency. Journal of Animal Science 90, 2142–51. Meyer K. (2007) WOMBAT–A tool for mixed model analyses in quantitative genetics by restricted maximum likelihood (REML). Journal of Zhejiang University - Science B 8, 815–21. Patterson H.D. & Thompson R. (1971) Recovery of inter-block information when block sizes are unequal. Biometrika 58, 545–54. Pearson J.V., Huentelman M.J., Halperin R.F. et al. (2007) Identification of the genetic basis for complex disorders by use of pooling-based genomewide single-nucleotide-polymorphism association studies. The American Journal of Human Genetics 80, 126–39. Peiris B.L., Ralph J., Lamont S.J. & Dekkers J.C.M. (2011) Predicting allele frequencies in DNA pools using high density SNP genotyping data. Animal Genetics 42, 113–16. Ramos A.M., Crooijmans R.P.M.A., Affara N.A. et al. (2009) Design of a high density SNP genotyping assay in the pig using SNPs identified and characterized by next generation sequencing technology. PLoS ONE 4, e6524. Saitou N. & Nei M. (1987) The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4, 406–25. Sangild P.T. (2003) Uptake of colostral immunoglobulins by the compromised newborn farm animal. Acta Veterinaria Scandinavica 44, S105–22. Sorensen D., Vernersen A. & Andersen S. (2000) Bayesian analysis of response to selection: a case study using litter size in Danish Yorkshire pigs. Genetics 156, 283–95. Staaf J., Vallon-Christersson J., Lindgren D., Juliusson G., Rosenquist R., H€ oglund M., Borg A. & Ringner M. (2008) Normaliza-
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
Neonatal passive transfer of immunity tion of Illumina Infinium whole-genome SNP data improves copy number estimates and allelic intensity ratios. BMC Bioinformatics 9, 409. Vallet J.L., Miles J.R. & Rempel L.A. (2013) A simple novel measure of passive transfer of maternal immunoglobulin is predictive of preweaning mortality in piglets. The Veterinary Journal 195, 91–7. Wang G., Lee H.-M., Englander E. & Greeley G.H. Jr (2002) Ghrelin —not just another stomach hormone. Regulatory Peptides 105, 75–81.
Supporting information Additional supporting information may be found in the online version of this article. Table S1. Information on all SNP markers selected for validation phase based on results from the Illumina BeadChip analysis using pooled DNAs.
Published 2014. This article is a U.S. Government work and is in the public domain in the USA., 45, 340–349
349
