We have developed and evaluated a cell-based, high-throughput assay to uncover new inhibitors of ABCG2. This assay builds upon our finding that D-luciferin, the substrate of fLuc, is a specific substrate of ABCG2, and uses BLI to screen for ABCG2 inhibitors (6
). The screening of 3,273 compounds identified 219 candidate ABCG2 inhibitors with at least a two-fold signal enhancement over controls, ~ 60% of which have been previously reported as ABCG2 inhibitors, including gefitinib, prazosin, and harmine. The ability to identify known ABC transporter inhibitors, both potent and weak, proved that the assay is sensitive and reliable. The results also demonstrate the ability of the assay to identify previously unknown ABCG2 inhibitors. Approximately 40% of the 219 potent and ~ 84% of the ~ 150 less potent compounds have never been reported previously as being inhibitors or substrates of an ABC transporter. The less potent compounds, in particular, may be difficult to identify with other methods.
The power of the BLI assay was further demonstrated by the confirmation of the majority (26/28) of the new ABCG2 inhibitors uncovered, indicating a low false-positive rate. Twenty eight candidate ABCG2 inhibitors with over five-fold signal enhancement were subjected to the MTX resensitization assay, and 16 of them were also tested with the BODIPY-prazosin dye uptake assay. All but two were confirmed by the MTX resensitization assay. The low false positive rate may be explained by a previous report, where a screen of more than 70,000 compounds by an assay using pure fLuc found no activator of the luciferase-coupled reaction that could enhance the luminescent signal (27
). The signal enhancement seen in our BLI assay is attributed only to the increased intracellular concentration of D-luciferin upon administration of putative ABCG2 inhibitors from our library. However, caution is still required because ABCG2 may not be the only membrane transporter that affects the permeability of D-luciferin. In addition, pore-forming agents or detergents that disrupt cell membranes may cause false positives because of the leakage of D-luciferin into cells. Among the two compounds identified in this screen that could not be confirmed by the MTX resensitization assay, calcimycin is a cation ionophore, and falls into this category.
The BLI-based assay is very sensitive with no false negatives obtained. While the results of the MTX resensitization assay were as expected, those of the BODIPY-prazosin dye uptake assay were intriguing. Only nine out of 16 compounds tested were confirmed by this fluorescence-based assay and seven (~ 44%) failed this assay altogether. All of the seven compounds were confirmed by the MTX resensitization assay (). Notably, MTX resistance is the hallmark of the ABCG2-inhibiting phenotype (28
), therefore, this discrepancy suggests a high false negative rate for the BODIPY-prazosin assay. To understand the discrepancy better, we analyzed the structures of the seven compounds that could not be confirmed by the BODIPY-prazosin assay. Metyrapone, acepromazine, piperacetazine and acetophenazine have an aromatic ketone functional group, which can act as an electron acceptor and deactivate the singlet state of BODIPY via
an intermolecular electron-transfer process (30
). Porphyrin in verteporin and benzopteridine in riboflavin can quench the fluorescence of BODIPY by way of photoinduced electron transfer (32
). It has been reported previously that the fluorescence of several dyes used to probe mitochondrial transmembrane potential can be quenched by some anticancer drugs, including adaphostin, MTX and amsacrine (34
). Accordingly, fluorescence-based assays must be cautiously applied. The implication of this finding is significant. Since fluorescence-based assays have seen the most use in discovering new ABCG2 inhibitors (13
), it is possible that many ABCG2 inhibitors that quench fluorescence have been missed. The BLI-based screening assay described here has the advantage of not being prone to such an artifact. That advantage was demonstrated by our search of the HDL for previously known ABCG2 inhibitors, which revealed that the BLI-based assay missed none of them.
The BLI-based assay is efficient due to the elimination of incubation and wash steps. Several hundred drugs can be screened in one day using the current system, with many thousands of drugs possible if the technique were automated. False negatives caused by cytotoxicity in extended incubation are not a concern.
Candidate ABCG2 inhibitors obtained from our screen of the HDL are categorized based on their therapeutic effects, and can be clustered into several classes, including drugs affecting the cardiovascular and central nervous systems, the gastrointestinal system, among others (Supplementary Table 3
). Previously, it has been reported that ABCG2 is expressed in brain, colon, small and large intestine, venous endothelium, and in capillaries, suggesting that the expression pattern indicates that ABCG2 plays a protective role in these tissues. However, further evidence is needed to support that hypothesis (4
). Our discovery that many drugs treating diseases associated with these tissues are ABCG2 inhibitors is consistent with the possible functions of ABCG2 indicated by its expression pattern. The involvement of ABCG2 in some of the processes represented by the classes of agents uncovered in our assay has been recognized previously. Many anti-HIV drugs inhibit ABCG2 (37
) and a bidirectional interaction between EGFR and ABCG2 has been reported (38
). The roles that ABCG2 may play in most of the other processes have not been studied extensively. Information obtained from this screen may shed light on new mechanisms of action of drugs designed for other purposes. Since most of the compounds in the HDL library have been approved by the FDA, certain candidates could be immediately used in conjunction with standard cancer chemotherapy to help combat the effects of ABC transporter expression, which often confound such treatment.