Proc. Natl. Acad. Sci. USAVol. 92, pp. 816-820, January 1995Biochemistry

Solution structure of pleckstrin homology domain of dynamin byheteronuclear NMR spectroscopyDAVID FUSHMAN*, SEAN CAHILL*, MARK A. LEMMONt, JOSEPH SCHLESSINGERt, AND DAVID COWBURN*t*The Rockefeller University, 1230 York Avenue, New York, NY 10021; and tDepartment of Pharmacology, New York University Medical Center,New York, NY 10016

Communicated by David Luckl The Rockefeller University, New Yorkl NY October 26, 1994

ABSTRACT The pleckstrin homology (PH) domain is arecognition motif thought to be involved in signal-transduction pathways controlled by a variety of cytoplasmicproteins. Assignments of nearly all 'H, 13C, and '5N reso-nances of the PH domain from dynamin have been obtainedfrom homonuclear and heteronuclear NMR experiments. Thesecondary structure has been elucidated from the pattern ofnuclear Overhauser enhancements, from 13C chemical shiftdeviations, and from observation of slowly exchanging amidehydrogens. The secondary structure contains one a-helix andeight 8-strands, seven of which are arranged in two contigu-ous, antiparallel 8-sheets. The structure is monomeric, incontrast to the well-defined intimate dimerization of thecrystal structure of this molecule. Residues possibly involvedin ligand binding are in apparently flexible loops. Steady-state15N{'H} nuclear Overhauser effect measurements indicateunequivocally the boundaries of this PH domain, and thestructured portion of the domain appears to be more extendedto the C terminus than previously suggested for other PHdomains.

The pleckstrin homology (PH) domain is a modular protein of'110 amino acids (1) that has been suggested to bind with

affinity to certain lipids (2, 3), or possibly to other phosphor-ylated ligands including 13,y subunits of guanine nucleotide-binding proteins (G proteins) and protein kinase C (4).Interactions between PH domains and their ligands arethought to mediate the formation of complexes involved inintracellular signal transduction, although well-characterizedinstances are as yet unknown.From the sequence analysis of 3BP2, a Src homology 3

binding protein, Mayer et al. (1) have shown that one domainof 3BP2 contains a section with considerable sequence ho-mology to several other proteins. Such a homology is sugges-tive that this domain may play a similar role to that of Srchomology regions 2 and 3 in recognizing different proteinareas in signal transduction (5). A very large number ofpossible PH domains have now been identified, but theircomparative similarity scores are quite low, and a definitiveclass of absolutely conserved sequences is still hard to define(6). The solution structures of the PH of spectrin (7) and of theN terminus of pleckstrin itself (2) have been reported, as hasthe crystal structure (8) of the target of this paper, the PHdomain of dynamin.One mutation in a putative PH domain has been linked to

genetic immunodeficiency (9). A hypothetical suggestion isthat PH domains (or a subclass) are associated with cell surfacereceptors not containing kinases and act as cofactors ormodifiers to recruited enzymic activities. Two clear examplesof such are the ,B-adenergic receptor kinase (reviewed in ref.10) and the insulin receptor substrate 1 (reviewed in ref. 11).It has been suggested that this may occur specifically to G

proteins (6), but there is no evidence to rule out direct PHinteractions in, for example, insulin receptor substrate 1binding to the insulin receptor or 4PS binding to the interleu-kin 4 receptor (11). The similarity of the structure of the PHN domain of pleckstrin to that of retinol binding protein (2)has led to the suggestion that lipids may be the natural ligandsand the observation that PH domains bind to lipid vesicles,especially those containing phosphatidylinositol 4,5-bis-phosphate (PIP2) (3). The lack of a large set of mutants of PHdomains with defined pathology may reflect the essential roleof PH domains, rather than the opposite.Dynamin is a 100-kDa GTPase involved in the initial stages

of endocytosis. In vitro, dynamin binds to microtubules, whichstimulate its activity. Binding of full-length dynamin to Srchomology 3 domains of proteins involved in signal transduc-tion has also recently been reported in vitro (12-14). Dynaminfamily members appear to be associated with many mechanicalfeatures of differentiation, the Golgi, and cell division (15-19)and, in association with kinesin and dynein, are associated withmicrotubules.

