Genetic background of extreme violent behaviour1
Supplementary Information
CRIME cohort
DNA was extracted from peripheral blood leukocytes or saliva using commercially available Puregene DNA isolation kit (Puregene, Gentra Systems, Minneapolis, USA) and Autopure LS equipment (Qiagen, Crawley, UK). A total of 579 CRIME cohort samples were genotyped at the Welcome Trust Sanger Institute, Cambridge, UK, with Illumina 670 K platform for 594,398 common single-nucleotide polymorphisms (SNPs).
GenMets and Corogene control cohorts
The GenMets and Corogene control samples were both genotyped with HumanHap610-Quad Genotyping BeadChip at the Cambridge Welcome Trust Sanger Institute.
Quality Control
The quality control of the data was conducted using PLINK v1.071 and SPSS (IBM Corp. Released 2011. IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp.).
In the criminal cohort, 29 samples were excluded from the data set: 3 duplicates, 2 id fails, and 24 heteterozygotic discrepancies that failed the Sanger quality control. Samples with a call rate (i.e., “missingness”) of less than 0.98 were removed (two individuals). Sex discrepancies were examined for rates of heterozygotic X-chromosomes, but no discrepancies were detected. All SNPs with a call rate < 0.98 and a minor allele frequency (MAF) < 0.01 were excluded from both criminal and control data sets. Also, SNPs with a Hardy-Weinberg equilibrium (HWE) test P 0.001 were removed. To avoid any ambiguity due to strand origins of the SNPs, all AT/TA and CG/GC SNPs were removed. The genome-wide identity by descent estimation was performed, with two individuals (1 sibling, 1 avuncular/half sibling) subsequently removed.
Data merging
After the initial QC for each set of criminal cases and controls, the three datasets were merged to include 546 criminal cases (506 males, 40 females; 544,255 SNPs), 2092 GenMets controls (1020 males, 1072 females; 549,049 SNPs) and 4030 Corogene controls (2496 males, 1534 females; 547186 SNPs). The resulting dataset consisted of 546 cases with 6122 controls and 569,250 markers. All SNPs that were not originally included in 3 datasets (47,754 SNPs) were removed. The genome-wide identity by descent estimation was performed. Subsequently, 1 case (avuncular/half sibling) and 32 controls (11 avuncular/half siblings, 18 siblings and 3 identical samples) were removed. The population structure was then analyzed by calculating the multidimensional scaling (MDS) clusters with PLINK and plotting the clusters with SPSS. Based on the MDS analysis, 146 population outliers were removed (39 criminals and 107 controls). After the QC, a total of 506 criminal offenders and 5983 controls as well as 521,486 SNPs were included in the dataset.
Statistical analyses
All of the association and haplotype analyses were performed using logistic regression in PLINK v1.07, with age and sex as covariates. In addition, the 10 MDS clusters were included as covariates in the GWAS analyses to adjust for population stratification. SPSS (IBM Corp. Released 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp.) was used in the gene x environment analysis to calculate the interaction term. The level of statistical significance was set to 0.025 (0.05/2) in the candidate gene study. Attributable fraction was calculated in replication cohorts (CRIME cohort for MAOA and Homicide Offender Cohort for CDH13) as described in Armitage and Berry2 (1993).
Power calculations
Assuming the additive model, genotype relative risk 1.3 and a perfect linkage disequilibrium between a trait locus and a marker, there was a 79.4% and 20.9% power (alpha = 0.05) to detect variants with a MAF 0.2 in the two models, violent offending and extremely violent offending, respectively. When increasing the genotype relative risk to as high as 1.8, the power also increased to 100% and 73.7% for violent offending and extremely violent offending, respectively. The power for the replication of the rs11649622 (CDH13 gene) in the sample of homicide offenders (n = 103) was 99.8% (with the assumptions of the additive model, genotype relative risk 2.72, and perfect disequilibrium between trait locus and a marker, for alpha 0.05, MAF 0.31). (
Genotyping with Sequenom MassArray
The flow-chart for this study procedure is shown in Supplementary Figure 1. Based on results from the GWAS for violent and extremely violent offenders, we selected seven variants for genotyping in the CRIME cohort (n = 749), and in the Health 2000 sample (n = 6600). We also genotyped the MAOA and HTR2B stop codon as candidate variants. After the re-analyses of violent offending and extremely violent offending data, all four CDH13 SNPs were selected for genotyping in the replication sample of homicide offenders (n = 114). The genotyping was performed at the Institute for Molecular Medicine Finland with Sequenom Mass array technology (Sequenom, San Diego, California).
