The Challenge of Exploiting ABCG2 in the Clinic

Abstract

ABCG2, or breast cancer resistance protein (BCRP), is an ATP-binding cassette half transporter that has been shown to transport a wide range of substrates including chemotherapeutics, antivirals, antibiotics and flavonoids. Given its wide range of substrates, much work has been dedicated to developing ABCG2 as a clinical target. But where can we intervene clinically and how can we avoid the mistakes made in past clinical trials targeting P-glycoprotein? This review will summarize the normal tissue distribution, cancer tissue expression, substrates and inhibitors of ABCG2, and highlight the challenges presented in exploiting ABCG2 in the clinic. We discuss the possibility of inhibiting ABCG2, so as to increase oral bioavailability or increase drug penetration into sanctuary sites, especially the central nervous system; and at the other end of the spectrum, the possibility of improving ABCG2 function, in the case of gout caused by a single nucleotide polymphism. Together, these aspects of ABCG2/BCRP make the protein a target of continuing interest for oncologists, biologists, and pharmacologists.

INTRODUCTION

The ABCG2 transporter was described nearly simultaneously by 3 different groups who gave it 3 different names. First, ABCG2 was termed BCRP, or breast cancer resistance protein, by Doug Ross’s group who cloned the transporter from an MCF-7 breast cancer subline selected with doxorubicin in the presence of verapamil¹. Shortly thereafter, the same gene was reported by Mike Dean’s group, who discovered the transporter mining an expressed sequence tag database and named the protein ABCP since it was an ABC transporter highly expressed in the placenta²; and by our group, who named it MXR for mitoxantrone resistance protein, from a colon cancer subline selected in mitoxantrone³.

Since the discovery of ABCG2, the number of reports characterizing substrates, inhibitors, polymorphisms and expression in normal and cancerous tissues has increased exponentially, much as was the case for P-glycoprotein (Pgp, MDR1, ABCB1) and the multidrug resistance associated protein (MRP1, ABCC1) after their discoveries. In this review, we will focus on the clinical challenges and opportunities ABCG2 presents with regard to cancer and cancer treatment.

Gene Expression and Protein Structure

The ABCG2 gene, located on chromosome 4, is over 66kb and contains 16 exons and 15 introns. In cell line models, high levels of ABCG2 expression are often accompanied by rearrangements involving chromosome 4 or by gene amplification^{4, 5}. The ABCG2 promoter has been reported to contain estrogen⁶, hypoxia⁷ and progesterone response⁸ elements that have been shown to control gene expression and a peroxisome proliferator-activated receptor γ (PPARγ) response element upstream of the ABCG2 gene has also been identified⁹. The promoter is methylated in selected cell lines; treatment with demethylating agents will increase expression. In other cell types, increased promoter acetylation following exposure to deacetylase inhibitors will increase gene expression¹⁰. Cytokines and growth factors¹¹,¹² as well as microRNAs^{13, 14} have been shown to have variable effects on gene expression.

The ABCG2 protein is made up of 655 amino acids and runs as a 72 kDa protein in reducing conditions. It is has one N-terminal nucleotide binding domain (NBD) and 6 C-terminal transmembrane segments comprising one transmembrane domain (TMD); this is in a reverse configuration compared to other ABC transporters where the NBD is at the C-terminus and the TMD is at the N-terminus. ABCG2 is considered a half-transporter as most transporters have at least 2 NBDs and 2 TMDs. ABCG2 is found in the G family of transporters, which is made up of only half transporters. As a half transporter, ABCG2 must dimerize to form a functional transporter.

Substrates and Inhibitors of ABCG2

ABCG2, much like Pgp, has proven to be a promiscuous transporter in that multiple compounds of different chemical classes are numbered among its substrates. It is of course best known for its ability to transport chemotherapeutic agents, with mitoxantrone, topotecan and SN-38 (the active metabolite of irinotecan) being among the most studied substrates. However, ABCG2 has also been shown to confer resistance to organic anions, such as the glucuronide conjugate of SN-38¹⁵. In this regard, there is overlapping substrate specificity with both Pgp and the MRPs. Other substrates include flavopiridol¹⁶; camptothecins such as irinotecan (and its active metabolite SN-38)^{17, 18}, 9-aminocamptothecin¹⁹ and diflomotecan²⁰; indolocarbazoles including edotecarin²¹ and becatecarin²²; antifolates such as methotrexate and some of its polyglutamylated forms²³, GW1843 and raltitrexed²⁴ as well as others; photosensitizers such as 2-(1-hexyloxethyl)-2-devinylpyropheophorbide a (HPPH)²⁵, benzoporphyrin derivative monoacid ring A²⁵ and pyropheophorbide a methyl ester²⁶; and kinase inhibitors such as gefitinib²⁷, imatinib²⁸, nilotinib²⁹ and JNJ-7706621³⁰. A number of other substrates unrelated to cancer treatment have also been described including uric acid³¹, HMG-CoA reductase inhibitors³², antivirals³³, antibiotics^{34, 35}, carcinogens^{36, 37} and dihydropyridines³⁸ as well as D-luciferin³⁹. A selected list of substrates is provided in Table 1.

