Introduction

Human and mouse MATE1 were initially discovered in 2005 as mammalian orthologs of the bacterial MATE family conferring multidrug resistance (Otsuka et al. 2005). Subsequently, kidney-specific human MATE2K (Masuda et al. 2006) as well as MATE orthologs in rats and rabbits was identified (Terada et al. 2006; Ohta et al. 2006; Zhang et al. 2007) and intensely characterized. Knowledge regarding their tissue distribution, membrane localization and function has been summarized in several reviews (e.g., Moriyama et al. 2008; Damme et al. 2011; Nies et al. 2012; Motohashi and Inui 2013a).

MATE transporters mediate the efflux of organic cations across the luminal membrane of renal proximal tubule cells and the canalicular membrane of hepatocytes in exchange with protons and are therefore considered as the long-searched-for proton-coupled transporters of tubular epithelia (Otsuka et al. 2005). Transported substrates include endogenous compounds such as creatinine, the vitamin thiamine (vitamin B1) as well as several drug agents such as the frequently clinically used antidiabetic metformin and the antibiotics cephalexin and cephradine. An altered MATE function or expression may contribute to the interindividual variability of drug disposition with consequences for drug response. Therefore, in recent years, a number of pharmacokinetic and pharmacogenetic studies in healthy volunteers as well as patients have been conducted particularly to elucidate the impact of MATEs on interindividual variability of metformin response. Moreover, the potential role of MATE proteins in renal drug–drug interaction is of increasing interest (Hillgren et al. 2013).

Here, we summarize the current state of knowledge of molecular and functional characteristics of the human MATE transporters, with a particular focus on tissue-specific expression, regulation, as well as substrate and inhibitor specificities. Furthermore, we summarize currently available data on the genetic variants of MATEs and discuss their potential functional impact on pharmacokinetics and drug therapy.

Gene organization

Human MATE genes

The human genome contains sequences for two distinct MATE genes, i.e., SLC47A1 and SLC47A2, both being located in tandem on chromosome 17p11.2 (Otsuka et al. 2005) (Fig. 1a). The reference transcript with the NCBI accession number NM_018242 encodes MATE1, a functional protein of 570 amino acids (NP_060712) (Otsuka et al. 2005). Two transcript variants of SLC47A1, SLC47A1_∆exon15 and SLC47A1_∆exon15-16, have been detected in liver, kidney and other tissues (Fig. 1b and see paragraph Tissue distribution and localization), which are predicted to encode proteins of 466 and 511 amino acids, respectively (Fig. 1c). The expression of the respective proteins has so far not been demonstrated. Moreover, the presence of further SLC47A1 transcript variants has been postulated based on mRNA and EST alignments (ENSEMBL accession numbers: ENST00000395585, ENST00000436810, ENST00000571335, ENST00000575023), but again, corresponding proteins have not been identified yet.

Fig. 1
figure 1

Organization of the human SLC47A1 and SLC47A2 genes. a Human SLC47A1 and SLC47A2 genes are located in tandem on chromosome 17. Chromosome banding is from Genecards (http://www.genecards.org); the reference tracks are from NCBI (http://www.ncbi.nlm.nih.gov/). b While cloning SLC47A1, two novel mRNAs expressed in human liver and kidney are identified that lead to alternatively spliced SLC47A1 isoforms lacking exon 15 (predicted protein 466 amino acids) or lacking exon 15 and 16 (predicted protein 511 amino acids) (unpublished data). The gel picture shows DNA fragments after PCR of kidney or liver cDNA and of plasmids encoding SLC47A1 reference sequence (MATE1 plasmid), SLC47A1 isoform lacking exon 15 (MATE1_∆exon15 plasmid) and SLC47A1 isoform lacking exon 15 and 16 (MATE1_∆exon15-16 plasmid). The same primer pair was used for all PCR reactions. c Sequence alignments of the already described SLC47A1 and SLC47A2 isoforms together with the two newly described SLC47A1 isoforms. Sequence alignments were constructed with Clustal Omega (Sievers et al. 2011) and visualized using Jalview version 2.9.0b2 (http://www.jalview.org) (Waterhouse et al. 2009)

