Research Article

An Oral Vaccine Against NMDAR1 with Efficacy in Experimental Stroke and Epilepsy

See allHide authors and affiliations

Science  25 Feb 2000:
Vol. 287, Issue 5457, pp. 1453-1460
DOI: 10.1126/science.287.5457.1453


The brain is generally considered immunoprivileged, although increasing examples of immunological responses to brain antigens, neuronal expression of major histocompatibility class I genes, and neurological autoimmunity have been recognized. An adeno-associated virus (AAV) vaccine generated autoantibodies that targeted a specific brain protein, the NR1 subunit of theN-methyl-d-aspartate (NMDA) receptor. After peroral administration of the AAV vaccine, transgene expression persisted for at least 5 months and was associated with a robust humoral response in the absence of a significant cell-mediated response. This single-dose vaccine was associated with strong anti-epileptic and neuroprotective activity in rats for both a kainate-induced seizure model and also a middle cerebral artery occlusion stroke model at 1 to 5 months following vaccination. Thus, a vaccination strategy targeting brain proteins is feasible and may have therapeutic potential for neurological disorders.

In the central nervous system (CNS), both the NMDA and non-NMDA subclasses of glutamate receptors are mediators of glutamatergic excitatory neurotransmission. NMDA receptors are of major interest, as they are involved in many processes necessary for brain development including neuronal migration (1), patterning of afferent termination (2), and several forms of long-term synaptic plasticity (3). Properties of the receptor include calcium permeability, voltage-dependent Mg2+ block, and slow channel kinetics (4). Molecular cloning has revealed three receptor subunit families (NR1, NR2, and NR3A) which form hetero-oligomeric complexes in native NMDA receptor channels (5, 6). NR1 subunits are essential for the formation of functional NMDA receptors, whereas addition of other subunits modifies receptor properties (6, 7). In addition to the role of NMDA receptors in brain plasticity and development, they have also been implicated as a mediator of neuronal injury associated with many neurological disorders including stroke, epilepsy, brain trauma, dementia, and neurodegenerative disorders (8).

Because of their central involvement in the cascade leading to neuronal death following a variety of cerebral insults, pharmacological NMDA receptor antagonists have been evaluated for potential clinical use. These drugs are effective in many experimental animal models of disease, and some of these compounds have moved into clinical trials (9). However, the initial enthusiasm for this approach has waned, because the therapeutic ratio for most NMDA antagonists is poor, with significant adverse effects at clinically effective doses, thus limiting their utility (9).

As an alternative approach to antagonize NMDA receptors, we investigated the hypothesis that a humoral autoimmune response targeting the NR1 subunit of NMDA receptor might have neuroprotective activity. Moreover, we also hypothesized that such autoantibodies would have minimal penetration into the CNS under basal conditions and thereby avoid the toxicity associated with traditional approaches, but following a cerebral insult, would pass into the brain more efficiently, antagonize the receptor, and thereby attenuate NMDA receptor–mediated injury.

Oral genetic vaccination with recombinant AAV. We used an oral genetic vaccine approach for immunization. Oral vaccines are scavenged by intestinal M cells, rapidly taken up by the antigen-presenting cells (APCs) in Peyer's patches and the lamina propria, and can induce strong humoral immune responses (10). We recently reported a highly efficient method to transduce gut lamina propria cells by means of a recombinant AAV vector (11), indicating its potential use as a systemic oral vaccine also applicable to CNS disorders. A mouse NMDAR1 cDNA was subcloned into an AAV plasmid and subsequently packaged and purified (11) to yield a high-titer recombinant AAV virus, AAVNMDAR1 (12) (Fig. 1, A and C). This vector (109 infectious units) was administered via an orogastric tube into the stomachs of a group of rats, with a control group receiving a similar dose of a recombinant AAV virus expressing β-galactosidase (AAVlac) as described (11). In rats treated with the oral AAVNMDAR1 vector, transduction of intestinal cells was confirmed by polymerase chain reaction (PCR) amplification of a portion of the cytomegalovirus (CMV) promoter from genomic DNA isolated from the proximal intestine (13). Similarly, the PCR product was also detected in AAVlac-immunized rat intestine, but not detected in DNA from the intestine of naı̈ve rats, or in the liver, spleen, or testis of any AAVNMDAR1- or AAVlac-immunized rats (Fig. 1B). NMDAR1 protein expression in the intestine was detected by immunohistochemistry with a commercially available NMDAR1 primary antibody and fluorescent secondary antibody at 1 month (Fig. 1D) and 5 months (Fig. 1F) after peroral dosing. Upon confocal microscopy imaging, expression was observed to be primarily in the lamina propria as previously described for AAVlac-treated animals (11). A total of 9 × 106 to 12 × 106 lamina propria cells were transduced at 4 weeks after immunization, with approximately ∼107 immunoreactive cells remaining at 5 months as previously reported when Escherichia coli lacZ was used as the transgene (11). Transduced cells were immunoreactive when the antibodies MRC OX-41 and OX-62 were used to define signal regulatory proteins expressed on cells of myeloid origin (SIRP) and dendritic cells, respectively (Fig. 2).

