Uncovering the Binding Mode of #-Secretase Inhibitors

Manuel Hitzenberger, and Martin Zacharias
ACS Chem. Neurosci., Just Accepted Manuscript • DOI: 10.1021/acschemneuro.9b00272 • Publication Date (Web): 21 Jun 2019
Downloaded from http://pubs.acs.org on June 25, 2019

Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscriptsis published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036

Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Uncovering the Binding Mode of -Secretase
Inhibitors
Manuel Hitzenberger and Martin Zacharias

Physics Department T38
Technical University of Munich
James-Frank-Str. 1
85748 Garching, Germany

Abstract

Knowledge of how transition state inhibitors bind to -secretase is of major importance
for the design of new Alzheimer’s disease therapies. Based on the known structure of secretase in complex with a fragment of the amyloid
precursor protein we have generated
a structural model of -secretase in complex with the e ective L-685,458 transition state
inhibitor. The predicted binding mode is in excellent agreement with experimental data,
mimicking all enzyme-substrate interactions at the active site and forming the relevant
transition state geometry with the active site aspartate residues. The model also indicates

17 the possible location and the amino acid residues of the proposed binding pockets S1′,
18 S2′ and S3′ near the active site that are occupied by chemical groups of the inhibitor. In
19 addition, we found that the stability of the complex is very likely sensitive to the pH value.
20 Comparative simulations on the binding of L-685,458 and the epimer L682,679 allowed
21 us to explain the strongly reduced a nity of the epimer for -secretase. The structural
model could form a valuable basis for the design of new or modi ed -secretase inhibitors.
24
25 Keywords: -secretase drug binding, enzyme inhibition, enzyme dynamics, -secretase tran-
26 sition state inhibition, ligand-receptor docking

29 Article

33 The intra-membrane cleaving hetero-tetramer -secretase (GSEC) processes the C-terminal
34 fragment of the amyloid precursor protein (APP), C99.1{6 Since some of the resulting cleav-
age products are strongly linked to Alzheimer’s disease (AD), GSEC is at the focus of many
37 drug-design research e orts.7{11 Ultimately, one wishes to control or modulate C99 processing
38 to prevent the progression of, or even cure, AD.12{17 A large number of experimental4,7{9,18{24 as well as theoretical25{34 studies have been
undertaken to investigate and characterize the
41 structure, biology and chemistry of GSEC and its substrates. Not long ago, the structurof GSEC (inactive mutation) in complex with Notch and
C83 substrates (a shortened APP
44 fragment) have been solved.23,24 Also recently, the binding mode of the pseudo-inhibitor
45 DAPT has been ascertained by computational methods.35
46
47 A putative structure of the complex between the low-a nity TSA inhibitor L-682-679 and
48 the PS1 homologue PSH has been published by Dang et al. in 2015.36 The atoms assigned to
49 the measured electron density, however, can not explain the mode of action of TSA inhibitors,
50
51 since none of the molecular features (especially not the the transition state mimichking OH
52 group) were in contact with the enzyme’s active site. Therefore, a detailed high resolution
53 54
51 structure or molecular model of how an e ective transition state inhibitor binds to presenilin
55 (PS), the catalytically active domain of GSEC, is still missing. Such a structure could be
56 very valuable for future modulator or inhibitor design strategies.
57 58
55 Based on the recent GSEC-substrate complex structures,23,24 we performed molecular dock-
59 ing, atomistic molecular dynamics (MD) simulations and free energy calculations to inves-
6 tigate the binding mode of the transition state analogue (TSA) inhibitor L-685,45816 (see
7
8 Figure 1). The predicted binding pose of the inhibitor matches the binding geometry of C99
9 residues L49, V50, M51 and L52 and also mimics the beta-sheet formation observed for sub-
10 11
8 strate binding near the active site of the enzyme.
12 We investigated all four possible presenilin active site (D257 and D385) protonation states
13 and found that L-685,458 binds stably to three of them. The binding a nity, however, de-
14
15 creased with increasing residual charge on the active site residues, indicating pH sensitivity.
16 The greatly reduced inhibitory performance16 of L-682,679, an epimer of L685,458 (see Figure
17 1 for comparison) can also be explained by our study.
39 Figure 1: Structures of the TSA inhibitors L-685,458 and L-682,679. The three side chains occupying the
40 proposed S1′, S2′ and S3′ pockets are termed “R1″, “R2″, “R3″, respectively. The other two large structural
41 42
39 moieties of the ligand, pointing towards PS TMDs 2 and 3 in the bound structure, are denoted as “H” (“head group”). Left panel: Skeletal
formula of the TSA inhibitor L-685,458. Right panel: Superposition of the
43 optimized geometries of L-685,458 (grey) and L-682,679 (cyan).
44 Starting structures for inhibitors L-685,458 and L-682,679 were generated by geometry op-
45
46 timization (on B3LYP37 level, with the TZVP38 basis set, see Supplementary Information).
47 Alignment of L-685,458 to GSEC-bound substrate residues V50, M51 and L52 (coordinates
48 taken from a short simulation based on PDB structure 6IYC24) resulted in a placement of the
49
50 inhibitor’s OH group, coinciding with the carbonyl oxygen of the scissile peptide bond in the
51 GSEC-substrate complex. Additionally, the R1-R3 side chains of the inhibitor pointed in the
52
53 same direction as the corresponding side chains of the APP fragment (V50, M51 and L52),
54 hence also reproducing the down-up-down sidechain sequence of the enzyme-bound substrate.
55 This placement, exactly mimicking the local substrate – enzyme interactions, appeared to be
56
57 unique since in extensive docking searches no other sterically feasible placement could be
58 identi ed with the OH group in a state of mimicking a transition state scenario.
59 To investigate the dynamics and stability of the docked geometry, MD simulations (1000 ns,

