CHAPTER 18 – SOLVING THE PROBLEM OF MULTIDRUG RESISTANCE ABC TRANSPORTERS IN CLINICAL ONCOLOGY

CHAPTER 18 – SOLVING THE PROBLEM OF MULTIDRUG RESISTANCE ABC TRANSPORTERS IN CLINICAL ONCOLOGY CHAPTER 18 – SOLVING THE PROBLEM OF MULTIDRUG RESISTANCE ABC TRANSPORTERS IN CLINICAL ONCOLOGY CHAPTER 18 – SOLVING THE PROBLEM OF MULTIDRUG RESISTANCE ABC TRANSPORTERS IN CLINICAL ONCOLOGY CHAPTER 18 – SOLVING THE PROBLEM OF MULTIDRUG RESISTANCE ABC TRANSPORTERS IN CLINICAL ONCOLOGY CHAPTER 18 – SOLVING THE PROBLEM OF MULTIDRUG RESISTANCE ABC TRANSPORTERS IN CLINICAL ONCOLOGY CHAPTER 18 – SOLVING THE PROBLEM OF MULTIDRUG RESISTANCE ABC TRANSPORTERS IN CLINICAL ONCOLOGY

I NTRODUCTION

Acquired drug resistance was first observed in

a laboratory model in 1950, in mouse leukemic

cells passaged in mice treated with

4-amino-N10-methyl-pteroylglutamic acid (Burchenal

et al., 1950) In 1972, Dano described drug

resistance due to the active outward transport

of chemotherapeutic agents (Dano, 1973)

Daunorubicin-selected resistant tumor cells

were found to have energy-dependent

trans-port of daunorubicin that could be inhibited by

vinblastine, vincristine, and other

anthra-cyclines Further, selection of cells for resistance

to vinblastine resulted in the same phenotype

Later, Biedler, Beck and Ling more fully

charac-terized the multidrug resistance phenotype

(Beck et al., 1979; Biedler and Peterson, 1981;

Riordan and Ling, 1979) Tumor cell lines

that were selected in the laboratory for

resis-tance to doxorubicin or vincristine became

cross-resistant to structurally unrelated

anti-cancer agents, displayed active outward drug

efflux, and were characterized by increased

expression of a 170 kDa cell membrane

glyco-protein that became known as P170 or

P-glyco-protein As critical as this discovery of the first

human ATP-binding cassette (ABC)

trans-porter was, it was the observation that drug

resistance could be reversed in vitro by several

different compounds, including verapamil,

that brought Pgp into prominence as a potential

target for improving cancer therapy (Tsuruo

et al., 1981) The first section of this chapter will

briefly review the mammalian ABC porters linked to multidrug resistance (dis-cussed in more detail in Chapters 5, 6 and19–21) Subsequently, the progress that hasbeen made in developing ABC transporters asclinical targets in anticancer therapy will bereviewed To date, 48 human ABC genes havebeen identified and classified into seven dis-

trans-tinct subfamilies (Dean et al., 2001) The

Human Gene Nomenclature Committee hasdesignated these subfamilies as ABCA through

ABCG (Klein et al., 1999) However, the

tradi-tional more familiar names will be used for themajority of the transporters described below

et al., 1986; Gros et al., 1986; Scotto et al., 1986) and

human P-glycoprotein, MDR1 (Ueda et al., 1987)

studies were aimed at exploring the structureand function of P-glycoprotein, and understand-ing its importance in human malignancy P-gly-coprotein (Pgp) is considered a ‘full’ transporter,

18

CHAPTER

comprising 12 transmembrane (TM) segments

divided between two domains, each linked

to an ATP-binding domain Both ATP-binding

domains contain Walker A and Walker B

sequences as well as the active transport family

signature motif ‘C’ characteristic of all ABC

transporters Current studies suggest that the

high-affinity binding of substrate to Pgp results

in ATP hydrolysis, which in turn causes a

confor-mational change in Pgp that shifts the substrate

to a lower-affinity binding site on the protein,

thereby releasing the substrate into either the

outer leaflet of the membrane or the extracellular

space (Ramachandra et al., 1998) Hydrolysis at

the second ATP-binding domain is required to

reset the protein conformation to allow binding

of a new substrate molecule (Sauna and

Ambudkar, 2001; Senior and Bhagat, 1998; van

Veen et al., 2000) Thus, Pgp has been viewed as a

‘two-cylinder engine’ (see also Chapters 4–6)

In vitro studies have shown that

overexpres-sion of Pgp in cancer cells confers high levels of

resistance to anthracyclines, Vinca alkaloids,

taxanes, etoposide, and probably hundreds, if

not thousands, of other compounds (Gottesman

and Pastan, 1993; Scala et al., 1997) Numerous

studies suggest that the principal physiological

role for Pgp is to protect the organism from

toxic substances This evidence includes the

identification of Pgp expression at sites that are

involved in drug excretion or at ‘sanctuary

sites’, including the epithelium of the

gastro-intestinal tract, the renal proximal tubule, the

canalicular surface of the hepatocyte, and

the endothelial cell surface comprising the

blood–brain barrier (Cordon-Cardo et al.,

1989, 1990; Thiebaut et al., 1987) Further

evi-dence is derived from in vivo knockout mouse

models in which the murine orthologue for

Pgp has been deleted or disrupted These mice

are healthy, reproduce normally, but display

altered sensitivity to, and excretion of,

com-pounds that are Pgp substrates (Borst and

Schinkel, 1996; Schinkel et al., 1994, 1997) In

human cancer, Pgp expression appears to be

due either to continuation of the phenotype

found in the normal tissue of origin or to

upregulation following exposure to anticancer

agents Numerous studies have attempted to

define the extent of Pgp expression in various

tumor types and correlate that information

with clinical endpoints such as response to

chemotherapy and survival In addition,

evi-dence establishing the importance of Pgp in

can-cer has been sought in clinical trials with Pgp

inhibitors As discussed below, these studies

have advanced our understanding of how toapproach Pgp and other ABC transporters astherapeutic targets, but have not yet generatedconvincing evidence for the use of inhibitors inclinical oncology

MULTIDRUG RESISTANCE PROTEIN1, MRP1 (ABCC1)

In 1992, MRP1 was identified as a second

human ABC drug transporter (Cole et al., 1992).

Cloned from a multidrug resistant human lungcarcinoma cell line, MRP1 has an additionalfive transmembrane segments (TMD0 orMSD1) located at the NH2-terminus of the pro-tein connected to a Pgp-like core by a linkerregion (L0 or CL3) (for further details, seeChapter 19) Mutational analyses have sug-gested that this linker region may be partlyresponsible for the organic anion affinity ofMRP1 but other regions of the protein clearly

participate as well (Bakos et al., 1998; Leslie

et al., 2001) (Chapter 19) Disruption of Mrp1

in murine embryonic stem cells results in a three- to fourfold increase in sensitivity toetoposide and teniposide, and a twofoldincrease in sensitivity to vincristine, doxoru-

bicin and daunorubicin (Lorico et al., 1996).

