Cross-breed comparisons identified a critical 591-kb region for bovine carcass weight QTL (CW-2) on chromosome 6 and the Ile-442-Met substitution in NCAPG as a positional candidate

Background Growth-related traits have been mapped on bovine chromosome 6 (BTA 6) in various bovine breed populations. We previously mapped a significant quantitative trait locus (QTL) for carcass and body weight (CW-2) between 38 and 55 cM on BTA 6 using a Japanese Black half-sib family. Additional QTL mapping studies detected four QTL for body or carcass weight that overlapped with CW-2 in Japanese Black and Japanese Brown half-sib families. To map the region in greater detail, we applied cross-breed comparisons of haplotypes that have been shown to be powerful in canine. Results We used 38 microsatellite markers to search for a shared Q (increasing carcass and/or body weight) haplotype within the 17-cM CW-2 region among five sires. Linkage disequilibrium mapping using maternal alleles of the offspring showed that an 815-kb shared Q haplotype was associated with body or carcass weight in both breeds. The addition of 43 single nucleotide polymorphism (SNP) markers narrowed the region to 591 kb containing 4 genes. The SNP changing Ile-442 to Met in NCAPG (chromosome condensation protein G) was significantly associated with carcass weight (p < 1.2 × 10-11) in a large Japanese Black population as well as in the five families. The Q allele of the SNP was also associated with a larger longissimus muscle area and thinner subcutaneous fat thickness in steers of all five families, indicating that the CW-2 locus is pleiotropic and favorable for marker-assisted selection of beef cattle. Conclusion A 591-kb critical region for CW-2 was identified. The SNP changing Ile-442 to Met in NCAPG (chromosome condensation protein G) can be used as a positional candidate of CW-2 for marker-assisted selection.


Background
Body size is one of the most visible animal characteristics and many genes can affect body size. In cattle, body size is correlated with meat quantity, an economically important trait that varies within and across breeds. We previously performed bovine quantitative trait locus (QTL) mapping for growth and carcass traits using Japanese Black paternal half-sib families constructed from a commercial population [1][2][3]. In these studies, a carcass weight QTL, CW-1 on bovine chromosome 14 (BTA 14), was detected in five families with significant linkages and successfully narrowed down to a 1.1-Mb region by identical-by-descent mapping [3] and linkage disequilibrium (LD) mapping [4]. In contrast with CW-1, another significant QTL for carcass and body weight, designated CW-2, on BTA 6 was replicated at a 1% chromosome-wise significance level, but the two significantly segregating sires had no apparent shared Q haplotypes [3]. LD mapping only narrowed the region down to 13.1 Mb (see Results). We thus changed strategies to narrow down the CW-2 region. We recently detected carcass weight QTL in regions that overlapped with CW-2 in Japanese Brown populations. The Japanese Black breed was established in 1948, basically from indigenous populations of the Japanese Islands, while the Japanese Brown breed originated from indigenous populations of the Korean Peninsula. Approximately 100 years ago, the Korean cattle were imported to Japan and crossed several times with Simmental bulls, followed by the establishment of the Japanese Brown breed in 1948. Therefore, it can be estimated that the two indigenous populations have survived separately for thousands of years. It will be interesting if a common carcass weight QTL is present across the entire Bos taurus population. Indeed, the CW-2 region was repeatedly highlighted for its association with growth-related traits in a mixed breed population [5], postweaning growth in a beef cattle population [6], and weight and body length at birth in a Charolais × Holstein cross-bred population [7]. These data indicate that CW-2 may be shared across breeds. If this is the case, cross-breed comparisons might be useful for fine-mapping to identify a shared hypothetical identicalby-descent haplotype.
In dogs, cross-breed comparisons have been proposed for fine-mapping to pinpoint disease-related genes [8]. Canine LD patterns reflect two bottlenecks in dog history (early domestication and recent breed creation) and longrange breed-specific haplotype blocks retain the underlying short-range ancestral haplotype blocks, suggesting that genetic risk factors are shared across breeds. This situation provides for an efficient mapping strategy: initial mapping within breeds and subsequent fine-mapping by cross-breed comparisons. The strategy was validated in coat color and hair ridge studies [9], in which cross-breed comparisons successfully refined the location of the coat color locus from 800 kb to 100 kb within the pigmentation-related gene. A recent study in cattle also revealed footprints of ancestral LD at short distances (< 10 kb) and these ancestral blocks are organized into larger blocks of a few hundred kilobases within a breed [10].
Here, we demonstrate that cross-breed comparisons are also efficient for fine-mapping of QTL in cattle. The CW-2 locus was narrowed down from a 17-cM region to a 591kb region shared by the Japanese Black and Japanese Brown breeds. Furthermore, a single nucleotide polymorphism (SNP) changing Ile-442 to Met in NCAPG (chromosome condensation protein G) was identified as a positional candidate. The NCAPG-LCORL locus was recently mapped as a locus associated with human adult height [11,12]. The findings of the present study indicate that genetic variations in NCAPG may affect body size in cattle as well as in humans.