6 including a POPC membrane and water) were performed for all four possible active site pro-
7
8 tonation states.
9 Creating starting structures for the S-epimer proved to be more challenging, because only
10 11
8 rotamers di erent from the energy-minimized structure could t in a similar fashion as L-
12 685,458. We selected two di erent ligand rotamers which were able to form many ligand-
13 enzyme interactions for further simulation: One, that was closer to the energy-minimum
14
15 structure of L-682,679 (“P1″) and an alternative one that was more similar to the L-685,485-
16 GSEC starting structure (“P2″) (Supporting Information, Figure 1). For each pose we gen-
17 erated simulation trajectories of 500ns, totaling 1000ns for every investigated PS protonation
18
19 state.
20 The simulations where both catalytic aspartates were protonated are referred to as either
21 22
19 “R-DP” (L685,458) or “S-DP” (L682,679). Simulations with a single protonated active site
23 are termed “R-D257″ and “S-D257″ (protonated D257) or “R-D385″ and “S-D385″ (proto-
24 nated D385), respectively. The states with non-protonated active site residues are denoted
25
26 as “R-NP” or “S-NP”.
27 In simulations with at least one protonated catalytic residue, inhibitor L-685,458 maintains
50 Figure 2: GSEC – L-685,458 interactions. Left panel: Interaction diagram (adapted from LigPlot+ output).
51 The green lines depict hydrogen bonds between the enyzme (blue bonds) and the ligand (grey bonds). Residues
52 involved in hydrophobic interactions are shown in red.
53 Right panel: Representative snapshot from simulation R-D385. Hydrogen bond colors correspond to the type
54 of the donor atom, GSEC backbone is depicted in green. Carbons of GSEC residues are colored in cyan, those
55 of L-685,458 are grey.
56
57 a stable hydrogen bonding network with PS residues G382, K380 and L432 throughout nearly
58 all of the accumulated 3 s of simulation time. Additionally, the OH group situated at the
59 epimeric center binds strongly to the catalytic residues of GSEC (see Figure 2 for details).

