Structure, Stability, and Function of hDim1 Investigated by NMR

Federica Simeoni, Andy Arvai, Paul Bello, Claire Gondeau, Karl-Peter Hopfner, Paolo Neyroz, Frederic Heitz, John Tainer, and Gilles Divita. Biochemist...
0 downloads 0 Views 491KB Size
Biochemistry 2003, 42, 9609-9618

9609

Structure, Stability, and Function of hDim1 Investigated by NMR, Circular Dichroism, and Mutational Analysis† Yu-Zhu Zhang,*,‡,⊥ Hong Cheng,‡ Kathleen L. Gould,§ Erica A. Golemis,*,‡ and Heinrich Roder*,‡ DiVision of Basic Science, Fox Chase Cancer Center, Philadelphia, PennsylVania 19111 and the Howard Hughes Medical Institute and Department of Cell and DeVelopmental Biology, Vanderbilt UniVersity School of Medicine, NashVille, Tennessee 37232 ReceiVed March 26, 2003; ReVised Manuscript ReceiVed June 9, 2003

ABSTRACT: The 142 amino acid Dim1p protein is a component of the U4/U6‚U5 tri-snRNP complex required for pre-mRNA splicing and interacts with multiple splicing-associated proteins. To gain further insight into the structural basis of its function, we determined the solution structure of the reduced form of the dominant negative human hDim1 (hDim11-128) using multidimensional NMR spectroscopy. This dominant negative hDim1 assumes a thioredoxin-like fold, confirming previous NMR and crystallographic results. However, in contrast to a recent crystal structure, the NMR solution structure for the reduced form of hDim11-128 presented here, along with thermodynamic data, indicates that the presence of a disulfide bond between Cys38 and Cys79 is structurally and functionally unimportant. Comparison of the truncated hDim11-128 with the full-length protein, using NMR and circular dichroism spectroscopy, indicates that the 14 C-terminal residues can undergo a local unfolding transition and assume alternative conformations, which appear to play a functional role. Other residues essential for hDim1 function are identified by using mutational and genetic approaches. The residues thus identified are not identical with those previously shown to govern Dim1 interaction with defined protein partners.

The Dim1 protein is extraordinarily well conserved throughout the eukaryotic kingdom, with 79% sequence identity over its entire length of 142 amino acids and full complementation function maintained between human (hDim1) and Schizosaccharomyces pombe (Dim1p) orthologs (1, 2). Dim1 functions at multiple levels in control of pre-mRNA processing. Dib1p, the Saccharomyces cereVisiae Dim1p ortholog, has been identified by mass spectrometry following purification as a component of the pre-mRNA splicing machinery in two independent studies aimed at identifying novel elements of the S. cereVisiae U4/U6‚U5 snRNP (3, 4). Dib1p was found to be required for the splicing of a nonmessenger RNA, the U3 RNA (4). Elimination of expression of the Dim1 ortholog dml-1 in Caenorhabditis elegans by RNA interference leads to embryonal lethality during gastrulation with an arrest phenotype consistent with global disruption of zygotic gene expression and compatible with a global defect in pre-mRNA splicing (5-7). These results † This work was supported by NIH Grants RO1 CA63366 (E.G.) and RO1 GM056250 (H.R.), Core Grant CA-06927 (to Fox Chase Cancer Center), an appropriation from the Commonwealth of Pennsylvania (to Fox Chase Cancer Center), NIH Training Grant T32 CA09035, ACS Fellowship PF-98-290-01-CSM (Y.Z.), and the HHMI (K.L.G.). The NMR facility was supported in part by a grant from the Kresge Foundation. * Corresponding authors. Tel.: 312-567-3484; fax: 312-567-3494; e-mail: [email protected] (Y. Z.). Tel.: 215-728-2860; fax: 215-7283616; e-mail: [email protected] (E. A. G.). Tel.: 215-728-3123; fax: 215-728-3574; e-mail: [email protected] (H. R.). ‡ Fox Chase Cancer Center. § Vanderbilt University School of Medicine. ⊥ Present address: Department of Biological, Chemical, and Physical Sciences, Illinois Institute of Technology, Chicago, IL 60616.

