The increasing evidence of fetal developmental effects on postnatal life, the still unknown fetal growth mechanisms impairing offspring generated by somatic nuclear transfer techniques, and the impact on stillbirth and dystocia in conventional reproduction have generated increasing attention toward mammalian fetal growth. We identified a highly significant quantitative trait locus (QTL) affecting fetal growth on bovine chromosome 6 in a specific resource population, which was set up by consistent use of embryo transfer and foster mothers and, thus, enabled dissection of fetal-specific genetic components of fetal growth. Merging our data with results from other cattle populations differing in historical and geographical origin and with comparative data from human whole-genome association mapping suggests that a nonsynonymous polymorphism in the non-SMC condensin I complex, subunit G (NCAPG) gene, NCAPG c.1326T>G, is the potential cause of the identified QTL resulting in divergent bovine fetal growth. NCAPG gene expression data in fetal placentomes with different NCAPG c.1326T>G genotypes, which are in line with recent results about differential NCAPG expression in placentomes from studies on assisted reproduction techniques, indicate that the NCAPG locus may give valuable information on the specific mechanisms regulating fetal growth in mammals.
FARM animal species are receiving a growing interest as animal models, because increasingly whole-genome sequences have become available (e.g., for cattle, horses, or chicken) as well as specific resource populations managed under standardized conditions (Andersson and Georges 2004). Fetal development plays a major role in postnatal physiology in mammals as demonstrated for humans, laboratory animals, and livestock (Park et al. 2008; Corson et al. 2009; Freathy et al. 2009). Additionally, in cattle, fetal growth has raised attention due to the still unknown fetal growth mechanisms impairing calves generated by somatic nuclear transfer techniques and because of its impact on stillbirth and dystocia in conventional reproduction. Calves generated by somatic cloning techniques are frequently affected by the large offspring syndrome, which is characterized by abnormally large fetal growth (Hill et al. 1999). The underlying mechanisms are still poorly understood. Besides artificially induced variation in fetal development, there is also substantial conventional genetic variation in fetal growth with a medium heritability in cattle (h2 = 0.2–0.6) (Hansen et al. 2004; Phocas and Laloe 2004). In conventional cattle production, fetal growth has an important impact, because it is highly correlated to incidence of dystocia and stillbirth (Hansen et al. 2004; Phocas and Laloe 2004), raising animal welfare issues. Thus, the identification of the factors modulating fetal growth would assist artificial reproduction and conventional animal breeding and also would contribute to an increased knowledge of pathways regulating fetal development. As exemplified by Candille et al. (2007), the detection of the background of naturally occurring genetic variation can elucidate previously unknown physiological pathways.
In the literature, several studies in different dairy and beef cattle populations concurrently reported that bovine chromosome 6 (BTA6) harbors quantitative trait loci (QTL) affecting fetal growth (Davis et al. 1998; Casas et al. 2000; Kneeland et al. 2004; Gutierrez-Gil et al. 2009; Maltecca et al. 2009). These results were in line with other studies providing evidence for QTL affecting stillbirth, dystocia, or calving ease on the same chromosome (Schrooten et al. 2000; Kühn et al. 2003; Holmberg and Andersson-Eklund 2006; Kolbehdari et al. 2008; Olsen et al. 2008; Schulman et al. 2008). This initiated our attempt to scrutinize BTA6 for the genetic background of variation in the fetal component regulating fetal growth. Fetal growth is affected by maternal factors providing resources for the fetus, by fetal factors determining the development of the fetus itself, and by feto–maternal interaction (Murphy et al. 2006). For dissecting these complex developmental processes, specific animal models are required, and use of foster mothers enables the specific investigation of the fetal component influencing fetal growth. Therefore, we took specific advantage of a unique cattle resource population (Kühn et al. 2002) generated from Charolais and German Holstein, representatives of beef and dairy cattle, respectively, which are known to differ substantially regarding fetal growth (Marlowe et al. 1984; Noakes 2001). This resource population was created by consistent application of embryo transfer to foster mothers, which separates genetically determined exclusively fetal effects from genetically determined maternal or embryo–maternal factors. Using this animal model, we identified a genetic locus affecting the fetal component of fetal growth on BTA6 by linkage and association analysis. Furthermore, merging our data with respective intraspecies and across-species information highlighted the non-SMC condensin I complex, subunit G (NCAPG) gene as strongly associated with fetal growth.
MATERIALS AND METHODS
Our study included 733 individuals from a resource population (Kühn et al. 2002) generated from the founder breeds Charolais and German Holstein: 5 Charolais males, 94 German Holstein females, 168 F1, and 466 F2. All F1 and F2 animals were generated by multiple ovulation and embryo transfer, which created full- and half-sib F1 and F2 families. Embryos were transferred randomly to virgin heifers kept in the experimental herd of the FBN Dummerstorf according to a standard protocol. Foster mothers were taken randomly from the German Holstein production population with no further specific genetic relationship between them. Female F2's were mated back to a German Holstein sire, which created a backcross (BC) population comprising 139 offspring available for genotyping.