6 Most of the ligand is situated between TMDs 8 and 9 (where Notch and C83 (C99) form
7
8 beta-sheets with the enzyme23,24) and only a small part of the molecule protrudes into the
9 cavity formed by TMDs 2, 3 and 5. This cavity, however is the region where the substrate
10 11
8 helices bind and the fact that this region is virtually not occupied by the inhibitor explains
12 why it is possible to simultaneously bind a TSA and the natural substrate.14,15,40 Therefore
13 only beta-sheet formation and subsequent access to the active site are impaired by the TSA
14
15 and not substrate binding itself. Figures 3 a) and b) illustrate that L-685,458 very closely
16 resembles the substrate when it is in position to be cleaved by the enzyme. Superposition
17 of L-685,468 (from simulation R-D385) and the GSEC-bound APP fragment (taken from a
18
19 short simulation of the complex between the natural substrate and the WT enzyme in the
20 same protonation state), shows that the TSA occupies exactly the same binding cavities as
21 22
19 the substrate, explaining the high a nity and inhibitory e ciency of L-685,458.
55 Figure 3: (a) Superposition of GSEC structures containing either the natural substrate or L-685,458 (a
56 rotated close up view shown in (b)). Superposition involves the backbone atoms of the substrate binding site
57 of PS. GSEC (green) and APP (orange) structures shown are from a short simulation of the native enzyme-
58 substrate complex. The inhibitor (gray) is taken from simulation R-D385. (c,d) Location of the side chain
59 binding pockets S1′, S3′ (occupied by R1 and R3, blue) and S2′ (occupied by R2, orange).

R1, R3 R-D385
R2
Head
R1, R3 S-D385
R2
Head
K380 100% (100%) T421 100% (100%) G384 100% (97%) A431 100% (97%) L425 100% (99%) T147 100% (99%)

A431 100% (99%) L381 100% (100%) D257 100% (96%) L271 99% (90%) A434 100% (99%) G384 100% (97%)
V272 100% (98%) L425 100% (100%) D385 100% (95%) L432 99% (83%) T421 100% (98%) D385 100% (96%)
L432 100% (96%) A434 100% (100%) L435 100% (90%) K380 96% (90%) V379 100% (92%) F388 95% (76%)
P433 100% (92%) L422 100% (99%) L268 99% (95%) V272 94% (83%) L422 100% (95%) L268 90% (82%)
I287 99% (91%) L432 100% (91%) F388 99% (90%) L268 84% (81%) L381 99% (99%) L435 83% (67%)
A434 96% (23%) K380 100% (88%) I253 98% (88%) P433 84% (50%) L432 99% (92%) Y256 77% (65%)
V261 95% (89%) V379 99% (86%) Y256 94% (76%) L381 66% (30%) K380 95% (67%) D257 74% (70%)
T281 94% (68%) L418 73% (25%) I143 91% (58%) I287 61% (48%) L85 88% (63%) G382 74% (47%)
L268 92% (71%) L85 72% (14%) G382 70% (36%) V261 56% (42%) L418 79% (36%) I253 71% (62%)
L282 89% (87%) L150 60% (45%) G382 55% (32%) I143 71% (39%)
L271 86% (48%) F283 60% (48%) F283 52% (37%) L271 70% (57%)
A275 83% (75%) T147 47% (27%) V379 45% (21%) V151 63% (46%)
D257 79% (4%) L271 44% (24%) L282 44% (26%) P267 62% (56%)
L381 75% (33%) A434 43% (0%) A434 31% (1%) A260 61% (36%)
F283 59% (31%) I287 38% (25%) A275 31% (16%) F283 54% (44%)
L425 51% (0%) M146 33% (23%) D257 27% (3%) L282 46% (36%)
G382 43% (4%) A260 32% (6%) L150 40% (29%)
Q276 26% (23%) L383 40% (29%)