In this paper, the assignments and structure of dynamin PHdomain is derived from nuclear Overhauser effects (NOEs),amide exchange data, and chemical shifts obtained by a varietyof homonuclear and heteronuclear NMR methods. In light ofthe binding (3) and sequence homology data, the structurereveals part of a putative ligand binding site and conservedstructural features.

MATERIALS AND METHODSSample Preparation. Recombinant dynamin PH domain

was obtained by expression of a pET vector product asdescribed (8, 20). The sequence of the 125-residue protein isshown in Fig. 1. The first residue is not from the naturalsequence. Uniform 15N labeling was achieved by growing thecells in minimal medium with 18 mM 15NH4Cl. Selective 15Nlabeling for methionine and lysine were obtained similarly.Solutions used for NMR contained 1-4 mM protein in phos-phate-buffered saline at pH 6.0 (uncorrected for isotopeeffects), 50mM sodium azide, 4 mM [U-2H]dithiothreitol, and10% 2H20 in the H2O samples. Sample precipitation wasobserved at temperatures >40°C. 2H2O samples were pre-pared by lyophilizing three times and dissolving the residue in99.996% 2H20. The internal 'H chemical shift reference usedwas sodium 2,2-dimethyl-2-silapentane-5-sulfonate, and indi-rect referencing was used for 15N (22).NMR Spectroscopy. NMR experiments were run on a

General Electric model OMEGA 500 spectrometer or Brukermodel DMX-500. Quadrature detection was achieved by theStates or States-time proportional phase incrementation

Abbreviations: PH, pleckstrin homology; NOE, nuclear Overhausereffect; NOESY, NOE spectroscopy; HMQC, heteronuclear multiplequantum coherence spectroscopy; HOHAHA, homonuclear Hart-mann-Hahn spectroscopy; PIP2, phosphatidylinositol 4,5-bisphos-phhate; RMSD, root-mean-square deviation.ITo whom reprint requests should be addressed.

816

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

July

12,

202

0

Proc. Natt Acad Sci USA 92 (1995) 817

ACa

log ( tmiHX)

(ACa.- A#) 108 (1jl2 HX)

d4 N(i,i+1)dNN (i.i+ l)

dat N(i,i+3J

da N(i,i+4)15NI IH) NOE

10 20 30 40 50 60 70 s0 90 100 110 120

MCSO StLtOMNIl,LVR IVWxnLSLalgiermuLwvre

I_ n

- H u'L M"Iwu uuj _ -

rffh n Ml n rmn rn fT[l nnirm rim- n- ami ff -nbhUUm4Jl n1 U fMliWinI-

liii 111111 dliii 111Illl IiIIII hIIIiin II11 ih iiiiiiiiiiiiiI III quill ii III 11111 IIIIIIIImIi 1111fill11111l

I nfn[L.if n n m n fllf &nIh rmi O1, n nlnmb[IM U 111

I IM m__ j n9 rth.n n m n rwm rh

RMSD I, ni C_l - d - -

A8 I p2 J3 I4 P5 6

1/2 2/3 3/4 4/5 5/6

J7 acl P86/7 7/1 1/8

FIG. 1. Amino acid sequence and summary of observed and calculated NMR data related to secondary structure. The amino acid sequenceof the protein is shown at the top, corresponding to residues 524-658 in the complete sequence of dynamin. Each subsequent row gives the valuefor the individual residues, with an appropriate scale range indicated by the left hand vertical bar. There are no observed values for positions 84-87.ACa, differences in chemical shifts between observed 13Ca and expected values for amino acid type. Positive values are commonly correlated withan a-helix, and negative values are correlated with ,3-strands (range, +5 to -5 ppm). ACO, differences in chemical shifts between observed 13C3and expected values for amino acid type. Negative values are commonly correlated with an a-helix, and positive values are correlated with p-strands(range, +5 to -5 ppm). log(tl/2 HX), logio of measured half-times of exchange for amide hydrogens with deuterium, with a maximum set to 104min (range, 0-4). (ACa - ACt) log(tl/2 HX), product of log(tl/2) and (ACa - AC3) as a qualitative indicator of positions of secondary structuralelements. Positive values are correlated with an a-helix; negative values are correlated with 3-strands (range, +30 to -30). daN (i, i + 1), logiointensities of Ha, - H'i+1 NOEs for each observed pair. Values are normalized between spectra from XEASY analysis, with 10,000 as largest (range,0-4). dNN (i, i + 1), logio intensities of H'i - H'j+1 NOEs for each observed pair. Values are normalized between spectra from XEASY analysis,with 10,000 as largest (range, 0-4). daN(i, i + 3) and daN(i, i + 4), bar diagrams of all observed i, i + 3 and i, i + 4 NOEs, characteristic of a-helix.Intensities are not significant. 15N{'H} NOEs, the ratio of the recorded intensities of spectra with and without heteronuclear NOEs. Missing valuesare from unassigned peaks or where overlap precludes precise integration. Positive values of 0.7-0.8 are appropriate to the expected Tc for a moleculeof this size, while negative values indicate extensive flexibility (21). The expected theoretical range is 0.83 to -3.6; the range +1 to -1 is shown.RMSD, calculated RMSD for 16 calculated structures (bar graph) (range 0-8 A). Line segments are the above 15N{lH} NOEs multiplied by -1and translated to a scale similar to the RMSDs. Heteronuclear NOEs are three-point smoothed from the previous row, and values unobservedare omitted. The concurrent rise and fall of the bars of RMSD and lines of heteronuclear NOE suggest that the high RMSD segments are moreflexible than the rest of the molecule (see text). At the bottom, bars indicate the observed and calculated elements of secondary structure and theiridentities.

