Large-Sample Genome-Wide Association Study of Resistance to Retained Placenta in U.S. Holstein Cows

A genome-wide association study of resistance to retained placenta (RETP) using 632,212 Holstein cows and 74,747 SNPs identified 200 additive effects with p-values < 10−8 on thirteen chromosomes but no dominance effect was statistically significant. The regions of 87.61–88.74 Mb of Chr09 about 1.13 Mb in size had the most significant effect in LOC112448080 and other highly significant effects in CCDC170 and ESR1, and in or near RMND1 and AKAP12. Four non-ESR1 genes in this region were reported to be involved in ESR1 fusions in humans. Chr23 had the largest number of significant effects that peaked in SLC17A1, which was involved in urate metabolism and transport that could contribute to kidney disease. The PKHD1 gene contained seven significant effects and was downstream of another six significant effects. The ACOT13 gene also had a highly significant effect. Both PKHD1 and ACOT13 were associated with kidney disease. Another highly significant effect was upstream of BOLA-DQA2. The KITLG gene of Chr05 that acts in utero in germ cell and neural cell development, and hematopoiesis was upstream of a highly significant effect, contained a significant effect, and was between another two significant effects. The results of this study provided a new understanding of genetic factors underlying RETP in U.S. Holstein cows.


Introduction
Retained placenta in dairy cattle refers to the failure of timely separation of the placenta from the dam after calving.This disease creates a number of problems in health and fertility including inflammation, fever, decreased milk yield, longer calving intervals, higher incidence of metritis and lower conception rate [1].In U.S. Holstein cows, retained placenta has an incidence rate of 3.6%, and is a low-heritability trait with 1% heritability [2].In humans, retained placenta was found to have inherited risk where the mother's having retained placenta increased the risk of this disease in the next generation [3].Only limited research was available on the association between retained placenta and genetic variants.A genome-wide association study (GWAS) using U.S. Holstein bulls and single nucleotide polymorphism (SNP) markers included retained placenta but found no significant effects [4].Another study on retained placenta using Canadian Holstein bulls identified several chromosome regions with higher heritability estimates than in other chromosome regions [5].Starting in 2018, national genetic and genomic evaluations for resistance to retained placenta (RETP) started for U.S. Holstein cattle [6] and have since accumulated a large sample of Holstein cows with RETP phenotypic observations and genotypes of genome-wide SNPs, providing a unique opportunity with unprecedented statistical power for identifying genetic variants associated with RETP.Using this large sample, this study aimed to identify genetic variants and chromosome regions affecting RETP in U.S. Holstein cows using a GWAS approach.

Overview of GWAS Results
The GWAS of RETP using 614,035 first-lactation Holstein cows and 74,747 SNPs identified 200 additive effects and no dominance effects with log 10 (1/p) > 8 (Figure 1a, Table S1).The 200 significant additive effects were distributed on twelve chromosomes (Figure 1a) : 5, 6, 7, 9, 10, 15, 17, 19, 24, 25, 29 and 31, where Chr31 is the X-Y nonrecombining region of the X chromosome.Allelic effects of the 200 SNPs (Figure 1b) showed that negative allelic effects had larger effect sizes than the positive effect sizes for most chromosomes.The average of the negative allelic effects was −0.089 and the average of the positive allelic effects was 0.077.Chr05 had the most positive allelic effect in C5H12orf75, and the X-Y nonrecombining region of the X chromosome (Chr31) had the most negative allelic effect between MAGED1 and MAGED4B.The detailed descriptions of candidate genes of the significant effects will use gene symbols for most genes and the full gene names are provided in Table S2.
Int. J. Mol.Sci.2024, 25, x FOR PEER REVIEW 2 of 12 sample, this study aimed to identify genetic variants and chromosome regions affecting RETP in U.S. Holstein cows using a GWAS approach.

