Global Epidemiological Impact of PNPLA3 I148M on Liver Disease

ABSTRACT

The prevalence of metabolic dysfunction‐associated steatotic liver disease (MASLD) has increased exponentially over the past three decades, in parallel with the global rise in obesity and type 2 diabetes. It is currently the most common cause of liver‐related morbidity and mortality. Although obesity has been identified as a key factor in the increased prevalence of MASLD, individual differences in susceptibility are significantly influenced by genetic factors. PNPLA3 I148M (rs738409 C>G) is the variant with the greatest impact on the risk of developing progressive MASLD and likely other forms of steatotic liver disease. This variant is prevalent across the globe, with the risk allele (G) frequency exhibiting considerable variation. Here, we review the contribution of PNPLA3 I148M to global burden and regional differences in MASLD prevalence, focusing on recent evidence emerging from population‐based sequencing studies and prevalence assessments. We calculated the population attributable fraction (PAF) as a means of quantifying the impact of the variant on MASLD. Furthermore, we employ quantitative trait locus (QTL) analysis to ascertain the associations between rs738409 and a range of phenotypic traits. This analysis suggests that these QTLs may underpin pleiotropic effects on extrahepatic traits. Finally, we outline potential avenues for further research and identify key areas for investigation in future studies.

Abbreviations

ALD

alcohol‐related liver disease

BMI

body mass index

CLD

chronic liver disease

edQTL

RNA editing quantitative trait locus

eQTL

expression quantitative trait locus

GWAS

genome‐wide association study

HCC

hepatocellular carcinoma

HCV

hepatitis C virus

lncQTL

lncRNA expression quantitative trait locus

MASH

metabolic dysfunction‐associated steatohepatitis

MASLD

metabolic dysfunction‐associated steatotic liver disease

Mb

megabase

MENA

Middle East and North Africa

MetALD

metabolic and alcohol‐related/associated liver disease

mQTL

methylation quantitative trait locus

PAF

population attributable fraction

PNPLA3

patatin‐like phospholipase domain‐containing 3

QTL

quantitative trait locus

sICAM‐1

soluble intercellular adhesion molecule‐1

SLD

steatotic liver disease

SNP

single nucleotide polymorphism

sQTL

splicing quantitative trait locus

T2D

type 2 diabetes

tuQTL

transcript usage quantitative trait locus

Summary.

• PNPLA3 I148M is the most common and impactful genetic risk factor for progressive MASLD (including MASH, cirrhosis and HCC).

• PNPLA3 I148M risk allele frequency varies across the globe and is lowest in Africa and highest in Latin America.

• PNPLA3 I148M accounts for a substantial proportion of chronic liver disease and cirrhosis cases, and the population attributable fraction varies with allele frequency.

• Obesity, diet and other environmental factors amplify the effects of PNPLA3 I148M on MASLD and may further increase the global impact of PNPLA3 I148M.

• The presence of QTLs associated with rs738409 may explain pleiotropic effects on extrahepatic traits.

1. Introduction

Metabolic dysfunction‐associated steatotic liver disease (MASLD) has reached epidemic proportions over the past three decades, following the global rise in obesity and type 2 diabetes (T2D), and is rapidly becoming the most common cause of liver‐related morbidity and mortality [1]. Recent estimates indicate that MASLD affects over 30% of adults worldwide [1] and over 55% of individuals with obesity and T2D [2, 3]. If the current trends in obesity and metabolic syndrome continue, the burden of liver disease is projected to increase even further [4]. Although obesity has played a cardinal role in the epidemic rise of MASLD, individual differences in susceptibility are strongly influenced by genetic factors, with heritability estimates ranging between 20% and 70% [5]. In addition, MASLD prevalence varies dramatically between geographical regions and racial and ethnic groups [6]. A 2004 investigation in a multiethnic population‐based cohort from Texas, USA, found that Hispanics had nearly twofold higher prevalence of hepatic steatosis (45%) compared to Blacks (24%) and roughly 1.4‐fold higher prevalence compared to Whites (33%), despite the fact that Blacks and Hispanics had an equally elevated burden of obesity and T2D compared to Whites [7]. This observation was consistent with earlier findings from hospital‐based cohorts that showed an underrepresentation of Blacks among patients with cryptogenic cirrhosis (attributed to MASLD) [8, 9] and prompted the search for genetic underpinnings of steatotic liver disease (SLD).