(TPPI) method. The water signal was suppressed by usingselective on-resonance irradiation during a relaxation delay of0.7-1.3 s and inNOE spectroscopy (NOESY) experiments alsoduring the mixing period. Experiments were run at 25°C withsweep widths of 8000 and 1500 Hz for 1H and 15N, respectively,unless indicated otherwise.Homonuclear experiments were run in both H20 and 2H20

using standard pulse sequences and phase cycling. A range oft1 increments from 200 to 512, each consisting of 2048 complexpoints was typically acquired with 32-128 scans per increment.In 2H20 double-quantum filtered two-dimensional correlatedspectroscopy spectra, 4096 points were acquired. Z-filteredhomonuclear Hartmann-Hahn spectroscopy (HOHAHA) ex-periments used an MLEV-17 sequence for spin lock withmixing times of 8, 33, and 65 ms in 2H20 and between 8 and90 ms in H20. NOESY spectra were recorded in 2H20 withmixing times (Tm) of 20, 50, 100, 150, and 200 ms and in H20with Tm values of 50, 100, 150, and 200 ms. A three-dimensionalNOESY heteronuclear multiple quantum coherence spectros-copy (NOESY-HMQC) and a small number of triple reso-nance experiments were performed on a Bruker AMX-600 atBruker Instruments, including CBCA(CO)NH, HN(CO)CA,and HNCA sets (23, 24).Other heteronuclear experiments consisted of HMQC,

HMQC-J, heteronuclear single quantum coherence spectros-copy, 15N-filtered HOHAHA and NOESY, two- and three-

dimensional NOESY-HMQC, and three-dimensionalHOHAHA-HMQC experiments. The Tm values in NOESY-HMQCs were 100 ms, and in three-dimensional HOHAHA-HMQC, the spin lock was 30 ms. The three-dimensionalspectra were recorded with a 32 x 100 x lk hypercomplexmatrix, with 32 scans per increment. The amide hydrogenexchange rates were measured by following the intensity ofcrosspeaks in HMQC experiments after exchanging a fullprotonated, lyophilized sample with 99.996% 2H20. Hetero-nuclear NOEs were measured using standard methods withwater presaturation (21). This method is recognized to pro-duce small errors from presaturation, but these are insignifi-cant for the qualitative uses of this paper.

Signal processing typically consisted of a single zero-fillingand polynomial baseline correction in wl. NOESY crosspeakintensity was estimated from the volume and maximum heightrelative to the height of crosspeaks between hydrogens sepa-rated by a fixed distance, using XEASY and CALIBA packages(26).

Spin System and Sequential Assignment. Typing and linkingof spin systems was accomplished by integrating fragmentsderived from the CBCA approach (23) with more conven-tional NOE-based assignment (27).

Structural calculations used DIANA with REDAC (28), withstructurally significant constraints of 752 upper and 187 lowerdistance bounds (from - 1500 NOEs), 38 hydrogen bonds

n ni mu n 131111111 rillll n n 11 11111 11111 ti - nniimin

Biochemistry: Fushman et al.