Overexpression of MRP1 confers resistance toetoposide, doxorubicin and vincristine; andMRP1 has also been shown to transport glu-tathione conjugates, glucuronides and sulfates

(Cole et al., 1994; Jedlitschky et al., 1994, 1996; Leslie et al., 2001) Further, MRP1 is able to

co-transport certain natural product substrates,such as vincristine with glutathione, without

covalent conjugation of the drug (Borst et al., 2000b; Hipfner et al., 1999; Leslie et al., 2001; Loe et al., 1998) Additional evidence has been

presented suggesting that MRP1 is able totransport irinotecan and its active metabolite,7-ethyl-10-hydroxy camptothecin (SN-38), com-pounds that are glucuronidated in normal

metabolism (Chen et al., 1999) Together, these

studies indicate that MRP1 is able to transportboth unmodified and modified xenobiotics.Recently, it was also discovered that MRP1 canconfer resistance to methotrexate, an antifolateantineoplastic agent not usually associatedwith the multidrug resistance phenotype

(Hooijberg et al., 1999) Like Pgp, MRP1 is

thought to provide protection to normal sues, and to be involved in drug disposition

tis-(Wijnholds et al., 2000b) Unlike Pgp, low-level

expression of MRP1 is ubiquitous throughout

the body with higher levels expressed in the

lung and kidney (see Chapter 19)

OTHERMRPS

Multiple MRP (ABCC) family members have

been identified (Borst et al., 2000b; Dean et al.,

2001) (see Chapters 20 and 21) MRP1, 2, 3 and

6 have the highest homology with one another,

with 17 predicted TM segments: a Pgp-like core

encoding two ATP-binding domains and two

membrane-spanning domains, with an

addi-tional NH2-proximal five TM segment region

(TMD0) (described in the previous section)

(Borst et al., 2000b; Leslie et al., 2001) In

con-trast, MRP4 (ABCC4), MRP5 (ABCC5), ABCC11

and ABCC12 lack the TMD0 characteristic of

MRP1, MRP2, MRP3 and MRP6 MRP4

and MRP5 have been shown to transport

nucleo-sides (Chen et al., 2001; Dean et al., 2001;

Jedlitschky et al., 2000; Wijnholds et al., 2000a),

while the functions of ABCC11 and ABCC12

are not yet known MRP2 (ABCC2), also

known as cMOAT (canalicular multispecific

organic anion transporter), has been identified

as the bilirubin glucuronide transporter

(Buchler et al., 1996; Paulusma et al., 1996) (see

Chapter 20) The Dubin–Johnson syndrome in

humans, as in the TR⫺and EHBR rat models, is

characterized by mutations in MRP2(ABCC2)

which result in the absence of the protein in the

canilicular membranes of the liver (Buchler

et al., 1996; Paulusma et al., 1996; Toh et al., 1999).

Patients accumulate an excess of bilirubin

glu-curonide and unconjugated bilirubin, resulting

in hyperbilirubinemia and hepatic

inflamma-tion Mutations in MRP6 (ABCC6) have been

linked to the connective tissue disorder

pseu-doxanthoma elasticum but have no known role

in drug resistance (see Chapters 21 and 28)

The question of whether MRP2 can confer

multidrug resistance has been addressed by

in vitro transfection studies, with both sense and

antisense MRP2 cDNA constructs Both types

of studies support the conclusion that MRP2 is

able to transport cisplatin as well as the MRP1

substrates etoposide, doxorubicin, vincristine

and methotrexate (Cole et al., 1994; Cui et al.,

1999; Koike et al., 1997; Masuda et al., 1997).

However, the prevalence of increased

expres-sion of MRP2 as a mechanism of resistance to

cisplatin and other anticancer drugs is not yet

known (Kool et al., 1997; Taniguchi et al., 1996)

(see Chapter 20)

Like MRP1, MRP3 (ABCC3) has been shown

to transport etoposide, doxorubicin, vincristine

and methotrexate (Hooijberg et al., 1999; Kool et

al., 1999; Zeng et al., 1999) MRP3 is expressed

at relatively high levels in human liver, ized to the basolateral surface of the hepatocyte

local-(Konig et al., 1999), where, like MRP1, it may be

involved in the transport of organic anionsback into the bloodstream Studies with MRP4and MRP5 have demonstrated transport of cyclicnucleotides, and resistance to 6-mercaptopurineand 6-thioguanine, two anticancer purine ana-

logues (Chen et al., 2001; Jedlitschky et al., 2000;

Wijnholds et al., 2000a) Taken together, the

findings suggest that the MRP subfamily ofABC transporters has a role, with some possi-ble built-in redundancy, in drug disposition

That function may be subverted by a cancer cell

in becoming drug resistant However, to date,conclusive links to clinical drug resistance havenot been established for MRP family membersother than MRP1 (see also Chapter 21)

SPGP/BSEP (ABCB11)

Structurally homologous to MDR1/Pgp, the

‘sister of P-glycoprotein’ was originally clonedfrom the hamster in a search for genes with

homology to MDR1 (Childs et al., 1995)

Subse-quently recognized as the bile salt exporter tein (BSEP), SPGP/BSEP (ABCB11) plays animportant role in biliary homeostasis (Gerloff

pro-et al., 1998) While evidence for a role for

SPGP-BSEP in drug resistance is limited, it is ing to note that paclitaxel is also a substrate for transport by this protein Overexpression

interest-of SPGP/BSEP in human ovarian SKOV3 cellsconferred a fourfold resistance to paclitaxel

(Childs et al., 1998) Sensitization by PSC833,

cyclosporin A and verapamil (typical Pgp/

MDR1 inhibitors) was observed

ABC2 (ABCA2)

Active outward efflux has also been observed

in SKEM cells, a human ovarian carcinoma cellline selected for estramustine resistance (Laing

et al., 1998) Estramustine is not known to be a

substrate for Pgp, and the resistant SKEM cellshave a phenotype distinct from that associatedwith overexpression of Pgp Amplification of

ABCA2 was detected in these cells, and

antisense-mediated downregulation of ABCA2sensitized the resistant cells to estramustine

(Laing et al., 1998) ABC2/ABCA2 belongs to

the ABCA subfamily, which also includes

ABCA1, the transporter linked to cholesterol

transport, and ABCR (ABCA4), the transporter

linked to retinal integrity (Broccardo et al.,

1999) (see Chapters 23 and 28)

MXR/BCRP/ABCP (ABCG2)

