medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1 Genome-Wide Association Studies of Coffee Intake in UK/US Participants of European 2 Ancestry Uncover Gene-Cohort Influences 3 Hayley H A Thorpe1,2, Pierre Fontanillas3, Benjamin K Pham4, John J Meredith4, Mariela V 4 Jennings4, Natasia S Courchesne-Krak4, Laura Vilar-Ribó5, Sevim B Bianchi4, Julian Mutz6, 5 23andMe Research Team2, Sarah L Elson2, Jibran Y Khokhar1,2, Abdel Abdellaoui7, Lea K 6 Davis8,9,10, Abraham A Palmer4,11, & Sandra Sanchez-Roige4,9 1 7 Department of Anatomy and Cell Biology, Schulich School of Medicine and Dentistry, Western 8 University, London, ON, Canada. 2 9 Department of Biomedical Sciences, Ontario Veterinary College, University of Guelph, Guelph, 10 ON, Canada 3 11 23andMe, Inc., Sunnyvale, CA, USA 4 12 Department of Psychiatry, University of California San Diego, La Jolla, CA, USA 5 13 Psychiatric Genetics Unit, Group of Psychiatry, Mental Health and Addiction, Vall d’Hebron 14 Research Institute (VHIR), Universitat Autònoma de Barcelona, Barcelona, Spain 6 15 Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, Psychology & 16 Neuroscience, King's College London, London, UK 7 17 Department of Psychiatry, Amsterdam UMC, University of Amsterdam, Amsterdam, The 18 Netherlands 8 19 Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN, 20 USA 9 21 Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, USA 10 22 Department of Psychiatry and Behavioral Sciences, Vanderbilt University Medical Center, 23 Nashville, TN, USA 11 24 Institute for Genomic Medicine, University of California San Diego, La Jolla, CA, USA 1 NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 25 Corresponding author: Sandra Sanchez-Roige, PhD, Department of Psychiatry, University of 26 California San Diego, San Diego, CA, USA. Email: sanchezroige@ucsd.edu 27 2 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 28 Abstract 29 Coffee is one of the most widely consumed beverages. We performed a genome-wide 30 association study (GWAS) of coffee intake in US-based 23andMe participants (N=130,153) and 31 identified 7 significant loci, with many replicating in three multi-ancestral cohorts. We examined 32 genetic correlations and performed a phenome-wide association study across thousands of 33 biomarkers and health and lifestyle traits, then compared our results to the largest available 34 GWAS of coffee intake from UK Biobank (UKB; N=334,659). The results of these two GWAS were 35 highly discrepant. We observed positive genetic correlations between coffee intake and 36 psychiatric illnesses, pain, and gastrointestinal traits in 23andMe that were absent or negative in 37 UKB. Genetic correlations with cognition were negative in 23andMe but positive in UKB. The only 38 consistent observations were positive genetic correlations with substance use and obesity. Our 39 study shows that GWAS in different cohorts could capture cultural differences in the relationship 40 between behavior and genetics. 3 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 41 Introduction 42 Coffee is a leading global food commodity that has psychoactive properties that are largely 43 due to the presence of caffeine1. While rates of use and daily intake varies widely by geographic 44 region, it is estimated that approximately 60-85% of adults in Europe and the United States 45 consume between 0.6 to 5.5 cups of coffee daily2-4. Intake of coffee and its bioactive constituents 46 is associated with benefits on cognitive function5 and lower risk of liver disease6,7 (but see8), 47 Parkinson’s and other neurodegenerative diseases6,7,9, cardiovascular disease6,7, type II 48 diabetes6,7, and certain cancers6,7,10. However, coffee intake is also associated with higher risks 49 for some adverse outcomes, including increased risk of other substance use and misuse11-14, 50 some cancers (e.g., lung cancer7,10,15), poor lipid profile6,7, pregnancy loss6,7, gastrointestinal 51 maladies16, and worse cardiovascular outcomes following excessive intake17. Given the 52 widespread and regular intake of coffee across the globe, addressing the full spectrum of 53 correlations with health and disease is an important but challenging task. 54 Genetic studies offer a compelling avenue to investigate the relationships between coffee 55 intake and other complex traits. Twin studies that calculate genetic contributions to daily coffee 56 intake estimate it to be 36-56% heritable, suggesting that coffee intake should be amenable to 57 genetic analysis. Whereas phenotypic correlations, which depend on measuring two or more traits 58 in the same cohort, can arise from genetic and environmental factors, genetic correlations assess 59 genetically driven relationships using the results from genome-wide association studies (GWAS) 60 and can therefore examine correlations between two or more traits, even if they were measured 61 in entirely non-overlapping cohorts. In the past decade, over a dozen GWAS (N=1,207-407,072) 62 have examined coffee intake18-34. Several of these GWAS have found associations with single 63 nucleotide polymorphisms (SNPs) within or near genes that metabolize caffeine (Supplementary 64 Table 1), such as CYP1A1 and CYP1A218-20,23-26,30,32,33. Some of these loci are also associated 65 with other complex traits, including liver disease35-37, cancers38-41, and alcohol consumption42-44. 4 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 66 This pleiotropy could suggest that these other associations are mediated by coffee intake or that 67 these loci also influence these traits via alternative independent mechanisms. Genetic 68 correlations have also been reported between coffee intake with other substance use45,46, reduced 69 gray matter volumes18, psychiatric illness45, osteoarthritis47, sleep48, body mass index (BMI)49, 70 type II diabetes49, and migraine50. However, some genetic correlations were conducted under a 71 priori justification (e.g., other substance use traits, sleep) and as such may fail to capture the full 72 scope of genetic correlations between coffee with other traits. Thus, a data-driven examination of 73 trait associations with coffee intake remains unexplored. 74 While coffee is the primary source of caffeine for many, other common dietary sources of 75 caffeine include tea, soft drinks, and chocolate. Consequently, when we refer to coffee intake, we 76 mean explicit measures of coffee intake (e.g., measured as cups/day) and not caffeine intake 77 unless otherwise specified. Intake of other caffeine sources also varies by geographic region 78 based on beverage sales2. For example, tea (rather than coffee) is the preferred source of caffeine 79 in the United Kingdom (UK; tea vs. coffee: ~50% vs. 20%) compared to the United States (US; 80 ~10% vs. 30%)2. As some genetic studies used data from the UK Biobank (UKB) only18,47,48,51-53 81 or combined cohorts across regions with different patterns of caffeinated beverage intake 82 (Supplementary Table 1)32,33,46, this distinction may limit generalizability or introduce 83 environmental and cultural confounds that affect the genetic associations between coffee intake 84 and other traits. 85 In this study, we used survey responses from US-based 23andMe, Inc. research 86 participants of European ancestry (N=130,153) and performed a GWAS of a single item “How 87 many 5-ounce (cup-sized) servings of caffeinated coffee do you consume each day?”. Using 88 genetic correlations and phenome- and laboratory-wide association studies (PheWAS, 89 LabWAS), we explored the relationships between coffee intake and thousands of biomarkers, 90 health features, and lifestyle traits to provide a fuller inventory of genetic correlations with coffee 5 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 91 intake. We compared our findings from the 23andMe cohort to those from the UKB using publicly 92 available GWAS summary statistics of coffee intake (“How many cups of coffee do you drink each 93 day? (Include decaffeinated coffee)”, N=334,659, http://www.nealelab.is/uk-biobank/). Although 94 we had originally intended to perform a meta-analysis, our results revealed a lower-than-expected 95 genetic correlation between coffee intake in the two cohorts; therefore, we instead used these 96 datasets to explore cohort differences in coffee intake across these two distinct populations. 97 Results 98 GWAS in the 23andMe US-based cohort replicated seven loci implicated in coffee intake 99 Participant demographics of the 23andMe cohort are described in Supplementary Table 100 2. The cohort was 65% male, had a mean age of 52.8 ± 16.9 years old, and an average BMI of 101 28.38 ± 6.54 (range: 14.0-69.1), similar to the US average of 27.5 (95% CI: 25.5-29.4)54. The 102 average coffee intake in the cohort was 1.98 (± 2.35 SD) cups per day, similar to the coffee intake 103 distributions in UKB (2.14 ± 2.09 SD; see Supplementary Figure 1 and Supplementary Table 104 3 for distributions). 105 We conducted a GWAS of 14,274,006 imputed genetic variants assuming an additive 106 genetic model that included age, sex, the first five genetic principal components, and indicator 107 variables for genotype platforms as covariates (Supplementary Table 4; Supplementary 108 Material for additional genotyping and GWAS details). The genomic control inflation factor of the 109 GWAS was λ=1.09, suggesting no substantial inflation due to population stratification. SNP- 110 heritability of coffee intake via Linkage Disequilibrium score regression (LDSC) was 7.57% ± 0.59 111 (Supplementary Table 5). 112 We identified seven genome-wide significant (p<5.00E-08) independent (r2<0.1) loci that 113 were associated with coffee intake (Figure 1, Table 1; Supplementary Figures 2-8 for locus 114 zoom plots). These associations replicated prior coffee or caffeine GWAS findings 6 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 115 (Supplementary Table 6). For example, rs2472297 (p=3.60E-65, chr15q24.1) is in the intergenic 116 region between CYP1A1 and CYP1A2, and has been previously associated with coffee and 117 caffeine intake18,20,24,26,28,30,32,33,55. CYP1A1 and CYP1A2 encode members of the cryptochrome 118 P450 superfamily of enzymes involved in xenobiotic metabolism22. rs2472297 has also been 119 previously associated with traits like alcohol consumption56,57, clozapine pharmacokinetics58, 120 kidney function59-62, and the concentration of biomarkers in urine63-68. We also identified 121 rs4410790 (p=5.20E-55, chr7p21.1), which is located upstream of the AHR gene encoding a 122 transcription factor that regulates CYP1A1/CYP1A2 and is activated by polycyclic aromatic 123 hydrocarbons, which are present in coffee22,69. Prior studies associated rs4410790 and caffeine 124 intake from tea20, as well as with traits like caffeine metabolism28, bitter beverage intake26, and 125 urine biomarkers64,66-68,70. Lastly, rs199612805 (p=1.80E-10, chr22q11.23), which is located near 126 ADORA2A, was also implicated in coffee intake. This variant was recently associated with 127 caffeine intake from tea and coffee in the UKB20. ADORA2A encodes an adenosine G-protein 128 coupled receptor that is inhibited by caffeine to produce stimulating effects71. The remaining four 129 SNPs – rs34645063, rs28634426, rs117824460, and rs11474881 – were in linkage disequilibrium 130 (LD) with SNPs previously identified by other coffee or caffeine GWAS18,20,24,26,28,30,55. rs34645063 131 (p=3.30E-09, chr6q16.1) is a deletion/insertion polymorphism between MMS22L and POU3F2. 132 rs34645063 is in LD (R2=0.74) with rs754177720 and is also associated with caffeine intake from 133 coffee or tea20. rs28634426 (p=2.10E-10, chr7q11.23) is an intronic variant of STYXL1 in LD with 134 rs17685 (R2=0.78) and rs1057868 (R2=0.76), which were previously implicated by coffee 135 GWAS18,20,24,26,30,55. rs117824460 (p=1.70E-08, chr19q13.2) is an intronic variant of CYP2A6, and 136 is in LD (R2=0.05) with rs56113850, which was implicated in coffee intake20,26 and caffeine 137 metabolism28. CYP2A6 encodes a cryptochrome P450 superfamily enzyme member that 138 metabolizes nicotine72; rs117824460 has also been associated with smoking traits57,73 and serum 139 albumin74, C-reactive protein75, and liver alkaline phosphatase levels66. The final significant 140 variant we identified, rs11474881 (chr20q13.33, p=1.10E-08), is an intronic variant of the 7 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 141 PCMTD2 gene; rs11474881 is in LD (R2=0.98) with rs6062679, which was previously implicated 142 in coffee and tea intake and bitter beverage consumption26. We used three additional multi- 143 ancestral cohorts to replicate these findings (Table 1; Supplementary Table 3). Of the SNPs that 144 passed QC, all replicated in a larger sample of 23andMe research participants of European 145 ancestry (N=689,661), all replicated in those with African American ancestry (N=32,312), and one 146 replicated in those of Latin American ancestry (N=124,155). 