Overview of GWAS Results
The GWAS of RETP using 614,035 first-lactation Holstein cows and 74,747 SNPs identified 200 additive effects and no dominance effects with log10(1/p) > 8 (Figure 1a, Table S1).The 200 significant additive effects were distributed on twelve chromosomes (Figure 1a) : 5, 6, 7, 9, 10, 15, 17, 19, 24, 25, 29 and 31, where Chr31 is the X-Y nonrecombining region of the X chromosome.Allelic effects of the 200 SNPs (Figure 1b) showed that negative allelic effects had larger effect sizes than the positive effect sizes for most chromosomes.The average of the negative allelic effects was −0.089 and the average of the positive allelic effects was 0.077.Chr05 had the most positive allelic effect in C5H12orf75, and the X-Y nonrecombining region of the X chromosome (Chr31) had the most negative allelic effect between MAGED1 and MAGED4B.The detailed descriptions of candidate genes of the significant effects will use gene symbols for most genes and the full gene names are provided in Table S2.

Additive Effects of Chr09
Chr09 had twenty-three significant effects including the #1 effect in LOC112448080, #2 upstream of AKAP12, #3 in CCDC170, and #5 in RMND1 (Tables 1 and S1, Figure 2a).These effects were in the 87.61-88.73Mb region about 1.12 Mb in size.An interesting aspect of this region was the estrogen receptor 1 (ESR1) gene fusions observed in humans.This region had PLEKHG1, MTHFD1L, AKAP12, RMND1, and CCDC170 genes upstream of ESR1 (Table 1), and four of these five upstream genes were involved in ESR1 fusions associated with human breast cancer, ESR1-CCDC170, ESR1-AKAP12, ESR1-MTHFD1, and ESR1-PLEKHG1 fusions [7][8][9][10].In addition, ESR1 had gene fusions with multiple other genes [7].These human ESR1 fusions indicated the potential involvement of ESR1 fusions in RETP of Holstein cows.The protein encoded by ESR1 regulates the transcription of many estrogen-inducible genes that play a role in growth, metabolism, sexual development, gestation, and other reproductive functions; and is expressed in many nonreproductive tissues [11].An SNP in ESR1 had a significant dominance effect on daughter pregnancy rate in Holstein cows [12].The well-documented ESR1 functions including

Additive Effects of Chr09
Chr09 had twenty-three significant effects including the #1 effect in LOC112448080, #2 upstream of AKAP12, #3 in CCDC170, and #5 in RMND1 (Tables 1 and S1, Figure 2a).These effects were in the 87.61-88.73Mb region about 1.12 Mb in size.An interesting aspect of this region was the estrogen receptor 1 (ESR1) gene fusions observed in humans.This region had PLEKHG1, MTHFD1L, AKAP12, RMND1, and CCDC170 genes upstream of ESR1 (Table 1), and four of these five upstream genes were involved in ESR1 fusions associated with human breast cancer, ESR1-CCDC170, ESR1-AKAP12, ESR1-MTHFD1, and ESR1-PLEKHG1 fusions [7][8][9][10].In addition, ESR1 had gene fusions with multiple other genes [7].These human ESR1 fusions indicated the potential involvement of ESR1 fusions in RETP of Holstein cows.The protein encoded by ESR1 regulates the transcription of many estrogeninducible genes that play a role in growth, metabolism, sexual development, gestation, and other reproductive functions; and is expressed in many non-reproductive tissues [11].An SNP in ESR1 had a significant dominance effect on daughter pregnancy rate in Holstein cows [12].The well-documented ESR1 functions including ESR1 fusions pointed to the possible involvement of ESR1 in the significant effects of RETP in the 87.61-88.73Mb region of Chr09.6)).'al+' is the positive allele, 'al-' is the negative allele, 'ae+' is the allelic effect of the positive allele, and 'ae−' is the allelic effect of the negative allele (Equation ( 7)).'f_al+' is the frequency of the positive allele.'f_al−' is the frequency of the negative allele.'effect' is the additive effect of the SNP as the difference between allelic effects of 'allele 1' and 'allele 2' (Equation ( 6)).'al+' is the positive allele, 'al-' is the negative allele, 'ae+' is the allelic effect of the positive allele, and 'ae−' is the allelic effect of the negative allele (Equation ( 7)).'f_al+' is the frequency of the positive allele.'f_al−' is the frequency of the negative allele.

