Identification and prioritization of NUAK1 and PPP1CC as positional candidate loci for skeletal muscle strength phenotypes

Abstract

Muscle strength is an important determinant in elite sports performance as well as in the activities of daily living. Muscle metabolism also plays a role in the genesis, and therefore prevention, of common pathological conditions and chronic diseases. Even though heritability estimates between 31 and 78% suggest a significant genetic component in muscle strength, only a limited number of genes influencing muscle strength have been identified. This study aimed to identify and prioritize positional candidate genes within a skeletal muscle strength quantitative trait locus on chromosome 12q22-23 for follow-up. A two-staged gene-centered fine-mapping approach using 122 single nucleotide polymorphisms (SNPs) in stage 1 identified a familybased association (n = 500) between several tagSNPs located in the ATPase, Ca²⁺ transporting, cardiac muscle, slow twitch 2 (ATP2A2; rs3026468), the NUAK family, SNF1-like kinase, 1 (NUAK1; rs10861553 and rs3741886), and the protein phosphatase 1, catalytic subunit, gamma isoform (PPP1CC; rs1050587 and rs7901769) genes and knee torque production (P values up to 0.00092). In stage 2, family-based association tests on additional putatively functional SNPs (e.g., exonic SNPs, SNPs in transcription factor binding sites or in conserved regions) in an enlarged sample (n = 536; 464 individuals overlap with stage 1) did not identify additional associations with muscle strength characteristics. Further in-depth analyses will be necessary to elucidate the exact role of ATP2A2, PPP1CC, and NUAK1 in muscle strength and to find out which functional polymorphisms are at the base of the interindividual strength differences.

muscular fitness is an important factor in the elite sports performance as well as in the activities of daily living (31). Muscle metabolism also plays an important role in the genesis, and therefore prevention, of many common pathological conditions and chronic diseases like cardiovascular disease (41, 42), osteoporosis (27), metabolic syndrome (58), obesity (16), and diabetes (42, 62). Given the fact that the Western society is aging, the progressive loss of muscle mass and function that occurs with aging (sarcopenia) will become a major health issue as it has devastating effects on quality of life (e.g., institutionalization, dependence of the elderly) and eventually on survival (35).

Several studies suggest that apart from environmental influences such as training, social status, or nutrition, a large genetic component is involved in the interindividual variation in muscle strength. Heritability estimates range between 31 and 78% and differ between muscle groups, contraction velocities, and contraction angles (25, 33, 44, 49–51, 53). Despite this significant genetic determination of muscle strength, only a limited number of studies showed significant association of genetic variants in quantitative trait loci (QTL) with muscle strength phenotypes, and even fewer associations have consistently been replicated (reviewed in Ref. 11).

Studies investigating genetic linkage with human muscle strength measures are even scarcer. To date only Tiainen et al. (52) and our research group (14, 15, 23, 24, 26, 48) have reported genetic linkage results on skeletal muscle strength characteristics. Our research group was the first to perform linkage analyses on a unique collection of young male Caucasian siblings drawn from the Leuven Genes for Muscular Strength study (LGfMS). These analyses focused on chromosomal regions harboring genes involved in the myostatin signaling pathway, as described elsewhere (23, 26). Single and multipoint microsatellite marker-based linkage analyses revealed that the chromosomal regions 12q12–14 [logarithm of the odds (LOD) 3.4, P = 0.00004], 12q22-23 (LOD 2.6, P = 0.0002) and 13q14.2 (LOD 2.74, P = 0.0002) showed significant or suggestive linkage to knee muscle strength (23, 26). It can, however, not be excluded that genes other than the original myostatin pathway candidate genes cause these linkage findings and therefore further fine mapping of these regions is warranted. In a recent report we followed up upon the chr12q12-14 region and identified the activin receptor 1 B (ACVR1B) gene as a strength determining gene using a comprehensive two-staged fine-mapping strategy (60).

Here we present the results of a similar two-staged fine mapping of the 12q22-23 region containing 143 known genes. The purpose of this study was to prioritize these genes and identify genes and polymorphisms that are associated with knee muscle strength characteristics. Therefore the study focused on 51 prioritized candidate genes and identified the NUAK1 and ATP2A2/PPP1CC loci to be related to variation in muscle strength.

MATERIAL AND METHODS

Study Design

A two-staged approach was designed to fine map the chr12q22-23 linkage region, as shown in Fig. 1.

Fig. 1.

Study design. The 2 stages of the study are displayed. Sample description, genotyping details, and statistical analyses are presented from top to bottom. LGfMS, Leuven Genes for Muscular Strength study; SNP, single nucleotide polymorphism; QTDT, quantitative transmission disequilibrium testing; FBAT, family-based association test; HBAT, haplotype-based family-based association test.

Fine mapping stage 1.