23 Table 1: Interaction frequencies (in percent of sampled frames) between PS residues and R1-R3 of the bound
24 25
23 inhibitor as well as the head group of the ligand (data taken from simulations R-D385 and S-385). Only amino acids that are within 5A of one of the side chains in at least 25% of sampled frames are shown here. Values in
parentheses correspond to the interaction frequencies with a 4A cut-o .
26 The residues in bold font can also be found within 5 A of the corresponding C83 side chains (V50, M51, L52)
27 in the PDB structure 4IYC,24 while italics indicate that the residue is within 5 A of the respective side chain
28 in the -secretase – Notch structure23 (Notch residues L1755, L1756, S1757).
29 *) Sites of known FAD mutations.
32 The structure of the bound inhibitor also revealed the location and constitution of the
33 S1′, S2′ and S3′ binding pockets in GSEC, proposed by Wolfe et al.4 Interestingly, instead
34
35 of three separated pockets, we identi ed only two: One S2′ pocket, which is very con ned
36 and situated close to TMDs 8 and 9 (see also Figure 3c and d). This pocket is permanently
37 occupied by the smaller side chain R2. The putative binding pockets S1′ and S3′, on the
38
39 other hand are part of the same spacious, hydrophobic cavity in which side chains R1 and R3
40 bind simultaneously (cf. with Figure 3c and d). These two ndings o er an explanation for
41 42
39 experimental results according to which the S2′ pocket is very sensitive to the size of the side
43 chain, while S1′ and S3′ are not.4,41 Based on the interaction frequencies observed during
44 the simulations it is possible to obtain a statistical overview over the nature of the residues
45
46 constituting pockets S1′, S2′ and S3′ (Table 1).
47 The interaction frequencies given in Table 1 also indicate that the S1′,S3′ pocket is a little
48 bit more transient than the the more rigid S2′ pocket where every residue is either in 72 to
49
50 100% of sampled frames within 5A of the side chain or in less than 25% (R-D385 simulation).
51 As can be expected, the interaction frequencies calculated for the less stable S-D385 bind-
52 ing mode are generally lower than in the corresponding R-D385 simulation, since the ligand
53
54 is more mobile in the S-epimer case. This is, however, not true for the S2′ pocket, where
55 binding seems to be of similar stability in both simulations. It reinforces the notion that
56
57 this interaction sites is an important stabilizing factor for the ligand-receptor complex and
58 may also serve a pronounced role in the stabilization of the transition state in the natural
59 substrate-enzyme complex.

6 The constitution of the S1′, S2′ and S3′ sites found in the inhibitor simulations is very similar
7
8 to the environment of the corresponding residues in the Cryo-EM structures of both the C83-
9 and the Notch-Gsec complex, further suggesting that these binding pockets are very distinct
10 11
8 (all residues constituting the S1′, S2′ and S3′ pockets in the Cryo-EM structures can be found
12 in Supporting Information, Table 1).
13 In simulations featuring L-685,458 only R-NP exhibited permanent, partial dissociation of
14
15 the ligand. Here, the hydrogen bonds to G382 and K380 were disrupted after just 80ns of
16 simulation time. This is likely due to electrostatic repulsion of D257 and D385 (both are neg-
17 atively charged in the R-NP simulation), thereby tearing apart the beta sheet and attracting
18
19 a larger number of water molecules into the binding site.
20 According to experimental measurements,16 the a nity of GSEC for L-682,679 is much lower
21 22
19 than for L-685,458 (IC50 value of >10000 nM, vs 17 8 nM). In accordance with experiment,
23 also in our simulations the enzyme-ligand stability was greatly reduced in the L-682,679-
24 GSEC complex: In all simulations, partial dissociation events took place. Most frequently
25
26 the beta-sheet was disrupted but also dissociation from the active site aspartates took place
27 in more than one case (Supporting Information, Figure 2 shows a representative snapshot of
28 L-682,679 binding).
29
30 The di erences in a nity of the two epimers is also re ected by molecular mechanics gener-
31 alized born surface area (MMGBSA) calculations, conducted on either the rst 100ns (500
32 frames) of each simulation or the complete 1000ns trajectories (500 frames): The binding
33
34 energies of the L-685,458 (R-epimer)-GSEC complexes are consistently more favorable for all
35 simulations where at least one of the aspartates was protonated. The results are given in
36
37 Table 2 and indicate that the R-DP simulation features by far the most stable enzyme-ligand
38 complex if the complete trajectories are taken into account. Due to the large system size, no
39 conformational entropy change estimations have been applied, therefore absolute numbers are
40
41 not comparable to experiment. However, the predicted relative binding free energy di erence
42 between L-685,458 and L-682,679 amounts to 5-10 kcal/mol and agrees qualitatively well with
43 experiment (1000 fold di erence in Kd corresponds to a binding free energy di erence of 4.2
44
45 kcal/mol)
46 Furthermore, we calculated the mean number of water molecules within 4A of the ligand
47 48
45 to estimate the level of desolvation. The results show, that there is a large gap in solvent
49 accessibility between the high a nity (R-DP, R-D257, R-385) and the low a nity (R-NP,
50 S-DP, S-D257, S-D385, S-NP) complexes (Table 2), with the former being signi cantly more
51
52 desolvated.
53 Calculation of the root mean square uctuations of the ligand painted the same picture: Sim-
54 ulations of the high a nity complexes, featured smaller ligand uctuations, indicating more
55
56 stable binding (ligand RMSD plots for each simulation are given in the Supporting Informa-
57 tion, Figure 3). In summary, especially at pH values below pH 7 L-685,458 is expected to bind
58 very stable to PS, closely mimicking the substrate, while the predicted a nity of L-682,679