Figure 1

Vector diagram and gut gene expression. (A) NMDAR1 plasmid construct and (B) PCR analysis. (C) NMDAR1 immunocytochemistry of HEK 293 cells (arrows). (D through F) NMDAR1 immunohistochemistry (blue; 1:250 dilution, Chemicon) with propidium iodide counterstaining (orange) using methods as described (11) to show the lamina propria (lp) and epithelial (ep) cell layers at (D) 4 weeks and (F) 5 months after AAVNMDAR1 and (E) 4 weeks after AAVlac administration. Scale bar: (D), 53 μm; (E) and (F), 25 μm.

Figure 2

NMDAR1 protein expression in lamina propria. Double immunofluorescent labeling (19) using monoclonal antibodies to NMDAR1 [red in (B), blue in (D)], and (A) SIRP (blue, 1:50 dilution, Chemicon) and (C) dendritic cells (red, 1:5, Chemicon). Sections were incubated with unlabeled secondary anti-mouse IgG (1:250) between primary antibody incubation steps. Primary antibodies were detected with Cy3- or Cy5-conjugated anti-mouse antibodies (1:250, Jackson Immunoresearch), and immediately prior to confocal imaging (19), sections were counterstained with acridine orange [green, 0.01% in PBS (pH 6.2) for 1 min], washed in PBS, and mounted in Vectashield (Vector). Bar, 10 μm.

Blood was taken and the serum analyzed for presence of specific antibodies. In AAVlac-treated animals, immunoglobulin G (IgG) antibodies were detected at 4 weeks, with titers further increasing at 16 weeks (14). When a purified β-galactosidase enzyme preparation was used, a commercial monoclonal antibody recognized two protein species with molecular weights of 116 and 85 kD on immunoblots (Fig. 3A). Some animals had antibodies that bound preferentially to the 85-kD species, whereas other animals had serum antibodies with higher affinity to the 116-kD protein (Fig. 3A). Immunoblotting of AAVNMDAR1 serum to a hippocampal membrane extract yielded the expected 117-kD band, consistent with the size of the native NMDAR1 receptor subunit and similarly recognized by commercial antibodies (Fig. 3B) (14). The sera from naı̈ve control (n = 4) or AAVlac-immunized (n = 12) animals showed no binding to the brain extract. When brain extracts were prepared under more severe denaturing conditions, there was degradation of the native receptor as shown by monoclonal antibody binding to 32- and 67-kD fragments (Fig. 3C). Individual AAVNMDAR1 rats showed different patterns of binding to some of these fragments (Fig. 3C). These data indicate that AAV can serve as an oral vaccine and that immunized animals generated antibodies against a range of NMDAR1 epitopes.

Figure 3

Immunoblot analysis of AAVlac and AAVNMDAR1 serum and AAVNMDAR1 serum epitope mapping. (A) Purified β-galactosidase (1 μg per lane) was separated on a 10% acrylamide gel under reducing conditions and screened with monoclonal β-galactosidase antibody (1:5,000, Gibco BRL) and naı̈ve, AAVlac, and AAVNMDAR1 serum (1:200) using standard immunoblot methods (14). (B) Polyclonal NMDAR1 antibody (1:250, Chemicon) and serum (1:200) from the three groups were screened against 20 μg of hippocampal extract for the presence of NMDAR1 antibodies. (C) Polyclonal (1:250) and monoclonal (1:3000, Chemicon) NMDAR1 antibodies and sera (1:200) were screened against 20 μg of denatured hippocampal extract. N35, N64, and N52 represent serum from three AAVNMDAR1 animals showing specific affinities for different NMDAR1 breakdown products. (Dthrough H) Epitope map profiles of five (rats N11, N19, N21, N52, N64) AAVNMDAR1-treated animals. Specificity is measured as a ratio between the AAVNMDAR1 signal and mean AAVlac signals for each peptide.

