Disposition of metformin: Variability due to polymorphisms of organic cation transporters

Abstract

Considerable interindividual variability in clinical efficacy is recognized in the treatment of type 2 diabetes mellitus with the biguanide metformin. Metformin is a substrate of organic cation transporters, which play important roles in gastrointestinal absorption, renal and biliary elimination, and distribution to target sites of substrate drugs. This raises the question of whether genetic variations in these transporters affect efficacy and risk of adverse events associated with metformin use. In this review, the pharmacogenetics of metformin is discussed in the light of the most recent literature. Overall, results from healthy volunteers support the notion that metformin pharmacokinetics can be affected by polymorphisms in genes encoding organic cation transporters. When considering the glycemic response to metformin in patients, however, the likely multifactorial nature of metformin response masks the effects of transporter polymorphisms observed in some clinical studies.

Key words::

• Diabetes mellitus type 2
• genetic polymorphism
• individualized medicine
• metformin
• organic cation transport proteins
• pharmacogenetics
• SLC22A1 transporter
• SLC22A2 transporter
• SLC47A1 transporter
• SLC47A2 transporter

Key messages

• Considerable interindividual differences in pharmacokinetics and clinical efficacy of metformin are recognized.

• Experimental studies in transgenic mice suggest that organic cation transporters are important determinants of metformin disposition and pharmacologic action.

• Several non-synonymous SNPs in the human genes encoding organic cation transporters were identified and have been found to cause a loss-of-function phenotype of the affected transporter in vitro.

• Clinical studies suggest that polymorphisms in organic cation transporter genes, such as the p.270A>S variant in the organic cation transporter 2 (SLC22A2) gene, may affect renal clearance of metformin at least in homozygous carriers of the reduced-function allele.

• An association of defined polymorphisms in organic cation transporters with glycemic response to metformin in patients has not yet been clearly established.

Introduction

Metformin is the principal biguanide drug that is used worldwide as an antihyperglycemic agent in patients with type 2 diabetes mellitus. In addition to type 2 diabetes, metformin is considered a therapeutic option for other diseases associated with insulin resistance, such as polycystic ovary syndrome (PCOS) and gestational diabetes. Metformin is already being prescribed to patients with PCOS, although it has not been approved for this indication (1). Its use in gestational diabetes is the subject of current clinical studies (2).

A major action of metformin is suppression of hepatic glucose production. Although it has been used in the clinic since 1957, the direct molecular target of metformin remains unknown. However, there has been considerable progress in defining its pharmacological effects. Inhibition of hepatocyte glucose production by metformin is mediated by activation of the enzyme adenosine monophosphate-activated protein kinase (AMPK), a master sensor, integrator, and regulator of cellular and body energy homeostasis (3). AMPK activation is also implicated as a mechanism for the induction of skeletal muscle glucose uptake. One potential pathway by which metformin activates AMPK involves the upstream serine-threonine kinase 11 (STK11, also known as LKB1) (4). A recent report provides evidence that metformin is able to inhibit hepatic gluconeogenesis also in an LKB1-and AMPK-independent manner via a decrease in hepatic energy state (5).

In addition to its efficacy in lowering glucose levels, metformin has the clinical advantages of inducing mild weight reduction and only a minimal risk of hypoglycemia. The United Kingdom Prospective Diabetes Study in 1998 and subsequent studies and meta-analyses on the effects of metformin on outcomes confirm that metformin is one of the main therapeutic options in type 2 diabetes mellitus, particularly in overweight or obese patients, because it may prevent some macrovascular and microvascular complications and mortality (6–8). This favorable action profile has made metformin one of the most widely prescribed antidiabetic drugs worldwide. In the US market, metformin recently became one of the top ten prescribed products; more than 52 million prescriptions for its generic formulations were filled in 2009 (9).

Metformin has some side-effects, principally gastrointestinal, which occur in 20%–30% of patients and require discontinuation of the drug in less than 5% of patients (10). Very rarely, metformin causes lactic acidosis. The frequency of this side-effect has been estimated to be 4–5 cases per 100,000 patient years (11,12). In cases where metformin was implicated as the cause of lactic acidosis, metformin plasma levels of >5 μg/mL were generally found (13), which is ˜4–5-fold higher than the maximum therapeutic steady-state plasma concentration (14). This suggests that metformin toxicity is associated with its accumulation in the body.

