Multidrug resistance in brain tumors: roles of the blood-brain barrier.

Abstract

Malignant brain tumors and brain metastases present a formidable clinical challenge against which no significant advances have been made over the last decade. Multidrug resistance (MDR) is one of the main factors in the failure of chemotherapy against central nervous system tumors. The MDR1 gene encoding P-glycoprotein (P-gp), a drug efflux pump which plays a significant role in modulating MDR in a wide variety of human cancers, is highly expressed in the blood-brain barrier (BBB). The BBB controls central nervous system exposure to many endogenous and exogenous substances. The exact molecular mechanisms by which the BBB is involved in the resistance of brain tumors to chemotherapy remain to be identified. The purpose of this review is to summarize reports demonstrating that P-gp, one of the most phenotypically important markers of the BBB, is present in primary brain tumors and thus plays a crucial role in their clinical resistance to chemotherapy.

Full text

Page 1

Cancer and Metastasis Reviews 20: 13–25, 2001.
© 2001 Kluwer Academic Publishers. Printed in the Netherlands.
Multidrug resistance in brain tumors: Roles of the blood–brain barrier
Anthony R´ egina1, Michel Demeule1, Alain Laplante1, Julie Jodoin1, Claude Dagenais1, France Berthelet2,
Albert Moghrabi1and Richard B´ eliveau1
1Laboratoire de M´ edecine Mol´ eculaire, Hˆ opital Sainte-Justine-Universit´ e du Qu´ ebec ` a Montr´ eal, C.P. 8888,
Succursale Centre-ville;2D´ epartement de Pathologie, Hˆ opital Notre-Dame, Montr´ eal, Qu´ ebec, Canada
Key words: blood–brain barrier, brain tumors, multidrug resistance, P-glycoprotein, endothelial cells
Abstract
Malignant brain tumors and brain metastases present a formidable clinical challenge against which no significant
advanceshavebeenmadeoverthelastdecade.Multidrugresistance(MDR)isoneofthemainfactorsinthefailureof
chemotherapyagainstcentralnervoussystemtumors.TheMDR1geneencodingP-glycoprotein(P-gp),adrugefflux
pump which plays a significant role in modulating MDR in a wide variety of human cancers, is highly expressed
in the blood–brain barrier (BBB). The BBB controls central nervous system exposure to many endogenous and
exogenous substances. The exact molecular mechanisms by which the BBB is involved in the resistance of brain
tumors to chemotherapy remain to be identified.
The purpose of this review is to summarize reports demonstrating that P-gp, one of the most phenotypically
important markers of the BBB, is present in primary brain tumors and thus plays a crucial role in their clinical
resistance to chemotherapy.
1. Introduction
The treatment of brain cancer is one of the most chal-
lenging areas of oncology. Brain cancer, including pri-
mary and metastatic brain tumors, account for over
120,000 new patients each year in the United States.
Braintumors,andparticularlyglioblastomamultiform,
have retained their dismal prognoses despite advances
in neurosurgical techniques and in radiation and drug
therapies. One of the difficulties encountered is the
anatomical localization of the tumor, exacerbated by
the invasion of surrounding tissue. This often results
in incomplete surgical resection of the tumor, with the
tumor usually recurring within a few centimeters of
the margins of the resection. Radiotherapy is limited
by low brain tolerance as well as by the infiltration of
tumor cells into healthy brain. Adjuvant chemotherapy
isthusessentialforthetreatmentofsometypesofbrain
tumors. However, another major problem which lim-
itstheefficacyofconventionalbraintumorchemother-
apyisthephenomenonofmultidrugresistance(MDR).
MDR is generally described as an acquired or intrinsic
resistance to a variety of structurally unrelated syn-
thetic and natural cytotoxins. Many primary brain
tumors are intrinsically chemoresistant; limited pene-
tration of cytotoxic drugs across the blood–brain bar-
rier (BBB) may also contribute to the poor response
of brain tumors to chemotherapy. Brain metastases
respond poorly to chemotherapy, which may be due
to differences in chemosensitivity between primary
and metastatic tumors. One reason for this is that in
metastases of previously treated tumors, clonal selec-
tion of chemoresistant tumor cells may already have
taken place. Small-cell lung cancer (SCLC) provides
an example of this, where the response rate of intra-
cerebral metastases to cyclophosphamide, etoposide
and vincristine without radiotherapy is 53% as com-
pared to a 79% response rate in the primary tumor [1].
A growing body of evidence links P-glycoprotein
(P-gp) expression in untreated tumors to a poor prog-
nosis for chemotherapy with P-gp substrate drugs.
Nevertheless, while the connection has been firmly
established for several hematological cancers, it
remains controversial for most solid tumors. This