Epitope mapping. To further define the range of epitopes, 94 overlapping 16–amino acid oligomers (16-mers) were synthesized that covered the entire 938 amino acids of the native NMDAR1 protein (15). Serum from AAVNMDAR1, AAVlac, and naı̈ve rats was screened against this panel of 16-mers. None of the naı̈ve (n = 4) or AAVlac (n = 14) serum screened had specific binding to any of the peptides. In contrast, AAVNMDAR1-immunized rat serum showed specific binding to peptides that corresponded to functional domains within the extracellular segments of the receptor (Fig. 3). These included peptide 49 (amino acids 483 through 498), which represents the NH2-terminal side of M1, the first transmembrane domain (rat N19, Fig. 3E), and two peptides corresponding to the extracellular domain between M3 and M4, peptide 69 (amino acids 681 through 696; rat N21, Fig. 3F) and peptide 72 (amino acids 711 through 726; rat N64, Fig. 3H). Each of these three peptides contain critical residues for glycine binding (16). The most common pattern observed in serum from 7 of the 19 AAVNMDAR1 rats screened was specific binding to peptides adjacent to the M4n region (peptide 80, amino acids 791 through 807) and/or peptide 65 (amino acids 641 through 657), corresponding to the M3c domain (rat N11, Fig. 3D). Two additional rats had antibodies that bound to peptides 54 and 55 (amino acids 541 through 566; rat N52, Fig. 3G), which map to the preM1 domain. The preM1, M4n, and M3c regions form part of the extracellular vestibule of the NMDA receptor channel, where amino acid substitutions at key residues have a significant influence on channel permeability (17).

Cell-mediated immunity. A splenocyte proliferation assay was performed to evaluate a cell-mediated immune component associated with the vaccine. No difference in proliferative response was found between splenocytes isolated from AAVlac and AAVNMDAR1 animals challenged with the appropriate antigen (Table 1) (18). In addition, immunohistochemistry with CD8+ and major histocompatibility complex (MHC) class I antibody markers on hippocampal sections from naı̈ve, AAVlac, and AAVNMDAR1 rats showed no difference in CD8+ or MHC class I immunoreactivity (Table 1).

Table 1

Splenocytes isolated from vaccinated rats were incubated with antigen [Hippocampal extract, recombinant NH2-terminal NMDAR1 (N-term.), and recombinant COOH-terminal NMDAR1 (C-term.)], positive (Concavalin A) and negative (RPMI) controls with proliferation quantitated using thymidine incorporation (18). ND, not done. NC, not changed from negative control. MHC I (1:100, Pharmingen) and CD8+ (1:200, Chemicon) immunohistochemistry was carried out (19) and sections were scored by densitometry and cell counting. (−), no immunoreactive cells. Cellular infiltration into the hippocampus, cortex and striatum was determined by HE staining. Neg., no inflammatory cells.

View this table:

Experimental epilepsy. We used the kainate model of temporal lobe epilepsy (19) to determine potential anti-epileptic and neuroprotective efficacy of AAVNMDAR1 vaccination. Animals received AAVNMDAR1 (n = 9) or AAVlac (n = 8), and a group of naı̈ve control rats received no vector (n = 17). At 1 month following vaccination, animals were administered kainic acid (10 mg/kg body weight) intraperitoneally (i.p.), and a blinded observer determined over 2 hours whether there were any signs of the characteristic progression through various behavioral seizure stages, including immobility and staring, “wet-dog-shakes,” unilateral and bilateral forelimb clonus, and facial clonus (19). Electroencephalographic (EEG) recordings (Fig. 4A) showed the first signs of electrographic seizure activity within 10 min after drug administration in control animals, with 13 of 17 naı̈ve and six of the eight AAVlac animals developing facial and forearm clonus and proceeding to status epilepticus (SE). In contrast, only two of the nine AAVNMDAR1-vaccinated animals developed seizures and SE (Table 2). The remaining seven AAVNMDAR1 rats showed neither EEG changes nor any behavioral changes after kainate treatment (P = 0.007, chi square analysis with expected SE frequency of 68% reduced to 22% in the NMDAR1-immunized group).