Full size image

For SLC47A2, four transcript variants are currently known, which give rise to proteins of different length and function (Fig. 1c): the originally identified MATE2 (isoform 1, 602 amino acids, NP_690872) (Otsuka et al. 2005), MATE2K lacking 36 amino acids due to alternative splicing of exon 7 (isoform 2, 566 amino acids, NP_001093116) (Masuda et al. 2006), MATE2B (219 amino acids, BAF37007) (Masuda et al. 2006) and MATE2 isoform 3 (580 amino acids, NP_001243592). MATE2 and MATE2K are both functional proteins, whereas MATE2B is not functional (Masuda et al. 2006; Tanihara et al. 2007; Asaka et al. 2007; Komatsu et al. 2011). No information is currently available for MATE2 isoform 3, whose existence has been inferred from sequencing of candidate full-ORF clones (Strausberg et al. 2002) but has not yet been experimentally proven.

MATE genes in other species

Orthologs of human MATE proteins have been found in many other species. In the current genome builds of the Ensemble project covering a total of 68 species (http://www.ensembl.org), 35 and 38 species have direct orthologs to human MATE1 and human MATE2, respectively. Figure 2a shows a phylogram of MATE proteins of species that are commonly used as preclinical models in drug development, i.e., mouse, rat, rabbit and the cynomolgus monkey.

Fig. 2
figure 2

Phylogram of MATE proteins of different species. a Sequence alignments were constructed with Jalview version 2.9.0b2 (http://www.jalview.org) (Waterhouse et al. 2009) using the BLOSUM62 algorithm. b NCBI accession numbers of the sequences used to generate the sequence alignments in a. Refseq: reference sequence. See http://www.ncbi.nlm.nih.gov/books/NBK21091/table/ch18.T.refseq_status_codes/?report=objectonly for description of the status codes

Full size image
Fig. 3
figure 3

Models of human MATE proteins. a Membrane topology was predicted using the web tool Topcons (http://topcons.net/) (Tsirigos et al. 2015). Lower red and higher blue lines indicate intracellular and extracellular regions of the proteins, respectively. Grey and white boxes indicate transmembrane helices. The Phyre2 server (Kelley et al. 2015) was used to model the three-dimensional structure of human MATE1. The highest scoring model for human MATE1 (c) was achieved when modeled on the bacterial MATE protein NorM from Vibrio cholerae (b). In c and d, genetic variants that have been identified in several ethnic populations and that have functional consequences are highlighted (color figure online)

Full size image

All MATE1 proteins are closely related and clustered in one group. Mouse Mate1 was originally cloned based on Genbank accession number AAH31436 (Otsuka et al. 2005; Hiasa et al. 2006) corresponding to cDNA clone BC031436. Mouse Mate1 was later designated as “mMate1a” because the novel variant mMate1b was identified as the true counterpart of human MATE1 (Kobara et al. 2008), which is now the validated reference sequence (NM_026183, Fig. 2b). At present, mMate1a is considered as nonexistent because the original cDNA clone apparently contained nucleotides, which are not present in the mouse C57BL/6J genome build 37 and resulted in frame-shift errors.

Of note, the so-called murine Mate2 proteins described by an independent research group are more closely related to human MATE1 than to human MATE2/2K. As previously suggested (Hiasa et al. 2007; Yonezawa and Inui 2011b), it would be reasonable to rename them, for example, as mouse Mate3 and rat Mate3. On the other hand, there are apparently no counterparts of human MATE2/2K in the mouse and the rat.

Protein characteristics

Post-translational modifications

Several post-translational modifications have been predicted or experimentally proven for MATE1 and MATE2. The online tool PhosphoSite (Hornbeck et al. 2015) predicts two phosphorylation sites for MATE1 (Thr17-P, Tyr299-P) and four for MATE2 (Ser544-P, Ser586-P, Thr588-P, Thr594-P); the latter ones have also been identified by a high-throughput phosphoproteomics approach (Raijmakers et al. 2010). N-terminal acetylation was experimentally shown for MATE1 (Van Damme et al. 2012). The functional consequences of these post-translational modifications are currently unknown. Neither MATE1 nor MATE2/2K is apparently glycosylated as predicted by UniProt (UniProt Consortium 2015).

