In this study, we sought to identify brain proteins associated with coffee consumption subtypes by PWAS and TWAS. Our study focused on the overlapping PWAS and overlapping TWAS genes in discovery and replication cohort, as well as genes that were shared by PWAS and TWAS. Subsequently, LDSC was used to demonstrate the genetic association and overlapping genetic architecture between coffee consumption subtypes and plasma proteins and peripheral metabolites.
For all the results of PWAS and TWAS, we found some meaningful genes. The first gene to note is ALDH2. Among the genetic variants encoding several alcohol-metabolizing enzymes, the alcohol dehydrogenase (ADH) and aldehyde dehydrogenase (ALDH) are two of the most genetically related to the risk of alcohol dependence (39). Among them, ALDH2 is known for its restriction of alcohol consumption and protection of alcoholism, and mutations and impaired enzyme activity of ALDH2 will lead to the accumulation of acetaldehyde (40). The point is similar to the GO terms of ALDH2 shown in GCBI: GO:0006066 ~ alcohol metabolic process, GO:0006068 ~ ethanol catabolic process. ALDH2 dysfunction has been confirmed to be involved in a wide range of human diseases, such as breast cancer, gastric cancer, colorectal cancer, Alzheimer's disease, and cardiovascular disease (41–43). Of note, these diseases have evidence that they are related to coffee consumption (44, 45). In summary, we have sufficient evidence to show that ALDH2 is related to coffee consumption to a certain extent. Similarly, for the ARPC2, we also found some related GO terms from GCBI, such as GO:0034314 ~ Arp2/3 complex-mediated actin nucleation, GO:0070358 ~ actin polymerization-dependent cell motility and so on. Previous studies have also confirmed that these functions affect alcohol dependence and alcohol consumption through various biological pathways. (46, 47). Specially, from GCBI, we found UBE3B related GO terms, including GO:0006511 ~ ubiquitin-dependent protein catabolic process and GO:0000209 ~ protein polyubiquitination. From previous researches, we can find that the ubiquitination system may play a major role in the biology of synaptic plasticity (48, 49), and the density of UBE3B immunoreactive pyramidal neurons is decreased in schizophrenia subjects (50). In other words, UBE3B is associate with the human brain's nervous system. We're still trying to figure out whether the UBE3B relationship with the human brain's nervous system affects coffee consumption.
To determine the relationship between the findings of our study and coffee consumption subtypes, we explore the related diseases of the results in our study and summarized the top 8 diseases by using GCBI and searching literature (Table 4). In these diseases, a total of 9 genes are associated with breast cancer, such as ARPC2, ALDH2 and MADD. For example, ZHONGLE et al. have reported that ARPC2 promotes proliferation and metastasis of breast cancer (51), and the relationship between coffee consumption and breast cancer has also been confirmed, the study of Cristina et al. mentioned that drinking coffee is negatively associated with the risk of breast cancer in postmenopausal women (52). In addition, in our study, there are 5 genes related to gastric cancer, including ALDH2, ARPC2, C14orf159 and so on. ALDH2 polymorphism changes the risk of gastric cancer (53), and ARPC2 plays an analgesic effect in gastric cancer, providing a new target for gastric cancer treatment (54). It is worth noting that the relationship between gastric cancer and coffee consumption has also been confirmed, and coffee may have a potentially beneficial effect on the incidence and mortality of gastric cancer (55). Moreover, our findings are also related to Alzheimer's disease, colorectal cancer, intellectual disability, melanoma, lung cancer, cardiovascular disease. Interestingly, there is evidence that these diseases are related to coffee consumption (5, 55).
In addition, a total of 11 common plasma proteins are significantly genetically related to coffee consumption subtypes both in discovery and replication cohorts, including Neural cell adhesion molecular L1-like protein (CHL1), Procollagen-lysine,2-oxoglutarate 5-dioxygenase 3 (PLOD3) and so on. Among those proteins, CHL1was associated with the development and progression of tumors more than once (56, 57). which is consistent with our study result. In addition, GCBI also suggests that CHL1 is associated with schizophrenia, depression, and a previous study suggested that coffee and caffeine consumption were significantly associated with decreased risk of depression (58).
5 overlapping blood metabolites (n-Butyl Oleate, myo-inositol, X-11423, 1-arachidonoylglycerophosphoinositol*, 1-palmitoylglycerophosphoethanolamine) were associated with the coffee consumption-related traits both in the discovery and replication cohort. Myo-inositol (MI) has been widely studied as an insulin sensitizing factor. It has the functions of increasing insulin sensitivity, reducing hyperandrogenism, and improving the menstrual cycle (59). Therefore, it is often used to treat and improve polycystic ovary syndrome (PCOS) and gestational diabetes mellitus (GDM) (59, 60).
On the whole, our study has several strengths. Firstly, this study was conducted in two independent GWAS datasets, increasing the credibility of the study. Secondly, most previous studies only used PWAS for analysis. In this study, PWAS and TWAS were carried out simultaneously, and the results of TWAS were used to verify the results of PWAS. Third, for PWAS analysis, we combined two reference human proteomes with two independent GWAS datasets to increase the accuracy of the study results. Similarly, for TWAS analysis, we also integrate two RNA expression weights with two independent GWAS datasets. Finally, we analyzed the genetic correlation between coffee consumption subtypes and human blood metabolites and human plasma proteomes from the perspectives of metabolomics and proteomics, which has never been reported in previous studies.
Nonetheless, there are two limitations of this study that should be noted. First, the GWAS datasets were all from European samples, and although large samples have been included, we need to be cautious about the promotion of research results in other ethnic groups. Second, although we used two reference human brain proteomes in both discovery and replication cohorts at the same time, this study still is limited by the number of reference brain proteomes(NROS/MAP=376, NBanner=152). Larger reference human brain proteomes are needed to alleviate this issue.