The antidiabetic response to metformin varies considerably from patient to patient. Based on clinical trial experience, metformin reduces HbA_1C values by 0.8% to 3% (15). In patients using metformin monotherapy as their first-ever antihyperglycemic agent, less than two-thirds of patients achieve a desired fasting glucose level or the HbA_1C goal of <7% (16,17). In routine clinical practice, the non-response rate to metformin in terms of achieving an HbA_1C of <7% within 1 year may be upwards of 50% (18). This incomplete response rate coupled with the waning effectiveness of metformin over time that occurs with most oral antidiabetic drugs (known as secondary failure due to the loss in insulin secretory capacity) highlights the need for personalized interventions to maintain tight glycemic control.

Role of organic cation transporters in metformin pharmacokinetics

It has been postulated that the observed variability in metformin response and the occurrence of lactic acidosis may be partially explained by variability in the pharmacokinetic disposition of metformin (19). The bioavailability of metformin is not complete, and large interindividual differences in bioavailability after oral administration in the range of 20%–70% have been described (20–22). Because metformin is not metabolized, this broad range probably reflects differences in absorption rather than first-pass metabolism. Furthermore, an inverse relationship between an orally administered dose of metformin and its bioavailability was observed (20,22), suggesting the involvement of a carrier-mediated saturable absorption process (23,24). Metformin is rapidly distributed following absorption but does not bind to plasma proteins. Its volume of distribution is large and can exceed 250 L, with the intestines, kidneys, and liver being the major organs of distribution (25,26). The clearance of metformin is primarily dependent on the single pathway of renal elimination, as metformin does not undergo relevant hepatic metabolism or biliary excretion. Because the renal clearance of metformin is much higher than the glomerular filtration rate, active tubular secretion by the kidney is the principal mechanism of metformin elimination. Accordingly, renal metformin clearance correlates with excretory kidney function (i.e. creatinine clearance rate) (20,27). Data suggest that substantial variation in metformin renal clearance exists, and genetic factors were found to contribute highly, by more than 90%, to the interindividual variation (28,29). Taken together, the pharmacokinetic characteristics of metformin and the inherited differences in metformin renal elimination suggest that active transport processes are critical to the disposition of metformin.

Because metformin is positively charged at physiological pH, it is not expected to diffuse freely through cell membranes but to cross biological membranes via carrier proteins. Indeed, metformin is a substrate for organic cation transporters (OCTs) in both the kidney and the liver (Figure 1). In humans, OCT1 (gene name SLC22A1) is expressed in the basolateral membrane of hepatocytes and is the primary mediator of hepatic metformin uptake (26,30,31). In contrast, OCT2 (gene name SLC22A2) is primarily expressed in the kidney and mediates uptake of metformin into the proximal tubule cells (31,32). In addition, metformin has been identified as a substrate for the multidrug and toxin extrusion (MATE) antiporters (33). MATE1 (gene name SLC47A1) is strongly expressed in the liver (canalicular membrane), kidney (brush border membrane), and skeletal muscle (34). MATE2 (gene name SLC47A2) is predominantly expressed in the brush border membrane of renal proximal tubules (34). Thus, MATE transporters may contribute to renal excretion of metformin.

Figure 1. Localization and functional role of organic cation transporters in the pharmacokinetics of metformin. OCT = organic cation transporter; MATE = multidrug and toxin extrusion antiporter; PMAT = plasma membrane monoamine transporter. Transporters in bold-face indicate major routes of metformin disposition in humans.