Single nucleotide polymorphisms (SNPs) were selected under the 1-LOD confidence interval of a previously defined multipoint microsatellite-based linkage peak (26). Thereto, an empirical two-step fine-mapping strategy, in which candidate genes are prioritized using a bioinformatics approach and the top genes are chosen for further SNP selection with a linkage disequilibrium-based method, was applied. A detailed description of this method and its application to another linkage region can be found elsewhere (60, 61). In brief, the 143 genes within the 1-LOD confidence interval were ranked according to their similarity to genes known to be involved in (the regulation of) muscle strength using Endeavour software (3). These known genes include candidate genes from the original myostatin pathway (23, 26), structural elements of muscle (actin and myosin related genes, together with troponin, titin, and nebulin), and genes identified based on Gene Ontology (GO) terms “contraction,” “muscle development,” and “regulation (negative and positive) of contraction.” Rankings were made for each of these five reference sets and combined into one global ranking.

We considered the top 20% of the genes according to every reference set for further analyses as well as those genes ranked within the top 20% in the global ranking. Of these, genes without evidence for skeletal muscle expression in publicly available databases were excluded. A total of 51 genes (36% of the original 143 genes in this region) were selected for further SNP analysis. Large-scale validation experiments in the original Endeavour article (3) revealed that the gene of interest, in this particular case the true strength determining gene, is ranked among the top 36% in ∼90% of the cases. Since we were able to discard 64% of the genes with a 90% certitude of still including the gene of interest, the use of Endeavour software is clearly beneficial for our study.

Within these selected 51 genes tagging SNPs were determined based on CEPH genotypes available from the HapMap (47) using Tagger (13), implemented in Haploview (7). Aggressive tagging and a r² threshold of 0.8 were applied on the SNPs with a minor allele frequency (MAF) >0.05. A list of the candidate genes and corresponding polymorphisms is provided as supplemental material (Supplemental Table S1, available online).¹ Linkage analyses and family-based association analyses were performed to test association of these polymorphisms with knee strength measurements (cfr. Statistical Analyses below).

Fine mapping stage 2.

An extended sample was genotyped for the stage 1 SNPs as well as additional SNPs in the most relevant genes (i.e., the genes with the highest significance and/or significance over different strength measurements) from the family-based association tests (ATP2A2, PPP1CC and NUAK1). These additional polymorphisms included other tagSNPs and polymorphisms likely to have functional consequences such as exonic SNPs, SNPs located in intron/exon boundaries, SNPs in (putative) transcription factor binding sites or in highly conserved regions. Assessment of functionality was based on queries from SNPselector (63), SNPseek, and publicly available genetic databases (Entrez Gene, UCSC genome browser, Ensembl). A list of these SNPs can be found in Table 1.

Table 1.