Topology and structure

Because crystal three-dimensional structures of human MATE proteins are not available, their membrane topology can only be predicted by different computational methods. Commonly based on hydrophobicity of each amino acid, they calculate the probability for a stretch of amino acids being located in the membrane. These algorithms suggest that human MATE proteins have 13 transmembrane helices with an extracellular C-terminus (Zhang et al. 2007) (Fig. 3a). This has been experimentally proven initially for rabbit Mate1 (Zhang and Wright 2009) and subsequently also for MATE1 from human and mouse (Zhang et al. 2012). The first 12 transmembrane helices constitute the functional core, while the 13th transmembrane helix is apparently not necessary for function, but may be important for turnover of the protein (Zhang and Wright 2009; Zhang et al. 2012). An experimental verification of MATE2/2K topology is still lacking.

Despite the lack of three-dimensional structures of human MATE proteins, structures can be predicted based on the homology modeling using available bacterial X-ray structures. When using the PHYRE2 server (Kelley et al. 2015), which enables a Web-based protein structure prediction, the highest scoring homology model of human MATE1 was achieved when modeled on the bacterial MATE protein NorM from Vibrio cholerae (He et al. 2010) (Fig. 3b, c). Missense variants leading to reduced or abolished function of human MATE1 (see paragraph Genetic variants in human MATEs and their clinical significance for drug pharmacokinetics and drug response) that have been described in different ethnic populations are shown on the predicted three-dimensional model (Fig. 3c) and topology model (Fig. 3d).

Tissue distribution and localization

In their initial description of human MATE1 and MATE2, Otsuka et al. (2005) analyzed expression of both transporters in 8 tissues by Northern blot analysis and identified kidney, liver, skeletal muscle and kidney, respectively, as the major sites of expression. A subsequent systematic quantitative real-time PCR analysis of 21 human tissues showed that MATE1 is ubiquitously expressed with additionally high MATE1 mRNA expression in the adrenal gland and testis (Masuda et al. 2006). More recently, MATE1 transcripts were also detected in synovial fibroblasts (Schmidt-Lauber et al. 2012) and bladder urothelium (Bexten et al. 2015). A systematic analysis of 48 human normal tissues (Fig. 4a) as well as of 20 normal human tissues with their corresponding tumor tissues (Fig. 4b) using tissue microarrays confirmed high MATE1 expression in adrenal gland, kidney and liver and showed a widespread expression in a large variety of additional tissues and tumors such as cervix, endometrium, uterus, testis and thyroid gland. Moreover, the transcript variant SLC47A1_∆exon15 was also present in all investigated tissues although at about tenfold lower levels than the reference variant (Fig. 4c). The kidney is the major site of MATE2 (Otsuka et al. 2005; Komatsu et al. 2011) and MATE2K expression (Masuda et al. 2006), though MATE2K transcripts were detected in almost all other investigated 20 tissues at low abundance as well (Masuda et al. 2006).

Fig. 4
figure 4

Tissue distribution of SLC47A1. a Expression profiling of SLC47A1 in 48 normal human tissues was performed using quantitative RT-PCR (TaqMan technology) and a commercial array comprising cDNAs from 48 normal human tissues normalized to glyceraldehyde-3-phosphate dehydrogenase (TissueScan RT-PCR array, Origene Technologies, Rockville, MD) as described (Nies et al. 2009 and unpublished data). Expression levels were below the detection limit in the following 16 tissues and cell types: colon, duodenum, esophagus, lymphocytes, mammary gland, muscle, optic nerve, pancreas, placenta, rectum, seminal vesicles, tonsil, ureter, urinary bladder, uvula, vagina. b Expression profiling of SLC47A1 by TaqMan technology using a commercial array comprising cDNAs from 18 normal and corresponding tumor tissues normalized to β-actin (Origene Technologies) was performed as described (Schaeffeler et al. 2011 and unpublished data). c Comparison of the transcript levels of SLC47A1 and the newly identified SLC47A1_Δexon15 isoform in 18 normal human tissues by TaqMan technology using the commercial array described in b (unpublished data). 1 pituitary gland, 2 cervix, 3 retina, 4 thymus, 5 small intestine, 6 heart, 7 stomach, 8 skin, 9 lymph node

