Mycophenolic acid pharmacogenomics in kidney transplantation

Mycophenolic acid (MPA) is a potent antiproliferative drug prescribed to prevent acute rejection in kidney transplantation. MPA reversibly inhibits the enzymes involved in the synthesis of guanosine nucleotides, thus preventing DNA replication of immune cells. Consequently, the repression of both cell and humoral immunity induces renal allograft tolerance. MPA is an effective and safe immunosuppressive drug, but some patients show variability in drug concentration, acute rejection, graft dysfunction, or MPA-related adverse events. Although the pharmacogenomics of immunosuppressive drugs has been widely investigated, MPA has been explored to a lesser extent. This review of MPA pharmacogenomic studies, included pharmacokinetics, adverse events, and main clinical outcomes of MPA treatment in kidney transplantation. Associations of variants in genes encoding MPA metabolizing enzymes, transporters, and targets with drug efficacy and safety are described. Most pharmacogenetic studies have focused on small sample sizes and few simultaneously analyzed genetic variants. Some studies reported significant associations of pharmacokineticsand pharmacodynamics-related genes with MPA exposure, acute rejection, graft dysfunction, hematological events, and gastrointestinal complications. However, even large cohorts did not replicate the findings, possibly due to divergent study design, immunosuppressive scheme, follow-up time, and other factors. Finally, the heterogeneity of aspects between studies limit conclusions on pharmacogenetic biomarkers of MPA in kidney transplantation. Page 2 Genvigir et al. J Transl Genet Genom 2020;4:[Online First] I http://dx.doi.org/10.20517/jtgg.2020.37


INTRODUCTION
Mycophenolic acid (MPA) is a potent antiproliferative drug broadly prescribed to prevent acute rejection in kidney transplantation. MPA is a reversible inhibitor of inosine-5´-monophosphate dehydrogenase (IMPDH), an important enzyme involved in the de novo synthesis of guanosine nucleotides, which are essential for the proliferation of T and B cells [1,2] . Consequently, guanosine nucleotide depletion by MPA prevents DNA replication, leads to repression of both cell-and humoral-mediated immunity and induces tolerance to allograft in kidney transplantation [1,3,4] .
IMPDH activity results from the expression of two isoforms, IMPDH type I and type II, which are encoded by IMPDH1 and IMPDH2, respectively [2] . Both genes are constitutively expressed in most tissues, though activated T and B lymphocytes have a higher expression of IMPDH2 [1] . Also, IMPDH type II has a higher affinity for MPA compared to IMPDH type I, which makes it a selective and potent antiproliferative drug for T and B cells [1,3] [ Figure 1].
MPA is available either as an ester prodrug or as a sodium salt, which are equivalent in terms of therapeutic effect and MPA exposure [5] . The prodrug 2-morpholinoethyl ester, named mycophenolate mofetil (MMF), is converted to the active metabolite MPA by carboxylesterases 1 and 2 (CES1 and CES2) [6] .
MPA is considered a safe drug even though some adverse events may occur, such as gastrointestinal complications, myelotoxicity, susceptibility to infections and neoplasms [1,3] . Designed to reduce gastrointestinal adverse events, the enteric-coated mycophenolate sodium salt (EC-MPS) has a delayed release formulation that delivers mycophenolate in the small intestine [7] . Due to enteric coating, EC-MPS causes a slower absorption than MMF, and more variable time for MPA to reach maximal concentration [8] .
MPA is extensively converted (about 90%) to the inactive 7-O-glucuronide (MPAG), by UDPglucuronosyltransferases (UGTs), mainly in the liver but also in intestine and to a minor extent in the kidney [9] . This process is mediated in the liver and kidney by UGT1A9, whereas UGT1A8 plays a key role in the intestine with minor contribution from the other UGTs [9] [ Figure 1]. Since MPA and MPAG are bound to serum albumin, an accumulation of MPAG may compete with free MPA for albumin and lead to increased free MPA concentration in plasma [9,10] . UGT2B7, with minor contribution of UGT1A8, mediates the biotransformation of MPA to its acylglucuronide form (AcMPAG) [ Figure 1]. This metabolite can inhibit the activity of IMPDH and is therefore pharmacologically active [11] . The cytochrome P450 (CYP) family also plays a role in MPA biotransformation. The 6-O-desmethyl-MPA (DM-MPA) is a phase I metabolite of MPA produced in the liver by the activity of CYP3A4, CYP3A5, and to a lesser extent by CYP2C8 [12] .
MPAG and AcMPAG, but not MPA, are excreted in the bile. The incorporation of these metabolites into hepatocytes is mediated by organic anion transport polypeptides (OATPs) [13,14] , which are membrane influx transporters encoded by genes of the SLCO family. From the hepatocyte, MPAG and AcMPAG are excreted in the bile via ATP-binding cassette subfamily C member 2 (ABCC2), also named multidrug resistanceassociated protein 2 (MRP2) [15,16] . ABC transporters are ATP-dependent drug efflux pumps [17] . Another important member is ABCB1 (also known as P-glycoprotein or multidrug resistance protein 1-MDR1), which seems to play a role in MPA disposition [18] . MPAG excreted in the bile undergoes extensive enterohepatic recirculation, being hydrolyzed to MPA in the small intestine and reabsorbed [ Figure 1]. The enterohepatic recirculation contributes 10%-60% of the total MPA exposure and causes a secondary peak 6-12 h after oral MPA administration [1] . MPA is also eliminated in the urine as MPA in negligible amounts, and mostly as MPAG, possibly mediated by ABCC2 [19] .
The more common MPA-related adverse events are diarrhea and other gastrointestinal complications. MMF can also cause myelotoxicity, especially leukopenia and anemia. Susceptibility to viral infections, such as cytomegalovirus (CMV) and BK polyomavirus (BKV), also increases in patients taking an MMFcontaining triple therapy regimen. Patients with low tolerance to MPA may require dose reduction, temporary interruption, or permanent discontinuation of the MMF treatment [1] .
The contribution of pharmacogenomics in the response to immunosuppressive drugs has been widely investigated. Several clinical studies have reported the influence of gene polymorphisms on the efficacy and safety of MPA, suggesting their potential contribution to transplant patient management [1][2][3] .
This review explores the pharmacogenomic studies that investigated polymorphisms in genes involved in the pharmacokinetics and pharmacodynamics of MPA in kidney transplantation and the main clinical outcomes and adverse events.