Quality Control
In the Sequenom genotyping, the thresholds for inclusion of samples and SNPs in all data sets (i.e.; CRIME, homicide offenders, and Health 2000, which included the Genmets cohort) were as follows: for both samples and SNPs the genotype call rate was > 0.95, MAF > 0.01, and HWE test P < 0.001. After the QC, there remained 509 violent offenders, including 78 extremely violent offenders, 103 homicide offenders, 1877 Genmets controls, and 4210 other Health 2000 controls (Table 1). All the SNPs passed the QC criteria.
MAO-A repeat allele genotyping
Altogether 720 individuals of CRIME cohort, 114 homicide offenders, and 2002 controls were genotyped for MAO-A u-VNTR polymorphism with the primer set 6FAM-5'- ACAGCCTGACCGTGGAGAAG-3' and 5'- GAACGGACGCTCCATTCGGA-3', as in Sabol et al. (1998). Amplification was performed in 5-μl reaction containing 10 ng of genomic DNA, 200 μM of dNTP, 1 μM of primers, and 0.25 units of Phusion Hot Start II DNA Polymerase (Thermo Scientific) in 1× buffer with 1.5 mM of MgCl₂ and 3% DMSO. Cycling parameters were initial denaturation for 1 min at 98°C, and then 30 cycles of 10 sec at 98°C, 30 sec at 60°C, 20 sec at 72°C, and a final extension 5 min at 72°C. The PCR products were analyzed using a genetic analyzer (ABI310; Applied Biosystems). Data were analyzed using computer software (ABI GeneScan; Applied Biosystems). The following PCR products were observed in this assay: 318 (3-repeat allele), 336 (3.6-repeat allele), 348 (4-repeat allele), and 378 (5-repeat allele) base pairs, thus in actual runs the amplicon sizes calculated by the software were 3 base pairs smaller than expected when ran together with the GeneScan 400HD size standard, except for the 336 base pairs PCR-product that was considered equivalent to 3.5-repeat allele. A rare 2-repeat allele was not observed in this assay. The alleles were classified by their function as in Sabol et al.3 1998, namely 3- and 5-repeat alleles were classified as a low-activity allele and 3.5-, and 4-repeat alleles as high-activity allele.
Quality Control
There were 15 samples of the CRIME cohort, 17 homicide offenders, and 50 controls with no genotyping results. In gender check, one CRIME sample, one homicide offenders’ sample, and 6 controls gave discrepant results and were thus removed from the data.
References
1. Purcell S,NealeB, Todd-BrownK, ThomasL, FerreiraMAR, BenderDet al. PLINK: a toolset for whole-genome association and population-based linkage analysis. Am J Hum Gen2007; 81: 559-575.
2. ArmitageP, BerryG. Statistical Methods in Medical Research. 3rd Edition. Blackwell Science Ltd, Oxford, 1994, pp 519-522.
3. Sabol SZ, Hu S, Hamer D. A functional polymorphism in the monoamine oxidase A gene promoter. Hum Genet 1998; 103: 273-279.
Genetic background of extreme violent behaviour1
Supplementary Table 1 a. Clinical and demographic characteristics among MAOA genotyped offenders
No violent crimes (ASP, substance abuse, maltreated) / One violent crime (ASP, substance abuse, maltreated) / 2–4 violent crimes (ASP, substance abuse, maltreated) / 5–9 violent crimes (ASP, substance abuse, maltreated) / More than 10 violent crimes (ASP, substance abuse, maltreated) / Controls (ASP, substance abuse, maltreated)Males / 173 (99, 153, 28) / 78 (54, 71, 17) / 194 (156, 183, 43) / 120 (92, 116, 41) / 76 (63, 74, 20) / 1110 (NA, NA, NA)
Females / 27 (26, 23, 7) / 10 (10, 9, 2) / 17 (15, 16, 5) / 7 (7, 7, 3) / 3 (3, 3, 3) / 842 (NA, NA, NA)
All / 200 (125, 176, 35) / 88 (64, 80, 19) / 211 (171, 199, 48) / 127 (99, 123, 44) / 79 (66, 77, 23) / 1952 (NA, NA, NA)
Age (mean ± SD) / 36.8 ± 10.5 / 38.1 ± 11.0 / 34.4 ± 8.7 / 36.5 ± 8.7 / 37.1 ± 8.6 / 54.9 ± 12.3
Genetic background of extreme violent behaviour1
Supplementary Table 1 b. The allele frequencies and ORs for MAOA low activity genotype in the CRIME cohort compared with general population.
