Address correspondence and reprint requests to Dr. S.M. Sisodiya at Department of Clinical and Experimental Epilepsy, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, U.K. E-mail: email@example.com
Summary: Purpose: The molecular basis of drug resistance in epilepsy is being explored. Two proteins associated with drug resistance in cancer, P-glycoprotein and multidrug resistance–associated protein 1, are upregulated in human epileptogenic pathologies. Other proteins associated with resistance in cancer include major vault protein (MVP) and breast cancer resistance protein (BCRP). We hypothesized that these proteins would also be upregulated in human epileptogenic pathologies.
Methods: Hippocampal sclerosis (HS), focal cortical dysplasia (FCD), and dysembryoplastic neuroepithelial tumor (DNT) were studied by using immunohistochemistry for MVP and BCRP. Nonepileptogenic control and histologically normal brain adjacent to epileptogenic tissue were used for comparison.
Results: MVP and BCRP were expressed ubiquitously in brain capillary endothelium. Ectopic upregulation of MVP was seen in hilar neurons in HS, dysplastic neurons in FCD, and lesional neurons in DNT. Only in HS cases were rare extralesional neurons immunoreactive. Glial upregulation was not seen. There was no qualitative upregulation of BCRP.
Conclusions: These results show that more than one resistance protein may be upregulated in a given epileptogenic pathology and may contribute to drug resistance. Determination of the types, amounts, and distribution of such proteins will be necessary for rational treatment for drug resistance in epilepsy.
Despite advances in the understanding, investigation, and therapy of epilepsy, about one third of patients with epilepsy remain refractory to drug treatment (1). The causes of resistance to antiepileptic drug (AED) treatment are not well understood. Surrogate markers of resistance, such as remote symptomatic etiology, have been identified (2–4). More recently, attention has focused on the possible mechanisms of drug resistance, drawing on parallels with drug resistance in cancer. There is now a body of evidence that certain drug transporter molecules, known to mediate multidrug resistance (MDR) in cancer, may also contribute to clinical drug resistance in epilepsy. In human or animal epileptogenic brain tissue, increased encoding messenger RNA (mRNA) or protein quantity and abnormal location of at least three such transporters, P-glycoprotein (P-gp, also known as MDR1 P-glycoprotein or ABCB1), multidrug resistance–associated protein 1 (MRP1, also known as ABCC1), and multidrug resistance–associated protein 2 (MRP2, also known as ABCC2), have been shown (5–11). These molecules have the capacity to transport a variety of AEDs, including phenytoin (PHT) (5,12), phenobarbitone (PB) (13), topiramate (TPM) (14), valproate (VPA) (15), lamotrigine (LTG), and felbamate (FBM) (16), and more controversially, carbamazepine (CBZ) (12,17). Increased transcription of P-gp–encoding mRNA occurs in response to seizures in animal models, with parallel increases in protein expression and AED transport (12). An association is found between a polymorphism in the gene encoding P-gp and clinical drug resistance (18). It seems likely, therefore, that these molecules contribute to MDR in human epilepsy. P-gp, MRP1, and MRP2 belong to the ABC [adenosine triphosphate (ATP)-binding cassette] superfamily of proteins. Another member of this family, breast cancer resistance protein (BCRP, also known as MXR or ABCG2), has also been shown in diverse cell lines in vitro to be associated with transport of, and resistance to, anticancer drugs (19–25). High levels of BCRP expression have been noted in a wide variety of untreated human solid tumors (26). Relatively high levels of BCRP expression have been associated with poor prognosis in childhood acute myeloid leukemia (27).
Cancer studies have identified other cellular components associated with MDR. Among these are intracellular ribonucleoprotein particles called vaults. Vaults are barrel-shaped, ∼35 × 65 nm in dimension, and so named because their structure resembles cathedral roof vaults (28). Cells that possess vaults may have ≤100,000 copies/cell. Vaults are highly conserved across the evolutionary spectrum in terms of components, size, and shape. Each ovoid vault has thin solid walls but is hollow and can split into two halves across an invaginated waist; each half can unfurl into a flower-like structure with eight petals. Vaults are the largest ribonucleoprotein complexes known, with a mass of 13 megadaltons. In mammals, vaults are composed of multiple copies of three proteins and a unique untranslated RNA species (vault RNA, vRNA). Of the three proteins, a 104-kDa component constitutes 70% of vault mass: this is called major vault protein (MVP).