147 We also report several notable nominal associations with coffee intake (p<1.00E-06, 148 Supplementary Table 6). rs72790130 (p=5.50E-08, chr16q23.3) and rs2155645 (p=9.80E-07, 149 chr11q23.2) are intronic variants of two cell adhesion molecule genes, CDH13 and NCAM1, 150 respectively. Both genes have been previously implicated in substance use traits by other 151 GWAS20,57,76-82 and candidate gene studies in humans and animal models83-88. rs2155645 is also 152 in LD with rs2298527 (R2=0.43), which was previously implicated in daily caffeine intake from 153 coffee20. rs11204734 (p=2.90E-07, chr1q21.3) is an intronic variant of ARNT; its protein 154 heterodimerizes with AHR and binds to xenobiotic response elements to regulate transcription of 155 CYP1A1 and CYP1A222. Finally, rs340019 (p=2.10E-07, chr15q22.2) is an intronic variant of 156 RORA, which is involved in circadian rhythm and metabolic regulation, among other functions89. 157 rs340019 is in LD with rs12591786 (R2=0.25), which was implicated in daily caffeine intake from 158 cups of coffee and tea20. 8 159 Table 1. Significant (p<5.00E-08) GWAS results for coffee intake from 23andMe research participants (N=130,153) of European 160 ancestry (EA). Replication (EA Rep) was conducted in an additional cohort of 23andMe participants of EA (N=689,661), and in those 161 with African American ancestry (AA; N=32,312) and Latin American ancestry (LA; N=124,155); *SNPs that did not pass QC in 162 replication. See Supplementary Table 6 for additional information. AA EA EA LA Allele p EA Rep p AA p LA p Nearest SNP BP Cytoband Effect Rep EA s value value value EA value gene(s) EAF EAF F F 7502788 3.60E 3.09E 0.0 0.1 CYP1A1, rs2472297 C/T chr15q24.1 0.23 0.08 1.28E-234* 0.24 5.47E-47* 0 -65 -05 7 4 CYP1A2 1728457 5.20E 4.52E 0.5 0.5 AGR3, rs4410790 C/T chr7p21.1 0.38 -0.06 7.58E-212* 0.38 9.89E-60* 7 -55 -15 2 5 AHR rs19961280 2484399 chr22q11.2 1.80E 0.01 1.28E 0.0 0.0 ADORA2 D/I 0.01 -0.10 2.48E-29 6.32E-13 5 1 3 -10 5 -07 8 2 A, UPB1 7567559 2.10E 0.3 0.2 rs28634426 G/T chr7q11.23 0.20 0.03 3.08E-16 0.24 0.08 8.17E-06 STYXL1 4 -10 2 7 9 9859107 3.30E 0.6 0.5 MMS22L, rs34645063 D/I chr6q16.1 0.47 -0.02 5.23E-16 0.48 0.05 4.35E-06* 5 -09 5 8 POU3F2 6289295 chr20q13.3 1.10E 0.4 0.6 rs11474881 D/I 0.55 -0.02 2.36E-12 0.55 .21 0.02* PCMTD2 6 3 -08 5 2 CTC- rs11782446 4137148 1.70E 0.0 0.0 A/G chr19q13.2 0.03 -0.06 2.00e-07 0.03 0.08 0.08 490E21.1 0 0 -08 1 2 2 163 10 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 164 165 Figure 1. GWAS and secondary analyses of coffee intake from the 23andMe cohort. A) 166 Manhattan plot displays seven genome-wide significant loci for coffee intake in the 23andMe 167 cohort (N=130,153). The horizontal line represents the threshold for significance (p=5.00E-08). 168 Nearest protein-coding genes (<1Mb) to significant loci are labeled. Quantile-quantile plot shown 169 in upper left corner. For more details, see Table 1 and Supplementary Table 6. B) Overlap of 170 genes identified by MAGMA, H-MAGMA, S-PrediXcan, and S-MultiXcan. Genes identified by all 171 four methods are displayed. C) Genes predicted to affect coffee intake identified by S-MultiXcan 172 according to the most significantly associated biological systems. For more details, see 173 Supplementary Table 9. D) Genes implicated in coffee intake by S-PrediXcan according to brain 174 regions. Upregulated genes are shown in red, downregulated shown in blue. For more detail, see 175 Supplementary Table 10. 11 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 176 Gene-based and tissue enrichment analyses suggest coffee intake is primarily 177 associated with gene expression in the brain 178 We used gene- and transcriptome-based analyses (MAGMA, H-MAGMA, S-MultiXcan/S- 179 PrediXcan) and identified 165 target candidate genes that may be most relevant to coffee intake. 180 MAGMA identified 31 genes implicated in coffee intake in physical proximity to GWAS loci 181 (Supplementary Table 7). H-MAGMA, which maps SNPs to genes via chromatin interaction from 182 human brain tissue, implicated 143 unique gene-tissue pairs showing expression specific to cell 183 type (75.16% neuron [31.30% cortical neuron, 33.04% iPSC derived neurons; 35.65% midbrain 184 dopamine neurons], 24.83% astrocyte) and developmental (48.00% fetal, 52.00% adult) 185 (Supplementary Table 8). Finally, S-MultiXcan predicted significant transcriptional regulation of 186 40 genes implicated in coffee intake dispersed across 20 tissues (Figure 1C; Supplementary 187 Table 10). Of the top biological systems implicated by S-MultiXcan, nine were attributed to the 188 nervous system (brain N=5; tibial nerve N=4), eight to the digestive system (esophagus N=6; 189 pancreas N=1; small intestine N=1), and six to the reproductive system (testis N=4; prostate N=2; 190 Figure 1C). Fifty percent of these genes were predicted to be downregulated in the digestive and 191 reproductive systems, whereas 66.67% of nervous system genes were predicted to be 192 upregulated. Cortical enrichment was further supported by S-PrediXcan (Figure 1D), showing 193 that SNPs associated with coffee intake most frequently correlated with predicted gene 194 expression in overall cortical and frontal cortical regions (N=4/tissue), as well as the putamen 195 (N=5). Overall, four genes (SCAMP2, SCAMP5, MPI, and FAM219B) were identified by all four 196 methods, and six of the 165 discovered genes (FBXO28, NEIL2, HAUS4, IGDCC4, RP11- 197 298I3.5, RP11-298I3.5) were not within 1Mb of SNPs identified by prior GWAS of coffee or 198 caffeine traits (Supplementary Table 11; Figure 1B; Table 2). These novel genes have been 199 associated with substance use (e.g., HAUS4 and smoking initiation73), educational outcomes 12 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 200 (e.g., HAUS4 and educational attainment90), and biomarkers (e.g., FBXO28 and mean platelet 201 volume91; IGDCC4 and mean corpuscular volume92). 202 Next, we used MAGMA gene-set analysis to identify biological pathways that may be most 203 strongly associated with coffee intake. This analysis revealed significant enrichment (p=4.75E- 204 07) in pathways related to the metabolism of xenobiotics or foreign substances (i.e., chemicals) 205 (Supplementary Table 12). 206 MAGMA tissue-based enrichment analyses suggested that coffee intake was only 207 significantly associated with brain tissue (Supplementary Figure 9A). More specifically, 208 differential expression by coffee intake was enriched (p<9.25E-04) in the frontal cortex, overall 209 cortex, cerebellum, and cerebellar hemispheres (Supplementary Figure 9B; Supplementary 210 Table 12), consistent with the S-PrediXcan findings (Supplementary Table 9). 211 Genetic correlation and polygenic score analyses of coffee intake in US- and UK-based 212 cohorts reveal discrepant associations with health and psychiatric traits, but consistent 213 positive associations with substance use and obesity 214 To boost statistical power and identify novel genes associated with coffee intake, we 215 sought to meta-analyze our data (metaGWAS) with those from the UKB using METAL93. As a 216 preliminary step to determine the appropriateness of a meta-analysis, we examined the genetic 217 correlation between coffee intake in the 23andMe and UKB cohorts. Surprisingly, the two datasets 218 were only moderately correlated (rg=0.63, p=3.54E-43), although all top loci (p<5.0E-05) shared 219 direction of effect and had similar effect strengths (Supplementary Figure 10). In addition, the 220 estimated LDSC SNPh2 heritability of coffee intake of our metaGWAS was slightly lower than for 221 both the univariate GWAS (metaGWAS SNPh2=4.09% ± 0.26 vs. 23andMe SNPh2=7.57% ± 0.59 222 vs. UKB SNPh2=4.85% ± 0.33; Supplementary Table 5). We interpreted these results as an 13 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 223 indication of cohort heterogeneity and proceeded to analyze genetic associations with coffee 224 intake in each cohort independently. 14 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 225 15 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 226 Figure 2. Discordant genetic and phenotypic associations with genetic disposition to 227 coffee intake in US and UK cohorts. A) Comparison of genetic correlations across psychiatric 228 (light gray), anthropologic (medium gray), and health (dark gray) traits between 23andMe (lanes 229 1 and 2) and UKB (lanes 3 and 4). Lanes 1 and 3 show rg values calculated by LDSC, and lanes 230 2 and 4 show FDR-corrected p values. Only traits for which at least one cohort was FDR- 231 significant are displayed. For a full list of correlations and trait names, see Supplementary Table 232 14. Most signals persisted after conditioning for dietary sugar, cigarettes per day, and Alcohol 233 Use Disorder Identification (AUDIT) Consumption scores using mtCOJO94 (Supplementary 234 Tables 15-16; Supplementary Figures 11-12). Genetic correlations for traits denoted with * 235 could not be calculated in both cohorts; ** denotes reverse coding. B) Phenomic associations 236 (panel 1: PheWAS (p<3.62E-05), panel 2: LabWAS (p<1.57E-04)) identified from PGS of coffee 237 intake from 23andMe and UKB summary statistics. Only traits for which at least one cohort was 238 FDR-significant are displayed (saturated bars=FDR significant; desaturated bars=FDR non- 239 significant). neuro.=neurological; gen.=genitourinary; neopl.=neoplasms; sense=sense organs; 240 derma.=dermatologic; imm.=immune. For full trait names and more detail, see Supplementary 241 Table 18-19. 242 243 To further understand these discrepancies, we performed a series of genetic correlation 244 and polygenic score analyses. First, we examined the genetic architecture of coffee intake 245 measured in 23andMe and UKB by comparing patterns of LDSC genetic correlation (rg) with 317 246 traits across 20 health, psychiatric, and anthropologic categories from publicly available GWAS 247 summary statistics (Figure 2A; Supplementary Table 14). After accounting for multiple testing, 248 75 traits were genetically correlated with coffee intake in the 23andMe cohort and 74 traits in the 249 UKB cohort. These associations could be underpinned by other unmeasured factors, like sugar 250 intake from coffee sweeteners or smoking and alcohol use95. However, these patterns of genetic 251 correlations persisted after conditioning on dietary sugar intake, cigarettes smoked per day, and 252 alcohol consumption measured by the Alcohol Use Disorder Identification Test (AUDIT; 253 Supplementary Tables 11-12; Supplementary Tables 15-16). Strikingly, of the traits significant 254 in at least one cohort, only 34 (29.57%) were significant in both datasets, and only 58.82% of the 255 traits significant in both datasets shared the same direction of correlation. 16 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 256 Among the traits that were significant and consistent in direction for both cohorts, we 257 observed positive genetic correlations between coffee intake and substance use phenotypes. For 258 example, we identified positive genetic correlations with smoking initiation (23andMe: rg=0.50, 259 p=4.74E-47; UKB: rg=0.12, p=1.89E-06), drinks per week (23andMe: rg=0.39, p=3.38E-28; UKB: 260 rg=0.21; p=1.39E-14), and cannabis initiation (23andMe: rg=0.28, p=1.34E-08; UKB: rg=0.09, 261 p=5.61E-03). The strength of genetic correlations for substance use and misuse traits significant 262 in at least one cohort was stronger in 23andMe compared to the UKB (0.30±0.03 vs. 0.09±0.02; 263 Welch’s t(51.97)=5.96, p=2.23E-07). For example, associations with substance use disorder and 264 dependence traits were mostly observed in the 23andMe cohort and were weaker or not observed 265 in the UKB, such as for tobacco use disorder, opioid use disorder, cannabis use disorder, nicotine 266 dependence, and alcohol dependence (rg=0.24 to 0.44, p=6.54E-23 to 2.12E-03), as well as 267 externalizing (23andMe: rg=0.48, p=7.21E-41; UKB: rg=0.07, p=4.37E-03), which is highly 268 correlated with substance use and misuse96. Cluster analysis showed that genetic correlations for 269 coffee intake in both cohorts aligned more with general substance use than misuse 270 (Supplementary Figure 13). 271 Metabolic traits were largely congruent in their positive genetic correlations with coffee 272 intake in both cohorts. For example, BMI (23andMe: rg=0.19, p=1.61E-11; UKB: rg=0.25, p=7.85E- 273 26) and waist-to-hip ratio (23andMe: rg=0.12, p=4.33E-04; UKB: rg=0.13, p=3.96E-07) were 274 positively genetically correlated with coffee intake in both datasets. Also consistent across cohorts 275 were the lack of significant genetic correlations with most cardiovascular and cancer traits. 