MAFcases / MAF
control / OR / 95% CI / p-value
Non-violent offenders
Entire CRIME cohort
All / 0.4204 / 0.4319 / 1.133 / 0.820 / 1.566 / 0.4488
Males / 0.4128 / 0.4149 / 1.115 / 0.763 / 1.629 / 0.5754
Females / 0.4444 / 0.4430 / 1.206 / 0.639 / 2.279 / 0.5633
Maltreated / All / 0.4146 / 0.4319 / 1.129 / 0.583 / 2.184 / 0.7195
subpopulation / Males / 0.3333 / 0.4149 / 0.852 / 0.365 / 1.988 / 0.7110
Females / 0.5714 / 0.4430 / 1.881 / 0.609 / 5.81 / 0.2720
Violent offenders
Entire CRIME cohort
All / 0.5258 / 0.4319 / 1.708 / 1.329 / 2.195 / 2.93E-05
Males / 0.5192 / 0.4149 / 1.661 / 1.252 / 2.204 / 0.0004
Females / 0.5676 / 0.4430 / 1.898 / 1.094 / 3.291 / 0.0226
Maltreated / All / 0.5102 / 0.4319 / 1.623 / 1.102 / 2.39 / 0.0142
subpopulation / Males / 0.5207 / 0.4149 / 1.748 / 1.133 / 2.697 / 0.0116
Females / 0.4615 / 0.4430 / 1.211 / 0.513 / 2.861 / 0.6618
Extremely violent offenders
Entire CRIME cohort
All / 0.6220 / 0.4319 / 2.662 / 1.603 / 4.423 / 0.0002
Males / 0.6184 / 0.4149 / 2.617 / 1.540 / 4.445 / 0.0004
Females / 0.6667 / 0.4430 / 3.143 / 0.521 / 18.97 / 0.2118
Maltreated / All / 0.6538 / 0.4319 / 3.173 / 1.347 / 7.476 / 0.0083
subpopulation / Males / 0.6500 / 0.4149 / 3.125 / 1.181 / 8.270 / 0.0217
Females / 0.6667 / 0.4430 / 3.143 / 0.521 / 18.97 / 0.2118
Supplementary Table 2 a. The most significant 50 SNPs in association with violent offending (In build GRCh36/hg18). After regenotyping, the lowest p-value (p = 2 x 10-5) was observed for rs10123897 (SPIN1).
Supplementary 2 b. The most significant 50 SNPs in association with extremely violent offending (10 or more aggravated violent crimes) (In build GRCh36/hg18).After regenotyping, the lowest p-value (p = 4.2 x 10-6) was observed for rs11649622 (CDH13).
Genetic background of extreme violent behaviour1
Supplementary Table 3.GWAS results for the 6 variants within the MAOA gene.
CHR / SNP / BP / A1 / MAF violent offenders / MAF controls / OR / P / MAF extremely violent offenders / MAF controls / OR / p23 / rs1465108 / 43423153 / A / 0.4593 / 0.4128 / 1.102 / 0.4372 / 0.541 / 0.4128 / 1.608 / 0.07974
23 / rs909525 / 43438146 / G / 0.4777 / 0.4259 / 1.148 / 0.2652 / 0.5574 / 0.4259 / 1.659 / 0.06255
23 / rs2235186 / 43480372 / A / 0.4121 / 0.3755 / 1.073 / 0.5769 / 0.459 / 0.3755 / 1.368 / 0.2478
23 / rs2072743 / 43484465 / A / 0.4278 / 0.3932 / 1.098 / 0.4541 / 0.4754 / 0.3932 / 1.39 / 0.2236
23 / rs1137070 / 43488335 / A / 0.4121 / 0.3743 / 1.089 / 0.4987 / 0.459 / 0.3743 / 1.392 / 0.2226
23 / rs3027409 / 43491977 / C / 0.01575 / 0.01554 / 1.759 / 0.2234 / 0.01639 / 0.01554 / 1.675 / 0.6223
Supplementary Table 4.Resultsfrom the association analysis of extremely violent offending (10 or more convictions for aggravated violent crimes) for the regenotyped variants (N = 78 cases and N = 1877 controls, BP = base pair in reference build GRCh37/hg19).