The role of these transporters in the disposition of metformin was established in knock-out mouse models of Oct1 and Oct2. Oct1(−/−) and Oct2(−/−) mice are viable and display no obvious phenotypic abnormalities (35,36). However, Oct1(−/−) mice show dramatically reduced hepatic uptake of tetraethylammonium (TEA; a prototypical organic cation) and metformin (26,36,37), whereas renal excretion of metformin is virtually unchanged compared with wild-type mice (26). In mouse hepatocytes, deletion of Oct1 results in a reduction in the effects of metformin on AMPK phosphorylation and gluconeogenesis. Thus, OCT1/Oct1 seems to be an important determinant of metformin action. When mice are given metformin, the blood lactate concentration significantly increases in wild-type mice, whereas only a slight increase was observed in Oct1(−/−) mice. Thus, Oct1 is responsible for the hepatic uptake of metformin, and the liver seems to be the key organ responsible for lactic acidosis (38). In Oct1/2 double knock-out mice, renal secretion of TEA is abolished, and plasma levels of TEA are substantially increased (35). Considering the differences in renal OCT expression between mice (Oct1 and Oct2) and human (OCT2), a combined deficiency of Oct1 and Oct2 in mice is believed to better reflect the effect of OCT2 deficiency in humans. Thus, the latter study emphasizes the role of OCT2 in renal elimination of cationic drugs such as metformin. Targeted disruption of Mate1, which is expressed in renal proximal tubule cells where it mediates luminal secretion of organic cations into the urine, is expected to impair renal elimination of the cationic drug metformin. Indeed, after intravenous administration of metformin, a 4-fold increase in the AUC of metformin was observed in Mate1(−/−) mice compared to wild-type mice (39). The renal secretory clearance of metformin in Mate1(−/−) mice was only 14% of that in Mate1(+/+) mice. However, Mate1(−/−) mice develop a nephropathy with creatinine clearance declining to only the half of that in wild-type mice (39). Although this observation emphasizes the relevance of Mate1 to normal kidney function, the inherent kidney dysfunction observed in Mate1(−/−) mice obscures the role of Mate1 in the renal clearance of metformin.

Implication of the polymorphic expression of OCTs in metformin pharmacokinetics and pharmacodynamics

Organic cation transporter 1 (OCT1)

Given the importance of OCT1 in metformin uptake into the liver and thus metformin activity, several pharmacogenetic studies have focused on polymorphisms in the gene encoding OCT1, SLC22A1, as modifiers of the glycemic response. Human SLC22A1 is a highly polymorphic gene (Table I). Functional studies in cell-based models have established that among the naturally occurring protein variants, the p.61R>C, p.189S>L, p.220G>V, p.401G>S, p.420del, and p.465G>R mutants reduce OCT1 function and thus reduce or eliminate metformin uptake (37). The decrease in uptake by some variant proteins is probably due to their cytosolic retention and reduced expression on the plasma membrane, as was shown for the SLC22A1 p.465G>R and p.61R>C variant proteins in transfected cells (37,40). Recently, Nies et al. confirmed these experimental data in a clinical study by demonstrating that the presence of the SLC22A1 c.262T>C (p.61R>C) SNP in patients strongly correlates with decreased liver OCT1 protein expression (41). Consistent with the previously described knock-out studies, phosphorylation of AMPK following metformin administration is reduced in cells expressing the non-functional or reduced-function variants, compared with cells expressing wild-type OCT1. Based on this experimental observation, it was proposed that OCT1 mediates the first step in the response pathway to metformin and that genetic variation in SLC22A1 may modulate the response to metformin in humans.

Table I. Polymorphisms investigated in clinical studies, their impact on the in-vitro transport function, and their allele frequency in different ethnic populations.

