PHARMACOLOGY
`
`The synaptic vesicle protein SV2A is the binding site
`for the antiepileptic drug levetiracetam
`
`Berkley A. Lynch*†, Nathalie Lambeng‡, Karl Nocka§, Patricia Kensel-Hammes¶, Sandra M. Bajjalieh¶, Alain Matagne储,
`and Bruno Fuks‡
`
`Departments of *Molecular and Cellular Biology and §Bioinformatics and Target Discovery, UCB Research Inc., 840 Memorial Drive, Cambridge, MA 02139;
`Departments of ‡In Vitro Pharmacology and 储CNS Pharmacology, UCB S.A., Pharma Sector, Chemin du Foriest, B-1420 Braine L’Alleud, Belgium; and
`¶Department of Pharmacology, University of Washington, D429 HSB, Box 357280, Seattle, WA 98195-7280
`
`Edited by William A. Catterall, University of Washington School of Medicine, Seattle, WA, and approved May 17, 2004 (received for review
`December 10, 2003)
`
`Here, we show that the synaptic vesicle protein SV2A is the brain
`binding site of levetiracetam (LEV), a new antiepileptic drug with
`a unique activity profile in animal models of seizure and epilepsy.
`The LEV-binding site is enriched in synaptic vesicles, and photoaf-
`finity labeling of purified synaptic vesicles confirms that it has an
`apparent molecular mass of ⬇90 kDa. Brain membranes and
`purified synaptic vesicles from mice lacking SV2A do not bind a
`tritiated LEV derivative, indicating that SV2A is necessary for LEV
`binding. LEV and related compounds bind to SV2A expressed in
`fibroblasts, indicating that SV2A is sufficient for LEV binding. No
`binding was observed to the related isoforms SV2B and SV2C.
`Furthermore, there is a high degree of correlation between binding
`affinities of a series of LEV derivatives to SV2A in fibroblasts and
`to the LEV-binding site in brain. Finally, there is a strong correlation
`between the affinity of a compound for SV2A and its ability to
`protect against seizures in an audiogenic mouse animal model of
`epilepsy. These experimental results suggest that SV2A is the
`binding site of LEV in the brain and that LEV acts by modulating the
`function of SV2A, supporting previous indications that LEV pos-
`sesses a mechanism of action distinct from that of other antiepi-
`leptic drugs. Further, these results indicate that proteins involved
`in vesicle exocytosis, and SV2 in particular, are promising targets
`for the development of new CNS drug therapies.
`
`Epilepsy, a group of diseases characterized by recurrent
`
`spontaneous seizures, is a prevalent chronic neurological
`disorder (1). An often devastating disease, epilepsy is frequently
`resistant to conventional antiepileptic drugs (AEDs), even when
`they are used in polytherapy (1). Traditional AEDs mainly target
`ion channels or postsynaptic receptors. Given that traditional
`AEDs do not control seizures in all patients, identification of
`alternative molecular pathways and targets for therapeutic in-
`tervention in epilepsy is warranted. Levetiracetam [LEV; (S)-
`␣-ethyl-2-oxo-pyrrolidine acetamide; KEPPRA] is an antiepi-
`leptic drug approved by the Food and Drug Administration in
`the year 1999 as an adjunctive therapy for the treatment of
`refractory partial epilepsy in adults (2, 3). LEV possesses several
`properties that distinguish it from classical AEDs (4). It has a
`distinctive pharmacological profile in animal models of seizures
`and epilepsy, as demonstrated by its lack of activity in the acute
`seizure models traditionally used to screen for antiepileptic
`drugs (5). This lack of activity against acutely generated seizures
`contrasts LEV’s potent seizure protection in animal models of
`chronic epilepsy, including genetic and kindling models (5).
`Compared with traditional AEDs, LEV has the ability to inhibit
`neuronal hypersynchronization when epileptiform activity is
`evoked in rat hippocampal slices (6, 7). Also exceptional among
`AEDs is LEV’s ability to counteract the development of amyg-
`dala electrical kindling in rodents even after drug dosing is
`terminated (8).
`LEV’s pharmacological profile has been presumed to relate to
`a distinctive mechanism of action (9). Additionally, LEV does
`not seem to act by means of any of the three main mechanisms
`currently accepted for the antiseizure action of established
`
`AEDs: (i) ␥-aminobutyratergic (GABAergic) facilitation, (ii)
`inhibition of Na⫹ channels, or (iii) modulation of low-voltage
`activated Ca2⫹ currents (9, 10). Previous studies revealed that
`LEV binds saturably, reversibly, and stereospecifically to an
`unidentified binding site in rat brain (11). Screening of a large
`number of known AEDs and other neuroactive compounds
`failed to identify any with high affinity for the LEV-binding site
`(11), providing support for the novelty of the LEV-binding site.
