The ABCG2 multidrug transporter protein has several important roles in normal human tissues. In terms of localization, ABCG2 can be found mainly in the tissue interfaces (blood–brain barrier, placenta, liver, intestine), where this transporter performs the removal of numerous endogenous and exogenous harmful substances [1, 31]. Mutations or SNPs in the ABCG2 gene may influence the transport of these substrates, leading to pathological alterations in the human body. The reduced uric acid excretion by ABCG2 in the intestine leads to hyperuricemia and gout . ABCG2 also modulates the ADME-Tox properties of several drugs; thus an inadequate functioning of ABCG2 may dramatically change the side effects of these compounds .
ABCG2 has been reported to be present in several naturally occurring variants in the human populations. Polymorphisms in the ABCG2 gene are frequent in Japan and in other Eastern countries [32,33,34], where the most studied variants are the Q141K (MAF: 0.16–0.32), and Q126X (MAF: 0.001), resulting in reduced-function or non-functional protein expression. In Caucasian populations, the Q126X variant has not been found, while, in addition to the frequently occurring Q141K variant (MAF: about 0.1), the R236X (truncated, non-functional variant) and the M71V, Q141K, R147W, T153M, K360Δ, F373C, R383C, T421A, T434M, and S476P ABCG2 variants were also reported to occur in several individuals [20, 22, 35, 36].
The most common variant, Q141K, has been described to increase the risk of hyperuricemia and gout; moreover, it has a significant effect on earlier onset of gout and was found to be associated with family history of the disease . Recently it has been described that ABCG2 dysfunction is a strong independent risk factor for pediatric-onset hyperuricemia/gout .
It should be mentioned that the second most frequent variant of the ABCG2 is V12M. As we have previously reported, V12M does not alter the expression of the protein in the red blood cell membranes of the carriers . It has also been demonstrated that this variant does not affect the urate transport activity of the protein , while data from high-throughput studies [37, 38] and meta-analysis of several published results  indicated that this SNP may be in fact protective against gout.
In the present study we have performed a comprehensive, detailed study of naturally occurring ABCG2 variants. A total of nine previously identified ABCG2 variants were tested—the M71V, Q141K, R147W, T153M, K360Δ, F373C, R383C, T434M, and S476P variants were expressed and analyzed in several assay systems. We have devised a suitable (EGFP corrected) transient mammalian cell expression systems for the evaluation of the ABCG2 variants, allowing a relatively rapid and simple determination of both protein expression and function. We have also applied an Sf9 insect cell—baculovirus expression system, in which membrane protein expression is hardly affected by folding or trafficking problems, to directly study ABCG2 function. In addition, a transposon-based stable mammalian cell expression system, also correctable for EGFP expression, was applied to perform detailed studies on protein folding, trafficking, membrane localization, and transport activity. In these stable cell lines ABCG2 and EGFP are expressed from the same vector and, as the cDNAs are separated by an IRES sequence, EGFP expression does not influence the expression of ABCG2. By sorting out cells with low EGFP expression, we could generate stable HeLa cell lines with low inserted copies of ABCG2, closely corresponding to physiological ABCG2 expression. In addition to the present studies, these cell lines are useful for a detailed examination of the effects of various drugs, potentially modifying the expression, trafficking or function of the naturally occurring variants of ABCG2.
Based on the data presented in the results section, we conclude that the R147W and R383C ABCG2 protein variants are seriously damaged, practically not expressed in the mammalian cells. The high-resolution confocal microscopy results indicate that the small amounts of these ABCG2 variants do not reach the plasma membrane, are localized mainly in the ER, and thus have major folding problems. The M71V, Q141K, F373C, and T153M ABCG2 variants are expressed in variable amounts in the mammalian cells but are found only at low levels in the plasma membrane. As studied in detail, the F373C variant is mostly localized in the Golgi and other intracellular compartments, while we found no apparent accumulation of any of the ABCG2 variants in the lysosomes. Previous studies have shown that the expression of the frequent ABCG2-Q141K variant is altered at several levels—reduced mRNA maturation, protein folding, membrane trafficking, and increased degradation of this protein have been suggested [24,25,26,27]. For the M71V variant our group demonstrated a reduced folding and trafficking in mammalian cells , while the partially defective F373C and T153M variants have not been studied as yet at the cellular level.