During the postnatal period, the calves were housed in a loose stall barn and were fed a milk replacer diet according to requirements, which was gradually replaced by ad libitum access to hay and concentrates. For all calves, birth weight and subsequently weight at 121 days of age were recorded (Table 1). Effects that affected fetal and postnatal growth within the F2 generation or the BC generation were calculated by applying a linear modelwhere traits investigated were birth weight, weight at 121 days of age, and average daily gain in the time period birth to day 121; sexi is the fixed effect of sexi; and birth yearj is the fixed effect of the jth year of birth. The resulting residuals eij for the F2 individuals were obtained and employed in all further QTL and association analyses. For the analysis of the BC individuals, the parity of dam (first or second parity) was added to the model as a fixed effect and the respective residuals were used.
Markers, genotypes, and haplotypes:
Initially, the individuals from the P0, F1, and F2 generation were genotyped for 16 microsatellite markers (see supporting information, Table S1) evenly distributed across the chromosome according to standard protocols (Kühn and Weikard 2007). In the initial QTL confidence interval for fetal growth determined by bootstrapping (see below), 23 candidate SNPs were added along with 13 intragenic SNPs far outside of the targeted QTL region serving as negative controls (see Table S1). Genotyping of SNPs was performed on an Illumina Bead station (Oliphant et al. 2002) (http://www.illumina.com) as part of a targeted 384-SNP GoldenGate assay. The NCAPG c.1326T>G SNP was genotyped by PCR–RFLP, using restriction enzyme TasI (Fermentas) (amplification primers: forward 5′-ATTTAGGAAACGACTACTGG-3′ and reverse 5′-ATTTGTATTCTCTTATTATCATC-3′). First, all microsatellite markers and then additionally, all SNP markers were included to calculate a genetic map using CRIMAP Version 2.50 options [Green et al. 1990, incorporating modifications by Ian Evans and Jill Maddox (University of Melbourne)]. Markers unseparated by crossovers were ordered according to Weikard et al. (2006) and according to the bovine BTA6 assembly Build 4.0 (NCBI accession no. NC_007304). For those marker pairs sharing an identical position on our genetic map, a distance of 0.01 cM was assumed to provide a minimal distance required for QTL linkage mapping. Minor allele frequencies and tests for deviation from Hardy–Weinberg equilibrium for the SNPs were calculated in founder breeds (see Table S1). The most likely haplotypes of the individuals comprising microsatellites and SNPs were determined by a Markov chain Monte Carlo (MCMC) algorithm implemented in Qxpak (Perez-Enciso and Misztal 2004). Pairwise linkage disequilibrium between the 23 SNPs distributed across the initial QTL confidence interval was determined by calculating pairwise r2 values, using the Haploview program (Barrett et al. 2005).
As a robust starting point for QTL analysis, we applied a QTL regression analysis [QTL express (Seaton et al. 2002)] in the line cross model assuming alternatively fixed QTL alleles in the founder breeds. The significance thresholds were determined by 10,000 permutations and 95% confidence intervals by 10,000 bootstraps. Significantly segregating F1 individuals were determined by performing within-F1 parental half-sib analyses for all F1 sires and dams exceeding 20 offspring in a regression QTL model (QTL express) and permutation tests. All families exceeding a 5% chromosomewise threshold were considered as segregating at the QTL.
In addition to the QTL regression analysis, a variance component QTL model as implemented in Qxpak (Perez-Enciso and Misztal 2004) including an infinitesimal animal effect was calculated,where y is a vector of phenotypes, u represents a vector of individual infinitesimal polygenic effects, Z and Q represent the incidence matrices for the polygenic and the QTL effect, respectively, g is a vector of QTL effects (either random or fixed effects within lines depending on the model applied), and e is the vector of random residuals. Identity-by-descent probabilities were calculated on the basis of a MCMC algorithm implemented in Qxpak (Perez-Enciso et al. 2000). A likelihood-ratio test (LRT) was calculated to test for statistical significance. Significance thresholds for the LRT were determined according to Nezer et al. (2002), considering one chromosome with a length of 1.2 M and an average marker density for the variance component analyses of 0.02927 M. LRT significance thresholds for α = 0.05 and α = 0.0001 were 10.5 and 24.
The variance component approach used to model a random QTL effect (Perez-Enciso and Misztal 2004) further complemented the initial QTL analysis with the more likely relaxed hypothesis regarding QTL allele distribution in the founder breeds. For the variance component QTL model, only markers with >200 informative meioses in the F2 design were included, because it was indicated that the random QTL model may be instable, when information gaps occur in the genotype file. The confidence interval for the QTL investigated by the variance component model was determined by the LOD drop-off method as suggested by Lander and Botstein (1989), with one LOD interval corresponding to a 96% confidence interval (Mangin et al. 1994).
For the two-QTL analyses we used a variance component method, because this enabled modeling of the QTL effects as random or fixed. For the two-QTL analyses, we assumed a fixed QTL effect at a specific position at the maximum in the one-QTL test statistic and tested in 1-cM intervals across the chromosome, whether there was a second QTL position for which the resulting two-QTL model had a statistically significant increased likelihood compared to the one-QTL model with the fixed QTL effect. The second QTL was modeled with a fixed or a random QTL effect, respectively. LRT tests were performed to assess the statistical significance of the two-QTL hypothesis (LRT model with two QTL vs. model with one QTL).