`Testing a series of LEV analogs revealed a strong correlation
`between their affinities for the brain binding site and their
`antiseizure potencies in the audiogenic mouse model of epilepsy
`(11). This finding indicates a functional role for the unidentified
`brain binding site in the antiseizure actions of LEV.
`Detection of a LEV-binding site in brain provided the ratio-
`nale to search for the LEV-binding molecule. Further charac-
`terization of the binding site led to its classification as an integral
`membrane protein of widespread distribution in brain, localized
`in neurons and enriched in the synaptic vesicle membrane
`fraction (11, 12). SDS兾PAGE on the LEV-binding protein from
`brain membranes cross-linked to a tritiated photoaffinity LEV
`derivative determined that the protein has an apparent molec-
`ular mass of ⬇90 kDa (12). Among possible candidate proteins
`matching these characteristics is the synaptic vesicle protein 2
`(SV2).
`SV2, an integral membrane protein present on all synaptic
`vesicles, is a small gene family consisting of three isoforms,
`designated SV2A, SV2B, and SV2C. SV2A is the most widely
`distributed isoform, being nearly ubiquitous in the CNS, as well
`as being present in endocrine cells (13, 14). SV2B is brain
`specific, with a wide but not ubiquitous distribution, and SV2C
`is a minor isoform in brain. The brain distribution of the
`LEV-binding site, as revealed by autoradiography, matches the
`equivalent distribution of SV2A as determined by immunocy-
`tochemistry (14, 15). Both SV2A⫺/⫺ and SV2B⫺/⫺ homozygous
`knockout (KO) mice have been reported, as well as double A兾B
`KOs (16, 17). SV2A and SV2A兾B KOs exhibit a severe seizure
`phenotype whereas the SV2B KOs do not. Studies of the SV2
`KOs indicate that SV2 has a crucial role in the regulation of
`vesicle function, although not in vesicle biogenesis or synaptic
`morphology (16, 17).
`Here, we report studies designed to test the hypothesis that
`SV2 is the LEV-binding site. We demonstrate that the protein
`SV2A is the LEV-binding site in brain and that there is an
`excellent correlation between the binding affinity of LEV and
`derivatives in brain and to heterologously expressed human
`
`This paper was submitted directly (Track II) to the PNAS office.
`
`Abbreviations: AED, antiepileptic drug; LEV, levetiracetam; SV2, synaptic vesicle protein 2;
`hSV2A, human SV2A; KO, knockout; ␤-gal, ␤-galactosidase; DM, n-dodecyl-␤-D-maltoside;
`[3H]ucb 30889, (2S)-2-[4-(3-azidophenyl)-2-oxopyrrolidin-1-yl]butanamide; pIC50, ⫺log
`IC50.
`†To whom correspondence should be addressed. E-mail: berkley.lynch@ucb-group.com.
`
`© 2004 by The National Academy of Sciences of the USA
`PNAS 兩
`June 29, 2004 兩 vol. 101 兩 no. 26 兩 9861–9866
`
`www.pnas.org兾cgi兾doi兾10.1073兾pnas.0308208101
`
`ARGENTUM Exhibit 1126
` Argentum Pharmaceuticals LLC v. Research Corporation Technologies, Inc.
`IPR2016-00204
`
`Page 00001
`
`

`
`SV2A in fibroblasts. These data have implications for the
`mechanism of action of LEV as an antiepileptic drug, and
`potentially for future research into the contribution of presyn-
`aptic mechanisms to seizure initiation and propagation in the
`brain.
`
`Materials and Methods
`Reagents. LEV and derivatives were synthesized at UCB Pharma.
`[3H]ucb 30889, (2S)-2-[4-(3-azidophenyl)-2-oxopyrrolidin-1-
`yl]butanamide (32 Ci兾mmol; 1 Ci ⫽ 37 GBq), was custom
`labeled by Amersham Pharmacia.
