Nature Structural Biology
9, 247 - 251 (2002)
Published online: 4 March 2002; | doi:10.1038/nsb773
Structure of the C-terminal FG-nucleoporin binding domain of Tap/NXF1Richard P. Grant1, Ed Hurt2, David Neuhaus1
& Murray Stewart11 MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK. 2 Biochemie-Zentrum Heidelberg, Im Neuenheimer Feld 328, 69120 Heidelberg, Germany.
Correspondence should be addressed to Murray Stewart ms@mrc-lmb.cam.ac.ukThe vertebrate Tap protein is a member of the NXF family of shuttling transport receptors for nuclear export of mRNA. Tap has a modular structure, and its most C-terminal domain is important for binding to FG repeat-containing nuclear pore proteins (FG-nucleoporins) and is sufficient to mediate nuclear shuttling. We report the solution structure of this C-terminal domain, which is based on a distinctive arrangement of four  -helices and is joined to the next module by a flexible 12-residue Pro-rich linker. F617A Tap suppresses FG-nucleoporin binding by the most C-terminal domain that, together with the structure of the other modules from which Tap is constructed, provides a structural context for its nuclear shuttling function.
Nuclear trafficking is mediated by soluble transport factors that shuttle between the nuclear and cytoplasmic compartments through nuclear pore complexes (NPCs)1,
2,
3. Although the precise mechanism by which transport factors in complex with their cargo move through NPCs is controversial, a consensus is emerging that this step involves interactions with nuclear pore proteins containing characteristic phenylalanine-glycine (FG) repeats2,
3,
4,
5,
6. Vertebrate Tap (NXF1) and Saccharomyces cerevisiae Mex67p are members of the NXF family of mRNA nuclear export factors7,
8. Mex67p and Caenorhabditis elegans NXF1 are essential for the nuclear export of poly(A)+ mRNA7,
9. Addition of the human NXF1 homolog, Tap, to cultured human 293 cells stimulates the nuclear export of mRNAs that are otherwise exported inefficiently10, and facilitates the export of unspliced viral mRNA containing the constitutive transport element (review ref. 1). Mex67p binds Mtr2p in vitro and in vivo, and formation of this heterodimer is essential for both mRNA export and association of Mex67p with nucleoporins11. Similarly, the NTF2-like protein NXT1 is a critical cofactor in human and invertebrate cells10,
12,
13 and may act as a regulatory switch12,
13. The ability of recombinant Tap−NXT1 heterodimers to rescue the function of yeast Mex67p/Mtr2p double knockouts implicates Tap as a general cellular factor for the export of processed mRNA14.
Tap has a multidomain structure. A nuclear localization sequence and a noncanonical mRNA-binding domain followed by four LLR repeats are located in its N-terminal half15,
16. The C-terminal half contains a NTF2-homology domain (residues 371−551) that forms a heterodimer with NXT1, which binds nucleoporins17, followed by a second domain that is both necessary and sufficient for localization to NPCs in vivo and nucleocytoplasmic shuttling of fusion proteins15,
18. This most C-terminal domain of Tap also contains a key FG-nucleoporin binding site14,
15,
18,
19,
20, and sequence analysis21 suggests that it has a UBA (ubiquitin-associated) fold. We report here the solution structure of residues 551−619 of Tap. This fragment (TapC) encompasses the most C-terminal Tap domain, retains binding to FG-nucleoporins and forms a compact four-helix fold related to that of a UBA domain. The structure indicates that several previously described mutations in this region15,
20,
21 probably disrupt its structure rather than identifying residues directly involved in nucleoporin binding.
Tap C binds nucleoporin FG repeats
Although sequence analysis21 has indicated that the Tap NTF2-homology domain ends near residue 508, recent structural work17 has shown that it extends to residue 551, indicating that the second nucleoporin-binding domain is located in residues 551−619, which we refer to as TapC. The one-dimensional NMR spectrum of bacterially expressed TapC had a dispersed amide region consistent with a folded structure, and circular dichroism (CD) indicates that it is  60%  -helical (Fig. 1a). TapC shows saturable binding of FG-nucleoporins. For example, TapC binds NSP1−FF18 (a fragment of nucleoporin NSP1 containing 18 FxFG repeats)22 with an apparent Kd of 29  M (Fig. 1b). TapC also binds shorter peptides containing FxFG repeat cores, but does not bind peptides in which a core Phe was substituted by Ala, indicating that the binding is specific and involves primarily the repeat cores (Fig. 1c).
