Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Jun 5;18(1):175.
doi: 10.1186/s12888-018-1753-4.

Collective Interaction Effects Associated With Mammalian Behavioral Traits Reveal Genetic Factors Connecting Fear and Hemostasis

Affiliations
Free PMC article

Collective Interaction Effects Associated With Mammalian Behavioral Traits Reveal Genetic Factors Connecting Fear and Hemostasis

Hyung Jun Woo et al. BMC Psychiatry. .
Free PMC article

Abstract

Background: Investigation of the genetic architectures that influence the behavioral traits of animals can provide important insights into human neuropsychiatric phenotypes. These traits, however, are often highly polygenic, with individual loci contributing only small effects to the overall association. The polygenicity makes it challenging to explain, for example, the widely observed comorbidity between stress and cardiac disease.

Methods: We present an algorithm for inferring the collective association of a large number of interacting gene variants with a quantitative trait. Using simulated data, we demonstrate that by taking into account the non-uniform distribution of genotypes within a cohort, we can achieve greater power than regression-based methods for high-dimensional inference.

Results: We analyzed genome-wide data sets of outbred mice and pet dogs, and found neurobiological pathways whose associations with behavioral traits arose primarily from interaction effects: γ-carboxylated coagulation factors and downstream neuronal signaling were highly associated with conditioned fear, consistent with our previous finding in human post-traumatic stress disorder (PTSD) data. Prepulse inhibition in mice was associated with serotonin transporter and platelet homeostasis, and noise-induced fear in dogs with hemostasis.

Conclusions: Our findings suggest a novel explanation for the observed comorbidity between PTSD/anxiety and cardiovascular diseases: key coagulation factors modulating hemostasis also regulate synaptic plasticity affecting the learning and extinction of fear.

Keywords: Anxiety; Behavioral genetics; Epistasis; Post-traumatic stress disorder; Quantitative trait.