276 The majority of traits were only significant in one cohort or showed discrepancies in the 277 direction of association. For example, coffee intake measured in the 23andMe dataset was 278 positively genetically correlated with anxiety-related traits (rg=0.22 to 0.44, p=1.41E-05 to 8.53E- 279 03). In contrast, all significant genetic correlations between coffee intake and anxiety-related traits 280 in the UKB were negative (rg= -0.33 to -0.12, p=5.49E-06 to 8.12E-03), except clinically diagnosed 17 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 281 anxiety (rg=0.17, p=1.39E-05). We also identified significant positive genetic correlations with 282 cross-disorder, attention deficit hyperactivity disorder, schizophrenia, and anorexia (rg=0.12 to 283 0.27, p=1.00E-07 to 0.01) that were exclusive to the 23andMe dataset, whereas these 284 associations were not apparent or were negatively genetically correlated in the UKB (rg= -0.13 to 285 0.02, p=1.01E-04 to 0.55). Significant positive correlations with cognitive variables, such as 286 executive function and intelligence, were found in the UKB (rg= 0.13 to 0.23, p=4.55E-08 to 8.04E- 287 23), though these were negatively genetically correlated in 23andMe (rg= -0.17 to -0.10, p=7.83E- 288 08 to 2.06E-03). Certain correlations with physical health traits also differed between cohorts. 289 While correlations with most gastrointestinal traits in the 23andMe cohort were positive, such as 290 a positive genetic correlation with gastric ulcers (rg=0.41, p=3.58E-03), the corresponding genetic 291 correlations observed in the UKB were either non-significant or negative (rg= -0.22 to 0.12, 292 p=1.34E-06 to 0.88). Positive genetic correlations with chronic pain as well as back, hip, and knee 293 pain were observed in the 23andMe dataset (rg=0.12 - 0.26, p=9.02E-08 - 3.58E-03), yet only 294 negative genetic correlations with pain traits were reported in the UKB (rg=-0.22 to -0.12, p=6.23E- 295 04 - 2.54E-06). Across all health and psychiatric traits that were significant within each cohort, all 296 traits showed a positive genetic correlation with coffee intake in 23andMe participants. Only 297 41.3% of correlations were positive in the UKB. 298 We observed similar discrepancies when we extended our results to a health-care system 299 population (Figure 2B). We conducted PheWAS and LabWAS by testing the association between 300 polygenic scores (PGS) for coffee intake derived from 23andMe or the UKB with 1,655 medical 301 traits and biomarkers. We identified 31 PheWAS and LabWAS traits that met the 5% FDR 302 significance threshold using the 23andMe PGS, and 24 using the UKB PGS (Supplementary 303 Tables 17 and 18). Only two endocrine traits (i.e., obesity and morbid obesity) and two biomarkers 304 related to red blood cells were consistent in significance and direction of association. Otherwise, 305 all significant associations were observed when testing PGS generated from one cohort but not 18 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 306 the other. For instance, when coffee intake PGS were derived from 23andMe, among the top 307 positive PheWAS and LabWAS associations were substance use disorders and respiratory 308 conditions (e.g., chronic airway obstruction, emphysema, and respiratory failure) and absolute 309 monocyte count. Among the top negative associations derived from 23andMe PGS were those 310 with sense organs, neoplasms, certain respiratory conditions (i.e., allergic rhinitis, acute and 311 chronic tonsillitis, chronic tonsillitis and adenoiditis), and urea nitrogen serum/plasma. When 312 coffee intake PGS were derived from UKB, among the top positive PheWAS and LabWAS 313 associations were endocrine and musculoskeletal disorders, as well as the two metabolic 314 biomarkers, glycated hemoglobin A1c and glucose. The only significant negative PheWAS and 315 LabWAS associations from UKB-derived PGS were with anxiety disorders, and biomarkers 316 related to blood (mean corpuscular hemoglobin concentration) and metabolic (cholesterol and 317 triglycerides in serum or blood) traits. 318 Discussion 319 In this study, we contributed to the existing GWAS literature of coffee intake by analyzing 320 a US population of 130,153 participants. We uncovered seven loci associated with coffee intake, 321 most of which were in genes implicated in metabolic processes. Coffee related variants were 322 significantly enriched in the central nervous system. Despite prior evidence that coffee intake 323 confers health benefits, we found genetic correlations mostly with adverse outcomes in both 324 cohorts, particularly substance use disorders and obesity-related traits, in both cohorts. 325 Relationships with other medical, anthropologic, and psychiatric traits were inconsistent in the US 326 and UK cohorts, suggesting that differences between populations may affect coffee intake GWAS 327 results and its genetic relationships with other traits. 328 Our GWAS replicated prior associations with genes and variants implicated in coffee and 329 caffeine intake as well as other metabolic and xenobiotic processes28, including rs2472297 near 330 CYP1A1/CYP1A218,24,26,33,46 and rs4410790 near AHR18,23,24,26,27,46,97, even though our study 19 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 331 sample was smaller compared to other GWAS (N=125,776-373,52220,24,26). Gene-based analyses 332 uncovered 165 candidate genes, including four genes that overlapped across all four analyses: 333 MPI, SCAMP2, SCAMP5, and FAM219B, all of which have been implicated in a prior coffee 334 GWAS18. These overlapping genes have other associations with substance use and medical 335 biomarkers including blood pressure, hypertension, and LDL cholesterol91,98-104. We identified 336 gene enrichment in brain tissues across the frontal cortex, putamen, and hippocampus, consistent 337 with prior GWAS showing enrichment for SNPs associated with coffee and caffeine in the central 338 nervous system18,20,26. This is supported by brain imaging studies across cortical and subcortical 339 areas showing morphological105-108 and functional109,110 differences between those who habitually 340 drink coffee compared to those who do not. 341 One of the most striking observations of this study is the breadth and magnitude of positive 342 associations between coffee intake with substance use. It is widely believed that use of one 343 substance heightens risk for use of other substances and that there are common genetic risk 344 factors for any substance use111,112; coffee, which is not generally considered a drug of misuse, 345 does not appear to be exempt from this. We identified positive genetic correlations between coffee 346 intake and other substances (i.e., tobacco, alcohol, cannabis and opioid use), as well as relevant 347 personality traits like externalizing behavior. The genetics of coffee intake aligned with substance 348 consumption phenotypes, corroborating prior GWAS and twin studies113-115 (but see23), but not 349 with substance misuse. This is perhaps unsurprising because the phenotype probed by the 350 23andMe and UKB cohorts focuses on quantity rather than clinically-defined dependence. We 351 and others previously demonstrated that the genetic architectures of other substance intake 352 versus problematic use are unique43,111,116-119, and this is likely also true for coffee. 353 We found consistent positive genetic correlations with BMI and obesity in both 23andMe 354 and the UKB. This is in contrast to meta-analyses of randomized control trials and epidemiological 355 studies that found unclear effects by any coffee or decaffeinated coffee intake on waist 356 circumference and BMI-defined obesity, and a modest inverse relationship between coffee intake 20 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 357 and BMI120,121. Results for these studies are highly heterogeneous, likely due to interindividual 358 variability in the inclusion of sugary coffee additives, cultivation, roasting, and brewing conditions 359 affecting its chemical makeup9,122, and other habits surrounding coffee intake (e.g., concurrent 360 food intake or appetite suppression by nicotine if smoking concurrently123). This contentious 361 relationship may also be explained by the amount of coffee intake, as greater coffee intake seems 362 to attenuate the genetic associations with BMI and obesity49, possibly due to the appetite 363 suppressant effects of caffeine124. Alongside accounting for other dietary intake, detailed 364 accounting of coffee preparation, and consumptive habits formed with coffee intake, future 365 subgroup analyses may help explain discrepant associations between the genetics and 366 prevalence of coffee intake with BMI-related traits. 367 We did not recapitulate the beneficial phenotypic relationships between coffee intake and 368 a variety of health outcomes that are generally reported by health association studies6-8,10,125-137, 369 perhaps because our study focused on the genetic relationship between coffee intake and other 370 medical outcomes, or because our study focused on coffee intake and not caffeine intake. At the 371 genetic level, we find no evidence of a common genetic background that could explain the 372 beneficial effects of coffee on 29 cancers, Alzheimer’s disease/dementia/cognitive impairments, 373 Parkinson’s disease, diabetes, cirrhosis, most cardiovascular conditions, or gout. In fact, some of 374 these associations (e.g., cardiovascular traits and type II diabetes) were positive in the 23andMe 375 cohort but showed no significant associations in the UKB cohort. Similarly, phenome-wide 376 analysis did not support prior cancerous, metabolic, cardiovascular, or neurological health 377 advantages of coffee intake6-8,10,126-136,138. Although this may seem discrepant to phenotypic 378 associations that generally report health benefits of coffee intake, recent meta-analysis of over 379 100 phenotypic studies on coffee intake health outcomes suggest high levels of heterogeneity 380 across cohorts6,7, especially across geographically separated populations6. 381 We found many opposing relationships with the genetics of coffee intake between 382 23andMe and UKB. For example, genetic correlations with pain, psychiatric illnesses, and 21 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 383 gastrointestinal traits were positively genetically correlated with coffee intake in 23andMe, but 384 these associations were negative in the UKB. Inversely, the UKB analysis revealed that coffee 385 intake was positively genetically correlated with cognitive traits, such as executive function and 386 intelligence, corroborating prior evidence139-142, yet genetic correlations with these two traits were 387 negative in 23andMe. Multiple PheWAS associations were also discordant. When PGS were 388 derived from 23andMe, we observed heightened odds between genetic liability for coffee intake 389 and respiratory illnesses, ischemic heart disease, infection, and alcohol-related disorders. Higher 390 odds for musculoskeletal and sleep conditions were mostly associated with coffee PGS generated 391 from the UKB. Only 11 out of the 42 phenotypes associated with coffee intake PGS showed 392 negative associations, and none of these purported health “benefits” were consistently observed 393 in both cohorts. Whereas the coffee intake PGS from 23andMe was associated with lower odds 394 for ear conditions, skin neoplasms, allergic rhinitis, and tonsillitis, the PGS of coffee intake from 395 the UKB was associated with a lower risk of anxiety disorders. Also of note is that the number of 396 positive genetic correlations and PGS associations between coffee intake and these other traits 397 was greater when analyzed using data from the 23andMe cohort versus from the UKB, and the 398 strength of these associations was usually stronger. Partially consistent with this, one meta- 399 analysis of mortality found an inverse relationship between coffee intake and all-cause mortality 400 in European but not US studies143. 401 Our study shows that cultural, cohort, or geographic influences could affect the inferred 402 genetic architecture of coffee intake and its associations with other health and lifestyle outcomes. 