Additive Effects of Chr23
Chr23 had fifty-two significant effects, the largest number of significant effects among all chromosomes.These effects except one were distributed in the 21.29-35.47Mb region, about 14 Mb in size (Table S1), but the most interesting region was the 23.70-33.08Mb region about 9.4 Mb in size due to the genes potentially contributing to or known to be associated with kidney disease (Table 2, Figure 2b).The most significant effect of this chromosome (#4 overall) was in SLC17A1, followed by SNPs in or near MRPS18B, PPT2, and the region of ELOVL5 to BOLA-DQA2 (Table 2).SLC17A1 with the most significant effect was involved in urate metabolic process and urate transport [13] and elevated serum urate concentrations could contribute to kidney disease [14].The PKHD1 gene had six significant effects and was downstream of another five significant effects (Table S1).This gene was associated with a severe form of polycystic kidney disease named autosomal recessive polycystic kidney disease (ARPKD) that presents primarily in infancy and childhood [15,16].The ACOT13 gene had a significant effect (#16 overall) and was also reported as a candidate gene for ARPKD kidney disease [17].It was interesting that the two genes known to be associated with kidney disease, PKHD1 and ACOT13, were near the two ends of the Chr23 region with significant RETP effects (Table 2, Figure 2b).BOLA-DQA2 is a bovine MHC class II gene [18] and MHC class II molecules are critical for the initiation of the antigen-specific immune response [19].The last significant effect at the very downstream end of the Chr23 region with significant RETP effects was between PRL and HDGFL1, where PRL is the prolactin gene and is essential for lactation [20].'u' indicates the SNP is upstream of the gene.'effect' is the additive effect of the SNP as the difference between allelic effects of 'allele 1' and 'allele 2' (Equation ( 6)).'al+' is the positive allele, 'al-' is the negative allele, 'ae+' is the allelic effect of the positive allele, and 'ae−' is the allelic effect of the negative allele (Equation ( 7)).'f_al+' is the frequency of the positive allele.'f_al−' is the frequency of the negative allele.