Nucleotide change	Amino acid change	Activity (37,47,62)	European (ethnic subpopulation, number of tested individuals) (40,41)	African (ethnic subpopulation, number of tested individuals) (40)	Asian (ethnic subpopulation, number of tested individuals) (40,47,63)
SLC22A1 (OCT1)
c.262T>C	p.61R>C	⇓	0.10 (EC 150)	0 (AA 200)	0 (AS 60)
			0.07 (EA 200)		0(JA 116)
c.289C>A	p.97Q>K	⇓			0.02 (JA/HC 59)
c.350C>T	p.117P>L	⇓			<0.02 (JA 66)
c.616C>T	p.206R>C	⇓			<0.01 (JA 66)
c.659G>T	p.220G>V	⇓	0 (EA 200)	<0.01 (AA 200)	0 (AS 60)
c.848C>T	p.283P>L	⇓
c.859C>G	p.287R>G	⇓
c.1201G>A	p.401G>S	⇓	0.01 (EC 150)	<0.01 (AA 200)	0 (AS 60)
			0.01 (EA 200)		0 (JA 116)
c.1222A>G	p.408M>V	⇔	0.60 (EA 200)	0.74 (AA 200)	0.76 (AS 60)
c.1256delATG	p.420del	⇓	0.17 (EC 150)	0.03 (AA 200)	0 (AS 60)
			0.19 (EA 200)		0 (JA 116)
c.1393G>A	p.465G>R	⇓	0.04 (EC 150)	0 (AA 200)	0 (AS 60)
			0.04 (EA 200)		0 (JA 116)
Nucleotide change	Amino acid change	Activity (50–52)	European (50,51)	African (50)	Asian (50,51,64,65)
SLC22A2 (OCT2)
c.495G>A	p.165M>I	⇓	0 (EA 200)	0.01 (AA 200)	0 (AS 60)
c.596C>T	p.199T>I	⇓	0 (EC 100)		0 (HC 100)
					<0.01 (KA 150)
					0 (VA 100)
c.602C>T	p.201T>M	⇓	0 (EC 100)		0 (HC 100)
					<0.01 (KA 150)
					0.02 (VA 100)
c.808G>T	p.270A>S	(⇓)	0.16 (EA 200)	0.11 (AA 200)	0.13 (HC 400)
					0.14 (HC 100)
					0.17 (JA 116)
					0.09 (AS 60)
c.1198C>T	p.400R>C	⇓	0 (EA 200)	0.02 (AA 200)	0 (JA 116)
					0 (AS 60)
c.1294A>C	p.432K>Q	⇑	0 (EA 200)	0.01 (AA 200)	0 (JA 116)
					0 (AS 60)
Nucleotide change	Amino acid change	Activity (54,55)	European (54,55)	African (54,55)	Asian (54,55)
SLC47A1 (MATE1)
g.−66T>C	−/−	(promoter) ⇓	0.32 (EA 68)	0.45 (AA 68)	0.23 (CA 68)
c.404T>C	p.159T>M	⇓	0 (EC 253)	0 (AA 95)	0.01 (JA 95)
				0 (TA 95)
c.1012G>A	p.338V>I	⇓	0 (EC 253)	0.05 (AA 95)	0.01 (JA 95)
				0.10 (TA 95)

AA = African American; AS = Asian American; CA = Chinese American; EA = European American; EC = European Caucasian; HC = Han Chinese; JA = Japanese Asian; KA = Korean Asian; TA = Tanzanian African; VA = Vietnamese Asian; ⇓= reduced transport function; ⇑ = increased transport function; ⇔ = no change.

Shu et al. first reported that individuals with any of the four reduced-function alleles (p.61R>C, p.401G>S, p.420del, or p.465G>R) have a significantly decreased glucose-lowering response to metformin compared with reference allele carriers (Table II summarizes results from clinical studies). To determine the glycemic response, study participants (healthy volunteers) underwent oral glucose tolerance testing before and after two doses of metformin. Individuals with the reference sequence demonstrated a 7% reduction in the area under the glucose concentration-time curve (glucose AUC), whereas the glucose-AUC in variant allele carriers increased from base-line by 8% (Figure 2). Shu et al. also investigated the effect of SLC22A1 variants on the pharmacokinetics of metformin in the same study cohort. Unlike prior animal studies, which demonstrated no difference in the pharmacokinetic profile in wild-type or Oct1(−/−) mice (26), pharmacokinetics differed across the SLC22A1 genotype groups in humans, with a significantly higher metformin AUC (+19%), higher maximal plasma concentration (C_max; +15%), and lower oral volume of distribution (V/F; –47%) in individuals carrying a reduced-function allele. These effects may be at least partially explained by a much lower hepatic uptake of metformin in individuals with a variant allele. Of note, renal metformin clearance was unchanged across SLC22A1 genotypes in that study, whereas in a subsequent study Tzvetkov et al. noted an increase in renal metformin clearance in healthy volunteers carrying a SLC22A1 reduced-function allele (42).

Figure 2. Association of SLC22A1 (OCT1) variants with response to metformin in healthy volunteers. Oral glucose (75 g) tolerance test was performed before (base-line) and after two doses of metformin. The glucose AUC was calculated from the time course of plasma glucose concentrations. Comparison of healthy individuals with only reference SLC22A1 alleles (n = 8) and those with at least one reduced-function allele in SLC22A1: p.61R>C, p.401G>S, p.420del, or p.465G>R (n = 12). Data are from (37).

Table II. Published pharmacogenetic studies investigating associations between polymorphisms in cation transporter genes and transporter expression, metformin pharmacokinetics, or glycemic effects of metformin therapy.