5
System G(100ns) G(1 s) Mean # H2O RMSF in A
R-DP
R-D257 R-D385 R-NP -80.61 0.21
-81.77 6.16
-79.74 0.20
-74.31 0.36 -82.13 0.20
-67.74 0.56
-74.24 0.27
-55.51 0.35 7.76 2.22
10.90 3.30
11.88 2.96
31.25 6.42 0.89
1.48
1.44
1.72
S-DP P1
S-DP P2 -69.39 0.21
-57.08 0.33 -64.12 0.30 16.83 5.04 2.04
S-D257 P1
S-D257 P2 -60.22 0.24
-44.50 0.47 -58.53 0.32 22.37 7.38 2.45
S-D385 P1
S-D385 P2 -74.02 0.24
-73.11 0.19 -63.49 0.35 18.47 4.47 1.91
S-NP P1
S-NP P2 -58.51 0.25
-61.45 0.44 -58.27 0.33 22.27 7.25 2.00

22 Table 2: Summary of MMGBSA results, average water hydration numbers and ligand RMSFs. MMGBSA
23 results are given for the rst 100ns of each simulation ( rst results column) and all available simulation frames
24 (second results column). Errors are given as standard errors (MMGBSA) or standard deviation (H2 O and
25 RMSF)
26 *) P1 und P2 trajectories have been merged for evaluation.
27
28 is greatly reduced (in agreement with experiment).
29
30 Binding of the inhibitors to the enzyme active site involves slight conformational changes
31 of the inhibitors to accommodate to the binding regions. In order to estimate this contri-
32 bution we optimized the geometries of both ligands (bound state) to the nearest minimum
33
34 (B3LYP/TZVP-level in vacuo). We found that the energy di erence is higher for the low
35 a nity inhibitor L-682,679 (97.89 kcal/mol) compared to L-685,458 (91.39 kcal/mol). This
36 37
34 contribution further disfavors binding of L-682,679 and explains the observed tendency to
38 perturb the binding region and to initiate complex dissociation.
39 Based on the identi ed binding mode it is now possible to make predictions regarding the
40
41 possible improvement of inhibitory e ciency: For example, the amide tail of bound L-685,458
42 is close to the polar side chains of PS residues K380 and Q276 (minimum distances of approx.
43 2.8 and 3 A in simulation R-D385, respectively). Hence, a hydrogen bond acceptor instead of
44
45 the NH2 group at this position could further increase a nity and speci city of the inhibitor.
46 Another potential improvement is enabled by the proximity between T421 and the R2 side
47 48
45 chain: If adapted to contain a hydrogen bond acceptor R2 might be able to form a hydrogen
49 bond with T421 (see also Supporting Information, Figure 4).
50 In conclusion, the ligand binding mode predicted by this study is in excellent agreement with
51
52 available experimental data and can serve as a starting point to systematically explore pos-
53 sible GSEC ligands and to rationally modify existing inhibitors.
54 However, due to the similarity of the bound Notch and C83 structures23,24 and the stability
55
56 of the backbone-backbone interactions between the inhibitor and the enzyme it seems to be
57 very challenging to craft TSA -secretase inhibitors speci cally targeting only a single type
58 of substrate.