Table 1.

Select substrates of ABCG2

Mitoxantrone1

Camptothecins

Topotecan121

Irinotecan (and SN-38)122

9-aminocamptothecin19

Tyrosine Kinase Inhibitors

Imatinib28

Gefitinib27

Nilotinib29

Dasatinib123

Carcinogens

Aflatoxin B137

Benzo[a]pyrene sulfate124

2-amino-3-methylimidazo[4,5-f]quinoline (IQ)37

3-amino-1,4-dimethyl-5H-pyrido[4,3-b]indole (Trp-P-1)37

2-amino-1-methyl-6-phenylimidazo(4,5-b)pyridine (PhIP)36

Porphyrins and Photosensitizers

Pheophorbide a125

Pyropheophorbide a methyl ester26

Chlorin e626

Protoporphyrin IX25

Benzoporphyrin derivative monoacid ring A25

2-[1-Hexyloxyethyl]-2 Devinyl Pyropheophorbide-a (HPPH)25

Antibiotics

Cephalosporins126

Fluoroquinolones35

Antifolates23, 24

HMG-CoA reductase inhibitors32, 127

Thiosemicarbazones113, 128

Imidazoacridinones129

Naphthoquinones130

Pancratistatin113

Sulfasalazine131

Cimetidine132

Flavopiridol16

Uric Acid31

Glyburide133

Becatecarin22

Riboflavin134

The search for ABCG2 inhibitors began with the discovery that fumitremorgin C (FTC) could reverse resistance in the mitoxantrone-selected S1-M1-3.2 cell line from which ABCG2 was cloned⁴⁰. Some of the first ABCG2 inhibitors identified were also Pgp inhibitors, such as elacridar (GF120918)⁴¹, biricodar (VX-710)⁴², dofequidar (MS-209)⁴³ and tariquidar (XR-9576)⁴⁴. Other inhibitors include dihydropyridines⁴⁵, tyrosine kinase inhibitors⁴⁶, flavonoids^{47, 48}, rotenoids⁴⁹ and botryllamides⁵⁰. Other inhibitors of ABCG2 are provided in Table 2.

Table 2.

Select inhibitors of ABCG2

Diketopiperazines

Fumitremorgin C135

Ko14373

Tryprostatin A136

Immunosupressants

Cyclosporin A137

Tacrolimus138

Sirolimus138

Tyrosine Kinase Inhibitors

Gefitinib46

Erlotinib139

Imatinib140

Nilotinib29

Lapatinib141

Sunitinib142

Vandetanib143

P-glycoprotein Inhibitors

Elacridar41

Tariquidar44

Biricodar42

Dofequidar43

Flavonoids

Genistein144

Naringenin144

Kaempferol144

Chrysin47

6-prenylchrysin48

Antivirals145

Calcium Channel Blockers146

Botryllamides50

Novobiocin147

Curcumin148

Bisindolylmaleimides149

Indolocarbazoles149

Dimethoxyaurones150

Chalcone Derivatives151

Acridone Derivatives152

Nonprenylated Rotenoids49

Cannabinoids153

In the case of several molecularly targeted anticancer agents, it has been difficult to determine whether the drugs are substrates or inhibitors. Evidence for both types of interaction has been presented in several reports for imatinib, nilotinib, and gefitinib^{27–29, 51, 52}. It is clear that the drugs interact with ABCG2, and careful studies suggest a concentration dependence – a substrate at low concentrations and inhibitory properties at high concentrations. In patients, oral gefitinib has been shown to increase plasma levels of orally administered topotecan (discussed below), suggesting in vivo activity as an inhibitor.