Structure of the Tap C-terminal domain
The solution structure of TapC, determined by homonuclear NMR (Fig. 2a,b; Table 1), is well defined between Leu 563 and Met 618, whereas assignment for residues 551−562 was hindered by a lack of sequential and medium-range NOEs, consistent with this region being unstructured. There is also evidence for cis-trans isomerization-induced signal doubling of residues adjacent to Pro 551, 553, 555, 559 and 561. In addition, reduced NOE contacts in the region between residues 609 and 619 indicate that the structure in this region has a backbone root mean square (r.m.s.) deviation of 0.54 Å (compared with 0.49 Å for Leu 563−Met 618). Within this region, residues 613− 618 form an -helix, as demonstrated by distinctive d N(i, i+3) and dNN(i, i+1) NOE connectivities.
 | | |
| | |
The solution structure of TapC contains a single domain constructed from four helices packed against each other (Fig. 3a). The side chains of Leu 571, Phe 574, Met 580, Cys 588, Leu 589, Trp 594, Ala 600, Phe 603 and Leu 606 form a hydrophobic core, whereas the side chain of Trp 584 is rotated away from the domain core and contributes to NOEs with residues 609−619. The side chains of Val 615, Ala 616 and Phe 617 seem to stabilize packing of helix 4 (residues 613−618) against the other three helices. The arrangement of helices 1 (residues 566−577), 2 (residues 582−592) and 3 (residues 597−607) is similar to that seen in a UBA fold (Fig. 3b), with most pronounced structural homology to the second UBA domain (UBA-2) of DNA repair protein HHR23a (PDB 1DV0)23. Comparison of the central helical region (residues 566−607) of TapC and the first 42 residues of HHR23a UBA-2 gives a backbone r.m.s. deviation of 1.61 Å. This distant structural homology is in agreement with predictions based on sequence analysis21. Helix 4 in TapC does not have a counterpart in HHR23a UBA-2.
| | |
The scarcity of NOE peaks arising from the first 13 residues of TapC together with the signal doubling indicative of Pro cis-trans isomerization in this region indicates that residues 551−562 are relatively unstructured. The sequence of this region (Fig. 3d) is consistent with its having little structure and suggests that this Pro-rich stretch constitutes a flexible linker between the NTF2 and TapC domains. The TapC domain would, therefore, be mobile relative to the rest of the protein. The crystal structure of residues 371−619 of Tap17 is consistent with this and shows clear electron density only for the NTF2-like heterodimerization domain; the TapC domain is disordered in these crystals, consistent with its being highly mobile. We also found that seven N-terminal residues are removed rapidly by trypsin, leaving a protease-resistant core (residues 558−619, determined by mass spectroscopy).
FG-nucleoporin binding site on TapC
Previous studies15,
20,
21 identified a tripeptide (NWD/N, which forms part of the loop between helices 2 and 3) that is conserved in several UBA sequences and suggested that Trp 594 and Asp 595 might be implicated directly in the interaction between Tap and FG-nucleoporins. This hypothesis was supported by the observation that the Tap W594A and D595R mutants did not bind nucleoporins19,
21. However, the structure of this region is conserved only weakly (r.m.s. deviation 0.85Å) between TapC and HHR23a UBA-2 (Fig. 3a,b). Moreover, the side chain of Trp 594 is mainly buried in the hydrophobic core of the TapC domain, where its aromatic ring is stacked against Leu 563, Glu 566, Gln 586, Gln 590 and Leu 589 (Fig. 3c), and forms a key part of the network of hydrophobic interactions linking helices 1 and 2. Burying the side chain of Trp 594 in the hydrophobic core of TapC would not only make it inaccessible and, therefore, unlikely to be involved directly in binding to FG-nucleoporins, it also suggests that the conservation of Trp 594 between sequences was due to its importance for maintaining the fold of the domain. Asp 595 is in a tight loop and might also have a structural role. To distinguish between putative functional and structural roles for Trp 594 and Asp 595, we constructed W594A, D595A and D595R19,
21 mutants of TapC. The CD spectrum of bacterially-expressed TapC/W594A is very different from that of wild type, indicating that it retains little -helical secondary structure (Fig. 1a), consistent with the importance of Trp 594 in forming the hydrophobic core of the domain. Further support for a mostly structural role for Trp 594 comes from the observation that full-length Tap W594A also showed decreased NXT1 binding, whereas wild type NXT1 binding was observed when the entire C-terminal domain was deleted21. We were unable to assess the secondary structure of the D595R mutant because of its insolublity, which in itself suggests that the native fold has been disrupted. The more conservative D595A mutant has a CD spectrum similar to wild type, indicating that its secondary structure has not been changed substantially. In contrast to D595R21, D595A retains wild type levels of FG-nucleoporin binding (Kds 26 and 29 M, respectively). Overall our results indicate that Trp 594 and Asp 595 are unlikely to be involved directly in binding FG-nucleoporins.