A member of the ABCG subfamily, MXR/

BCRP/ABCP (ABCG2), is a ‘half transporter’

able to confer high levels of drug resistance to

mitoxantrone, topotecan, CPT-11 and its active

metabolite SN38, as well as anthracyclines

(Allikmets et al., 1998; Brangi et al., 1999; Doyle

et al., 1998; Litman et al., 2000; Miyake et al.,

1999) Thus, its substrate specificity appears

somewhat more limited than Pgp and MRP1 In

addition, flavopiridol, a new cell cycle inhibitor

in clinical trials, has been found to be a

sub-strate for ABCG2 (Robey et al., 2001) A single

ATP-binding domain followed by six TM

seg-ments comprising a single membrane-spanning

domain make up the half-size transporter

des-ignated ABCG2, which is thought to require

dimerization to form a functional unit Two

other members of this subfamily are involved

in sterol transport (ABCG5 and ABCG8) (see

Chapter 22), but a normal function for ABCG2

is not yet known (Dean et al., 2001) High levels

of ABCG2 are found in the syncytiotrophoblast

cells of the placenta, where the function could

be either transport of toxins out of, or

trans-port of nutrients into, the fetal circulation

(Maliepaard et al., 2001) In Pgp-deficient mice,

increased bioavailability and fetal penetration

of topotecan was observed following

coadmin-istration of topotecan and GF120918, a Pgp

inhibitor found to also inhibit ABCG2 (Jonker

et al., 2000) A murine transporter, Abcg3, with

high homology to human ABCG2 has been

described (Mickley et al., 2001) Its tissue

dis-tribution pattern is different from ABCG2,

suggesting the two transporters are not

coex-pressed Overexpression and amplification of

ABCG2 occurs during in vitro selection of cells

with mitoxantrone or topotecan (Knutsen et al.,

2000; Maliepaard et al., 1999) Recent studies

have also shown that the substrate specificity of

ABCG2 can be significantly altered by a

differ-ence in a single amino acid (Honjo et al., 2001).

OTHERABC TRANSPORTERS

For many of the ABC transporters listed above,

no conclusive direct evidence has been

obtained to suggest a role in clinical drug

resistance For some transporters, importantendogenous substrates are known to exist, anddrug transport is probably a secondary func-tion One question is whether the function of atransporter can be subverted to serve as amediator of multidrug resistance in tumorcells In one scenario, an ABC transporter notnormally expressed at high levels may beupregulated, induced, or redistributed to thecell surface, and in doing so, becomes capable

of conferring drug resistance In another nario, mutation of a transporter protein couldresult in a gain of function For example,ABCG2 confers resistance primarily to mitox-antrone and camptothecin analogues; however,mutation of amino acid 482 adds rhodamineand anthracyclines to the list of substrates it

sce-can transport (Honjo et al., 2001) Similarly,

only minor sequence changes are required

to improve the efficiency of drug transport

by MDR3/Mdr2 (ABCB4), a choline flippase or translocator closely related

phosphatidyl-to Pgp (MDR1) that normally transports

phos-pholipids into the bile (Borst et al., 2000a; Smit et al., 1993; Zhou et al., 1999) (see Chapter22) Mutations such as these have not beendemonstrated in clinical cancer to date With atleast 48 ABC transporters encoded in thehuman genome, this list of transporters with apotential role in drug resistance may yet beincomplete However, the list of substratesencompassed by the already described trans-porters is quite extensive, and includes some ofthe newest agents in the anticancer drug arma-mentarium It could be argued with consider-able conviction that no anticancer agent could

be identified for which a drug transportercould not be found

MVP/LRP

Not an ABC transporter, but included in manyclinical studies of multidrug resistance, MVP

(major vault protein) (also known as LRP, lung

resistance protein) is a component of the

multi-meric vault proteins which are found in thecytoplasm and in the nuclear membrane

(Scheffer et al., 2000b) Thought to mediate

redistribution of drugs away from the nucleus,the expression of vaults may be coordinatelyregulated with Pgp or MRP1 although directevidence that this is the case is lacking.MVP/LRP expression has been detected in lungcancer, acute leukemia and ovarian cancer Inseveral studies, expression of MVP/LRP has

been a better correlate of poor prognosis than

Pgp (den Boer et al., 1998; Izquierdo et al., 1995;

Table 18.1lists many of the compounds found

to be inhibitors of Pgp-mediated drug efflux

and drug resistance Characterized as both

competitive and non-competitive inhibitors,

these agents are able to increase

chemosensitiv-ity in in vitro models by several orders of

mag-nitude Early characterization of Pgp inhibitors

in vitro led to trials with what are now referred

to as first-generation inhibitors These

com-pounds were already used in clinical medicine

and found in the laboratory to be inhibitors of

Pgp and were used in combination with an

anticancer agent known to be a Pgp substrate

Several reviews that catalogue these trials are

available (Bradshaw and Arceci, 1998; Ferry

et al., 1996; Fisher and Sikic, 1995; Fisher et al.,

1996) These trials demonstrated the safety ofcombining a Pgp inhibitor with a chemothera-peutic agent, but fell far short of the goal ofdefining a role for Pgp inhibition in clinicaloncology This, in turn, meant that a role forPgp in conferring clinical drug resistance wasalso not confirmed

The failure of the first-generation Pgpinhibitor trials to support a role for inhibition

of this ABC transporter in clinical oncologycould be ascribed to several factors First, asPgp inhibitors, the first-generation agents werenot very potent, requiring micromolar concen-trations for effective inhibition Concentrationscomparable to those that were effective in labo-ratory models could seldom be obtained with-out toxicity in patients Second, the trials weredesigned to identify a ‘home run’; thus, theinhibitors were administered with the anti-cancer agents without first requiring either thattumors be clearly refractory to treatment, orthat randomization be incorporated into thetrial design Third, the trials never soughtphysical evidence that Pgp inhibition was

occurring in vivo Finally, assays were usually

not included to confirm the presence of Pgpexpression or function in the tumors

Second-generation Pgp inhibitors were typically analogues of first-generation agents,developed specifically for the purpose of Pgp inhibition These included R-verapamil(stereoisomer of verapamil) and PSC 833 (deriv-ative of cyclosporin D) These agents were morepotent than many of the first-generation agentsbut still did not achieve the success sought interms of efficacy Nor did they confirm a role forPgp inhibition in clinical oncology Trials withthese second-generation agents again confirmedthe safety of adding a Pgp inhibitor to therapywith conventional agents, with the caveat thatpharmacokinetic interactions necessitated alower dose of the anticancer agent in combina-tions, including PSC 833 Perhaps the mostimportant outcome of the completed Pgp rever-sal trials was the recognition that a distinctionneeded to be made between the efficacy of theinhibitor in blocking Pgp and the efficacy of theinhibitor in improving cancer treatment

Trials with third-generation agents are now

in progress, more than 25 years since the tification of the molecular target, Pgp, andmore than 20 years since the identification

iden-of the first Pgp inhibitor, verapamil Several iden-ofthese compounds are reported to have little or

no pharmacokinetic interactions, overcoming a

TABLE18.1 P-GLYCOPROTEIN INHIBITORS USED IN CLINICAL

DEVELOPMENTa

First-generation agents Verapamil

Quinidine Quinine Amiodarone Nifedipine Second-generation agents R-verapamil

PSC 833 Dexniguldipine Third-generation agents GF120918

VX710 R101933 XR9576 LY335979 OC144-093

aAgents shown represent only a partial list.

major problem linked to the use of PSC 833.