Page 2

14
review will focus on the possible role of the BBB and,
more specifically, on the multidrug transporter P-gp
in clinical resistance to chemotherapy against brain
tumorsandwilldiscusscurrentandfutureperspectives
in brain tumor therapeutics.
2. The BBB: cytoarchitecture and
physicochemical properties
The BBB and the blood–cerebrospinal fluid barrier
(BCSFB) are interfaces that separate the systemic cir-
culation from the brain parenchyma and cerebrospinal
fluid,respectively.Thesebarriershelptomaintainbrain
homeostasis not only by regulating the entry of blood-
borne substances into the CNS, but also by restrict-
ing the access of potential toxins to brain parenchyma.
As a result of the latter property, some drugs do not
accumulate within the brain to a concentration which
permits them to exert their desired pharmacological
effects.
The surface area of the BBB is estimated to be
5000-fold larger than that of the BCSFB and there-
fore the BBB is considered to be the main route for
the uptake of endogenous and exogenous ligands into
the brain parenchyma [2]. Brain microvessel endothe-
lial cells are responsible for the morphological and
functional characteristics of the BBB. As compared
to peripheral endothelial cells, brain endothelial cells
are sealed by continuous tight junctions and no fen-
estrations are present. Furthermore, brain endothelial
cells exhibit very low pinocytic activity. As a result,
solutes must cross the luminal membrane, the cyto-
plasmandtheabluminalmembraneofbrainendothelial
cells to gain access to the brain. Hence, carrier systems
are essential to shuttle nutrients (e.g., glucose, amino
acids), vitamins, hormones, monocarboxylic acids and
amines to the brain [3]. For large proteins with impor-
tant physiological functions (e.g., insulin and trans-
ferrin),receptor-mediatedendocytosisandtranscytosis
can mediate transport across the BBB [4].
The determinants of passive drug entry into the
CNS have been recently reviewed [5]. Due to the
physical nature of cell membranes, lipid solubility
is an important determinant of passive BBB perme-
ability. The ability of a compound to partition into
an organic solvent relative to a physiologic buffer
(e.g., octanol/buffer partition coefficient) is a com-
mon measure of lipophilicity. Nevertheless, the overall
hydrophilic/lipophilic balance of a molecule appears
to be a better predictor of BBB permeability than the
octanol/water partition coefficient. Steady-state brain
distributioninvivoisnotonlyafunctionofpassivedif-
fusion across the BBB, but depends also on the relative
affinity for plasma proteins and brain tissue. Ionization
at physiologic pH (pKa), the affinity and capacity of
transport systems, potential BBB/cerebral metabolism
and egress through the CSF route are also important
factors.Furthermore,thereareanumberofactiveefflux
transportmechanismspresentintheCNS,locatedboth
in the BBB and in the choroid plexus [6]. The efflux of
manyendogenoussubstancesandxenobioticsbytrans-
port systems reduces their effective penetration from
blood to brain, which ultimately decreases brain expo-
sure to these compounds. In this line of research, much
attention has been focused on the multidrug trans-
porters such as P-gp and multidrug-resistance associ-
ated protein (MRP) (Figure 1).
3. P-glycoprotein
3.1. Expression of P-gp in the BBB and
other tissues
P-gpisamembranetransporterpresentintheBBBthat
belongs to the ATP binding cassette (ABC) superfam-
ily [7]. It was initially discovered due to its expression
in different types of MDR tumors [8,9]. It was shown
to confer the MDR phenotype [10], by which a can-
cer cell exposed to a single anticancer drug becomes
simultaneously resistant both to that drug and to other
drugs of unrelated structure or function. P-gp is the
product of the MDR1 and MDR2 or MDR3 genes in
humans, as well as the mdr1a, mdr2 and mdr1b genes
in rodents. In humans, the MDR phenotype is caused
by the MDR1 gene, whereas it is caused by mdr1a and
mdr1b in rodents [11,12].
P-gpexpressionwasfirstdetectedinthehumanbrain
by in situ hybridization [13], and later by immunocy-
tochemistry using different monoclonal antibodies. It
was also established at that time that P-gp is present in
human tissues such as liver, kidney, intestine, adrenal
glandsandblood–tissuebarrierssuchasplacenta,testis
capillaries and brain capillaries [14,15].
3.2. Expression of P-gp in brain endothelial cells
The P-gp isoforms expressed in mouse brain capil-
laries were determined by Western immunoblotting
using antibodies specific to each isoform. The mdr1a