MVP is identical to lung resistance–related protein, or LRP (29). LRP was first associated with drug resistance in doxorubicin-resistant non–small cell lung carcinoma cells (30). Although the precise mechanism of vault-mediated drug resistance is unknown, several studies have shown that MVP upregulation is an excellent marker and predictor of MDR in both cancer cell lines and a broad variety of clinical tumors (28,31).
We proposed that BCRP and MVP upregulation in cells not normally expressing these proteins would be associated with the MDR phenotype in human epilepsy, and tested this hypothesis by using immunohistochemistry for BCRP and MVP in control specimens and surgically resected human brain material from patients with common causes of refractory epilepsy.
The study was approved by the Joint Research Ethics Committee of the Institute of Neurology and the National Hospital for Neurology and Neurosurgery. We studied formalin-fixed paraffin-embedded human brain tissue from neuropathological archives. All case tissue (n = 20) was from therapeutic surgical resection for epilepsy refractory to multiple drug treatment and was pathologic tissue surplus to diagnostic requirements. All samples were made anonymous. Routine staining (Nissl, hematoxylin and eosin) was performed to confirm the histologic diagnosis and to provide anatomic detail.
Disease surgical resection specimens had ideal fixation conditions, with no significant preresection hypoxia, being immersed in formalin immediately, and embedded within a week. Control tissue was of three types: (a) histologically normal brain tissue (n = 5; two frontal, three temporal samples) from surgical resections for trauma or decompression for increased intracranial pressure from five individuals with no history of epilepsy or exposure to AEDs; (b) histologically normal brain tissue (n = 20) within a resection specimen containing focal epileptogenic pathology (or pathologies); this normal tissue allowed control for age, sex, region of brain, exposure to AEDs, direct and indirect effects of seizures, and tissue preservation; (c) positive control tissue was surplus normal postmortem human lung, known to express MVP (32,33), or normal placenta or colon, known to express BCRP (34). All disease cases had been exposed to multiple (at least three) AEDs.
Tissue preparation, fixation, and preservation were the same in all cases. Five cases of hippocampal sclerosis (HS) were studied. All showed the typical patterns of cell loss involving CA1, with varying degrees of loss from CA3 and CA4 and relative sparing of CA2 sector. In all cases, additional lateral temporal neocortex was also studied. In an additional five cases, HS was associated with a resected second pathology (contusion, old infarct, Rasmussen encephalitis, ganglioglioma grade I, and meningioangiomatosis). Five cases of focal cortical dysplasia (FCD) were examined. All had adjacent normal cortical tissue for comparison. All cases had previously been characterized histologically by the presence of large dysplastic neurons (highlighted with silver and neurofilament stains), disordered laminar architecture, and variable numbers of balloon cell glia, often located in layer I or the deeper regions of the cortex and the underlying white matter. Four cases of dysembryoplastic neuroepithelial tumor (DNT) were examined, all of which were large excisions. Each case showed typical features of DNT with intracortical nodules of glial, neuronal, and oligodendrocyte-like cells, and with separate glioneuronal elements. Adjacent normal-appearing cortex was available in all four DNT cases. A single case of isolated ganglioglioma also was examined, with adjacent normal cortex. To confirm results for MVP, a second antibody against a separate epitope of MVP was used (see later) in some of the cases: five controls, two HS cases, five dual-pathology cases, four FCD cases, four DNT cases, and the single ganglioglioma case.
For detection of MVP, two monoclonal antibodies, MVP-37 (33) and LRP (clone 42, isotype IgG1 from BD Biosciences Cowley, Oxford, England) were used. These are known to detect separate epitopes of MVP in paraffin sections, and are not known to cross-react with any other drug-resistance proteins (33). For BCRP, monoclonal antibody BXP-21 was used (34): BXP-21 is not known to cross-react with other drug-resistance proteins, and no other cross-reactivity has been reported for this antibody.
Five-micrometer sections were dewaxed and rehydrated in graded alcohol.
For MVP-37, endogenous peroxidase activity was blocked in 0.6% hydrogen peroxide and methanol for 15 min. Sections were then treated with 6 M guanidine HCl in 50 mM Tris HCl (pH 7.5) for 2 h at 4°C, followed by protein blocking with 10% normal swine serum for 20 min. Sections were then incubated with primary antibody (1:2,000 dilution in 1% bovine serum albumin–phosphate-buffered saline (PBS) with 0.05% Tween 20) overnight at 4°C.
For LRP antibody, endogenous peroxidase activity was blocked in 0.6% hydrogen peroxide and distilled water for 10 min. Sections were microwaved in citrate buffer for 10 min and cooled for 15 min. Protein blocking was achieved by immersing sections in 10% nonfat milk diluted in PBS for 30 minutes. Sections were then incubated with primary antibody LRP (1:2000 dilution in 1% bovine serum albumin-PBS with 0.05% Tween 20) overnight at 4°C.