GENES RELATED TO MPA PHARMACOKINETICS
The pharmacogenomic studies involving genes related to MPA-metabolizing enzymes and transporters are summarized in Tables 1 and 2, respectively.

CES1 and CES2
CES1 and CES2 play crucial roles in the hydrolysis of the MMF prodrug to MPA. CES1 is highly expressed in the liver, while showing extremely low levels in the intestine. On the other hand, CES2 is mainly expressed in the intestine, and also observed in liver tissue [20] .
Fujiyama et al. [21] evaluated the influence of CES2 variants in 5´UTR (-1548A>G, rs3890213) and intronic regions (4595A>G, rs2303218 and 8721C>T, rs2241409) on MPA pharmacokinetics. They found no association of these variants with MPA plasma concentration-time area under the curve, partial (AUC 0-6 ) and total (AUC 0-12 ) , maximum plasma concentration (Cmax) and time required to reach the peak (Tmax) in 80 Japanese kidney recipients at 28 days after transplantation. These variants were also not associated with allograft rejection (AR) or diarrhea [21] .

CYP3A4, CYP3A5 and CYP2C8
The CYP enzymes CYP3A4, CYP3A5 and possibly CYP2C8 can catalyze the generation of the phase 1 metabolite DM-MPA, evident in human liver samples [12] , suggesting that variants in these genes may be involved in the clinical outcomes of MPA treatment.
CYP2C8 rs11572076 (G>A) variant was associated with increased risk of anemia (OR: 3.2, 95%CI: 1.7-6.2, P < 0.001) in a study that evaluated 2,724 variants in 978 adult kidney or simultaneous kidney-pancreas recipients [28] . The authors also found no association between this variant and leukopenia, a finding later ABCC2 rs1885301  [27] in a case-control study with children and young adult kidney transplant recipients. On the other hand, CYP2C8 rs11572076GG genotype was associated with reduced risk of leukopenia (GG vs. GA; OR: 0.14, 95%CI: 0.03-0.59, P = 0.008) with no association with AR, diarrhea or anemia [29] . One limiting and common aspect to these studies is the low frequency of the CYP2C8 rs11572076 variant in North American and Canadian (3%) and French (0.5%) patients [28,29] . Moreover, there is a minor involvement of CYP2C8 enzyme in MPA metabolism, which makes it difficult to interpret the findings.
In Brazilian patients on EC-MPS and tacrolimus treatment, our group showed that the CYP2C8*3 (rs11572080 + rs10509681) variant was associated with higher estimated glomerular filtration rate (eGFR), but not with AR, delayed graft function (DGF), or presence of adverse events [24,30] . The frequency of the CYP2C8*3 variant was also low (9%) in these cohorts.
The human UGT1A family (1A1, 1A3-1A10) is encoded by a single locus on chromosome band 2q37. UGT1A1 and UGT1A9 are expressed in the liver and extrahepatic tissues, where UGT1A9 is highly expressed in the kidney and shows lower expression levels in colon, adrenal and bladder [32] . UGT1A7, UGT1A8 and UGT1A10 are absent in the liver but are expressed in other tissues, such as the intestine [32] . The human UGT2 family is located on chromosome 4q13. UGT2B7 is expressed in a broad range of human tissues including liver, small intestine, and kidney [32] .