These compounds include XR 9576, R101933,

LY335979, OC144-093, and GF120918 (Dantzig

et al., 1999; Mistry et al., 2001; Newman et al.,

2000; Sparreboom et al., 1999; Starling et al.,

1997; van Zuylen et al., 2000a) The compounds,

evaluated in studies enlightened by lessons

from the first- and second-generation inhibitor

trials, offer the potential to finally discover the

importance of Pgp in clinical oncology

It seems logical that the efficacy of a Pgp

inhibitor in the clinic will be linked to the

importance of Pgp in drug resistance So clear is

this logic that investigators in this field have

largely relied upon the clinical trial process to

provide the answer to the question of whether

Pgp is significant in clinical oncology This may

have been the central flaw in the past decade of

clinical research In breast cancer, markers such

as the estrogen receptor, erbB2, aneuploidy, and

S-phase are measured in thousands of patients,

with steadily improving uniformity of

tech-nique, and correlated with clinical outcome In

contrast, in the field of multidrug resistance, we

have relied upon ‘drug resistance reversal

tri-als’ to answer the question of whether Pgp is

important in cancer treatment If a concerted

effort to identify the diseases in which Pgp

expression confers a resistant phenotype had

been made, we might have set the stage for

well-conceived clinical trials Instead, selection

of the tumor types and trial designs for clinical

studies has relied as much upon guesswork as

upon facts

Early studies of Pgp demonstrated frequentand high levels of expression in colon, kidney,

adrenocortical and hepatocellular cancers (Fojo

et al., 1987; Goldstein et al., 1989) Initially, there

was hope that Pgp could explain the profound

intrinsic drug resistance found in these cancers

However, the failure of these cancers to respond

to therapies with drugs not transported by Pgp

suggested that Pgp alone could not account for

the intrinsically drug-resistant phenotype, and

attention has turned to cancers that respond to

chemotherapy initially, but ultimately acquire

resistance Numerous clinical studies evaluating

or measuring Pgp expression and/or functionhave appeared, and Pgp expression has beencorrelated with clinical outcome However, thestudies have been largely retrospective, singleinstitution, small studies with insufficient power

to provide a definitive statistical outcome

One problem with designing a study ered to provide this information is that meth-ods for Pgp detection remain imperfect Weand others have previously delineated these

pow-issues (Beck et al., 1996; Herzog et al., 1992),

and they can be summarized as follows: (1)mRNA and protein methods that use wholetumor specimens risk contamination with nor-mal tissues, which may increase or decrease thePgp expression level detected; (2) Northernblot analysis for mRNA and immunoblotanalysis for protein expression are not sensitiveenough for the low levels frequently detected

in clinical samples; (3) polymerase chain

reac-tion (PCR) assays for MDR1 mRNA detecreac-tion

are commonly performed with methods thatfail to take into account the fact that quantita-tion is most accurate in the exponential phase ofamplification; (4) immunohistochemical assaysare best for direct examination of individualcancer cells, eliminating problems with normaltissue contamination, but are difficult to quan-titate; (5) antibodies used in immunohisto-chemistry studies are not as specific as needed;(6) Pgp is difficult to detect in formalin-fixedtissue; thus, investigators disagree as to whethermonoclonal antibody C219, one of the mostcommonly used antibodies, can detect Pgp inarchival samples

In an effort to address the discrepanciesamong reports concerning detection of Pgpexpression in clinical samples, Beck and co-workers assembled a workshop at St Jude’sChildren’s Hospital (Memphis, USA) to com-pare Pgp detection methods in use by investiga-

tors from around the world (Beck et al., 1996).

While specific recommendations were made,there is still disagreement on several levels Forexample, should cancer cells be scored as posi-tive for Pgp if membrane staining cannot beidentified? Studies requiring membrane stain-ing often report a far lower frequency of Pgpdetection in breast cancer There are also per-sistent issues of sensitivity Studies utilizing the

PCR method for MDR1 mRNA detection have a

higher frequency of MDR1/Pgp detection thanother mRNA detection methods This can beascribed to the ability of the amplificationprocess to detect mRNAs of low abundance

Another issue discussed at the St Jude’s

Workshop, and still not resolved, is the

develop-ment of a uniform standard for measuredevelop-ments

Since different PCR assays may run at different

efficiencies, it is difficult to know whether the

levels measured by one investigator are

compa-rable to those measured by another, unless

uni-form controls are run For example, in breast

cancer studies, one investigator reported levels

of MDR1 mRNA in tumors as comparable to

levels in normal tissues (Lizard-Nacol et al.,

1999) Since MDR1 mRNA levels in normal

breast tissue are very low, the investigators

con-cluded that levels of expression in breast cancer

were comparably low Use of one or more

stan-dard positive controls would aid in answering

this question across studies

Detection of MRP1 (ABCC1) and other drugtransporters has been less intensively investi-

gated (see Chapters 19–21) MRP1 has been

detected by the same methods used for Pgp:

immunohistochemistry for protein and reverse

transcriptase PCR (RT-PCR) or RNase

protec-tion for mRNA expression Nooter et al (1995)

examined 370 human cancer samples by RNase

protection High levels of MRP1 expression

were found in chronic lymphocytic leukemia

and prolymphocytic leukemia Occasionally,

high levels of expression were found in

esophageal carcinoma, in non-small cell lung

cancer, and in acute myelogenous leukemia

(AML) Predominantly low but ubiquitous

expression of MRP1 was found in the

remain-ing tumor types An additional 108 samples

evaluated by immunohistochemistry with the

monoclonal antibody MRPr1 confirmed these

findings The antibodies most commonly used

in immunohistochemical analyses, MRPr1,

MRPm6 and QCRL-1, recognize sequences

spe-cific for human MRP1 and to date, the

cross-reactivity problems that have plagued Pgp

detection have not arisen (Hipfner et al., 1998).