In 2008, the first genome‐wide association study (GWAS) of hepatic steatosis performed in a multiethnic cohort from Dallas, TX, found a common nonsynonymous variant, rs738409 C>G, encoding p.I148M, in a previously unknown gene, patatin‐like phospholipase domain‐containing 3 (PNPLA3), that was strongly associated with hepatic fat content, assessed quantitatively by ¹H‐MR spectroscopy [10]. The variant was common in all ancestry groups represented in the study (overall G allele frequency 23%), but had the highest frequency in Hispanics (49%), followed by Whites (23%) and Blacks (14%), mirroring the prevalence of hepatic steatosis in the three groups. Together with another missense variant identified in the same gene (rs6006460 G>T, p.S453I), which is specific to Blacks, PNPLA3 I148M was shown to explain up to 70% of interethnic differences in hepatic fat content in the study cohort [10].

Numerous studies performed since showed that PNPLA3 I148M is associated with a full spectrum of SLD, including hepatic steatosis, steatohepatitis or MASH (metabolic dysfunction‐associated steatohepatitis), fibrosis, cirrhosis, hepatocellular carcinoma (HCC), and liver‐related mortality, in patients with MASLD, alcohol‐related SLD (MetALD and ALD depending on the severity of alcohol consumption), hepatitis C and mixed disease etiologies [11, 12, 13, 14, 15, 16, 17, 18]. The association between PNPLA3 I148M and SLD has been replicated in various geographical regions and in paediatric populations as well as in adults (reviewed in [19]).

Unlike many other variants discovered in GWAS, that had modest disease odds ratios in the range of 1.1–1.2 per allele, PNPLA3 I148M has had an unusually large effect size, with odds ratios of 1.5–3 per allele for progressive forms of MASLD (including MASH, cirrhosis and HCC), and proportionally higher odds ratios among GG homozygotes, consistent with an additive genetic model for quantitative traits and multiplicative model for dichotomous outcomes [10, 11, 19]. This large effect size has led some authors to compare the impact of PNPLA3 I148M to alleles causing monogenic (Mendelian) disorders [20]. After 15 years of GWAS, which included increasingly large sample sizes, PNPLA3 I148M remains the most common and impactful genetic risk factor for SLD. Given both its high frequency and large effect size, I148M was estimated to have the highest population attributable fraction (PAF) for chronic liver disease (CLD) and cirrhosis in European populations, compared to other GWAS‐identified variants [21]. Here we review the contribution of PNPLA3 I148M to global burden and regional differences in MASLD, focusing on recent evidence emerging from population‐based sequencing studies and prevalence assessments.

2. Regional Variation in Prevalence of MASLD and Frequency of PNPLA3 I148M (rs738409)

PNPLA3 I148M is common in all parts of the world, but the risk (G) allele frequency varies widely across the globe (Figure 1). Data from large‐scale genome sequencing projects (including the 1000 Genomes Project [29] and gnomAD [30]) indicate that the allele frequency is lowest in Sub‐Saharan Africa (12%), followed by Europe (23%), South Asia (24%–30%), East Asia (35%–45%), and is highest in Central and South America (~50%). Substantial variation has also been observed across the Americas, a region with a complex genetic structure shaped by historical admixture events among indigenous Native American, European, and African populations as a result of European colonisation, migration and transatlantic slave trade [31]. Accordingly, the lowest frequency of PNPLA3 I148M has been reported in populations with a higher proportion of African ancestry (e.g., Dominican Republic, 25%) and highest in populations with a greater proportion of indigenous Native American ancestry (e.g., Mexico, 52%). Notably, some studies report that G allele is the major (more common) allele in Peru (72%) [29] and Guatemala (69%) [32].

FIGURE 1.

Worldwide allele frequency distribution of PNPLA3 rs738409 I148M. Allele frequencies are based on data from the 1000 Genomes Project and gnomAD database, downloaded from NCBI dbSNP: https://www.ncbi.nlm.nih.gov/snp/rs738409/. Estimates of SLD prevalence derived from recent meta‐analyses and population‐based studies [1, 22, 23, 24, 25, 26, 27, 28].