UGT1A1 and UGT1A7
Given the known minimal contribution of UGT1A1, UGT1A7 and UGT1A10 to MPA metabolism, there are few pharmacogenetic studies with these isoforms.
Satoh et al. [23] investigated the influence of the UGT1A1*28 (rs3064744) variant in 30 Japanese adult patients. The authors expected to find an association between this UGT1A1 polymorphism and MPA toxicity, but they did not obtain any significant result.
Regarding the clinical outcomes, Woillard et al. [34] confirmed that MMF treatment combined with tacrolimus and sirolimus increased the risk of diarrhea when compared with cyclosporine. Moreover, only in cyclosporine-treated patients, did the authors demonstrate that UGT1A8*2 variant (c.518G allele) was associated with lower risk of MMF-related diarrhea. The supposed mechanism for this protection involved the decrease in intestinal exposure to AcMPAG [34] . It is known that local (not systemic) exposure to AcMPAG contributes to the toxicity to the intestinal mucosa [3] . However, in North American kidney transplant recipients up two weeks, UGT1A8 c.518GG (UGT1A8*2/*2) genotype carriers had higher severity of gastrointestinal disorders (abdominal pain, acid reflux, indigestion, diarrhea and constipation) compared to UGT1A8 c.518C allele carriers [41] . This finding was not confirmed in final adjusted analysis (P = 0.069). Woillard et al. [34] found that the results were not similar when they studied the relationship of UGT1A8*2 with diarrhea alone or together with abdominal pain, nausea/vomiting and anorexia, which constituted overly heterogeneous phenotypes.
The occurrence of infections, but not diarrhea or blood disorders, was associated with high dose (2 g/day) MMF treatment and the UGT1A8*3/*3 (c.830AA) genotype [42] . The authors suggested that this finding was due to increased MPA levels and immunosuppression, in accordance with the lower enzyme activity associated with this variant.
As described in Table 1, UGT1A8 c.518C>G and c.830G>A variants were not associated with AR, diarrhea, leukopenia or anemia in other investigations with kidney recipients [27,29,39,43] .

UGT1A9
UGT1A9 but not UGT1A8 is expressed in the liver at high levels and is considered the most important hepatic UGT enzyme involved in MPAG formation from MPA [9,32] .
The variants UGT1A9 rs6714486 (c.-275T>A) and rs17868320 (c.-2152C>T) lead to increased protein expression and activity in liver microsomes [44] . Accordingly, these polymorphisms were associated with lower MPA exposure and enterohepatic recirculation (AUC 6-12 used as marker) in kidney patients on MPA and tacrolimus treatment [39,45,46] . Kuypers et al. [47] also reported this finding but only among patients treated with 2 g MMF, and van Schaik et al. [39] reported that the association was dependent on tacrolimus treatment. On the contrary, in patients treated with cyclosporine or macrolides, no association with MPA or MPAG pharmacokinetics was found by others [14,40,48,49] .
In pediatric kidney patients, UGT1A9 c.-331C allele was associated with leukopenia [43] , possibly because of increased systemic MPA exposure [48] , which was speculated but not evaluated by the authors.
Pazik et al. [51,52] brought to light the important role of UGT1A9 in reducing exposure to dietary toxins and carcinogens. They showed that UGT1A9 c.98TC genotype was associated with decreased graft function (proteinuria and diminished eGFR), possibly due to reduced detoxifying potential of the UGT1A9 c.98T>C encoded enzyme.