The binding of FG-nucleoporins to NTF2 (ref. 24), importin- 25 and the NXF1−NXT1 heterodimer17 involves primarily the binding of FxFG cores to hydrophobic surface patches. Our observation that peptides in which core Phe residues were altered to Ala fail to bind TapC indicate that its FG-nucleoporin binding site might have similar characteristics. When we examine the surface of TapC, we find two hydrophobic patches analogous to those on NTF2 and importin- that are, therefore, candidates for the binding site for FG-nucleoporin cores. The first (Fig. 4a) is near the turn between helix 2 and helix 3 and is centered on Tyr 596. It also contains Ala 573 and Met 570, with small contributions from Trp 594, Leu 589 and Phe 574. A second patch (Fig. 4b) contains Gly 579, Met 580, Ala 608, Gly 610, Ile 612, Phe 617 and Met 618, as well as the aliphatic regions of Lys 607 and Glu 614. To explore the involvement of these patches in nucleoporin binding, we made the TapC L563A, Y596D and F617A point mutants and assayed their binding to NSP1−FF18 (ref. 22). The L563A and Y596D mutations in the first region have little affect on the secondary structure assessed by CD and have apparent Kds (27 and 31 M, respectively) similar to wild type (29 M). By contrast, although the F617A mutant in the second hydrophobic patch seems to fold normally by CD (Fig. 1a), it shows no detectable binding to FG-nucleoporin repeats (Fig. 1b). Moreover, fluorescently labeled wild type TapC shows punctate nuclear rim staining in Triton-permeabilized19 HeLa cells, whereas this staining was substantially reduced with the F617A mutant (Fig. 4f,g). Residues near Phe 617 make long-range NOE contacts with Trp 584, consistent with the change in Trp fluorescence seen on nucleoporin binding. Results obtained using triple Ala mutants in full-length Tap are also consistent with this region being part of the FG-nucleoporin binding site20. Thus, the A23 (G610A/E611A/I612A) mutant was unable to facilitate CTE (consecutive transport element)-containing mRNA export, whereas A24 (P613A/E614A/V615A) was partially active. Other inactive mutants (A15 = K587A/C588A/L589A and A17 = N593A/W594A/D595A) contain residues in the hydrophobic core and, therefore, probably alter the domain structure.
| | Figure 4. Interaction between TapC and FG-nucleoporins. | | |  | a, Hydrophobic patch 1 (dark blue residues) centered on Tyr 596. b, Hydrophobic patch 2 contains Phe 617, which mutational analysis implicated in binding FG-nucleoporins. The aliphatic regions of polar side chains are shown in light blue. c, Tap C in the same orientation as (b) but with residues that had NH chemical shift changes >0.01 p.p.m. in the presence of a FxFG peptide shown in red. These residues are clustered around hydrophobic patch 2, consistent with its association with FxFG-nucleoporin binding. d, Part of the NH,H fingerprint region of the 800 MHz homonuclear double quantum (2Q) correlation spectrum, showing peaks assigned to residues in the vicinity of Phe 617 in the presence (red) and absence (black) of 1 mM FxFG peptide. There were no contributions from the added peptide in the region shown. e, Selected individual (NH and H) peaks from the 2Q spectrum in the presence (red) and absence(black) of FxFG peptide. For example, Phe 617 and Met 618 show substantial chemical shift changes, whereas Trp 594 and Tyr 596 show negligible changes. f, Wild type TapC shows nuclear rim staining in permeabilized cells. g, In contrast, the F617A mutant shows substantially reduced levels of staining.