Association linkage disequilibrium (LD) studies were performed to fine map the QTL in the initial QTL confidence interval 44–56 cM, applying the following model as implemented in the LD association analysis in Qxpak,where yi is the record of individual i, λihk is an indicator variable, which is 1 when the allele at the hth haplotype (1 or 2) of the ith individual is k and otherwise 0, gk is the allelic effect of allele k, ui is the infinitesimal genetic effect of individual i, and eihk is the residual. Multiple testing of a total of 52 markers had to be accounted for to avoid false positive associations. To avoid false positive associations due to multiple testing of 52 markers, a Bonferroni correction was calculated, a most stringent procedure given the fact that there was linkage and also linkage disequilibrium between markers. Those SNPs with the most significant effect were tested in an extended two-marker model, whether they still showed a significant association after accounting for the effects of one of the other significantly associated SNPs,where yi is the record of individual I, λihk is an indicator variable for locus 1, which is 1 when the allele at the hth haplotype (1 or 2) of the ith individual is k and otherwise 0, λimn is a respective indicator variable for locus 2, and gk and gm are the respective allelic effects for locus 1 and locus 2. A likelihood-ratio test (likelihood of model with locus 1 and locus 2 vs. likelihood of model with locus 1) was calculated to test for statistical significance.
Furthermore, SNPs that were significant in the association analyses were tested, if they had an effect on the QTL test statistic, when included in the fixed QTL model as implemented in Qxpak,where yi is the record of individual i, λihk is an indicator variable, which is 1 when the allele at the hth haplotype (1 or 2) of the ith individual is k and otherwise 0, gk is the allelic effect of allele k, βi is the additive QTL coefficient, and a is the additive QTL effect.
The most likely haplotypes of the Qq segregating F1 individuals were taken from the MCMC algorithm (see above) and assigned to the respective QTL alleles Q and q according to the effect estimation from the half-sib regression model. For this purpose, we selected offspring with extremely different transmission probabilities at position 50 cM for the parental target chromosome as calculated in the QTL Express analysis. Then, we determined from the flanking parental marker alleles, which of the two parental haplotypes the respective offspring inherited at the respective position. The haplotype, for which a positive effect on fetal growth was calculated in the regression analysis, was denoted “Q,” and the haplotype, for which a negative effect was calculated, was denoted “q.” All Q-carrying and all q-carrying haplotypes, respectively, were aligned and tested for concordance between QTL status and the candidate polymorphisms. The allele frequencies for the NCAPG c.1326T>G locus were assumed according to the respective allele count for the alleles in the founder populations (NCAPG c.1326G, 0.9 and NCAPG c.1326T, 0.1 for Charolais and NCAPG c.1326G, 0.2 and NCAPG c.1326T, 0.8 for German Holsteins; expected allele frequency in the F1 population, 0.55 for allele NCAPG c.1326G and 0.45 for allele NCAPG c.1326T). Following Seroussi (2009), the likelihood of the observed concordance would be pGNQ × pTNq, where pG is the allele frequency of the allele NCAPG c.1326G in the F1 population, pT is the allele frequency of the allele NCAPG c.1326T in the F1 population, and NQ and Nq are the number of concordant chromosomes carrying the Q or the q QTL allele, respectively. Furthermore, the concordance analysis in the significantly segregating F1 parents was complemented by analysis of individuals very likely homozygous QQ or qq, respectively, at the QTL for fetal growth. For this purpose, the random QTL effect at position 55 cM as well as the fixed QTL effect at position 50 cM for each individual was taken from the two-QTL model. The 10 individuals with the highest and lowest QTL effects were assumed QQ or qq at the respective QTL positions. Binomial tests were performed to test whether there was a significant concordance of divergent marker alleles in the two groups,where is the probability that allele 1 is observed at least a times in the sample of n Q-carrying chromosomes and allele 2 is observed at least b times in the n q-carrying chromosomes. Due to 10 individuals tested in each group, nQ and nq = 20; allele 1 is the marker allele concordant with the Q allele in the significantly segregating F1 parents, allele 2 is the marker allele concordant with the q allele in the significantly segregating F1 parents, a is the number of chromosomes carrying marker allele 1, and b is the number of chromosomes carrying the marker allele 2. The allele frequencies of the two different marker alleles are denoted p1 and p2 for alleles 1 and 2, respectively.
Resequencing and refined haplotypes:
To obtain additional information on the haplotype structure and linkage disequilibrium in the NCAPG region we resequenced 4916 bp comprising the region proximal to the NCAPG gene and several regions from the NCAPG and the LCORL gene. Both genes are located in the confidence interval of this study and are also highlighted in a QTL analysis on calving difficulty in another cattle breed and in whole-genome association studies addressing human height (Olsen et al. 2008; Weedon et al. 2008). Information on PCR primers is indicated in Table S2. Genomic PCR products were generated and subsequently sequenced with the DYEnamic ET Terminator Cycle Sequencing reaction and the MegaBACE 1000 DNA Analysis System (GE Healthcare). For sequencing, PCR primers were used. Targets for resequencing were the four F1 individuals, which showed statistically significant segregation of the QTL (probably Qq at the QTL), and the 20 individuals with the highest and lowest estimated QTL effects (supposed to be QQ and qq, respectively) selected as described in the previous paragraph. To obtain further information on the NCAPG c.1326T>G mutation history, we additionally included 6 unrelated individuals from the Bos indicus subspecies for genotyping by resequencing in the vicinity of the NCAPG c.1326T>G locus.