UGT2B7
UGT2B7 is the key enzyme involved in AcMPAG formation. Djebli et al. [53] investigated the in vivo and in vitro effect of UGT2B7 rs7438135 (c.-900G>A, otherwise termed c.-840G>A or c.−842G>A) and rs7439366 (802C>T) variants, which were in complete reverse linkage disequilibrium. The authors found that AcMPAG production was higher in the presence of UGT2B7 c.-900A allele compared to c.-900GG genotype in human liver microsomes. Moreover, the same authors showed that in patients treated with sirolimus (n = 40), but not with calcineurin inhibitors (CNIs), UGT2B7 c.-900AA (or 802CC) genotype was associated with higher AcMPAG AUC 0-9 at month 1 and 3 after transplantation [53] . Likewise, no association with exposure to AcMPAG was found in other studies with adult patients treated with tacrolimus or cyclosporine [48,54] .
The effect of UGT2B7 polymorphisms on MPA exposure in kidney recipients also was investigated. UGT2B7 c.-900G>A and 802C>T variants had no effect on MPA pharmacokinetics in patients treated with or without corticosteroids, which may induce glucuronidation and result in low MPA exposure [36,39,49,53,55] . Baldelli et al. [48] found that the UGT2B7 802TT genotype was associated with higher MPA peak concentration without differences in AUC 0-12 MPA. Likewise, UGT2B7 -79G>A variant, which is linked to the 802C>T and shows a decline in the transcriptional activity of the reporter gene in Caco-2 and HepG2 cells, was not associated with MPA exposure [39,56] . On the contrary, MMF apparent oral clearance (CL/ F) was significantly higher in pediatric patients with UGT2B7 802T allele compared to UGT2B7 802CC genotype carriers, early after transplantation (60 days) [40] . It is important to consider here that differences in metabolism between adults and children may modify the effect of the genetic variants [27,57] .

ABCB1, ABCC2 and ABCG2
ABCB1, ABCC2 and ABCG2 (breast cancer resistance protein, BCRP) are expressed in tissues such as intestine, liver and kidney, consistent with their critical role in the absorption, distribution, and elimination of many drugs [58] .
Wang et al. [18] showed that MPA is a substrate for ABCB1. MPA glucuronide metabolites are substrates for ABCC2, which is considered the main transporter of MPAG from the liver to the biliary system, although this role was suggested also for ABCG2 [15] . In the kidneys, ABCC2 also seems to play an important role in MPAG excretion [19] .
In studies with kidney recipients, no association of these ABCB1 variants with exposure to MPA, and MPAG or AcMPAG were observed up to one year after transplantation [23,24,59] . Bouamar et al. [59] found that this result was independent of cyclosporine or tacrolimus treatment. It is known that co-administration of cyclosporine leads to lower MPA exposure, by inhibiting enterohepatic recycling of MPA, compared to coadministration of tacrolimus [3] . The inhibition of ABCC2 and OATPs by cyclosporine possibly contributes to this drug-drug interaction [18,60] . Moreover, cyclosporine is a substrate of ABCB1 [25] .
Satoh et al. [23] found that the ABCB1 c.3435T allele was associated with greater requirement of MMF dose reduction due to diarrhea in adult patients treated with tacrolimus. Nevertheless, the authors did not find differences in dose-adjusted MPA AUC in ABCB1 c.3435T allele carriers compared to c.3435CC genotype individuals. Thus, they speculated that the interaction of ABCB1 with other transporters and enzymes might be involved in MPA-induced gastrointestinal toxicity in their small sample size (n = 30) [23] . ABCB1 c.3435TT genotype may decrease protein expression or activity, leading to increased drug absorption across the intestine and its higher systemic and intracellular concentrations [61] .