Figure 4

Kainate-induced seizure damage in the hippocampus. (A) EEG recordings and (B throughI) analysis of hippocampal damage in the hilus using (B, D, F, and H) fluorescent TUNEL labeling (green) or (C, E, G, and I) clusterin immunohistochemistry (green) combined with immunohistochemistry with the neuronal marker NeuN (orange) in the corresponding brains. In those rats that exhibited SE, an anticonvulsant dose of sodium pentobarbital (30 mg/kg i.p.) was administered after 45 min of continuous SE. Animals that did not develop seizures also received pentobarbital. Inset in (B) shows a high-power image of a double-labeled TUNEL and NeuN immunofluorescent neuron. (B and C) AAVlac rat developed SE. (D and E) AAVNMDAR1 rat developed no SE. (F and G) AAVNMDAR1 rat (N12) developed SE, with hippocampal injury. (H and I) AAVNMDAR1 rat (N7) developed SE, with no hippocampal injury. TUNEL and clusterin immunohistochemistry was performed as described (19, 20) on 16-μm slide-mounted sections and visualized using ExtrAvidin-FITC (Sigma) and anti-rabbit FITC (Jackson Immunoresearch), respectively, prior to NeuN (1:1000, Chemicon) immunohistochemistry and anti-mouse Cy3 detection (Jackson Immunoresearch). Clusterin (1:200) was provided by D. Christie, University of Auckland, New Zealand. Bar, 200 μm.

Table 2

Hippocampal injury was graded according to the extent of injury: (–), no injury; (++), moderate (<20 TUNEL or clusterin-immunofluorescent cells); (+++), extensive (>20 TUNEL or clusterin-immunofluorescent cells). Numbers in parentheses represent total number of rats in that group.

View this table:

To confirm that the AAVNMDAR1 vaccination was able to suppress kainate-induced seizures, we examined the hippocampi of all animals for any signs of seizure-induced neuropathology. Three days after kainate treatment, animals were euthanized before brains were removed and frozen. Coronal hippocampal sections were prepared for immunohistochemistry. No discernible TUNEL signal or clusterin immunofluorescence [a cell death marker (20)] was observed in the hippocampus of any animal (AAVlac, naı̈ve, or AAVNMDAR1) that did not develop SE (Fig. 4, D and E). In comparison, all AAVlac and naı̈ve animals that experienced SE had numerous TUNEL-positive and clusterin-immunofluorescent neurons in hippocampal CA1, CA3, and hilar regions (Fig. 4, B and C). Of the two AAVNMDAR1 animals that had SE, only one of these animals (rat N12) had some injury (Fig. 4, F and G), whereas the second animal (N7) had no injury whatsoever, despite more than 45 min of severe SE (Fig. 4, H and I;Table 2). N12, the only rat to have SE as well as brain injury following AAVNMDAR1 oral dosing, was also the only animal of the nine in this experiment that did not have detectable circulating antibodies to the NMDAR1 receptor subunit. Because seizure duration is a critical determinant of the extent of neuronal injury, with as little as 30 min duration being sufficient to cause cell death (21), these results suggest that the AAVNMDAR1 vaccination can not only produce resistance to seizures but also protect against excitotoxic injury.

We were interested in determining if the anti-epileptic and neuroprotective effect of the NMDAR1 vaccination was nonspecific and could be reproduced with immunization against any brain protein. Therefore, we generated an AAV vector expressing GAD-65 and perorally administered this vector to an additional group of rats. These rats all developed significant antibody titers to the native protein (Fig. 5A). At 1 month following vaccination, these rats were challenged with kainate as above (19). All of these rats (n = 6) rapidly progressed to motor seizures and SE (Table 2). Analysis of their brains showed extensive damage (22).