Fetal NCAPG expression pattern:
Fetal expression patterns of the NCAPG gene were analyzed by reverse transcription (RT)–PCR on a multitissue panel consisting of six fetal bovine tissues (skeletal muscle, kidney, liver, lung, brain, and placenta, all tissues from 123 days of pregnancy). Principally, total RNA was extracted using the Nucleospin II RNA kit (Macherey–Nagel). For muscle tissue, an initial purification step using Trizol (Invitrogen) was added. Genomic DNA was carefully eliminated from RNA preparations by repeated on-column RNAse-free DNaseI digestion. The cDNAs were synthesized using the Superscript III RNase H-reverse transcriptase (Invitrogen) with oligo(dT) primer according to the manufacturer's instructions. Expression of the NCAPG gene was analyzed using gene-specific primers (forward primer, 5′-TGGAGCAAGTAAAAGATAGAAC-3′; reverse primer, 5′-ACCACTTTTCTCCGTC CTCTG-3′), producing a DNA fragment of 244 bp.
Quantitative NCAPG expression analysis:
Quantitative NCAPG gene expression was analyzed by quantitative real-time PCR (qRT–PCR), using the LightCycler 480 system (Roche Diagnostics), on total RNA from fetal placenta taken at days 120–128 of gestation of six purebred German Holstein fetuses with different genotypes at the NCAPG c.1326T>G polymorphism (three with genotype NCAPG c.1326T/ NCAPG c.1326T, two with genotype NCAPG c.1326T/ NCAPG c.1326G, and one with the rare homozygous NCAPG c.1326G/ NCAPG c.G genotype). The fetuses were all paternal half-sibs generated by ordinary artificial insemination to lactating first parity cows. The RNA samples (1 μg) were reverse transcribed to cDNAs, using Superscript III RNase H-reverse transcriptase (Invitrogen) and the gene-specific primer 5′-CTCCATCATTTCTTCTCCTAAC-3′. The cDNA samples equivalent to 25 ng total RNA in the reverse transcription assay were subjected to qRT–PCR in duplicate, using the LightCycler 480 SYBR Green I Master kit (Roche Diagnostics) and gene-specific exon-spanning primers (forward primer, 5′-CTGGAGCAAGTAAAAGATAGAAC-3′; reverse primer, 5′-ACCACTTTTCTCC GTCCTCTG-3′) generating an amplicon of 245 bp. To generate a standard calibration curve a standard transcript was prepared from the 245-bp PCR amplicon and purified with the Nucleospin Extract kit II (Macherey–Nagel). The quantity of the NCAPG mRNA in the analyzed cDNA samples of fetal placenta was calculated from the standard calibration curve on the basis of a serial dilution of the standard amplicon covering four orders of magnitude (103–106 copies), which ran simultaneously with the samples.
Statistical analysis of mRNA abundance was performed by regressing the amount of transcript against the number of copies of the NCAPG c.1326T allele,where yi is the number of NCAPG transcripts in the equivalent of 25 ng fetal placental RNA of fetus i, b is the regression coefficient of the NCAPG c.1326T allele, NCAPGi is the number of copies of the NCAPG c.1326T allele of individual i, and ei is the residual effect.
Detection of a QTL affecting fetal growth:
Initially, a line cross QTL regression model including 16 microsatellite markers revealed a highly significant QTL (F value 90.16, P <0 .0001) affecting fetal growth in a 12-cM 95% confidence interval (44–56 cM) on BTA6 (Figure 1). The QTL accounted for 16.3% of the variance in the model as calculated as the relative reduction of the residual variance due to including the QTL in the model (Knott et al. 1996). The QTL seems to act in a strictly additive mode, because we did not find any evidence for dominance or imprinting effects. The QTL allele assumed to be originating from the Charolais breed was estimated to increase birth weight by 3.78 kg (SE 0.40). Whereas the QTL region exhibited a substantial effect on prenatal growth, the test statistic dropped substantially for the trait “weight at 121 days of age” and did not reveal any significant QTL for “average daily gain in the time period birth to day 121.” In consecutive half-sib regression analyses, four of eight F1 parents tested (three sires, one dam) proved to be significantly heterozygous at the QTL affecting fetal birth with the peaks of the test statistics between 48 and 52 cM. The QTL test statistic for the other four F1 half-sib families with >20 offspring did not exceed the statistical significance threshold. However, it has to be considered that those F1 families all had a lower number of offspring than those four F1 families with significant segregation of the QTL.
Association study of targeted SNPs with fetal and early postnatal growth:
Targeted candidate SNPs and control SNPs were used to refine the QTL for fetal growth in association analyses. In addition to the initially genotyped 16 microsatellites, we tested 23 SNPs (see Table S1) in and adjacent to the initial QTL confidence interval for their LD association with fetal growth. Whereas 16 loci were nominally significant, adjustment for multiple testing maintained 7 SNPs and 1 microsatellite showing a 5% experimentwise significance (Figure 2). The SNP showing the strongest association was NCAPG c.1326T>G with a nominal P-value <10−16. Because SPP1 c.621+338A>G is located adjacent to NCAPG c.1326T>G and displays LD to NCAPG c.1326T>G, the observed trait association (P < 10−10) of SPP1 c.621+338A>G was to be expected. Interestingly, there is a cluster of markers (rs29025709–DIK82) located 8 cM telomeric of NCAPG c.1326T>G comprising four markers that all also show a significant association with fetal growth, although three of them, e.g., DIK82, are not contained in the initial 95% confidence interval 44–56 cM. To test if the marker associations in the cluster SEPSECS–DIK82 were due to linkage disequilibrium generated by the pedigree relationships in our data set, five markers (the four markers with the highest significance from the one-marker association and also PPARGC1A c.1892+19T>C) were separately tested in a two-marker model for an association additional to the effect of one of the four other markers. We found that associations for SPP1 c.621+338A>G and STIM2 (rs29014935), but not for DIK82, vanished, when tested jointly with NCAPG c.1326T>G (Figure 2). Taking SPP1 c.621+338A>G into the two-marker model still yielded significant effects for NCAPG c.1326T>G, STIM2 (rs29014935), and DIK82. Joint analysis with DIK82 in a two-marker model showed no statistically significant effect for SPP1 c.621+338A>G, but STIM2 (rs29014935) and NCAPG c.1326T>G were still significantly associated. These results indicated that there might be more than one locus affecting fetal growth in the targeted chromosomal region of BTA6.