`
`WT and SV2 KO Animals. All animal experiments were approved by
`the local ethics committee for animal experimentation, accord-
`ing to U.S. and Belgian law. SV2A KO mice have been reported
`(16). SV2B KOs were bred with animals heterozygous for the
`SV2A gene disruption to produce SV2A⫹/⫺SV2B⫺/⫺ breeders,
`which were used to generate SV2A兾B KOs. WT C57BL兾6 and
`SV2 KO mouse brain membranes for Western blot and binding
`analyses were prepared as follows. Frozen whole brains were
`homogenized (10% wt兾vol) in 20 mM Tris䡠HCl buffer (pH 7.4)
`containing 250 mM sucrose (buffer A). The homogenates were
`spun at 30,000 ⫻ g at 4°C for 15 min, and the pellets were
`resuspended in the same buffer. After incubation at 37°C for 15
`min, the membranes were washed two times, resuspended in
`buffer A, and frozen.
`
`Photoaffinity Labeling Experiments. Purification of crude synapto-
`somal (P2) and synaptic vesicle (LP2) fractions from rat brain
`was performed as reported (12), except that the LP2 fraction was
`washed at 100,000 ⫻ g for 60 min. Photoaffinity labeling was
`performed as described (12), except for using synaptic vesicle
`fractions instead of brain membranes. The inclusion of 1 mM
`LEV in the reaction was shown to prevent the photolabeling. For
`size analysis of the labeled binding site, membranes were ex-
`tracted and run on SDS兾PAGE. The developed gel was sliced,
`and the slices were solubilized and counted as described (12).
`
`Synaptic Vesicle Purification from WT and KO Mice. Synaptic vesicles
`were purified by using the technique of Huttner et al. (18) as
`described (19), except that the final two centrifugation steps were
`done by using a SW 50 rotor.
`
`Cloning and Expression of SV2 Isoforms. Human SV2A (hSV2A)
`was PCR-amplified from a human fetal brain cDNA with added
`GATEWAY (Invitrogen) attB1 and attB2 flanking sequences.
`The resultant product was cloned by recombination into a
`pDONR201 vector (Invitrogen). hSV2B and hSV2C were PCR-
`amplified from first-strand cDNA synthesized from human adult
`brain total RNA (Ambion, Austin, TX) and recombined by TA
`cloning into a GATEWAY pENTR兾SD兾D vector (Invitrogen).
`All three isoforms were recombined into pd40 expression vectors
`(Invitrogen), supporting expression of the native, nonfused
`protein from a cytomegalovirus immediate-early promoter. For
`transfections, one of four vectors was used, either a control
`vector containing the ␤-galactosidase (␤-gal) gene, or hSV2A,
`hSV2B, or hSV2C in pd40 vectors. Transient transfections in
`COS-7 cells were performed on preconfluent cells by using
`either the reagent Lipofectamine 2000 (Invitrogen) or FuGENE
`6 (Roche), with equivalent results. For confluent cell assays, cells
`were grown, transfected, and assayed in 24-well plates. For
`suspension cell assays, cells were grown in, transfected in, and
`harvested from 10-cm plates (and stored at ⫺80°C for later use).
`Cells were used 48 h posttransfection.
`
`SV2A Solubilization and Immunoprecipitation. Rat brain membranes
`were diluted in a solubilization buffer [20 mM Tris䡠HCl (pH 7.4),
`0.25 M sucrose, and protease inhibitors (Complete, Roche)]
`
`9862 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0308208101
`
`containing 15 mM n-dodecyl-␤-D-maltoside (DM), incubated for
`2 h at 4°C and centrifuged at 4°C for 1 h at100,000 ⫻ g. Four
`micrograms of anti-SV2A antibodies (sc-11936, Santa Cruz
`Biotechnology) or normal goat IgG (sc-2028, Santa Cruz Bio-
`technology) was added to 1.5 mg of supernatant for overnight
`incubation at 4°C, and then incubated with protein A-Sepharose
`beads for 1 h at 4°C. After several washes, the immune pellets
`were collected by centrifugation and boiled for 5 min in SDS
`sample buffer containing 2-mercaptoethanol. Proteins were
`separated by SDS兾PAGE and transferred to nitrocellulose for
`immunoblotting with anti-SV2A antibodies (see above). Twenty
`micrograms of rat brain membranes without immunoprecipita-
`tion was immunoblotted as a control.
