Genetically Proxied Therapeutic Effect of Metformin Use, Blood Pressure, and Hypertension’s Risk: a Drug Target-Based Mendelian Randomization Study

In this work, we aim to evaluate the association of the genetically proxied effect of metformin on blood pressure (BP) and hypertension through a drug target-based Mendelian randomization (MR) analysis. Thirty-two instrumental variables for five metformin targets (i.e., AMP-activated protein kinase (AMPK), growth differentiation factor 15 (GDF15), mitochondrial glycerol 3 (MG3), mitochondrial complex I (MCI), and glucagon (GCG)) were introduced to the MR analysis on the datasets of hypertension, systolic and diastolic blood pressure (SBP and DBP). The MR analyses demonstrated that the MCI- and MG3-specific metformin’s use would significantly reduce SBP, DBP, and hypertension risk. The meta-analyses showed that the genetically proxied metformin’s use equivalent to a 6.75 mmol/mol reduction on HbA1c could decrease both the SBP (beta =  − 1.05, P < 0.001) and DBP (beta =  − 0.51, P = 0.096). Furthermore, metformin’s use was also implied to reduce the hypertension risk. The MG3- and MCI-dependent metformin’s effect may play key roles in the anti-hypertension function. Graphical Abstract Supplementary Information The online version contains supplementary material available at 10.1007/s12265-023-10460-z.


Introduction
Hypertension is the leading modifiable risk factor for cardiovascular diseases, affecting more than 1.5 billion people globally [1].Metformin, one of the most widely recommended first-line pharmacotherapies for managing type 2 diabetes (T2D), has been reported to improve cardiovascular outcomes which probably via lowering blood pressure (BP) [2].Fanghänel et al. [3] noted that metformin treatment lowered BP in patients with diabetes.Similar findings were reported in several randomized controlled trials (RCTs) involving patients with portal hypertension [4].A meta-analysis of 26 RCTs containing 4119 patients indicated that metformin could effectively decrease systolic blood pressure (SBP) [5].However, previous findings were inconsistent, with some studies indicating no remarkable effect of metformin on reducing BP [5].A meta-analysis based on 41 RCTs (3074 patients) reached the opposite conclusion that metformin had no significant effect on BP [6].Notably, these studies had intrinsic methodologic limitations, including small sample sizes and selection bias.Whether metformin had an effect on BP and hypertension is currently unclear.
Mendelian randomization (MR) analysis, a complementary and alternative approach to RCTs, is a powerful statistical method that uses the significantly associated single nucleotide polymorphisms (SNPs) as instrumental variables (IVs) to quantify potential causal effects.Drug exerts their effects by regulating the pharmacological targets' expression, and the naturally occurring human genetic variation can serve as a proxy for therapeutic drug targets' reaction [7].A recently developed extension to the MR paradigm, i.e., the drug target-based MR study, has been used to find the drug-repurposing candidates for various diseases [8].The objective of this study was to estimate the causal effect of metformin on hypertension in a larger European population using drug target-based MR.The present findings may potentially offer novel prevention and treatment strategies for hypertension.

The Genetic IVs for Five Metformin Targets
Zheng and colleagues have explored the causal relationship between metformin and Alzheimer's disease risk through a drug target-based MR approach [9].They have identified 32 IVs for five primary drug targets (i.e., AMP-activated protein kinase (AMPK), growth differentiation factor 15 (GDF15), mitochondrial glycerol 3 (MG3), mitochondrial complex I (MCI), and glucagon (GCG)) and 23 drug target-associated genes (DTGs) of metformin (see Fig. 1 and supplementary Table 7A of Zheng et al. [9]).In order to identify the valid IVs, their selection process not only complied with the three key assumptions of MR analysis but also performed additional tests.For instance, they confirmed the human tissue expression of the selected genes with the data from GTEX, eQTLGen, and Zheng et al. [10][11][12].Besides, as a positive control, the authors verified metformin's effect on reducing T2D risk and HbA1c using the MR method [9].According to the authors, one standard deviation (SD) unit lowering of the HbA1c equals to 6.75 mmol/mol reduction of HbA1c [9].Besides, the current study was also conducted under the guidance of the STROBE-MR checklist (see Table S1).