9 Materials and Methods section, Table 1, Figures 1-3. For further reference we also provide a
10 PDB le containing the structure of the predicted -secretase – L685,458 complex.

14 Author Informations
16
17 M.H. performed research, analyzed data, and wrote the article; and M.Z. designed research
18 and wrote the article.

21 Acknowledgments

24 We thank Harald Steiner and Dieter Langosch for helpful discussions. Financial support by
25 the DFG (German Research Foundation) grant FOR 2290 (project P7) is gratefully acknowl-
26 edged. Computer resources for this project have been provided by the Gauss Centre for
27
28 Supercomputing/Leibniz Supercomputing Centre under grant pr27za.

Table of Content Graphics
Figure 4: For Table of Contents Only

6 References

9 [1] D. Langosch, C. Scharnagl, H. Steiner, and M. K. Lemberg. Understanding intramem-
10 brane proteolysis: from protein dynamics to reaction kinetics. Trends Biochem. Sci.,
11 40:318{327, 2015.

13 [2] H. Steiner, A. Fukumori, S. Tagami, and M. Okochi. Making the nal cut: pathogenic
15 amyloid- peptide generation by -secretase. Cell Stress, 2:292{310, 2018.

17 [3] B. De Strooper, T. Iwatsubo, and M. S. Wolfe. Presenilins and -secretase: structure,
18 function and role in alzheimer’s disease. Cold Spring Harb. Perspect. Med 2, a006304,

20 2012.
21
22 [4] D. M. Bolduc, D. R. Montagna, M. C. Seghers, M. S. Wolfe, and D. J. Selkoe. The
23 amyloid beta forming tripeptide cleavage mechanism of -secretase. eLife, 5:e17578,
24
25 2016.
26
27 [5] Y. Qi-Takahara, M. Morishima-Kawashima, Y. Tanimura, G. Dolios, N. Hirotani,
28 Y. Horikoshi, F. Kametani, M. Maeda, T. C. Saido, R. Wang, and Y. Ihara. Longer
29 30
27 forms of amyloid protein: implications for the mechanism of intramembrane cleavage by
31 -secretase. J. Neurosci., 25:436{445, 2005.
32
33 [6] M. Takami, Y. Nagashima, Y. Sano, S. Ishihara, M. Morishima-Kawashima, S. Fu-
34 namoto, and Y. Ihara. -secretase: successive tripeptide and tetrapeptide release from
35
36 the transmembrane domain of -carboxyl terminal fragment. J. Neurosci., 29:13042{
37 13052, 2009.
38
39 [7] M. Szaruga, B. Munteanu, S. Lismont, S. Veugelen, K. Horr e, M. Mercken, T. C. Saido,
40
41 N. S. Ryan, T. De Vos, S. N. Savvides, R. Gallardo, J. Schmykowitz, F. Rousseau, N. C.
42 Fox, B. De Strooper, and L. Chavez-Guti errez. Alzheimer’s-causing mutations shift a
43 length by destabilizing -secretase a n interactions. Cell, 170:443{456, 2017.
44
45 [8] T. Xu, Y. Yan, Y. Kang, Y. Jiang, K . Melcher, and H. E. Xu. Alzheimer’s disease-
46
47 associated mutations increase amyloid precursor protein resistance to -secretase cleavage
48 and the a 40/a 42 ratio. Cell Discovery, 1:16026, 2016.
49
50 [9] L. Sun, R. Zhou, G. Yang, and Y. Shi. Analysis of 138 pathogenic mutations in presenilin-
51
52 1 on the in vitro production of a 42 and a 40 peptides in -secretase. Proc. Natl. Acad.
53 Sci., 114:E476{485, 2016.
54
55 [10] T. Iwatsubo, A. Okada, N. Suzuki, H.Mizusawa, N. Nukina, and Y. Ihara. Visualization
56 57
55 of a 42 (43) and a 40 in senile plaques with end-speci c a monoclonals: evidence that L-685,458
58 an initially deposited species is a 42 (43). Neuron, 13:45{53, 1994.