I

Dow

nloa

ded

by g

uest

on

July

12,

202

0

818 Biochemistry: Fushman et al.

chosen within strands of helix, with slowed exchange, andneighboring NOEs between strands, and 50 4,p constraintscorresponding to conservative ranges of allowed angles, inthose regions of strand or helix which were well defined (seeFig. 1). All peptide bonds were assumed to be trans. A finalselection of 16 structures from 200 starting structures was doneby using the lowest target functions (see Fig. 3 B and E). DLANAuses no assumptions about protein energetics, other than vander Waals repulsion; structures are unrefined and only ad-justed by rotation/translation for comparison purposes. Struc-tures were aligned using XPLOR and displayed and analyzedwith Biosym Technology's INSIGHTII package.

RESULTS AND DISCUSSIONThe 125-amino-acid section of dynamin used is that shown inFig. 1. This section was selected to obtain an entirely homo-geneous preparation with minimal secondary proteolysis, asdetermined by mass spectrometry. A combination of two- andthree-dimensional NMR experiments were used to identifyand sequentially assign nearly all of the 1H, 13C, and 15Nresonances in the backbone and C0 positions of the total125-amino-acid residues, as shown in Fig. 1. Line broadeningdue to chemical exchange precluded the identification ofNOEs or other connectivities from the amide resonances ofunassigned residues, located in the putative loop region,residues 84-87.

Elements of secondary structure were defined by severalNMR criteria as summarized in Fig. 1. Eight 13 strands werepredicted by negative AC0 shifts, restricted exchange, andstrong daN (i, i + 1) NOEs. Helical regions are identified bycoincidence of strong dNN (i, i + 1) NOEs, by NOEs tohydrogens three and four residues distant, and by CBCA shiftvalues (29). One a-helix is found for the sequence 103-115. Anetwork of hydrogen bonds implied by eight slowly exchangingamides stabilizes the helix. In Fig. 2, the network of homo-nuclear NOEs defining the sheets is shown.For calculated structures, there were no apparent sets of

different topologies. The root-mean-square deviation(RMSD) between the set and its average is 0.94 A for the coresegments (14-20, 32-37, 41-45, 55-57, 62-66, 75-81, 95-100,103-115, 117-120). The precision of any structure derivedwithout complete assignment of all hydrogens beyond the 13position is necessarily limited; however, the general fold of thedynamin PH domain in solution is available from these struc-tures.

In solution, dynamin PH consists of two major sheets, ofstrands 1-4 and 5-7. These sheets are composed of antiparallelstrands, and run continuously into each other 1 -- 2 -- 3 -- 4and 5 -- 6 -> 7. The sheets form a sandwich surrounding ahydrophobic core, with the strands of one sheet making an--'60° crossing angle with those of the other sheet (Fig. 3A).The end of the second sheet (137 strand) leads via a tight turn(7/1; Fig. 1) into a four-turn a-helix. A further tight turn leadsinto a short 13-strand, which crosses strand 135. The negativeheteronuclear 15N{1H} steady-state NOEs establish that res-idues 1-10 and 123-125 are essentially without structure. Thereduced positive NOEs in loops 1/2, 2/3, and 5/6 are readilyinterpreted as increased flexibility in these areas (Fig. 1),mimicked by the calculated RMSD for static structures (Figs.1 and 3B). The unassigned residues in loop 6/7 preclude asimilar interpretation of heteronuclear NOEs, but the techni-cal difficulties in observing these residues may arise fromconformational fluctuation in this region, leading to linebroadening from exchange on an intermediate time scale. Itseems likely that the overall changes in mobility reflected inincreased RMSD and decreased heteronuclear NOEs in loops1/2, 3/4, and 4/5 are effective also in loop 6/7, as suggestedin many other cases (31).

4

3

2

1

129 0 r

?~ cA 9 7

n 78 - g NOEsN""'I ~ I \ANN6

A76A H A FTA Lp '

0671(H 06 H 5~

120 119

FIG. 2. Schematic diagram of the antiparallel 13-sheet regions ofthe PH domain. Solid lines indicate observed long-range NOEs.Dashed bars represent NOEs that are expected but not unequivocallyidentified, due to resonance overlap or line broadening due tochemical exchange. Hydrogen bonds from slowly exchanging amideprotons that are supported by at least one interstrand NOE areindicated by dashed lines. Strand 8 is a mini ,B-strand, which crossesstrand 5 at approximately right angles. It is illustrated here as a parallelsheet interaction. Italic single letters are amino acid residue designa-tions.

