Supplementary Information I: Description of the Study Groups

Supplementary Material

Supplementary Information I: Description of the study groups

Dogs used for mutation screening of the coding DNA sequence

We analyzed the coding DNA sequence of the candidate genes in seven (htr1A and htr1B) or eight (htr2A and slc6A4) probands. The probands had no known common grandparents. For htr1A and htr1B we used two dogs that had bitten both humans and other dogs, one dog that had bitten humans and threatened other dogs, two dogs that had bitten only humans, and two dogs that had bitten only other dogs. Five dogs were male (all intact) and two dogs were female (both castrated). The mean age of the dogs at the time the impression of their owner was collected was 3.2 years. In all dogs, the aggression problems had started when the dog was younger than 5 years old. There were no indications that the aggressive behavior of the dogs could be explained by non-genetic factors such as poor socialization or traumatic experiences.All dogs except one were reported to be healthy by their owner. The unhealthy dog had epilepsy. For htr2A and slc6A4, we used two dogs that had bitten both humans and other dogs, one dog that had bitten humans and threatened other dogs, two dogs that had only bitten humans, one dog that had only bitten other dogs, one dog that had threatened people and other dogs, and one dog that had threatened people. Six dogs were male (two castrated) and two dogs were female (one castrated). The mean age of these dogs at the time the impression of their owner was collected was 2.9 years. In all dogs, the aggression problems had started when the dog was younger than 4 years old. There were no indications that the aggressive behavior of the affected dogs could be explained by non-genetic factors such as poor socialization or traumatic experiences. All dogs except one were reported to be healthy by their owner. The unhealthy dog had eye problems. Analysis of the CDS in three (htr1A), six (htr1B), or eight (htr2A and slc6A4) non-aggressive Golden Retrievers has been described by van den Berg et al. (2004; 2005). For htr1A and htr1B these non-aggressive dogs were unrelated to the probands. For htr2A and slc6A4 we used one non-aggressive sibling of each proband.

Dogs used for linkage analysis

As is usual in dog breeds, all probands were related to each other within a limited number of generations. For linkage analysis, small families consisting of probands and their close relatives (siblings, half-siblings, parents, grandparents and cousins) were extended by merging families if there were not more than three meioses separating aggressive dogs. In this way we created 23 families. Thirteen of the 23 families contained only one aggressive dog. Such families are not informative in an affecteds-only analysis, so they were excluded from further analysis. We also excluded one small family with two aggressive dogs because the aggressive behavior in these particular dogs may have had a medical cause. The remaining nine families were used for linkage analysis; their pedigrees are depicted in Figures S1-S9. DNA samples were available for 96 individuals in these families (31 affecteds and 65 unaffecteds).

Marker haplotype frequencies were determined in a group of 27 (htr1A and htr2A), 31 (htr1B), or 26 (slc6A4) parent dogs. We either had a DNA sample of these dogs, or we could deduce their haplotypes from haplotypes of their offspring. This group also included dogs from the 13 families that were not included in the linkage analysis. The owner impressions about human-directed aggression were converted into a dichotomous variable for linkage analysis. Probands were classified as “affected” if they scored at least 2 (i.e. threatening) on human-directed aggression (n=29). Dogs recruited as a relative of a proband were only classified as “affected” if they scored 3 (biting) on human-directed aggression (n=2). All other dogs were classified as “unaffected” (n=65).

There were 24 male affected dogs (11 castrated) and 7 female affected dogs (3 castrated). The mean age of the affected dogs at the time the impression of their owner was collected was 3.3 years. In all affected dogs, the aggression problems had started when the dog was younger than 5 years old. Owners of 15 affected dogs reported that the dog was already showing abnormal aggressive behavior when it was a puppy. We collected detailed information about the history of the dogs during the personal interviews with the owners and using mail questionnaires (van den Berg et al., 2003a; 2006). There were no indications that the aggressive behavior of the affected dogs could be explained by non-genetic factors such as poor socialization or traumatic experiences. Twenty-nine affected dogs were reported to be healthy by their owner. One dog (number 63 in family 5; see Figure S5) had eye problems and one dog (number 206 in family 3; see Figure S3) had hypothyroidy. The phenotypes of these two dogs were filled in as “unknown” in the linkage analysis.

Figure S0. Legend to the symbols in the pedigrees in Figures S1-S9. Squares represent males; circles represent females. Black symbols represent aggressive dogs and white symbols represent non-aggressive dogs. We used grey symbols for dogs of which we have no behavioral information. Arrows indicate probands, i.e. dogs that were referred to our clinic because of their aggressive behavior. Numbers and letters at the upper left corners of the symbols are identification numbers of the dogs. These numbers are in brackets if we had no DNA sample of the dog. The numbers beneath the symbols are the genotypes (htr1A) or haplotypes (htr1B, htr2A, and slc6A4) for the candidate genes (see Table S1 for a legend to the symbols).