ABCC2
The extensively studied variant of ABCC2 (on chromosome 10), the rs717620 (c.-24C>T), is located in the 5´-UTR. The effect of this polymorphism in several in vivo studies is contradictory, suggesting that it is highly tissue specific and dependent on regulatory factors, such as the epigenetics (miRNA expression) [58] . Naesens et al. [62] investigated the ABCC2 c.-24C>T variant, in linkage disequilibrium with rs3740066 (c.3972C>T), in adult kidney transplant recipients treated with MMF and tacrolimus for one year. At day 7, only non-carriers of the ABCC2 c.-24C>T variant and patients with mild liver dysfunction had significantly lower MPA exposure than those without liver disease. This difference was not observed in ABCC2 c.-24T allele carriers. The reasons were not clarified. In the same study, from day 42 post-transplantation, the ABCC2 c.-24T allele was associated with higher MPA C/D and AUC [62] . The authors suggested that the variant was associated with increased protein expression and/or activity and enhanced enterohepatic recirculation. In line with this, the ABCC2 c.-24TT genotype was also associated with higher MPA C/D (days 3-8) in 408 Chinese patients on cyclosporine or tacrolimus treatment, but this result was not confirmed after Bonferroni correction in the multiple comparisons analysis [38] .
van Schaik et al. [39] found that MPA AUC 0-12 was 17%-23% higher for ABCC2 c.-24C>T carriers, but this difference reached statistical significance only at 6 weeks post-transplantation. Conversely, the ABCC2 c.-24T allele was associated with a lower MPA AUC at month 1, in a study with 55 kidney transplant recipients on tacrolimus and sirolimus therapy [63] .
Božina et al. [64] did not find any association between the ABCC2 c.-24T variant and MPA pharmacokinetics in kidney recipient or donors, but the A allele of the ABCC2 rs2273697 (Val417Ile, c.1249G>A) in kidney donors was associated with a reduced peak (29%) and early (AUC 0-2 , 33%) exposure to MPA, suggesting an increased renal elimination. Moreover, the authors showed that the combination of cyclosporine and ABCC2 c.1249A allele reduced MPA AUC 0-2 by 49%, an additive effect of this immunosuppressant. Regarding clinical outcomes, Naesens et al. [62] found that ABCC2 c.-24T allele carriers had more frequent episodes of diarrhea that non-carriers in the first year after transplantation. The authors considered the limitation of the sample size but stated that an increased enterohepatic recirculation of MPAG associated with the ABCC2 c.-24C>T variant could lead to diarrhea. It is important to mention that the local exposure and not the systemic concentration of MPA is relevant for gastrointestinal adverse events. Moreover, AcMPAG, produced in gastrointestinal cells and also substrate of ABCC2, contributes to cell toxicity [3,15] .
Other studies failed to confirm associations between ABCC2 polymorphisms and clinical outcomes (AR, DGF, graft function) or MPA-related adverse events in pediatric or adult kidney recipien ts [

ABCG2
Brazilian, Chinese, and North American cohorts have been studied regarding two ABCG2 variants in pediatric and adult kidney transplant patients [ Table 2]. The ABCG2 c.421C>A (rs2231142) variant leads to the amino acid exchange Gln141Lys, whereas ABCG2 -20 + 11790G>A (rs4491984) is an intronic polymorphism.
The investigations pointed out that these variants did not influence MPA or MPAG exposure or clinical outcomes during treatment with different immunosuppressive therapies [24,27,30,37,38] .

SLCO1B1, SLCO1B3 and SLCO2B1
OATPs additionally contribute to drug disposition, with ABC proteins; however, OATPs do not depend directly on using cellular ATP [67,68] . OATP1B1 and OATP1B3 are mainly expressed in the liver, while OATP1A2 is expressed in the intestine, biliary cells of the liver, and distal nephron of the kidney [67] .
In studies with OATP-transfected human embryonic kidney (HEK) cells, MPAG uptake, but not MPA, was enhanced by OATP1B3 (SLCO1B3) and to a lesser extent by OATP1B1 (SLCO1B1). MPA or MPAG uptake was not influenced by OATP1A2 (SLCO1A2) [14] . In line with this, Michelon et al. [13] found that MPAG and AcMPAG, but not MPA, were substrates of OATP1B1. Likewise, SLCO1B1 variants have not been associated with AR, DGF, diarrhea, leukopenia or neutrophil count in the studies summarized in Table 2 [22,24,27,49,59,65] . Nevertheless, SLCO1B1 c.521C allele (SLCO1B1*5) was associated with reduced MPA-related adverse events, including leukopenia, anemia, thrombocytopenia, diarrhea, nausea, vomiting or infection in French patients on cyclosporine-free treatment [13] . Possibly because cyclosporine inhibits OATP1B1, the effect of the SLCO1B1 c.521C allele was not detectable in the therapy with this immunosuppressant [13,60] . Indeed, cyclosporine-based treatment and SLCO1B1 c.521C allele were independent factors related to risk reduction of adverse events (OR: 0.22, 95%CI: 0.08-0.56, P = 0.002 and OR: 0.38, 95%CI: 0.21-0.69, P = 0.001; respectively) [13] . It was also demonstrated in vitro the reduced uptake of MPAG and AcMPAG related to SLCO1B1 c.521C variant. The authors hypothesized that carriers of the SLCO1B1 c.521C allele had impaired hepatic uptake of MPAG and AcMPAG, which leads to less enterohepatic recycling and, consequently, reduced exposure to MPA and less adverse events [13] . However, this study did not evaluate MPA exposure, and other investigations did not find an association between the SLCO1B1 c.521T>C variant and MPA pharmacokinetics, as previously mentioned.
SLCO1B3 c.334T>G and c.699G>A are in linkage disequilibrium, and the SLCO1B3 c.334G-c.699A haplotype was associated with reduced MPAG uptake in HEK293 cells in vitro [14] . In a study with 70 patients on tacrolimus and sirolimus treatment, but not on cyclosporine, the SLCO1B3 c.334GG genotype was associated with lower MPA peak concentration and exposure (AUC 0-12 ) and higher MPAG/MPA ratio [14] . On the basis of in vitro and in vivo results, the authors suggested that reduced OATP1B3 activity would decrease hepatic uptake of MPAG, reducing reabsorption of MPA through enterohepatic cycling in SLCO1B3 c.334GG genotype carriers.
Other investigations did not find an association between SLCO1B3 c.334T>G (or c.699G>A) variant and exposure to MPA, MPAG, or AcMPAG during treatment with either cyclosporine, in line with the previous findings [14] , or tacrolimus [37,38,59] . On the contrary, the SLCO1B3 c.334GG (699AA) genotype was associated with higher MPA AUC 6-12 , considered a marker of MPA recirculation, in Japanese patients with tacrolimusbased therapy at day 28 after transplantation [65] .