Pharmacokinetic Effects of ABCG2 Single Nucleotide Polymorphisms

Several studies of single nucleotide polymorphisms (SNPs) in the ABCG2 gene have been reported⁵³. Among the over two dozen sequence variations reported for ABCG2, the C421A nucleotide change, resulting in a glutamine to lysine substitution in the translated protein, has received the most attention. This variant is found with high frequency in people of Chinese (35%) or Japanese (35%) descent, but less frequently in people of African-American (2–5%), Hispanic (10%), European (11–14%) or Middle Eastern (13%) descent⁵⁴. The Q141K change has been shown to result in decreased plasma membrane expression, decreased or reduced ATPase activity compared to wild-type ABCG2^55–57. The net effect is a reduction in transport of substrates. Whether this is primarily due to decreased plasma membrane expression or to impaired function of properly localized transporter has not been clearly delineated. In vitro models show varying degrees of impaired folding. Misfolding is detected by the endoplasmic reticulum associated degradation (ERAD) quality control system, resulting in ubiquitin-mediated proteasomal degradation⁵⁸. The clinical impact of this SNP has been demonstrated, and linked to altered pharmacokinetics of topotecan⁵⁹, 9-aminocamptothecin⁶⁰, diflomotecan⁶¹, sulfasalazine⁶², and statins such as atorvastatin and rosuvastatin⁶³, as well as to increased incidence of gefitinib-related diarrhea⁶⁴. Most striking, however, is the recent association of the Q141K SNPs in ABCG2 with gout, where impaired transport of uric acid leads to higher plasma concentrations and clinical manifestation of gout³¹.

Clinically Important Normal Tissue Expression

Soon after the discovery of ABCG2, several studies were published aimed at identifying the normal tissue expression profile of ABCG2. The location and expression of ABCG2 in normal tissues lends clues to its likely role in normal physiology. A schematic representation is provided in Figure 1. Normal tissue localization suggests a role in determining drug distribution and normal tissue expression in the blood-brain barrier, blood-testis barrier, and maternal-fetal barrier⁶⁵. Existence of these barriers has been confirmed in animal models lacking ABCG2. Other normal tissue roles, as suggested by expression in breast epithelium, have also been implicated⁶⁶. Discussion of ABCG2 expression in normal tissues with potential for clinical intervention is provided below.

Figure 1.

Schematic representation of ABCG2 expression in normal tissues. Arrows indicate direction of substrate transport. In the brain and testis, ABCG2 is expressed mainly in the blood vessel endothelium, forming part of the blood-brain and blood-testis barrier, respectively.

Gastrointestinal Tract

General studies examining expression of ABCG2 in normal human tissues revealed high levels of expression in the epithelium of the gut^{67, 68}. When Taipalensuu and colleagues examined tissue samples from various sites along the gastrointestinal tract, higher expression of ABCG2 compared to ABCB1 was found in the jejunum⁶⁹. ABCG2 gene expression was found to decrease along the length of the gastrointestinal tract, with highest levels found in the duodenum and lowest levels found in the rectum⁷⁰. Similarly, Hilgendorf et al also reported higher levels in the jejunum versus the colon⁷¹.

Animal models have repeatedly demonstrated the significance of ABCG2 in limiting oral drug availability. Jonker and colleagues demonstrated in Abcb1a/b-deficient mice that the oral bioavailability of topotecan was 9-fold higher in mice when oral topotecan was coadministered with the ABCG2 inhibitor elacridar⁷². Similar results were reported by Allen and colleagues who reported that coadministration of oral topotecan with the ABCG2 inhibitor Ko143 resulted in 4- to 6-fold higher plasma topotecan levels in Abcb1a/b-deficient mice compared to mice given oral topotecan alone⁷³. Pretreatment of Abcb1a/b-deficient mice with elacridar increased the AUC of the investigational analgesic GV196771 16.4-fold compared to wild-type mice⁷⁴. Studies in Abcg2-deficient mice reported higher plasma levels of orally administered compounds compared to wild-type mice including the aurora kinase inhibitor JNJ-7706621³⁰, erlotinib⁷⁵, sulfasalazine⁷⁶ and some dietary flavonoids⁷⁷. Subsequent clinical trials in humans have demonstrated increased oral bioavailability of topotecan when it is coadministered with the ABCG2 inhibitors elacridar^{78, 79} or gefitinib⁸⁰. Thus, ABCG2 plays a major role in controlling the oral bioavailability of substrate compounds. Clinically, this means that some fraction of interpatient variation in drug exposure for ABCG2 substrate drugs is likely to be accounted for by differences in ABCG2 expression and function. The Q141K polymorphism is likely to result in increased drug exposure for orally administered anticancer compounds that are also substrates for ABCG2.