`
`Western Blot Analysis of Protein Expression. Approximately 10 ␮g of
`total protein from WT and KO brain membranes was loaded on
`a Tris䡠glycine兾4–12% polyacrylamide gel and developed. After
`transfer to a nitrocellulose membrane, the blots were probed
`with either a monoclonal cross-reactive to all SV2 proteins (13)
`(obtained from the Developmental Studies Hybridoma Bank,
`supported by the National Institute of Child Health and Human
`Development and maintained by the Department of Biological
`Sciences of the University of Iowa, Iowa City) or with a poly-
`clonal antibody specific for the SV2A isoform (sc-11936, Santa
`Cruz Biotechnology, as above). For the synaptic vesicle purifi-
`cation experiment, 2 ␮g of each fraction was analyzed by
`immunoblot for SV2A, by using an SV2A-specific polyclonal
`antibody (14), and for the synaptic vesicle protein synaptophysin,
`by using an anti-synaptophysin monoclonal antibody from
`Chemicon. Analysis of the expression of SV2 isoforms in COS-7
`cells was performed by immunoblot against 13 ␮g (SV2A, SV2B,
`␤-gal), or 39 ␮g of total protein (SV2C), and probing with the
`SV2 monoclonal as described above.
`
`Binding of [3H]ucb 30889 to Brain Membranes, Heterologous SV2
`Proteins, and Immunoprecipitated SV2A. WT and KO membrane
`and synaptic vesicle preparation binding assays were performed
`as described (15), either by using 100 ␮g of protein (brain
`membranes) and 1.8 nM or by using 20 ␮g of protein (synaptic
`vesicle fractions) and 3.6 nM [3H]ucb 30889 per assay. Binding
`assays with solubilized and immunoprecipitated SV2A used
`similar conditions (1.8 nM [3H]ucb 30889). For binding exper-
`iments on the different SV2 isoforms, transfected cells were
`incubated for 2 h at 4°C in PBS with [3H]ucb 30889 (1.8 nM), in
`the presence or absence of 1 mM LEV. The assay was terminated
`by harvesting in a 24-well GF兾B filter plate (PerkinElmer) with
`rapid washing using 4°C PBS. IC50 curves of LEV, ucb 30889, and
`ucb L060 against hSV2A were determined by using confluent
`cells in 24-well plates and binding conditions as above, followed
`by rinsing the cells three times rapidly with 4°C PBS. After a final
`aspiration, 200 ␮l of 0.1 N NaOH was added to lyse the cells, and
`the samples were counted. For binding experiments on the larger
`series of LEV derivatives to hSV2A, aliquots of previously frozen
`transfected COS-7 cells, containing 2–3 ⫻ 104 cells, were incu-
`bated for 120 min at 4°C in 0.2 ml of a RPMI medium
`1640-Hepes 25 mM solution containing [3H]ucb 30889 (1.8 nM)
`and increasing concentrations of unlabeled competing drugs.
`The termination of the binding reaction by filtration and radio-
`activity counting was performed as described above. The ⫺log
`IC50s (pIC50s) were determined by nonlinear curve fitting.
`
`Audiogenic Seizure Mouse Model. Antiseizure activity of LEV and
`analogues was assessed in sound-susceptible mice (20) by ex-
`posing the mice to acoustic stimuli of 90 db, 10–20 kHz for 30 s,
`60 min after i.p. pretreatment. The reported ED50 values were
`obtained from testing of four to eight groups (n ⫽ 10) admin-
`istered different doses and reflect the potency of the compounds
`for inhibiting clonic convulsions.
`
`Lynch et al.
`
`Page 00002
`
`

`
`brain membranes, where it was demonstrated that labeling
`occurred to an ⬇90-kDa integral membrane protein (12). The
`synaptic vesicle localization of the binding site, along with the
`measured size and an integral membrane character led us to
`consider SV2 proteins as the primary candidate for the binding
`site.
`
`[3H]ucb 30889 Binds Only to Animal Brain Membranes and Synaptic
`Vesicles Containing the SV2A Isoform. To determine whether SV2
`is necessary for LEV and related compound binding, we mea-
`sured [3H]ucb 30889 binding to brain membranes from WT and
`SV2 KO mice. We found that [3H]ucb 30889 binds only to
`membranes from animals expressing SV2A (Fig. 2 A and B).