 Full Figure and legend (166K) |
| |
Further evidence for Phe 617 having a role in FxFG-nucleoporin binding was obtained from analysis of chemical shift changes in TapC resonances produced by addition of the DSGFSFGSK peptide (FxFG site in bold) (Fig. 4d,e). During this titration, large shift changes were observed for residues following Thr 604 that were near Phe 617 (Fig. 4c). However, the size and distribution of the shift movements suggest that there were probably also conformational changes associated with peptide binding. Therefore, even though the shift changes support a role for this region in FxFG nucleoporin binding, it is not possible to use these data to identify unequivocally individual residues that interacted directly. Changes in the rest of the molecule seem to be distributed randomly, except for a group of residues adjacent to this area (Trp 584 ring NH, Cys 588, Phe 603 and Leu 606) that may also contribute to nucleoporin binding.
Conclusions
The structure of the TapC domain, together with crystallography data for its other domains16,
17, allows a picture of the complete molecule to be assembled. Tap is constructed from a series of functional modules joined by flexible linkers. The binding site for CTE-containing mRNA is located in the N-terminal domain, which is constructed from a noncanonical RNA binding motif and four Leu-rich repeats15,
16. This module is linked to the NTF2-homology central domain (residues 371−551) that binds NXT1 (ref. 17), which is then connected to the TapC domain (residues 563−618) by a flexible Pro-rich linker (residues 552−562). Both the TapC and NTF2-like domains of Tap bind FG-nucleoporins10,
11,
12,
13,
14,
15,
16,
17. The TapC domain is sufficient for shuttling and NPC binding, and deletion of this domain reduces binding to the nuclear envelope15,
19. However, the nucleoporin-binding activity of the NTF2 domain is probably also functionally important, especially in the context of strengthening the interaction and possibly providing a regulatory function12,
13. Thus, a Mex67 mutant lacking the equivalent of the TapC domain is still able to mediate mRNA export11, and an analogous effect is also seen with NXF1, where the formation of a heterodimer with NXT1 seems to be important for FG-nucleoporin binding12,
13,
17. Both the NTF2 (ref. 17) and TapC domains (Fig. 4) of Tap seem to bind to the Phe-rich cores of FG-nucleoporins via hydrophobic surface patches analogous to those seen with NTF2 (ref. 24) and importin-25. Moreover, as observed with other transport factors, the Kds for the Tap−nucleoporin interactions are weak, which would imply a rapid off-rate and enable them to move rapidly through NPCs6,
24. However, the apparent Kd for the TapC domain alone (29 M) is weaker than that observed for NTF2 (ref. 24); thus, binding to the Tap NTF2 domain may also be required to make the Tap−nucleoporin interaction sufficiently strong to mediate effective transport. In this context, the flexibility of the Pro-rich linker (residues 552−562) could be important by enabling the TapC and NTF2-like domains to bind nucleoporins simultaneously. Such a mechanism could generate a level of fine control for the Tap−nucleoporin interaction in vivo and provide a rationale for the importance of the formation of the conserved Tap−NXT1 and Mex67p−Mtr2 heterodimers in mRNA export.
Methods
Expression and purification of TapC.
PCR using Turbo Pfu (Stratagene) was used to introduce NdeI and EcoRI restriction sites at the 5' and 3' termini of a 210-nucleotide fragment of human Tap cDNA corresponding to residues 550−619. Amplification products were digested with NdeI and EcoRI, cloned into pMW172 and transformed into BL21(DE3) RIL cells (Stratagene). Sequencing confirmed that no errors had been introduced. A single, ampicillin-resistant colony was picked into 1 l 2xTY medium containing 50 g ml-1 ampicillin and grown at 37 °C for 18 h with vigorous shaking. Cells were harvested and lysed by sonication. Clarified lysate was dialyzed extensively against 10 mM Tris-HCl, 1 mM EGTA and 1 mM PMSF, pH 8.5, at 4 °C; applied to a Q-Sepharose column in 20 mM Tris-HCl, 1mM EGTA and 1 mM PMSF, pH 8.5; and eluted with a 200 ml gradient to 500 mM NaCl. TapC fractions were pooled, concentrated and purified to homogeneity by gel filtration over Sephacryl S-100 in 20 mM Tris-HCl, 100 mM NaCl, 1 mM EGTA and 1 mM PMSF, pH 8.5. TapC protein was shown to be essentially pure by SDS-PAGE and soluble up to 6 mM in Tris-buffered MilliQ water. Mass spectroscopy yielded a molecular mass consistent with the predicted sequence Tap 551−619, and N-terminal sequencing confirmed that the initiator Met was absent. CD spectra were recorded using an ISA CD6 spectrophotometer.