These differences broadly parallel differences in worldwide prevalence of MASLD (Figure 1) [1, 22, 23, 24]. According to the most recent meta‐analyses, the highest prevalence of MASLD was estimated to be in Latin America (44.4%), followed by Middle East and North Africa (MENA) (39.4%) [23], South Asia (33.8%), South‐East Asia (33.1%), North America (31.2%), East Asia (29.7%), Asia Pacific 28.0% and Western Europe 25.1% [1]. The data for Sub‐Saharan Africa are more sparse, but previous meta‐analyses estimated MASLD prevalence to be around 13.5% [25]. Several studies have also reported that the prevalence of MASLD varies among Hispanic/Latino populations of different origin. For example, in the US, the prevalence of MASLD was reported to be highest among Hispanics of Mexican origin and lowest among those of Cuban, Dominican and Puerto Rican origin [26, 27, 33]. Substantial variation in prevalence of MASLD has also been reported among countries in Central and South America [28]. Differences in the frequency of the PNPLA3 148 M allele likely contribute at least partially to this variation in prevalence, although one study revealed that the proportion of Native American ancestry was an independent risk factor for MASLD in Hispanic/Latino adults, even after accounting for PNPLA3 genotype [34]. Furthermore, variation in the prevalence of obesity, extent of urbanisation, dietary composition and lifestyle habits likely also contribute to regional differences in MASLD prevalence.

One commonly used measure to quantify the impact of genetic variants on disease is the PAF or the fraction of disease cases that would not have occurred if the risk allele had been removed from the population. The PAF is a function of the effect size (relative risk of disease conferred by the variant) and the allele frequency (i.e., the proportion of individuals with the risk genotype(s)) [35]. We estimated the impact of PNPLA3 I148M on the global and regional prevalence of MASLD, using previously reported odds ratios [19] and allele frequencies. Data on the effect size of PNPLA3 I148M in different regions are sparse. However, most studies which performed ancestry‐stratified analyses found concordant effects of PNPLA3 I148M on MASLD across ancestries [36, 37]. While some studies reported significant heterogeneity in the effects of I148M on MASLD and cirrhosis across ancestries [37, 38], this may have been due to the small sample size for some ancestry groups [37] or heterogeneous characteristics of study cohorts, from which subjects of European and non‐European ancestries were derived [38]. Therefore, we used a range of odds ratios, similar to those reported in previous studies of European subjects [11, 19, 39].

Assuming a similar effect size of I148M on MASLD across different ancestries and regions, the variant would be expected to account for a much larger fraction of cases in Latin America vs. Africa due to regional variation in allele frequency (Table S1). For example, assuming an allelic odds ratio of 1.9 [19], PNPLA3 I148M accounted for 18.7% of MASLD cases in Africa, whereas it accounted for > 50% MASLD cases in Latin America. Assuming a larger odds ratio of 2.5 (as reported in some previous studies), the corresponding PAF increased to 28% in Africa and 67% in Latin America. These differences are a consequence of a higher total number of individuals with at‐risk genotypes among Latin American populations, but also a higher fraction of risk‐allele homozygotes, who have higher relative odds of MASLD than heterozygotes. Although high PAF estimates should be interpreted with caution [35], they illustrate the contribution of the PNPLA3 I148M allele to regional differences in the MASLD burden.

Data on the prevalence of MASH are less abundant because accurate diagnosis usually requires liver biopsy or accurate estimates of liver inflammation and fibrosis. Nevertheless, estimates from meta‐analyses suggest that similar regional differences exist in the prevalence of MASH, with the highest prevalence reported in Latin America (7.1%), followed by MENA (5.85%), South Asia (5.4%), South‐East Asia (5.3%), North America (5.0%), East Asia (4.8%), Asia Pacific (4.5%) and Western Europe (4.0%). These estimates were based on the average progression rates and MASH rates among patients with steatosis [4]. However, given the fact that PNPLA3 I148M is associated with MASH and fibrosis, one might expect to see an even greater burden of MASH and cirrhosis in countries with a higher prevalence of I148M.