This topology of the fold in solution is identical to thatobserved in the crystal structure of dynamin PH (8) and similarto that proposed in solution for other PH domains. Thedimerization in the crystal structure does not then appear tohave major effects on the structure. RMSDs between thecrystal structure and all solution structures are 131, 0.74; 132,0.60; 133,0.64; 134,0.47; 135,0.79; 16,0.53; 137,0.60; al, 0.44; and138, 1.28. The larger values for 135 and 138 RMSD betweensolution and crystal probably reflect minor changes caused bydimer formation, although these secondary structural ele-ments are not directly at the interface. This study definitivelyidentifies the length of the folded domain from the positiveheteronuclear 15N{ H} steady-state NOE determination forresidues 10-121. The sequence of the expressed spectrin PHwas significantly shorter in the C terminus (7). The sequenceof the expressed pleckstrin-N PH contains a "histidine tag"immediately after the equivalent of position 123 in dynaminPH (3). Such an unnatural insertion may alter the localconformation. In these other solution structures, the C-terminal contacts of the helix are not reported. In contrast, thestructures of dynamin PH contain a mini 1-strand (138), whichdefinitely establishes the position of the helix close to the restof the molecule. Assuming this to be a general feature of PHdomains, the long C-terminal helix then appears to have atotally distinct structure from that of proteins like retinolbinding protein; the latter class normally has a helix-turn-helix at the C terminus (25).

Nonetheless, the suggestion that PH domains may be in-volved in binding to lipids (2, 3) is intriguing, and dynamin PHhas a binding affinity to PIP2 (J. Zheng and D.C., unpublishedresults), similar to that of pleckstrin-N PH (3). However, so farthere is no evidence that lipids including PIP2 bind to thedynamin PH domain or to other PH domains specifically. In

Proc. NatL Acad ScL USA 92 (1995)

Dow

nloa

ded

by g

uest

on

July

12,

202

0

Biochemistry: Fushman et al. Proc. Natl. Acad Sci. USA 92 (1995)

0"0 o 0i

0 ,

0)|O0 OoC

"OC

819

Dow

nloa

ded

by g

uest

on

July

12,

202

0

820 Biochemistry: Fushman et al.

pleckstrin-N PH, the area of a possible PIP2 binding site hasbeen suggested to be in the 1/2 and 2/3 loops, especially thelysine and arginine side chains. Dynamin PH presents asimilarly charged surface (Fig. 3E), which can probably ac-commodate multiple phosphorylated ligands. In this respect,the demonstration of specificity and physiological significanceis needed to discriminate the possible role of PH in PIP2binding from the suggested, hypothetical role of PH in rec-ognizing, for example, phosphorylated serine and threoninesites (4). The wide sequence variability and range of associa-tion of PH domains with other functional domains argues fora broader role than recognizing a ubiquitous lipid. Thedetermination of larger number of PH domain structures mayassist in identifying common areas of possible ligand sites.

Note Added in Proof. Since submission of this article, two publicationshave appeared describing a 120-amino-acid residue segment corre-sponding to residues 2-121 of the sequence used in this paper. Onepaper (32) describes the 2.8 A crystal structure, and the other (33)describes the solution structure from NMR measurements. Thesestructures, that of ref. 8, and that of this paper are in generalagreement.

The provision of several spectral sets by Dr. Clemens Anklin(Bruker) is gratefully acknowledged. Mass spectra were generouslyprovided by Professor Brian Chait. We are grateful to K. M. Fergusonand Professor Paul B. Sigler for discussion and to them and StephenFesik for communication of results prior to publication. NMR re-sources at The Rockefeller University were purchased with grantsfrom the National Institutes of Health, the Keck Foundation, and theNational Science Foundation. M.A.L. is the Marion Abbe Fellow ofthe Damon Runyon-Walter Winchell Cancer Research Fund (DRG-1243). This work was supported at New York University by a grantfrom Sugen (to J.S.).

1. Mayer, B. J., Ren, R., Clark, K. & Baltimore, D. (1993) Cell 73,629-630.

2. Yoon, H. S., Hajduk, P. J., Petros, A. M., Oljniczak, E. T., Mead-ows, R. P. & Fesik, S. W. (1994) Nature (London) 369, 672-675.