While the physiologic role of ABCG2 is not regulating intestinal uptake of topotecan and erlotinib, the transporter may serve a protective role by transporting dietary carcinogens. A few examples of the dietary carcinogens that are subject to transport by ABCG2 are given in Table 1. One hour after oral administration of PhIP, plasma levels of PhIP were 3.1-fold higher in Abcg2^−/− mice compared to wild type mice, confirming the role of ABCG2 in limiting the uptake of the carcinogen³⁶. Similarly, ABCG2 was shown to limit the intestinal absorption of the carcinogens 2-amino-3-methylimidazo[4,5-f]quinoline (IQ), 3-amino-1,4-dimethyl-5H-pyrido[4,3-b]indole (Trp-P-1) and aflatoxin B1 in a mouse model³⁷.

Blood-Brain Barrier

Given the high levels of ABC transporter expression in brain microvessels, groups began looking for ABCG2 expression at the blood brain barrier soon after its discovery. ABCG2 was localized to the luminal surface of the endothelial cells of the human brain, much as described for Pgp^{68, 81}, suggesting that ABCG2 forms part of the blood-brain barrier along with other transporters such ABCC1 and ABCC2⁸². Aronica et al found ABCG2 to be localized to the luminal surface of blood vessels in the human brain, but did not find ABCG2 in glial cells or neuronal cells⁸³. In mice, Cisternino and colleagues reported that Abcg2 levels in brain microvessels were 700-fold higher than in the cortex. Interestingly, they also reported that Abcg2 expression was 3-fold higher in Abcb1-deficient mice, suggesting not only a compensatory role for Abcg2 in the brain, but also cooperativity between the two transporters.

These findings in tissue localization studies have been mirrored by functional studies in mouse models that show that not only is ABCG2 an important component of the blood-brain barrier, but more dramatically, there is marked cooperativity between Abcg2 and Abcb1 at the blood-brain barrier. In Abcg2-deficient mice, the brain penetration of imatinib 2 h after i.v. administration of imatinib (12.5 mg/kg) was 1.2-fold higher versus wild-type mice⁸⁴. This was not observed in a similar study where brain imatinib levels were examined 1 h after imatinib administration, suggesting that differences in brain levels mediated by transporters are time-sensitive⁸⁵. Area under the plasma and brain tissue concentration-time curve (AUC) levels of topotean in Abcb1a/b^−/−, Abcg2^−/− mice were 12-fold higher than wild-type mice, compared to mice deficient in Abcb1a/b or Abcg2 alone, suggesting that both transporters must be inhibited to achieve a maximal effect⁸⁶. Similar results were observed with lapatinib. While Polli et al found brain concentrations of lapatinib were similar in Abcg2^−/− mice and only 3- to 4-fold higher in Abcb1a/b^−/− mice compared to wild-type mice, brain levels were nearly 40-fold higher in Abcb1a/b^−/−, Abcg2^−/− mice compared to wild-type mice after IV administration⁸⁷. This was true at both high and low concentrations of IV lapatinib. Lagas et al reported that brain accumulation of sorafenib was 4.3-fold higher in Abcg2^−/− mice while it was 9.3-fold higher in Abcb1a/b^−/−, Abcg2^−/− mice compared to wild-type; administration of sorafenib to wild-type with elacridar, a dual Pgp and Abcg2 inhibitor, resulted in brain accumulation levels similar to those seen in mice deficient in both Abcb1a/b and Abcg2⁸⁸. This group also reported similar results for dasatinib⁸⁹, as did Chen et al⁹⁰. Using gefitinib as a PET tracer, Kawamura and colleagues demonstrated 8-fold higher radioactivity in the brains of Abcb1a/b−/−, Abcg2−/− mice compared to wild-type⁹¹. Clinically, the expression of ABCG2 at the blood-brain barrier likely limits the access of anticancer agents that are substrates. For example, the tyrosine kinase inhibitors gefitinib, erlotinib and lapatinib developed for lung or breast cancer likely do not gain significant access to the CNS. Since both of these tumor types have predilection for CNS metastases, a very important new line of investigation will be to identify mechanisms to increase CNS entry, to evaluate safety, and then to study efficacy. Based on the animal work, however, inhibition of ABCG2 alone will not be sufficient, and inhibitors of multiple transporters will be needed. An immediate need is to determine how closely the data in the knock-out mouse studies reflects the physiology of the human blood brain barrier.