Study population	Outcome (versus reference)
Healthy volunteers: carriers of at least one reduced-function SLC22A1 allele (p.61R>C, p.401G>S, p.420del, or p.465G>R) (n=12)versusreference allele carriers (n=8)	Metformin kinetics (at least one reduced-function allele versus ref.): AUC ⇑ (+ 19%) Cmax ⇑ (+15%) CL/F ⇓ (−38%) V/F ⇓(−47%) CLR ⇔	(66)
Healthy male volunteers (n=103): association of SNPs in SLC22A1, SLC22A2, SLC22A4, and SLC47A1 with metformin renal clearance	presence of SLC22A1 reduced-function alleles was associated with increased renal metformin clearance	(42)
Healthy volunteers: SL22A2 p.270A>S heterozygous (n=5) and p.270A>S homozygous (n=4)versusreference allele carriers (n=6)	Metformin kinetics (808 TT versus ref.): AUC ⇔ Cmax ⇔ CLR ⇓ (−26%) In the presence of cimetidine, metformin CL was decreased in all participants, but the decrease was lower in TT than GG carriers	(67)
Healthy volunteers: SLC22A2 p.199T>I heterozygous (n=3), p.201T>M heterozygous (n=2), p.270A>S heterozygous (n=6), and p.270A>S homozygous (n=6)versus reference allele carriers (n=9)	Metformin kinetics (p.270 SS versus ref.): AUC ⇑ (+73%) Cmax ⇑ (+63%) CL/F ⇓ (−32%) CLR ⇓ (−51%) In heterozygous allele carriers, only minor (p.270A>S) or no (p.199T>I , p.201T>M) changes were seen	(53)
Diabetic patients: heterozygous carriers of variant SLC47A1 (p.64G>D, p.125L>F, p.328D>A) and SLC47A2 (p.211G>V) alleles (n=7) versus reference allele carriers (n=41)	Metformin kinetics: CL/F ⇔	(57)
Tissue samples from human livers; heterozygous carriers of the SLC22A1 p.61R>C allele (n=29) versus reference allele carriers (n=118)	SLC22A1 mRNA expression ⇓	(41)
Human kidney tissue samples from donors with at least one SLC47A1 g.−66T>C variant allele (n=26) versus samples from donors with the reference genotype (n=12)	SLC47A1 mRNA expression ⇓ (−;34%)	(54)
Healthy volunteers: carriers of at least one reduced-function SLC22A1 allele (p.61R>C, p.401G>S, p.420del, or p.465G>R) (n=12) versus reference allele carriers (n=8)	Oral glucose (75 g) tolerance test before and after two doses of metformin: Effect of metformin on glucose AUC ⇓	(37)
Women with PCOS: carriers of at least one reduced-function SLC22A1 allele (p.61R>C, p.401G>S, p.420del, or p.465G>R) (n=66) versus reference allele carrier (n=84)	Oral glucose (75 g) tolerance test: Effect of metformin on glucose AUC after 6 months of treatment ⇔ Effect of metformin on insulin AUC after 6 months of treatment ⇓ (≥2 variant alleles versus control)	(46)
Patients with diabetes mellitus (n=1,531): SLC22A1 p.61R>C and p.420del variant allele carriers versus reference allele carriers	Variants did not affect the initial HbA1C reduction, the chance of achieving a treatment target, the average HbA1C on monotherapy up to 42 months, or the hazard of monotherapy failure	(43)
Diabetic patients: responders to metformin (n=24) versus non-responders (n=9); correlation of SLC22A1 and SLC22A2 polymorphisms (identified by screening the respective genes) with response to metformin	c.−43T>G and c.1222A>G were negative and positive predictors, respectively, for the efficacy of metformin (OCT1 mRNA levels tended to be lower in human livers with the c.1222A>G variant allele)	(44)
Diabetic patients: association of SNPs in SLC47A1 with reduction in HbA1C level after initiation of metformin treatment in a population-based cohort (n=116)	Among genotypes, the intronic SNP rs2289669 (c.922-158G>A) was significantly associated with an increased glucose-lowering effect	(56)
Diabetic patients: association of SNPs in SLC22A1 with reduction in HbA1C level after initiation of metformin treatment in a population-based cohort (n=102)	Among genotypes, the SNP rs622342 was significantly associated with an increased glucose-lowering effect	(45)

