Determinants of Menarche Timing: Insights from Twin Cohorts and Population Genetic Studies = 초경은 어떻게 결정되고, 무엇이 그 시기를 달라지게 하는가? : 쌍둥이 코호트와 인구 기반 유전 연구를 통한 통합적 고찰

국문 초록 (Abstract)

목적: 본 연구는 쌍둥이 연구, 극단 표현형 기반 전장유전체연관분석(extreme phenotype GWAS), 그리고 변이성 전장유전체연관분석(variability GWAS, vGWAS)을 상호보완적으로 활용하여 초경 연령(Age at Menarche, AAM)의 유전적 구조를 규명하고자 하였다.

방법: 쌍둥이 쌍 자료를 분석하여 조기 초경과 지연 초경(early menarche, EM; late menarche, LM) 간 유전 및 환경 요인의 이질성을 평가하고, 일란성 쌍둥이 내 표현형 불일치에 영향을 미치는 요인을 탐색하였다. 또한 극단 표현형 GWAS를 통해 EM과 LM 간의 유전 좌위 및 생물학적 경로를 비교하였으며, vGWAS를 수행하여 평균 효과와 독립적으로 표현형 변이성에 기여하는 유전 요인을 규명하였다.

결과: 쌍둥이 분석 결과, LM에서 EM보다 높은 유전력이 관찰되었다. AAM 불일치는 전 세계 일란성 쌍둥이에서 빈번하게 나타났으며, 아동기 신체활동과 신장이 지연 초경과 유의하게 연관되었다. 극단 표현형 GWAS에서는 초경 시기 양 극단이 서로 다른 생물학적 기전을 통해 조절됨이 확인되었다. 조기 초경은 전사 활성 가속 및 발달 프로그램 조절 기전과 관련되었으며, 지연 초경은 DNA 복구, 유전체 안정성, 세포 보호 과정과 연관된 경로가 강조되었다. 또한 조기 초경 관련 유전 신호는 대사적 교란 요인에 더 민감한 반면, 지연 초경 관련 유전 신호는 보다 안정적으로 유지되었다. vGWAS 분석에서는 평균 효과와 독립적으로 초경 시기의 변이성에 영향을 미치는 유전 좌위가 확인되었으며, 특히 지연 초경 관련 변이에서 표현형 분산 효과가 유의하게 관찰되었다. 이는 지연 초경이 양적 조절 기전에 의해, 조기 초경은 보다 결정론적 발달 기전에 의해 조절될 가능성을 시사한다.

결론: 초경 연령은 서로 독립적인 다수의 유전 시스템에 의해 조절되며, 조기 초경과 지연 초경은 서로 다른 병인 및 유전 구조를 갖는 별개의 생물학적 현상임을 확인하였다. 조기 초경은 환경 및 대사 요인에 민감한 맥락 의존적 기전에 의해 조절되는 반면, 지연 초경은 보다 견고한 유전적 기전에 의해 표현형 변이를 형성한다. 본 연구 결과는 전통적인 GWAS 접근법에 대한 재고를 요구하며, 사춘기 발달 이상에 대한 임상적 관리 전략 수립에 중요한 시사점을 제공한다.

다국어 초록 (Multilingual Abstract)

Objective: To investigate the genetic architecture of age at menarche (AAM) using complementary twin study, extreme phenotype genome-wide association study (GWAS), and variability GWAS (vGWAS) approaches.

Methods: This study analyzed twin pairs to understand the heterogeneity of genetic and environmental architecture for early and late menarche (EM, LM) and factors associated with twin discordance. Extreme phenotype GWAS compared genetic loci and functional pathways between EM and LM. Finally, vGWAS analysis of AAM identified genetic determinants of phenotypic variability.

Results: Twin analysis revealed higher heritability for LM compared to EM. AAM discordance was prevalent globally among monozygotic twin pairs, with childhood activity and height predicting later menarche. Extreme GWAS identified fundamentally distinct biological pathways underlying timing extremes: early menarche involved transcriptional acceleration and developmental programming mechanisms, while late menarche emphasized DNA repair, genome stability, and protective cellular processes. Early menarche genetics showed greater sensitivity to metabolic confounding, while late menarche maintained more robust genetic signals. vGWAS analysis identified loci influencing timing variability independently of mean effects. Late menarche variants correlated significantly with variance effects, while early menarche variants did not, indicating quantitative versus deterministic developmental control mechanisms respectively.

Conclusions: AAM is controlled by multiple independent genetic systems. Early and late menarche represent distinct etiology with different genetic architectures: early timing involves context-sensitive mechanisms prone to metabolic confounding, while late timing employs robust mechanisms that create phenotypic variability. These findings challenge traditional GWAS approaches and have implications for clinical management of pubertal disorders.

목차 (Table of Contents)