The Information on the Outcome Datasets of Hypertension and Blood Pressure
The GWAS datasets for hypertension were acquired from two independent cohorts.Hypertension cohort 1 came from the UK biobank and contained 462,933 participants (119,731 hypertension patients and 343,202 healthy controls) [13].Hypertension cohort 2 derived from the Finnish biobank project FinnGen (https:// www.finng en.fi/ fi).Its sample size was 218,754, with 55,917 being hypertension patients and 162,837 being healthy controls [14].Their GWAS summary statistics data were acquired through the platform of the IEU OpenGWAS project (https:// gwas.mrcieu.ac.uk/ datas ets, GWAS ID: "ukb-b-14057," "finn-b-I9_HYPTENS") [15].The GWAS data for SBP and diastolic blood pressure (DBP) were provided by the International Consortium of Blood Pressure (ICBP) [16].The study included 757,601 participants of European descent.The basic information of the included four outcome datasets was summarized in Fig. 1.

MR Analysis and Meta-analysis
In the MR analysis, the inverse-variance weighted (IVW) (when IVs > 2) and Wald ratio (when IVs ≤ 2) approaches were selected to be the major analytical tools.Other sensitivity methods such as MR Egger, weighted median, weighted mode, and simple mode were employed to assess the robustness of the conclusions of the MR analysis with IVs > 2. The fixed-effect and random-effect statistical models were employed for the meta-analysis.All analyses were conducted with R packages MRPRESSO (version 1.0), ieugwasr (version 0.1.5),TwoSampleMR (version 0.5.6), and meta (version 6.0-0) [15,17,18].A P value less than 0.05 was considered statistically significant.

Test the Influence of Potential Confounders
To test whether the metformin target-associated blood lowering effect was influenced by the confounders such as body mass index (BMI), smoking, and alcohol consumption, the multiple variable Mendelian randomization (MVMR) was performed.Their GWAS summary statistics data were acquired through the platform of the IEU OpenGWAS project (https:// gwas.mrcieu.ac.uk/ datas ets, GWAS ID: "ukb-b-19953," "ieu-b-4877," and "ukb-b-5779").

The MR Analyses of 5 Drug Targets on SBP, DBP, and Hypertension
The MR estimates of metformin drug targets on SBP, DBP, and 2 hypertension cohorts were summarized in Table 1.When the IVs for exposure and its potential proxy SNPs cannot be found in the outcome dataset, they were excluded from further analyses.Therefore, only four targets were kept for the analyses on SBP.We noticed that the MG3-and MCI-dependent metformin effects were confirmed to reduce both SBP and DBP significantly.Consistently, the MCI-specific metformin influence was also identified to remarkably diminish the risk of hypertension in both tested hypertension cohorts.Besides, in Fig. 1 The workflow of the current work hypertension cohort 1, the MG3-and AMPK-dependent metformin effects also demonstrated obvious anti-hypertension function.As shown in Supplementary Table S2-S5, the pleiotropy tests of AMPK-and MCI-specific metformin's effect on SBP, DBP, and hypertension indicated no significant results.We also did not detect remarkable heterogeneity results in the AMPK study.However, the test of MCI demonstrated significant heterogeneity results (Supplementary Table S2-S5).The sensitivity analyses on AMPK and MCI suggested that the effects were robust to various MR approaches (Supplementary Table S6-S7).

Metformin's Effect on the SBP and DBP
In the MR analysis, the GCG-specific SNP and its proxy were not found in the GWAS dataset of SBP, and therefore, it was excluded from the meta-analysis.As shown in Fig. 2A, the metformin's targets could decrease both the SBP and DBP.In the analysis of SBP, four metformin targets were all confirmed to influence blood pressure reduction.The MG3-, MIC-, and AMPK-specific metformin's effects were statistically significant.Similarly, these four targets also demonstrated the ability to reduce DBP, with the MG3-and MCI-specific effects being proven to be statistically significant (see Fig. 2B).Notably, the MG3-specific influence was confirmed to have the most substantial effect among the four targets.