`Western analysis of brain membranes from WT and SV2 KOs,
`using a monoclonal antibody cross-reactive to all
`isoforms
`(SV2A, SV2B, and SV2C), or using a polyclonal specific to
`SV2A, confirmed the expected genotype, with no SV2A immu-
`noreactivity in the SV2A or SV2A兾B KOs (Fig. 2 A). There was
`no binding of [3H]ucb 30889 to brain membranes from animals
`lacking SV2A and no significant binding in the SV2A兾B double
`homozygote, which still expresses SV2C (Fig. 2B). Membranes
`from animals lacking only SV2B show binding to [3H]ucb 30889
`that is roughly equivalent to that seen in WT (Fig. 2B). However,
`we do measure a statistically significant difference between total
`binding to the WT and SV2B KO membranes (at P ⬍ 0.05 by
`two-tailed t test). We have confirmed this observation by Scat-
`chard analysis of binding to the WT and SV2B KO (data not
`shown), but no difference was observed between the affinities of
`[3H]ucb 30889 for the brain membranes from the WT or SV2B
`KO animals. Taken together, these results indicate that the
`observed binding in WT mouse brain is completely dependent
`on the presence of SV2A. However, the results raise the possi-
`bility that the presence of SV2B might affect the binding to
`SV2A. If there is a residual binding to SV2C in brain, it is so low
`that it cannot be responsible for the observed binding site in WT
`
`Photolabeling of the LEV-binding site in synaptic vesicles from rat
`Fig. 1.
`brain. After photoaffinity labeling with [3H]ucb 30889, SDS-solubilized pro-
`teins were run on an SDS兾PAGE gel, and then the gel was cut into thin slices
`and counted for 3H content. Photoaffinity labeling of crude synaptosomal (P2)
`(F) and synaptic vesicle (LP2) (䊐) fractions by using [3H]ucb 30889 is shown.
`Photoaffinity labeling identifies a protein of ⬇90 kDa enriched in the synaptic
`vesicle fraction.
`
`Results
`Photoaffinity Labeling of LEV-Binding Site in Synaptic Vesicles Labels
`a 90-kDa Protein. [3H]ucb 30889 is a photoactivable derivative of
`LEV that shows a higher affinity for the LEV-binding site in
`brain but otherwise behaves as a surrogate for LEV (15). This
`photoaffinity ligand was used to label the LEV-binding site in a
`synaptic vesicle fraction purified from rat brain membranes by
`using density centrifugation (Fig. 1). Gel electrophoresis (SDS兾
`PAGE) of the photo-cross-linked sample revealed that the
`radioligand labeled a protein with an approximate molecular
`mass of 90 kDa, consistent with results previously seen in crude
`
`PHARMACOLOGY
`
`Binding of [3H]ucb 30889 to WT and KO brain membranes and synaptic vesicles. (A) Western blot of brain membranes from WT and homozygous KO
`Fig. 2.
`mice probed with an anti-SV2 monoclonal antibody (cross-reactive to all isoforms) or with an anti-SV2A-specific polyclonal antibody. Lanes 1, WT; lanes 2, SV2A⫺/⫺
`KO; lanes 3, SV2B⫺/⫺ KO; lanes 4, SV2A⫺/⫺兾B⫺/⫺ double KO. (B) Binding of [3H]ucb 30889 to brain membranes from WT, SV2A⫺/⫺, SV2B⫺/⫺, and SV2A⫺/⫺兾SV2B⫺/⫺
`KO mice. Binding is observed only to membranes from animals expressing SV2A. 䊐, [3H]ucb 30889 alone; I, [3H]ucb 30889 plus 1 mM LEV . Error bars are the
`SD of experiments performed with five WT brains and four KO brains, with three replicates within each experiment. (C) Purification of synaptic vesicles enriches
`for the synaptic vesicle proteins and LEV binding. Shown are blots of mouse brain homogenate (H), crude synaptosomes (P2), plasma and heavy membranes (LP1),
`and synaptic vesicles (LP2) (2 ␮g of each fraction) that were probed for the synaptic vesicle proteins SV2A (Upper) and synaptophysin (Lower). The synaptic vesicle
`fraction from WT animals displays enrichment of both synaptic vesicle proteins and LEV-binding proteins, whereas material from SV2A KOs shows enrichment
`of synaptophysin only. (D) Binding to the different fractions using [3H]ucb 30889 shows significant binding only to the WT LP2 fraction, containing SV2A-rich
`synaptic vesicles. Shown are [3H]ucb 30889 alone (open bars) and [3H]ucb 30889 plus 1 mM LEV (filled bars). Shown are representative examples of two
`experiments. Error bars are the SD of two replicates.
`
`Lynch et al.
`
`PNAS 兩
`
`June 29, 2004 兩 vol. 101 兩 no. 26 兩 9863
`
`Page 00003
`
`

`
`Binding of [3H]ucb 30889 to heterologously expressed SV2 isoforms.