Page 3

15
Figure 1. Schematic view of a brain capillary. Endothelial cells are sealed by continous tight junctions and surrounded by a basal lamina.
Pericytes are present at the periphery of the vessel. Astrocyte foot processes are in close contact. Hydrophilic molecules cannot diffuse
through the plasma membrane, so specific transporters present on both the luminal and the abluminal membranes and carrier-mediated
endocytosis allow the uptake of nutrients and hormones (A,B). A very small amount of water-soluble molecules use the paracellular
pathwaybydiffusingthroughthetightjunctions(D).Effluxpumps(P-gp,MRPs)impedethebrainpenetrationoflipophilicsubstrates(E).
encoding isoform is expressed in brain capillaries iso-
latedfrommouse,whereasthemdr1bandmdr2encod-
ing isoforms are not observed there [16]. Analysis of
mdr RNA by RT-PCR showed that mdr1a and mdr1b
are both present in rat brain, where mdr1a was specif-
ically detected in brain capillaries and mdr1b in brain
parenchyma [17].
Recently published data from our laboratory
showed that P-gp is predominantly expressed in brain
endothelial cells [18]. Endothelial cells were iso-
lated from brain and other tissues using magnetic
microbeadscross-linkedtoanantibodydirectedagainst
the platelet-endothelial cell adhesion molecule-1
(PECAM-1). The endothelial cells were immobilized
in a cell separation column mounted onto a magnet,
while the other cells were eluted (negative fraction).
The enrichment of endothelial cells in the preparation
was about 68-fold, assayed by immunodetection of
endothelialnitricoxidesynthase(eNOS)inthepositive
andnegativefractions[18].P-gpwasstronglyenriched
(59-fold) in this endothelial cell fraction from brain
and was absent from the negative fraction, in which
the glial fibrillary acidic protein (GFAP), an astrocytic
marker, was present. It was also shown in this study
that the P-gp isoform found in brain endothelial cells
is encoded by mdr1a [18].