Polymorphisms selected in NUAK1, ATP2A2, and PPP1CC and putative function

SNP	No.	Coordinate*	Minor allele	Minor Allele Frequency	Putative Function
NUAK1
rs10861553	1	104976944	C	0.48	Tagging SNP, genotyped stage 1
rs6539247	2	104984032	T	0.44	Conserved
rs3741883	3	104985068	C	0.22	Tagging SNP, genotyped stage 1
rs55774704	4	104985440	A	0.0	Nonsynonymous [G/D]
rs3741885†	5	104985987			Conserved
rs3741886	6	104990989	C	0.071	Tagging SNP, genotyped stage 1
rs934085†	7	104993301			Conserved
rs17038085†	8	104996331			Conserved
rs17038089	9	104998855	T	0.041	Conserved
rs7133815	10	105000270	C	0.19	Conserved
rs12146713	11	105000935	C	0.065	Conserved
rs1560757	12	105001319	G	0.20	Tagging SNP
rs11112856	13	105002687	A	0.16	Conserved, located in transcription factor binding site (TFBS)
rs34594928†	14	105004761			Frameshift mutation
rs12582194	15	105006063	A	0.25	Tagging SNP, genotyped stage 1
rs1366041	16	105008642	G	0.066	Conserved
rs967872	17	105016425	G	0.45	Tagging SNP, genotyped stage 1
rs17038111	18	105017017	C	0.030	Conserved
rs1560000	19	105017277	C	0.41	Conserved
rs11112868	20	105022585	G	0.46	Conserved
rs1427785	21	105024538	C	0.42	Conserved
rs4964439†	22	105025580			Conserved
rs10861556†	23	105027775			Conserved
rs12297025	24	105030377	A	0.11	Tagging SNP, genotyped stage 1
rs10492351	25	105032945	G	0.082	Conserved
rs1215604	26	105041118	G	0.49	Tagging SNP
rs2434081	27	105052584	C	0.38	Conserved, located in TFBS
rs2559602	28	105054291	C	0.48	Conserved
rs1215597	29	105056202	A	0.49	conserved
rs55663911	30	105056410	G	0.0	Nonsynonymous [L/S]
ATP2A2
rs7304243	1	109146710	C	0.32	Conserved
rs3026421	2	109169412	T	0.0	Located in (TFBS)
rs3026432	3	109203486	T	0.0	Conserved, located in CpG island
rs3026433†	4	109203575			Located in CpG island
rs3026434	5	109204058	G	0.00096	Synonymous SNP [E/E], conserved, located in CpG island
rs3026436	6	109205386	G	0.0	Conserved
rs3026452	7	109212232	G	0.038	Tagging SNP
rs929518	8	109220198	A	0.0010	Tagging SNP
rs4630352	9	109231802	A	0.28	Tagging SNP
rs3026468	10	109242464	A	0.075	Tagging SNP, genotyped stage 1
rs1803737	11	109250126	C	0.0	Nonsynonymous [A/V], located in TFBS
rs3026473	12	109254251	A	0.0	Conserved
rs35616357	13	109262096	C	0.0	Located in TFBS
rs2302372	14	109263086	T	0.0	Synonymous SNP [D/D], conserved, located in TFBS
rs12312588	15	109264520	A	0.0	Synonymous SNP [L/l]
rs1860561	16	109267624	A	0.18	Tagging SNP
rs12297171	17	109268602	T	0.0030	Synonymous SNP [L/l], conserved
rs1042649	18	109268757		0.0	Conserved
rs1042650	19	109268768	A	0.0	Conserved
rs1042653	20	109268818	A	0.0	Conserved, located in TFBS
rs1803736	21	109268904	T	0.0	Conserved, located in TFBS
rs3026489	22	109268961	C	0.025	Tagging SNP
rs1803738	23	109269281	C	0.0	Conserved, located in TFBS
rs1803735	24	109269449	T	0.0	Conserved
rs3211481	25	109269506	T	0.0	Conserved, located in TFBS
rs3026490	26	109269892	A	0.0	Conserved
rs12638	27	109273087	T	0.0	Conserved, located in TFBS
rs3415	28	109273140	C	0.0	Conserved
rs9540	29	109273238	G	0.011	Conserved, located in TFBS
PPP1CC
rs7132423	1	109609452	C	0.0	Located in regulatory region
rs35343396	2	109610578	A	0.0	Intergenic region
rs12826018	3	109636538	C	0.0	Located in TFBS
rs11065703	4	109637378	G	0.0	Intergenic region
rs17682482	5	109639171	G	0.064	Tagging SNP
rs3191028	6	109642234	T	0.0	Located in 3′ UTR, conserved
rs1050587	7	109642544	G	0.076	Located in 3′ UTR, exonic splice enhancer
rs11558236	8	109642611	G	0.0010	Located in 3′ UTR, conserved
rs11558238	9	109643266	A	0.0	Nonsynonymous [K/*], conserved
rs11558239	10	109643272	G	0.0	Nonsynonymous [K/E], conserved
rs1973505	11	109644386	A	0.12	Tagging SNP, synonymous [C/C], genotyped stage 1
rs28702860	12	109644818	C	0.0	Nonsynonymous [N/D], conserved
rs11558237†	13	109646916			Nonsynonymous [F/S], conserved
rs7960761	14	109648798	G	0.076	Tagging SNP, genotyped Stage 1
rs11558235	15	109652822		0.0	Nonsynonymous [C/R], conserved
rs11558234†	16	109654001			Synonymous [S/S], conserved
rs11558240	17	109654064		0.0	Synonymous [N/N], conserved
rs7301769	18	109655409	G	0.078	Tagging SNP, genotyped stage 1
rs1476470	19	109751414	T	0.45	Tagging SNP, conserved
rs11065729	20	109761871	C	0.14	Intergenic region, conserved
rs12312903†	21	109764018			Intergenic region, conserved
rs12312907†	22	109764023			Intergenic region, conserved
rs12303008	23	109769214	G	0.028	Intergenic region
rs2339635	24	109769255	G	0.34	Intergenic region

^*Based on National Center for Biotechnology Information build 36;

^†failed assay. Informative single nucleotide polymorphisms (SNPs) are indicated in boldface.

A two-staged fine-mapping strategy similar to the one described for this study has already been successfully used in the fine mapping of a linkage region on chr12q12-14 and identified ACVR1B as a muscle strength-related gene (60).

Study Sample

Siblings analyzed in stage 1 and stage 2 of this study were selected from the LGfMS project (748 men, aged 17–36 yr) based on family size and DNA availability. The recruitment protocol and subject characteristics have been described elsewhere (23, 26). The medical and ethical committee of the Katholieke Universiteit Leuven approved the study, and all participants gave their written informed consent.

For stage 1, data from 500 brothers of 234 Caucasian families (26 singles, 160 duos, 38 trios, and 10 quads) were included in the statistical analyses, resulting in an overlap of 169 subjects with the preceding multipoint microsatellite-based linkage study (26). For stage 2, subjects for whom genotyping failed in stage 1 were excluded and additional subjects for whom new DNA was collected between the two stages were included. As a result, the stage 2 sample consisted of 536 siblings (239 families, 7 singles, 176 duos, 47 trios, and 9 quads) with an overlap of 464 individuals with stage 1.