3. Harlan, J. E., Hajduk, P. J., Yoon, H. S. & Fesik, S. W. (1994)Nature (London) 371, 168-170.

4. Gibson, T. J., Hyvonen, M., Musacchio A. & Saraste M. (1994)Trends Biochem. Sci. 19, 349-353.

5. Musacchio, A., Gibson, T., Rice, P., Thompson, J. & Saraste, M.(1993) Trends Biochem. Sci. 18, 343-348.

6. Shaw, G. (1993) Biochem. Biophys. Res. Commun. 195, 1145-1151.

7. Macias, M. J., Musacchio, A., Ponstingl, H., Nilges, M., Saraste,M. & Oschkinat, H. (1994) Nature (London) 369, 675-677.

8. Ferguson, K. M., Lemmon, M. A., Schlessinger, J. & Sigler, P. B.(1994) Cell 79, 199-209.

9. Rawlings, D. J., Saffran, D. C., Tsukasa, S., Largaespada, D. A.,Grimaldi, J. C., Cohen, L., Mohr, R. N., Bazan, J. F., Howard,M., Copeland, N. G., Jenkins, N.A. & Witte, 0. N. (1993)Science 261, 358-361.

10. Lefkowitz, R. J. (1993) Cell 74, 409-412.11. Keegan, A. D., Nelm, K., White, M., Wang, L.-M., Pierce, J. H.

& Paul, W. E. (1994) Cell 76, 811-820.12. Herskovits, J. S., Shpetner, H. S., Burgess, C. C. & Vallee, R. B.

(1993) Proc. Natl. Acad. Sci. USA 90, 11468-11472.13. Gout, I., Dhand, R., Hiles, I. D., Fry, M. J., Panayotou, G., Das,

P., Truong, O., Totty, N. F., Hsuan, J., Booker, G. W., Campbell,I. D. & Waterford, M. D. (1993) Cell 75, 25-36.

14. Booker, G. W., Gout, I., Downing, A. K., Driscoll, P. C., Boyd, J.,Waterfield, M. D. & Campbell, I. D. (1993) Cell 73, 813-822.

15. Izutsu, K. (1991) Hum. Cell 4, 100-108.16. Hollenbeck, P. J. (1990) Nature (London) 347, 229.17. Yeh, E., Driscoll, R., Coltrera, M., Olins, A. & Bloom, K. (1991)

Nature (London) 349, 713-715.18. Chen, M. S., Burgess, C. C., Vallee, R. B. & Wadsworth, S. C.

(1992) J. Cell Sci. 103, 619-628.19. Bloom, G. S. (1992) Curr. Opin. Cell Biol. 4, 66-73.20. Lemmon, M. A. & Ladbury, J. E. (1994) Biochemistry 33, 5070-

5076.21. Kay, L. E., Torchia, D. A. & Bax, A. (1989) Biochemistry 28,

8972-8979.22. Live, D., Davis, D., Agosta, W. & Cowburn, D. (1984) J. Am.

Chem. Soc. 106, 1939-1941.23. Grzesiek, S. & Bax, A. (1993) J. Biomol. NMR 3, 185-204.24. Bax, A. & Grzesiek, S. (1993) Acc. Chem. Res. 26, 131-138.25. Flower, D. R., North, A. C. T. & Attwood, T. K. (1993) Protein

Sci. 2, 753-761.26. Giintert, P., Braun, W. & Wiithrich, K. (1991) J. Mol. Biol. 217,

517-530.27. Wiithrich, K. (1986) NMR of Proteins and Nucleic Acids (Wiley,

New York).28. Guntert, P. & Wuthrich, K. (1992) J. Biomol. NMR 1, 447-456.29. Vuister, G. W., Kim, S.-J., Wu, C. & Bax, A. (1994) Biochemistry

33, 10-16.30. Sutcliffe, M. J. (1993) Protein Sci. 2, 936-944.31. Powers, R., Clore, G. M., Garrett, D. S. & Gronenborn, A. M.

(1993) J. Magn. Reson. B101, 325-327.32. Timm, D., Salim, K., Gout, I., Guruprasad, L., Waterfield, M. &

Blundell, T. (1994) Nature Struct. Biol. 1, 782-788.33. Downing, K., Driscoll, P. C., Gout, I., Salim, K., Zvelebil, M. J.

& Waterfield, M. D. (1994) Curr. Biol. 4, 884-891.

Proc. Natl. Acad Sci. USA 92 (1995)

Dow

nloa

ded

by g

uest

on

July

12,

202

0