CHR / SNP / BP / A1 / A2 / MAF cases / MAF controls / OR / P / GENE16 / rs11649622 / 83605265 / A / G / 0.3077 / 0.1796 / 2.715 / 4.19E-06 / CDH13
9 / rs17787449 / 79418448 / A / G / 0.2436 / 0.1287 / 2.524 / 5.04-05 / PRUNE2
16 / rs4075942 / 83549820 / G / A / 0.1859 / 0.0936 / 2.502 / 0.000232 / CDH13
16 / rs12919501 / 83511016 / G / A / 0.2372 / 0.1261 / 2.215 / 0.000495 / CDH13
16 / rs7190768 / 83741891 / A / G / 0.4936 / 0.3471 / 1.779 / 0.00181 / CDH13
Genetic background of extreme violent behaviour1
Supplementary Table 5. The distribution of MAOA and CDH13 (rs rs11649622) genotypes among the extremely violent offenders. The ORs for combined homozygosity for low activity MAOA and rs11649622 minor allele (i.e., 1,1 vs. 2,2 genotypes compared between cases and controls) were 13.45 (95% CI 3.86–46.94) for all individuals and 16.63 (95% CI 4.39–62.89) for males.
ALL INDIVIDUALS
CASES / MAOACDH13 / 1,1 / 1,2 / 2,2
1,1 / 4 (5.1%) / 0 / 3 (3.8%) / 7 (9.0%)
1,2 / 19 (24.4%) / 2 (2.6%) / 13 (16.7%) / 34 (43.6%)
2,2 / 25 (32.1%) / 0 / 12 (15.4%) / 37 (47.4%)
48 (61.5%) / 2 (2.6%) / 28 (35.9%) / 78 (100%)
CONTROLS / MAOA
CHD13 / 1,1 / 1,2 / 2,2
1,1 / 14 (0.8%) / 16 (0.9%) / 17 (0.9%) / 47 (2.6%)
1,2 / 183 (10.2%) / 113(6.3%) / 248 (13.8%) / 544 (30.4%)
2,2 / 383 (21.4%) / 252 (14.1%) / 565 (31.5%) / 1200 (67.0%)
580 (32.4%) / 381 (21.3%) / 830 (46.3%) / 1791 (100%)
MALES
CASES / MAOACDH13 / 1,1 / 1,2 / 2,2
1,1 / 4 (5.3%) / 0 / 3 (4.0%) / 7 (9.3%)
1,2 / 18 (24.0%) / 0 / 13 (17.3%) / 31 (41.3%)
2,2 / 25 (33.3%) / 0 / 12 (16.0%) / 37 (49.3%)
47 (62.7%) / 0 / 28 (37.3%) / 75 (100%)
CONTROLS / MAOA
CHD13 / 1,1 / 1,2 / 2,2
1,1 / 8 (0.8%) / 0 / 12 (0.12%) / 20 (2.0%)
1,2 / 135 (13.5%) / 0 / 171 (17.0%) / 306 (30.5%)
2,2 / 278 (27.7%) / 0 / 399 (39.8%) / 677 (67.5%)
421 (42.0%) / 0 / 582 (58.0%) / 1003 (100%)
Genetic background of extreme violent behaviour1
Supplementary Figure 1.The flow-chart of the study procedure.
Genetic background of extreme violent behaviour1
SupplementaryFigure 2 a. Q-Q-plot of GWAS analysis of violent behavior.
Supplementary Figure 2 b. Q-Q-plot of GWAS analysis of extremely violent behavior.
Supplementary Figure 3. Linkage Disequilibrium structure of the CDH13SNPs in the CRIME cohort.