Online Appendix
Type 2 Diabetes Risk Alleles are Associated with Reduced Size at Birth
Rachel M Freathy, Amanda J Bennett, Susan M Ring, Beverley Shields, Christopher J Groves, Nicholas J Timpson, Michael N Weedon, Eleftheria Zeggini, Cecilia M. Lindgren, Hana Lango, John R. B. Perry, Anneli Pouta, Aimo Ruokonen, Elina Hyppönen, Chris Power, Paul Elliott, David P Strachan, Marjo-Riitta Järvelin, George Davey Smith, Mark I McCarthy, Timothy M Frayling and Andrew T Hattersley.
1
Supplementary Methods
Genotyping. Genotyping of the ALSPAC and EFSOCH samples was performed by KBiosciences (Hoddesdon, UK), using their own system of fluorescence-based competitive allele-specific PCR (KASPar). (www.kbioscience.co.uk). Genotyping of the NFBC1966 and 1958BC samples was performed by the authors in Oxford using TaqMan® SNP Genotyping Assays (Applied Biosystems, Foster City, California, USA; assay identifiers available from the authors): ABsolute QPCR Rox Mix (ABgene, Thermo Fisher Scientific, Epsom, UK) was used in conjunction with the TaqMan® assays to carry out PCR reactions and genotypes were ascertained using a 7900HT Sequence Detection System (ABI Prism, Applied Biosystems, Foster City, California, USA).
The per-SNP genotyping call rate was calculated for each study group. For 29 of the 30 SNP assays, the median [range] call rate was: 96.9% [92.5 to 98.9%]. The call rate of the remaining SNP assay (rs4402960 in 1958BC) was 85.7%. The percentage of duplicate samples included for genotyping of each study group was 2% (ALSPAC children), 6% (ALSPAC mothers), 13% (EFSOCH children), 6% (EFSOCH mothers), 7% (NFBC1966) and 1% (1958BC). Concordance between duplicate samples was ≥ 99% for all SNP assays except rs4402960 in 1958BC (97%). There was no evidence of deviation from Hardy-Weinberg equilibrium (all P > 0.01). Our investigations into the suboptimal assay suggested no obvious reason for the low call rate. There was no evidence of deviation from Hardy-Weinberg equilibrium (P = 0.62) and the minor allele frequency of 0.317 was very similar to that in the other UK cohorts, suggesting no allele-specific or genotype-specific effect that might introduce bias. Finally, when we repeated the meta-analysis for SNP rs4402960 with the 1958BC data excluded, the results changed very little, so we decided to include the data from this assay.
Risk allele frequencies for type 2 diabetes varied little among the UK-based samples (rs10946398: 31-32%; rs10811661: 83-85%; rs1111875: 58-60%; rs4402960: 31-32%; rs13266634: 68-71%). In the NFBC1966 sample, the frequencies were often outside the UK ranges (rs10946398: 39%; rs10811661: 85%; rs1111875: 52%; rs4402960: 30%; rs13266634: 63%), but were very similar to allele frequencies observed previously in other samples of Finnish ancestry [1]. Therefore, differences between these and the UK allele frequencies are very likely to reflect underlying population differences, rather than artefacts caused by our sampling or genotyping.
Exclusions due to extreme birth weight values. The numbers of individuals with birth weights outside 4 SD of the sex-adjusted study mean (after excluding multiple births and babies born before 36 full weeks of gestation) were as follows: ALSPAC = 2; EFSOCH = 0; NFBC1966 = 3; 1958BC = 2.
1
Reference
1. Scott LJ, Mohlke KL, Bonnycastle LL, Willer CJ, Li Y, et al. (2007) A genome-wide association study of type 2 diabetes in Finns detects multiple susceptibility variants. Science 316: 1341-1345.