Strength Measurements

Knee flexion and extension strength tests were performed on a Cybex NORM dynamometer (Lumex, Ronkonkoma, NY) and are described in detail elsewhere (23, 26). Maximal isometric knee strength and dynamic knee strength at 60°/s, 120°/s and 240°/s were retained for analyses in this paper. Torque was measured at specific angles, i.e., 60° for knee extension (quadriceps) and 30° for knee flexion (hamstrings) (0° = full extension of the knee) as the force-length relationship of a muscle predicts that optimal strength is generated at longer muscle length.

SNP Genotyping

Genomic DNA was extracted either from EDTA whole blood by a standard salting-out method or from saliva collected in Oragene DNA Self-Collection Kits (Oragene, Kanata, Canada) using the protocol provided by the manufacturer.

In stage 1 a total of 122 SNPs was genotyped on an Illumina Bead Array platform (Illumina, San Diego, CA). Genotypes were determined using GenCall 5.2.0 software and visually checked. The overall sample success rate, locus success rate, and genotype call rate were 99.8% (499/500), 97.5% (119/122), and 98.9% (58,722/59,381), respectively. These SNPs capture 48% of all available alleles at r² > 0.8. Per gene, coverage is highly variable because the number of SNPs per gene was set relative to the Endeavour ranking so that for the top genes more polymorphisms were selected than for the 20–30th gene. For example for MYBPC1 which ranked first in the global ranking eight tagSNPs were selected, resulting in a coverage of 57% of the alleles with MAF >0.05. In contrast, for UBE3B that ranked 30th, only one tagSNP was selected and only 17% of alleles with MAF >0.05 was captured.

In the second stage of the study, genotyping of 83 SNPs was performed using the matrix-assisted laser desorption ionization time-of-flight mass spectrometry iPLEX Gold platform (Sequenom, San Diego, CA). Genotypes determined by Sequenom's Typer 4.0 software were visually checked prior to statistical analysis. Overall sample success rate, locus success rate, and genotype call rate were 98.9% (530/536), 84.3% (70/83), and 94.9% (35,209/37,100), respectively.

Statistical Analyses

Pedstats v0.6.4 (57) was used to check pedigrees for Mendelian (in)consistency and to test Hardy-Weinberg equilibrium. Unlikely genotypes were identified and discarded using the error-checking option of Merlin v1.0.1.(1).

Haploview (7) was used to examine the linkage disequilibrium (LD) structure across the region of interest and to calculate r² between all pairs of SNPs within a candidate region.

Linkage analyses were performed using Merlin v1.2.1.(1). We conducted family-based association tests (FBAT v1.7.3)(21, 30) and used the haplotype version of this test (HBAT)(22) to obtain empirical P values by means of the Monte-Carlo permutation procedures implemented in HBAT (10,000 permutations were performed). Trait offsets were specified as the sample mean of the trait.

Covariates could not be included directly in the FBAT/HBAT analyses. Therefore a multiple regression of the strength measurements on age, stature, and fat-free mass was performed using SAS v9.1.3 and v9.2 (SAS Institute, Cary, NC), and the resulting residuals were used as input phenotypes for FBAT/HBAT. SAS 9.2 software was also used to calculate descriptive statistics.

RESULTS

Descriptive Statistics

In this two-staged genetic fine-mapping study 500 and 536 individuals from the LGfMS were included in stage 1 and stage 2, respectively, with an overlap of 464 individuals between both stages. Somatic and strength characteristics of these individuals are given in Table 2. Study samples for the two stages did not differ significantly for any of these traits.

Table 2.

Somatic characteristics and knee strength statistics

	Stage 1		Stage 2
	Mean	SD	Mean	SD
n	500		536
Age, yr	24.2	4.4	24.2	4.5
Weight, kg	74.1	9.6	74.0	9.8
Stature, cm	180.7	6.3	180.6	6.4
Flexion
Isometric, Nm	148.7	26.8	149.0	27.4
Torque at 60°/s, Nm	121.3	22.2	121.7	22.6
Torque at 120°/s, Nm	105.6	18.9	105.6	18.9
Torque at 240°/s, Nm	74.9	16.7	75.0	16.5
Extension
Isometric, Nm	261.3	48.3	260.9	48.7
Torque at 60°/s, Nm	196.7	34.5	196.8	34.9
Torque at 120°/s, Nm	159.1	25.5	159.5	26.0
Torque at 240°/s, Nm	107.3	19.2	107.6	19.8

All flexion and extension measurements were taken at an angle of 30 and 60°, respectively.

Stage 1 Genotyping

To reduce the number of genes and SNPs for the fine-mapping analyses, an empirical two-step selection approach was used. First, the 143 genes in the 12q22-23 region were ranked according to their similarity to five different reference sets of genes known to be involved in (the regulation of) muscle strength. Selection of the top 20% genes for each of these reference sets and exclusion of genes that are not expressed in skeletal muscle tissue resulted in the selection of 51 candidate genes. Second, tagging SNPs were determined on CEPH genotypes downloaded from the HapMap (47) website (MAF > 0.05, r² > 0.8, aggressive tagging). The most informative SNPs per gene (i.e., tagging the most other SNPs) were selected, resulting in the selection of 117 SNPs in 40 candidate genes. For the 11 remaining genes no polymorphisms meeting our criteria [Illumina designability rank = 1 (i.e., high success rate of assay design); MAF > 0.05] were present. Five additional SNPs in the gaps in between the 43 genes were selected (61). A list of these SNPs can be found as supplemental material (Supplemental Table S1, available online).