403 Geographic regions may have an observable influence on GWAS results144. We observed no 404 significant differences in subtle geographic differences on coffee intake correlations using location 405 data available in the UKB (Supplementary Figure 14), suggesting cultural differences may 406 contribute more to the cohort variations we report here. There is considerable variation in how or 407 with whom one may consume coffee that could be subject to cultural influence. Caffeinated 408 beverage sales, for instance, suggest that coffee and carbonated caffeinated beverages are more 22 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 409 preferred in the US than the UK2, whereas tea is the preferred source of caffeine in the UK and 410 may modify coffee intake2 (Supplementary Figure 15). Higher levels of coffee intake or caffeine 411 from high caloric beverages in the US cohort may partially explain the higher number and 412 magnitude of negative health associations observed in the 23andMe analysis. Even across coffee 413 beverage subtypes, the concentration of caffeine, other coffee chemical constituents, and 414 manufacturing byproducts (e.g., plastics and metals from packaging) varies and thus may be 415 important parameters in health associations122,145,146. A recent investigation revealed the volume 416 of ground or instant coffee is important to the potential health effects of its intake147; instant coffee 417 (~60 mg of caffeine per cup) is more commonplace in the UK whereas fresh brewed coffee (~85 418 mg of caffeine per cup 20) is preferred in the US2. Cultural differences in coffee intake could help 419 explain the divergent patterns of health and lifestyle associations between UK and US 420 participants, though the relative contributions of culture, geography, and their interactions to these 421 differences will need further exploration. 422 There are multiple caveats to consider when interpreting our findings. Firstly, our study 423 does not address causality between coffee intake and other health and lifestyle traits. Mendelian 424 randomization (MR) studies have attempted to address the exposure-outcome relationships 425 between two traits by using genetic instruments (i.e., SNPs identified by GWAS) as proxies for 426 exposure and associating them with an outcome of interest. For example, MR using genetic 427 markers associated with coffee intake suggest that coffee consumption has no causal effect on 428 obesity and endocrine disorders, despite observational studies suggesting protective effects of 429 coffee148. Similarly, MR studies of coffee and other substance use (e.g., tobacco, alcohol, 430 cannabis) are also contentious149,150, with evidence that inconsistencies may be driven by gene- 431 cohort confounds such as those we found in this study151. Secondly, the phenotype examined by 432 23andMe was exclusively caffeinated coffee intake, with one cup defined as 5 ounces, whereas 433 the UKB also included decaffeinated coffee and did not explicitly define the volume of one cup. 434 The caffeine content within coffee was also not directly measured. However, secondary analysis 23 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 435 using summary statistics of estimated caffeine intake from any coffee subtype in the UKB20 yielded 436 remarkably similar patterns of genetic correlations as those derived from our GWAS of cups of 437 coffee consumed (Supplementary Figure 16, Supplemental Table 14). This analysis 438 presumably mitigated the relative contribution of decaffeinated coffee (3mg of caffeine per cup 439 versus 60 to 85mg per cup of caffeinated coffee20) to the revealed genetic associations, so we do 440 not believe the cohort discrepancies are driven by the inclusion of decaffeinated coffee drinkers 441 in the UKB. Another consideration is the possible health effects of non-caffeine coffee 442 components, which are comparatively under-investigated9, such as other coffee bean 443 phytochemical and drink additives. Furthermore, while it is unlikely that the discrepancies in 444 genetic associations are driven by age, which is similar between cohorts (approximately 53 years 445 old in 23andMe versus 5720 years old in UKB), these cohorts skew older than the population 446 average. They are also of above average socioeconomic status152 and are of European descent, 447 limiting generalizability of our findings to a larger population. Some studies also show sex- 448 dependent differences in coffee and caffeine metabolism and health associations with 449 intake138,153,154, which was not examined in our study. 450 Overall, we present striking differences in genetic associations of coffee intake across two 451 large cohorts of European ancestry. While some genetic signals replicate across diverse cohorts, 452 such as our GWAS findings and the associations between coffee intake with substance use and 453 obesity traits, other associations may be obscured by cohort or cultural differences related to the 454 phenotype in question. Our study provides a cautionary perspective on combining large cohort 455 datasets gathered from unique geo-cultural populations. 456 24 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 457 Acknowledgements 458 We would like to thank the research participants and employees of 23andMe for making 459 this work possible. The following members of the 23andMe Research Team contributed to this 460 study: Stella Aslibekyan, Adam Auton, Elizabeth Babalola, Robert K. Bell, Jessica Bielenberg, 461 Katarzyna Bryc, Emily Bullis, Daniella Coker, Gabriel Cuellar Partida, Devika Dhamija, Sayantan 462 Das, Teresa Filshtein, Kipper Fletez-Brant, Will Freyman, Karl Heilbron, Pooja M. Gandhi, Karl 463 Heilbron, Barry Hicks, David A. Hinds, Ethan M. Jewett, Yunxuan Jiang, Katelyn Kukar, Keng- 464 Han Lin, Maya Lowe, Jey C. McCreight, Matthew H. McIntyre, Steven J. Micheletti, Meghan E. 465 Moreno, Joanna L. Mountain, Priyanka Nandakumar, Elizabeth S. Noblin, Jared O'Connell, Aaron 466 A. Petrakovitz, G. David Poznik, Morgan Schumacher, Anjali J. Shastri, Janie F. Shelton, 467 Jingchunzi Shi, Suyash Shringarpure, Vinh Tran, Joyce Y. Tung, Xin Wang, Wei Wang, Catherine 468 H. Weldon, Peter Wilton, Alejandro Hernandez, Corinna Wong, Christophe Toukam Tchakouté. 469 We would also like to thank The Externalizing Consortium for sharing the GWAS summary 470 statistics of externalizing. The Externalizing Consortium: Principal Investigators: Danielle M. Dick, 471 Philipp Koellinger, K. Paige Harden, Abraham A. Palmer. Lead Analysts: Richard Karlsson Linnér, 472 Travis T. Mallard, Peter B. Barr, Sandra Sanchez-Roige. Significant Contributors: Irwin D. 473 Waldman. The Externalizing Consortium has been supported by the National Institute on Alcohol 474 Abuse and Alcoholism (R01AA015416 -administrative supplement), and the National Institute on 475 Drug Abuse (R01DA050721). Additional funding for investigator effort has been provided by 476 K02AA018755, U10AA008401, P50AA022537, as well as a European Research Council 477 Consolidator Grant (647648 EdGe to Koellinger). The content is solely the responsibility of the 478 authors and does not necessarily represent the official views of the above funding bodies. The 479 Externalizing Consortium would like to thank the following groups for making the research 480 possible: 23andMe, Add Health, Vanderbilt University Medical Center’s BioVU, Collaborative 481 Study on the Genetics of Alcoholism (COGA), the Psychiatric Genomics Consortium’s Substance 25 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 482 Use Disorders working group, UK10K Consortium, UK Biobank, and Philadelphia 483 Neurodevelopmental Cohort. 484 MVJ, SBB, and SSR are supported by funds from the California Tobacco-Related Disease 485 Research Program (TRDRP; Grant Number T29KT0526 and T32IR5226). SBB and AAP were 486 also supported by P50DA037844. BP, JJM, and SSR are supported by NIH/NIDA DP1DA054394. 487 HHAT is funded through a Natural Science and Engineering Research Council PGS-D 488 scholarship and Canadian Institutes of Health Research (CIHR) Fellowship. JYK is supported by 489 a CIHR Canada Research Chair in Translational Neuropsychopharmacology. LKD is supported 490 by R01 MH113362. NSC is funded through an Interdisciplinary Research Fellowship in 491 NeuroAIDs (Grant Number R25MH081482). JM is funded by the National Institute for Health and 492 Care Research (NIHR) Maudsley Biomedical Research Centre at South London and Maudsley 493 NHS Foundation Trust and King’s College London. The content is solely the responsibility of the 494 authors and does not necessarily represent the official views of the National Institutes of Health. 495 The datasets used for the PheWAS and LabWAS analyses described were obtained from 496 Vanderbilt University Medical Center’s BioVU which is supported by numerous sources: 497 institutional funding, private agencies, and federal grants. These include the NIH funded Shared 498 Instrumentation Grant S10RR025141; and CTSA grants UL1TR002243, UL1TR000445, and 499 UL1RR024975. Genomic data are also supported by investigator-led projects that include 500 U01HG004798, R01NS032830, RC2GM092618, P50GM115305, U01HG006378, 501 U19HL065962, R01HD074711; and additional funding sources listed 502 at https://victr.vumc.org/biovu-funding/. PheWAS and LabWAS analyses used CTSA (SD, 503 Vanderbilt Resouces). This project was supported by the National Center for Research 504 Resources, Grant UL1 RR024975-01, and is now at the National Center for Advancing 505 Translational Sciences, Grant 2 UL1 TR000445-06. 506 Author Contributions 26 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 507 SSR and AAP conceived the idea. PF and SLE contributed formal analyses and curation 508 of 23andMe data. HHAT contributed to formal analyses, investigation, and data visualization. BP, 509 AA, and NCK contributed to formal data analysis and data visualization. JJM and LVR contributed 510 to formal analyses. JM contributed to data visualization. HHAT and SSR wrote the manuscript. 511 All authors reviewed and edited the manuscript. 512 Competing interests 513 PF and SLE are employees of 23andMe, Inc., and hold stock or stock options in 23andMe. 514 AAP is on the scientific advisory board of Vivid Genomics for which he receives stock options. 515 Materials and Correspondence 516 Correspondence and material requests should be directed to Sandra Sanchez-Roige, 517 sanchezroige@ucsd.edu. 518 Data availability 519 We will provide 23andMe summary statistics for the top 10,000 SNPs upon publication. 520 Full GWAS summary statistics will be made available through 23andMe to qualified researchers 521 under an agreement with 23andMe that protects the privacy of the 23andMe participants. 522 Please visit (https://research.23andme.com/collaborate/#dataset-access/) for more information 523 and to apply to access the data. 27 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 524 Methods 525 Study cohorts, coffee intake and univariate GWAS 526 23andMe. Univariate GWAS was conducted in a sample of 130,153 male and female 527 research participants of the genetics testing company 23andMe, Inc, as previously described155. 528 Participants provided informed consent and volunteered to participate in the research online, 529 under a protocol approved by the external AAHRPP-accredited IRB, Ethical & Independent (E&I) 530 Review Services. As of 2022, E&I Review Services is part of Salus IRB 531 (https://www.versiticlinicaltrials.org/salusirb). During 4 months in 2015 and 14 months between 532 2018-2020, participant responses to the question “How many 5-ounce (cup-sized) servings of 533 caffeinated coffee do you consume each day?” were collected as part of a larger survey. 534 Participants categorized as of European descent by genotype data were included in the univariate 535 GWAS (see Supplementary Material)156. Participant demographics are presented in 536 Supplementary Table 2. 537 DNA extraction and genotyping were performed from saliva samples by clinical laboratories 538 CLIA-certified and CAP-accredited by the Laboratory Corporation of America. 23andMe, Inc. 539 conducted all quality control, imputation, and univariate genome-wide analyses as previously 540 described (see Supplementary Table 4 for SNPs analyzed following quality control and 541 imputation)157,158. Variants were imputed based on an imputation panel combining 1000 Genomes 542 Phase 3, UK10K and the Human Reference Consortium. The 23andMe pipeline removes variants 543 of low genotyping quality (failed a Mendelian transmission test in trios (p<1.00E-20), failed Hardy- 544 Weinberg test (p<1.02E-20), failed batch effects test (ANOVA p<1.00E-20), or had a call rate 545 <90%) or imputation quality (r2<0.50 averages across genotyping arrays or a minimum r2<0.3 or 546 variants with apparent batch effects (p<1.00E-50))157,159. The 23andMe GWAS pipeline performs 547 linear regression and assumes an additive model for allelic effects43,155,160,161. Only unrelated 548 participants were included in the GWAS, which was determined using a segmental identity-by- 28 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 549 descent (IBD) estimation algorithm. Two individuals that shared more than 700 cM IBD at one or 550 both genomic segments (~20% of the genome) were classified as related. This is the minimum 551 threshold expected to identify first cousins in an outbred population. Age (inverse-normal 552 transformed), sex, the top five principal genotype components, and indicator variables for 553 genotype platforms were applied as covariates and p-values were corrected for genomic control. 