Binding to nucleoporin FG repeats.
Samples were prepared at 0.08 mg ml-1 in 350 l of 20 mM HEPES, 110 mM K acetate and 2 mM Mg acetate, pH 7.35. Binding of NSP1-FF18 (ref. 22) to TapC was measured by the change in Trp fluorescence at 30 °C, with excitation and emission filters set at 288 and 340 nm, respectively, using a Perkin-Elmer LS50B spectrofluorimeter and slit widths of 2.5 nm. To assess nuclear envelope staining, HeLa cells were permeabilized with 0.5% Triton X-100 as described19 and then incubated with 0.5 mg ml-1 wild type or F617A TapC that were labeled with Oregon Green or TritC for 5 min at room temperature. Fluorescent micrographs were obtained using a MRC1024 confocal microscope.
Structure determination.
We experienced difficulty in expressing high levels of 15N-labeled TapC in bacteria. However, because TapC could be prepared at high concentration and initial experiments indicated that the 2D spectra were well resolved and amenable to homonuclear analysis, assigning proton resonances was possible following the conventional strategy of spin-system assembly and sequential assignment for protein samples at natural isotopic abundance26. NOESY (mixing times 25, 50, 100, 150 and 200 ms), TOCSY (mixing time 60 ms), double-quantum filtered COSY and double quantum correlation (2Q preparation period 30 ms) spectra of TapC at 3 mM in 25 mM Na phosphate buffer, 50 mM NaCl, 10% (v/v) D2O, pH 6.0, were acquired at 295, 300 or 310 K using Bruker DRX500, DMX600 or Avance800 spectrometers. For 2D NOESY experiments, the spectral width in both dimensions was 9,615 Hz, whereas it was 7,246 Hz for double quantum-filtered COSY and TOCSY experiments, and 7,245 and 14,492 Hz in F2 and F1, respectively, for double quantum correlations. In all cases, the time domain data sizes were 1,024(complex)  512(real), and TPPI was used to achieve quadrature detection in F1. Water suppression was achieved using phase-coherent presaturation. Data processing was carried out using Xwin-NMR (Bruker), and time-domain data were multiplied in both dimensions by phase-shifted sine-bell or squared sine-bell functions. Spectra were referenced relative to internal 0.1 mM sodium [2,3-2H4]3-trimethylsilyl propionate. Distance constraints were classified as strong (0−2.8 Å), medium (0−3.5 Å) or weak (0−5.0 Å) based on intensities. Calibration was based primarily on dNN crosspeaks in helices (corresponding to 2.8 Å) and dN(i, i+3) crosspeaks in helices (corresponding to 3.5 Å); the upper bound for the very weak category was set to 5.0 Å to allow for spin diffusion. For each category, the lower distance limit was set to zero to not interfere with the initial phase of the simulated annealing protocol, during which all van der Waals radii are greatly reduced27. A small number of constraints were defined as ambiguous28 to allow for possible ambiguities in positional assignments within a residue — for example, constraints involving CH2 or C H2 of a particular spin system. Using simulated annealing protocols in X-PLOR 3.1 (refs 29,30), 50 structures were calculated starting from initial coordinates with randomized  and  torsion angles. In all structure calculations, the force field comprised only geometric terms (bond lengths and angles, van der Waals repulsive terms, and planarity constraints for peptide bonds and aromatic rings); electrostatic, van der Waals attractive and hydrogen bonding terms were not included. Structure calculations were carried out using r-6 summation of distance constraints involving groups of equivalent atoms28. Atomic r.m.s. deviations and energy-ordered r.m.s. deviation profiles were calculated using CLUSTERPROSE31.
Coordinates.
Atomic coordinates for the final ensemble of 26 structures have been deposited in the Protein Data Bank (accession code 1GO5).
Received 1 November 2001; Accepted 7 February 2002; Published online: 4 March 2002.