`Fig. 3.
`(A) Western blot with cross-reactive anti-SV2 monoclonal showing roughly
`equivalent amounts of hSV2A (lane A), hSV2B (lane B), and hSV2C (lane C).
`There is no detectable SV2 immunoreactivity in the ␤-gal-transfected cells
`(␤-gal, fourth lane). In the case of hSV2C, three times as much total protein was
`loaded as used for the other samples. (B) Binding of [3H]ucb 30889 to hSV2A,
`hSV2B, or hSV2C (three times the number of cells as other samples) transiently
`expressed in COS-7 cells. Significant binding is observed only to hSV2A, not to
`hSV2B or hSV2C. Shown are [3H]ucb 30889 alone (open bars) and [3H]ucb
`30889 plus 1 mM LEV (filled bars).
`
`animals based on the binding capacity in total mouse brain
`membranes.
`To determine whether LEV binds synaptic vesicle-associated
`SV2, we compared [3H]ucb 30889 binding across synaptic vesicle
`preparations from WT and SV2A KO mice (Fig. 2 C and D).
`Western blot analysis of the fractions obtained across the
`purification confirms the coenrichment of the synaptic vesicle
`proteins synaptophysin and SV2A with [3H]ucb 30889 binding.
`In contrast, the equivalent SV2A KO samples show enrichment
`of synaptophysin in the synaptic vesicle-enriched fraction LP2,
`but an absence of SV2A expression and [3H]ucb 30889 binding
`throughout all samples. The results confirm the synaptic vesicle
`localization of the LEV-binding site (12) and support the identity
`of the LEV-binding site as SV2A.
`
`hSV2A Is Sufficient for [3H]ucb 30889 Binding. To determine whether
`SV2A is solely responsible for the brain binding of LEV, we
`analyzed binding to SV2A expressed heterologously in nonneu-
`ral cells (Fig. 3). hSV2A, hSV2B, and hSV2C were transiently
`expressed in the fibroblast cell line COS-7, and the expression
`was verified by Western blot analysis (Fig. 3A). Note that the
`heterogeneous pattern of staining of all three isoforms in the
`Western blot, which differs somewhat from that observed in
`brain (Fig. 2 A), has been reported in the literature and attrib-
`uted to heterogeneity in the SV2 protein’s glycosylation (21). We
`generally observe that hSV2A and hSV2B have similar levels of
`expression in the COS-7 system, but that hSV2C expresses at a
`lower level. We thus corrected for the lower expression of hSV2C
`in the binding assay by increasing the amount of material added
`(3⫻) to yield approximately equivalent amounts of immunore-
`active protein in all three isoforms (as shown in Fig. 3B). There
`are significant levels of binding of [3H]ucb 30889 in cells
`expressing hSV2A, and this binding is displaced by excess LEV,
`indicating that it is specific. No statistically significant binding
`was observed under identical conditions to COS-7 cells trans-
`fected with hSV2B or hSV2C, or to COS-7 cells transfected with
`
`9864 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0308208101
`
`Binding of [3H]ucb 30889 to immunoprecipitated SV2A. (A) Western
`Fig. 4.
`blot of immunoprecipitated SV2A. Immunoprecipitation was performed with
`a goat antibody against SV2A, or normal goat IgG. Western blotting with the
`SV2A antibody revealed the presence of the protein in the former condition
`only. Mb, membrane; IP, immunoprecipitate. (B) Binding of [3H]ucb 30889 to
`SV2A purified by IP from a detergent-soluble fraction of rat cortical mem-
`branes. Binding occurs only to the DM-solubilized membrane or anti-SV2A
`immunoprecipitate, not to the control IgG immunoprecipitate. Shown are
`[3H]ucb 30889 alone (open bars) and [3H]ucb 30889 plus 1 mM LEV (filled bars).
`Shown is a representative example of three experiments.
`
`a vector encoding ␤-gal (Fig. 3B), consistent with the interpre-
`tation that SV2A is the sole LEV-binding SV2 isoform.
`
`[3H]ucb 30889 Binds to Immunoprecipitated SV2A. We next deter-
`mined whether [3H]ucb 30889 binds to native SV2A immuno-
`precipitated from membrane extracts. Rat cortical membranes
`were solubilized with DM and SV2A immunoprecipitated with
`a selective goat polyclonal antibody. Immunoprecipitation was
`confirmed by a Western blot analysis, which showed the presence
`of large amounts of SV2 immunoreactivity in the starting extract,
`and in the immunoprecipitated sample, but not in the IgG
`control sample (Fig. 4A). Binding to [3H]ucb 30889 was observed
`in the anti-SV2A immunoprecipitated sample, but not in the
`negative control (Fig. 4B). This result suggests a direct interac-
`tion between [3H]ucb 30889 and the immunopurified SV2A
`protein.