554 We conducted replication in three multi-ancestral cohorts (European ancestry N=689,661; 555 African American ancestry N=32,312; Latin American ancestry N=124,155; daily mg of caffeine 556 from coffee, transformed by log10(x+75)). Demographic information on these cohorts is shown in 557 Supplementary Table 3. Ancestry was determined by analyzing local ancestry (see 558 Supplementary Material)156. 559 UK Biobank. Summary statistics of coffee intake (N=334,659) were generated from UK 560 Biobank (UKB) participants. Participants provided informed consent, were of White British 561 descent, and answered the question “How many cups of coffee do you drink each day? (Include 562 decaffeinated coffee)”. Other previously published GWAS of coffee intake with publicly available 563 summary statistics were not included in our meta-analysis due to differences in the way that coffee 564 intake was measured (e.g., “How often do you drink coffee?”, “How much coffee do you consume 565 per year?”30), or differences in ascertainment (e.g., Parkinson’s disease only31). Secondary 566 analysis was also conducted with GWAS summary statistics of caffeine intake from coffee 567 (N=373,522) in the UKB that was calculated based on the number of cups of caffeinated and 568 decaffeinated coffee consumed20. For further information about the UKB data collection and 569 GWAS summary statistics for coffee intake and caffeine consumed from coffee, see 570 http://www.nealelab.is/uk-biobank/ (field 1498, both sexes) and Said et al. 202020, respectively. 571 GWAS meta-analysis. We performed sample size weighted meta-analysis of the 23andMe 572 and UKB cohorts using METAL (version 2020-05-05)93 as previously described43. A total of 29 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 573 491,347 participants of European ancestry and 9,551,852 SNPs passing quality control were 574 included in this meta-analysis. 575 576 Gene-based analyses (MAGMA, H-MAGMA, SPrediXcan/S-MultiXcan) 577 The web-based platform Functional Mapping and Annotation of Genome-Wide 578 Association Studies (FUMA v1.3.8) was used to further explore the functional consequences of 579 lead SNPs and identify prior associations in the literature. GWA significant lead SNPs of coffee 580 cups per day consumed from the UKB were identified using FUMA. SNPs were annotated based 581 on ANNOVAR categories, Combined Annotation Dependent Depletion scores, RegulomeDB 582 scores, expression quantitative trait loci (eQTLs), and chromatin state predicted by ChromHMM. 583 Novel coffee intake candidate genes were identified as genes not in linkage disequilibrium or 584 within 1Mb of GWAS-significant SNPs uncovered by other GWAS of coffee and caffeine traits 585 (e.g., coffee/caffeine intake or caffeine metabolism). These SNPs were identified using the EBI 586 GWAS Catalog (https://www.ebi.ac.uk/gwas/). 587 MAGMA gene-based and pathway analyses. We used “Multi-marker Analysis of GenoMic 588 Annotation” (MAGMA, v1.08) to conduct gene-based associations on the 23andMe GWAS 589 summary statistics of coffee intake. SNPs were annotated to protein-coding genes using FUMA 590 and Ensembl build v92, which accounts for SNP LD using multiple regression methods. The 591 default settings were used, and LD was estimated using the 1000 Genomes European reference 592 sample. The significance of associations across 19,773 genes were adjusted using Bonferroni 593 correction for multiple testing (one-sided p<2.53E-06; Supplementary Table 7). Gene-set 594 analysis was subsequently conducted on 10,678 gene-sets and Gene Ontology terms curated 595 from the Molecular Signatures Database (MsigDB v7.0). Tissue-specific gene expression profiles 596 were also assessed in 54 tissue types and 30 general tissue types across the body with average 597 gene expression in each tissue type used as a covariate in the analysis (Supplementary Table 30 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 598 13). Using Genome-Tissue Expression (GTEx, v8) RNA-seq data, gene expression values were 599 log2 transformed values of the average Reads Per Kilobase Million (RPKM) for each tissue type 600 (RPKM>50 were listed as 50). Significance was determined following Bonferroni correction (one- 601 sided p<9.26E-04 for 54 tissue types; one-sided p<1.67E-03 for 30 general tissue types). 602 H-MAGMA. To identify neurobiologically relevant target genes, we incorporated coffee 603 intake GWAS data with chromatin interaction profiles from human brain tissue using Hi-C coupled 604 MAGMA (H-MAGMA)162. 605 S-PrediXcan and S-MultiXcan. We performed a transcriptome-wide association study 606 (TWAS) using the MetaXcan package (ver0.7.5)163,164 consisting of S-PrediXcan and S-MultiXcan 607 to identify specific eQTL-linked genes associated with coffee intake. eQTLs are genomic loci that 608 contribute to heritable variation in mRNA levels that might influence the expression of a particular 609 gene or its neighbors. This approach uses genetic information to predict gene expression levels 610 in various brain tissues and tests whether the predicted gene expression correlates with coffee 611 intake. S-PrediXcan uses precomputed tissue weights from the GTEx project database 612 (https://www.gtexportal.org/) as the reference transcriptome dataset via Elastic net models. As 613 input data, we included our GWAS summary statistics, transcriptome tissue data, and covariance 614 matrices of the SNPs within each gene model (based on HapMap SNP set; available to download 615 at the PredictDB Data Repository) from all available tissues (N=49). We applied a Bonferroni 616 correction for multiple testing across all tissues (N=21,565; Supplementary Table 10). 617 LDSC heritability and genetic correlations 618 Linkage Disequilibrium Score regression (LDSC; https://github.com/bulik/ldsc) was used 619 to calculate heritability (SNP-h2) and genetic correlations between habitual coffee intake and other 620 phenotypes165. SNP-h2 was calculated from publicly available, pre-computed LD scores 621 (“eur_w_ld_chr/”). LDSC was also used to calculate genetic correlations (rg) between habitual 31 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 622 coffee intake and 317 selected traits informed by prior literature across the following categories: 623 substance use and misuse, anxiety, bipolar disorder & depression, cancer, cardiovascular, diet, 624 gastrointestinal, lifestyle, metabolic, neurological, pain, personality, other anthropologic, other 625 health, and other psychiatric traits. 626 mtCOJO 627 We conditioned summary statistics on traits that are correlated with coffee intake using 628 mtCOJO94 to determine if genetic associations are retained after controlling for their effects. 629 Conditioning on GWAS summary statistics for dietary sugar intake166, Alcohol Use Disorder 630 Identification Test (AUDIT) consumption43, and cigarettes smoked per day73 were examined at a 631 p<1.0E-5 and clump-r2 threshold of 0.10. 632 Phenome and laboratory-wide association studies 633 We conducted phenome-wide association analyses (PheWAS) and Laboratory-wide 634 association analyses (LabWAS) to test the association between polygenic scores (PGS) for 635 coffee intake and liability across thousands of medical conditions from hospital-based cohorts. 636 These analyses were conducted using data from the Vanderbilt University Medical Center 637 (VUMC). The project was approved by the VUMC Institutional Review Board (IRB #160302, 638 #172020, #190418). VUMC is an integrated health system with individual-level health data from 639 electronic health record (EHR) data for about 3.2 million patients. The VUMC biobank contains 640 clinical data from EHR as well as biomarkers obtained from laboratory assessments. A portion of 641 the individuals from VUMC also have accompanying array genotyping data. This cohort, with over 642 72,821 patients, is called BioVU167,168. 643 For each of the unrelated genotyped individuals of European ancestry from BioVU, we 644 computed polygenic scores for coffee intake using the PRS-CS “auto” version167. Genotyping and 645 quality control for this cohort have been extensively described168,169. 32 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 646 Phenome-wide association analyses (PheWAS). To identify associations between the 647 PGS for coffee and clinical phenotypes, we performed a PheWAS. We fitted a logistic regression 648 model to each of 1,338 case/control disease phenotypes (“phecodes”) to estimate the odds of 649 each diagnosis given the coffee PGS, while adjusting for sex, median age of the longitudinal EHR, 650 and the first 10 PCs. Analyses were conducted using the PheWAS v0.12 R package170. We 651 required the presence of at least two International Disease Classification codes mapped to a 652 PheWAS disease category (Phecode Map 1.2; https://phewascatalog.org/phecodes) and a 653 minimum of 100 cases for inclusion of a phecode. The disease phenotypes included 145 654 circulatory system, 120 genitourinary, 119 endocrine/metabolic, 125 digestive, 117 neoplasms, 655 91 musculoskeletal, 85 sense organs, 76 injuries & poisonings, 65 dermatological, 76 respiratory, 656 69 neurological, 64 mental disorders, 42 infectious diseases, 42 hematopoietic, 34 congenital 657 anomalies, 37 symptoms, and 31 pregnancy complications. 658 Laboratory-wide association analyses (LabWAS). We also examined laboratory results in 659 BioVU, which we refer to as LabWAS. We implemented the pipeline already established by 169 660 Dennis, et al. . Broadly, LabWAS uses the median, inverse normal quantile transformed age- 661 adjusted values from the QualityLab pipeline in a linear regression to determine the association 662 with the input coffee intake PGS variable. We controlled for the same covariates as for the 663 PheWAS analyses, excluding median age because the pipeline corrects for age using cubic 664 splines with 4 knots. 665 Cluster analysis 666 Previous studies have shown that consumption and misuse/dependence phenotypes 667 have a distinct genetic architecture43,111,116-119. To explore whether the coffee intake analysis 668 clustered closer to substance intake or misuse/dependence phenotypes, we used an 669 unsupervised machine learning hierarchical clustering algorithm known as agglomerative nesting 670 (AGNES)167 on a genetic correlation matrix of all traits. AGNES initially forms single-item clusters 33 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 671 that are fused together into intermediate groups until all traits are included in a single cluster171. 672 Clusters are formed with Ward’s method such that the total within cluster variance is minimized 673 while maintaining the fewest number of clusters based on cluster dissimilarity. Dissimilarity is 674 assessed through Euclidean Distance of each pairwise genetic correlation with another trait. The 675 product of AGNES is a dendrogram, formed with multiple brackets called “branches”. AGNES 676 was implemented in R using the cluster package (ver2.1.4)167. 677 Clustering was conducted with summary statistics of cigarettes per day73, former 678 smoker73, smoking initiation73, problematic opioid use161, ICD10 F17 nicotine dependence172, 679 alcohol dependence173, AUDIT consumption116, AUDIT problems116, cannabis initiation79, 680 cannabis use disorder118, drinks per week73, externalizing psychopathology96, Fagerström Test 681 for Nicotine Dependence (FTND)174, general risk tolerance175, age of smoking initiation73, and 682 opioid use disorder161. The genetic correlations of cigarettes per day, former smoker, and smoking 683 initiation were reverse coded to show the intuitive effects against the other traits in the 684 dendrogram. 685 34 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 686 References 687 1 International Coffee Organization. Annual Review Coffee Year 2019/2020. (International 688 Coffee Organization, London, 2021). 689 2 Reyes, C. M. & Cornelis, M. C. Caffeine in the Diet: Country-Level Consumption and 690 Guidelines. Nutrients 10, doi:10.3390/nu10111772 (2018). 691 3 Landais, E. et al. Coffee and Tea Consumption and the Contribution of Their Added 692 Ingredients to Total Energy and Nutrient Intakes in 10 European Countries: Benchmark 693 Data from the Late 1990s. Nutrients 10, doi:10.3390/nu10060725 (2018). 694 4 Rehm, C. D., Ratliff, J. C., Riedt, C. S. & Drewnowski, A. Coffee Consumption among 695 Adults in the United States by Demographic Variables and Purchase Location: Analyses 696 of NHANES 2011-2016 Data. Nutrients 12, doi:10.3390/nu12082463 (2020). 697 5 Camandola, S., Plick, N. & Mattson, M. P. Impact of Coffee and Cacao Purine 698 Metabolites on Neuroplasticity and Neurodegenerative Disease. Neurochem Res 44, 699 214-227, doi:10.1007/s11064-018-2492-0 (2019). 700 6 Grosso, G., Godos, J., Galvano, F. & Giovannucci, E. L. Coffee, Caffeine, and Health 701 Outcomes: An Umbrella Review. Annu Rev Nutr 37, 131-156, doi:10.1146/annurev-nutr- 702 071816-064941 (2017). 