GENES RELATED TO MPA PHARMACODYNAMICS
Polymorphisms in IMPDH1 or IMPDH2 that result in increased IMPDH activity are likely to enhance T and B cell proliferation and decrease the response to MPA in kidney transplantation [1,2] . As a result, less sensitivity to MPA requires an increase in the dose of MMF to avoid the risk of acute rejection, but this clinical approach exposes patients to higher blood concentrations of MPA and increases the likelihood of adverse events [1] .
The pharmacogenomics studies involving genes related to MPA pharmacodynamics are summarized in the Table 3.
The IMPDH1 rs2278293 and rs2278294 (intronic) and rs2228075 (synonymous) variants were investigated in kidney recipients from several populations treated with MMF [ Table 3].
Wang et al. [71] examined the contribution of 17 variants in IMPDH1, including rs2228075, rs2278293 and rs2278294, to acute rejection or toxicity in 191 adult kidney recipients from the US on MMF therapy. Other studies reported a lack of association of both IMPDH1 rs2278293 and rs2278294 variants with AR in adult patients from the Transgene Study [13] and the Collaborative Transplant Study (the largest cohort, over 1000 kidney recipients) on MMF treatment [73] . Similar results were also found in adult kidney recipients treated with EC-MPS from the multicenter Dominos study [29] . The IMPDH1 rs2278293 was also not associated with AR in a kidney transplantation study from Brazil [22] .
Shah et al. [73] reported the lack of an association of the variants rs2278293 and rs2278294 in IMPDH1 with MMF long-term dose tolerated and dose achieved in adult kidney recipients from the Collaborative Transplant Study. A short-term follow-up study also reported no direct association of both IMPDH1 rs2278294 and rs2278293 polymorphisms with subclinical AR in 82 Japanese adult kidney recipients. However, the interaction of the rs2278293 A allele with high MPA night-time exposure range (AUC > 60 µg.h/mL and C 0 ≥ 1.9 µg/mL) increased the risk of subclinical AR [74] .
Two studies investigated the missense variant IMPDH1 rs2228075 (Ala525Ala, c.1320G>A) and the risk for AR, though no significant association was found in MMF-treated adult kidney recipients from the US [71] and from the CAESAR Study [25] The Collaborative Transplant Study explored the variants IMPDH1 rs2278293 and rs2278294 in a large cohort of kidney recipients, and found no association with long-term graft function (one year) and graft survival (five years) [73] .
The influence of IMPDH1 variants on hematological and gastrointestinal adverse events related to MMF therapy was also explored in several studies. The IMPDH1 rs2278294 G>A was reported to be associated with increased risk of leukopenia (A allele: OR: 1.66, 95%CI: 1.11-2.48, P = 0.0139) in adult kidney recipients from the Apomygree and FDCC studies [72] . Moreover IMPDH1 rs2278294 G and rs2228075 G alleles were associated with delayed time to leukopenia in children and young adult patients from USA [27] . Although these intronic variants were predicted not to affect IMPDH1 function, the authors suggested that these alleles may affect the sensitivity of the enzyme to MPA.
Other studies reported lack of association of IMPDH1 rs2278293, rs2278294 or rs2228075 with leukopenia in adult kidney recipients on MMF therapy from the US [71] , Apomygre and FDCC studies [72] , and Transgene  Study [13] . Likewise both rs2278294 and rs2278293 polymorphisms were not associated with leukopenia or anemia in adult patients from the Dominos Study within six months of EC-MPS treatment [29] . Eight other variants in IMPDH1 were also identified by Wang et al. [71] , but no association with leukopenia was found.
Woillard et al. [29] also investigated the influence of IMPDH1 rs2278923 and rs2278924 polymorphisms on gastrointestinal adverse events, such as diarrhea, but no association was found in adult kidney recipients from Dominos Study on EC-MPS therapy. Likewise, no association was found between these variants and susceptibility to CMV and other infections in adult kidney recipients on MMF therapy from the Apomygre and FDCC studies [72] .
Pazik et al. [75] explored the influence of IMPDH1 rs2278923 and rs2278924 on time-dependent change in body mass index (BMI) of adult kidney recipients from Poland, and found an association of the rs2278294 G allele with slower BMI gain over five years post-transplantation.