Figure 5

Antibody passage across an intact BBB. (A) NMDAR1 antibody (1:5000), and naı̈ve, AAVlac, and AAVNMDAR1 serum (1:500) was screened against 20 μg of hippocampal membrane extract, and AAVGAD65 serum (1:200) against 20 μg hippocampal cytosolic fraction by immunoblot analysis (14). CSF (1:5) from the same animals was screened against 30 μg of hippocampal extract for NMDAR1 antibodies. (Bthrough D) Anti-rat IgG immunoreactivity in the hilus and CA2–CA3 region of an AAVNMDAR1-immunized rat [(B) arrows]. (C) A high-powered image of the CA3 region in (B) is compared to the same region under basal conditions [(D) arrows]. (Ethrough J) Immunohistochemistry with 50 μg/ml (E and I) AAVNMDAR1- and (F and H) AAVlac-purified IgG, and (G and J) NMDAR1 antibody (1:250, Chemicon) and diaminobenzidine (DAB) detection (19) in the (E through G) CA3 and (H through J) hilar region. Scale bar: (B), 500 μm; (C through G), 200 μm; (H through J), 80 μm.

To confirm passage of NMDAR1 antibodies into the brain, after 4 weeks, a group of vaccinated rats (n = 4) underwent nontraumatic sampling of cerebrospinal fluid (CSF) from the cisterna magna (23). Using immunoblot analysis and a highly sensitive chemiluminescent detection method, we were able to detect NMDAR1 autoantibodies at low levels in the CSF by immunoblotting, suggesting autoantibody passage into the brain (Fig. 5A). NMDAR1 autoantibodies were not detected in CSF from naı̈ve (n = 4) or AAVlac-immunized (n = 4) rats. After kainate treatment, an approximate tenfold increase in CSF levels of NMDAR1 autoantibodies was observed (Fig. 5A). In addition, IgG immunoreactivity (24) was specifically increased in the hippocampus in the CA3 and dentate hilar neurons, which are also susceptible to kainate-induced neuronal damage but were protected in NMDAR1-immunized rats (Fig. 5, B through D). Furthermore, immunohistochemistry with IgG purified from AAVNMDAR1, AAVlac, or naı̈ve serum was conducted on hippocampal sections from naı̈ve animals that had been perfused with phosphate-buffered saline (PBS) to remove endogenous IgG present in blood vessels. Only AAVNMDAR1-purified IgG showed selective immunoreactivity in hippocampal CA3, dentate hilar, and CA1 neurons, consistent with the pattern of expression of NMDAR1 as determined with a commercial polyclonal antibody (Fig. 5, E through J).

Experimental stroke. We used the endothelin-1 model of middle cerebral artery occlusion (MCAO) (25) to determine the antistroke and ischemic neuroprotection efficacy of AAVNMDAR1 vaccination. In this approach, a fine needle is used to direct stereotactic delivery of endothelin-1 into the vicinity of the middle cerebral artery (MCA), which induces vasospasm leading to complete occlusion and generation of an extensive infarct in the MCA territory (including the striatum and cortex on the injected side). The model has been used previously to test novel antistroke drugs, including the NMDA receptor antagonist MK-801 (25, 26). Three groups of rats [naı̈ve (n = 10), AAVlac (n = 8), and AAVNMDAR1 (n = 10)] underwent MCAO 5 months after vaccination. At 3 days after endothelin-1 administration, rats were euthanized, the brains were removed, and the infarct volume was determined by a blinded investigator (Fig. 6) (25). The total infarct volume was 55.6 ± 20.2 mm3 in naı̈ve rats and 66.4 ± 12.4 mm3 in the AAVlac animals, whereas in the AAVNMDAR1 rats, the infarct was reduced by ∼70% to 19.6 ± 6.2 mm3 (P = 0.002, Student'st-test) (Fig. 6H). When the cortex and striatum were analyzed independently, the protection in terms of infarct volume was 65 and 80%, respectively, compared to the AAVlac group and 69 and 74%, respectively, compared to the naı̈ve controls (Fig. 6H).

Figure 6

Reduced ischemic damage in cortex and striatum of AAVNMDAR1-vaccinated rats after MCAO. HE staining of the striatum and/or cortical regions contralateral (A,C, and E) and ipsilateral (B,D, and F) to MCAO in (A through D) AAVlac- and (E and F) AAVNMDAR1-vaccinated animals. (G) Isolectin B4 (1:50, Sigma) and DAB detection (19) of microglial infiltration in the ipsilateral hemisphere. (H) Total infarct volume (mean ± SEM) in the naı̈ve (white bars), AAVlac- (blue bars), and AAVNMDAR1-vaccinated groups (red bars). *P < 0.01. Scale bar: (A, B, and E through G), 600 μm; (C and D), 100 μm.