Combined linkage and association study:
In a second QTL mapping step, we included the additional SNPs into the QTL analysis to increase the mapping resolution and obtained a QTL with fixed effect at position 50 cM (LRT = 92.7) (Figure 3). Its 1-LOD drop-off interval spanned from 47 to 51 cM. The extension of the QTL model including one of the four markers with the most significant association to fetal growth tested whether one of those markers accounts for the QTL with fixed effect in the region 47–51 cM. The drop in the QTL test statistic was most prominent for NCAPG c.1326T>G and shifted the peak of the test statistic telomeric toward 58 cM (Figure 3). Marker DIK82 also resulted in a marked decrease of the test statistic when included in the model and, in combination with NCAPG c.1326T>G, resulted in an almost flat test statistic with a maximum of 1.74 in the region 40–65 cM. Thus, the combined LD information from NCAPG c.1326T>G and DIK82 completely erases the QTL with fixed effect in the region 44–56 cM. In contrast, SPP1 c.621+338A>G did not further decrease the test statistic when jointly included with NCAPG c.1326T>G in the QTL model (data not shown). STIM2 (rs29016265), like SPP1 c.621+338A>G, resulted in a substantial decrease in the test statistic, although the maximum of the test statistic still showed a statistically significant QTL (data not shown).
Initial data from association mapping suggested that two QTL affecting fetal growth could be segregating in the middle part of BTA6. This was supported by results from a QTL analysis assuming a QTL with random effect, which indicated a highly significant QTL (Figure 4) with a confidence interval spanning the region 55–57 cM. Thus, the confidence intervals obtained from the fixed effect and the random QTL effect analyses did not overlap. Although this incongruence obtained by the LOD drop-off methods should not be overinterpreted, it justified the consecutive two-QTL analyses. When the second QTL was modeled with a line-specific fixed QTL effect in addition to a first QTL with a line-specific fixed effect at 50 cM, no significant second QTL was observed (Figure 4). However, when the second QTL was modeled as random in addition to a first QTL with a line-specific fixed effect at 50 cM, the test statistic showed a highly significant QTL (LRT two QTL vs. one QTL = 70.4) in the interval 54–56 cM and further peaks at positions 58, 63, and 72 cM. The confidence interval for the second QTL with presumably random QTL effects could be narrowed down to 2 cM. Further extending the model to include three different QTL did not provide any significant proof that a third QTL is segregating in the interval 40–65 cM.
To obtain further information on the position and the potential causal background of the QTL detected, we performed concordance analyses of the parental haplotypes from those F1 parents that were significantly segregating Qq for the fetal growth QTL. All four parental F1 chromosomes carrying the Q allele shared the allele NCAPG c.1326G, whereas all F1 chromosomes carrying the q allele shared the allele NCAPG c.1326T (Figure 5). The likelihood of this concordance being observed by chance is 0.36 × 10−3. It has to be highlighted that all four segregating F1 parents are unrelated. The other four F1 parents with >20 offspring that did not reach the chromosomewise statistical significance threshold in the QTL regression analysis all shared a NCAPG c.1326G allele at the parental haplotype associated with increased fetal growth. Additionally, they also shared a NCAPG c.1326T allele at the parental haplotype associated with decreased fetal growth. Thus, the NCAPG c.1326T>G–QTL haplotypes in the not significantly segregating F1 parents showed the same concordance as the respective haplotypes in those F1 parents, for which we had obtained statistically significant segregation. Looking at all P0 individuals, we found that no NCAPG c.1326T>G allele is fixed either in Charolais or in German Holsteins, although the allele frequencies in Charolais and German Holstein for NCAPG c.1326T>G differed substantially. This is in contrast to SPP1 c.621+338A>G, which also displayed a discordant distribution of alleles on chromosomes carrying alternative QTL alleles. As outlined in the previous sections, however, QTL and association studies showed that NCAPG c.1326T>G erased the effects of SPP1 c.621+338A>G when included into the respective analyses, while this was not observed vice versa.
To further extend the concordance analyses, we obtained the putative QTL effects for the fixed effect and the random effect QTL at 50 and 55 cM, respectively, from the two-QTL analyses. From the 10 F2 individuals with the lowest fixed QTL effects (presumably qq) at position 50 cM, 9 had a homozygous NCAPG c.1326T/T and 1 had a heterozygous NCAPG c.1326T/G genotype. Eight of the 10 F2 individuals with the highest fixed QTL effects (presumably QQ) at position 50 cM had a homozygous NCAPG c.1326G/G and 2 had a heterozygous NCAPG c.1326T/G genotype. No other locus including SPP1 c.621+338A>G showed a similarly biased distribution of genotypes. Statistical analysis revealed that the likelihood to obtain the NCAPG c.1326T>G distribution of alleles by chance is 2.31 × 10−9. For the QTL with a random effect at position 55 cM, no analogously discordant allele distribution was observed for any marker.