Since immunostaining does not recognize a partially degraded protein in the lysosomes, we have explored the effects of the selective proteasome inhibitor MG132 on the cellular fate of the ABCG2 variants (see Fig. 4b). We found that in the case of the WT protein, proteasome inhibition results in an increase in the expression of the fully glycosylated protein, and the presence of a small amount of the non-glycosylated ABCG2, normally eliminated by the proteasomes. In contrast, in the case of the ABCG2-R147W and R383C mutant variants, there is practically no expression of the fully glycosylated protein, while the appearance of a major non-glycosylated band when the proteasome activity is inhibited indicates that the misfolded, non-glycosylated protein (retained in the ER) is eliminated by the proteasomes. The ABCG2-M71V mutant variant shows lower expression than the WT protein, and proteasome inhibition results in the appearance of a large amount of the non-glycosylated protein, indicating partial ER retention and proteasomal degradation of this variant. These experiments indicate that the WT ABCG2 goes through a cellular processing with only a minor portion retained in the ER and degraded by the proteasomes, while the variants variably mis-folded are mostly retained in the ER and rapidly eliminated by the proteasomes.
An important finding in the present work is that the K360Δ, T434M, and S476P ABCG2 variants are properly expressed and functional in the plasma membrane and thus, in spite of being missense variants, may not cause any pathological alterations. In the case of the ABCG2-T434M protein, the Hoechst dye extrusion experiments showed similar or lower transport capacity, while in the ATPase experiments, we measured a higher activity. These findings may indicate a slight functional and/or trafficking problem with this variant.
While this manuscript was under review, Toyoda et al.  reported an analysis of the variants R147W, T153M, K360Δ, F373C, T421A, T434M, S476P, S572R, D620N at the clinical level and performed a detailed cellular study using HEK cells, membrane vesicles, and a Xenopus oocyte expression system. The expression and localization of these variants, as reported in this paper, in most cases are in accordance with our current results and clearly establish the potential clinical role of the non-functional variants. Still, the approaches were somewhat different, e.g. our stable cell lines, expressing low levels of an untagged ABCG2, may be more informative regarding mild changes in expression and trafficking, as the GFP tag on the ABCG2 protein may alter these parameters (see in , Suppl. Figure 5). A potential result of these differences is that Toyoda et al.  concluded that the T434M and S476P variants (expressed with a GFP tag) did not have full activity in urate transport experiments. In contrast, we could not observe such differences in our transport and ATPase experiments with the untagged protein (using several well-established substrates of the ABCG2 protein, including Hoechst 33342, quercetin and prazosin).
In Fig. 5a, the positions of the studied amino acid variants are depicted in the recently published atomic level model of the human ABCG2 protein . The two non-expressing mutant variants (R147W and R383C), and two of the variants which showed lower expression (Q141K, F373C) are localized to the so-called connector region (see Fig. 5b). This region has been shown to be crucially important in the stabilization and the structural rearrangements during the ABCG2 transport cycle [26, 39], and an artificial mutation in the R383 position (R383A) was reported to damage the function and trafficking of this variant . All these data further emphasize the key role of this region within the ABCG2 protein and predict a damaging effect of mutations in this region. As shown by us previously , the M71V variant, with reduced expression and trafficking, is localized in the NBD, and may directly alter the folding and maturation of the protein.
An interesting ABCG2 variant is the K360Δ protein, in which a lysine is deleted in a flexible intracellular loop, which is actually not seen in any of the atomic level models [28, 41,42,43]. This lysine is close to other lysines (K357, K358), previously identified as ubiquitination sites of the ABCG2 protein [44, 45]. The high expression of this variant in some of our experiments may suggest that the ubiquitin-mediated degradation of this variant is reduced.
In summary, the studies presented in this paper indicate that two naturally occurring ABCG2 variants, R147W and R383C, are detrimental to both expression and function, and thus heterozygous and especially homozygous or compound heterozygous patients carrying these ABCG2 variants may be especially susceptible to gout formation and drug toxicity, while less likely to have drug resistant tumors (in addition they may show a Junior-blood group feature—see ). Moreover, any potential treatment promoting the trafficking of these variants to the plasma membrane will not correct the defective function of these variants.
Individuals carrying the M71V, Q141K, T153M, and F373C ABCG2 variants, especially in a compound heterozygous form, may have similar phenotypic and pathological features, which is less efficient trafficking and function. In the case of these individual variants, a therapy promoting trafficking to the plasma membrane may significantly improve ABCG2 function. The expression of T434M and S476P proteins are not damaged; however, they may show somewhat altered transport function. In contrast, the ABCG2-K360Δ is a fully functional variant, with a potential of increased plasma membrane expression.
While the present experiments help to clarify the cellular phenotype of the specific ABCG2 variants examined, the transient expression system (yielding parallel ABCG2 and EGFP expression) used in this study provides a relatively simple and rapid screening tool. This system can be used for assessing the cellular expression, thus it allows to estimate potential medical effects (e.g. drug sensitivity or susceptibility to gout) in individuals carrying these mutations. The results presented here may also help to elucidate the specific role of the affected regions in the folding and functioning of the ABCG2 protein.