It should be noted that in addition to I148M, several other coding PNPLA3 variants have been identified that may contribute to global MASLD prevalence. One of these is the missense variant rs6006460 G>T, encoding S453I, which was shown to be associated with a lower prevalence of hepatic steatosis, independent of I148M [10, 37, 40]. In contrast to the I148M variant, S453I is common in African‐ancestry populations (allele frequency 10%–11%) but virtually absent in other ancestry groups (Figure S1). The two variants appear to act in an additive fashion, with no documented interaction between rs738409 and rs6006460 [10]. Nonetheless, given a similar frequency of S453I and I148M among African populations, it is possible that the protective S453I variant may contribute to the lower overall prevalence of MASLD in African‐ancestry individuals. Another variant is PNPLA3 rs2294918 G>A (encoding E434K), which was shown to influence the risk of MASLD by altering the expression of PNPLA3 and interacting with the I148M variant in one study [41]. In particular, the risk G (434E) allele was shown to increase the expression of both 148I and 148M proteins (without changing protein activity), while the 434K allele reduced the expression of 148M and 148I, thus dampening the deleterious effect of 148M and the protective effect of 148I [41]. Unlike I148M and S453I variants, the risk rs2294918‐G allele frequency is highest in Africa (86%) and Asia (85%) and lowest in Europe (59%) [30], which may also contribute to the impact of I148M on MASLD burden in these regions, although more data are needed to evaluate the magnitude of its effect.

3. What Factors Affect the Impact of I148M?

While available data indicate that the effect sizes of PNPLA3 I148M on MASLD are similar across different geographical regions, the risks have been shown to be amplified by obesity and other metabolic factors (as discussed elsewhere in this issue). For example, one study found that the odds ratio of cirrhosis for GG homozygotes versus CC homozygotes was increased from 2.4 among lean individuals (those with BMI < 25 kg/m²) to 5.8 in the obese (those with BMI > 35 kg/m²) [42]. Thus, the contribution of PNPLA3 I148M to the global burden of SLD (including MASLD and MetALD) may also depend on the prevalence of obesity and other metabolic risk factors in different world regions. For example, this may account for the exceptionally high prevalence of MASLD in the MENA region, which has reported some of the highest prevalence rates for obesity and T2D [23]. This may also account for the lowest reported prevalence of MASLD in Peru (12.5%), which has the highest known frequency of PNPLA3 I148M, but where only 15% of the population is obese [28]. Given the projected trends in obesity rates, especially in the developing world, one may therefore expect that the impact of PNPLA3 I148M will be even greater in populations with a higher burden of obesity [4].

It is noteworthy that the majority of genetic studies, including GWAS, employed an additive model of inheritance. While this is a reasonable assumption for complex traits, this approach neglects possible synergistic effects, such as the genotype‐by‐environment interactions (G × E) described above, and genotype‐by‐genotype interactions (epistasis), which can sometimes manifest as non‐linear (or non‐additive) effects. A number of intriguing examples of these effects have been documented. For example, Lazo et al. identified evidence of an interaction between the presence of the PNPLA3 G allele and the association between moderate alcohol consumption and hepatic steatosis, demonstrating that the effect of genotype depended on level of alcohol consumption [43]. Furthermore, Cherubini et al. identified a robust interaction between the PNPLA3 p.I148M variant and female sex, providing evidence that epistasis contributes to the genetic architecture of common traits such as MASLD [44].

In addition, there have been interesting studies focusing on gene‐diet interaction effect(s), for example, a recent study assessing a gene‐diet interaction between rs738409, nutrient intake and severity of liver histology [45]. Vilar‐Gomez et al. [45] showed that the PNPLA3 rs738409 G‐allele may modulate the effect of specific dietary nutrients on the risk of fibrosis in patients with NAFLD. Other human studies have investigated the direct effect of PNPLA3 rs738409 on the development of liver fibrosis in relation to liver histological features. In particular, a large proportion of the indirect effect of rs738409 on fibrosis severity was reported to be mediated by portal inflammation [46].

Recent studies have highlighted the influence of genetic variants, including variants influencing risk and protection against MASLD histological severity, including rs738409, on liver microbial DNA composition [47]. For example, Pirola et al. [47] found that members of the Gammaproteobacteria class were significantly enriched in carriers of the rs738409, including the genera Enterobacter.