Finally, backcross calves generated from mating a German Holstein sire to all female F2's for their first and second parity were recruited for an additional paternal half-sib regression QTL analysis. It showed a statistically significant QTL for fetal growth (F = 10.43, P < 0.05) within the family, indicating that the German Holstein sire should be Qq at a QTL affecting fetal growth. Further genotyping revealed that the sire was heterozygous NCAPG c.1326T/G, with the allele NCAPG c.1326G on the paternal haplotype increasing birth weight.
Resequencing and refined haplotypes:
In the chromosomal segments that were included in resequencing, we found 26 polymorphisms in the F1 and F2 individuals (see Table S3). Alignment of haplotypes for the Qq individuals and the supposedly QQ and qq individuals, respectively, demonstrated that there is a large common haplotype block for the Q-carrying chromosomes encompassing the NCAPG c.1326G allele. This haplotype block spans at least 193 kb. This block is shared by all Q-carrying F1 chromosomes (see Figure 5) and by 18 of 20 supposedly Q-carrying F2 chromosomes. In contrast, no continuous shared identity-by-state haplotype block could be identified across all the q-carrying F1 chromosomes. We observed different haplotypes for all four q-carrying F1 chromosomes and 10 different haplotypes in the supposedly q-carrying F2 chromosomes. None of the polymorphisms, except NCAPG c.1326T>G, which were heterozygous in all segregating F1 parents and that showed a concordance with the QTL alleles, affected the amino acid sequence of the encoded protein or are located in a prominent regulatory region of the target gene. Resequencing in six unrelated B. indicus individuals showed that all were homozygous for the NCAPG c.1326T allele (see Table S3).
Expression analysis in a multitissue panel demonstrated NCAPG mRNA expression in all fetal tissues investigated including skeletal muscle, kidney, liver, lung, and brain and also in the fetal and maternal part of the placentomes (Figure 6A). Quantitative real-time analysis in the fetal part of placentomes originating from paternal half-sib fetuses with different NCAPG c.1326T>G genotypes suggested a differential NCAPG gene expression (Figure 6B). Regression of the number of NCAPG transcripts on the number of NCAPG c.1326T alleles showed an allele substitution effect of 11,675 copies (P = 0.09), suggesting that the NCAPG c.1326T allele is associated with increased NCAPG expression level.
In our study, we provide evidence for a QTL with highly significant effect on fetal growth in a unique cattle resource population. Systematic use of foster mothers and embryo transfer clearly breaks down any genetically determined correlation between biological mother and fetus and enables a stringent dissection of genetic factors modulating exclusively the fetal component of fetal growth.
Confirmation of a locus affecting fetal growth on BTA6:
The QTL position affecting fetal growth on BTA6 determined in our study matches the suggested position of former studies on fetal growth and correlated traits very well (Casas et al. 2000; Kühn et al. 2003; Kneeland et al. 2004; Holmberg and Andersson-Eklund 2006). Recently, Gutierrez-Gil et al. (2009) mapped a highly significant, additively acting QTL for birth weight and body length at birth on BTA6 in a cross from Charolais and Holsteins explaining 27% of the variance with the Charolais originating QTL allele associated with increased birth weight. The authors reported a confidence interval from position 27 to 65 cM, which includes the NCAPG gene and is very similar to the confidence interval from another recent QTL analysis on bovine fetal growth in an experimental dairy cattle population (Maltecca et al. 2009).
Collecting evidence for a causal role of the NCAPG locus on fetal growth:
The final aim of studies dissecting the genetic background of complex traits is the detection of genetic variants that might be of causal effect. Whereas causality in livestock species is difficult to obtain, several lines of evidence can be collected to support any assumption on causal variants (Glazier et al. 2002; Rebbeck et al. 2004). The results described in this study merged with several other reports from the literature provide substantial arguments in favor of a functional role of the NCAGP locus on fetal growth:
Coincidence of QTL position and position of trait-associated NCAPG c.1326T>G SNP: Within the confidence interval of the highly significant QTL affecting fetal growth, we identified the NCAPG c.1326T>G locus displaying a highly significant association with the phenotype.
Concordance between QTL and NCAPG c.1326T>G alleles: There is a significant concordance of QTL alleles and NCAPG c.1326T>G alleles in segregating F1 individuals and in the extreme F2 individuals with presumably QQ and qq QTL genotypes. As outlined by Seroussi (2009), full haplotype concordance of a representative sample of chromosomes is likely to point to the causative gene. The concordance between QTL alleles and NCAPG c.1326T>G alleles is further confirmed by an analogous concordance also in those F1 parents that did not reach the formal statistical significance threshold in the within-family QTL analyses. Lack of full concordance between the assumed QTL genotype and the NCAPG c.1326T>G genotype in the F2 individuals might be due to the effects of the second QTL affecting fetal growth in close proximity to the NCAPG gene.
NCAPG c.1326T>G explained a major proportion of the QTL: The QTL model including NCAPG c.1326T>G decreased the QTL test statistic at the position 50 cM below the significance threshold. In combination with locus DIK82, NCAPG c.1326T>G completely erases the QTL effect in the fixed QTL model.