Full size image

Human MATE1 protein expression and localization have been studied in kidney, liver, placenta, adrenal gland, testis and prostate by different groups and techniques including immunoblotting, immunohistochemistry and quantitative proteomics (Otsuka et al. 2005; Masuda et al. 2006; Tanihara et al. 2007; Ha Choi et al. 2009; Kusuhara et al. 2011; Komatsu et al. 2011; Motohashi et al. 2013; Ahmadimoghaddam et al. 2013; Wang et al. 2015) (Fig. 5). In human kidney, MATE1 is localized together with MATE2 and MATE2K in the brush-border membrane of the proximal tubule epithelial cells (Otsuka et al. 2005; Masuda et al. 2006; Komatsu et al. 2011; Motohashi et al. 2013). All three MATE transporters are therefore considered to contribute to the renal tubular secretion of cationic drugs, which may enter the cells via organic cation transporter 2 (OCT2)-mediated uptake across the basolateral membrane (Fig. 5a). In human and murine hepatocytes, MATE1 is apparently localized on the canalicular (apical) membrane (Otsuka et al. 2005; Kusuhara et al. 2011). Recombinant human MATE1 is also localized in the apical membrane when expressed in vitro in polarized Madin–Darby canine kidney cells (Sato et al. 2008; König et al. 2011). Hepatic MATE1 protein levels have been recently quantified by a quantitative proteomics approach and varied about fivefold in a cohort of 55 individuals (Wang et al. 2015).

Fig. 5
figure 5

Localization of MATE1 in different human tissues. a Scheme of the localization of OCT2 and MATEs on the basolateral and luminal membranes, respectively, of proximal tubule epithelial cells. b, c Immunolocalization of MATE1 in the brush-border membrane of kidney, in cells of the adrenal gland, in Sertoli cells of the testis (arrows), and the glandular epithelial cells of the prostate (arrow) using a MATE1-specific rabbit antibody (HPA021987, Sigma-Aldrich) (unpublished data). OC+, organic cation

Full size image

Methods to study MATE function

Cell models

In proximal tubule epithelial cells, MATE proteins are located on the apical membrane facing the tubular lumen (Fig. 5a). They are physiologically functioning as efflux transporters moving substances out of the cells in exchange with protons, which are secreted into the tubular lumen by sodium/proton exchangers. Yet, MATE function can easily be studied in vitro using cell lines stably expressing a respective recombinant MATE protein. Commonly, cells are prepulsed with ammonia to acidify the cytosol so that MATE function can be measured as uptake of the compounds of interest. Detailed descriptions of this method and other designs to measure MATE-mediated transport have been described in detail (Otsuka et al. 2005; Tanihara et al. 2007; Hillgren et al. 2013). Moreover, primary renal cell lines and polarized tubule cell monolayers are increasingly used as models to study tubular excretion of drugs (Fisel et al. 2014).

Knockout mouse models

Knockout mouse models are powerful tools for assessing the role of transporters in the context of multiple transporters, metabolizing enzymes, plasma protein binding and blood flow (Degorter and Kim 2011). Because neither Mate2 nor Mate2K is expressed in mouse kidney (Hiasa et al. 2007) (Fig. 2a), but renal expression of mouse Mate1 is high, Mate1 knockout mice can be considered as a model to study MATE1 and MATE2/2K deficiency in humans (Tsuda et al. 2009a; Yonezawa and Inui 2011b). Mate1 knockout mice are viable and fertile without any overt phenotypical or histological alterations suggesting that other transporters may compensate for Mate1 function in the kidney (Tsuda et al. 2009a; Li et al. 2011). Mate1 knockout mice have been used to elucidate the pharmacokinetics and particularly renal elimination of MATE substrates such as endogenous compounds and xenobiotics, including the antidiabetic drug metformin, the anticancer agents cisplatin and flutamide, the antibacterial drug cephalexin and the herbicide paraquat (Tsuda et al. 2009a; Watanabe et al. 2010; Nakamura et al. 2010; Li et al. 2011; Toyama et al. 2012; Li et al. 2013a; Nakano et al. 2015). For example, treatment of Mate1 knockout mice with metformin resulted in significantly increased hepatic levels of metformin and led to lactic acidosis suggesting that homozygous variant or compound heterozygous carriers of MATE polymorphisms resulting in significantly reduced or even abolished transporter function may be at risk to develop metformin-induced lactic acidosis (Toyama et al. 2010, 2012). Despite the usefulness of the Mate1 knockout mouse model, species differences in substrate affinity between mouse and human MATE1 need to be considered (Hillgren et al. 2013).