Renal Excretion

In our earlier studies examining ABCG2 expression in normal tissues, we localized ABCG2 to the cortical tubules of the kidney, and postulated a role for ABCG2 in the renal excretion of compounds⁶⁸. Subsequent studies reported expression of ABCG2 at the brush border membrane of proximal tubules⁹². Mouse models confirmed the role of ABCG2 in the renal excretion of sulfate conjugates of E3040 and edaravone^{93, 94}.

One recently identified substrate points to an important physiologic role of ABCG2 in the kidney. Several genome-wide association studies have independently found SNPs in ABCG2 to be associated with higher uric acid levels and gout^95–97. Studies in Xenopus oocytes injected with ABCG2 mRNA confirmed ABCG2 was a transporter of uric acid, as uric acid transport rates in oocytes injected with ABCG2 encoding the Q141K SNP were over 50% lower than rates in oocytes expressing wild-type ABCG2³¹. SNPs in ABCG2 are postulated to account for approximately 10% of gout cases in white males³¹.

From an oncologic perspective, this observation raises the question of whether an increased incidence of tumor lysis syndrome might be observed in patients harboring the Q141K SNP. Indeed, it is striking to note the number of case reports in the literature of tumor lysis syndrome in the context of diseases or drugs not conventionally thought associated with this complication – sorafenib in hepatocellular carcinoma, imatinib in chronic myelogenous leukemia, flavopiridol in chronic lymphocytic leukemia^98–101. Given that these compounds are all substrates of ABCG2, and inhibitors at higher concentrations, it is tempting to speculate that these isolated cases occur in patients with the Q141K SNP.

ABCG2 Expression in Cancer and Drug Resistance

Well over 10 years after the initial reports of the ABCG2 gene, its role in clinical drug resistance is still murky. The majority of data concerning the contribution of ABCG2 to drug resistance in cancer has been gathered for leukemia and table 3 summarizes results of studies measuring ABCG2 expression in leukemia. While most reports that examined ABCG2 expression in acute lymphoblastic leukemia (ALL) reported relatively low levels, higher levels seemed to be found in acute myelogenous leukemia (AML), but even here there are conflicting reports. In AML, early studies seemed to suggest a link between expression of ABCG2 and clinical outcome, but as many reports have suggested that no link exists (see Table 3). More recently, groups have been looking at the purported “stem cell” population, and some studies have reported relatively high levels of expression of ABCG2 in stem cells or in the most primitive progenitor population^102–104. The presence of a leukemic stem cell is well accepted, but the existence of a hierarchical model including stem cells in solid tumors is more controversial¹⁰⁵. Some studies have identified stem cells based upon their ability to efflux Hoechst dye. These cells have increased levels of ABCG2 and are thus more drug resistant. However, selection of cells based upon a Hoechst dim phenotype is not sufficient for identification of the putative stem cell population because ABCG2 is also a differentiated phenotype expressed, as noted above, in multiple cell types. If a hierarchical model is confirmed for solid tumors, and a distinct long-lived cell population exists that contributes to repopulation of cancer after therapy, as the hypothesis suggests, then targeting ABCG2 or other protective mechanisms will be needed to eradicate this cell population.

Table 3.