Concordant segregating QTL–NCAPG c.1326T>G haplotypes within breed: Within the German Holstein population, we observed a purebred sire significantly segregating for a QTL affecting fetal growth. This sire also showed a concordance between the QTL allele and the NCAPG c.1326T>G allele identical to that of the F1 parents of the resource population.
NCAPG c.1326T>G–QTL linkage disequilibrium conserved across distantly related breeds: Setoguchi and colleagues (Setoguchi et al. 2009) provided strong evidence for a highly significant association of the NCAPG c.1326T>G locus with growth in two unrelated Japanese cattle breeds. These populations have entirely different historical and geographical origin compared to each other (Minezawa 2003) and also compared to the European Charolais and German Holstein breeds used in our study (McKay et al. 2008). Conservation of linkage disequilibrium between alleles of the NCAPG c.1326T>G locus and QTL alleles affecting bovine growth across such divergent populations provides further support for a functional role of the NCAPG locus.
The NCAPG c.1326T allele associated with low fetal growth is presumably the ancient allele: The large haplotype block shared by the Q-carrying haplotypes compared to the lack of continuous haplotype block in the q-carrying haplotypes demonstrates that the q allele should be the ancient variant. This fits into the results from resequencing B. indicus individuals, where only the NCAPG c.1326T allele was found. Thus, during evolution presumably the NCAPG c.1326T allele had already been present, when the B. taurus and B. indicus subspecies separated. NCAPG c.1326G has been demonstrated to be associated with increased fetal growth in our population and should be associated with increased calving problems. Thus, in natural populations the mutation would be particularly disadvantageous due to lack of assistance during delivery of the calf.
Fine mapping in other breeds confirmed positional candidacy of NCAPG: Further positional indication advocating NCAPG as a background for divergent fetal growth comes from a recent fine-mapping study in yet another cattle breed. Olsen et al. (2008) identified a 150-kb candidate QTL interval for calving difficulty, which is highly correlated to fetal growth, 20 kb adjacent to the NCAPG gene in Norwegian Red dairy cattle.
Across-species support for positional candidacy of NCAPG: Comparative data provide further support for the hypothesis that the NCAPG locus will affect mammalian growth. Concurrently, the NCAPG/LCORL region was one of only seven loci, which showed a highly significant effect on human height in two powerful major genomewide association metaanalyses in humans (Gudbjartsson et al. 2008; Weedon et al. 2008).
Indication of a functional role of NCAPG in fetal development: The lines of support for an effect of the NCAPG locus on (fetal) growth accumulating from genetic studies are finally complemented by physiological data. NCAPG is a non-structural maintenance of chromosomes (SMC) subunit, which has a catalytic function in the mammalian condensin I complex and is important during mitotic cell division (Dej et al. 2004). The NCAPG c.1326T>G SNP is a nonsynonymous mutation in exon 9 of the NCAPG gene generating the replacement of isoleucine by methionine at amino acid position 442 of the NCAPG protein. Comparison to other species supports the hypothesis that NCAPG c.1326T indeed is the ancient variant (Figure 7). The respective part of the NCAPG protein is highly conserved in mammals compared to nonmammalian species and is part of a HEAT repeat (Protsweep, http://genius.embnet.dkfz-heidelberg.de/menu/biounit/open-husar) (Del Val et al. 2007), which is suggested to play important roles in chromosome dynamics (Neuwald and Hirano 2000). Furthermore, NCAPG interacts with DNA methyltransferase DNMT3B and thus links regulation of DNA methylation and mitotic chromosome condensation in mammalian cells (Geiman et al. 2004). Our own results demonstrated that NCAPG is indeed expressed in a variety of fetal tissues including the cotyledons of the placentome, which is a prerequisite for NCAPG modulating the fetal component of fetal growth.
Suggestive differential NCAPG expression in fetuses with different NCAPG c.1326T>G genotype: Our results, although not statistically significant, suggest a decreased NCAPG expression level associated with the NCAPG c.1326G allele, which increases fetal growth. A reduced NCAPG mRNA expression associated with the NCAPG c.1326G is in good agreement with transcriptome analyses in placentomes from cloned fetuses: NCAPG is one of the genes most differentially expressed in the placentomes of embryos originating from somatic nuclear cloning vs. placentomes originating from embryos generated by conventional artificial insemination (Everts et al. 2008). Somatic cloning of bovine embryos is known to frequently result in extremely large offspring and significant abnormalities of the placenta (Hill et al. 1999; Constant et al. 2006). Thus, our expression data add a first functional indication that the NCAPG gene may indeed represent a causal gene affecting fetal growth.
Exclusion of other functional and positional candidate SNPs and genes:
Two prominent mutations (ABCG2c.1743A>G and OPN3907), in genes [ATP-binding cassette, sub-family G (WHITE), member 2 (ABCG2), and secreted phosphoprotein 1 (SPP1)] that closely neighbor NCAPG and are located within or near the QTL confidence interval ranging from 47 to 51 cM are controversially discussed as a causal background for genetic variation in milk composition (Cohen-Zinder et al. 2005; Schnabel et al. 2005; De Koning 2006). Furthermore, Allan et al. (2007) described a significant association of the OPN3907 polymorphism with birth weight in an experimental cattle population. However, genotypes of the four F1 parents significantly segregating for the QTL affecting fetal growth in our study revealed that all individuals are homozygous for both polymorphisms. Thus, both polymorphisms can be excluded as a cause for the QTL investigated in this study.