suggested an essential, highly conserved function for Dim1 family proteins in eukaryotes. Analyses of the interaction properties of Dim1 proteins have provided direct connections to the pre-mRNA splicing machinery. It has been shown that hDim1 interacts with hnRNP F and hnRNP H′ (5), proteins that function in tissuespecific enhancement of pre-mRNA splicing (8, 9) and contain RNA recognition motif (RRM)-like sequences that confer the ability to directly interact with poly(rG) (10). hDim1 also interacts with Npw38/PQBP-1 (5, 11, 12). This protein possesses a specific RNA binding activity for poly(rG) (11), colocalizes with the SRm160/SRm300 splicing co-activators when coexpressed with Dim1 (5, 13, 14), and under some circumstances, regulates RNA polymerase IIdependent transcription (15). Finally, Dib1p has been found to interact strongly with Prp6p (16), a protein required for the accumulation of U4/U6‚U5 tri-snRNP (17, 18). Together, these studies suggest that hDim1 functions at multiple control points in the splicing of pre-mRNAs, potentially as part of a large spliceosomal complex involving numerous proteinprotein interactions. Which of these contacts is critical for the essential function of Dim1 is not yet clear. Past studies by our group using NMR1 coupled with molecular modeling (2) and by others using X-ray crystallography (19) have indicated that hDim1 adopts a thioredoxin fold with a C-terminal (14 amino acid) extension. Analysis 1 Abbreviations: NMR, nuclear magnetic resonance; TOCSY, total correlation spectroscopy; HSQC, heteronuclear single-quantum correlation; NOE, nuclear Overhauser effect; NOESY, NOE spectroscopy; rmsd, root-mean-square deviation; DTT, dithiothreitol; GuHCl, guandidine hydrochloride.

10.1021/bi034486i CCC: $25.00 © 2003 American Chemical Society Published on Web 07/26/2003

9610 Biochemistry, Vol. 42, No. 32, 2003 of amino acid conservation throughout evolution and mutational studies (2) indicates that the ability to take part in redox reactions through formation of a disulfide bond is probably not important for in vivo Dim1 activity, arguing against a simple analogy to thioredoxin. However, deletion of the C-terminal extension to make hDim11-128 results in the creation of a dominant negative form of Dim1 (2). This suggests that regulation of the activity of the thioredoxinlike core by the C-terminal extension may be important for Dim1 function. The functional importance of the C-terminal extension is further supported by the fact that the dim1-35 mutant initially isolated in S. pombe (1), which contains a point mutation at the junction between this extension and the thioredoxin-like core (G126D), manifests a temperaturesensitive phenotype leading to loss of cellular viability. However, at present, the role of this extension for Dim1 function is poorly understood. Dim1 provides a useful model for analysis of the folding and function of proteins in complexes involving many protein-protein interactions. In this study, we have taken a systematic approach to elucidating the structure and function of hDim1. (1) We report the solution structure for the reduced dominant negative form of Dim1 and compare it to the crystal structure of the oxidized full length Dim1 protein. (2) The consequences of deleting residues 129-142 for Dim1 folding and stability are investigated using CD spectroscopy. (3) The in vivo complementation properties of a systematic series of alanine mutants pinpoint surface residues critical for protein function. (4) For selected Dim1 mutants with reduced in vivo function, we study their impact on folding and stability. (5) Finally, we compare residues shown to be required for complementation to those necessary for interaction with partner proteins, hnRNP F and Npw38/PQBP-1. Together, these studies identify functionally important regions on Dim1 and can be used to guide further studies of Dim1 in pre-mRNA splicing. MATERIALS AND METHODS Protein Expression, Purification, Characterization, and Mutagenesis. Protein expression and purification protocols for full-length hDim1 and hDim11-128 have been described (2). Mutants hDim1 G126D, hDim1-8 (R86A, K88A), and hDim1-12 (R124A/K125A/R127A) were expressed and purified by similar procedures. The mutagenesis strategy used to create alanine scan derivatives of hDim1 has been reported (5). For expression in S. pombe, a series of alanine scan mutants discussed in ref 5 were cloned into the vector pMNS21L. We have previously found that extended storage of hDim1 samples (1-2 months at 4 °C or ∼1 week at room temperature) results in complete cleavage of the polypeptide chain at positions 128 or 129 (2, 5). The mechanism of the cleavage awaits further study because the flanking amino acids do not match the recognition sequence for any known peptidase. Truncated forms (1-128/129) of the mutant proteins were also obtained after prolonged storage of the purified proteins. The short peptides were removed by using a centrifugal filter device (Ultrafree; Millipore, Bedford, MA) with a 5 kDa molecular weight cutoff. The molecular weight of the purified truncated proteins was determined by mass spectroscopy, which showed two peaks corresponding to