Leukemia studies examining ABCG2 expression

Author	Cancer	Method	Samples	Conclusion
Ross et al 2000154	AML ALL	RT-PCR	20 AML samples 1 ALL sample	High ABCG2 levels in 33% of samples.
Sargent et al 2001155	AML	IHC: BXP-34 antibody	20 AML samples	27% of samples had >10% of cells stain positive.
van der Kolk et al 2002 156	AML	FC: BXP-21, mitoxantrone transport	20 paired de novo and relapsed/refractory AML samples	Higher ABCG2 protein levels and lower mitoxantrone accumulation in CD34+CD33- cells; lower expression and higher mitoxantrone accumulation in CD34-CD33+ cells
Sauerbrey et al 2002157	Childhood ALL	RT-PCR	47 de novo childhood ALL samples 20 relapsed childhood ALL samples	No correlation between ABCG2 expression and response.
Abbott et al 2002158	AML	RT-PCR	40 de novo AML samples	Intermediate to high levels of ABCG2 in 83% of samples.
Steinbach et al 2002159	Childhood AML	RT-PCR	59 de novo AML	In non-remission group, ABCG2 levels were 10- fold higher compared to remission group. In 9 patients with initial stage and relapse samples, ABCG2 higher at relapse.
van den Heuvel-Eibrink et al 2002160	AML	RT-PCR	20 paired de novo and relapsed/refractory AML samples	ABCG2 levels higher at relapse than at diagnosis.
van der Pol et al 2003 161	AML	FC: BXP-21, BXP-34, BODIPY-prazosin transport	45 AML patients	No changes in ABCG2 expression or function at relapse.
Nakanishi et al 2003162	Acute Leukemia	RT-PCR	21 AML samples	ABCG2 expression in blast samples correlated with in vitro resistance to flavopiridol.
Plasschaert et al 2003 163	ALL	FC: BXP-34, mitoxantrone transport	46 de novo ALL samples	Higher ABCG2 expression and function in B-cell lineage ALL.
Stam et al 2004164	Childhood ALL	RT-PCR	26 de novo ALL samples	ABCG2 expression significantly less in infants compared to older children.
Suvannasankha et al 2004165	AML	RT-PCR FC: BXP-21, BXP-34, 5D3, mitoxantrone transport	31 pretreatment blast samples	Increased 5D3 staining correlated with decreased probablility of achieving CR.
Suvannasankha et al 2004166	ALL	RT-PCR FC: BXP-21, BXP-34, 5D3, mitoxantrone transport	30 de novo ALL samples	Significant correlation between BXP-21 staining and shorter DFS.
Benderra et al 2004167	AML	RT-PCR	149 de novo AML samples	ABCG2 expression was prognostic factor of CR, 4-year DFS and 4-year overall survival. Predictive in patients receiving daunorubicin, and mitoxantrone but not idarubicin.
Uggla et al 2005168	AML	RT-PCR	40 de novo AML samples	Of responders, patients with lower ABCG2 expression had longer survival than patients with high ABCG2 expression.
Galimberti et al 2005169	AML	RT-PCR	51 de novo AML samples	ABCG2 expression was found in 56% of samples; 48.2% had intermediate levels 27.6% had high levels.
Raaijmakers et al 2005 103	AML	RT-PCR FC: BXP-21, mitoxantrone Transport	22 de novo AML samples	Highest ABCG2 protein and function in most primitive CD34+CD38- AML population versus more differentiated CD34+CD38+ and CD34- populations.
Benderra et al 2005170	AML	FC: BXP-34, mitoxantrone transport	85 de novo AML samples	ABCG2 function correlated with prognosis and was higher in non-responders and in patients with poor DFS and poor OS. Higher ABCG2 function associated with lower CR rate.
Wilson et al 2006171	AML
Damiani et al 2006172	AML	FC: BXP-34	73 de novo AML samples with normal karyotype	ABCG2 protein overexpressed in 33% of samples
van den Heuvel-Eibrink 2007173	AML of older age	RT-PCR	154 de novo AML samples	Higher ABCG2 expression associated with secondary AML. ABCG2 expression associated with ABCB1 expression. CD-34 related ABCB1/ABCG2 coexpression significantly associated with a lower CR rate.
Kourti et al 2007174	Childhood ALL	RT-PCR	31 bone marrow samples	High ABCG2 expression in 32% of samples.
Jiang et al 2007106	CML	RT-PCR	18 CML samples	Lin-CD34+CD38-cells express high ABCG2 ABCB1 and low OCT1.
de Figueiredo-Pontes et al 2008102	AML	FC: BXP-21	26 de novo AML samples	Higher ABCG2 expression in leukemic stem cells (CD34CD38-CD123+) versus more mature (CD34+CD38+) cells.
Shman et al 2008175	Childhood AML	RT-PCR	29 childhood AML samples	Higher ABCG2 expression in CD34+ cells.
Ho et al 2008104	AML	RT-PCR	31 de novo AML samples	No difference in ABCG2 expression in total blasts from patients achieving CR versus NR, but higher ABCG2 expression in CD34+CD38- AML subset in NR versus CR samples.
Fedasenka et al 2008176	Childhood ALL	RT-PCR	19 childhood B-cell ALL samples	ABCG2 levels lower in patients with minimal residual disease but not statictically significant.
Cortez et al 2009177	Childhood ALL	RT-PCR	118 de novo B-cell ALL 22 de novo T-cell ALL	ABCG2 expression in the whole population and ABCG2 and ABCC1 expression in high-risk group associated with PFS. Low ABCG2 expression associated with higher treatment-related death in the whole population and low ABCG2 and ABCC1 expression related to treatment-related death in high-risk group.
Svirnovski et al 2009178	Pediatric ALL, AML Adult CLL	Semi-quantitative RT-PCR FC: BXP-21, mitoxantrone transport	65 de novo Childhood ALL 42 relapse Childhood ALL 53 de novo Childhood AML 16 relapse Childhood AML 16 de novo adult CLL 23 treated adult CLL	Low ABCG2 protein expression in pediatric acute leukemias and adult CLL. Functional ABCG2 in 100% of relapsed ALL, 33% of de novo AML, 17% of relapsed AML and 50% of CLL samples. No difference in ABCG2 expression between de novo and relapsed samples.