Conflict of interest statement

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Figures

Fig. 1
Fig. 1
Continuous discriminant analysis for quantitative traits. Paired genotype (a)-phenotype (y) data for individuals are divided into training and test sets. The training set is used to model the conditional distribution Pr(a|y), while including the interaction effects between all m SNPs. Parameters with large magnitudes that often result from insufficient data are made unfavorable by the penalizer λ. Bayes’ rule is then used to obtain Pr(y|a) and applied to predict phenotype values for individuals in the test group. The correlation R between the predicted and actual phenotypes is optimized with respect to λ. Because of the training/test set division, R2 is in general not equal to r2, the proportion of phenotype variance explained by genetic predictors. The latter can be estimated by using the optimized penalizer and repeating the inference
Fig. 2
Fig. 2
Regularized inference of genotype-quantitative trait associations for two different sample sizes and varying penalizer values. a–d Simulated data with n = 100, where the overall dependence of prediction score R (correlation between predicted and actual phenotype values for test individuals) is shown in a, and b–d show the comparisons between predicted and true parameter values (single-SNP parameter h and interaction J for each SNP and SNP pairs, respectively) for three different penalizer λ values. Closer to the diagonal is better. Note that the condition λ = 0.1 in c optimizing R (see a) gives the best fit. e–g Analogous results for n = 104. The sample size is large enough such that overfitting under small λ is negligible. The number of SNPs was m = 5 and the dominant model was used. Parameters were generated randomly from normal distributions: h(0) ~ N(−0.3, 0.12), h(1)~ N(0.3, 0.12), J(0) ~ N(0, 0.052), and J(1) ~ N(0.1, 0.052). The phenotype values {yk} for k = 1, …, n were generated from N(0, 1) and, for each individual, the conditional genotype distribution given by Eqs. (2–3) was used to generate genotypes
Fig. 3
Fig. 3
Collective inference performance. Ridge regression (RR) and CDA were compared using simulated data. We first sampled phenotype values of n individuals from the standard normal distribution. Restricting ourselves to the number of SNPs (m ≤ 20) allowing for the enumeration of all possible genotypes, we then assigned single-SNP and interaction parameters for m = 10 (a), m = 15 (b), and m = 20 SNPs (c) from normal distributions hi(0) ~ N(0, 0.01), Jij(0) ~ N(0, 0.01), hi(1) ~ N(0, 0.01), and Jij(1) ~ N(0.1, 0.01), under the dominant model. We next calculated the genotype distribution conditional on phenotypes for all possible genotypes, and chose a genotype for each individual based on this distribution. We repeated this sampling for 100 replicates. For each data set, we applied RR and CDA collective inference, using a single penalizer λ determined by optimizing R by cross-validation (right column). Power was defined as the proportion of replicates for which P < 0.05
Fig. 4
Fig. 4
Quantile-quantile plots of behavioral traits for pathway-based SNP groups from CDA inference. a Fear conditioning (FC) in cued and context tests. b Prepulse inhibition (PPI). c Forced swim test. d Elevated plus maze. e Sleep (total duration and difference in sleep lengths between lighted and dark periods) f Fear of noise and humans/objects in dogs. Data sets used are outbred mice (a–d) and Labrador Retrievers (e–f). Colored symbols and filled symbols represent pathways with false discovery rate < 0.05 and with significance exceeding Bonferroni-corrected threshold, respectively
Fig. 5
Fig. 5
Top-ranked pathways associated with quantitative behavioral traits. Results for mice (a–g) and dogs (h) are shown. a Fear conditioning (FC) cue test. b FC context test. c Prepulse inhibition (PPI). d Elevated plus maze. e Forced swim test. f Sleep duration in 24 h. g Differences in sleep length in light (L) and dark (D) periods. h Fear of noise in dogs. Dashed red lines represent Bonferroni-corrected significance thresholds. Groups of pathways belonging to different classes are labeled with colored texts. ABC, ATP-binding cassette; Activ., activation/activates; alkyl., alkylation; assemb., assembly; biol., biology; biosynth., biosynthesis; catab., catabolism; Cdk, cyclin-dependent kinase; cell., cellular; CL, cardiolipin; clear., clearance; cleav., cleavage; cmplx., complex; cont., containing; demethyl., demethylates; devel., development/developmental; DSCAM, Down syndrome cell adhesion molecule; enab., enables; EPH, erythropoietin-producing human hepatocellular receptor; ER, endoplasmic reticulum; exec., execution; expr., expression; FA, fatty acid; facilit., facilitative; FZD, frizzled protein; gCOO, γ-carboxylation/carboxylated; glycosyl., glycosylation; GPCR, G protein-coupled receptor; homeo., homeostasis; IFN, interferon; IL, interleukin; ind., induces/induced; indep., independent; inhib., inhibition/inhibits; inter., interaction; interconv., interconversion; intermed., intermediate; ISG15, interferon-stimulated gene 15; LPC, lysophosphatidylcholine; LRRFIP1, leucine-rich repeat flightless-interacting protein 1; MAPK, mitogen-activated protein kinase; mech., mechanism; med., mediated; metab., metabolism; misc., miscellaneous; neg., negative; NFkB, nuclear factor kappa B; NLRP, NACHT, LRR and PYD domains-containing protein; oxidat., oxidation; PAO, polyamine oxidase; phosph., phosphorylation; PI, phosphatidylinositol; PIP2, phosphatidylinositol phosphate 2; pol I, polymerase I; prog., programmed; propept., propeptide; prot., protein; R., receptor/receptors; Rap, Ras-related protein; reg., regulation/regulates; remod., remodeling; remov., removal; repl., replication; resp., response; RSK, ribosomal 6 kinase; rxn., reaction; sig., signal/signaling; stimul., stimulation; synap., synaptic; synth., synthesis; sys., system; TDG, thymine-DNA glycosylase; term., terminal/terminates/termination; TET, ten-eleven translocation methylcytosine dioxygenase; TFAP, transcription factor activating enhancer binding protein; TNFR, tumor necrosis factor R; transcr., transcription; transm., transmembrane; transp., transports/transporter/transportation; TSR, thrombospondin repeat; ubiquit., ubiquitination; UFA, unsaturated fatty acid
Fig. 6
Fig. 6
Independent-SNP and collective association levels of variants contributing to pathways. Those highly ranked for mouse behavioral traits are shown. a Manhattan plot for fear conditioning, showing single-SNP p-values for linear regression. The mouse SNPs for genes in two pathways, Effects of PIP2 hydrolysis and γ-carboxylation of protein precursors are shown in color. Horizontal lines show the collective inference p-values for these two pathways. b–c Detailed views of two loci contributing to pathways in a. d–f Prepulse inhibition and three pathways, Platelet homeostasis, Serotonin clearance from synaptic cleft, and Metallothioneins bind metals. The collective p-values of the latter two pathways (bottom horizontal lines) are indistinguishable. Filled rectangles represent the coding regions of genes indicated
Fig. 7
Fig. 7
Independent-SNP and collective association levels of variants contributing to pathways associated with elevated plus maze. a Manhattan plot and variants in three pathways, Interconversion of polyamines, Hydrolysis of lysophosphatidylcholine, and Interleukin-10 signaling. b–c Detailed views of two loci contributing to pathways in a. Filled rectangles represent the coding regions of genes indicated
Fig. 8
Fig. 8
Broad-sense heritability of pathways compared to proportion of additive variance explained. a Fear conditioning (cue test) in mice. b Prepulse inhibition (PPI) in mice. The top-ranked pathways in Fig. 5a,c are shown in the same order. CDA values represent r2 estimated using regularization conditions determined from cross-validation applied to half of the whole cohort and repeating the inference for the other half. Error bars represent the 95% c.i. The GCTA and LDAK outcomes represent the proportion of variance explained by the same set of SNPs but without interaction effects. For pathways in which the GCTA/LDAK p-values were higher than 0.05, the proportion of variance was set to zero. The CDA p-values are all smaller than 10−3 (Fig. 5)

Similar articles

See all similar articles

References

    1. Bendesky A, Kwon YM, Lassance JM, Lewarch CL, Yao S, Peterson BK, et al. The genetic basis of parental care evolution in monogamous mice. Nature. 2017;544:434–439. doi: 10.1038/nature22074. - DOI - PMC - PubMed
    1. Weber JN, Peterson BK, Hoekstra HE. Discrete genetic modules are responsible for complex burrow evolution in Peromyscus mice. Nature. 2013;493:402–405. doi: 10.1038/nature11816. - DOI - PubMed
    1. Sousa N, Almeida OF, Wotjak CT. A hitchhiker's guide to behavioral analysis in laboratory rodents. Genes Brain Behav. 2006:5 Suppl 2:5–24. - PubMed
    1. Wang GD, Xie HB, Peng MS, Irwin D, Zhang YP. Domestication genomics: evidence from animals. Annu Rev Anim Biosci. 2014;2:65–84. doi: 10.1146/annurev-animal-022513-114129. - DOI - PubMed
    1. Ilska J, Haskell MJ, Blott SC, Sanchez-Molano E, Polgar Z, Lofgren SE, et al. Genetic characterization of dog personality traits. Genetics. 2017;206:1101–1111. doi: 10.1534/genetics.116.192674. - DOI - PMC - PubMed

Publication types

MeSH terms

Feedback