Cynomolgus monkey

In a recent study, cynomolgus monkey was evaluated as a surrogate model for studying human organic cation transporters, including MATE1 and MATE2K. This animal model was suggested to have some utility for in vitro–in vivo extrapolations involving the inhibition of renal OCT2 and MATEs and may be a promising tool for the risk assessment of human drug–drug interactions (Shen et al. 2016).

Substrates and inhibitors

So far, more than 1000 compounds have been investigated whether they interact with human MATE1 and MATE2K. In contrast, similar comprehensive analyses have not been performed for MATE2, and the prototypical probe substrate tetraethylammonium (TEA) is currently the only known transported substrate (Komatsu et al. 2011). Several structure–activity relationship models have been proposed to predict whether a certain compound is a substrate, an inhibitor or both of MATE1 and MATE2K (Kido et al. 2011; Astorga et al. 2012; Wittwer et al. 2013; Morrissey et al. 2016). Moreover, recently a quantitative structure–pharmacokinetic relationship model was developed to predict renal clearance of MATE substrates (Dave and Morris 2015). In general, MATE substrates are cationic by nature or positively charged at physiological pH 7.4, hydrophilic, and have a low molecular weight. Examples include the endogenous substrate creatinine, the vitamin thiamine, the prototypical probe substrates 1-methyl-4-phenylpyridinium (MPP) and TEA, the antidiabetic drug metformin and the histamine H2 receptor antagonist cimetidine as well as the herbicide paraquat (Fig. 6). The substrate specificity of MATE is similar to OCT2 (encoded by SLC22A2), localized on the basolateral membrane of kidney proximal tubule cells (Fig. 5a), thereby supporting the concept that OCT2 and MATE transporters work in concert in the renal elimination of endogenous compounds and xenobiotics (Otsuka et al. 2005; Masuda et al. 2006; Tanihara et al. 2007; Damme et al. 2011; Morrissey et al. 2013; Hillgren et al. 2013; Motohashi and Inui 2013a, b; Chu et al. 2016). A similar functional vectorial transport can be assumed for the antiviral drugs acyclovir and ganciclovir, which are taken up from blood into the proximal tubule epithelial cells via the organic anion transporter 1 (OAT1, encoded by SLC22A6) and then excreted into the tubular lumen via MATE1 (Takeda et al. 2002). A functional interplay between the organic anion transporter 3 (OAT3, encoded by SLC22A8), another basolateral membrane transporter of renal proximal tubule cells, and MATE1 leading to extensive renal secretion and insufficient drug exposure was probably the reason for failure of a novel oxazolidinone antibiotic in phase I clinical trials (Lai et al. 2010).

Fig. 6
figure 6

Structures of selected MATE substrates were downloaded from the publicly available ChEBI database (https://www.ebi.ac.uk/chebi/init.do)

Full size image

Zwitterionic compounds (e.g., cephalexin, cephradine, oxaliplatin) and anionic compounds (e.g., estrone sulfate) are also transported by MATEs indicating a broader substrate range than OCT2. While the transport capacity of several substrates for MATE1 and MATE2K is quite similar, some drug agents are preferentially transported by MATE1 (e.g., cephalexin, cephradine, fexofenadine) or by MATE2K (e.g., oxaliplatin and verapamil). Table 1 gives an overview of MATE1 and MATE2K substrates and their selectivities toward the MATE transporters.

Table 1 Overview of substrates for human MATE1 and MATE2K
Full size table

Inhibitors of MATE1 and MATE2K are generally characterized by a positive charge at pH 7.4, a high LogP value and a median molecular weight of 349 (range 284–558) (Astorga et al. 2012; Wittwer et al. 2013). Table 2 summarizes inhibitor selectivities for MATE1 and MATE2K. A higher selectivity of inhibitors for MATEs over OCT2, as is the case for the H2 receptor antagonist cimetidine and the antimalarial agent pyrimethamine (Yonezawa and Inui 2011b), may explain clinically relevant drug–drug interactions (see also paragraph Role of MATEs for toxicity and drug–drug interactions).

Table 2 Selected inhibitors of human MATE1 and MATE2K and their selectivity for either transporter
Full size table

Of interest, the recent development of 11C-labeled metformin as positron electron tomography tracer and its application in mice will enable the noninvasive testing of physiological MATE function and MATE-mediated drug–drug interactions in future clinical investigations (Hume et al. 2013; Shingaki et al. 2015; Jensen et al. 2016).