Zhang et al. 1-128 and 1-129 fragments, but no peak corresponding to full-length protein. Circular Dichroism Spectroscopy. CD spectra were acquired on an Aviv 62A spectrometer (Aviv, Lakewood, NJ). GuHCl-induced unfolding transitions were measured by recording the ellipticity at 222 nm on a series of Dim1 samples (11.5 µM in 50 mM sodium phosphate buffer, pH 7.82) prepared by mixing aliquots of equimolar stock solutions of native and fully unfolded (6.5 M GuHCl) Dim1. The unfolding equilibrium of reduced Dim1 was measured in the presence of 2 mM DTT. CD spectra over the range from 190 to 300 nm were recorded on oxidized and reduced (1 mM DTT) samples of wild-type and mutant Dim1 (4 µM) in 10 mM potassium phosphate, pH 7.82. Protein concentration was determined by measuring the absorption at 280 nm in 6.5 M GuCl. All CD measurements were performed in 2-mm quartz cuvettes thermostated at 20 °C. NMR Data Collection. NMR data were collected on a Bruker DMX-600 spectrometer equipped with a 5-mm x,y,zshielded pulsed-field gradient triple-resonance probe. Stable hDim11-128 samples at concentrations suitable for NMR studies can be prepared only at pH higher than 7.5. Better spectra are obtained at physiological temperature (37 °C) than at room temperature (20 °C). Hence, all NMR experiments were carried out at pH 7.65 and 37 °C. NMR data processing and sequential assignments have been described (2). Side-chain assignments were obtained from 3D 15N TOCSY-HSQC (20) and 3D HCCH-TOCSY (21, 22). Distance restraints were obtained from cross-peak volumes in 3D 15N NOESY-HSQC and 3D 13C-edited aliphatic NOESY (20). Cross-peaks were categorized as weak (1.8-5 Å), medium (1.8-3.5 Å), or strong (1.4-2.7 Å) in each data set based on the ratio of the peak volume of cross-peaks to that of the relevant diagonal peak. The ratio of weak, medium, and strong peaks was approximately 57:23:13. 3 JHNHR coupling constants were obtained from a 3D HNHA experiment (23, 24). Backbone φ angles were constrained to -60° ( 30° for 3JHNHR < 5.5 Hz, -120° ( 40° for 3JHNHR > 9 Hz, -120° ( 50° for 3JHNHR between 8-9 Hz, -120° ( 60° for 3JHNHR between 7-8 Hz. For 3JHNHR between 5.5-7 Hz, the backbone φ angles were constrained to -80° ( 30° and omitted if such restraints were violated. Hydrogen bonds were assigned to slowly exchanging amide protons from a series of 46 min HSQC spectra, with the first spectrum initiated about 2 h after the start of buffer exchange, using a Millipore 5K filter. In the first HSQC spectrum of the hydrogen exchange experiment, 42 crosspeaks were observed. Among these, 14 were confirmed to be involved in hydrogen bonds by the measurement of 3hJNC′ coupling across hydrogen bond from a 3D HNCO experiment optimized for magnetization transfer across the weak coupling between H-bonded amide nitrogen and proton-acceptor carbonyl carbon (25). Each hydrogen bond was specified as two distance restraints (HN-O distance and N-O distance). In the first round of structure calculation, only the hydrogen bonds identified by both experiments were used. Structure Calculation. Structure calculations were performed on a Silicon Graphics O2 workstation, using the simulated annealing protocol within CNS1.0 (see http:// cns.csb.yale.edu). In each round of structure calculation, random structures were generated from the initial structure and subjected to 1000 steps of high-temperature dynamics

Solution Structure and Mutational Analysis of Dim1 at 50 000 K for a total of 15 ps. Each structure was then cooled to 0 K over a period of 15 ps (1000 steps at 15 fs/ step). A second, slow-cool annealing was carried out over a total of 15 ps (3000 steps at 5 fs/step) from 2000 to 0 K. Restrained energy minimization was then performed for 2000 steps. The initial structure for the first round of calculation was an extended peptide generated from the sequence. The initial structure for subsequent calculations was one of the structures with low CNS energy from the previous round. In the final round, a family of 120 structures was obtained with root mean square deviation (rmsd) for bond lengths