Table S1. Legend to the codes that are used for genotypes (htr1A) and haplotypes (htr1B, htr2A, and slc6A4) in the pedigrees of the Supplementary Information I.

code in pedigrees / allele or haplotype
htr1A / 0 / unknown
1 / 297
2 / 303
htr1B / 0 / unknown
1 / 143-A-G-T-G
2 / 143-C-A-T-G
3 / 143-A-G-C-C
4 / 139-A-G-C-C
5 / 139-A-G-T-G
6 / 147-A-G-T-G
htr2A / 0 / unknown
1 / 132-C
2 / 130-C
3 / 132-T
4 / 128-C
slc6A4 / 1 / C-G
2 / T-A

Figure S1. Family 1.

Figure S2. Family 2

Figure S3. Family 3. We used “phenotype unknown” for dog 206 because this dog suffered from medical problems.

Figure S4. Family 4

Figure S5. Family 5. There are four meioses between dog 63 and the dogs in the sibship of 98. We made an exception to our “3-meioses-rule” because DNA samples from both parents were available for these two sibships. We used “phenotype unknown” for dog 63 because this dog suffered from medical problems.

Figure S6. Family 6

Figure S7. Family 7

Figure S8. Family 8

Figure S9. Family 9

Dogs used for association analysis

Case group

We selected all pedigree dogs that were aggressive to humans and that did not share grandparents with other aggressive dogs (n= 14). From the aggressive dogs that did share grandparents with other aggressive dogs, we selected dogs so that the number of dogs could be maximized, while keeping the degree of relatedness between the dogs as low as possible. We allowed dogs to share one grandparent. In addition, we allowed a dog’s parent to be the grandparent of another dog. However, we never allowed a dog to be a grandparent or parent of more than two dogs. If there was a choice between multiple dogs with the same degree of relatedness, we would select dogs with the highest estimated breeding value for human-directed aggression (Liinamo et al.2007). In this way, we selected 42 aggressive dogs with a pedigree. We added eight aggressive dogs without a pedigree to this group. These eight dogs were all euthanatized because of their aggressiveness and had high estimated breeding values for human-directed aggression.

There were 33 male (10 castrated) and 17 female (9 castrated) dogs in the aggressive group. The mean age of the aggressive dogs at the time the impression of their owner was collected was 3.7 years. In 44 dogs, the aggression problems had started when the dog was younger than 5 years old. Six dogs were 7 years or older at the time the impression of their owner was collected. For these dogs, the age of onset of the aggression problems was not recorded. We collected detailed information about the history of the dogs during the personal interviews with the owners and using mail questionnaires (van den Berg et al. 2003a; 2006). There were no indications that the aggressive behavior of the affected dogs could be explained by non-genetic factors such as poor socialization or traumatic experiences. The dogs had no diseases that could explain the aggressive behavior according to the owners.

Control group

For the control group, we first selected non-aggressive Golden Retrievers from our database. We selected dogs that did not share grandparents (n=4). We then selected dogs so that the number of dogs could be maximized from the non-aggressive dogs that did share grandparents with other non-aggressive dogs. If there was a choice between multiple dogs with the same degree of relatedness, we selected dogs with the lowest estimated breeding value for human-directed aggression (Liinamo et al.2007). The within-group relatedness was similar in the cases and controls. The number of unrelated non-aggressive dogs in our database was limited (n=25). There were 18 male (4 castrated) and 7 female dogs (3 castrated) in the non-aggressive group. The mean age of the non-aggressive dogs at the time the impression of their owner was collected was 4.3 years.

We completed the non-aggressive group with 25 dogs from the random group (15 males, 10 females). The random group consists of 194 Dutch pedigree Goldens that were born between July 2002 and February 2003. This sample was collected with the purpose of estimating the prevalence of elbow dysplasia in the Dutch Golden Retriever population. From each litter that was born between July 2002 and February 2003 and registered at the Dutch Kennel Club, we asked the owners of each third, sixth, and ninth puppy to participate in the random group. We have no behavioral phenotypes of these dogs.

Dogs used for quantitative genetic analysis

The quantitative genetic analysis was performed on 328 dogs with varying types of phenotypes available (n= 162 probands and n=166 relatives). Of these 328 dogs, 242 had a known pedigree containing at least three generations, through which 895 additional ancestral dogs without own phenotypes could be traced (see Liinamo et al. 2007 for further information). For 86 phenotyped dogs pedigree data was not available. The majority of the other dogs were connected with other families in the data through their pedigrees. There were 115 male probands (49 castrated) and 47 female probands (20 castrated). The mean age of the probands at the time the impression of their owner was collected was 3.3 years. DNA was available for 139 probands. The majority of the probands were referred because they were aggressive towards humans; a few were referred because they were aggressive towards other dogs. Owner impressions were available for156 probands. Original CBARQ items were filled out for 84 probands. Shortened CBARQ scores were available for 111 probands. CBARQ factors were available for 117 probands. There were 75 male relatives (15 castrated) and 91 female relatives (32 castrated). The mean age of the relatives at the time the impression of their owner was collected was 5.2 years. DNA was available for 142 relatives. Owner were available for 164 relatives. Original CBARQ items were filled out for 134 relatives. Shortened CBARQ scores were available for 150 relatives. CBARQ factors were available for 155 relatives.