REFERENCES
- Cullen, B.R. Mol. Cell. Biol. 20, 4181−4187 (2000). | Article | PubMed | ISI | ChemPort |
- Ryan, K.J. & Wente, S. Curr. Opin. Cell Biol. 12, 361−371 (2001). | Article | ISI |
- Stewart, M. et al. FEBS Lett. 498, 145−149 (2001). | Article | PubMed | ISI | ChemPort |
- Damelin, M. & Silver, P.A. Mol. Cell 5, 133−140 (2000). | Article | PubMed | ISI | ChemPort |
- Rout, M.P. et al. J. Cell Biol. 148, 635−652 (2000). | Article | PubMed | ISI | ChemPort |
- Ribbeck, K. & Görlich, D. EMBO J. 20, 1320−1330 (2001). | Article | PubMed | ChemPort |
- Segref, A. et al. EMBO J. 16, 3256−3271 (1997). | Article | PubMed | ChemPort |
- Herold, A. et al. Mol. Cell. Biol. 20, 8996−9008 (2000). | Article | PubMed | ISI | ChemPort |
- Tan, W., Zolotukhin, A.S., Bear, J., Patenaude, D.J.& Feldber, B.K. RNA 6, 1762−1772 (2000). | Article | PubMed | ISI | ChemPort |
- Braun, I.C., Herold, A., Rode, M., Conti, E. & Izaurralde, E. J. Biol. Chem. 276, 20536−20543 (2001). | Article | PubMed | ISI | ChemPort |
- Strässer, K., Bassler, J. & Hurt, E. J. Cell Biol. 150, 695−706 (2000). | Article | PubMed | ISI | ChemPort |
- Weingand, H.L. et al. Mol. Cell. Biol. 22, 245−256 (2002). | Article | PubMed |
- Levesque, L. et al. J. Biol. Chem. 276, 44953−44962 (2001). | Article | PubMed | ISI | ChemPort |
- Katahira, J. et al. EMBO J. 18, 2593−2609 (1999). | Article | PubMed | ChemPort |
- Kang, Y. & Cullen, B.R. Genes Dev. 13, 1126−1139 (1999). | PubMed | ISI | ChemPort |
- Liker, E., Fernandez, E., Izaurralde, E. & Conti, E. EMBO J. 19, 5587−5598 (2000). | Article | PubMed | ChemPort |
- Fribourg, S., Braun, I.C., Izaurralde, E. & Conti, E. Mol. Cell 8, 645−656 (2001). | Article | PubMed | ISI | ChemPort |
- Schmitt, I. & Gerace, L. J. Biol. Chem. 276, 42355−42363 (2001). | Article | PubMed | ISI | ChemPort |
- Bachi, A. et al. RNA 6, 136−158 (2000). | Article | PubMed | ISI | ChemPort |
- Kang, Y., Bogerd, H.P. & Cullen, B.R. J. Virol. 74, 5863−5871 (2000). | Article | PubMed | ISI | ChemPort |
- Suyama, M. et al. EMBO Rep. 1, 53−58 (2000). | Article | PubMed | ChemPort |
- Clarkson, W.D., Kent, H.M. & Stewart, M. J. Mol. Biol. 263, 517−524 (1996). | Article | PubMed | ISI | ChemPort |
- Withers-Ward, E.S. et al. Biochemistry 39, 14103−14112 (2000). | Article | PubMed | ISI | ChemPort |
- Bayliss, R. et al. J. Mol. Biol. 293, 579−593 (1999). | Article | PubMed | ISI | ChemPort |
- Bayliss, R., Littlewood, T.D. & Stewart, M. Cell 102, 99−108 (2000). | Article | PubMed | ISI | ChemPort |
- Wüthrich, K. NMR of proteins and nucleic acids (John Wiley, New York; 1986).
- Hommel, U., Harvey, T.S., Driscoll, P.C. & Campbell, I.D. J. Mol. Biol. 227, 271−282 (1992). | Article | PubMed | ISI | ChemPort |
- Nilges, M. J. Mol. Biol. 245, 645−660 (1995). | Article | PubMed | ISI | ChemPort |
- Nilges, M., Gronenborn, A.M., Brünger, A.T. & Clore, G.M. Protein Eng. 2, 27−38 (1988). | PubMed | ISI | ChemPort |
- Brünger, A.T. X-PLOR Version 3.1(Yale University, New Haven; 1992).
- Diamond, R. Acta Crystallogr. D 51, 127−135 (1995). | Article | PubMed | ISI | ChemPort |
Acknowledgments
We are grateful to our colleagues in Cambridge and Heidelberg, especially R. Bayliss, T. Littlewood, K. Strässer and A. Weeds for their helpful comments and criticisms. We thank J.C. Yang for help in recording NMR spectra, S. Peak-Chew for N-terminal sequencing and mass spectroscopy, and G. Wong for preparing proteins. Supported in part by the Human Frontiers Science Program.
Competing interests statement:
The authors declare that they have no competing financial interests. |
