More than 100 loci have been identified for age at menarche by genome-wide association
studies; however, collectively these explain only B3% of the trait variance. Here we test two
overlooked sources of variation in 192,974 European ancestry women: low-frequency proteincoding
variants and X-chromosome variants. Five missense/nonsense variants (in ALMS1/
LAMB2/TNRC6A/TACR3/PRKAG1) are associated with age at menarche (minor allele frequencies
0.08–4.6%; effect sizes 0.08–1.25 years per allele; Po5 10 8). In addition, we
identify common X-chromosome loci at IGSF1 (rs762080, P ¼ 9.4 10 13) and FAAH2
(rs5914101, P ¼ 4.9 10 10). Highlighted genes implicate cellular energy homeostasis, posttranscriptional
gene silencing and fatty-acid amide signalling. A frequently reported mutation
in TACR3 for idiopathic hypogonatrophic hypogonadism (p.W275X) is associated with 1.25-
year-later menarche (P ¼ 2.8 10 11), illustrating the utility of population studies to estimate
the penetrance of reportedly pathogenic mutations. Collectively, these novel variants explain
B0.5% variance, indicating that these overlooked sources of variation do not substantially
explain the ‘missing heritability’ of this complex trait.