Page 4

16
3.3. Subcellular localization of P-gp in
brain endothelial cells
The data above are in agreement with a previous study
that showed that P-gp was strongly enriched in the
luminal membranes of rodent brain vascular endothe-
lium[19].Thiswasdoneusingamethodthatallowsthe
isolation of luminal membranes of blood vessels. P-gp
in luminal membranes from brain, and from other tis-
sues, was immunodetected with monoclonal antibody
C219 and polyclonal antibody Ab-1. This procedure
provided a 17-fold enrichment of P-gp compared with
isolated brain capillaries and a 400–500-fold enrich-
mentcomparedwithmembranesfromthewholebrain.
Intheseisolatedendothelialcells,GFAPshowedavery
weak enrichment (1.4-fold over brain capillaries) indi-
cating minimal contamination by astrocytes. Studies
using other techniques such as immunoelectron and
confocal microscopy also indicated that P-gp is local-
ized at the luminal side of the BBB endothelium rather
than the abluminal side [20,21]. Taken together, these
results support the finding that brain endothelial cell
P-gp is mainly located in the luminal membranes.
This model of luminal localization of P-gp in
endothelial cells has been challenged [22] by an alter-
native model contending that immunoreactive P-gp is
located on astrocyte foot processes. While this postu-
late is interesting, it contradicts the findings of many
other studies demonstrating P-gp localization on the
luminal side of human brain endothelium [13,20,21].
Clearly, further experiments are needed to confirm the
cellular localization of P-gp at the human BBB. One
possibleexplanationfortheseconflictingdatacouldbe
a difference between species.
In addition to being found in the luminal membrane
of brain endothelial cells, P-gp expression has been
also reported in caveolae [23–25]. Caveolae are flask-
shaped plasma membrane invaginations involved in
many cellular events, such as signal transduction, lipid
and protein sorting, endocytosis and potocytosis [26].
In capillary endothelial cells, caveolae play a role in
the transport of macromolecules across the cells by
transcytosis [27]. We have observed that P-gp is local-
ized in the caveolae of rat brain capillaries. Caveolae
wereisolatedbyflotationoflow-densitymicrodomains
on a sucrose gradient, producing a subcellular frac-
tion where caveolin-1, the structural protein of cave-
olae, was enriched. P-gp contains a caveolin-binding
motif that is present within proteins reported to
interact with caveolin. The caveolin-binding motif
mediates the association of caveolin-binding proteins
with the scaffolding domain of caveolin [28]. We
showed, by co-immunoprecipitation studies, that P-gp
interacts with caveolin-1 [25]. Association of P-gp
with caveolin-1 could enable P-gp to move between
the plasma membrane and caveolae. The scaffolding
domain of caveolin-1 is involved in interactions with
numerous proteins and it has been reported to regu-
late the signaling molecules localized in caveolae such
as eNOS, protein kinase C, epidermal growth factor
receptor and the insulin receptor [29]. The effect of
caveolin-1 on P-gp activity is still unknown and the
function of P-gp in caveolae of endothelial cells of the
BBB remains to be determined.
3.4. Functions of the P-gp in the BBB
In brain capillaries, P-gp plays an important role in
preventing many hydrophobic molecules (drugs or
other substrates) from crossing the BBB and reach-
ing the CNS. An increase in the brain concentrations
of P-gp substrates was measured in mdr1a knock-out
mice, which lack P-gp in the BBB. These mice were
50–100timesmoresensitivetothepesticideivermectin
than were wild-type mice. The accumulation of this
neurotoxic drug in brains of mdr1a(−/−) mice was
increased 80–100-fold compared to wild-type mice
[30]. Similar results were obtained with mdr1a knock-
outmiceexposedtodigoxin,ondansetron,loperamide,
paclitaxel, vinblastine and doxorubicin.
4. Multidrug-resistance associated protein
A second protein expressed in many MDR tumors
is MRP1. Like P-gp, MRP1 is a member of the
ABC superfamily and is expressed in many MDR
cell lines which do not express P-gp [31–33]. In
humans, seven MRP homologues have been identified
to date [34]. MRPs are distributed throughout most
human tissues [35]. MRP1 is expressed and localized
on the basal membrane of Sertoli cells in the testes
and on the basolateral membrane of lung epithelial
cells [36,37]. It has been demonstrated by Western
blot analysis that MRP1 is expressed in the enriched
membrane fraction of human and rat choroid plexus,
and MRP1 has been localized to the basal mem-
brane of primary cultured rat choroid plexus epithe-
lial cells [38]. The expression of MRP1 at the BBB
is still under debate. Western blot analysis and RT-
PCR suggest that MRP1 is expressed in isolated rat
brain capillaries, primary cultured rat, porcine and