Of these 122 SNPs, 119 were successfully genotyped, one of which was excluded due to Hardy-Weinberg disequilibrium (rs3817552 in MYBPC1, P < 0.001).

Figure 2 shows the empirical P values for family-based association for all of the 118 polymorphisms. Three genes were repeatedly highly suggestive (P < 0.01, −log P value >2) over different strength measurements: NUAK1 (NUAK family, SNF1-like kinase, 1), PPP1CC (protein phosphatase 1, catalytic subunit, gamma isoform), and ATP2A2 (ATPase, Ca²⁺ transporting, cardiac muscle, slow twitch 2) and were therefore selected for further follow-up.

Fig. 2.

Family-based association results using single SNP markers. Empirical P values from HBAT analyses were calculated after 10,000 permutations. Isometric strength for knee extension (×) and flexion (bold line) and dynamic strength at 60°/s (+), at 120°/s (circle) and at 240°/s (♦) are shown. Significant results (P < 0.01, −log P value > 2.0) were marked with associated genes. All flexion and extension measurements were taken at an angle of 30 and 60°, respectively.

The effect of age, stature, and estimated fat-free mass on muscle strength was also evaluated. None of the polymorphisms significantly associated with muscle strength were associated with age, stature, or fat-free mass (data not shown). FBAT analyses of the strength characteristics corrected for these confounding factors showed a reduced significance for ATP2A2, PPP1CC and NUAK1 and increased significance of TXNRD1 (thioredoxin reductase 1) located next to NUAK1 compared with the uncorrected P values. These results are available as Supplemental Fig. S1.

Stage 2 Genotyping

Table 1 and Fig. 3 show the putative function and the location of the 30 SNPs selected in NUAK1. Six assays failed during analyses and two SNPs were 100% homozygous, resulting in 22 informative SNPs, all in Hardy-Weinberg equilibrium (P > 0.001). Three SNPs (rs10861553, rs3741886, and rs2434081) showed significant associations with all flexion strength measurements (P values ranging from 0.040 to 0.00092) (Table 3). Including age, stature, and fat-free mass as covariates resulted in P values ranging between 0.09 and 0.006 for the associations between the latter three SNPs and the strength measurements as shown in Supplemental Table S2.

Fig. 3.

Location of the selected SNPs in NUAK1. A: the locations of the SNPs are indicated by the arrows with numbers (cfr. Table 1). Exons and introns are depicted as boxes and lines, respectively. White boxes are exons and black boxes are untranslated regions. B: pairwise r² was estimated and is displayed by a gray-shaded square with the hundredths of r² value inside (Haploview). Darker shades of gray indicate higher degrees of linkage disequilibrium (LD).

Table 3.

Empirical P values for family-based association analyses for NUAK1

	rs10861553	rs6539247	rs3741883	rs3741886	rs17038089	rs7133815	rs12146713	rs1560757	rs11112856	rs12582194	rs1366041	rs967872	rs17038111	rs1560000	rs11112868	rs1427785	rs12297025	rs10492351	rs1215604	rs2434081	rs2559602	rs1215597
										Knee extensors
Isometric, Nm	0.061	0.22	0.99	0.013	0.064	0.16	0.45	0.72	0.43	0.97	0.091	0.81	0.34	0.39	0.86	0.58	0.17	0.66	0.30	0.42	0.85	0.60
Concentric, 60°/s, Nm	0.15	0.50	0.98	0.032	0.74	0.053	0.50	0.43	0.19	0.45	0.14	0.67	0.14	0.10	0.67	0.058	0.45	0.34	0.86	0.17	0.99	0.65
Concentric, 120°/s, Nm	0.30	0.57	0.93	0.22	0.30	0.082	0.82	0.42	0.37	0.31	0.70	0.77	0.046	0.27	0.96	0.083	0.45	0.077	0.84	0.17	0.34	0.65
Concentric, 240°/s, Nm	0.16	0.20	0.74	0.13	0.83	0.40	0.11	0.56	0.80	0.86	0.61	0.75	0.085	0.59	0.44	0.44	0.54	0.35	0.53	0.88	0.040	0.41
										Knee flexors
Isometric, Nm	0.00092	0.026	0.29	0.0065	0.11	0.35	0.079	0.77	0.71	0.61	0.21	0.33	0.16	0.76	0.19	0.42	0.10	0.64	0.17	0.018	0.71	0.92
Concentric, 60°/s, Nm	0.00228	0.59	0.75	0.0049	0.12	0.52	0.11	0.74	0.65	0.34	0.53	0.34	0.67	0.95	0.34	0.58	0.44	0.57	0.15	0.0059	0.14	0.69
Concentric, 120°/s, Nm	0.00132	0.13	0.97	0.0038	0.23	0.99	0.20	0.74	0.66	0.21	0.55	0.38	0.10	0.50	0.17	0.60	0.20	0.87	0.28	0.030	0.52	0.82
Concentric, 240°/s, Nm	0.11	0.50	0.73	0.023	0.41	0.99	0.15	0.38	0.38	0.37	0.54	0.071	0.20	0.32	0.055	0.16	0.14	0.091	0.57	0.39	0.49	0.96

All flexion and extension measurements were taken at an angle of 30 and 60°, respectively. Boldface indicates significant associations at P < 0.05.