Carcass and/or body weight QTL in Japanese Black and Japanese Brown cattle
Carcass and/or body weight QTL on BTA 6 were mapped in three Japanese Black half-sib families (Sires A-C) and two Japanese Brown half-sib families (Sires D and E; Figure 1; Additional file 1). Sires A and B had the same Q haplotype (Q1) in the QTL region, whereas Sire C had a different Q haplotype (Q2) (Figure 2A). Sire E is an offspring of Sire D and inherited the Q haplotype of Sire D (Q3). In these QTL analyses, the highest F-statistic was obtained in Sire A's family, whose QTL was previously designated as CW-2 [3]. Because the 95% confidence intervals (CI) of the QTL in the other families overlapped with the CW-2 region (38-55 cM), we assumed that these QTL had an identical genetic origin and attempted to find a shared haplotype among the three Q haplotypes. By comparing the three Q haplotypes using 38 microsatellite markers with an interval of less than 1.18 Mb, three possible identical-by-state (IBS)Q regions were detected between 38 and 55 cM: DIK9006-BMS2508 (663 kb), DIK9010-DIK9011 (537 kb), and DIK4852-DIK9017 (815 kb) (Figure 2A; Additional file 2).

Identical-by-state haplotype associated with carcass and body weight
To examine whether any of the possible IBS regions were associated with carcass or body weight, we performed LD mapping between 38 and 54 cM using the maternal alleles of the offspring from Sire C (Japanese Black) and Sires D and E (Japanese Brown) ( Figure 3). The offspring steers were divided into two groups according to their paternally inherited haplotypes (Q or q) in this region. The average carcass or body weight was then compared between the steers that harbored two consecutive Q alleles in the maternal alleles and the other steers in each group. In Sire F-statistic profiles for carcass or body weight (Sire C only) on BTA 6 Figure 1 F-statistic profiles for carcass or body weight (Sire C only) on BTA 6. Marker locations were obtained from the Shirakawa-USDA linkage map [20]. Bold and thin lines indicate a fixed-effect QTL model and a fixed-effect QTL model accounting for the effect of SNP-9 of NCAPG, respectively. The boxes on the x-axis indicate the 95% CI of the QTL. The horizontal lines indicate the thresholds for chromosome-wise 0.1% (---), 1% (--), and 5% (---) significance levels. C's family, the frequency of two consecutive Q2 alleles among the maternal alleles was less than 3% in most regions, while the frequency of two consecutive Q1 alleles was greater than 17% between DIK1058 and DIK9019. Therefore, average body weight was compared between the steers harboring Q1 alleles in the maternal alleles and those harboring non-Q1 alleles, and between the steers harboring maternal Q2 alleles and those harboring non-Q2 alleles, in the regions where corresponding Q frequencies for two consecutive markers were higher than 5%.
Body weight was significantly higher (p < 10 -3 ) in the regions between DIK1058 and DIK9019 in the maternal Q1-harboring steers than in the non-Q1-harboring steers in the paternal q-inherited offspring group ( Figure 3A). This finding allowed us to narrow down the CW-2 region to the 13.1 Mb flanked by DIK1058 and DIK9020, which contained all possible IBS regions. As for Q2, Sire C was homozygous in the telomeric part at the 95% CI ( Figures  1 and 2), which allowed us to narrow down the QTL to the centromeric side of DIK9017.
ctg6030 ctg6020 LD mapping using maternal alleles of the offspring in Japanese Black and Japanese Brown families 1.E-03 1.E-02 1.E-01 In Japanese Brown families, offspring from Sires D and E were combined to maximize the number of offspring. Carcass weight was higher in DIK4852-DIK9016 (p = 0.011) and in DIK9019-MNB202 (p = 0.0052 at DIK9019-DIK9020) in the maternal Q3-harboring steers than in the non-Q3-harboring steers among the offspring with the paternally inherited q haplotype ( Figure 3B). The two regions were in linkage disequilibrium. Among 161 steers that inherited the paternal q haplotype, 50 harbored maternal Q3 in DIK4852-DIK9016, 31 of which harbored maternal Q3 in DIK9019-DIK9020. On the other hand, all of the steers that harbored maternal Q3 in DIK9019-DIK9020 also harbored maternal Q3 in DIK4852-DIK9016. The DIK4852-DIK9016 region was included in the narrowed-down CW-2 region and corresponded to one of the possible IBS regions, whereas the DIK9019-MNB202 region was included in a homozygous region in Sire C. An association of the DIK4852-DIK9016 region with carcass weight was also observed among offspring with the paternally inherited Q haplotype.
These results strongly suggest that the causal alleles in the three QTL haplotypes Q1, Q2, and Q3 are identical, and that the IBS region between DIK4852 and DIK9017 might be responsible for the QTL.