For both antibodies, labeling was detected by using an LSAB+ kit (DAKO Corporation, Carpinteria, CA, U.S.A.). Staining was visualized with a NovaRed kit (Vector, Burlingame, CA, U.S.A.). Nuclei were counterstained with hematoxylin. The sections were dehydrated in alcohol, cleared in xylene, and coverslipped. Normal human postmortem lung tissue was used as a positive control. Negative controls were treated identically except that the primary antibody was replaced with normal mouse immunoglobulin G (IgG) for MVP-37 and IgG1 for LRP.
For BXP-21, sections were microwaved in 1 mM EDTA (pH 8) for 15 min and cooled for another 15 min. Protein blocking was carried out with 10% normal swine serum followed by blocking of endogenous biotin by using a biotin/avidin-blocking system. Sections were incubated with BXP-21 monoclonal antibody (1:1,000 dilution) overnight at 4°C. Labeling was detected by using LSAB+ Kit, and visualized with NovaRed as described earlier. For negative controls, primary antibody was replaced by normal mouse IgG.
All cases and controls were processed simultaneously in the same experiment. All cases were examined by six authors, including separately by three neuropathologists (M.T., P.vdV., B.N.H.). Blinding was not possible for neuropathology. Experimental sections were assessed only if positive and negative (obtained with omission of primary) controls reacted as predicted. No labeling was noted in negative controls sections labeled with an irrelevant primary (e.g., normal mouse IgG).
Immunolabeling for MVP
In normal lung tissue, bronchiolar epithelial labeling was observed as expected with both antibodies (Fig. 1A). In the five control brain sections, endothelial labeling in blood vessels was observed throughout each section (Fig. 1B). Labeling was not seen in any cases in normal cortical or white matter glia. Only occasional neurons in one control case were immunolabeled with either antibody (Fig. 1C). Three cases had evidence of acute neuronal injury: no such neurons were labeled with either antibody. No immunolabeling was observed in any negative control sections, processed with nonspecific mouse IgG in place of MVP-37 or LRP.
In all the following pathological cases, endothelial immunopositivity was observed in all sections, confirming that immunoreaction had not failed technically in any case. The absence of immunolabeling within particular cell populations in these sections thus reflected absence of protein or a level of protein expression below the limit for detection with our immunohistochemistry protocol. For MVP, positive immunolabeling always appeared cytoplasmic, either in cell body or processes, as illustrated in the figures and as previously documented for MVP (32,33). Within a given pathology, no large differences were observed in proportions of a given cell population labeled.
Focal cortical dysplasia
In these five cases, the dysplastic pathology was multifocal within the histologic section, areas of maximum cellular abnormality being surrounded by areas containing fewer or no giant or dysplastic neurons. In all five cases, strongly positive immunolabeling was seen in dysplastic neurons and their processes (Fig. 2A–D). Not all abnormal neurons were immunolabeled, even within a given locality (Fig. 2A–D). Some balloon cell glia also were positively labeled in cases where they were apparent (Fig. 2E). Notably, normal-appearing neurons within the dysplasia were not seen to label (Fig. 2A–D), and no normal neurons outside the dysplastic region were labeled. The same patterns of immunoreactivity were confirmed with the second anti-MVP antibody, LRP (Fig. 2B).
In all five pure HS cases, a proportion of remaining hilar neurons were immunolabeled with MVP-37 (Fig. 3A and B). In some cases, clear labeling of hilar neuronal processes was observed (Fig. 3B). Only occasional granule cells and rare hippocampal pyramidal neurons were immunopositive. Only very rare normal-appearing neurons in the associated extralesional lateral temporal neocortex were immunolabeled in one of these five cases (Fig. 3C). Gliosis was seen in all cases in the hippocampus on routine staining, but labeling in normal or reactive glia with MVP-37 was not seen in any case in either hippocampus or temporal cortex. Very occasional labeling of microglial cells was noted in one case.