The functionality and position of the 24 and 29 polymorphisms selected in PPP1CC and ATP2A2, respectively, are depicted in Table 1 and Fig. 4. Five assays failed during analyses and 28 appeared to be 100% homozygous, giving a total of 20 informative SNPs, of which one was not in Hardy-Weinberg equilibrium (rs4630352). Results from the family-based association analyses can be found in Table 4. Significant associations and trends toward significance over several strength measurements were present for rs3026468 in ATP2A2 and for rs1050587, rs7960761, and rs7301769 in PPP1CC. However, as the LD between rs3026468 and the significant PPP1CC SNPs is rather high (r² values between 0.91 and 0.95), it is possible that they all point toward the same “strength increasing allele.”

Fig. 4.

Location of the selected SNPs in ATP2A2 and PPP1CC. A: the locations of the SNPs are indicated by the arrows with numbers (cfr. Table 1). Exons and introns are depicted as boxes and lines, respectively. White boxes are exons and black boxes are untranslated regions. B: pairwise r² was estimated and is displayed by a gray-shaded square with the hundredths of r² value inside (Haploview). Darker shades of gray indicate higher degrees of LD.

Table 4.

Empirical P values for family-based association analyses for ATP2A2 and PPP1CC

				ATP2A2										PPP1CC
	rs7304243	rs3026434	rs3026452	rs929518	rs4630352	rs3026468	rs1860561	rs12297171	rs3026489	rs9540	rs17682482	rs1050587	rs11558236	rs1973505	rs7960761	rs7301769	rs1476470	rs11065729	rs12303008	rs2339635
Knee extensors
Isometric, Nm	0.81	0.49	0.95	0.54	0.54	0.016	0.020	0.11	0.92	0.34	0.14	0.023	0.49	0.14	0.055	0.042	0.20	0.84	0.69	0.66
Concentric, 60°/s, Nm	0.99	0.51	0.98	0.49	0.80	0.022	0.36	0.78	0.67	0.68	0.075	0.081	0.53	0.21	0.091	0.079	0.39	0.71	0.80	0.72
Concentric, 120°/s, Nm	0.20	0.49	0.45	0.45	0.36	0.017	0.16	0.55	0.58	0.58	0.17	0.14	0.51	0.68	0.16	0.14	0.48	0.46	0.35	0.32
Concentric, 240°/s, Nm	0.28	0.52	0.25	0.50	0.625	0.0015	0.21	0.48	0.63	0.84	0.017	0.0091	0.47	0.17	0.0091	0.0098	0.68	0.42	0.55	0.76
Knee flexors
Isometric, Nm	0.094	0.51	0.86	0.52	0.64	0.0044	0.42	0.13	0.79	0.56	0.13	0.012	0.47	0.23	0.025	0.028	0.010	0.88	0.39	0.11
Concentric, 60°/s, Nm	0.16	0.46	0.19	0.50	0.17	0.019	0.40	0.51	0.36	0.24	0.39	0.058	0.49	0.93	0.079	0.091	0.15	0.81	0.86	0.99
Concentric, 120°/s, Nm	0.38	0.52	0.051	0.53	0.54	0.27	0.49	0.12	0.61	0.64	0.98	0.56	0.50	0.36	0.75	0.73	0.54	0.091	0.73	0.40
Concentric, 240°/s, Nm	0.26	0.51	0.11	0.50	0.58	0.17	0.85	0.66	0.69	0.70	0.69	0.46	0.52	0.44	0.53	0.53	0.99	0.72	0.74	0.79

All flexion and extension measurements were taken at an angle of 30 and 60°, respectively. Bold face indicates significant difference at P < 0.05.

Inclusion of age, stature, and fat-free mass as covariates reduced the evidence for association between PPP1CC and the muscle strength characteristics (Supplemental Table S3). Three SNPs show significant associations with a single strength phenotype (rs17682482, extensor strength at 240°/s, P = 0.048; rs1050587 - isometric flexor strength, P = 0.047 and rs11065729, flexor strength at 120°/s, P = 0.050). In ATP2A2, significant associations remain present between rs3026468 and extensor strength (0°/s and 240°/s, P values of 0.045 and 0.033, respectively). Associations between rs7304243, rs3026452, and rs1860561 and a single strength phenotype are also observed.