One difficulty in determining the role of any ABC transporter in drug resistance is that studies often examine only one transporter, much as the case in mouse studies where examining mice lacking Abcg2 only showed modest increases in CNS uptake of some drugs⁸⁷. When gene expression profiles were generated from 170 pretreatment blast samples obtained from patients with AML in a study by Wilson and colleagues, unsupervised clustering classified patients into 6 distinct groups. The group with the poorest disease free survival rates was found to have high expression of ABCG2, usually accompanied by ABCB1 expression. This group also had the highest rate of resistant disease (77%). The fact that this group contained only 22 patients, only 13% of the total number of patients, suggests that these patients might be missed in studies that feature only a small cohort of patients, arguing for studies to feature larger numbers of patients and determine expression of multiple transporters.

Intriguing results from one study were recently reported by Jiang et al., who found higher ABCG2 and ABCB1 expression as well as lower OCT1 expression in lin-CD34+CD38- cells in CML samples¹⁰⁶. As imatinib requires OCT1 to enter the cell, CML cells expressing lower levels of OCT1 would be expected to uptake lower levels of imatinib, and this would be compounded by expression of ABCB1 and ABCG2, both of which have been shown to transport imatinib. Thus, increased expression of ABCB1 and ABCG2 along with decreased expression of OCT1 might play a role in drug resistance in CML, protecting the CML stem cell and preventing imatinib from reaching its target.

In solid tumors, ABCG2 expression has been studied multiple times in only breast cancer and non-small cell lung cancer. Here again, conflicting results have been reported. In breast cancer, ABCG2 expression was reported to be generally low^{107, 108} and only one study reported a correlation between ABCG2 expression and response¹⁰⁹. Some studies in lung cancer do seem to suggest a correlation between ABCG2 expression and response^110–112; however, all these studies involve platinum-based therapies. Since platinum compounds are not known to be transported by ABCG2, it is unclear exactly how ABCG2 expression could account for the resistance observed in ABCG2-positive tumors. More studies must be conducted before the role of ABCG2 in clinical drug resistance, if any, can be elucidated.

What is striking about some of the clinical studies is a lack of validated controls. Some clinical studies do not include any kind of validated cell line model as a frame of reference when referring to gene expression or protein levels as “high” or “low”. Or, only cell lines expressing extremely high levels of ABCG2 are used--levels which one would not expect to find in clinical samples. We have previously suggested the inclusion of cell lines expressing validated low but detectable expression of ABCG2, such as MCF-7 breast cancer cells, A549 lung cancer cells, HCC-2998 colon cancer cells or SF295 CNS cancer cells⁶⁵, and these cell lines were also found to have low but detectable levels of pheophorbide a transport¹¹³. As shown in table 4, gene expression levels of ABCG2 in drug-selected cell lines, even those selected with relatively low levels of drug, can have significantly higher ABCG2 expression than unselected lines. If methods used in clinical studies cannot detect ABCG2 expression or function in cell lines known to express low but detectable levels of ABCG2, how can the conclusions of these studies be accepted?

Table 4.

Relative ABCG2 gene expression in unselected and drug-selected cancer cell lines.

Cell line	ABCG2 expression
MCF-7	10
MCF-7 FLV100	535.2
MCF-7 FLV250	592.8
MCF-7 FLV500	783.2
MCF-7 FLV1000	1837.6
SF295	41.8
NCI-H460	264
A549	115.8
HCC-2998	52.3

Gene expression was determined by RT-PCR analysis and normalized to the level of expression in MCF-7 parental cells, arbitrarily assigned a value of 10 units. MCF-7 FLV100, 250, 500 and 1000 cells are maintained in 100, 250, 500 and 1000 nM flavopiridol, respectively. MCF-7 AdVp3000 cells are maintained in 3000 ng/ml adriamycin with 2.5 µM verapamil.