Ile-442-Met Mutation as a positional candidate
The possible IBS region between DIK4852 and DIK9017 corresponds to the 17. To assess the association of these SNPs with carcass weight, 187 steers selected from the 4.7% extremes of 7990 Japanese Black steers were genotyped. SNP-9 of NCAPG had the strongest association (p < 1.2 × 10 -11 , Fisher's exact test for independence; frequency of G-allele, 0.14; Figure 4A; Additional file 3), and it was in strong linkage disequilibrium with the four intronic SNPs (Figure 4BC). These data indicated that SNP-9 is the most appropriate CW-2 marker among the examined SNPs as a candidate for the causative sequence variation, although three other genes are in the same LD block and cannot be excluded as candidate genes for CW-2.
To further confirm the association of SNP-9 with carcass weight in Japanese Brown as well as Japanese Black populations, the steers of all five families and the cows of Sire D's offspring were genotyped with SNP-9 and average carcass weight was compared between genotypes in each family. Because steers and cows sometimes show different trait characteristics, the cows were separately evaluated from the steers in Sire D family (Figure 5). The G/G-genotype (442-Met/Met of NCAPG) and G/T-genotype (442-Met/Ile of NCAPG) had significantly greater average carcass weight than the T/T-genotype (442-Ile/Ile of NCAPG) in the steers of all five families and the cows of Sire D's offspring (p < 0.01, t-test). The carcass weight of the G/G-genotype tended to be higher than that of the G/T-genotype, but the difference was significant only in the families of Sires B and E. The effect of SNP-9 was also confirmed by QTL analyses including the genotype of SNP-9 as a fixed effect. The QTL peaks were all extinguished by including the SNP-9 genotype ( Figure 1). These data confirmed that SNP-9 of NCAPG is a positional candidate for the quantitative trait nucleotide for CW-2.

Pleiotropy of CW-2
QTL mapping studies detected QTL for the longissimus muscle area (LMA) and subcutaneous fat thickness (SFT) in the regions overlapping with CW-2 at less than 5% chromosome-wise significance levels in families of Sires A and B (Additional file 4). A QTL for LMA was also detected in Sire D's family at the 1% chromosome-wise significance level (Additional file 4). These QTL peaks were also extinguished in a fixed effect QTL model accounting for the effect of SNP-9 (Additional file 4). Therefore the SNP-9 genotypes were tested for LMA and SFT in each family ( Figure 5). The G/G-or G/T-genotypes had a significantly greater average LMA than the T/T-genotype in the steers of the five families and the cows of Sire D's offspring (p < 0.01, t-test), although the correlation between carcass weight and LMA traits was not high (r = 0.26-0.54). On the other hand, the G/G-genotype had a significantly Association study in the IBS region using a Japanese Black population    0.039, t-test). This is a particularly interesting finding because the correlation between carcass weight and SFT traits was r = 0.18-0.39 in each family (Additional file 5), meaning that greater carcass weight was correlated with thicker SFT. For example, a Q allele of another carcass weight QTL CW-1on BTA 14 is associated with greater carcass weight and thicker SFT [1]. In CW-2, however, the SNP-9 genotype was associated with thinner SFT in the steers of the five families, indicating that the CW-2 locus is pleiotropic. The results further support the notion that the CW-2 QTL in the Japanese Black population is the same as the QTL in a respective position in the Japanese Brown population.