Supplementary Information II: Additional data concerning the association studies

Figure S10: Power estimations for the association study at various marker allele frequencies. Estimations were made using Shaun Purcell’s genetic power calculator ( harvard.edu/~purcell/gpc/) with the following assumptions: number of cases = number of controls = 49; high risk allele frequency (A) = 0.1; prevalence = 0.01. The mean pairwise D’

between the SNPs flanking a candidate gene was used as an estimate of the local D’ in these estimations (mean D’=0.89 for htr1A; 0.80 for htr1B; 0.72 for htr2A; and 0.88 for slc6A4).

Calculations were performed for two different genotype relative risks: rr2 (genotype relative risk Aa=2; genotype relative risk AA=4) and rr5 (genotype relative risk Aa=5; genotype relative risk AA=10).

Table S2.

Single nucleotide polymorphisms (SNPs) used for the association study, Hardy-Weinberg p-values in controls and cases, and results of chi square or two-tailed Fisher’s exact tests for comparisons of case and control genotype frequencies. These results are based on genotype data from 49 aggressive cases and 49 control dogs.

SNP name 1 / HW
p-value
controls 2 / HW
p-value
cases 2 / Genotype comparison
p-value
BICF2P546848 / 0.80 / 0.61 / 1.0
BICF2P1051894 / 1.0 / 1.0 / 0.27
BICF2P1398268 / 0.62 / 0.80 / 0.60
BICF2S23127755 / 1.0 / 0.57 / 0.68
BICF2P1200391 / 0.81 / 0.70 / 0.69
BICF2P590055 / 0.91 / 0.66 / 0.95
BICF2S22939125 / 1.0 / 0.73 / 0.74
BICF2S23215863 / 0.75 / 1.0 / 0.83
BICF2P25993 / 0.69 / 0.20 / 0.78
Htr1A
BICF2S23442706 / 0.73 / 1.0 / 0.36
BICF2P519607 / 1.0 / 0.42 / 0.62
BICF2P1341930 / 1.0 / 0.75 / 0.75
BICF2P1159241 / 0.029 / 0.032 / 1.0
BICF2P555130 / 0.024 / 0.0041 / 0.55
BICF2S23326229 / 0.42 / 0.28 / 0.57
Htr1B
BICF2P670331 / 1.0 / 0.37 / 0.36
BICF2S23153760 / 1.0 / 0.16 / 0.37
BICF2P1426522 / 1.0 / 0.97 / 1.0
BICF2P27571 / 1.0 / 0.80 / 0.90
BICF2S23444066 / 0.97 / 0.84 / 1.0
TIGRP2P164447_rs8805986 / 1.0 / 0.16 / 0.32
BICF2P812153 / 1.0 / 0.20 / 0.14
BICF2P855402 / 1.0 / 0.20 / 0.043
BICF2G630315581 / 1.0 / 0.73 / 0.46
BICF2G630315746 / 1.0 / 0.78 / 0.63
BICF2P1168502 / 1.0 / 1.0 / 1.0
Htr2A
BICF2P164280 / 0.70 / 0.25 / 0.12
BICF2G630316047 / 0.43 / 0.69 / 0.36
BICF2S23125159 / 0.23 / 0.77 / 0.83
BICF2S23661838 / 1.0 / 0.64 / 0.72
BICF2S22954191 / 0.43 / 1.0 / 0.73
BICF2P813837 / 0.97 / 0.27 / 0.37
BICF2S23018060 / 0.17 / 0.30 / 0.90
BICF2S23551918 / 0.50 / 0.83 / 0.81
Slc6A4
BICF2S23325050 / 0.86 / 0.43 / 0.76
BICF2S23124809 / 1.0 / 0.64 / 0.75
BICF2P950384 / 1.0 / 0.21 / 0.30
BICF2S245135 / 0.59 / 0.48 / 0.95
BICF2S2347312 / 0.57 / 0.43 / 0.60
BICF2S23154457 / 0.85 / 0.42 / 0.53
BICF2S23141984 / 0.44 / 1.0 / 0.80

1 htr1A,htr1B,htr2A = respectively serotonin receptor 1A, 1B, and 2A gene; slc6A4 = serotonin transporter The genes are included in the table to show their position relative to the SNPs.

2 HW= Hardy-Weinberg