Vaccination effects on behavior. NMDA receptor antagonists administered systemically at anti-epileptic and neuroprotective doses typically result in some impairment in motor behavior (27). To determine whether vaccination was associated with any changes in motor behavior, rats were tested on circular track and line-crossing mobility paradigms; these tests failed to demonstrate a difference between the groups (repeated measures ANOVA, P = 0.87 and P = 0.32, respectively) (22).

Discussion. The brain, retina, and testis are generally perceived as immune-privileged sites, as defined traditionally by the prolonged survival of allogeneic or xenogeneic tissue transplanted into these organs. Although we now know that this immune privilege is relative and not absolute, many foreign antigens within the CNS escape immune surveillance and certain self-antigens in the brain may not induce immune tolerance; for example, in paraneoplastic disorders (PND), circulating autoantibodies interact with neuronal antigens (28). It is believed that brain antigens previously sequestered from the immune system via the blood-brain barrier (BBB) are presented to a naı̈ve immune system when ectopically expressed in cancer cells and then elicit a humoral immune response (29). A second immune or inflammatory process may result in compromise of the integrity of the BBB, enabling a breach of the immune privilege of the brain and passage of antibodies to the targeted neurons, resulting in the characteristic disease phenotype. If this hypothesis is correct, it should be possible to vaccinate against selected brain antigens to induce a state of autoimmunity. Our study is analagous to the clinical example of PND in which a neuronal antigen is ectopically expressed (in the gut in our study, as opposed to a cancer cell in a malignant tumor), thereby inducing autoimmunity and influencing brain function.

Recently, Schenk et al. reported that repeated immunization with amyloid-β attenuated Alzheimer disease-like pathology in the PDAPP mouse model, possibly mediated via antibody facilitating the clearance of plaques (30). Also of relevance to our experiments on stroke and epilepsy are the demonstration of autoantibodies to GluR3 and GAD-65 in Rasmussen's encephalitis (31) and Stiffman syndrome (32), respectively. The Rasmussen's association was discovered when rabbits developed seizures following vaccination with a GluR3 receptor fragment. The autoantibodies were shown to have agonist properties at the receptor and thereby suggested a mechanism for hyperexcitability leading to the epilepsy associated with Rasmussen's encephalitis (33). Antibodies from the plasma of patients with Stiffman syndrome have been shown to bind to GAD-65 and inhibit γ-aminobutyric acid (GABA) synthesis in vitro (34). A GABA-deficient state may therefore lead to disinhibition and the rigidity characteristic of Stiffman syndrome. The higher than predicted (100% versus 68%) SE frequency we observed in the GAD-immunized rats supports this hypothesis.

Serum antibody titers were obtained in vaccinated animals, and we were able to detect low levels of NMDAR1 antibodies in the CSF under basal conditions. The lack of any detectable cellular immune response suggests that the likely mechanism of action of the vaccination is mediated through a humoral response consistent with the observed association between antibody titer and the neuroprotective phenotype.

NMDA receptor antagonists influence several experimental animal models of human neurological disease. The systemic administration of kainate is a well-established model of temporal lobe epilepsy. Moreover, previous studies have shown that NMDA receptor antagonism can partially inhibit seizure activity in this model in which, by activation of both pre- and postsynaptic receptors, kainate induces excessive glutamate release and NMDA-mediated neuronal excitation and injury (35). Our study produced a 78% anticonvulsant effect, comparing favorably with previous pharmacological studies, in which similar levels of anti-epileptic activity are typically associated with significant motor impairment (27).

In partial stroke models, some of the most reproducible and potent data on neuroprotection has come from the use of NMDA receptor antagonists, particularly the noncompetitive compound MK-801 (25,26). However, even with this antagonist at doses that depress motor activity, tissue rescue is limited to ∼50% in the cortex with no infarct reduction in the striatum (26). Moreover, there is a narrow time window of only a few hours in which MK-801 and other promising new antistroke drugs can be given for rescue, with effective protection requiring continued maintenance dosing (36).

One human study in Russia with 270 patients showed the presence of NMDA receptor autoantibodies in the serum after a stroke (37). Neurological recovery occurred in individuals with significant antibody titers, but poor prognosis or death was associated with the absence of NMDA receptor autoantibodies. This data further supports the clinical relevance of the vaccine.