703 7 Poole, R. et al. Coffee consumption and health: umbrella review of meta-analyses of 704 multiple health outcomes. BMJ 359, j5024, doi:10.1136/bmj.j5024 (2017). 705 8 Kositamongkol, C. et al. Coffee Consumption and Non-alcoholic Fatty Liver Disease: An 706 Umbrella Review and a Systematic Review and Meta-analysis. Front Pharmacol 12, 707 786596, doi:10.3389/fphar.2021.786596 (2021). 708 9 Socala, K., Szopa, A., Serefko, A., Poleszak, E. & Wlaz, P. Neuroprotective Effects of 709 Coffee Bioactive Compounds: A Review. Int J Mol Sci 22, doi:10.3390/ijms22010107 710 (2020). 35 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 711 10 Zhao, L. G. et al. Coffee drinking and cancer risk: an umbrella review of meta-analyses 712 of observational studies. BMC Cancer 20, 101, doi:10.1186/s12885-020-6561-9 (2020). 713 11 Freedman, N. D., Park, Y., Abnet, C. C., Hollenbeck, A. R. & Sinha, R. Association of 714 coffee drinking with total and cause-specific mortality. N Engl J Med 366, 1891-1904, 715 doi:10.1056/NEJMoa1112010 (2012). 716 12 Svikis, D. S. et al. Coffee and energy drink use patterns in college freshmen: 717 associations with adverse health behaviors and risk factors. BMC Public Health 22, 594, 718 doi:10.1186/s12889-022-13012-3 (2022). 719 13 Swan, G. E., Carmelli, D. & Cardon, L. R. Heavy consumption of cigarettes, alcohol and 720 coffee in male twins. J Stud Alcohol 58, 182-190, doi:10.15288/jsa.1997.58.182 (1997). 721 14 Treur, J. L. et al. Associations between smoking and caffeine consumption in two 722 European cohorts. Addiction 111, 1059-1068, doi:10.1111/add.13298 (2016). 723 15 Tang, N., Wu, Y., Ma, J., Wang, B. & Yu, R. Coffee consumption and risk of lung cancer: 724 a meta-analysis. Lung Cancer 67, 17-22, doi:10.1016/j.lungcan.2009.03.012 (2010). 725 16 Nehlig, A. Effects of Coffee on the Gastro-Intestinal Tract: A Narrative Review and 726 Literature Update. Nutrients 14, doi:10.3390/nu14020399 (2022). 727 17 Butt, M. S. & Sultan, M. T. Coffee and its consumption: benefits and risks. Crit Rev Food 728 Sci Nutr 51, 363-373, doi:10.1080/10408390903586412 (2011). 729 18 Kang, J. et al. Increased brain volume from higher cereal and lower coffee intake: 730 shared genetic determinants and impacts on cognition and metabolism. Cereb Cortex 731 32, 5163-5174, doi:10.1093/cercor/bhac005 (2022). 732 19 Cornelis, M. C. & van Dam, R. M. Genetic determinants of liking and intake of coffee and 733 other bitter foods and beverages. Sci Rep 11, 23845, doi:10.1038/s41598-021-03153-7 734 (2021). 735 20 Said, M. A., van de Vegte, Y. J., Verweij, N. & van der Harst, P. Associations of 736 Observational and Genetically Determined Caffeine Intake With Coronary Artery 36 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 737 Disease and Diabetes Mellitus. J Am Heart Assoc 9, e016808, 738 doi:10.1161/JAHA.120.016808 (2020). 739 21 Jin, T. et al. Interactions of Habitual Coffee Consumption by Genetic Polymorphisms 740 with the Risk of Prediabetes and Type 2 Diabetes Combined. Nutrients 12, 741 doi:10.3390/nu12082228 (2020). 742 22 Vogel, C. F. A., Van Winkle, L. S., Esser, C. & Haarmann-Stemmann, T. The aryl 743 hydrocarbon receptor as a target of environmental stressors - Implications for pollution 744 mediated stress and inflammatory responses. Redox Biol 34, 101530, 745 doi:10.1016/j.redox.2020.101530 (2020). 746 23 Matoba, N. et al. GWAS of 165,084 Japanese individuals identified nine loci associated 747 with dietary habits. Nat Hum Behav 4, 308-316, doi:10.1038/s41562-019-0805-1 (2020). 748 24 Kennedy, O. J. et al. Coffee Consumption and Kidney Function: A Mendelian 749 Randomization Study. Am J Kidney Dis 75, 753-761, doi:10.1053/j.ajkd.2019.08.025 750 (2020). 751 25 Jia, H. et al. GWAS of habitual coffee consumption reveals a sex difference in the 752 genetic effect of the 12q24 locus in the Japanese population. BMC Genet 20, 61, 753 doi:10.1186/s12863-019-0763-7 (2019). 754 26 Zhong, V. W. et al. A genome-wide association study of bitter and sweet beverage 755 consumption. Hum Mol Genet 28, 2449-2457, doi:10.1093/hmg/ddz061 (2019). 756 27 Nakagawa-Senda, H. et al. A genome-wide association study in the Japanese 757 population identifies the 12q24 locus for habitual coffee consumption: The J-MICC 758 Study. Sci Rep 8, 1493, doi:10.1038/s41598-018-19914-w (2018). 759 28 Cornelis, M. C. et al. Genome-wide association study of caffeine metabolites provides 760 new insights to caffeine metabolism and dietary caffeine-consumption behavior. Hum 761 Mol Genet 25, 5472-5482, doi:10.1093/hmg/ddw334 (2016). 37 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 762 29 Pirastu, N. et al. Non-additive genome-wide association scan reveals a new gene 763 associated with habitual coffee consumption. Sci Rep 6, 31590, doi:10.1038/srep31590 764 (2016). 765 30 Coffee and Caffeine Genetics Consortium et al. Genome-wide meta-analysis identifies 766 six novel loci associated with habitual coffee consumption. Mol Psychiatry 20, 647-656, 767 doi:10.1038/mp.2014.107 (2015). 768 31 Hamza, T. H. et al. Genome-wide gene-environment study identifies glutamate receptor 769 gene GRIN2A as a Parkinson's disease modifier gene via interaction with coffee. PLoS 770 Genet 7, e1002237, doi:10.1371/journal.pgen.1002237 (2011). 771 32 Amin, N. et al. Genome-wide association analysis of coffee drinking suggests 772 association with CYP1A1/CYP1A2 and NRCAM. Mol Psychiatry 17, 1116-1129, 773 doi:10.1038/mp.2011.101 (2012). 774 33 Sulem, P. et al. Sequence variants at CYP1A1-CYP1A2 and AHR associate with coffee 775 consumption. Hum Mol Genet 20, 2071-2077, doi:10.1093/hmg/ddr086 (2011). 776 34 Yang, A., Palmer, A. A. & de Wit, H. Genetics of caffeine consumption and responses to 777 caffeine. Psychopharmacology (Berl) 211, 245-257, doi:10.1007/s00213-010-1900-1 778 (2010). 779 35 Raja, A. M. et al. Genetic Susceptibility to Chronic Liver Disease in Individuals from 780 Pakistan. Int J Mol Sci 21, doi:10.3390/ijms21103558 (2020). 781 36 Gao, H. et al. Association of GCKR Gene Polymorphisms with the Risk of Nonalcoholic 782 Fatty Liver Disease and Coronary Artery Disease in a Chinese Northern Han Population. 783 J Clin Transl Hepatol 7, 297-303, doi:10.14218/JCTH.2019.00030 (2019). 784 37 Cai, W. et al. Genetic polymorphisms associated with nonalcoholic fatty liver disease in 785 Uyghur population: a case-control study and meta-analysis. Lipids Health Dis 18, 14, 786 doi:10.1186/s12944-018-0877-3 (2019). 38 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 787 38 Xiao, X. et al. Functional POR A503V is associated with the risk of bladder cancer in a 788 Chinese population. Sci Rep 5, 11751, doi:10.1038/srep11751 (2015). 789 39 Menashe, I. et al. Large-scale pathway-based analysis of bladder cancer genome-wide 790 association data from five studies of European background. PLoS One 7, e29396, 791 doi:10.1371/journal.pone.0029396 (2012). 792 40 Rotunno, M. et al. Phase I metabolic genes and risk of lung cancer: multiple 793 polymorphisms and mRNA expression. PLoS One 4, e5652, 794 doi:10.1371/journal.pone.0005652 (2009). 795 41 Aldrich, M. C. et al. CYP1A1/2 haplotypes and lung cancer and assessment of 796 confounding by population stratification. Cancer Res 69, 2340-2348, doi:10.1158/0008- 797 5472.CAN-08-2576 (2009). 798 42 Thompson, A. et al. Functional validity, role, and implications of heavy alcohol 799 consumption genetic loci. Sci Adv 6, eaay5034, doi:10.1126/sciadv.aay5034 (2020). 800 43 Sanchez-Roige, S. et al. Genome-Wide Association Study Meta-Analysis of the Alcohol 801 Use Disorders Identification Test (AUDIT) in Two Population-Based Cohorts. Am J 802 Psychiatry 176, 107-118, doi:10.1176/appi.ajp.2018.18040369 (2019). 803 44 Clarke, T. K. et al. Genome-wide association study of alcohol consumption and genetic 804 overlap with other health-related traits in UK Biobank (N=112 117). Mol Psychiatry 22, 805 1376-1384, doi:10.1038/mp.2017.153 (2017). 806 45 Yin, B., Wang, X., Huang, T. & Jia, J. Shared Genetics and Causality Between 807 Decaffeinated Coffee Consumption and Neuropsychiatric Diseases: A Large-Scale 808 Genome-Wide Cross-Trait Analysis and Mendelian Randomization Analysis. Front 809 Psychiatry 13, 910432, doi:10.3389/fpsyt.2022.910432 (2022). 810 46 Coffee et al. Genome-wide meta-analysis identifies six novel loci associated with 811 habitual coffee consumption. Mol Psychiatry 20, 647-656, doi:10.1038/mp.2014.107 812 (2015). 39 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 813 47 Xu, J. et al. Assessing the Association between Important Dietary Habits and 814 Osteoporosis: A Genetic Correlation and Two-Sample Mendelian Randomization Study. 815 Nutrients 14, doi:10.3390/nu14132656 (2022). 816 48 Qi, X. et al. Evaluating the Effects of Diet-Gut Microbiota Interactions on Sleep Traits 817 Using the UK Biobank Cohort. Nutrients 14, doi:10.3390/nu14061134 (2022). 818 49 Wang, T. et al. Habitual coffee consumption and genetic predisposition to obesity: gene- 819 diet interaction analyses in three US prospective studies. BMC Med 15, 97, 820 doi:10.1186/s12916-017-0862-0 (2017). 821 50 Yuan, S., Daghlas, I. & Larsson, S. C. Alcohol, coffee consumption, and smoking in 822 relation to migraine: a bidirectional Mendelian randomization study. Pain 163, e342- 823 e348, doi:10.1097/j.pain.0000000000002360 (2022). 824 51 Laaboub, N. et al. Associations Between High Plasma Methylxanthine Levels, Sleep 825 Disorders and Polygenic Risk Scores of Caffeine Consumption or Sleep Duration in a 826 Swiss Psychiatric Cohort. Front Psychiatry 12, 756403, doi:10.3389/fpsyt.2021.756403 827 (2021). 828 52 Song, M. Y. & Park, S. Association of Polygenetic Risk Scores Related to Immunity and 829 Inflammation with Hyperthyroidism Risk and Interactions between the Polygenetic 830 Scores and Dietary Factors in a Large Cohort. J Thyroid Res 2021, 7664641, 831 doi:10.1155/2021/7664641 (2021). 832 53 Kim, E. J. et al. Coffee Consumption and Incident Tachyarrhythmias: Reported Behavior, 833 Mendelian Randomization, and Their Interactions. JAMA Intern Med 181, 1185-1193, 834 doi:10.1001/jamainternmed.2021.3616 (2021). 835 54 Ellison-Barnes, A., Johnson, S. & Gudzune, K. Trends in Obesity Prevalence Among 836 Adults Aged 18 Through 25 Years, 1976-2018. JAMA 326, 2073-2074, 837 doi:10.1001/jama.2021.16685 (2021). 40 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 838 55 Cole, J. B., Florez, J. C. & Hirschhorn, J. N. Comprehensive genomic analysis of dietary 839 habits in UK Biobank identifies hundreds of genetic associations. Nat Commun 11, 1467, 840 doi:10.1038/s41467-020-15193-0 (2020). 841 56 Zhou, H. et al. Genome-wide meta-analysis of problematic alcohol use in 435,563 842 individuals yields insights into biology and relationships with other traits. Nat Neurosci 843 23, 809-818, doi:10.1038/s41593-020-0643-5 (2020). 844 57 Liu, M. et al. Association studies of up to 1.2 million individuals yield new insights into 845 the genetic etiology of tobacco and alcohol use. Nat Genet 51, 237-244, 846 doi:10.1038/s41588-018-0307-5 (2019). 847 58 Pardinas, A. F. et al. Pharmacogenomic Variants and Drug Interactions Identified 848 Through the Genetic Analysis of Clozapine Metabolism. Am J Psychiatry 176, 477-486, 849 doi:10.1176/appi.ajp.2019.18050589 (2019). 850 59 Liu, H. et al. Epigenomic and transcriptomic analyses define core cell types, genes and 851 targetable mechanisms for kidney disease. Nat Genet 54, 950-962, doi:10.1038/s41588- 852 022-01097-w (2022). 853 60 Stanzick, K. J. et al. Discovery and prioritization of variants and genes for kidney 854 function in >1.2 million individuals. Nat Commun 12, 4350, doi:10.1038/s41467-021- 855 24491-0 (2021). 856 61 Hellwege, J. N. et al. Mapping eGFR loci to the renal transcriptome and phenome in the 857 VA Million Veteran Program. Nat Commun 10, 3842, doi:10.1038/s41467-019-11704-w 858 (2019). 859 62 Wuttke, M. et al. A catalog of genetic loci associated with kidney function from analyses 860 of a million individuals. Nat Genet 51, 957-972, doi:10.1038/s41588-019-0407-x (2019). 861 63 Yin, X. et al. Genome-wide association studies of metabolites in Finnish men identify 862 disease-relevant loci. Nat Commun 13, 1644, doi:10.1038/s41467-022-29143-5 (2022). 41 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 863 64 Casanova, F. et al. A genome-wide association study implicates multiple mechanisms 864 influencing raised urinary albumin-creatinine ratio. Hum Mol Genet 28, 4197-4207, 865 doi:10.1093/hmg/ddz243 (2019). 866 65 Tin, A. et al. Target genes, variants, tissues and transcriptional pathways influencing 867 human serum urate levels. Nat Genet 51, 1459-1474, doi:10.1038/s41588-019-0504-x 868 (2019). 869 66 Pazoki, R. et al. GWAS for urinary sodium and potassium excretion highlights pathways 870 shared with cardiovascular traits. Nat Commun 10, 3653, doi:10.1038/s41467-019- 871 11451-y (2019). 