1
Supplementary Table 1. Meta-analysis of birth weight by maternal genotype across studies with maternal genotype available (ALSPAC and EFSOCH)
Locus / Total N in meta-analysis / Per-allele effect size, in g (95% CI) / P value / Heterogeneity statistics: I2 (P value)*CDKAL1 (rs10946398) / 7760 / 4 (-11, 19) / 0.60 / 1.4 (0.31)
CDKN2A-2B (rs10811661) / 7817 / 17 (-1, 35) / 0.06 / 0.0 (0.54)
HHEX-IDE (rs1111875) / 7799 / -14 (-28, 0) / 0.05 / 54.8 (0.14)
IGF2BP2 (rs4402960) / 7821 / 2 (-13, 17) / 0.83 / 61.9 (0.11)
SLC30A8 (rs13266634) / 7655 / 8 (-7, 23) / 0.29 / 0.0 (0.64)
*I2 is the percentage of total variation among the study estimates that is due to between-study heterogeneity. The P value is from Cochran’s Q test.
1
Supplementary Table 2. Birth weight meta-analysis results before and after adjustment of maternal and fetal genotype effects for one another (using ALSPAC and EFSOCH samples, where genotype was available for both mother and child)
Fetal genotypeSNP (locus) / [1] Meta-analysis of fetal genotype against birth weight, adjusted for sex and gestation / [2] Meta-analysis of fetal genotype against birth weight, adjusted for sex, gestation and maternal genotype
N in meta-analysis / Per-risk allele effect size in g (95% CI) / P value / Heterogeneity statistics: I2 (P value)* / N in meta-analysis / Per-risk allele effect size in g (95% CI) / P value / Heterogeneity statistics: I2 (P value)*
rs10946398 (CDKAL1) / 5399 / -25 (-43, -7) / 0.005 / 40.3 (0.20) / 5399 / -36 (-56, -16) / 0.0005 / 0.0 (0.90)
rs10811661 (CDKN2A/B) / 5493 / 6 (-15, 28) / 0.56 / 35.5 (0.21) / 5493 / -2 (-27, 23) / 0.89 / 72.5 (0.06)
rs1111875 (HHEX/IDE) / 5480 / -25 (-42, -9) / 0.003 / 42.4 (0.19) / 5480 / -29 (-48, -10) / 0.003 / 0.0 (0.63)
rs4402960 (IGF2BP2) / 5507 / -7 (-25, 10) / 0.42 / 0.0 (0.44) / 5507 / -10 (-31, 10) / 0.32 / 0.0 (0.99)
rs13266634 (SLC30A8) / 5342 / 3 (-15, 21) / 0.73 / 0.0 (0.76) / 5342 / -7 (-28, 14) / 0.51 / 0.0 (0.99)
Maternal genotype
SNP (locus) / [1] Meta-analysis of maternal genotype against birth weight, adjusted for sex and gestation / [2] Meta-analysis of maternal genotype against birth weight, adjusted for sex, gestation and fetal genotype
N in meta-analysis / Per-risk allele effect size in g (95% CI) / P value / Heterogeneity statistics: I2 (P value)* / N in meta-analysis / Per-risk allele effect size in g (95% CI) / P value / Heterogeneity statistics: I2 (P value)*
rs10946398 (CDKAL1) / 5399 / 3 (-14, 21) / 0.70 / 79.6 (0.03) / 5399 / 21 (1, 42) / 0.04 / 70.8 (0.06)
rs10811661 (CDKN2A/B) / 5493 / 15 (-6, 37) / 0.16 / 0.0 (0.41) / 5493 / 16 (-9, 41) / 0.21 / 64.1 (0.10)
rs1111875 (HHEX/IDE) / 5480 / -8 (-24, 9) / 0.37 / 63.1 (0.10) / 5480 / 7 (-12, 26) / 0.50 / 27.2 (0.24)
rs4402960 (IGF2BP2) / 5507 / 2 (-16, 19) / 0.87 / 60.4 (0.11) / 5507 / 7 (-13, 27) / 0.51 / 46.8 (0.17)
rs13266634 (SLC30A8) / 5342 / 17 (-1, 35) / 0.07 / 0.0 (0.56) / 5342 / 20 (0, 41) / 0.06 / 0.0 (0.61)
*I2 estimates the percentage of total variation between the two study estimates that is due to between-study heterogeneity. The P value is from Cochran’s Q test.
1