IMPDH2
Wu et al. [70] identified 25 variants (24 novel) in IMPDH2, including the missense rs72624919 (Ser485Cys), in 288 healthy subjects. The functional analysis was carried out with Ser485Cys and Leu263Phe (rs121434586, c.787C>T), a missense variant previously described [76] . The IMPDH2 Phe263, but not the Cys485, reduced markedly protein level and enzymatic activity, and altered the structure and function of the enzyme. They also detected the intronic variant IMPDH2 rs72639214 (IVS1-93), which was associated with reduced IMPDH2 basal mRNA expression in lymphoblastoid cell lines, as well as the intronic variant IMPDH2 rs11706052 A>G (c.819 + 10T>C) also known as 3757T>C [70] .
The IMPDH2 rs11706052 (3757T>C) was found to be associated with increased IMPDH activity in PBMC and increased IMPDH plasma concentration, six and 12 h, respectively, after MMF oral intake by 101 adult kidney recipients [78] . In this line of evidence, Winnicki et al. [79] reported that IMPDH2 3757T>C reduced the antiproliferative effect of MPA on lymphocytes (50% inhibition) isolated from 20 healthy volunteers. The authors suggested that this variant is associated with a poor response to MPA therapy. However, in a large cohort from the Collaborative Transplant Study, the IMPDH2 rs11706052 had no impact on MMF dose tolerated (one year) or dose achieved (three years) [73] .
Further studies did not confirm the association of the IMPDH2 rs11706052 with AR in adult patients from different populations [29,71,78,80] , including large cohorts [72,73] , and pediatric patients [27] . Lack of association with AR was also described for IMPDH2 variants rs121434586 (Leu236Phe, c.787C>T) and rs4974081 (-3624A>G) in adult kidney recipients on MMF treatment from a US cohort [71] and the Apomygre and FDCC studies [72] .
Shah et al. [73] also reported the lack of an association of the IMPDH2 rs11706052 with long-term graft function (one year) and graft survival (five years) in a large cohort of kidney recipients on MMF therapy from the Collaborative Transplant Study.
Some studies also explored the influence of polymorphisms in IMPDH2 on adverse events related to MPA. Pazik et al. [80] described an association of the IMPDH2 rs11706052 (3757T>C) polymorphism with increased lymphocyte counts and reduced risk of lymphopenia (3757C allele: OR: 0.32, 95%CI: 0.11-0.90, P = 0.032), but not with neutropenia in adult kidney recipients from Poland. Other studies did not confirm the influence of the rs11706052 variant on leukopenia in adult kidney recipients on MMF [71,72] or EC-MPS treatment [29] . In the same way, a retrospective study reported no association of rs11706052 with leukopenia or time to leukopenia in children and young adults from the US [27] .
Other variants in IMPDH2, such as rs121434586 (Leu236Phe, c.787C>T) and rs4974081 (-3624A>G), were studied, but no association with leukopenia was found in adult patients [71,72] , children, and young adults [27] on MMF treatment. The influence of the IMPDH2 rs11706052 on anemia was also explored but no association was found in adult patients from the Dominos Study on EC-MPS therapy [29] .
Two studies explored MPA-related gastrointestinal adverse events and found no association of IMPDH2 rs11706052 with diarrhea in adult patients on MMF [80] or EC-MPS treatment [29] .
Gensburger et al. [72] found a lack of association between IMPDH2 rs11706052 and rs4974081 and the susceptibility to CMV and other infections in adult kidney recipients on MMF therapy from the Apomygre and FDCC studies. This was also reported for IMPDH2 rs11706052 and the incidence of serious infections in adult patients on MMF treatment from Poland [80] . Pazik et al. [75] also studied how the presence of IMPDH2 rs11706052 could change the BMI of adult kidney recipients on MMF treatment in a timedependent manner, but no association was found with BMI change over five years post-transplantation.