The AAVNMDAR1-vaccinated rats generated autoantibodies, which were polyclonal. Epitope mapping of the antibodies showed a diverse range of epitopes in multiple regions of the extracellular domains. Individual animals had antibodies which bound to 16-mers that mapped to the extracellular vestibule of the channel including the preM1, M3c, and M4n domains (16) as well as epitopes in the NH2-terminal region and within the extracellular loop lying between M3 and M4, which directly mapped to glycine-binding sites (16). Several animals did not appear to have an immunodominant epitope using the set of overlapping 16-mers, a result consistent with antibodies that are dependent on the conformational state of the protein. Further analysis of epitopes and protein fragments and correlation with neuroprotection will be necessary to define which sites of the NMDAR1 protein are the best targets for neuronal protection and vaccine development.

A major limitation of the successful translation of promising NMDA receptor antagonists to the clinic has been the significant profile of CNS-adverse effects (9). A potential advantage of a vaccine or antibody approach to NMDA receptor antagonism is that the receptor blockade is minimal under resting physiological conditions, under which high serum titers of antibodies do not pass the BBB efficiently. However, after a neuronal insult, the BBB has increased permeability to serum antibodies, and transport and subsequent binding to the target protein can occur (38). This “on demand” or selective delivery of the neuroprotective agent, limited both spatially to the site of injury and to the precise timing of injury, is one of the most promising features of our approach. Of particular interest, glutamate itself has been reported to alter BBB permeability (39). Hence, our vaccine strategy might induce an autoprotective loop, whereby a cerebral insult increases brain extracellular glutamate and increases BBB permeability locally, resulting in facilitated passage of the autoantibody and antagonism of glutamate receptors. It has been shown that the BBB is compromised after systemic kainate administration or MCA occlusion (40). Whether this vaccine approach would also be effective in neurological disorders in which NMDA receptors have been implicated but the BBB remains impermeable (possibly including neurodegenerative diseases and AIDS dementia) is still unclear.

We showed neuroprotection in the rats immunized against NMDAR1 but not in the control groups including naı̈ve, lacZ-, and GAD-65–immunized rats. Moreover, the induction of autoantibodies that bound functional domains of the NMDA receptor and that yielded a staining pattern in hippocampal slices confined to cells known to express native NR1 receptors suggests that the immunization effect was specific to the NMDA receptor. Despite this binding specificity, it remains possible that effects unrelated to receptor antagonism (for instance, antibody-mediated up-regulation of neuroprotective genes) contributed in part to the dramatic neuroprotective phenotype we observed. Furthermore, the passage into the CNS of antibodies binding to other neuronal surface proteins might also protect against neurotoxic insults and stroke. However, the epileptogenic phenotype of our GAD-65–immunized rats, all of whom progressed to SE following kainate administration, as well as previous data showing seizures in rabbits immunized against GluR3 (31), make a nonspecific vaccination effect less likely.

In contrast to the motor impairment associated with systemic administration of most pharmacological NMDA receptor blockers, the AAVNMDAR1-vaccinated rats had no impairment in locomotor function. Our results therefore suggest the possibility of vaccination for individuals at risk of stroke and other cerebral insults without impairment of neurological function. In support of this idea, we have collected preliminary data suggesting that far from impairing cognition, the action of the autoantibodies may parallel those of partial NMDA receptor agonists and facilitate learning and memory (41). Oral recombinant AAV immunization is associated with a humoral response persisting over many months and possibly years. One possible limitation is that the chronic elevation of these autoantibodies may eventually have untoward effects on brain function. We are continuing to look at long-term effects of the immunization on more complex behaviors mediated by NMDA receptors. These experiments and additional safety studies need to be investigated before any translation of this technology to the clinic.

The clinical utility of current antistroke drugs has been limited by a narrow therapeutic time window for rescue and clinical benefit. A stroke vaccine that generates autoantibodies whose access to the brain and neuroprotective activity is spatially and temporally regulated may hold promise as a prophylactic measure to protect the brain. Moreover, the ability of systemic immunization to generate autoantibodies which bind to, and thereby alter, the function of native brain proteins opens up a new avenue for modulating the nervous system and treating neurological and psychiatric disorders.

  • * To whom correspondence should be addressed. E-mail: matthew.during{at}


View Abstract

Stay Connected to Science

Navigate This Article