For other ABC transporters, there is minimalexperience to judge the sensitivity and speci-

ficity of detection methods A panel of specific

monoclonal antibodies has been generated for

detection of other members of the MRP (ABCC)

subfamily but their epitope sequences have not

yet been precisely defined (Scheffer et al.,

2000a) ABCG2 mRNA expression has been

assayed by RT-PCR in single studies in breast

cancer and in leukemia (Kanzaki et al., 2001;

Ross et al., 2000) Polyclonal and monoclonal

antibodies have been developed to detect

ABCG2 (MXR/BCRP), but reports have not yet

appeared describing expression in tumor tissue

of patients with acute myelogenous leukemia(AML) express Pgp at the time of diagnosis, andexpression is observed in cells from about 50% of

patients at the time of relapse (Table 18.2).

Certain subtypes of AML are also noted to havehigher frequencies of detection, including sec-ondary leukemias While not invariable, mosttrials report that Pgp expression is correlatedwith a reduced complete remission rate, and agreater incidence of refractory disease (Filipits

et al., 1998; Legrand et al., 1999; Leith et al.,

1999; Michieli et al., 1999; van der Kolk et al.,

2000) Complete response rates in the range of50–70% are reported in Pgp-negative leukemia,compared to 30–50% in Pgp-positive leukemia

Because of the high correlation between CD34

expression and Pgp expression (Campos et al.,

1992), some investigators have argued that Pgp,rather than conferring the resistant phenotypethrough drug efflux, may instead be a pheno-typic marker of a poor prognosis subset of

leukemia patients However, ex vivo studies

using leukemic cells from patients have shownthat Pgp expression does correlate with reduced

accumulation of daunorubicin (Broxterman et al., 1999; Michieli et al., 1999) In addition, leukemic

cells obtained from patients receiving bicin after administration of a Pgp inhibitor haveshown increased daunorubicin accumulation

daunoru-(Tidefelt et al., 2000) In a recently reported trial, Broxterman et al (2000) found that the prognos-

tic value of Pgp could be mitigated by ing idarubicin, an anthracycline not subject toPgp-mediated efflux, for daunorubicin Onefinal observation supporting a role for Pgp indrug resistance in AML is derived from trials inwhich a Pgp inhibitor was used (cyclosporin A

substitut-or PSC 833) in combination with chemotherapy

Leukemic cells obtained from patients in relapsefollowing treatment with either cyclosporin A

or PSC 833 have decreased expression of

TABLE18.2 EXPRESSION OFP-GLYCOPROTEIN IN

ACUTE MYELOGENOUS LEUKEMIA

Author/year Methoda Population n Categoryb Positive Clinical correlate

Expression studies

Studies with clinical correlations CRc (%)

et al., 2000 accum Intermediate 33 34 61

et al., 1999 secondary Negative 29 48 79 (p⫽ 0.02)

Studies appearing after the 1994 Consensus Conference on MDR Detection Methods (Beck et al., 1996).

aAll immunohistochemical assays were performed on fresh cytospins of leukemic cells Rh123 indicates functional assay with the Pgp substrate rhodamine 123 CsA indicates differences in the functional assay with or without the addition of cyclosporin A Accum indicates accumulation of either rhodamine 123 or calcein AM in the functional assay OS, overall survival.

bEach set is listed high to low levels of transporter expression or function.

cCR, complete remission.

d Expression also correlates with resistant disease, p⬍0.005.

Pgp or MDR1 mRNA (Kornblau et al., 1997; List

et al., 1993, 1996) While it cannot be absolutely

concluded that circumvention of Pgp explained

the clinical outcome, the absence of a correlation

between clinical response and Pgp expression in

this trial stands in contrast to the results obtained

by numerous investigators from different

institu-tions (Table 18.2) Taken together, the clinical

data support an important role for Pgp in drug

resistance in AML

MRP1 and LRP expression have also beenevaluated in leukemia patients MRP1 has been

detected at high levels in chronic lymphocytic

leukemia and in prolymphocytic leukemia

(Nooter et al., 1996b) Levels in AML are less

fre-quently elevated (10–34%) (Legrand et al., 1999;

Leith et al., 1999) These studies are divided

as to whether MRP1 confers a poor prognosis in

a subset of AML patients The non-ABC protein

LRP/MVP (see above) has been detected in

AML and in several series has been found to be

of greater prognostic value than Pgp (Dorr

et al., 2001; Filipits et al., 1998; List et al., 1996; Xu

et al., 1999) In these studies, the well-known

prognostic value of Pgp expression in AML is

not detectable, thus creating a discrepancy that

is difficult to reconcile with earlier data Two of

these studies included patients who had

received Pgp inhibitors, which conceivably

con-founded the analysis (Dorr et al., 2001; List et al.,

1996) The largest trial to date, reported by Leith

et al (1999), found no correlation between

LRP/MVP expression and prognosis in a

popu-lation of previously untreated patients Finally,

low levels of BCRP/MXR (ABCG2) have been

observed in AML samples, with one-third

hav-ing levels as high as 2.6 times those found in the

drug sensitive MCF-7 breast cancer cell line

(Ross et al., 2000).

BREAST CANCER

Detection of Pgp in clinical samples from

patients with solid tumors has been much more

difficult than in hematologic malignancies

These difficulties relate to the lack of specificity

of the antibodies, to the heterogeneity of clinical

samples, and to the lack of standard laboratory

methods Studies published after the 1994

St Jude’s Workshop (see above) have frequently

incorporated the recommendations,

particu-larly relating to the need to use more than one

detection methodology (Beck et al., 1996) This

includes using multiple antibodies or RT-PCR

as a second method for Pgp or MDR1 mRNA

detection, respectively Despite this effort, theresults remain variable as observed by Trock

et al (1997) in a meta-analysis of 31 studies In

the meta-analysis study, 41% of breast tumors

expressed MDR1/Pgp, the frequency of

detectable expression increased after therapy,and expression was associated with a greaterlikelihood of treatment failure However, therewas considerable heterogeneity among thestudies, with the reported incidence rangingfrom 0% to 80% This heterogeneity persists in

studies reported since 1996 As shown in Table 18.3, the detection rate using immunohisto-chemistry still ranges from 0% to 71%, and frus-tratingly, even when the same antibody is being

used (Faneyte et al., 2001; Yang et al., 1999) Most

studies report some expression of Pgp in breastcancers, and many report membrane staining

(Bodey et al., 1997; Chevillard et al., 1996;

Hegewisch-Becker et al., 1998; Schneider et al.,

2001), considered by most investigators to bethe truest indicator of functional Pgp expres-sion Results with RT-PCR methods have beenmuch less revealing, with studies suggesting noincrease in expression relative to normal tissue

(Arnal et al., 2000; Dexter et al., 1998; Faneyte

et al., 2001; Lizard-Nacol et al., 1999) The

dis-crepancy of these results with those obtained

by immunohistochemical methods may be due

to the greater sensitivity of PCR as describedearlier

Several studies have also attempted to relatePgp expression in breast cancer with clinicaldrug resistance Pgp expression has beenobserved to increase in locally advanced breastcancer following therapy, with the incidenceincreasing from 26% to 57% in one study

(Chung et al., 1997) and from 14% to 43% in another (Chevillard et al., 1996) Among 359

samples, including primary cancer, locallyadvanced, and recurrent disease, the incidence

of Pgp expression was 11% in samples obtainedfrom untreated patients, and 30% in samplesfrom patients who had previously receivedtreatment Although the 1997 meta-analysisconcluded that patients with tumors express-ing Pgp were more likely to experience treat-ment failure, several small recent studies havenot been able to confirm a significant impact ofPgp expression on response rate or overall sur-

vival (Honkoop et al., 1998; Linn et al., 1997;

Wang et al., 1997).