Discussion
Here we show that cross-breed comparisons are useful for fine-mapping of the QTL in cattle, as recently demonstrated in the fine-mapping of a monogenic trait locus in dogs [9]. We detected a carcass weight QTL on BTA 6 in two breed populations, Japanese Black and Japanese Brown. The genetic distance between the two cattle breeds can be thousands of years, because their ancestors have survived separately on the Japanese Islands and the Korean Peninsula. Linkage disequilibrium mapping using the two breeds narrowed the CW-2 QTL region to a 591kb section. Comparison of the Q haplotypes inherited in different lineages is, therefore, a useful strategy for narrowing down a QTL.
One of the critical issues should be whether the CW-2 Q in the Japanese Black population is identical to the QTL in the Japanese Brown population. In this study, the effects of the CW-2 on carcass weight, LMA, and SFT traits were similar in the two breed populations. Namely, the CW-2 Q allele was associated with greater carcass weight, larger LMA, and thinner SFT. Importantly, although greater carcass weight is generally associated with a thicker SFT, in both populations the CW-2 Q allele was associated with a thinner SFT. In addition, QTL for growth-related traits were mapped around the CW-2 region in various cattle populations [5][6][7], suggesting that the CW-2 Q allele is widely distributed in Bos taurus populations.
The 591-kb critical region contains four genes, LOC523874, C6H4orf30, NCAPG, and LOC540095 (the bovine ortholog of LCORL). The NCAPG-LCORL region was recently identified as a QTL for human adult height [11,12]. An SNP causing Ile-442 to Met in NCAPG was detected as a positional candidate for the quantitative trait nucleotide for CW-2. The NCAPG gene encodes chromosome condensation subunit G, which is a catalytic subunit of the mammalian condensin I complex. Disruption of the NCAPG ortholog (cnd3) causes defective chromosome condensation in fission yeast [16], while in Drosophila, the NCAPG ortholog (Cap-G) is essential for chromosome condensation in the metaphase of single, unreplicated sister chromatids, and also has a role during the interphase in regulating heterochromatic gene expression [17]. In Hela cells, NCAPG interacts with HSF2 to mediate HSPA1A (hsp70i) bookmarking [18]. The NCAPG expression is upregulated in more aggressive metastatic melanomas than in less aggressive primary melanomas [19]. These findings indicate that NCAPG may affect cell proliferation and growth through regulating cell cycle and chromosome condensation. Further studies are required to reveal the molecular mechanism that links NCAPG to growth and body size.

Conclusion
We identified a 591-kb critical region for CW-2. Among the four genes located within the region, NCAPG is a candidate for the causative gene because it has an amino acid substitution that is significantly associated with carcass weight. The SNP for the amino acid substitution provides a useful CW-2 marker for marker-assisted selection. The pleiotropic characteristic of CW-2, especially the effect on SFT, is clearly different from a previously fine-mapped carcass weight QTL, CW-1 [1]. Therefore, the two QTL probably affect carcass weight in different ways and both will be useful for breeding beef cattle.

DNA samples and phenotype data
Paternal half-sib families of Sires A through E were constructed from carcass data and pedigree records collected by the Japan Wagyu Register Association (Kyoto, Japan) and Kumamoto Union of Livestock and Agriculture Cooperation (Kumamoto, Japan). Offspring of each sire were reared in different herds in a prefecture. The offspring of each sire were collected over a period ranging from 2 to 5 years. The average slaughter age was 878 days for the offspring of Japanese Black Sires A through C and 750 days for the offspring of Japanese Brown Sires D and E. Sire DNA was obtained from semen. Offspring DNA samples were collected from adipose tissues around the kidney at the slaughterhouses or blood at the individual farmer's houses. Six traits were analysed in this study: body weight at slaughter (Sire C only), cold carcass weight, LMA, rib thickness (thickness of a muscle layer in a rib of beef), SFT, and marbling. These traits were systematically measured by certified graders and recorded at the slaughterhouses in Japan.

QTL mapping
The genome or chromosome screen was conducted using the microsatellite markers on the Shirakawa-USDA linkage map [20]. QTL analyses were performed with the interval mapping method using a linear regression model for half-sib families [21,22], as described previously [1].
Linear regression analysis was performed using the following model: where y is the vector of phenotypic value, X is the design matrix of fixed effects composed of sex, slaughter year, age (day), and probability of having the Q phase at a given location(Prob(Q)), b is the vector of fixed effects, and e is the residual error. b was estimated by the least squares method. An F-statistic value at each position was calculated from the residual sum square regressed with Prob(Q), and the total residual sum square without Prob(Q). The analysis was performed at 2-cM intervals along each chromosome. To evaluate whether the QTL effect was well estimated, the information content was calculated as a variance of Prob(Q) divided by 0.25, which was the maximum possible value of Prob(Q) [23]. The allele substitution effect from q to Q was calculated as an estimator of the cofactor for Prob(Q) in the b. The contribution ratio was calculated as a proportion of the trait variance explained by the paternal allele substitution from q to Q. A threshold for significance of the F-statistic value was obtained by 10,000 random permutations of the phenotypic data [24]. The 95% CI of the QTL locations was calculated using the bootstrapping method [25]. Briefly, a set of offspring was chosen so as to be the same number as the original half-sibs by resampling from the original half-sibs. Resampling was repeated 10,000 times. The position of the F-statistic peak in each bootstrapping was collected. The CI was determined by the distribution of the peaks. Therefore, the CI may be fragmented into separated regions, and not a single contiguous region.