Additive Effects of Chr05 and Chr17
Chr05 had thirty-five significant additive effects (Table S1).The most significant effect of Chr05 (#17 overall) was at 27,005,657 bp between the KRT18 and KRT8 genes, 3394 bp downstream of KRT18 and 34,610 bp upstream of KRT8, noting that neither KRT18 nor KRT8 had any SNP inside the gene.Gene Ontology (GO) analysis showed that KRT8 was involved in embryonic placenta development (Table S3), and this biological function could be directly relevant to retained placenta.A SNP upstream of KRT18 (in EIF4B-KRT18) at 26,984,066 bp was also significant (#97 overall).It should be noted that another SNP in EIF4B-KRT18 at 26,964,045 bp had a sharply negative recessive genotype for age at first calving (AFC) [21].The effects in or near KITLG were also interesting due to KITLG's known biological functions.KITLG was upstream of the second-most significant effect of Chr05 (#25 overall), contained a significant effect and was between another two significant effects (Table 3, Figure 3a).The KITLG gene encodes the ligand of the tyrosine-kinase receptor encoded by the KIT locus, and this ligand is a pleiotropic factor that acts in utero in germ cell and neural cell development, and hematopoiesis [22].The 'in utero' biological functions of KITLG could be directly relevant to RETP.The POC1B gene had the third most significant effect of Chr05 (#28 overall).This gene has an important role in basal body and cilia formation [23].'u' indicates the SNP is upstream of the gene.'d' indicates the SNP is downstream of the gene.'effect' is the additive effect of the SNP as the difference between allelic effects of 'allele 1' and 'allele 2' (Equation ( 6)).'al+' is the positive allele, 'al-' is the negative allele, 'ae+' is the allelic effect of the positive allele, and 'ae−' is the allelic effect of the negative allele (Equation ( 7)).'f_al+' is the frequency of the positive allele.'f_al−' is the frequency of the negative allele.
Int. J. Mol.Sci.2024, 25, x FOR PEER REVIEW 5 of 12 known biological functions.KITLG was upstream of the second-most significant effect of Chr05 (#25 overall), contained a significant effect and was between another two significant effects (Table 3, Figure 3a).The KITLG gene encodes the ligand of the tyrosine-kinase receptor encoded by the KIT locus, and this ligand is a pleiotropic factor that acts in utero in germ cell and neural cell development, and hematopoiesis [22].The 'in utero' biological functions of KITLG could be directly relevant to RETP.The POC1B gene had the third most significant effect of Chr05 (#28 overall).This gene has an important role in basal body and cilia formation [23].
Chr17 had fifty-one significant effects (Table S1).The most significant effect of Chr17 (#6 overall) was in LOC112442091, and the second-most significant effect of Chr17 (#16 overall) was between LOC112441992 and LOC112442089.All the above three genes had unknown biological functions.The third-most significant effect of Chr17 was between ANAPC10 and HHIP.A 388,434 bp region between IL15 and ZNF330, 31,334 bp downstream of IL15 and 1245 bp upstream of ZNF330, had the fourth-most significant effect of Chr17 (#26 overall) and seven other significant effects (Table 3, Figure 3b).IL15 is a cytokine that regulates T and natural killer cell activation and proliferation [24], and ZNF330 is predicted to enable zinc ion binding activity [25].Chr17 had fifty-one significant effects (Table S1).The most significant effect of Chr17 (#6 overall) was in LOC112442091, and the second-most significant effect of Chr17 (#16 overall) was between LOC112441992 and LOC112442089.All the above three genes had unknown biological functions.The third-most significant effect of Chr17 was between ANAPC10 and HHIP.A 388,434 bp region between IL15 and ZNF330, 31,334 bp downstream of IL15 and 1245 bp upstream of ZNF330, had the fourth-most significant effect of Chr17 (#26 overall) and seven other significant effects (Table 3, Figure 3b).IL15 is a cytokine that regulates T and natural killer cell activation and proliferation [24], and ZNF330 is predicted to enable zinc ion binding activity [25].