Additionally, intriguing interactions have been observed between PNPLA3 rs738409 and a loss‐of‐function variant in HSD17B13 (rs72613567), which is linked to a decreased susceptibility to chronic liver disease and the progression from steatosis to steatohepatitis. For example, the rs72613567:TA variant mitigated liver injury associated with the risk‐increasing PNPLA3 I148M allele and resulted in an unstable and truncated protein with reduced enzymatic activity in a sample of 46 544 participants in the DiscovEHR study [48]. Furthermore, the analysis of a potential PNPLA3‐HSD17B13 interaction assessed in 1153 non‐Hispanic whites with biopsy‐proven nonalcoholic fatty liver disease enrolled in the nonalcoholic steatohepatitis. Clinical Research Network studies indicated that the protective impact of the HSD17B13 rs72613567 TA‐allele on the likelihood of developing steatohepatitis and fibrosis may be constrained to specific demographic subgroups, including individuals aged 45 years and above, females, those with class 2 obesity or diabetes, and those with the PNPLA3 rs738409 CC genotype [49]. In Japanese patients with MASLD, carriage of the HSD17B13 rs6834314‐G allele (a variant in linkage disequilibrium with rs72613567) had the effect of moderating the impact of the PNPLA3 rs738409 GG genotype on the development of advanced hepatic fibrosis [50]. Finally, a study focused on chronic hepatitis C showed that carriage of the combination PNPLA3 minor allele and HSD17B13 major allele may represent a risk factor for HCC among HCV‐infected patients [51].

4. Molecular Impact and Variant Functionality

The elucidation of the mechanism linking genetic variants to diseases has been facilitated by the widespread application of expression quantitative trait loci (eQTLs) analysis. This analysis has enabled the identification of causative variants that colocalise with GWAS variants, thus providing explanations for some of the observed relationships.

The analysis of gene expression regulation using eQTLs can be addressed by exploring cis‐eQTLs or trans‐eQTLs, which are classified depending on their genomic location [52]. Cis‐eQTLs affect local genes, whereas trans‐eQTLs affect distant genes, even on different chromosomes.

Although eQTLs have been extensively employed for the identification of disease‐associated genetic variants, they only account for 20%–50% of disease association variants. Cis‐eQTLs, where gene expression levels are influenced by a gene‐proximal single nucleotide polymorphism (SNP) located within 1 megabase (Mb), have been widely employed for this purpose. Nevertheless, cis‐eQTLs typically account for a relatively small proportion of disease heritability, which suggests the existence of additional pathways to disease. Trans‐eQTLs typically exhibit reduced effect sizes in comparison to cis‐eQTLs, necessitating greater statistical power to detect them effectively. This is because the SNP in question is typically situated at a greater genetic distance (exceeding 5 megabases) from the gene in question or on a different chromosome [53].

Furthermore, quantitative trait loci (QTLs) can influence not only gene expression but also gene splicing. Splicing quantitative trait loci (sQTLs) are significantly enriched in functional elements of the genome [52]. Collectively, genetic variants that influence gene expression offer a potential molecular rationale for their observed effects.

We conducted a quantitative trait locus (QTL) analysis to identify associations between rs738409 and various phenotypic traits. This analysis employed a range of resources, including FIVEx [54], an interactive eQTL/sQTL browser accessible at https://fivex.sph.umich.edu, and the QTLbase, which compiles and curates genome‐wide summary statistics for human molecular traits across > 70 tissue/cell types. Figure 2 shows significant cis‐eQTLs associated with the rs738409 (22:43928847_C/G), including tissue, sample size, effect, and p‐value. A comprehensive list of QTLs is shown in Table S2. Of note, there are two significantly associated blood mQTLs (methylation quantitative trait locus), five sQTL (splicing quantitative trait locus) one located in PNPLA3 and four in SAMM50, six blood pQTLS (protein quantitative trait locus) in nearby loci (P17405 [SMPD1]; Q02083 [NAAA]; P53365 [ARFIP2]; P07686 [HEXB]; P00326 [ADH1C]; and Q9NZK5 [ADA2]), two edQTL (RNA editing quantitative trait locus) in the liver, one lncQTL (lncRNA expression quantitative trait locus) in thyroid gland and nine stem cell‐iPSC tuQTL (transcript usage quantitative trait locus). The rs6006460, on the other hand, only has two blood‐monocyte eQTLs in SAMM50 and one mQTL in brain‐prefrontal cortex (Table S2).