DISCUSSION

This paper described a two-staged fine mapping of a previously identified linkage peak on chr12q22-23 (23, 26). FBATs showed an association between knee muscle torque production and rs10861553 and rs3741886 in NUAK1, rs3026468 located in ATP2A2, and rs1050587 and rs7301769 in PPP1CC. Additional putatively functional polymorphisms were not associated with knee muscle strength.

After stage 1, the ATP2A2, PPP1CC, and NUAK1 genes were selected as candidate genes for follow-up. The ATP2A2 gene encodes the sarcoplasmic reticulum calcium ATPase 2. It is expressed predominantly in cardiac and slow-twitch skeletal muscle (34, 64) and is involved in the calcium handling necessary for the muscle contraction-relaxation cycle (40). Darier's disease patients, who have an altered ATP2A2 expression level due to mutations in the ATP2A2 gene, have a prolonged contraction time and half-relaxation time of the adductor pollicis muscle (36).

NUAK1 and PPP1CC both play a role in glycogen metabolism. NUAK1, an AMP-activated protein kinase, is expressed in skeletal muscle. It could have both a structural and a regulatory role in the reaction of the muscle on muscle contraction. Hoppe et al. (20) suggested that unc82 (the Caenorhabditis elegans homolog of NUAK1) can regulate some aspects of thick filament organization during changes in muscle length. Fisher et al. (18) indicated that NUAK1 could play a regulatory role in muscle metabolism and be active in glucose transport-related pathways that are controlled similarly by muscle contraction and insulin.

PPP1CC (also known as PPP1G) is a subunit of the protein phosphatase 1 and is universal in skeletal muscle (37). Glycogen-associated phosphatases such as the protein phosphatase 1 are responsible for the dephosphorylation and subsequent inactivation of glycogen synthase. PPP1CC is not required for insulin-stimulated glycogen synthesis in skeletal muscle but appears to be a component of the response to contractile action (6, 38).

Since we did not a priori select our candidate genes based on their specific function, it is surprising that two genes from complementary pathways in glycogen metabolism were identified in our study.

In the second stage of the study, we genotyped additional tagging SNPs and SNPs likely to influence gene function in an extended sample. In NUAK1, two tagSNPs and one conserved SNP in a transcription factor binding site showed associations with multiple muscle strength measurements. For ATP2A2 and PPP1CC 3 tagSNPs and one SNP in a splice enhancer showed association.

Because muscle strength can be influenced by factors such as age, stature, and fat free mass, these factors were included as covariates in the analyses. As FBAT/HBAT analyses cannot directly incorporate confounding factors in the analyses, we performed a multiple regression of the strength measurements on age, weight, and fat-free mass. The resulting residuals were then used as input phenotypes for the FBAT/HBAT analyses. For both stage 1 and stage 2 analyses, inclusion of the covariates resulted in reduced significance of the association findings. For NUAK1 evidence for association remained present for rs10861553 (TagSNP), rs2434081, and rs2559602 (two conserved SNPs), while a trend toward association became apparent for rs12146713, a conserved SNP. For the ATP2A2/PPP1CC locus, only rs3026468, a tagSNP, remained significant over different strength measurements.

As most associated SNPs are intronic tagSNPs direct effects on protein function are not expected. However, it is possible that these tagSNPs are located in a currently unknown transcriptional regulatory element or in a regulatory miRNA, often found in intronic sequences (9, 32, 39, 46, 55). Alternatively, these SNPs could be in LD with neighboring, yet untyped, polymorphisms. These polymorphisms can be currently unknown polymorphisms within the NUAK1 or ATP2A2 genes or polymorphisms located in neighboring genes. The latter seems surprising given the low levels of intragenic LD. However, several studies show that even though strong nonrandom association between intragenic polymorphisms is present, a nonnegligible proportion of intragenic SNPs is in weak LD with each other (4, 5, 45, 54). Moreover, LD has been reported between quite distant markers (2, 12, 45). Screening of the LD structure around rs10864553 showed that no other genes are located within this region. The specific source of the association therefore remains unclear. For the ATP2A2/PPP1CC locus, it is possible that one of the hypothetical proteins in the region (HSU79274, FLJ21127, or MGC15619) has a yet unknown function in muscle strength.

In an effort to strengthen our association results with gene expression data, we genotyped the associated SNPs in a sample (n = 23) for which knee strength data and muscle biopsies were available (56). Unfortunately, due to the small sample size, no significant SNP effects on muscle strength or ATP2A2/PPP1CC/NUAK1 expression could be detected (data not shown).