Additive Effects of Other Chromosomes
For the remaining nine chromosomes with significant additive effects, only Chr15 and Chr24 had substantial numbers of significant effects, fifteen effects for Chr15 and eleven effects for Chr24, whereas chromosomes 6, 7, 10, 19, 25, 29 and 31 each had 1-3 effects.Examples of these effects are summarized in Table 4, whereas effects not discussed in the main text can be found in Table S1.6)).'al+' is the positive allele, 'al-' is the negative allele, 'ae+' is the allelic effect of the positive allele, and 'ae−' is the allelic effect of the negative allele (Equation ( 7)).'f_al+' is the frequency of the positive allele.'f_al−' is the frequency of the negative allele.
The fifteen effects of Chr15 were distributed over the 26-67 Mb region and the four most significant effects of Chr15 were in SLC1A2 (#48), IL10RA (#89), SPON1 (#90), and downstream of DCDC1 ($95) covering a large distance of the 26-67 Mb region (Figure 4a).SLC1A2 encodes a membrane-bound protein as the principal transporter that clears the excitatory neurotransmitter glutamate and glutamate clearance is necessary for proper synaptic activation and to prevent neuronal damage from excessive activation of glutamate receptors [26].IL10RA encodes a protein that is a receptor for interleukin 10, and is structurally related to interferon receptors [27].SPON1 is predicted to be an extracellular matrix structural constituent and to be involved in cell adhesion [28].The four most significant effects of Chr24 were in a narrow 30.38-31.01Mb region about 0.63 Mb in size (Figure 4b), downstream of SS18, in TAF4B (2 effects) and upstream of TAF4B with overall rankings of #15, #18, #33 and #37, respectively, where SS18 is involved in positive regulation of transcription by RNA polymerase II [29], and TAF4B is involved in initiation of transcription of genes by RNA polymerase II [30].Chr31, the X-Y nonrecombining region of the X-chromosome, had three significant effects (Figure 4c).The top two effects of this chromosome were in MAGED1-MAGED4B (#73) and downstream of NUDT11 (#75).The SNP in MAGED1-MAGED4B had the most negative allelic effect among all SNPs with a low allele frequency of 0.07.Chr06 had three significant effects (#104, #108 and #175 overall) in the SLC4A4-GC-NPFFR2 region (Figure 4d), which had been reported to have highly significant effects on milk yield, fertility and somatic cell score [12,31].Chr10 had two significant effects slightly above the log 10 (1/p) = 8 cutoff value for declaring significance, one in RASGRP1, which activates the Erk/MAP kinase cascade and regulates T-cells and B-cells development, homeostasis and differentiation [32], and one in LIPC which Enables phospholipase A1 activity and triglyceride lipase activity [33].Chr07 had one significant effect (#78 overall) downstream of OR7A5.Olfactory receptors interact with odorant molecules in the nose, to initiate a neuronal response that triggers the perception of a smell.Chr19 had two significant effects, one in ABCA10 (#96 overall) and one in ABCA9 (#125 overall).

Gene Ontology Analysis
Gene Ontology (GO) analysis was conducted to understand the potential biological functions of candidate coding genes of the 200 significant additive effects, and the results are summarized in Table S3 and Figure S1.The GO results provided many more details about the biological functions of the candidate genes than described thus far in this article, e.g., ESR1 and KITLG each had over one hundred biological functions (Table S3).The GO results also identified a few genes involved in embryonic placenta development, including KRT8 of Chr05 downstream of the #17 effect (Table 2), EDNRA of Chr17 upstream of the #61 effect, GCM1 of Chr23 upstream of the #157 and #168 effects, and MAP3K4 of Chr09 with the #155 effect (Table S1).However, the GO results did not include some of the information we collected from journal articles and the National Center for Biotechnology Information (NCBI), e.g., the ESR1 fusions, the in utero biological functions of KITLG, and the association of PKHD1 and ACOT13 with kidney disease.Although the GO analysis identified over 2000 biological functions of the candidate genes (Table S3), none of those biological functions was identified to have a direct effect on retained placenta.In contrast, the GWAS results of this study provided Holstein-specific and high-confidence evidence for the potential associations between the candidate genes and RETP.The combination of the GWAS results of this study with the GO results as well as the biological information of the candidate genes collected elsewhere should provide useful functional annotations of the candidate genes and indications of the potential genetic mechanisms of the significant SNP effects affecting RETL in Holstein cows.

Holstein Population and SNP Data
The Holstein population in this study had 632,212 cows with RETP phenotypic observations and 78,964 original and imputed SNPs.With the requirement of 0.05 minor allele frequency, 74,747 SNPs were used in the GWAS analysis.The SNP positions were those from the ARS-UCD1.3cattle genome assembly.Genes containing or in proximity to highly significant SNP effects were identified as candidate genes affecting RETP.The RETP phenotypic values used in the GWAS analysis were the phenotypic residuals after removing fixed nongenetic effects available from the December 2023 U.S. Holstein genomic evaluation data.