`
`Compound Affinities to SV2A Correlate with Affinities to the LEV-
`Binding Site in Brain and to Antiseizure Potencies. To characterize in
`more detail the nature of the binding interaction of LEV and
`derivatives to the SV2A protein, we measured their relative
`affinities for hSV2A. In experiments testing the ability of
`unlabeled compounds to displace [3H]ucb 30889 from hSV2A
`expressed in COS-7 cells, the affinities (pIC50s) of ucb 30889,
`LEV, and LEV’s enantiomer, ucb L060 (7.2, 5.7, and 3.6,
`respectively), show the same rank order and similar values (Fig.
`5A) to those previously reported in studies of LEV binding to rat
`brain membranes (11, 15). Critically, ucb L060 binds with
`significantly less affinity to hSV2A than does LEV. The stereo-
`selectivity for LEV over its opposite stereoisomer, ucb L060, is
`a key characteristic of the binding site in brain and was confirmed
`by these studies.
`
`Lynch et al.
`
`Page 00004
`
`

`
`PHARMACOLOGY
`
`other hand, do not display seizures or other gross phenotypes
`(17) (SV2C KO mice have not been reported). The severe
`disability of the homozygous SV2A KO mice has prevented us
`from performing certain desirable experiments, such as testing
`the effects of LEV on seizures in these animals, or indeed, fully
`characterizing the seizures by using implantable electrodes. Both
`inhibitory (16) and excitatory (V. Lopantsev and S.M.B., un-
`published results) neurotransmission is reduced in the absence of
`SV2A. Likewise, studies in adrenal chromaffin cells from SV2A
`KO mice revealed reduced calcium-stimulated exocytosis, sug-
`gesting that the release probability of synaptic vesicles is reduced
`in the absence of SV2A (22). Also, hippocampal neurons
`cultured from mice lacking both SV2A and SV2B demonstrate
`altered activity-dependent synaptic depression (17). These ex-
`amples of altered neurotransmission were all observed in the
`absence of changes in either synapse or synaptic vesicle density
`or morphology (16, 17, 22). Together these observations suggest
`that SV2A acts as a modulator of vesicle fusion although it is
`possible that SV2A has additional functions at the presynaptic
`terminal.
`The molecular action of SV2A is unknown. The SV2s are
`twelve transmembrane integral membrane glycoproteins with a
`significant homology to the major facilitator superfamily (MFS)
`of transporters found in both bacteria and eukaryotes (23).
`Given their universal presence in synaptic vesicles, it has been
`proposed that the SV2s might transport a common constit-
`uent of the vesicles, such as calcium or ATP (14). However, no
`transport function has yet been found to rely on an SV2 protein.
`SV2A interacts with the presynaptic protein synaptotagmin,
`considered the primary calcium sensor for regulating calcium-
`dependent exocytosis of synaptic vesicles, and may affect syn-
`aptotagmin’s function (19, 24). Also, SV2 proteins seem to con-
`tribute the majority of sugar side chains to the lumen of synaptic
`vesicles and perhaps are the source of a neurotransmitter-binding
`matrix hypothesized to exist in the synaptic vesicle interior (25).
`Given the growing recognition that many, if not most, proteins
`have multiple functions (26), it is possible that SV2A has several
`functions, including those discussed above, but perhaps others as
`well. The availability of a compound that acts through SV2A
`should provide a powerful tool to probe the structure and
`function of this protein and may contribute to the understanding
`of the regulation of neurotransmitter release.
`The identification of the synaptic vesicle protein SV2A as the
`binding site in brain for LEV has important implications not only
`for the antiepileptic mechanism of action of LEV, but also for
`future drug discovery in epilepsy and other neurological condi-
`tions. The correlation between antiseizure properties of LEV
`derivatives and their affinity for SV2A strongly suggests a
`mechanistic link between the two. There are reports of other
`effects of LEV, including the partial inhibition of N-type voltage-
`gated Ca2⫹ channels and the reduction of inhibition of ␥-amino-
`butyric acid (GABA)- and glycine-gated currents, induced by
`Zn2⫹ and ␤-carbolines (9). Currently, it is unclear whether these
`effects are mediated by the observed interaction with SV2A, or
`by alternate mechanisms. In either case, we believe that the
`correlation between SV2A binding and drug potency suggests
`that LEV is modulating one or more of the functions of SV2A,
`and correspondingly contributing to its efficacy in treating
`epilepsy.