Whether MRP1 is found in breast cancer atlevels capable of conferring drug resistance is

not resolved As mentioned previously, MRP1

mRNA is expressed ubiquitously in normal

TABLE18.3 EXPRESSION OFP-GLYCOPROTEIN IN BREAST CANCER

Dexter, et al., 1998 IHC 31 F JSB-1 6%

et al., 1998 30% PriorRx resistance

human tissues and, consequently, finding

expression in tumor tissue is not surprising

Detection of MRP1 mRNA in 100% of samples

by RT-PCR at levels comparable with normal

tissue levels reinforces this point (Dexter et al.,

1998; Filipits et al., 1996) One study reported a

correlation between relapse-free survival in

breast cancer patients and MRP1 expression as

detected by immunohistochemistry (Nooter

et al., 1997) In this series, which comprised

breast cancer samples from 259 patients,

MRP1 expression was detected in 34%

OVARIAN CANCER

The problem of variability continues when

expression studies in solid tumors other than

breast cancer are reviewed Thus, for ovarian

cancer, the reported incidence of Pgp positivity

ranges from 17% to 71% The methodologies

described in these studies are more variable

than those in the breast cancer studies The

study reporting 71% positivity is the outlier,

and was the only one to use immunoblotting as

a detection method with a polyclonal antibody

not used in the immunohistochemical studies

(Joncourt et al., 1998) Two other groups used

antibodies not widely accepted, but potentially

deserving of further testing since good

detec-tion methods for Pgp in archival material have

not been established (Schneider et al., 1998;

Yokoyama et al., 1999b) If the true incidence of

Pgp positivity in ovarian cancer at diagnosis is

less than 20%, it can readily be appreciated that

a drug resistance reversal trial would need to

either select the subset of patients which would

be most likely to benefit from a Pgp inhibitor,

or expand the size of the trial sufficiently to

detect a difference in fewer than one-fifth ofpatients

LUNG CANCER

The majority of breast and ovarian cancer trials have used immunohistochemical methods

to evaluate Pgp expression In contrast, in lung

cancer, Pgp/MDR1 mRNA quantitation

meth-ods are more prevalent Most studies measuring

MDR1 mRNA do so by RT-PCR, and report

approximately a 25% incidence of expression

(Table 18.4), with a range of 15% to 50% MRP1expression is reported at a much higher fre-quency, 70% to 80% in small cell lung carcinoma(SCLC) and 100% in non-small cell lung carci-noma (NSCLC), perhaps not surprising in view

of its relatively high level of expression in mal lung tissue However, few studies havecompared both histologies In studies of lung

nor-cancer cell lines, increased levels of both MRP1 and MRP3, but not MRP2, correlated with

reduced sensitivity to doxorubicin, vincristine,

etoposide and cisplatin (Young et al., 2001),

sug-gesting that these transporters may play a role inthe intrinsic resistance of lung cancer

SARCOMA

Investigators have also considered Pgp sion to be important in sarcomas However,examination of the literature reveals that differ-ent detection methodologies with varyingresults have been reported An early study insoft tissue sarcomas noted a marked impact

expres-of Pgp expression on relapse-free survival and

overall survival (Chan et al., 1990) These

inves-tigators used a unique immunohistochemical

TABLE18.3. (continued)

et al., 1999 96% PostRx breast tissue

Studies were included if they clearly defined the methodology used for MDR1/Pgp detection, and if they delineated

a cut-off for positivity M, membrane staining required for positivity.

Fixation method: C, cytospin – acetone or paraformaldehyde; F, frozen section; P, paraffin-embedded, formalin fixed.

IHC, immunohistochemistry; PCR, polymerase chain reaction.

RR, response rate; OS, overall survival; DFS, disease-free survival; CR, complete response; Rx, therapy.

aJSB-1, C219, C494 gave concordant results.

bLow levels equivalent to those in normal breast tissue.

cLevels defined in relationship to P-glycoprotein expression levels in KB8-5 cells.

TABLE18.4 EXPRESSION OFP-GLYCOPROTEIN, MRP, ANDLRP

IN SELECTED SOLID TUMORS

Reference n Hista Method Pgp (%) MRP (%) Laboratory or clinical

correlation

Ovarian cancer

van der Zee et al., 1995 89 IHC 15 48% positive postRx: p⬍ 0.001

Yokoyama et al., 1999b 58 IHC 27.6 22.4 MRP with RR: p⬍0.01

Lung cancer

Nooter et al., 1996a 35 NSCLC IHC 74 Membrane staining in 34%

lung; 32% high

Narasaki et al., 1996 6 SCLC RT-PCR 100 SCLC levels comparable to nl lung

11 NSCLC RT-PCR 100 NSCLC levels below nl lung

Savaraj et al., 1997 31 SCLC RNAblot 26 RR: p⬍ 0.01; OS: 10 mo vs 2

Yokoyama et al., 1999a 159 NSCLC IHC 60 OS: 74% vs 48%, p⬍ 0.05

Wright et al., 1998 109 NSCLC IHC 87 73% intermediate/high levels

Oshika et al., 1998 107 NSCLC IHC 44 Cancer cell cytoplasm/nl

bronchial epithelium

Sarcoma

Kuttesch et al., 1996 76 RMS IHCb 41 CR or OS: p⬎ 0.05

OS: p⬍ 0.0000267

Levine et al., 1997 65 STS IHC 48 DFS: 32% vs 18% p⫽ 0.039;

OS: 54% vs.14%, p⫽ 0.07

Baldini et al., 1995 92 Osteo IHC 30 RFS: 80% vs 42% p⫽ 0.002

increase post-Rx

Chan et al., 1997 62 Osteo IHC 44 RFS: 87% vs 0%; OS: 94% vs

35% p⬍ 0.00001

Wunder et al., 2000 123 Osteo RT-PCR 65 High in 36%; DFS: p⬎ 0.05

Perri et al., 2001 53 ES IHC 64 Pgp 3⫹32%; DFS, OS: p ⬎ 0.05

Abbreviations: Rx, therapy; RFS, relapse-free survival; OS, overall survival; DFS, disease-free survival; RR, response rate; CR, complete response; PD, progressive disease; NSCLC, non-small cell lung cancer; SCLC, small cell lung cancer; STS, soft tissue sarcoma; Osteo, osteosarcoma; RMS, rhabdomyosarcoma; SS, synovial sarcoma; ES, Ewing’s sarcoma; IHC, immunohistochemistry; RT-PCR, reverse transcriptase–polymerase chain reaction.