GWAS Analysis
The GWAS analysis used an approximate generalized least squares (AGLS) method.The AGLS method combines the least squares (LS) tests implemented by EPISNP1mpi [34,35] with the estimated breeding values from a routine genetic evaluation using the entire U.S. Holstein population.The statistical model was: where y = column vector of phenotypic deviation after removing fixed nongenetic effects such as heard-year-season (termed as 'yield deviation' for any trait) using a standard procedure for the CDCB/USDA genetic and genomic evaluation; µ = common mean; I = identity matrix; g = column vector of genotypic values; X g = model matrix of g; b = (µ, g ′ ) ′ , X = (I, X g ); a = column vector of additive polygenic values; Z = model matrix of a; and e = column vector of random residuals.The first and second moments of Equation ( 1) are: E(y) = Xb and var(y) = V = ZGZ ′ + R = σ 2 a ZAZ ′ + σ 2 e I, where σ 2 a = additive variance, A = additive relationship matrix, and σ 2 e = residual variance.The problem of estimating the b vector that includes SNP genotypic values in Equation ( 1) is the requirement of inverting the V if the generalized least squares (GLS) method is used, or solving the mixed model equations (MME) [36].Either the GLS or MME method for each of the genome-wide SNPs is computationally demanding for our sample size.To avoid these computing difficulties, the GWAS used the method of approximate GLS (AGLS) that replaces the polygenic additive values (a) with the best linear unbiased prediction based on pedigree relationships [12,21,31,37].The significance tests for additive and dominance SNP effects used the t-tests of the additive and dominance contrasts of the estimated SNP genotypic values [34,38].The t-statistic of the AGLS was calculated as: where L j = additive or dominance contrast; var(L j = standard deviation of the additive or dominance contrast; s a = row vector of additive contrast coefficients = P 11 /p 1 0.5P 12 (p 2 − p 1 )/(p 1 p 2 ) −P 22 /p 2 ; s d = row vector of dominance contrast coefficients = −0.5 1 0.5 ; v 2 = (y − X b) ′ (y − X b)/(n − k) = estimated residual variance; ĝ = column vector of the AGLS estimates of the three SNP genotypic effects of g 11 , g 12 , and g 22 from Equation (4); (X ′ X) − gg = submatrix of (X ′ X) − corresponding to ĝ; and where p 1 = frequency of A 1 allele, p 2 = frequency of A 2 allele of the SNP, P 11 = frequency of A 1 A 1 genotype, P 12 = frequency of A 1 A 2 genotype, P 22 = frequency of A 2 A 2 genotype, n = number of observations and k = rank of X.
Additive effects of each SNP were estimated using three measures, the average effect of gene substitution, allelic mean, and allelic effect of each allele based on quantitative genetics definitions [38,39].The allelic mean (µ i ), the population mean of all genotypic values of the SNP (µ), the allelic effect (a i ), and the average effect of gene substitution of the SNP (α) are: where P 11.1 = P 11 /p 1 , P 12.1 = P 12 /p 1 , P 12.2 = P 12 /p 2 , and P 22.2 = P 22 /p 2 .The additive effect measured by the average effect of gene substitution of Equation ( 7) is the distance between the two allelic means or effects of the same SNP, and is the fundamental measure for detecting SNP additive effects as shown by the t-statistic of Equation ( 2).However, the allelic effect of Equation ( 7) is not comparable across SNPs because the allelic effect is affected by the genotypic mean of the SNP defined by Equation (6).To compare allelic effects across SNPs, we replaced the SNP genotypic mean (µ) in Equation ( 6) with the average of all SNP genotypic means (µ all ): Equation ( 8) was used only for the purpose of graphical display of allelic effects.

Gene Ontology (GO) Analysis
To understand the potential functions of selected candidate genes, the Gene Ontology (GO) analysis was performed using the OmicShare platform (www.omicshare.com/tools,accessed on 15 May 2024).

Conclusions
The GWAS results in this study indicated that RETP in U.S. Holstein cows was affected by multiple genetic variants with additive effects.Although the exact genetic mechanism underlying RETP remained unknown, these significant effects involved genes with a variety of biological functions reported elsewhere including ESR1 gene fusions, immunity, genetic effects on fertility, health and milk production, kidney disease, lactation, KIT ligand in utero, and basal body and cilia formation.The SNP effects detected in this study along with known biological functions of genes with or near the SNP effects provided a new understanding of genetic factors underlying RETP in U.S. Holstein cows and provided comparative information about the genetic mechanism of retained placenta in other species.