Metformin's Effect on the Risk of Hypertension
In the MR analysis, we evaluated the effects of metformin on two independent hypertension cohorts.Though the random effect model meta-analysis suggested no remarkable effect of metformin target on reducing hypertension's risk, the common effect model indicated a significant effect of metformin targets on decreasing hypertension risk (see Fig. 2C and D).Among five targets, the AMPK-and MCI-specific effects of metformin were consistent in two MR analyses and played key roles in reducing hypertension's risk.In the MR analysis on hypertension cohort 2, the MG3-specific metformin's effect showed no significant effect on the risk of hypertension.However, it was identified to reduce the risk of hypertension in cohort 1 significantly.The GCG-specific metformin's effect was confirmed to have little effect on hypertension's risk.The GDF15-specific metformin effect was implied to increase hypertension risk in two MR analyses.

The Analyses of Potential Confounders
In order to rule out the potential influence of some potential confounding factors (e.g., BMI, smoking, and alcohol consumption) on the metformin targets-specific HbA1c lowering effect, the MVMR analyses were conducted.The basic information of three potential confounder datasets was summarized in Supplementary Table S8.However, during the process of extracting the IVs of exposures for MVMR, we noticed that the valid IVs representing the BMI, smoking, and alcohol consumption had

Discussion
To the best of our knowledge, the present work is the first MR study to investigate metformin's genetically predicted therapeutic effects on BP and hypertension.Basing on the data from large-scale GWAS studies in two independent hypertension cohorts of 175,648 patients, we observed that genetically proxied metformin use leads to a 13% reduction of hypertension risk, which may be owed to its ability to decrease SBP and DBP.Among five drug targets, metformin may exert its anti-blood pressure activities majorly through regulating MG3 and MCI.Collectively, these findings reported a suggestive benefit of metformin in reducing BP and provided novel evidence to guide hypertension prevention.
Metformin is recommended as the first-line treatment for T2D and is proposed to be beneficial for cardiovascular outcomes.Several earlier RCTs have suggested that metformin can significantly reduce BP in diabetic and hypertensive individuals.In addition to these human studies, several rodent studies have also provided evidence for BP-lowering effect of metformin.Kosegawa et al. observed that the SBP levels were significantly lower in the metformin group than in the controls group [19].The effect of metformin on reducing BP in streptozotocin-induced diabetic rats was also observed in another animal trial conducted by Majithiya and Balaraman [20].These results from RCTs and experimental studies are consistent with our findings.However, in contrast, some studies did not confirm a significant effect of metformin on reducing BP.Although the reason is unclear, the conflictive results might be due to multiple causes, such as small sample sizes, selection bias, differences in study populations, and strategies of BP measurement.Further prospective RCTs with multi-ethnic, larger sample sizes are needed.
Our findings suggest that MG3 and MCI may be the potential target for the effect of metformin on reducing BP.The MG3 target-associated gene GPD2 encoded the mitochondrial glycerol 3-phosphate dehydrogenase (mGPDH), which is further partly used as a glycerol backbone for lipid molecules.Recent studies have shown that mGPDH was associated with kidney disease and macrophage inflammatory activities [21,22].We hypothesized that metformin might reduce BP through these mechanisms [23].Furthermore, we also found that the inhibition of expression of an MCI-related gene could be associated with the effect of metformin on reducing BP.Intriguingly, the previous report has also confirmed that the inhibition of MCI would diminish hypertension in mice model [24].
The current work still has limitations.First, though the 5 drug targets may recapitulate the major effect of metformin use, the possibility that some undiscovered targets would greatly affect the metformin response cannot be totally precluded.Second, future studies focusing on the validation of the included DTGs are still needed since the present DTGs derive from the database and systemic literature review and thus lack the direct evidences from biological tests [9].Third, our study analyzed only the data from European participants which may limit the generalizability of the conclusion in other populations.

Conclusion
In summary, the present study provided novel evidences to support the causal effect of metformin on reducing hypertension risk.We revealed that metformin might decrease BP through the inhibition of MG3 and MCI.These findings implied the repurposing of metformin for cardiovascular diseases.However, randomized trials with large samples are still warranted to provide more robust conclusions.
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Table 1
MR estimates of 5 targets-specific effects of metformin on SBP, DBP, and hypertension P value more than 5E − 8, see TableS9-S11).It indicates that the MVMRs were unfeasible here and those factors were hardly to influence our abovementioned conclusions.