• I. INTRODUCTION 10
• 1.1 BACKGROUND OF MENARCHE 10
• 1.2 EPIDEMIOLOGY AND BIOLOGICAL MECHANISMS 11
•  1.2.1 Secular and cross-population trends 11
•  1.2.2 Neuroendocrine and metabolic regulation 12
•  1.2.3 Genetic determinants 13
•  1.2.4 Environmental and lifestyle determinants 13
• 1.3 PREVIOUS FINDINGS AND CLINICAL IMPLICATIONS 15
• 1.4 CURRENT KNOWLEDGE GAPS 16
• 1.5 RESEARCH QUESTIONS AND HYPOTHESES 17
•  1.5.1 Research questions 17
•  1.5.2 Hypotheses 17
•  1.5.3 Analytical framework 18
• II. WHAT EXPLAINS THE ONSET OF MENARCHE? : INSIGHTS FROM TWIN COHORT STUDIES 21
• 2.1 INTRODUCTION 21
• 2.2 METHODS 23
•  2.2.1 Study population and measurements 23
•  2.2.2 Ascertainment and standardization of age at menarche 23
•  2.2.3 Pairwise twin questions on developmental risk factors (CTP) 23
•  2.2.4 Partitioning AAM variance into three compartments 24
•  2.2.5 Full equation of population-level marginal contribution 24
•  2.2.6 Changes in the contributions from three components across menarche timing 26
•  2.2.7 Cotwin-control analysis for discordance and developmental risk factors for EM 27
• 2.3 RESULTS 28
•  2.3.1 Study participants and AAM characteristics 28
•  2.3.2 Classical genetic analysis across the peripubertal age 34
•  2.3.3 Contributions from risk categories to AAM variance 37
•  2.3.4 Developmental risk factors for AAM discordance and earlier menarche in MZT 39
• 2.4 DISCUSSION 42
•  2.4.1 Frequent AAM discordance in twins and the role of recall error 42
•  2.4.2 Relatively small C: implications for prioritizing AAM determinants 43
•  2.4.3 Developmental factors associated with EM 43
•  2.4.4 Early vs. late menarche: late menarche is not simply the reverse of early menarche 43
•  2.4.5 Limitations of the study 44
• 2.5 CONCLUSION 44
• III. GENETIC ARCHITECTURE OF EXTREME PUBERTAL TIMING: DISTINCT PATHWAYS UNDERLYING EARLY AND LATE MENARCHE 46
• 3.1 INTRODUCTION 46
• 3.2 METHODS 48
•  3.2.1 Study populations and phenotype 48
•  3.2.2 Genotyping, imputation, and quality control procedures 48
•  3.2.3 Genome-wide association study design and statistical models 50
•  3.2.4 Trans-ancestry meta-analysis procedures 51
•  3.2.5 Genomic locus definition and variant annotation 51
•  3.2.6 Statistical fine-mapping and credible set construction 52
•  3.2.7 Gene prioritization strategy 52
•  3.2.8 Tissue expression and pathway enrichment analysis 53
•  3.2.9 eQTL integration and colocalization analysis 54
•  3.2.10 Mendelian randomization analysis and drug target tractability assessment 55
• 3.3 RESULTS 56
•  3.3.1 Study characteristics and phenotype distribution 56
•  3.3.2 Genome-wide association results and heritability estimation 59
•  3.3.3 Gene-based association results 64
•  3.3.4 MAGMA tissue expression analysis results 67
•  3.3.5 Developmental expression analysis 67
•  3.3.6 Partitioned heritability analysis results 68
•  3.3.7 Gene prioritization results 71
•  3.3.8 Expression of prioritized genes 74
•  3.3.9 Pathway enrichment analysis 76
•  3.3.10 Colocalization and Mendelian randomization 79
• 3.4 DISCUSSION 83
•  3.4.1 Differential impact of adult BMI adjustment on genetic discovery 83
•  3.4.2 Tissue-specific architecture reveals different sites of genetic action 83
•  3.4.3 Pathway enrichment provides biological context 84
•  3.4.4 Mendelian randomization evidence is limited and asymmetric 85
•  3.4.5 Integration with twin-based heritability findings 85
•  3.4.6 Limitations 86
•  3.4.7 Future directions for mechanistic understanding 86
• 3.5 CONCLUSION 87
• IV. VARIANCE QUANTITATIVE TRAIT LOCI ANALYSIS REVEALS INDEPENDENT GENETIC CONTROL OF AGE AT MENARCHE VARIABILITY 88
• 4.1 INTRODUCTION 88
• 4.2 METHODS 91
•  4.2.1 Study samples and phenotype definition 91
•  4.2.2 Genome-wide vQTL analysis 91
•   Levene’s test 91
•   Regression-based framework 92
•  4.2.3 Colocalization of vQTL and EM/LM GWAS signals 92
•  4.2.4 Directional consistency of EM and LM GWAS signals 93
•  4.2.5 Regression of vQTL signals on EM/LM mean effects 93
•  4.2.6 Stratified analyses of vQTL loci 94
• 4.3 RESULTS 95
•  4.3.1 Discovery of vQTL loci of AAM 95
•  4.3.2 Colocalization of vQTL and EM/LM GWAS signals 98
•  4.3.3 Relationship between variance and mean effects 98
•  4.3.4 Stratified analyses by early-life body size and height 101
• 4.4 DISCUSSION 103
•  4.4.1 Variance as an independent dimension of genetic architecture 103
•  4.4.2 Asymmetric relationship with phenotypic extremes 103
•  4.4.3 Limited evidence for gene–environment interactions 104
•  4.4.4 Candidate genes and biological mechanisms 105
•  4.4.5 Limitations 105
• 4.5 CONCLUSION 106
• V. CONCLUSION 107
• 5.1 SUMMARY OF FINDINGS 107
• 5.2 INTEGRATIVE INTERPRETATION: A DUAL-CONDITION FRAMEWORK FOR MENARCHE 108
• 5.3 WHAT THIS STUDY EXTENDS BEYOND PREVIOUS RESEARCH 109
•  5.3.1 Moving from a single liability model to heterogeneity in developmental architecture 109
•  5.3.2 Clarifying the role of metabolic cues vs intrinsic timing mechanisms 109
•  5.3.3 Introducing variance-sensitive genetic architecture into pubertal timing research 110
• 5.4 IMPLICATIONS 110
•  5.4.1 For genetics of complex traits 110
•  5.4.2 For clinical translation 110
•  5.4.3 Public-health relevance 111
• 5.5 FINAL CONCLUSION 111
• VI. REFERENCES 112