Regulation

Transcriptional regulation

The 5′ regions of the rat and human SLC47A1 genes contain, instead of a TATA box, two GC-rich regions. These are critical for basal transcriptional activity due to binding of Sp1 as a general transcription factor (Kajiwara et al. 2007). Additionally, the rate of human SLC47A1 transcription is also regulated by AP-1 and AP2-rep, both of which bind in the 5′ UTR close to the transcriptional start site (Ha Choi et al. 2009). Finally, Nkx-2.5, SREBP-1 and USF-1 were identified to bind in the 5′ gene region of the SLC47A1 gene; they may also function as possible regulators of transcription (Kim et al. 2013). Of note, genetic variants in either of these transcription factor binding sites result in altered promoter activity in vitro and partially also in altered MATE1 mRNA expression levels (see Paragraph Genetic variants in human MATEs and their clinical significance for drug pharmacokinetics and drug response).

The role of nuclear receptors, i.e., ligand-activated transcription factors, in regulation of human SLC47A1 and SLC47A2 is still poorly understood, and limited data are available for mice. Studies using hepatocyte nuclear factor 4α (Hnf4α) knockout mice showed that hepatic Mate1 expression depends on the presence of Hnf4α (Lu et al. 2010). Similar results were obtained for renal Mate1 expression (Martovetsky et al. 2013). Whether HNF4α is also important for human MATE1 expression is currently unknown. The nuclear factors aryl hydrocarbon receptor (Ahr), constitutive androstane receptor (Car), nuclear factor erythroid-2-related factor 2 (Nrf2), peroxisome proliferator-activated receptor alpha (Pparα) and pregnane X receptor (Pxr) are not involved in the hepatic regulation of mouse Mate1 (Lickteig et al. 2008).

The observation that the pro-inflammatory cytokines TNFa, IL-1b and IL-6 may decrease MATE1 mRNA and protein expression in human rheumatoid arthritis synovial fibroblasts suggests additional and as yet unexplored signaling pathways of MATE1 regulation (Schmidt-Lauber et al. 2012).

Moreover, MATE proteins may also be post-transcriptionally regulated as recently suggested by in vitro studies using mouse Mate1, whose transport activity was negatively regulated by ischemia/reperfusion-inducible protein (Li et al. 2013b).

Effects of gender

The effect of gender on Mate1 expression was systematically studied in mice, rats and rabbits. In mice, Mate1 mRNA levels were significantly higher in female than male livers but, on the contrary, significantly lower in female kidneys than in males (Lickteig et al. 2008). This gender difference is apparently not caused by different estrogen levels (Meetam et al. 2009). No gender differences were observed in rats (Nishihara et al. 2007; Ma et al. 2015) and rabbits (Zhang et al. 2007) for renal Mate1 expression. Of interest, in a preliminary analysis of SLC47A1 and SLC47A2 mRNA levels using the publicly available TCGA data set (http://cancergenome.nih.gov/; (Cancer Genome Atlas Research Network 2013), we identified no gender differences in non-tumor human kidney (unpublished data).

Ontogeny

Although the critical importance of membrane transporters in pharmacotherapy of adults has been recognized in recent years, much less is known about the ontogeny of transporters from birth to adulthood (Mooij et al. 2015; Brouwer et al. 2015; Elmorsi et al. 2015). While availability of human data is limited, studies have been performed in mice and rats where increasing levels of renal Mate1 mRNA expression from the fetus through the postnatal–juvenile period were observed, finally reaching adult levels (Sweeney et al. 2011; Ahmadimoghaddam et al. 2013). Summaries of rodent studies on hepatic and renal Mate1 expression are given in several recent reviews (Klaassen and Aleksunes 2010; Brouwer et al. 2015; Elmorsi et al. 2015). However, it is unclear whether these data are transferable and predictive for the human situation. Only one human study comprising only a very small set of samples showed increasing hepatic MATE1 mRNA expression levels from neonates to older children up to adults (Klaassen and Aleksunes 2010).