CLINICAL IMPLEMENTATION
Increasing knowledge in pharmacogenomics and its clinical implementations depend on several factors, including the robustness of the studies, sample size, and reproducibility of the results in different populations. The Pharmacogenomics Knowledge Base (PharmGKB; www.pharmgkb.org) aims to collect, encode and disseminate understanding of human genetic variations on drug responses [81,82] , and most of the existing pharmacogenetic information is compiled in this database [83] . For clinical annotations, PharmGKB curators determine "levels of evidence" score that is a measure of confidence in the variantdrug associations using well-defined criteria based on careful literature review. This score has four levels, from 1 to 4 (1A, 1B, 2A, 2B, 3 and 4), 1A being the highest scientific evidence [81,82] .
Although the data reviewed here highlight the importance of pharmacogenomics in the variability of the response to MPA, these clinical associations are not strong enough to be used for clinical translation [85] and more evidence is needed to clarify the level of contribution of pharmacogenomics in kidney transplant patients treated with MPA.
Moreover, it is important to mention that three regulatory agencies, Food and Drug Administration (FDA, US), Pharmaceuticals and Medical Devices Agency (PMDA, Japan) and Health Canada (Santé, Canada) (HCSC), recommend drug labeling for MPA as an "actionable PGx" (genetic testing not required) for patients with rare hereditary deficiency of hypoxanthine-guanine phosphoribosyl-transferase, such as Lesch-Nyhan and Kelley-Seegmiller syndrome, because it may cause an exacerbation of disease symptoms characterized by the overproduction and accumulation of uric acid [81,82,86] .

FINAL CONSIDERATIONS
The genotype-phenotype associations reviewed here showed that genetic influence on MPA treatment seems to be small, especially due to inconsistency between studies. However, many factors need to be considered.
The strategies applied to MPA dosing were different between studies, which included fixed dosing ("onedose-fits-all") and dosing to a therapeutic range (therapeutic drug monitoring-TDM). MPA AUC 0-12 has been recommended as the best marker for dose adjustment [87] . However, unlike other drugs such as tacrolimus, MPA trough level showed a poor correlation with AUC [87][88][89] . TDM use for MPA is drawing much attention, but it is still controversial [3] . It is known that TDM can contribute to correct the variability in MPA exposure [89,90] . Therefore, TDM of MPA can minimize the genetic influence on the efficacy and safety of therapy.
Moreover, the studies included in this review are heterogeneous in some aspects, such as population (sample size, ethnics, age, etc.), inherent characteristics of the clinical approach, immunosuppressive scheme and follow-up time, time after transplantation, definition of the analyzed events (for example, AR, DGF or adverse events), and analytical methods for measurement of the MPA exposure.
Together, these heterogeneities demonstrate the limitations of this review for pointing out the pharmacogenetic biomarkers useful for clinical applications of MPA in kidney transplantation.

CONCLUSION
Individualized treatment can contribute to improve efficacy and decrease the toxicity of immunosuppressive drugs. Here, the influence of genetic variants on MPA pharmacokinetics and pharmacodynamics in kidney transplant recipients was reviewed. The combination of multiple drugs, the different sample sizes, and the lack of association consistency between studies have been important challenges of MPA pharmacogenomics. Together they limit the conclusions and clinical applications of MPA pharmacogenomics in kidney transplantation. Currently, further pharmacogenomic studies are needed to elucidate the contribution of genetic background to the effectiveness and safety of MPA therapy.