Studies in diabetic patients were also conducted to study pharmacodynamics. Zhou et al. investigated whether the two most common reduced-function SNPs in SLC22A1, p.61R>C and p.420del, decrease the glycemic response in patients with type 2 diabetes (43). In that study, a series of drug response models for metformin were assessed, including short-and mid-term HbA_1C reduction, reaching the treatment target of HbA_1C of <7%, and time to monotherapy failure, in a large population-based study of 1,531 patients recruited to GoDARTS (Genetics of Diabetes Audit and Research Tayside). The study result was essentially negative, i.e. the SLC22A1 variants p.61R>C and p.420del did not affect the initial HbA_1C reduction, the chance of achieving a treatment target, the average HbA_1C on monotherapy up to 42 months, or the hazard of monotherapy failure (43). In another study, Shikata et al. analyzed variants of SLC22A1 and SLC22A2 in 33 Japanese diabetic patients with variable treatment efficacy to metformin (44). None of the identified polymorphisms in either gene was associated with metformin responder status, which was defined by an absolute reduction of >0.5% in HbA_1C within the first 3 months of metformin therapy. However, when clinical variables were included and multivariate statistics were applied, two SNPs in SLC22A1, c.-43T>G and c.1222A>G, were negative and positive predictors, respectively, for the metformin responder status. The biological basis of these associations remains somewhat elusive, as the c.1222A>G SNP is not associated with altered metformin transport in cellular models, and expression studies in human liver tissue samples did not reveal any allelic differences in OCT1 expression. Becker et al. analyzed associations between 11 tagging SNPs in SLC22A1 and changes in the HbA_1C level in a subcohort of 102 incident metformin users from the population-based Rotterdam study (45). No significant associations were observed except for the intronic rs622342 A>C SNP. For each minor C allele of this variant, the reduction in HbA_1C levels in diabetic patients was 0.28% less. However, the mechanism by which this SNP, which is not in linkage disequilibrium with known reduced-function SLC22A1 SNPs or associated with renal metformin clearance (42), affects the glycemic response to metformin remains unclear.

Similar to patients with type 2 diabetes, great variability in the clinical response to metformin has also been observed in women with PCOS. To test whether SNPs in SLC22A1 contribute to the variability in treatment response, Gambineri et al. conducted a prospective study in 150 patients of European descent with PCOS who were treated with metformin for 6 months (46). Carriers of at least one of the four SLC22A1 reduced-function alleles (p.61R>C, p.401G>S, p.420del, or p.465G>R) had a significantly decreased cholesterol-and triglyceride-lowering response to metformin compared with reference allele carriers. Glucose tolerance testing at base-line and at the end of the treatment revealed that the SLC22A1 genotype is a determinant of the insulin response to metformin, but not of the glucose AUC. Given that the lipid-lowering effect is one aspect of the hepatic action of metformin and that OCT1 plays a trigger role in hepatic metformin uptake, this observation is in agreement with the hypothesis that polymorphisms in SLC22A1 contribute to the variable response to metformin.

The genotypic frequencies of the non-synonymous, reduced-function polymorphisms in SLC22A1 vary substantially among different races or ethnicities. In general, they are rare in Asian populations and most frequent in Caucasians (Table I). Although this observation may be biased because genetic variants in SLC22A1 have been investigated and identified largely in European populations, a recent screen for genetic variants in SLC22A1 in Chinese and Japanese populations did not substantially change the conclusion (47). Given the important role of OCT1 in hepatic uptake and action of metformin, inter-ethnic differences in the frequency of SLC22A1 reduced-function SNPs may be associated with inter-ethnic differences in the pharmacodynamic profile of metformin (48). Significant heterogeneity in metformin efficacy by ethnic group, however, was not observed (49).

Organic cation transporter 2 (OCT2)

Given the importance of OCT2 in metformin pharmacokinetic disposition, some pharmacogenetic studies have focused on polymorphisms in the gene encoding OCT2, SLC22A2, as a modifier of metformin renal clearance. To date, almost 500 variable sites in SLC22A2 have been identified. Thirteen cause non-synonymous amino acid changes, and most are present at frequencies of less than 1%. Some variants such as p.165M>I, p.199T>I, p.201T>M, and p.400R>C lead to clearly reduced activity compared to the OCT2 reference (50,51), whereas the p.270A>S (c.808G>T) variant has more subtle effects on transporter function in vitro (52). Most pharmacogenetic studies performed in healthy volunteers have focused on the last-mentioned SNP because it is the only common coding polymorphism in SLC22A2, with an allele frequency of about 15% regardless of the ethnic background. Overall, these studies suggest that the SLC22A2 c.808G>T SNP has no significant impact on renal metformin clearance in heterozygous carriers (Figure 3). A marked reduction in metformin clearance by almost 40%, however, was observed in volunteers with the homozygous variant TT genotype. For this polymorphism, the results obtained so far (Figure 3) fit well with a recessive model of genotype–phenotype interaction, i.e. changes in renal metformin clearance occur only when both c.808 alleles are dysfunctional. Thus, it is possible that patients with the SLC22A2 c.808 GG or GT genotype will require a different dosage of metformin to achieve optimal glucose control compared with homozygous TT allele carriers. However, data are lacking to show whether SLC22A2 c.808G>T genotype-dependent changes in metformin pharmacokinetics (demonstrated only in healthy volunteers so far) translate into changes in metformin glycemic response in patients.

