Background
Asialoglycoprotein receptor 1 (ASGR1) is the major subunit of asialoglycoprotein receptor (ASGPR), a liver-specific lectin that plays a role in the homeostasis of glycoprotein [
1]. Variants in
ASGR1 are associated with lower non-high-density lipoprotein (non-HDL)-cholesterol and a lower risk of coronary artery disease (CAD) [
2,
3]. Anti-ASGR1 neutralizing antibodies in mice show synergistic effects on serum cholesterol relative to some currently used lipid modifiers (i.e. statins and ezetimibe) [
4], highlighting ASGR1 as a possible therapeutic target for lowering cholesterol and preventing cardiovascular diseases (CVD) [
5].
Beyond lipid modification, the consequences of inhibiting ASGR1 are uncertain, raising the possibility of potentially important non-lipid effects. A loss-of-function variant in
ASGR1 (an intronic 12-base-pair deletion (del12)) confers a larger effect on CAD risk than is predicted by its effect on non-HDL-cholesterol in humans [
2], suggesting non-lipid pathways also contribute to its athero-protective properties. Concerns have also been raised about the possibility of ASGR1 inhibitors having adverse effects on the liver or the biliary system [
5]. ASGR1-deficient pigs have lower non-HDL-cholesterol but develop mild to moderate hepatic injury [
6]. ASGR1 inhibitors may increase the risk of cholelithiasis through elevating biliary cholesterol excretion [
4], similar to adenosine triphosphate-binding cassette transporters G5/8 (ABCG5/8) [
7]. Furthermore, the overall effect of ASGR1 inhibitors on all-cause mortality remains unclear.
To address the gap, we performed a drug-target Mendelian randomization (MR) study [
8], to assess genetically mimicked effects of ASGR1 inhibitors comprehensively in comparison with currently used lipid modifiers. First, we assessed genetically mimicked effects of ASGR1 inhibitors on all-cause mortality. Second, we investigated genetically mimicked effects of ASGR1 inhibitors on 25 traits selected
a priori as known (lipid traits and CAD) [
2,
3] or suspected (liver function and cholelithiasis) [
5,
6] effects of inhibiting ASGR1, as well as adiposity and type 2 diabetes because these are well-known effects of statins [
9] and have been suggested as general consequences of lowering LDL-cholesterol [
10,
11]. Third, we conducted a phenome-wide association study (PheWAS), i.e. a wide-ranging genotype-to-phenotype scan [
12], to examine the likely effects of ASGR1 inhibitors on a comprehensive range of health outcomes. Fourth, we conducted colocalization analysis to assess the plausibility of any associations found. We also replicated and assessed these associations by sex, where possible, because sex-specific effects are evident for some lipid modifiers [
13,
14] and for LDL-cholesterol [
15].
Discussion
Consistent with previous studies [
2,
3], we found genetically mimicked ASGR1 inhibitors associated with lower apoB, TG, total cholesterol and CAD risk. Our study adds by providing novel genetic evidence suggesting ASGR1 inhibitors reduce all-cause mortality, identifying non-lipid effects of ASGR1 inhibitors on liver function, erythrocyte traits, calcium, IGF-1 and CRP and confirming our findings using colocalization and replication.
Genetically mimicked ASGR1 inhibitors were positively associated with lifespan, possibly outperforming currently used lipid modifiers (Fig.
2). Correspondingly, a previous genetic analysis showed the del12 mutation in
ASGR1 has a greater magnitude of effect on CAD risk than other variants lowering non-HDL-cholesterol [
2]. These differences may be related to ASGR1 inhibitors reducing apoB and TG more than currently used lipid modifiers. Alternatively, other mechanisms may play a role, for example, adverse effects on weight gain and type 2 diabetes risk could detract from beneficial effects of statins on lifespan [
9].
When examining associations of genetically mimicked ASGR1 inhibitors with the 25
a priori outcomes, we found an association with higher ALP, consistent with previous MR studies [
2,
3]. ALP is a glycoprotein known to bind ASGPR, and thus inhibiting ASGR1 decreases the clearance of ALP from the circulation [
39]. Previous studies suggested ASGR1 deficiency is associated with higher ALT, AST and GGT in pigs [
6], but a loss-of-function
ASGR1 variant has little association with AST, ALT and bilirubin in humans although a mild increase in GGT and decrease in albumin cannot be excluded [
2]. Using a large sample to increase statistical power, we showed genetically mimicked ASGR1 inhibitors were associated with higher GGT and lower albumin, which was supported by colocalization. These findings suggest a potentially adverse effect of ASGR1 inhibitors on liver function. However, effects of liver function on CAD seem limited [
40‐
42], although higher GGT and albumin may increase CAD risk [
41,
42]. Mild-to-moderate elevations in aminotransferase are common in statin users, but statin-induced liver injury is rare even for those with elevated baseline liver enzymes [
43,
44]. We did not find an association of genetically mimicked ASGR1 inhibitors with cholelithiasis, in contrast to a previous hypothesis that inhibiting ASGR1 upregulates ABCG5/8 and subsequently promotes cholelithiasis [
4,
7].