Figure 1 .
Figure 1.Graphical view of additive effects.(a) Manhattan plot of additive effects of all chromosomes.(b) allelic effects of the 151 significant SNPs.'u' indicates the SNP is upstream of the gene.'d' indicates the SNP is downstream of the gene.

Figure 1 .
Figure 1.Graphical view of additive effects.(a) Manhattan plot of additive effects of all chromosomes.(b) allelic effects of the 151 significant SNPs.'u' indicates the SNP is upstream of the gene.'d' indicates the SNP is downstream of the gene.
Int. J. Mol.Sci.2024, 25, x FOR PEER REVIEW 3 of 12 ESR1 fusions pointed to the possible involvement of ESR1 in the significant effects of RETP in the 87.61-88.73Mb region of Chr09.

Figure 2 .
Figure 2. Additive effects of Chr09 and Chr23.(a) Statistical significance of additive effects of Chr09 SNPs.(b) Statistical significance of additive effects of Chr23 SNPs.'u' indicates the SNP is upstream of the gene.

Figure 2 .
Figure 2. Additive effects of Chr09 and Chr23.(a) Statistical significance of additive effects of Chr09 SNPs.(b) Statistical significance of additive effects of Chr23 SNPs.'u' indicates the SNP is upstream of the gene.

Figure 3 .
Figure 3. Additive effects of Chr05 and Chr17.(a) Statistical significance of additive effects of Chr05 SNPs.(b) Statistical significance of additive effects of Chr17 SNPs.'d' indicates the SNP is downstream of the gene.

Figure 3 .
Figure 3. Additive effects of Chr05 and Chr17.(a) Statistical significance of additive effects of Chr05 SNPs.(b) Statistical significance of additive effects of Chr17 SNPs.'d' indicates the SNP is downstream of the gene.

Figure 4 .
Figure 4. Additive effects of Chr15, Chr24, Chr31 and Chr06.(a) Statistical significance of additive effects of Chr15 SNPs.(b) Statistical significance of additive effects of Chr24 SNPs.(c) Statistical significance of additive effects of Chr31 SNPs.(d) Statistical significance of additive effects of Chr06 SNPs.'u' indicates the SNP is upstream of the gene.'d' indicates the SNP is downstream of the gene.

Figure 4 .
Figure 4. Additive effects of Chr15, Chr24, Chr31 and Chr06.(a) Statistical significance of additive effects of Chr15 SNPs.(b) Statistical significance of additive effects of Chr24 SNPs.(c) Statistical significance of additive effects of Chr31 SNPs.(d) Statistical significance of additive effects of Chr06 SNPs.'u' indicates the SNP is upstream of the gene.'d' indicates the SNP is downstream of the gene.

Table 1 .
Significant additive effects of Chr09 for RETP.u' indicates the SNP is upstream of the gene.'d' indicates the SNP is downstream of the gene.'effect' is the additive effect of the SNP as the difference between allelic effects of 'allele 1' and 'allele 2' (Equation ( '

Table 1 .
Significant additive effects of Chr09 for RETP.
'u' indicates the SNP is upstream of the gene.'d' indicates the SNP is downstream of the gene.

Table 3 .
Significant additive effects of Chr05 and Chr17 for RETP.

Table 3 .
Significant additive effects of Chr05 and Chr17 for RETP.

Table 4 .
Significant additive effects of seven selected chromosomes for RETP.
'd' indicates the SNP is downstream of the gene.'effect' is the additive effect of the SNP as the difference between allelic effects of 'allele 1' and 'allele 2' (Equation (

Table 4 .
Significant additive effects of seven selected chromosomes for RETP.