`Determining the effect of LEV on SV2A function is compli-
`cated by both the absence of proven SV2A functions, and also by
`the unusual pharmacology of LEV, which lacks effects on the
`electrophysiology of normal brain tissue and neurons (4, 27), and
`on standard amino acid neurotransmitter release as studied by
`microdialysis in normal brain (28). The apparent lack of effect
`on normal electrophysiology has implications for any hypothesis
`of LEV’s effects on the function(s) of SV2A. We do not
`anticipate LEV to affect SV2A functions that are critical to
`June 29, 2004 兩 vol. 101 兩 no. 26 兩 9865
`PNAS 兩
`
`Fig. 5. Binding affinities of selected LEV derivatives for human SV2A. (A) IC50
`curves of LEV, ucb L060, and ucb 30889 against hSV2A transiently expressed in
`COS-7, using [3H]ucb 30889. Shown are LEV (‚), ucb 30889 (I), and ucb L060
`(F). Error bars are SEM, n ⫽ 3. (B) Correlation of binding of a series of LEV
`compounds to mouse brain and to hSV2A. Shown are pIC50s measured by using
`[3H]ucb 30889. There is a high degree of correlation between binding affinity
`of these compounds to hSV2A and to mouse brain membranes. pIC50 values
`are the mean of two independent experiments, where each determination lies
`within 0.2 log units of the mean.
`
`Testing the binding of LEV and additional analogs to hSV2A
`expressed in COS-7 cells revealed that pIC50s are highly corre-
`lated with the values obtained in mouse brain (r2 ⫽ 0.98) (Fig.
`5B) and rat brain membranes (data not shown). There was also
`a clear correlation between the affinities of these compounds for
`hSV2A in COS-7 and the potency of their antiseizure protection
`in the mouse audiogenic model of epilepsy (r2 ⫽ 0.84) (Fig. 6).
`These data are consistent with the previous report of a corre-
`lation between binding of LEV analogs in rat brain and antisei-
`zure potency (11). We also investigated the binding of other
`AEDs, including valproate, carbamazepine, phenytoin, ethosux-
`imide, felbamate, gabapentin, tiagabine, vigabatrin, and zoni-
`samide. None of these AEDs, at concentrations up to 100 ␮M,
`competed with [3H]ucb 30889 for binding to SV2A (data not
`shown), consistent with previous studies testing AEDs against
`the LEV-binding site in rat brain (11, 15).
`
`Discussion
`We have identified SV2A as the binding site for the antiepileptic
`drug LEV and as a potential target for CNS therapy. Intrigu-
`ingly, SV2A KO mice seem normal at birth, but develop an
`unusually strong seizure phenotype by 1.5 weeks of age, and
`usually die within 3 weeks after birth (16). SV2B KOs, on the
`
`Fig. 6. Correlation between binding affinity and antiseizure potency of LEV
`derivatives. Correlation of binding of a series of LEV-related compounds to
`hSV2A assayed in transiently transfected COS-7 cells (pIC50s measured by using
`[3H]ucb 30889), and of antiseizure potencies shown as the ⫺log ED50s (pED50s)
`in the mouse audiogenic seizure model. There is a good correlation between
`antiseizure potency in audiogenic mice and affinity to hSV2A.
`
`Lynch et al.
`
`Page 00005
`
`

`
`normal physiology of the neuron but rather think LEV might
`modulate a function of SV2A present only under pathophysio-
`logical conditions. It is thus unlikely that LEV alters synaptic
`release in normal brain and neurons, and indeed, there is no
`evidence for such an effect.
`The strong seizure phenotype observed in the SV2A KO
`animals supports the interpretation that SV2 can influence
`mechanisms of seizure generation or propagation. The fact that
`the SV2A KO mice exhibit seizures, while LEV inhibits seizures,
`suggests that LEV is not acting simply as an antagonist of SV2A
`function (or LEV would presumably act as a proconvulsant
`instead of an anticonvulsant). It is possible that LEV binding
`enhances a function of SV2A that inhibits abnormal bursting in
`epileptic circuits, a function whose loss in the SV2A KOs results
`in seizures. Alternately, the