MATE expression under pathophysiological conditions

Given the important role of MATEs in the renal excretion of endogenous compounds and drugs, altered expression of MATEs under pathophysiological conditions may be clinically relevant. In rat models of chronic renal failure or acute kidney injury, induced by ischemia/reperfusion, Mate1 protein levels are decreased in the proximal tubules (Nishihara et al. 2007; Matsuzaki et al. 2008). Because liver diseases may result in altered expression of renal transporters in different animal models (Ikemura et al. 2009), the effect of cholestasis, induced by bile duct ligation, on renal organic cation transporters was studied in rats. In contrast to increased levels of the uptake transporter Oct2, the expression of Mate1 protein was not affected by acute cholestasis (Kurata et al. 2010). The observed increased renal tubular secretion of cimetidine was, therefore, attributed to elevated levels of Oct2 rather than Mate1. In contrast, in a rat model of acute liver injury, induced by ischemia/reperfusion, renal Oct2 and Mate1 levels were both decreased resulting in decreased systemic and tubular secretory clearances of cimetidine (Ikemura et al. 2013). Moreover, renal Mate1 and Oct2 mRNA levels were decreased in a diabetic mouse model and a mouse model of nonalcoholic steatohepatitis leading to increased plasma half-life and decreased oral clearance of metformin (Clarke et al. 2015). These studies show that MATE expression may change under different pathological conditions with consequences on drug disposition. Clinical studies are necessary to investigate whether these changes also occur in humans.

Role of MATEs for toxicity and drug–drug interactions

Cisplatin nephrotoxicity

MATEs, together with OCT2, play a key role in the renal elimination of platinum drugs (Yonezawa and Inui 2011a). Because cisplatin is efficiently transported by OCT2 into the proximal tubule epithelial cells but not effluxed into urine by MATEs, it may accumulate within the cells increasing the risk of nephrotoxicity (Yokoo et al. 2007; Terada and Inui 2008; Nakamura et al. 2010; Fisel et al. 2014). This is apparently not the case for oxaliplatin because this is a substrate for MATEs (Table 1). However, only ~30 % of patients treated with cisplatin develop nephrotoxicity suggesting the involvement of additional transporters such as copper transporters (SLC31A1) in cisplatin disposition (Harrach and Ciarimboli 2015).

MATE-mediated drug–drug interactions

Because the H2 receptor antagonist cimetidine and the antimalarial agent pyrimethamine inhibit MATEs >tenfold more potent than OCT2, they may cause drug–drug interactions when co-administered with other MATE substrates such as metformin (Hillgren et al. 2013). Table 3 summarizes clinical studies investigating pharmacokinetic consequences on interacting drugs with MATE. In general, the renal clearance of the affected drug decreases, while drug exposure is increased. However, further clinical studies are warranted whether mostly moderate changes in plasma levels subsequently result in clinically relevant pharmacodynamic consequences.

Table 3 Clinical studies investigating potential MATE-mediated drug–drug interactions
Full size table

Genetic variants in human MATEs and their clinical significance for drug pharmacokinetics and drug response

Identification of genetic variants and their effects in vitro

Pharmacokinetic studies with the Mate1 knockout mouse model clearly show an important role of Mate1 for renal drug elimination. It is therefore obvious to elucidate the impact of genetic variants in human MATEs on drug response and/or the development of adverse drug reactions. Due to a variety of large-scale next-generation sequencing projects, such as the 1000 Genomes project (The 1000 Genomes Consortium et al. 2012) and the NHLBI-GO Exome Sequencing Project (http://evs.gs.washington.edu/EVS), the number of SLC47A1 and SLC47A2 genetic variants is increasing and in particular rare variants are discovered. For example, the Exome Aggregation Consortium (ExAC) strives to aggregate exome sequencing data from a variety of large-scale exome sequencing projects and currently covers data from >60,000 unrelated individuals (Exome Aggregation Consortium et al. 2015). The ExAC database lists 207 and 206 missense variants for MATE1 and MATE2K, respectively, most of them with minor allele frequencies <0.01 % (http://exac.broadinstitute.org). In general, the allele frequencies of missense variants are low and usually do not exceed 2 % in different ethnic populations. Several regulatory region and missense variants have been analyzed for their functional consequences in luciferase assays and transport assays in vitro, respectively (Table 4). The altered promoter activity of some regulatory variants could be explained by altered binding of transcription factors such as Sp1, AP-1, Nkx-2.5, SREBP-1 and USF-1, to the SLC47A1 gene promoter (Kajiwara et al. 2007; Ha Choi et al. 2009; Kim et al. 2013) and MZF-1 to the SLC47A2 gene promoter (Choi et al. 2011). Several missense variants showed a complete loss of function in vitro, i.e., MATE1-Gly64Asp, MATE1-Val480Met and MATE2K-Gly211Val, which was attributed to an abolished plasma membrane expression of the respective transporter (Kajiwara et al. 2009; Chen et al. 2009). By mapping the missense MATE1 variants on the three-dimensional structure and topology models of human MATE1 (Fig. 3c, d), it becomes evident that most of them are located within the transmembrane regions or in the last intracellular loop. These locations are apparently crucial for a proper MATE1 function. This has also been shown by site-directed mutagenesis studies, in which cysteine, histidine and glutamate residues in the transmembrane regions of human MATE1 and MATE2K have been identified being essential for substrate binding and transport activity (Asaka et al. 2007; Matsumoto et al. 2008; Damme et al. 2011).