Figure 3. Effects of polymorphisms in genes encoding the organic cation transporter 2 (OCT2, SLC22A2) on renal clearance (CL_R) of metformin in healthy volunteers. Synopsis/meta-analysis (analysis method: fixed effect inverse variance model) of published studies (42,53,67,68). Modified from (69).

Song et al. identified few heterozygous carriers of the rare SLC22A2 c.596C>T (p.199T>I) and c.602C>T (p.201T>M) variants (which exhibit reduced transport of metformin in cellular assays) and compared metformin clearance in these individuals with that in wild-type allele carriers (53). These studies suggest that the presence of one reduced-function SLC22A2 c.596C>T or c.602C>T allele leads to a reduction in renal metformin clearance comparable to that observed in homozygous SLC22A2 c.808 TT carriers. The small number of tested individuals, however, precludes definite conclusions. Because these SNPs occur only in some ethnic subpopulations and at a frequency of ≤1%, their clinical impact in view of a population-based pharmacogenetic screening approach is rather limited.

Multidrug and toxin extrusion antiporters (MATE1, MATE2)

As described above, the MATE1 and MATE2 transporters, which are located in the canalicular membrane of hepatocytes and in the brush border of the renal epithelium, are responsible for the final step of the excretion of cationic compounds into bile and urine, respectively. Metformin is a substrate of these transporters, and therefore these transporters may be involved in metformin excretion. Several polymorphisms have been identified and characterized in the gene encoding MATE1, SLC47A1. SLC47A1 mRNA expression, for example, is significantly lower in human kidney samples from individuals who are homozygous or heterozygous for the SLC47A1 g.−66T>C SNP in the basal promoter in comparison with samples from homozygous reference allele carriers (54). In-vitro experiments suggest that the reduced transcriptional activity of SLC47A1 g.−66T>C results from a reduction in the binding potency of the transcriptional activator, activating protein-1, and an enhanced binding potency of the repressor, activating protein-2 repressor, to the mutant basal promoter region (54). In cellular models, two non-synonymous SNPs in SLC47A1, p.159T>M (c.404T>C) and p.338V>A (c.1012G>A), cause a significant loss in transporter activity for metformin (55). Unfortunately, no studies have been reported that examine the effect of the experimentally characterized promoter or reduced-function variants on the pharmacokinetics and pharmacodynamics of metformin in patients.

Based on data from the population-based Rotterdam cohort study, a significant association between the tagging rs2289669 G>A SNP in SLC47A1 and metformin response in a study sample of 116 incident metformin users was observed (56). The SNP was associated with an increased glucose-lowering effect. For each minor A allele, the HbA_1C reduction was 0.3% larger (56). However, the small sample size, the limited clinical phenotype, and the lack of a replication study limit the informative value of this study. In another clinical study, Toyama et al. observed no differences in metformin disposition (evaluated by recording the plasma concentration–time profile after oral administration of metformin) between diabetic patients carrying a heterozygous variation in SLC47A1 or SLC47A2 and patients with the reference genotype (57). Similarly, no differences were observed in metformin pharmacokinetics between heterozygous Mate1(+/−) knock-out mice and Mate1(+/+) wild-type mice, whereas metformin kinetics were markedly affected in homozygous Mate1(−/−) knock-out mice. Thus, the authors hypothesized that heterozygous variants in SLC47A1 and SLC47A2 do not substantially contribute to the interindividual variation in metformin pharmacokinetics (57). The postulated recessive model of the SLC47A1/2 genotype–phenotype interaction seems to be plausible particularly because two MATE efflux transporters with substrate redundancies are expressed, such that one MATE transporter can compensate at least partially for an inherently reduced expression or function of another MATE transporter.

