Epigenetic and genetic components of height regulation

Adult height is a highly heritable trait. Here we identified 31.6 million sequence variants by whole-genome sequencing of 8,453 Icelanders and tested them for association with adult height by imputing them into 88,835 Icelanders. Here we discovered 13 novel height associations by testing four different models including parent-of-origin (|β|=0.4–10.6 cm). The minor alleles of three parent-of-origin signals associate with less height only when inherited from the father and are located within imprinted regions (IGF2-H19 and DLK1-MEG3). We also examined the association of these sequence variants in a set of 12,645 Icelanders with birth length measurements. Two of the novel variants, (IGF2-H19 and TET1), show significant association with both adult height and birth length, indicating a role in early growth regulation. Among the parent-of-origin signals, we observed opposing parental effects raising questions about underlying mechanisms. These findings demonstrate that common variations affect human growth by parental imprinting.

Locus plot depicting variants at the ADAMTSL4 locus (hg38) associating with adult height. chr1:150553749[C] is labelled as a diamond and shown in purple, and other variants are colored according to correlation (r 2 ) with that marker (legend at top-left). -log 10 P-values are shown along the left y-axis and correspond to the variants depicted in the plot. The right y-axis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.
Locus plot depicting variants at the ADAMTS17 locus associating with adult height. The leading variant, rs72755233 [A], is labelled as a diamond and shown in purple, other variants are colored according to correlation (r 2 ) with the leading marker (legend at top-left). -log 10 P-values are shown along the left y-axis and correspond to the variants depicted in the plot. The right y-axis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.
Locus plot depicting variants at the ADAMTS10 locus associating with adult height. The leading variant, rs62621197 [T], is labelled as a diamond and shown in purple, other variants are colored according to correlation (r 2 ) with the leading marker (legend at top-right). -log 10 P-values are shown along the left yaxis and correspond to the variants depicted in the plot. The right y-axis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.
Locus plot depicting variants at the GH1 locus (hg38) associating with adult height. The missense variant, chr17:63918372 [C], is labelled as a diamond and shown in purple, and other variants are colored according to correlation (r 2 ) with that marker (legend at top-right). -log 10 P-values are shown along the left y-axis and correspond to the variants depicted in the plot. The right y-axis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.
Locus plots depicting variants at the LECT2 locus (hg38) associating with adult height using the (a) additive, and (b) recessive model. Genome-wide significance is observed under the recessive model, but not under the additive model. In both plots, the leading variant under the recessive model, rs62623707 [G], is labelled as a diamond and shown in purple, and other variants are colored according to correlation (r 2 ) with this marker (legend at top-right). -log 10 P-values are shown along the left y-axis and correspond to the variants depicted in the plot. The right y-axis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.

Supplementary Fig. 14.
Locus plot depicting rs7482510[G] at 11p15.5 (hg38) associating with adult height using the (a) paternal and (b) maternal models. Genome-wide significance is observed for rs7482510[G] when paternally inherited but not under the maternal model. rs7482510[G] is labelled as a diamond and shown in purple and other variants are colored according to correlation (r 2 ). -log 10 P-values are shown along the left y-axis and correspond to the variants depicted in the plot. The right y-axis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.
Locus plot depicting rs143840904[T] at the KCNQ1 locus (hg38) associating with adult height using the (a) paternal and (b) maternal models. Genome-wide significance is observed for rs143840904[T] when maternally inherited but not under the paternal model. rs143840904[T] is labelled as a diamond and shown in purple and other variants are colored according to correlation (r 2 ). -log 10 P-values are shown along the left y-axis and correspond to the variants depicted in the plot. The right y-axis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.

Supplementary Fig. 16.
Locus plot depicting variants at the RTL1 locus (hg38) associating with adult height using the (a) paternal and (b) maternal models. Genome-wide significance is observed under the paternal model, but not under the maternal model. In all plots, the leading variant under the paternal model, rs41286560 [T], is labelled and shown in purple, and other variants are colored according to correlation (r 2 ) with that marker. -log 10 P-values are shown along the left y-axis and correspond to the variants depicted in the plot. The right yaxis shows calculated recombination rates at the chromosomal location, plotted as a solid grey line.
Scatter plot showing how the 697 variants reported by GIANT 1 effect adult height (x-axis) and length at birth (y-axis) in our data. The effect allele (minor or major) of each variant was concluded to be the one that has a positive effect on adult height in our data. The black solid line, y = -0.0006 + 0.37x + e, represents weighted linear regression with R 2 = 0.05 using MAF(1-MAF) as weights. Different colors of points represent their P-values. Red points represent variants that associate with both phenotypes with P<0.05, green points represent variants that associate with both phenotypes with P>0.05, blue points represent variants that associate only with adult height with P<0.05 and purple points represent variants that associate only with birth length with P<0.05.
LD score regression plot for height, assuming an additive, recessive, paternal and maternal genotypic model respectively. Each point represents an LD score quantile. The x-axis coordinates represent the mean LD score of a quantile and the y-axis the mean χ 2 statistic of the quantile. The blue line is the LD score regression line.
Frequency of height associated phenotypes in the OMIM database*. Tabulated is the frequency of the phenotypic term "Abnormality of body height" and its descendant terms two subclasses down in the human phenotype ontology. Only phenotypic terms that are a typical feature of a OMIM disease were regarded (typical = occourring in > 50 % of patients).

Supplementary Discussion
Novel signals detected under the additive model Nine novel height association signals were identified using the additive model. One is common (MAF>5%), four are low frequency (MAF = 0.5-5%) and four are rare (MAF < 0.5%) ( Table 1).
Five are in genes linked to monogenic growth defects (NPR2, ACAN, GH1, ADAMTS10 and ADAMTS17) (Supplementary Table 3). The nine novel variants are shortly described below. and was not observed outside of Iceland (Table 1, Supplementary Data 5, Supplementary Table 4 and Supplementary Fig. 5). This amounts to heterozygous carriers being 10.1 cm shorter than the population average. Other mutations in NPR2 are known to cause Maroteaux-type acromelic dysplasia under a recessive mode of inheritance 4 (Supplementary Table 3). The observed association with less height is consistent with the reported heterozygous effect on height (-1.8 SD) in relatives of a proband with this dysplasia 5 and findings in individuals with idiopathic short stature 6,7 . We observed no homozygous carriers of p.Gly869Val in our data.