In the PheWAS, we found genetically mimicked ASGR1 inhibitors were positively associated with erythrocyte traits, IGF-1 and CRP, but inversely with calcium. Previous MR studies suggest that higher reticulocyte count and possibly haemoglobin, haematocrit and red blood cell count increase CAD risk [
35,
42,
45]; higher IGF-1 increases the risk of CAD and some cancers [
46‐
49]; and CRP has a neutral role in CAD, cancer and lifespan [
50‐
52], whilst lower calcium decreases CAD risk and increases lifespan [
53,
54]. An inverse association of genetically mimicked statins with calcium has also been reported [
13]. Given the strong associations of genetically mimicked ASGR1 inhibitors with lower CAD risk and longer lifespan, these non-lipid effects would appear to be mainly of etiological interest.
Non-lipid effects of ASGR1 inhibitors generally differed from those of currently used lipid modifiers. Notably, genetic mimics of ASGR1 inhibitors were not associated with the higher BMI or type 2 diabetes risk seen for statins [
9], possibly because of different mechanisms. Statins inhibit cholesterol synthesis via 3-hydroxy-3-methylglutaryl–coenzyme A reductase (HMGCR), PCSK9 inhibitors increase LDL-receptors, and ezetimibe decreases cholesterol absorption [
55]. ASGR1 inhibitors decrease cholesterol synthesis by downregulating HMGCR and increase cholesterol clearance by upregulating LDL-receptors [
6,
56]. However, ASGR1 inhibitors also reduces lipogenesis by activating adenosine monophosphate (AMP)-activated protein kinase (AMPK) and thereby inhibiting sterol regulatory element-binding protein 1 (SREBP1) [
4]. AMPK plays an essential role in cellular energy homeostasis [
57], which may offset any detrimental effects of inhibiting HMGCR on BMI and type 2 diabetes [
9]. AMPK is also involved in the regulation of erythrocyte survival [
58], which might explain the effects of ASGR1 inhibitors on erythrocyte traits. ASGR1 inhibitors promote cholesterol excretion by upregulating liver X receptor α [
4], which may cause hepatic steatosis and elevate liver enzymes [
59]. It is also possible that endoplasmic reticulum stress-induced hepatocyte apoptosis drives the potentially adverse effect of ASGR1 inhibitors on liver function [
6].
Colocalization analysis identified rs186021206 as the SNP with the largest posterior probability for both LDL-cholesterol and CAD, which substantiates its use as a genetic mimic of ASGR1 inhibitors. Colocalization generally substantiated the findings, although the posterior probabilities were < 0.80 for lifespan, HDL-cholesterol and CAD, probably due to insufficient power given the conditional posterior probabilities for colocalization were all > 0.80 [
26]. However, the posterior probabilities for two independent variants associated with each trait were > 0.99 for AST and SHBG, suggesting the associations of genetically mimicked ASGR1 inhibitors with AST and SHBG could be confounded by linkage disequilibrium [
18].
This is the first study comprehensively investigating genetically mimicked effects of ASGR1 inhibitors in comparison with currently used lipid modifiers on lifespan and a range of potentially relevant health outcomes substantiated by an agnostic search for novel effects and colocalization. Nevertheless, this study has several limitations. First, MR should fulfil the instrumental variable assumptions of relevance, independence and exclusion restriction, that is genetic instruments should be strongly related to the exposure, share no common cause with the outcome and be independent of the outcome given the exposure [
8]. To satisfy the relevance assumption, we checked the
F-statistics for all the SNPs were > 10, suggesting weak instrument bias was unlikely. We used well-established, functionally relevant SNPs to mimic each lipid modifier to reduce the possibility of pleiotropic effects on the outcomes through pathways unrelated to the drug targets [
60]. We expressed the effects of each lipid modifier in effect sizes of LDL-cholesterol reduction. This presentation does not imply that any consequences of lipid modifiers work through LDL-cholesterol but provides an interpretable means of quantifying the MR estimates for comparability. The small number of independent genetic mimics for each lipid modifier considered precluded the use of pleiotropy robust MR methods and limited the power to identify potential effects of ASGR1 inhibitors. We used colocalization to assess the validity of the genetic mimic and any associations found for ASGR1 inhibitors. However, we cannot exclude the possibility that some effects of ASGR1 inhibitors have been missed. Second, the
ASGR1,
HMGCR,
PCSK9 and
NPC1L1 variants may affect the prescription of lipid modifiers and thus mitigate their genetic effects on lifespan. However, such complementary mechanisms would not explain the positive associations of genetically mimicked ASGR1 inhibitors with lifespan. Third, PheWAS is comprehensive but agnostic. Nevertheless, it provides insights about unknown effects of ASGR1 inhibitors, which has implications for drug development including identifying potential side-effects and elucidating mechanisms. Replication using other large GWAS excluding UK Biobank participants would be worthwhile, when available. Fourth, genetic associations for binary phenotypes were obtained using linear regression in the UK Biobank (
http://www.nealelab.is/uk-biobank/), which can inflate false positives when the case number is small and the SNP MAF is rare. However, results were validated using two common
ASGR1 variants. Fifth, MR could be open to selection bias, particularly from recruiting survivors [
61]. However, the UK Biobank participants were relatively young likely obviating selective survival to recruitment on genetic endowment for CAD. We used parental attained age as a measure of all-cause mortality, which reduces selection bias from only recruiting survivors. Sixth, associations in people of European ancestry may not apply to other populations. However, causal effects should act consistently across settings unless the mediating mechanisms differ [
62], for example, genetically mimicked effects of ASGR1 inhibitors on liver function were replicated in East Asians. Finally, MR assesses the lifelong effects of inhibiting ASGR1 which may not directly reflect quantitative effects of ASGR1 inhibitors in the short term. Further investigation is needed to confirm these findings in clinical practice.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.