872 67 Zanetti, D. et al. Identification of 22 novel loci associated with urinary biomarkers of 873 albumin, sodium, and potassium excretion. Kidney Int 95, 1197-1208, 874 doi:10.1016/j.kint.2018.12.017 (2019). 875 68 Haas, M. E. et al. Genetic Association of Albuminuria with Cardiometabolic Disease and 876 Blood Pressure. Am J Hum Genet 103, 461-473, doi:10.1016/j.ajhg.2018.08.004 (2018). 877 69 Jimenez, A., Adisa, A., Woodham, C. & Saleh, M. Determination of polycyclic aromatic 878 hydrocarbons in roasted coffee. J Environ Sci Health B 49, 828-835, 879 doi:10.1080/03601234.2014.938552 (2014). 880 70 Teumer, A. et al. Genome-wide association meta-analyses and fine-mapping elucidate 881 pathways influencing albuminuria. Nat Commun 10, 4130, doi:10.1038/s41467-019- 882 11576-0 (2019). 883 71 Fredholm, B. B., Chen, J. F., Masino, S. A. & Vaugeois, J. M. Actions of adenosine at its 884 receptors in the CNS: insights from knockouts and drugs. Annu Rev Pharmacol Toxicol 885 45, 385-412, doi:10.1146/annurev.pharmtox.45.120403.095731 (2005). 886 72 Tanner, J. A. & Tyndale, R. F. Variation in CYP2A6 Activity and Personalized Medicine. 887 J Pers Med 7, doi:10.3390/jpm7040018 (2017). 42 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 888 73 Saunders, G. R. B. et al. Genetic diversity fuels gene discovery for tobacco and alcohol 889 use. Nature 612, 720-724, doi:10.1038/s41586-022-05477-4 (2022). 890 74 Sakaue, S. et al. A cross-population atlas of genetic associations for 220 human 891 phenotypes. Nat Genet 53, 1415-1424, doi:10.1038/s41588-021-00931-x (2021). 892 75 Koskeridis, F. et al. Pleiotropic genetic architecture and novel loci for C-reactive protein 893 levels. Nat Commun 13, 6939, doi:10.1038/s41467-022-34688-6 (2022). 894 76 Deak, J. D. et al. Genome-wide association study in individuals of European and African 895 ancestry and multi-trait analysis of opioid use disorder identifies 19 independent 896 genome-wide significant risk loci. Mol Psychiatry 27, 3970-3979, doi:10.1038/s41380- 897 022-01709-1 (2022). 898 77 Pasman, J. A. et al. Genetic Risk for Smoking: Disentangling Interplay Between Genes 899 and Socioeconomic Status. Behav Genet 52, 92-107, doi:10.1007/s10519-021-10094-4 900 (2022). 901 78 Xu, K. et al. Genome-wide association study of smoking trajectory and meta-analysis of 902 smoking status in 842,000 individuals. Nat Commun 11, 5302, doi:10.1038/s41467-020- 903 18489-3 (2020). 904 79 Pasman, J. A. et al. GWAS of lifetime cannabis use reveals new risk loci, genetic 905 overlap with psychiatric traits, and a causal influence of schizophrenia. Nat Neurosci 21, 906 1161-1170, doi:10.1038/s41593-018-0206-1 (2018). 907 80 Schumann, G. et al. KLB is associated with alcohol drinking, and its gene product beta- 908 Klotho is necessary for FGF21 regulation of alcohol preference. Proc Natl Acad Sci U S 909 A 113, 14372-14377, doi:10.1073/pnas.1611243113 (2016). 910 81 Stringer, S. et al. Genome-wide association study of lifetime cannabis use based on a 911 large meta-analytic sample of 32 330 subjects from the International Cannabis 912 Consortium. Transl Psychiatry 6, e769, doi:10.1038/tp.2016.36 (2016). 43 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 913 82 Wain, L. V. et al. Novel insights into the genetics of smoking behaviour, lung function, 914 and chronic obstructive pulmonary disease (UK BiLEVE): a genetic association study in 915 UK Biobank. Lancet Respir Med 3, 769-781, doi:10.1016/S2213-2600(15)00283-0 916 (2015). 917 83 King, C. P. et al. Cdh13 and AdipoQ gene knockout alter instrumental and Pavlovian 918 drug conditioning. Genes Brain Behav 16, 686-698, doi:10.1111/gbb.12382 (2017). 919 84 Drgonova, J. et al. Cadherin 13: human cis-regulation and selectively-altered addiction 920 phenotypes and cerebral cortical dopamine in knockout mice. Mol Med 22, 537-547, 921 doi:10.2119/molmed.2015.00170 (2016). 922 85 Yang, J. et al. The contribution of rare and common variants in 30 genes to risk nicotine 923 dependence. Mol Psychiatry 20, 1467-1478, doi:10.1038/mp.2014.156 (2015). 924 86 Weber, M. et al. Increased polysialic acid neural cell adhesion molecule expression in 925 human hippocampus of heroin addicts. Neuroscience 138, 1215-1223, 926 doi:10.1016/j.neuroscience.2005.11.059 (2006). 927 87 Barker, J. M., Torregrossa, M. M. & Taylor, J. R. Low prefrontal PSA-NCAM confers risk 928 for alcoholism-related behavior. Nat Neurosci 15, 1356-1358, doi:10.1038/nn.3194 929 (2012). 930 88 Mackowiak, M. et al. Cocaine decreases the expression of PSA-NCAM protein and 931 attenuates long-term potentiation via glucocorticoid receptors in the rat dentate gyrus. 932 Eur J Neurosci 27, 2928-2937, doi:10.1111/j.1460-9568.2008.06255.x (2008). 933 89 Kojetin, D. J. & Burris, T. P. REV-ERB and ROR nuclear receptors as drug targets. Nat 934 Rev Drug Discov 13, 197-216, doi:10.1038/nrd4100 (2014). 935 90 Lee, J. J. et al. Gene discovery and polygenic prediction from a genome-wide 936 association study of educational attainment in 1.1 million individuals. Nat Genet 50, 937 1112-1121, doi:10.1038/s41588-018-0147-3 (2018). 44 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 938 91 Chen, M. H. et al. Trans-ethnic and Ancestry-Specific Blood-Cell Genetics in 746,667 939 Individuals from 5 Global Populations. Cell 182, 1198-1213 e1114, 940 doi:10.1016/j.cell.2020.06.045 (2020). 941 92 Thompson, A., King, K., Morris, A. P. & Pirmohamed, M. Assessing the impact of alcohol 942 consumption on the genetic contribution to mean corpuscular volume. Hum Mol Genet 943 30, 2040-2051, doi:10.1093/hmg/ddab147 (2021). 944 93 Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and efficient meta-analysis of 945 genomewide association scans. Bioinformatics 26, 2190-2191, 946 doi:10.1093/bioinformatics/btq340 (2010). 947 94 Zhu, Z. et al. Causal associations between risk factors and common diseases inferred 948 from GWAS summary data. Nat Commun 9, 224, doi:10.1038/s41467-017-02317-2 949 (2018). 950 95 Kendler, K. S., Schmitt, E., Aggen, S. H. & Prescott, C. A. Genetic and environmental 951 influences on alcohol, caffeine, cannabis, and nicotine use from early adolescence to 952 middle adulthood. Arch Gen Psychiatry 65, 674-682, doi:10.1001/archpsyc.65.6.674 953 (2008). 954 96 Karlsson Linner, R. et al. Multivariate analysis of 1.5 million people identifies genetic 955 associations with traits related to self-regulation and addiction. Nat Neurosci 24, 1367- 956 1376, doi:10.1038/s41593-021-00908-3 (2021). 957 97 Wojcik, G. L. et al. Genetic analyses of diverse populations improves discovery for 958 complex traits. Nature 570, 514-518, doi:10.1038/s41586-019-1310-4 (2019). 959 98 Vuckovic, D. et al. The Polygenic and Monogenic Basis of Blood Traits and Diseases. 960 Cell 182, 1214-1231 e1211, doi:10.1016/j.cell.2020.08.008 (2020). 961 99 Klarin, D. et al. Genetics of blood lipids among ~300,000 multi-ethnic participants of the 962 Million Veteran Program. Nat Genet 50, 1514-1523, doi:10.1038/s41588-018-0222-9 963 (2018). 45 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 964 100 Feitosa, M. F. et al. Novel genetic associations for blood pressure identified via gene- 965 alcohol interaction in up to 570K individuals across multiple ancestries. PLoS One 13, 966 e0198166, doi:10.1371/journal.pone.0198166 (2018). 967 101 Sung, Y. J. et al. A Large-Scale Multi-ancestry Genome-wide Study Accounting for 968 Smoking Behavior Identifies Multiple Significant Loci for Blood Pressure. Am J Hum 969 Genet 102, 375-400, doi:10.1016/j.ajhg.2018.01.015 (2018). 970 102 Astle, W. J. et al. The Allelic Landscape of Human Blood Cell Trait Variation and Links to 971 Common Complex Disease. Cell 167, 1415-1429 e1419, doi:10.1016/j.cell.2016.10.042 972 (2016). 973 103 Meeks, K. A. C. et al. Genome-wide analyses of multiple obesity-related cytokines and 974 hormones informs biology of cardiometabolic traits. Genome Med 13, 156, 975 doi:10.1186/s13073-021-00971-2 (2021). 976 104 German, C. A., Sinsheimer, J. S., Klimentidis, Y. C., Zhou, H. & Zhou, J. J. Ordered 977 multinomial regression for genetic association analysis of ordinal phenotypes at Biobank 978 scale. Genet Epidemiol 44, 248-260, doi:10.1002/gepi.22276 (2020). 979 105 Pham, K. et al. High coffee consumption, brain volume and risk of dementia and stroke. 980 Nutr Neurosci 25, 2111-2122, doi:10.1080/1028415X.2021.1945858 (2022). 981 106 Haller, S., Montandon, M. L., Rodriguez, C., Herrmann, F. R. & Giannakopoulos, P. 982 Impact of Coffee, Wine, and Chocolate Consumption on Cognitive Outcome and MRI 983 Parameters in Old Age. Nutrients 10, doi:10.3390/nu10101391 (2018). 984 107 Araujo, L. F. et al. Association of Coffee Consumption with MRI Markers and Cognitive 985 Function: A Population-Based Study. J Alzheimers Dis 53, 451-461, doi:10.3233/JAD- 986 160116 (2016). 987 108 Perlaki, G. et al. Coffee consumption may influence hippocampal volume in young 988 women. Brain Imaging Behav 5, 274-284, doi:10.1007/s11682-011-9131-6 (2011). 46 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 989 109 Magalhaes, R. et al. Habitual coffee drinkers display a distinct pattern of brain functional 990 connectivity. Mol Psychiatry 26, 6589-6598, doi:10.1038/s41380-021-01075-4 (2021). 991 110 Laurienti, P. J. et al. Dietary caffeine consumption modulates fMRI measures. 992 Neuroimage 17, 751-757 (2002). 993 111 Deak, J. D. & Johnson, E. C. Genetics of substance use disorders: a review. Psychol 994 Med 51, 2189-2200, doi:10.1017/S0033291721000969 (2021). 995 112 Vanyukov, M. M. et al. Common liability to addiction and "gateway hypothesis": 996 theoretical, empirical and evolutionary perspective. Drug Alcohol Depend 123 Suppl 1, 997 S3-17, doi:10.1016/j.drugalcdep.2011.12.018 (2012). 998 113 Chang, L. H. et al. Investigating the genetic and causal relationship between initiation or 999 use of alcohol, caffeine, cannabis and nicotine. Drug Alcohol Depend 210, 107966, 1000 doi:10.1016/j.drugalcdep.2020.107966 (2020). 1001 114 Hettema, J. M., Corey, L. A. & Kendler, K. S. A multivariate genetic analysis of the use of 1002 tobacco, alcohol, and caffeine in a population based sample of male and female twins. 1003 Drug Alcohol Depend 57, 69-78, doi:10.1016/s0376-8716(99)00053-8 (1999). 1004 115 Swan, G. E., Carmelli, D. & Cardon, L. R. The consumption of tobacco, alcohol, and 1005 coffee in Caucasian male twins: a multivariate genetic analysis. J Subst Abuse 8, 19-31, 1006 doi:10.1016/s0899-3289(96)90055-3 (1996). 1007 116 Mallard, T. T. & Sanchez-Roige, S. Dimensional Phenotypes in Psychiatric Genetics: 1008 Lessons from Genome-Wide Association Studies of Alcohol Use Phenotypes. Complex 1009 Psychiatry 7, 45-48, doi:10.1159/000518863 (2021). 1010 117 Gelernter, J. & Polimanti, R. Genetics of substance use disorders in the era of big data. 1011 Nat Rev Genet 22, 712-729, doi:10.1038/s41576-021-00377-1 (2021). 1012 118 Johnson, E. C. et al. A large-scale genome-wide association study meta-analysis of 1013 cannabis use disorder. Lancet Psychiatry 7, 1032-1045, doi:10.1016/S2215- 1014 0366(20)30339-4 (2020). 47 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1015 119 Sanchez-Roige, S., Palmer, A. A. & Clarke, T. K. Recent Efforts to Dissect the Genetic 1016 Basis of Alcohol Use and Abuse. Biol Psychiatry 87, 609-618, 1017 doi:10.1016/j.biopsych.2019.09.011 (2020). 1018 120 Ramli, N. N. S., Alkhaldy, A. A. & Mhd Jalil, A. M. Effects of Caffeinated and 1019 Decaffeinated Coffee Consumption on Metabolic Syndrome Parameters: A Systematic 1020 Review and Meta-Analysis of Data from Randomised Controlled Trials. Medicina 1021 (Kaunas) 57, doi:10.3390/medicina57090957 (2021). 1022 121 Lee, A. et al. Coffee Intake and Obesity: A Meta-Analysis. Nutrients 11, 1023 doi:10.3390/nu11061274 (2019). 1024 122 Ludwig, I. A. et al. Variations in caffeine and chlorogenic acid contents of coffees: what 1025 are we drinking? Food Funct 5, 1718-1726, doi:10.1039/c4fo00290c (2014). 1026 123 Schwartz, A. & Bellissimo, N. Nicotine and energy balance: A review examining the 1027 effect of nicotine on hormonal appetite regulation and energy expenditure. Appetite 164, 1028 105260, doi:10.1016/j.appet.2021.105260 (2021). 1029 124 Schubert, M. M. et al. Caffeine, coffee, and appetite control: a review. Int J Food Sci Nutr 1030 68, 901-912, doi:10.1080/09637486.2017.1320537 (2017). 1031 125 Cornelis, M. C., Bennett, D. A., Weintraub, S., Schneider, J. A. & Morris, M. C. Caffeine 1032 Consumption and Dementia: Are Lewy Bodies the Link? Ann Neurol 91, 834-846, 1033 doi:10.1002/ana.26349 (2022). 1034 126 Di Maso, M., Boffetta, P., Negri, E., La Vecchia, C. & Bravi, F. Caffeinated Coffee 1035 Consumption and Health Outcomes in the US Population: A Dose-Response Meta- 1036 Analysis and Estimation of Disease Cases and Deaths Avoided. Adv Nutr 12, 1160- 1037 1176, doi:10.1093/advances/nmaa177 (2021). 