FIGURE 2.

cis‐eQTLs associated with variant: rs738409 (22:43928847_C/G) and various phenotypic traits. This analysis employed a range of resources, including FIVEx [54], and the QTLbase.

In addition to the effect of the variant on liver phenotypes, the presence of these QTLs may explain pleiotropic effects on extrahepatic traits [55, 56, 57]. For instance, blood eQTLS and mQTLs might explain the potential involvement of rs738409 in COVID‐19 outcomes [58, 59, 60, 61]. There are many other examples of diseases associated with rs738409 or other variants in PNPLA3, for example, kidney diseases [62] or even rare conditions such as cystathioninuria and microcephaly‐congenital cataract‐psoriasiform dermatitis syndrome among many other diseases [56].

The rs738409 has been shown to be associated with immune‐related traits [56], including plasma soluble intercellular adhesion molecule‐1 (sICAM‐1) levels in a large GWAS of 22 435 healthy women from the Women's Genome Health Study [63].

Finally, a recent report identified a colocalisation between a GWAS signal for type 2 diabetes and the missense lead variant rs738409, as well as a trans‐pQTL for ADP ribosylation factor interacting protein 2 (ARFIP2) [64]. This latter protein is strongly downregulated in individuals with type 2 diabetes, as reported by Gudjonsson et al. [64] in the AGES study. This observation prompts a number of inquiries, including the manner by which a missense variant in PNPLA3 affects the circulatory levels of ARFIP2, the role of ARFIP2 in the interpretation of PNPLA3 function and the relationship of ARFIP2 to type 2 diabetes.

Therefore, not only a complete understanding of the liver‐related traits associated with rs738409 is required but also a thorough understanding of the gene and protein interactions, the active protein ligands and, most importantly, an accurate and comprehensive assessment of the impact of the pleiotropic effects of the variant.

5. Conclusion, Unanswered Questions and Future Research Opportunities

In conclusion, PNPLA3 I148M is the most common and impactful risk variant for progressive MASLD and likely other forms of steatotic liver disease. It affects all regions and contributes significantly to global variation in MASLD. The effect of PNPLA3 I148M is magnified by obesity; thus, its impact on burden of MASLD may be even greater if the obesity rates continue to rise. Despite the well‐established role of PNPLA3 I148M in chronic liver disease, several unanswered questions remain.

One puzzling question is how a variant of such high impact on disease risk became so common.

The effect size of PNPLA3 rs738409 is unusually large given its frequency, compared to most other common variants identified in GWAS, and contradicts population genetic theories that posit that variants of large effects are kept at a low frequency as a result of negative selection [65]. This begs a question of whether the risk allele conferred a selective advantage in cold climates faced by early humans. One recent study found no evidence of selection in the last 10 000 years of human history [66], but this conclusion remains to be confirmed.

Another area requiring further study is whether rs738409 can be an explanation for extrahepatic manifestations of MASLD and MASH. In fact, this variant yields disparate outcomes contingent on the specific tissue under examination. The rs738409 cis‐eQTLs in the aortic and coronary arteries have been associated with positive effects, whereas other cis‐eQTLs in the skin and brain have been linked to negative effects (Figure 2). Furthermore, it would be of interest to ascertain whether the QTLs of rs738409 exert an influence on the global prevalence and incidence of MASLD.

Finally, given the high prevalence and effect size of PNPLA3 I148M, one might ask whether genetic testing should be considered as part of patient management, as the field moves towards precision medicine [67].

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

TABLE S1. Attributable fraction of PNPLA3 rs738409 I148M on liver disease.

TABLE S2. QTLs associated with rs738409 and rs6006460.

FIGURE S1. Worldwide allele frequency distribution of PNPLA3 rs6006460 S453I. Allele frequencies are based on data from the 1000 Genomes Project and gnomAD database, downloaded from NCBI dbSNP: https://www.ncbi.nlm.nih.gov/snp/rs6006460. Estimates of SLD prevalence derived from recent meta‐analyses and population‐based studies [1, 26–31, 33].

Data Availability Statement

All data used in the manuscript are publicly available from cited online databases or included in Supporting Information.