Another explanation of this limited set of significant association findings is the expected number and effect size of associated SNPs with the current sample size. Depending on the approach taken in power analyses [family-based, in PBAT, with a model including MAF, mode of inheritance, LD between marker and trait locus and effect size, or based on r² of the SNP effect, without the influence of the MAF in Genetic Power Calculator (43)], different outcomes on the power of this study are found. To address this issue, we applied similar assumptions to the power calculations as reported in Hagberg et al. (19). Given that 122 SNPs were tested in the first phase and 83 SNPs in the second phase, the alpha-level was adapted to 0.0004 and 0.0006 respectively (Bonferroni correction: 0.05/122 = 0.0004 - 0.05/83 = 0.0006). With a sample size of 500 (and 536) subjects, this study has 83% (and 87%) power to detect a SNP with an effect size of 4%. This probably represents an effect size that can be expected for SNP variants at the higher end of the effect-size distribution for complex traits and corresponds well to the top four listed SNP effects in genome-wide association (GWA)-detected variants for training-induced responses in VO_2max in the HERITAGE family study, which was also based on a similar sample size of 473 whites (10). Power calculations in PBAT for continuous traits show less optimal power as the stage 2 sample could detect a locus with an effect of 15% explained variance with 80% power, if the locus MAF exceeds 0.08. This estimate corresponds well to the lower bound of MAF for the SNPs that were associated at P < 1.5 × 10⁻⁴ with change in VO_2max in Bouchard et al. (10). Given that the chromosomal region was identified by linkage analysis and SNPs have been selected in muscle-related prioritized genes, we might hypothesize the detection of association effects of larger size (e.g., in the 4% range).

In addition, the availability of numerous genes and environmental factors that possibly influence the phenotypes under study and the presence of gene-gene and gene-environment interactions will also affect the power of the study. However, a strategy similar to the one described here has successfully identified the ACVR1B gene in the chr12q12-14 linkage region (23, 26) as a strength-determining gene (60).

It can be argued that the currently applied fine-mapping strategy is biased toward known genes. Indeed, we did not apply a true hypothesis-free approach for this genomic region as is done in a GWA. In the first stage a prioritizing bioinformatics tool, Endeavour (3), was used to select a set of genes based on their ranked position as candidate genes. Endeavour uses multiple heterogeneous data sources [literature, functional annotation (GO), microarray expression, expressed sequence tag expression, protein domains, protein-protein interactions, pathway membership, cis-regulatory modules, transcriptional motif, and sequence similarity] based on a set of reference genes. We applied five ranking procedures using different sets of reference genes and therefore feel strongly that the selected genes were well chosen and allowed us to apply a cost-effective strategy for SNP selection in those genes. However, the outcome of the in silico-based gene selection needs to be placed within the specific available database information at the time of selection.

In addition, it should be noted that not all genetic variation was captured using our stage 1 SNP selection method because we focused on SNPs with a MAF >5%. It is therefore plausible that genes and/or polymorphisms other than the ones we investigated are involved in the linkage signal. Moreover, genes for which no significant associations were found should not be considered negatively associated as not all SNPs were covered (e.g., rare alleles). However, we were aware of these limitations from the beginning of our study (as stated in Ref. 60) and acknowledge that additional research will be necessary to determine which other genes and polymorphisms are responsible for the linkage signal. The analyses should also take into account that since the onset of this study a considerable amount of new information about human genetic variants has become available (e.g., 1,000 genomes project).

Even though dense SNP panels (current study, 122 SNPs, average marker distance 0.28 cM) are shown to induce higher genetic informativeness than sparser microsatellite markers (original study, 11 microsatellites, average marker distance 4.9 cM) (17, 59), we could not confirm the original linkage peak using our SNP selection. This difference could be related to sample size differences or regional informativeness of closely chosen markers. Therefore we reran the analyses on the subset of individuals that had both microsatellite and SNP data available (169 siblings from 57 families). In this subset, a clear linkage peak is present in all analyses. An example of these analyses can be found in the supplemental material (Supplemental Fig. S1, available online). In addition, this model was compared with a model including both linkage and association, and results support an association between rs10861553 and rs3741886 in NUAK1 and strength of the knee flexors (data not shown).

In the original linkage study (26) insulin-like growth factor 1 (IGF1) and myosin binding protein C, slow type (MYBPC1) were suggested as possible candidate genes as IGF1 has a role in muscle growth, repair, and hypertrophy (8), and MYBPC1 is involved in muscle contraction (29). However, for none of the eight SNPs in MYBPC1 and the five SNPs in IGF1 associations reached our significance threshold.

In summary, we conducted a genetic fine-mapping study to identify and prioritize a small number of genes within a list of muscle-related prioritized genes underlying a previously reported linkage peak for knee muscle strength on chr12q22-23. TagSNPs in the NUAK1, ATP2A2, and PPP1CC genes were associated with knee muscle strength. Despite detailed analyses of putatively functional polymorphisms in these genes, no additional associations could be detected. Further in-depth analyses are necessary to elucidate the exact role of these loci in human muscle strength and to determine which (functional) polymorphisms are at the base of the linkage and association results.

GRANTS

A. Windelinckx and W. Huygens were funded by the Research Fund of the K. U. Leuven (OT/04/44 and OT/98/39, respectively). G. De Mars was funded by Research Foundation Flanders (FWO) Grant G.0496.05; he passed away on February 28th 2010. Maarten Peeters is a postdoctoral researcher for the Research Foundation Flanders. The fine mapping phase of the Leuven Genes for Muscular Strength Study is funded by OT/04/44 and the FWO (G.0496.05).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).