Conclusion

Several non-synonymous, promoter and deletion-type variants in the SLC22A1, SLC22A2, and SLC47A1 organic transporter genes exhibit reduced transporter activity of metformin in experimental studies. The best-studied SNP in humans with respect to the pharmacokinetics of metformin is the SLC22A2 p.270A>S SNP, which shows a recessive mode of genotype–phenotype interaction with an almost 40% reduction in renal metformin clearance in homozygous carriers of the minor allele. At the level of pharmacokinetics in healthy individuals, the results are less controversial; however, the results are inconsistent when focusing on the glycemic response to metformin in patients with type 2 diabetes. In clinical studies, metformin has a decreased effect on glucose tolerance in healthy individuals who carry reduced-function polymorphisms of SLC22A1. However, these findings were not confirmed in a retrospective study of 1,531 metformin-treated diabetic patients, which showed no differential effect in HbA_1C reduction when considering the two most frequent index variants at this locus.

Why have we so far failed to establish genotypes of organic cation transporter genes (SLC22A1, SLC22A2, SLC47A1, SLC47A2) as validated predictors of pharmacokinetics and clinical response to metformin in patients? Possibly, the effects of single genetic polymorphisms in transporter genes are too small against a noisy background caused by environmental factors, age, gender, and co-morbidities. In fact, in a study of young female patients with PCOS who presumably had few co-morbidities (i.e. with more tight control of non-genetic covariates), the SLC22A1 genotype was found to be a significant determinant of lipid and insulin responses to metformin (46).

Estimates of the contribution of genes and environment to the variation in renal metformin clearance suggest a strong genetic influence in Caucasian and Asian populations (genetic component r_GC>0.9) (28,29). However, caution is needed in interpretation of these data obtained from two small populations of young healthy subjects. Both studies clearly failed to reach satisfactory statistical power (58). Moreover, the value of heritability applies only to the population in which it was established. Assuming a given extent of genetic control, the heritability of renal clearance of metformin tends to be larger the more uniform the tested population. Although the exact value is not known, an r_GC value >0.9 certainly under-estimates the environmental component of renal metformin clearance in a general population, as indicated by calculations of Tzvetkov et al. Considering only the two non-genetic parameters kidney function and age, these parameters are explaining 42% and 9% of the variation in renal clearance of metformin, respectively (42). Three decades ago, Sirtori et al. and Tucker et al. observed that renal metformin clearance is highly correlated with creatinine clearance (20,27). Cholestasis was also identified as an important non-genetic factor that is associated with markedly reduced OCT mRNA and protein expression in the human liver (41), potentially affecting hepatic disposition of metformin.

Moreover, variations in genes other than those encoding OCTs and MATEs are likely to modulate the pharmacokinetics or response to metformin therapy. For example, some recent studies suggest a role for STK11, a molecular target gene of metformin, in the treatment response to metformin (59,60). Another candidate is SLC29A4, which encodes the plasma membrane monoamine transporter (PMAT). PMAT transports metformin, is expressed in human intestine, and may play a role in the intestinal absorption of metformin (61). Unlike warfarin in which 40% of the variation in treatment response is due to two genes, VCORCI and CYP2C9, the pharmacogenetics of metformin seems to not be one of the few cases in which one or a few genes determines a large proportion of the variation of the response to a drug. One reason for the ‘polygenic phenotype’ of metformin pharmacokinetics may be the functional redundancy of some human transporters. Metformin is a substrate of human MATE1 and MATE2 transporters, which are expressed at the renal brush border membrane. Another example is OCT1 and OCT3, which are localized at the basolateral hepatocyte membrane (41). Although OCT3 is expressed in the human liver at lower levels than OCT1, OCT3 transports metformin with a higher efficacy than OCT1 and thus may counterbalance impaired OCT1 function (41).

In conclusion, investigation of polymorphisms in organic cation transporters and attempts to link the variants with the pharmacokinetics and pharmacodynamics of metformin have provided valuable information about the role of these transporters in the disposition of metformin in humans. Knowing the biology of the polymorphisms in genes defining the disposition of metformin (a prototypic cationic drug) may bring us a step closer to clinical application in terms of individualized drug therapy.

Acknowledgements

I thank MF Fromm for critical reading of the manuscript.

Declaration of interest: The author states no conflict of interest and has received no payment in preparation of this manuscript.