In the hippocampal sections of all five cases with a second (dual) pathology, a proportion of remaining hilar neurons were immunolabeled, as in the pure HS cases. Also as for the pure HS cases, only occasional granule cells and rare hippocampal pyramidal neurons were immunopositive (Fig. 3D). In all these cases, the second pathology was in a different block from the HS, and histologically normal adjacent tissue was present. Endothelium was immunopositive in all sections, and was the only immunoreactive cell type in adjacent normal-appearing cortex. In the ganglioglioma dual-pathology case, a proportion of dysplastic neurons were immunolabeled; in the meningoangiomatosis case, intense labeling was seen in the abnormal meningoendothelial and angiomatous components; some neurons adjacent to an old infarct were labeled in one case (Fig. 3E), and some adjacent to an old contusion were labeled in another. In the case with stage 4 Rasmussen encephalitis, a number of immunopositive neurons were observed adjacent to devastated, chronically scarred cortex (Fig. 3F), and very occasional small aggregates of microglial cells labeled. The same patterns of immunoreactivity were observed with LRP antibody (shown for hippocampus in dual pathology cases; Fig. 3G and H).
Dysembryoplastic neuroepithelial tumors and ganglioglioma
In only two of four cases was nonendothelial immunopositivity noted, and this was only within the lesion. In these cases, only lesional neurons (Fig. 4A and B) were immunopositive for MVP; oligodendroglia-like and glial cells did not appear to label (Fig. 4A and B). No immunolabeling other than of endothelial cells was seen in adjacent normal tissue, examined in all four cases.
A proportion of dysplastic neurons in the ganglioglioma were strongly immunopositive (Fig. 4C), whereas in adjacent cortex, immunolabeling was seen only in endothelium.
The same patterns of immunoreactivity were observed with LRP antibody for both pathologies (not illustrated for DNT; ganglioglioma, Fig. 4D).
Immunolabeling for BCRP
For this antigen, positive immunoreaction was obtained in positive controls as expected. In the five control brain sections, endothelial labeling in blood vessels was observed throughout each section. Labeling was not seen in any cases in cortical or white-matter glia. Three cases had evidence of acute neuronal injury: no such neurons were labeled. No immunolabeling was observed in any negative control sections, processed with nonspecific mouse IgG in place of BXP-21. In all the disease cases, strong immunolabeling was noted in capillary endothelium, without obvious differences between disease groups. No immunolabeling of normal-appearing or reactive glia, or normal-appearing or dysplastic neurons was noted in any disease cases. As no obvious differences were observed between cases and controls, these findings are not illustrated.
We have shown that MVP and BCRP are detectable by routine immunohistochemistry in brain capillary endothelium from both normal brain tissue and brain tissue containing a variety of pathologies. As expected, labeling for BCRP is seen in capillary endothelium, and appears to be membranous (26,34); brain capillary endothelial expression has recently been reported (35). For MVP, as expected, the observed positive immunolabeling is apparently cytoplasmic (28,33). This is the first report of the detection of MVP in normal human cerebral parenchyma, rather than in brain-derived cell lines, or cerebellar Purkinje cells (31). The endothelial localization of both proteins is compatible with a role in host defence and the generation of privileged compartments, although the proteins could contribute to direct barrier functions only if they were appropriately located in the cell membrane. BCRP and MVP (and vaults) have previously been demonstrated in greatest concentration in normal tissues most exposed to xenotoxins (31).
For BCRP, there is no evidence for ectopic upregulation. Thus we did not find any immunoreactivity in reactive or normal-appearing glia, or in dysplastic or normal-appearing neurons in epileptogenic or adjacent brain tissue. Expression was noted in capillary endothelium, but with our methods, we could not demonstrate definite upregulation in disease cases in comparison to control cases or adjacent normal tissue: immunoreaction was strong in all cases. This does not exclude such tissue-restricted upregulation, but may require other methods for its demonstration.
We have shown upregulation of MVP in epileptogenic pathologies from cases of drug-resistant epilepsy in comparison to histologically normal nonepileptogenic brain tissue obtained from surgical resection undertaken for other reasons, or normal-appearing cortex adjacent to epileptogenic pathology and within the same section. Completely normal human brain tissue is impossible to obtain in a comparable setting, and this applies equally to normal hippocampus. We consider adjacent tissue as ideal for comparison, as it has undergone identical stress in vivo, and identical treatment from resection onward (8). We noted identical patterns of immunoreactivity with both anti-MVP antibodies used: as these antibodies were chosen because they react against separate MVP epitopes, we suggest that the results truly reflect the presence of MVP. In the rest of the discussion, the results for both antibodies are considered together.