Page 5

17
bovine brain endothelial cells, and immortalized rat
and mouse brain endothelial cells [17,39,40]. How-
ever, MRP1 expression is greater in rat and porcine
brainhomogenateandprimaryratandporcinecultures
than in isolated brain capillaries [17,41]. No expres-
sion of MRP1 was detected in isolated human brain
capillaries by immunohistochemistry whereas expres-
sion of both P-gp and platelet-endothelial cell adhe-
sion molecule-1 (PECAM-1, an endothelial marker)
was clearly detected [42]. Recently, the expression of
MRP1, 4, 5 and 6 was demonstrated in both primary
cultured bovine brain endothelial cells and a capillary-
enriched fraction. Low levels of MRP3 were detected
in cultured cells but not in the capillary-enriched
fraction [43].
MRPsareorganicaniontransporterswhichtransport
anionic drugs and neutral drugs conjugated to acidic
ligands such as glutathione (GSH), glucuronate or sul-
fate.However,MRP1,MRP2andMRP3canalsocause
resistance to neutral organic drugs that are not known
to be conjugated to acidic ligands, such as doxoru-
bicin,daunorubicin,epirubicin,vincristine,vinblastine
and etoposide, by transporting these drugs together
with free GSH [44]. Recently, it has been demon-
strated that MRP1, MRP2 and MRP3 are also involved
in resistance to methotrexate, which is an important
chemotherapeutic agent in the treatment of malignan-
cies that disseminate to the brain [45,46].
5. The BBB in brain tumors
Brain tumors can be divided into three groups accord-
ingtothedevelopmentoftheirvasculature:(1)primary
tumors that develop in situ and derive their blood sup-
plyfromneuralvessels;(2)secondarytumorsthatarise
elsewhere in the body and metastasize to the brain and
also derive their vasculature from neural vessels; and
(3) meningiomas, which arise from extra-axial tissues
within the cranium or spinal canal and which initially
derive their vascular supply from the meninges.
5.1. Evidence for BBB disruption
The primary evidence against a major role of the BBB
in the low efficacy of chemotherapy is the increased
microvascular permeability in gliomas that leads to
brain edema. Indeed, in primary and metastatic brain
tumors a number of alterations in the brain capillary
ultrastructure have been described [47]. Interendothe-
lial junctions in brain tumor vessels are abnormal.
A relationship between suppression of the tight-
junction associated protein claudin-1 and the alter-
ationoftight-junctionmorphologyinvesselsofhuman
glioblastomas has been shown recently [48]. Fenestra-
tions have been described in primary non-glial tumor
vessels and are common in most if not all metastatic
tumor vessels. An increase in the number of pinocytic
vesicles has also been reported [49].
5.2. Evidence for BBB integrity
The most obvious evidence of BBB integrity is the
low concentrations found for most chemotherapeutic
agents in primary brain tumors and brain metastases.
Some ultrastructural studies showed that, even in high-
grade tumors, the neo-vasculature preserved partial
BBBfunctionatthecellularlevel.Arecentstudyfound
that ZO-1, a protein associated with tight junctions,
was present in the endothelial cells of microvessels
in all astrocytic tumors studied [50]. Fenestrations are
extremely rare in glial tumor vessels, even in glioblas-
toma multiforme [49]. Brain tumors grow by increas-
ing tumor mass and by infiltrating surrounding normal
brain. Changes in the ultrastructure and morphology
of peritumoral capillaries have been investigated [51].
In this study, peritumoral capillaries presented struc-
turally normal tight junctions. No differences were
observed between normal and peritumoral capillaries
with respect to diameter, wall thickness, endothelial
thickness or endothelial vesicle density.
In conclusion, the specific morphological character-
istics of the BBB are maintained at the cellular level.
6. P-gp expression in brain tumors
6.1. Primary brain tumors
6.1.1. Whole tumors
The exact mechanism by which brain tumors resist
chemotherapy remains unknown. One of the main
arguments contradicting involvement of the BBB in
resistance to chemotherapy is the lack of correlation
observed between the expression level of P-gp and
the resistance levels of some tumors [52]. However, a
recent report shows that even low levels of P-gp and
MRP1 expression, which may be difficult to detect in
tumors, can significantly affect their sensitivity to a
widerangeofclinicallyimportantP-gpsubstratedrugs
[53]. This suggests that P-gp expression, even at a low