1038 127 Mentis, A. A., Dardiotis, E., Efthymiou, V. & Chrousos, G. P. Non-genetic risk and 1039 protective factors and biomarkers for neurological disorders: a meta-umbrella systematic 1040 review of umbrella reviews. BMC Med 19, 6, doi:10.1186/s12916-020-01873-7 (2021). 48 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1041 128 Chen, X., Zhao, Y., Tao, Z. & Wang, K. Coffee consumption and risk of prostate cancer: 1042 a systematic review and meta-analysis. BMJ Open 11, e038902, doi:10.1136/bmjopen- 1043 2020-038902 (2021). 1044 129 Hong, C. T., Chan, L. & Bai, C. H. The Effect of Caffeine on the Risk and Progression of 1045 Parkinson's Disease: A Meta-Analysis. Nutrients 12, doi:10.3390/nu12061860 (2020). 1046 130 Sartini, M. et al. Coffee Consumption and Risk of Colorectal Cancer: A Systematic 1047 Review and Meta-Analysis of Prospective Studies. Nutrients 11, 1048 doi:10.3390/nu11030694 (2019). 1049 131 Loftfield, E. et al. Association of Coffee Drinking With Mortality by Genetic Variation in 1050 Caffeine Metabolism: Findings From the UK Biobank. JAMA Intern Med 178, 1086-1097, 1051 doi:10.1001/jamainternmed.2018.2425 (2018). 1052 132 Wu, L., Sun, D. & He, Y. Coffee intake and the incident risk of cognitive disorders: A 1053 dose-response meta-analysis of nine prospective cohort studies. Clin Nutr 36, 730-736, 1054 doi:10.1016/j.clnu.2016.05.015 (2017). 1055 133 Liu, Q. P. et al. Habitual coffee consumption and risk of cognitive decline/dementia: A 1056 systematic review and meta-analysis of prospective cohort studies. Nutrition 32, 628- 1057 636, doi:10.1016/j.nut.2015.11.015 (2016). 1058 134 Kennedy, O. J. et al. Systematic review with meta-analysis: coffee consumption and the 1059 risk of cirrhosis. Aliment Pharmacol Ther 43, 562-574, doi:10.1111/apt.13523 (2016). 1060 135 Crippa, A., Discacciati, A., Larsson, S. C., Wolk, A. & Orsini, N. Coffee consumption and 1061 mortality from all causes, cardiovascular disease, and cancer: a dose-response meta- 1062 analysis. Am J Epidemiol 180, 763-775, doi:10.1093/aje/kwu194 (2014). 1063 136 Ding, M., Bhupathiraju, S. N., Satija, A., van Dam, R. M. & Hu, F. B. Long-term coffee 1064 consumption and risk of cardiovascular disease: a systematic review and a dose- 1065 response meta-analysis of prospective cohort studies. Circulation 129, 643-659, 1066 doi:10.1161/CIRCULATIONAHA.113.005925 (2014). 49 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1067 137 Abalo, R. Coffee and Caffeine Consumption for Human Health. Nutrients 13, 1068 doi:10.3390/nu13092918 (2021). 1069 138 Hou, C. et al. Medical conditions associated with coffee consumption: Disease-trajectory 1070 and comorbidity network analyses of a prospective cohort study in UK Biobank. Am J 1071 Clin Nutr 116, 730-740, doi:10.1093/ajcn/nqac148 (2022). 1072 139 Paz-Graniel, I. et al. Association between coffee consumption and total dietary caffeine 1073 intake with cognitive functioning: cross-sectional assessment in an elderly 1074 Mediterranean population. Eur J Nutr 60, 2381-2396, doi:10.1007/s00394-020-02415-w 1075 (2021). 1076 140 Dong, X., Li, S., Sun, J., Li, Y. & Zhang, D. Association of Coffee, Decaffeinated Coffee 1077 and Caffeine Intake from Coffee with Cognitive Performance in Older Adults: National 1078 Health and Nutrition Examination Survey (NHANES) 2011-2014. Nutrients 12, 1079 doi:10.3390/nu12030840 (2020). 1080 141 Haskell-Ramsay, C. F. et al. The Acute Effects of Caffeinated Black Coffee on Cognition 1081 and Mood in Healthy Young and Older Adults. Nutrients 10, doi:10.3390/nu10101386 1082 (2018). 1083 142 Johnson-Kozlow, M., Kritz-Silverstein, D., Barrett-Connor, E. & Morton, D. Coffee 1084 consumption and cognitive function among older adults. Am J Epidemiol 156, 842-850, 1085 doi:10.1093/aje/kwf119 (2002). 1086 143 Kim, Y., Je, Y. & Giovannucci, E. Coffee consumption and all-cause and cause-specific 1087 mortality: a meta-analysis by potential modifiers. Eur J Epidemiol 34, 731-752, 1088 doi:10.1007/s10654-019-00524-3 (2019). 1089 144 Abdellaoui, A., Dolan, C. V., Verweij, K. J. H. & Nivard, M. G. Gene-environment 1090 correlations across geographic regions affect genome-wide association studies. Nat 1091 Genet 54, 1345-1354, doi:10.1038/s41588-022-01158-0 (2022). 50 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1092 145 De Toni, L. et al. Phthalates and heavy metals as endocrine disruptors in food: A study 1093 on pre-packed coffee products. Toxicol Rep 4, 234-239, 1094 doi:10.1016/j.toxrep.2017.05.004 (2017). 1095 146 Sakaki, J. R., Melough, M. M., Provatas, A. A., Perkins, C. & Chun, O. K. Evaluation of 1096 estrogenic chemicals in capsule and French press coffee using ultra-performance liquid 1097 chromatography with tandem mass spectrometry. Toxicol Rep 7, 1020-1024, 1098 doi:10.1016/j.toxrep.2020.08.015 (2020). 1099 147 Chieng, D. et al. The impact of coffee subtypes on incident cardiovascular disease, 1100 arrhythmias, and mortality: long-term outcomes from the UK Biobank. Eur J Prev Cardiol 1101 29, 2240-2249, doi:10.1093/eurjpc/zwac189 (2022). 1102 148 Nordestgaard, A. T. Causal relationship from coffee consumption to diseases and 1103 mortality: a review of observational and Mendelian randomization studies including 1104 cardiometabolic diseases, cancer, gallstones and other diseases. Eur J Nutr 61, 573- 1105 587, doi:10.1007/s00394-021-02650-9 (2022). 1106 149 Treur, J. L. et al. Smoking and caffeine consumption: a genetic analysis of their 1107 association. Addict Biol 22, 1090-1102, doi:10.1111/adb.12391 (2017). 1108 150 Verweij, K. J. H., Treur, J. L. & Vink, J. M. Investigating causal associations between use 1109 of nicotine, alcohol, caffeine and cannabis: a two-sample bidirectional Mendelian 1110 randomization study. Addiction 113, 1333-1338, doi:10.1111/add.14154 (2018). 1111 151 Ware, J. J. et al. Does coffee consumption impact on heaviness of smoking? Addiction 1112 112, 1842-1853, doi:10.1111/add.13888 (2017). 1113 152 Sanchez-Roige, S. & Palmer, A. A. Emerging phenotyping strategies will advance our 1114 understanding of psychiatric genetics. Nat Neurosci 23, 475-480, doi:10.1038/s41593- 1115 020-0609-7 (2020). 51 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1116 153 Jee, H. J., Lee, S. G., Bormate, K. J. & Jung, Y. S. Effect of Caffeine Consumption on 1117 the Risk for Neurological and Psychiatric Disorders: Sex Differences in Human. Nutrients 1118 12, doi:10.3390/nu12103080 (2020). 1119 154 Nehlig, A. Interindividual Differences in Caffeine Metabolism and Factors Driving 1120 Caffeine Consumption. Pharmacol Rev 70, 384-411, doi:10.1124/pr.117.014407 (2018). 1121 155 Sanchez-Roige, S. et al. CADM2 is implicated in impulsive personality traits by genome- 1122 and phenome-wide association studies in humans, with further support from studies of 1123 Cadm2 mutant mice. Translational Psychiary 13, 167, doi:10.1038/s41398-023-02453-y 1124 (2023). 1125 156 Durand, E. Y., Do, C. B., Mountain, J. L. & Macpherson, J. M. Ancestry Composition: A 1126 Novel, Efficient Pipeline for Ancestry Deconvolution. bioRxiv. doi:10.1101/010512 1127 (2014). 1128 157 Hyde, C. L. et al. Identification of 15 genetic loci associated with risk of major depression 1129 in individuals of European descent. Nat Genet 48, 1031-1036, doi:10.1038/ng.3623 1130 (2016). 1131 158 Eriksson, N. et al. Web-based, participant-driven studies yield novel genetic associations 1132 for common traits. PLoS Genet 6, e1000993, doi:10.1371/journal.pgen.1000993 (2010). 1133 159 Lam, M. et al. Large-Scale Cognitive GWAS Meta-Analysis Reveals Tissue-Specific 1134 Neural Expression and Potential Nootropic Drug Targets. Cell Rep 21, 2597-2613, 1135 doi:10.1016/j.celrep.2017.11.028 (2017). 1136 160 Sanchez-Roige, S. et al. Genome-wide association study of delay discounting in 23,217 1137 adult research participants of European ancestry. Nat Neurosci 21, 16-18, 1138 doi:10.1038/s41593-017-0032-x (2018). 1139 161 Sanchez-Roige, S. et al. Genome-wide association study of problematic opioid 1140 prescription use in 132,113 23andMe research participants of European ancestry. Mol 1141 Psychiatry 26, 6209-6217, doi:10.1038/s41380-021-01335-3 (2021). 52 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1142 162 Sey, N. Y. A. et al. A computational tool (H-MAGMA) for improved prediction of brain- 1143 disorder risk genes by incorporating brain chromatin interaction profiles. Nat Neurosci 1144 23, 583-593, doi:10.1038/s41593-020-0603-0 (2020). 1145 163 Barbeira, A. N. et al. Exploring the phenotypic consequences of tissue specific gene 1146 expression variation inferred from GWAS summary statistics. Nat Commun 9, 1825, 1147 doi:10.1038/s41467-018-03621-1 (2018). 1148 164 Barbeira, A. N. et al. Integrating predicted transcriptome from multiple tissues improves 1149 association detection. PLoS Genet 15, e1007889, doi:10.1371/journal.pgen.1007889 1150 (2019). 1151 165 Bulik-Sullivan, B. K. et al. LD Score regression distinguishes confounding from 1152 polygenicity in genome-wide association studies. Nat Genet 47, 291-295, 1153 doi:10.1038/ng.3211 (2015). 1154 166 Meddens, S. F. W. et al. Genomic analysis of diet composition finds novel loci and 1155 associations with health and lifestyle. Mol Psychiatry 26, 2056-2069, 1156 doi:10.1038/s41380-020-0697-5 (2021). 1157 167 Ge, T., Chen, C. Y., Ni, Y., Feng, Y. A. & Smoller, J. W. Polygenic prediction via 1158 Bayesian regression and continuous shrinkage priors. Nat Commun 10, 1776, 1159 doi:10.1038/s41467-019-09718-5 (2019). 1160 168 Roden, D. M. et al. Development of a large-scale de-identified DNA biobank to enable 1161 personalized medicine. Clin Pharmacol Ther 84, 362-369, doi:10.1038/clpt.2008.89 1162 (2008). 1163 169 Dennis, J. et al. Genetic risk for major depressive disorder and loneliness in sex-specific 1164 associations with coronary artery disease. Mol Psychiatry 26, 4254-4264, 1165 doi:10.1038/s41380-019-0614-y (2021). 53 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1166 170 Carroll, R. J., Bastarache, L. & Denny, J. C. R PheWAS: data analysis and plotting tools 1167 for phenome-wide association studies in the R environment. Bioinformatics 30, 2375- 1168 2376, doi:10.1093/bioinformatics/btu197 (2014). 1169 171 Kaufman, L. & Rousseeuw, P. J. Finding groups in data : an introduction to cluster 1170 analysis. (Wiley, 1990). 1171 172 Watanabe, K. et al. A global overview of pleiotropy and genetic architecture in complex 1172 traits. Nat Genet 51, 1339-1348, doi:10.1038/s41588-019-0481-0 (2019). 1173 173 Walters, R. K. et al. Transancestral GWAS of alcohol dependence reveals common 1174 genetic underpinnings with psychiatric disorders. Nat Neurosci 21, 1656-1669, 1175 doi:10.1038/s41593-018-0275-1 (2018). 1176 174 Quach, B. C. et al. Expanding the genetic architecture of nicotine dependence and its 1177 shared genetics with multiple traits. Nat Commun 11, 5562, doi:10.1038/s41467-020- 1178 19265-z (2020). 1179 175 Karlsson Linner, R. et al. Genome-wide association analyses of risk tolerance and risky 1180 behaviors in over 1 million individuals identify hundreds of loci and shared genetic 1181 influences. Nat Genet 51, 245-257, doi:10.1038/s41588-018-0309-3 (2019). 1182 176 Barkley-Levenson, A. M., Lagarda, F. A. & Palmer, A. A. Glyoxalase 1 (GLO1) Inhibition 1183 or Genetic Overexpression Does Not Alter Ethanol's Locomotor Effects: Implications for 1184 GLO1 as a Therapeutic Target in Alcohol Use Disorders. Alcohol Clin Exp Res 42, 869- 1185 878, doi:10.1111/acer.13623 (2018). 1186 177 Distler, M. G. et al. Glyoxalase 1 increases anxiety by reducing GABAA receptor agonist 1187 methylglyoxal. J Clin Invest 122, 2306-2315, doi:10.1172/JCI61319 (2012). 1188 178 Dawson, G. R. & Tricklebank, M. D. Use of the elevated plus maze in the search for 1189 novel anxiolytic agents. Trends Pharmacol Sci 16, 33-36, doi:10.1016/s0165- 1190 6147(00)88973-7 (1995). 54 medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-ND 4.0 International license . 1191 179 Henn, B. M. et al. Cryptic distant relatives are common in both isolated and 1192 cosmopolitan genetic samples. PLoS One 7, e34267, doi:10.1371/journal.pone.0034267 1193 (2012). 1194 180 Fuchsberger, C., Abecasis, G. R. & Hinds, D. A. minimac2: faster genotype imputation. 1195 Bioinformatics 31, 782-784, doi:10.1093/bioinformatics/btu704 (2015). 1196 181 Pruim, R. J. et al. LocusZoom: regional visualization of genome-wide association scan 1197 results. Bioinformatics 26, 2336-2337, doi:10.1093/bioinformatics/btq419 (2010). 55 A) λ=1. 094 Obs v er i edquantes l 10E-00 10E-02 10E-04 10E-06 Theor eti calquant il es B) C) D) medRxiv preprint doi: https://doi.org/10.1101/2023.09.09.23295284; this version posted September 11, 2023. The copyright holde preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint i A) It is made available under a CC-BY-ND 4.0 International license . pr pr g r g UKB 23andMe g B) a Bet Endocr ine/ Met abol ic Muscul oskel etalI nfect iousDi sease Ment alDi sor der s Neur o. Respi rat ory Gen.Neopl . Sense Der ma. Ci rcul ator y Bl ood I mm. Met abol ic