Upregulation is seen in cells within a variety of different pathologies (HS, dual pathology, FCD, DNT, ganglioglioma) that cause refractory epilepsy in diverse ways. In all these conditions, upregulation is seen in lesional neuronal populations, both normal-appearing neurons in HS and DNT and dysplastic neurons in FCD, and in FCD balloon cell glia, a less-differentiated cell population. These findings suggest upregulation is not simply nonspecific. For a given cell type upregulating MVP in a given pathology, only a proportion of that cell population is immunopositive, again arguing against a nonspecific upregulation. No normal glia were immunopositive, in normal or abnormal brain tissue, in contrast to an in vitro study of cultured cell lines, in which MVP was detected by using immunofluorescence in cultured normal human astrocytes (36). This discrepancy may be due to methodologic differences, and because we have studied native astrocytes in situ, whereas the study of Berger et al. (36) used cultured astrocytes that had been passaged for an unspecified number of cycles in vitro (36). In appropriate circumstances, MVP upregulation can occur after a single cell passage (37).
We have not directly shown an upregulation of vaults, the ribonucleoprotein complex presumed to be the biologically active superstructure containing MVP. However, MVP upregulation is mirrored by an upregulation in minor vault proteins (38) and association of constituent proteins and vRNA into complete vaults (39). Upregulation of MVP per se, as a marker of increased vault numbers, is associated in vitro with demonstrable MDR (36,37,40–42). Absolute vault numbers, and by implication MVP levels, appear to relate directly to the degree of drug resistance in cell lines (39).
Studies of clinical tumors have shown a correlation between MVP upregulation and MDR phenotype across a wide variety of cancers, including acute myeloid leukemia, multiple myeloma, childhood acute lymphoblastic leukemia, lung cancer, neuroblastoma, and ovarian carcinoma (28). The associations are clear, although the underlying mechanism(s) are not, and it may be that MVP/vault upregulation alone is insufficient to generate MDR (38). Knockout mice lacking MVP do not demonstrate increased sensitivity to cytostatics, but this does not exclude a role for MVP in drug resistance in cancer cells or refractory epilepsy (43). Multiple drug resistance in cancer is often multifactorial. However, it is known that MVP upregulation is not simply a marker of MDR mediated by P-gp or MRP1 (28,44), resistance-mediating molecules previously shown to be upregulated in epileptogenic pathologies (10). Indeed, in the few comparative studies undertaken, MVP upregulation was a superior prognostic predictor compared with P-gp or MRP1 (44). It is worth noting that P-gp upregulation was seen only in glia, whereas neuronal MRP1 upregulation was seen only in FCD and not in HS or DNT (10).
We do not know the cause of MVP regulation in epilepsy. However, MVP upregulation in lesional versus extralesional cortex suggests that the upregulation is linked to the epileptogenic tissue specifically. This is seen in the dual-pathology cases and best demonstrated in the FCD cases: disparate foci of dysplasia showed MVP upregulation, whereas intervening or adjacent cortex did not. Because adjacent normal cortex does not show upregulation, we do not think that MVP upregulation is an epiphenomenon reflecting AED exposure; cellular stress such as ultraviolet exposure (45) or heat shock (37) do not cause MVP upregulation in vitro, but we cannot exclude local seizure-generated cellular stress as a cause for upregulation in vivo. We and others reported similar results for ABC drug-resistance mediators P-gp and MRP1 (8,12). The findings are in keeping with an association of MVP with MDR in epilepsy also.
The mechanism(s) whereby MVP/vaults might effect an MDR phenotype is not known. Cancer studies suggest that vaults may alter intracellular anticancer drug disposition (41), or that they may influence nucleocytoplasmic trafficking (40), or axonal transport (46,47). The possible mechanism in epilepsy needs further exploration.
At least two different classes of drug resistance–associated molecules are, therefore, present in refractory epilepsy. ABC proteins P-gp, MRP1, and MRP2 are upregulated and probably contribute to MDR by drug transport: MVP is upregulated in normal and dysplastic neurons in epileptogenic pathologies; its mechanism of action is unknown. These observations suggest that the MDR phenotype in epilepsy may have a complex multifactorial basis. It is not necessarily the case, however, that the entire known range of resistance-mediating proteins are all upregulated in brain tissue from individuals with refractory epilepsy—no parenchymal upregulation of BCRP was noted. Careful determination of the range of proteins upregulated, their subcellular location, relative expression, and substrate specificities may identify useful therapeutic windows or determine novel treatment strategies in drug-resistant epilepsy.
Acknowledgment: This work was supported by grants from the Institute of Neurology and the National Hospital for Neurology and Neurosurgery, Glaxo-Wellcome plc through the Epilepsy Research Foundation, U.K., the Patrick Berthoud Trust, the University of London Central Research Fund and the British Council. We are grateful to Nicholas Win and Steve Durr for technical assistance.