Microsatellite development and genotyping
Microsatellites were searched for in the genomic sequences and the primers were designed using Primer 3 [26]. Twenty microsatellite markers from DIK9001 to DIK9020 were developed in this study. The markers were anchored to the BAC fingerprint map [27] by a BLAST search (BLASTN) against bovine BAC sequences (HTGS) [28] or by screening bovine BAC libraries, RPCI-42 [29] and CHORI-240 [30]. Marker information, such as primer sequences, genomic positions, and locations in the BAC fingerprint map, is shown in Additional file 2. Genotyping was performed using polymerase chain reaction (PCR) with a fluorescent-labeled reverse primer, followed by electrophoresis using ABI 3730 DNA analyzer (Applied Biosystems, Foster City, CA) and analysis using GeneMapper software (Applied Biosystems). The sires and their offspring were genotyped to determine the phase of the sires' chromosomes.

Linkage disequilibrium mapping
Linkage disequilibrium mapping was performed using the maternal alleles of the offspring steers. First, offspring were divided into two groups according to paternally inherited haplotypes (Q or q) of the CW-2 region (38-55 cM). When a crossover was observed or suspected within the region in the paternal haplotype, the offspring were eliminated. The numbers of paternal Q-and q-inherited steers were 126 and 117 in Sire C's family, 106 and 94 in Sire D's family, and 104 and 67 in Sire E's family, respectively. The maternal allele was determined by subtracting the paternal allele from the genotype. The markers for which the number of alleles was less than four (Sire C family) or three (families of Sires D and E), or heterozygosity was less than 0.4 were eliminated. The effect of pairwise Q alleles on body or carcass weight was evaluated in each group using a t-test. Phenotypic values for carcass and body weight were adjusted based on slaughter year and age in the family. The average of the adjusted carcass or body weight was compared between the steers that harbored two consecutive Q alleles in the maternal alleles and the other steers, in each group. In Sire C family, the test was performed for Q1 and Q2 alleles in the regions where corresponding Q frequencies were higher than 5%. In Japanese Brown families, offspring from Sires D and E were combined to maximize the number of offspring. Carcass weight was not adjusted between families, because the effect of sire was not significant (p = 0.21). Frequencies of two consecutive Q3 alleles in the maternal alleles were more than 4.3% in the entire region.

SNP discovery and genotyping
Six genes (LAP3, MED28, LOC523874, C6H4orf30, NCAPG, and LOC540095) were located between DIK4852 and DIK9017. SNPs in the coding regions of the genes were searched for by resequencing Sire A. SNPs chosen from a 120 K SNP collection [31] and a commercially available bovine 50 K SNP chip (Illumina, San Diego, CA) were also used. The sires and their offspring were genotyped to determine the phases of the sires' chromosomes. SNPs were genotyped by direct sequencing of the genomic PCR products. As for SNP (Indel) 19, PCR was performed using a fluorescent-labeled reverse primer and genotyped in the same way as a microsatellite marker. SNP information, such as alleles, genomic positions, and primer sequences used for amplification and sequencing, is shown in Additional file 3.

Association study
A population consisting of 7990 Japanese Black steers was constructed from DNA samples collected at one slaughterhouse over 7 years. These cattle were reared in different herds throughout Japan. To maximize the power and minimize the pedigree effect of a selected population, steers were selected from the 4.7% extremes, from many differ-y Xb e = + , ent sires, and did not include more than 5 half-sibs in each extreme. The selected population consisted of 90 offspring (> 570 kg) from 41 sires and 97 offspring (< 410 kg) from 52 sires, among which 18 sires were common. The 187 steers were genotyped with SNPs. The association of each SNP with carcass weight was assessed by Fisher's exact test for independence using a 2 × 2 contingency table consisting of an observed number of each allele in each extreme population.

Calculation of LD
SNPs were confirmed not to have deviated from Hardy-Weinberg equilibrium (p > 0.05). Haplotype frequencies consisting of two SNPs were estimated using the expectation-maximization algorithm of Dempster et al. [32]. An LD coefficient (r 2 ) between two SNPs was calculated from the estimated haplotype frequencies.