Table 4 Allele frequencies of SLC47A1 and SLC47A2 sequence variants in different ethnic populations and functional consequences
Full size table

Genotype–phenotype correlations and clinical consequences of genetic variants

Because some MATE1 and MATE2K variants altered metformin transport function in in vitro experiments (Table 4) and the lack of Mate1 changed metformin pharmacokinetics in the knockout mouse model (Tsuda et al. 2009a), several studies addressed the association of SLC47A1 and SLC47A2 genotypes with pharmacokinetic/pharmacodynamic parameters and treatment outcome of metformin (Table 5). In most studies, the variants had no effects on the pharmacokinetic parameters of metformin. However, the SLC47A1 regulatory region variant rs2252281 and the intronic variants rs2289669 and rs8065082 were repeatedly associated with reduced metformin response (Becker et al. 2009; Jablonski et al. 2010; Stocker et al. 2013; Tkac et al. 2013; He et al. 2015), while the common SLC47A2 promoter variant rs12943590 was associated with greater metformin response (Choi et al. 2011; Stocker et al. 2013) in some studies. Yet, in other studies, the effect of the SLC47A1 variant rs2289669 could not be replicated (Christensen et al. 2011; Klen et al. 2014), indicating that other metformin transporters (Becker et al. 2010; Stocker et al. 2013; Christensen et al. 2013) and/or non-genetic factors (Maruthur et al. 2014; Emami Riedmaier et al. 2015) need to be considered as well.

Table 5 Genotype–phenotype correlations of SLC47A1 and SLC47A2 sequence variants
Full size table

Notably, the SLC47A1 missense variant rs111653425 (MATE1-Ala465Val), which occurs at a low allele frequency of 1.8 % in Icelanders (Table 4) and whose in vitro functional studies are missing, has been significantly associated with increased serum creatinine levels.

Conclusions

Since their initial discovery in 2005 (Otsuka et al. 2005), there is an increasing interest to elucidate the functional role of MATE1 and MATE2K as important transport proteins for renal and hepatic organic cation excretion. Both MATE transporters are now considered as the long-searched-for proton-coupled transporters in the luminal membrane of proximal tubule epithelial cells. Intensive functional studies by several groups have revealed the partial overlapping substrate and inhibitor specificities of MATE1 and MATE2K as they are capable of transporting a wide range of organic cations including a number of clinically relevant drugs. Since several clinical studies have suggested that MATEs may be involved in clinically relevant drug–drug interactions, both transporters are recommended to be tested during the drug development process for clinically relevant drug–drug interactions (Hillgren et al. 2013). However, despite the convincing functional evidence of MATE transporters in in vitro and in knockout mouse experiments, the role of genetic variation for drug pharmacokinetics and for drug therapy in humans, particularly for response to the intensely studied antidiabetic agent metformin, appears to be limited and does not resemble major effects observed in Mate1 knockout mouse models. The reason for this discrepancy may be that other factors may contribute substantially to the expression and function of MATE proteins in human such as epigenetic regulation (e.g., DNA methylation, microRNAs; Ivanov et al. 2014; Fisel et al. 2016). Moreover, in addition to MATEs, other drug transporters may affect drug pharmacokinetics as well. Thus, comparable to most intensively studied membrane transporters from the ABC family (e.g., P-glycoprotein/ABCB1; Wolking et al. 2015), a more comprehensive view needs to be considered to fully understand the role of MATEs for drug therapy.