Clinical Interventions with ABCG2

With its normal tissue expression profile and its potential role in clinical drug resistance, ABCG2 presents several points of possible clinical intervention. In cancer therapeutics, a strategy of ABCG2 inhibition could be employed to increase oral biovavailability of chemotherapy agents, increase brain penetration, or overcome cancer cell resistance. With regard to oral bioavailability, a limited number of clinical trials have already demonstrated the feasibility of coadministering dual ABCG2/Pgp inhibitors with oral chemotherapy^78–80. Depending on the compounds being orally administered, inhibiting multiple ABC transporters would most likely be necessary. Coadministering an ABC transporter inhibitor with chemotherapeutic agents that are transporter substrates could theoretically decrease interpatient variability caused by transporter expression and increase ease of administration of oral agents. Additionally, dose-limiting diarrhea that is normally found when administering oral topotecan was not observed when topotecan was coadministered with elacridar⁷⁸. However, it is not certain that the clinical benefits of IV therapies will be maintained when the drugs are administered orally, since peak levels and drug exposure can vary depending on route of administration. Thus, further clinical trials will be needed to quantify clinical benefit achieved by increasing oral bioavailability.

It should be noted that the impact of inhibiting ABCG2 in the intestine may also be limited by the fact that high local concentrations after oral administration of some substrates may be enough to overcome the transporter’s effect. Several orally administered drugs including gefitinib, nilotinib, and imatinib have been shown to be substrates only at low concentrations and to behave as inhibitors at higher concentrations^{27, 29}. Since concentrations in the gut would be expected to be relatively high, additional inhibition of transporters in the gut may not be necessary.

Concerns of high local substrate concentrations likely do not apply to the blood-brain barrier where very high endothelial ABCG2 levels have been observed. Further, it will probably be necessary to inhibit multiple ABC transporters at the blood-brain barrier to increase brain penetration. This has been repeatedly demonstrated in mouse models where the maximum brain penetration of tyrosine kinase inhibitors as well as topotecan was achieved in mice lacking the genes for both Abcg2 and Abcb1/2^86–89. However, such proof in humans is lacking. While dual inhibitors of ABCG2 and Pgp such as elacridar, tariquidar and dofequidar are readily available and inhibit the blood-brain barrier in mice, what is lacking is a validated probe of ABCG2 (or Pgp) function in the human blood-brain barrier. Without such a probe to prove inhibition of transporters at the blood-brain barrier, results of clinical trials combining a tyrosine kinase inhibitor with an ABC transporter inhibitor would be difficult to interpret. However, several imaging agents have been studied that may be useful in this regard. N-desmethyl-loperamide is being explored for imaging Pgp in the CNS in patients¹¹⁴ and animal studies suggest that radiolabeled gefitinib may be useful in this regard⁹¹.

Clinical trials attempting to reverse ABCG2-mediated resistance would be an even more difficult undertaking, due partly to the fact that a role for ABCG2 in clinical resistance has not been clearly proven in any disease and also due to the lack of clinical success with Pgp inhibitors. If any clinical trials are attempted, they would probably be conducted with AML, as the most data for ABCG2 expression seems to have been gathered with this disease, but, as mentioned above, conflicting data have been been reported. Indeed, we would argue that such trials should not be undertaken prior to the development of correlative assays to aid in the interpretation of clinical trial data. First, validated assays of ABCG2 and Pgp must be developed and routinely used to determine levels of expression of transporters. Additionally, functional imaging assays must be developed to prove that given inhibitors can alter drug accumulation in tumors and evaluate drug penetration into sanctuary sites such as the brain.

Clinical intervention for the Q141K SNP would require a completely opposite approach. If this SNP results in decreased transport of carcinogens in the gut and decreased urate transport in the kidneys, increasing expression or increasing transport efficiency of ABCG2 in these sites may provide increased protection from carcinogens or decreased incidence of gout. It has been shown that, for mutated forms of CFTR, a “pharmacologic chaperone” can be used to “rescue” the mutant protein and result in increased membrane expression^{115, 116}. Our laboratory has shown that mitoxantrone can function as a pharmacologic chaperone for ABCG2, increasing levels, improving processing, and increasing cell surface expression^117–119. The ideal pharmacologic chaperone for ABCG2 would be a compound that is able to increase protein expression or function while exhibiting no toxicity and which does not act as an inhibitor of the protein. Some studies have shown that some dietary flavoniods can increase protein expression along with gene expression of ABCG2 in the Caco-2 cell line model¹²⁰; however, caution must be exercised as a number of flavonoids have also been shown to inhibit ABCG2.

In short, this is a very exciting time for studies of ABC transporters in general. Our understanding of the importance of the transporters in human physiology and disease has increased far beyond that original discovery of Pgp, with its sole role as a mediator of drug resistance in cancer. While the studies of the 1990’s failed to prove the Pgp hypothesis in drug resistance, we believe it would be unwise to assume that there will never be a role for ABC transporters as a target in cancer therapeutics.