aPgp positive excluding diffuse weak staining, categorized as Pgp negative.

bDetection also by RT-PCR, 51% no correlation with survival.

methodology incorporating detection of Pgp

in paraffin-embedded tissue with monoclonal

antibody C219 Similar results were reported

in osteosarcoma using this same methodology

(Chan et al., 1997) However, as shown in

Table 18.4, these findings have been both

substantiated and disputed in studies using

other methodologies Thus the importance

of Pgp in this tumor remains uncertain

(Baldini et al., 1995; Coley et al., 2000; Kuttesch

et al., 1996; Perri et al., 2001; Wunder et al.,

2000)

CONCLUSIONS

In light of the recent progress in the

identifica-tion and characterizaidentifica-tion of new ABC

trans-porters, it seems fitting to re-examine recent

literature relating to the role of Pgp in clinical

drug resistance in five tumor types in which

Pgp has been thought to be important Few

studies emphasize serial clinical samples,

which have the potential to show the

acquisi-tion of multidrug resistance coincident with

increased expression of Pgp Sadly, recent

stud-ies reporting the incidence of expression of Pgp

in clinical samples appear to be as discordant

as older studies What can we conclude from

the available data? It seems that, despite efforts

to bring uniformity to the methods used to

measure Pgp levels in clinical samples, it is still

difficult to discern which studies are valid

and which are not The most valid data appear

to be those obtained in AML, where

immuno-staining, MDR1 mRNA measurments, and

functional studies all confirm Pgp expression

in a subset of patients presenting with this

disease Pgp expression in AML is associated

with a decreased complete response rate and

overall survival In solid tumors, carefully

per-formed studies repeatedly find some fraction

of samples positive, although correlations with

response and survival are more variable Taken

together, the studies reviewed here suggest an

incidence of Pgp expression of 30% in de novo

leukemia, 50% in relapsed/refractory/secondary

leukemias, 40% in breast cancer, 20% in ovarian

cancer, 25% in lung cancer, and 30% in

sar-coma This rate of positivity can be regarded as

sufficient to indicate an important role for Pgp

in clinical oncology However, it suggests that

subsets of patients need to be selected for

mul-tidrug resistance reversal trials, since some

tumors do not develop Pgp as a mechanism of

be non-toxic in preclinical development ever, the administration of PSC 833 to patientsrequired reduction of anticancer drug doses by25–70%, in order to prevent toxicity This dosereduction was determined empirically as thedose of the inhibitor was increased in phase Itrials The greatest impact appeared to be ondosing with paclitaxel and vinblastine

How-Dose reductions were due to a delay in ance of the anticancer drug, and were initiallythought to be innocuous, since it was assumedthat the delayed clearance would result in a com-parable area under the concentration versus timecurve (AUC) If all that mattered in cancerchemotherapy was the duration of drug expo-sure above a certain threshold, then treatingpatients with doses that resulted in equivalenttoxicity would mean equivalent efficacy Thisassumption proved to be entirely wrong, andprovided an important pharmacology lesson to anumber of clinical scientists working in mul-tidrug resistance Some studies do report equiva-lent AUCs However, decreased clearance means

clear-a longer hclear-alf-life If the AUC is cclear-alculclear-ated toinfinity, the long terminal half-life may accountfor a significant portion of the AUC calculation,missing the fact that the maximal concentration –and potentially the effective concentration – is infact reduced Indeed, two studies reported actualreductions in the AUC, one with paclitaxel andthe other with doxorubicin, associated with 30%

and 65% dose reductions, respectively (Advani

et al., 2001; Fracasso et al., 2000) The assumption

also did not take into account changes in clearance of metabolites Thus, the AUCs for

6-hydroxy-paclitaxel and doxorubicinol were

increased by 222% and 259%, respectively, in a

phase I trial in which patients received both

doxorubicin and paclitaxel with PSC 833

(Advani et al., 2001) Similarly, the AUC for

dox-orubicinol increased following administration

of the Pgp inhibitor GF120918, while the AUC

for doxorubicin was not significantly affected

(Sparreboom et al., 1999).

At least three potential mechanisms arethought to underlie the pharmacokinetic

interactions observed: (1) liver and renal Pgp

inhibition; (2) inhibition of drug-metabolizing

cytochrome P450s; and (3) impaired bile flow

The relative contribution of each of these

mech-anisms to the observed pharmacokinetic

inter-actions is not known However, an estimate of

the magnitude of the Pgp interaction can be

obtained by referring to studies of knockout

mice in which the murine orthologue of Pgp

has been deleted These studies have shown

that the absence of Pgp in the mouse results in

a delay in clearance of a number of

com-pounds, which is associated with an increase in

serum drug levels Thus, vinblastine levels

were increased 1.7-, 2.4-, 2.3-, and 2.1-fold in

plasma, liver, kidney and lung, respectively, in

Pgp-deficient mice (van Asperen et al., 1996).

Doxorubicin levels were only affected in the

liver, where they were 4.5-fold higher than in

the wild-type mice (van Asperen et al., 1999).

Given the relatively high levels of Pgp that are

normally found in the kidney and liver, these

results suggest that alternate mechanisms for

doxorubicin transport and/or metabolism

exist in these tissues Greater increases in

doxo-rubicin levels were observed in the central

nervous system (CNS) due to the absence of

Pgp in the endothelial cells in the brain;

how-ever, redundancy must exist in the human

blood–brain barrier, since no toxicity

attribut-able to increased CNS penetration of anticancer

agents has been observed in the clinical trials

with Pgp inhibitors

The cytochrome P450 (CYP) mixed-functionoxidases are a multigene family encoding

enzymes that play a critical role in the

metabo-lism of many drugs and xenobiotics PSC 833

and cyclosporin A can inhibit the metabolism

of numerous compounds that are substrates for

the CYP3A4 isoenzyme, thus contributing to

pharmacokinetic interactions (Relling, 1996)

Numerous anticancer agents, including

pacli-taxel, are substrates for this isoform of

cytochrome P450 (Kivisto et al., 1995; Wacher

impaired bile salt transport and persistent

intrahepatic cholestasis (Wang et al., 2001) Further, mutations in ABCB11/BSEP have been

found in patients afflicted with persistentfamilial intrahepatic cholestasis, type 2 (PFIC2)