Gutierrez-Gil et al. (2009) pointed out that the QTL affecting fetal growth they described in the interval 27–65 cM on BTA6 seems to primarily affect bone growth. This would support homology of the bovine QTL on BTA6 to the NCAPG/LCORL locus on human chromosome 4 (HSA4) affecting height (Gudbjartsson et al. 2008; Weedon et al. 2008), which is also determined by bone growth. Interestingly, there is an evolutionary conserved chromosomal breakpoint between BTA6 genes leucine aminopeptidase 3 (LAP3) and integrin-binding sialoprotein (IBSP) (Weikard et al. 2006). While on BTA6 (NCBI Build 4.0) both loci are located immediately next to each other in close vicinity to NCAPG, LAP3 was mapped near NCAPG on HSA4, but IBSP is located very distantly from LAP3 at the telomeric end of the chromosome (NCBI build 36.3). Thus, if indeed there is a species-spanning homologous locus affecting mammalian bone growth, IBSP as well as secreted phosphoprotein 1 (SPP1, also known as osteopontin) and matrix extracellular phosphoglycoprotein (MEPE), all well described for their effects on bone metabolism (Beck et al. 2000; Nampei et al. 2004; Karadag and Fisher 2006), but located centromeric of IBSP on BTA6, should be excluded as candidate genes for this locus. Our results demonstrate that comparative QTL information originating from different species can help to refine the genetic background of a complex trait in the target species as suggested by Burgess-Herbert et al. (2008).
Resequencing in the vicinity of the NCAPG c.1326T>G revealed a large number of variants within the NCAPG and the neighboring LCORL gene. However, none showed both concordance with the QTL alleles and a potential functional relevance. It has to be granted that in spite of many positional and functional indicators in favor of NCAPG c.1326T>G, it still cannot formally be excluded that variants other than NCAPG c.1326T>G might represent the causal background of the QTL affecting fetal growth. Next-generation sequencing (Shendure and Ji 2008) of the entire genomic region in a well-characterized data set from divergent bovine populations might add further evidence in this respect. This will have to be complemented by refined functional in vivo and in vitro analyses in specific cell models.
Indication of a second locus affecting fetal growth:
Association analyses combined with QTL/association analyses and also one-QTL mapping suggested a second QTL with effect on fetal growth distinct from the NCAPG locus. The variance component model provided statistically significant detection of a second QTL, but only if this QTL was modeled with random effects. In contrast, the two-QTL model was not statistically more likely than a one-QTL model, when a fixed effect was assumed for the second QTL. Interestingly, the test statistics provided by Setoguchi et al. (2009) also give an indication that a second QTL might be segregating in the Japanese Brown population at the position 60 cM. Whereas for the locus at 50 cM, there is a mounting indication that the causal mutation might be located within or near the NCAPG gene, the genetic background of the putative second QTL at position 55 cM remains open. Although peroxisome proliferator-activated receptor gamma, coactivator 1 alpha (PPARGC1A) and stromal interaction molecule 2 (STIM2) are candidate genes for fetal growth due to their function in energy metabolism and embryonic development (Puigserver and Spiegelman 2003; Eid et al. 2008), our data do not convincingly support a causal role of either gene regarding genetic modulation of fetal growth. Further research is necessary to elucidate the background of the second QTL affecting fetal growth. However, our results together with a wealth of other data on lactation suggest that the middle part of BTA6 seems to represent a true hotspot of genetic variation affecting growth and lactation traits.
While the QTL at positions 50 and 55 cM obviously exert strong effects on fetal growth, we could not detect effects on growth in the consecutive life period until day 121. Hadjipavlou and Bishop (2009) described that QTL for growth might act at distinct times of life and may have no effect at other times. Analogously, we found an indication of NCAPG effects only in later life in our resource population (data not shown). Thus, a comprehensive analysis of the NCAPG locus regarding effects on growth and associated metabolic traits will shed further light on NCAPG c.1326T>G effects in postnatal life. This knowledge will be a prerequisite for applying NCAPG c.1326T>G in selection programs. Our results demonstrate that in spite of probably large differences in allele frequencies, the causal QTL alleles affecting fetal growth are still segregating within the Charolais and German Holsteins representing beef and dairy cattle breeds, respectively. Thus, both breeds may benefit from a targeted selection on NCAPG c.1326T>G.
Our QTL and association studies showed that the causal mutation underlying the identified locus modulating fetal growth has a major effect. The substantial size of the effect will facilitate the further functional investigation of NCAPG for its role in affecting growth during normal and aberrant fetal development in conventional reproduction or in assisted reproduction techniques.
We thank Jill Maddox (University of Melbourne, Australia) for providing the modified CRIMAP Version 2.50. Also, we thank our colleagues at the FBN Dummerstorf involved in the generation and care of the SEGFAM F2 resource population for their continuous support of our work. We thank the German Federal Ministry of Education and Research for the financial support of this work within the scope of the FUGATO QUALIPID project (FKZ 0313391C).
Supporting information is available online at http://www.genetics.org/cgi/content/full/genetics.109.106476/DC1.
Communicating editor: T. R. Magnuson
- Received June 20, 2009.
- Accepted August 21, 2009.
- Copyright © 2009 by the Genetics Society of America