(Strautnieks et al., 1998) Since bile salts are the

major driving force for bile flow, inhibition ofbile salt export results in cholestasis Hyper-bilirubinemia may be explained by reduced lev-els of MRP2 (ABCC2), which has been observed

in several forms of cholestasis (Kullak-Ublick

et al., 2000) (see Chapter 20) Drug excretion isalso impaired, requiring dose reduction of drugsexcreted primarily in the bile when adminis-tered to patients with cholestasis, includingdoxorubicin, vincristine and paclitaxel (Panday

et al., 1997; Rollins and Klaassen, 1979) Notably,

inhibition of BSEP by PSC 833 or cyclosporin Aresults in reduced bile salt transport, and a

reduction in bile flow (Bohme et al., 1993, 1994; Stieger et al., 2000) A reduction in oxidized

glutathione (GSSG) excretion into the bile wasalso observed, suggesting some impairment ofMRP2 function at the canalicular membrane

(Song et al., 1998), and potentially explaining

the hyperbilirubinemia observed followingadministration of PSC 833 or cyclosporin A.Thus, by inhibiting BSEP and reducing bileflow, PSC 833 and cyclosporin A may slowexcretion of drugs from the liver

TABLE18.5 CLINICAL RESULTS FROM PHASEI ANDII TRIALS WITH

SECOND- AND THIRD-GENERATIONP-GLYCOPROTEIN

ANTAGONISTS

PSC 833 trials

Phase I Mitoxantrone 44% 37 12 CR (32%) Advani et al., 1999

Poor risk AMLc Etoposide 58%

AraC

AraC Phase I Mitoxantrone 40% 30 15 CR (50%) Chauncey et al., 2000

Phase I/II Daunorubicin (72 h) 0 43 21 CR (49%) Dorr et al., 2001

Poor risk AMLc AraC Phase I Mitoxantrone 25% 23 6 CR (26%) Visani et al., 2001

Poor risk AMLc Etoposide 62.5%

Phase II Paclitaxel 60% 58 5 PR (8.6%) Fracasso et al., 2001

Ovarian carcinoma 3 h infusion

Phase I trials with third-generation agents

3 h infusion

Abbreviations: CR, complete response; PR, partial response; AML, acute myelogenous leukemia; N/A, not available.

aDose reduction required at the MTD, compared to MTD in the absence of antagonist.

bDose reduction relative to MTD for doxorubicin, 50 mg m⫺2; administered on a q 3-week schedule.

cPoor risk AML: includes variable proportions of patients with relapsed, refractory, or secondary AML.

dUnpublished data.

various combination chemotherapy regimens.

These included mitoxantrone and etoposide,

with or without cytosine arabinoside (AraC),

for acute leukemia; VAD (vinblastine and

dexa-methasone) for myeloma; and paclitaxel and

cisplatin for ovarian cancer Results from

ongo-ing or recently completed randomized trials are

not yet available The dose reductions required

in each trial are shown; for doxorubicin, the

results are calculated relative to a dose of

50 mg m⫺2 on a 3-weekly schedule However,

single agent doxorubicin has been administered

at doses as high as 80 mg m⫺2 every 3 weeks

(Edmonson et al., 1993).

High response rates are found only in theAML trials These studies were typically

undertaken in poor risk populations, including

patients with relapsed or refractory leukemia,

or elderly patients with secondary leukemia

Nearly 50% of patients on the trial combining

PSC 833 with daunorubicin and AraC

experi-enced a complete response (Dorr et al., 2001).

These same investigators had noted a complete

response rate of 69% with the same regimen

combined with cyclosporin A (List et al., 1993).

No dose reductions were made in this trial, and

pharmacokinetic studies in the PSC 833 trial

revealed that one-half of patients had no

detectable pharmacokinetic interaction (Dorr

et al., 2001) The authors concluded that

sys-tematic dose reductions would potentially

have led to undertreatment of half of enrolled

patients Perhaps tellingly, response rates were

lower on two trials with mitoxantrone,

etopo-side and AraC (26% and 32%), where dose

reductions were required to prevent severe

tox-icity (Advani et al., 1999; Visani et al., 2001),

although heterogeneity in AML subtypes may

have contributed to these differences as well

Findings with paclitaxel are also illustrative

Dose reduction of paclitaxel was required

whether administered as a 3-hour or a 96-hour

infusion (Chico et al., 2001; Fracasso et al., 2001).

In our study combining a 96-hour infusion of

paclitaxel with PSC 833, both clinical evidence

and pharmacokinetic studies suggested that

third of patients were undertreated,

one-third were overtreated, and only one-one-third of

patients had appropriate doses of paclitaxel

when administered at the maximum tolerated

dose determined in combination with PSC 833

(Chico et al., 2001) In the phase I trial with a

3-hour infusion of paclitaxel, reduced doses were

given to allow equal toxicity following addition

of PSC 833, and it was assumed that the AUCs

would be comparable to AUCs in the absence of

PSC 833 However, the AUCs were reduced by

an average of 41% (range 24–59%) in patientsreceiving 30–50% dose reductions of paclitaxel

(Fracasso et al., 2000) In light of these

observa-tions, the 8.6% response rate observed withpaclitaxel plus PSC 833 in refractory ovariancancer may be significant, given that the

70 mg m⫺2 dose administered every 3 weeksrepresented a 60% dose reduction from the

standard dose (Fracasso et al., 2001).

From one perspective, it could be argued thatthe addition of a Pgp inhibitor to a combinationchemotherapy regimen could not be expected

to have a large impact, particularly on responserates Indeed, both the effect of Pgp expression

on clinical outcome and the impact of PSC 833have been measured in the penumbra ofchemotherapy combinations that frequentlyinclude potent additional agents For example,Pgp inhibitors can only be expected to have animpact on the 30–50% of leukemias expressingPgp, and can only be expected to enhance the contribution of the anthracycline to theclinical response The same can be said for theovarian cancer trials The Pgp expression stud-ies reported thus far suggest that fewer than20% of ovarian cancers express this transporter.Thus, when a Pgp inhibitor is added to a regi-men combining paclitaxel with cisplatin, theincremental benefit provided by the inhibitor

to the combination would only be a fraction of

a fraction, and thus could only be detected in arandomized trial encompassing large numbers

of patients Furthermore, benefit may only beseen in survival analyses, if the main role of theinhibitor is to prevent the emergence of a resis-tant clone

Hindsight is, of course, 20/20 Given the likelihood that any benefit of Pgp inhibitionwas lost in the dose reductions required in thePSC 833 trials, a defensible conclusion can bereached that Pgp inhibition has not yet beenadequately tested These trials, as in the first-generation inhibitor studies, provided valuableinsight, including convincing evidence thatPgp could be inhibited in patients Asdescribed below, surrogate assays emergedduring the course of these trials, confirmingincreased drug retention in Pgp-bearing nor-mal tissues, and in some tumors

SURROGATE ASSAYS

In assessing the outcome of drug resistancereversal trials, it became apparent that assays