Solution structure of the carboxyl-terminal LIM domain from quail cysteine-rich protein CRP2.

Proteins of the cysteine-rich protein (CRP) family (CRP1, CRP2, and CRP3) are implicated in diverse processes linked to cellular differentiation and growth control. CRP proteins contain two LIM domains, each formed by two zinc-binding modules of the CCHC and CCCC type, respectively. The solution structure of the carboxyl-terminal LIM domain (LIM2) from recombinant quail CRP2 was determined by multidimensional homo- and heteronuclear magnetic resonance spectroscopy. The folding topology retains both independent zinc binding modules (CCHC and CCCC). Each module consists of two orthogonally arranged antiparallel beta-sheets, and the carboxyl-terminal CCCC module is terminated by an alpha-helix. 15N magnetic relaxation data indicate that the modules differ in terms of conformational flexibility. They pack together via a hydrophobic core region. In addition, Arg122 in the CCHC module and Glu155 in the CCCC module are linked by an intermodular hydrogen bond and/or salt bridge. These residues are absolutely conserved in the CRP family of LIM proteins, and their interaction might contribute to the relative orientation of the two zinc-binding modules in CRP LIM2 domains. The global fold of quail CRP2 LIM2 is very similar to that of the carboxyl-terminal LIM domain of the related but functionally distinct CRP family member CRP1, analyzed recently. The carboxyl-terminal CCCC module is structurally related to the DNA-binding domain of the erythroid transcription factor GATA-1. In the two zinc-binding modules of quail CRP2 LIM2, flexible loop regions made up of conserved amino acid residues are located on the same side of the LIM2 domain and may cooperate in macromolecular recognition.

Tetrahedral zinc-binding domains are important structural elements in a wide variety of proteins, and more than 10 different classes of such Zn(II)-binding motifs have been identified and biochemically characterized, many of them in proteins specifically interacting with nucleic acids (1,2). The four coordinating ligands in the tetrahedral zinc-binding sites are composed of cysteine sulfur, histidine imidazole nitrogen, or, occasionally, oxygen from a glutamate or aspartate side chain. The LIM 1 motif defines one class of zinc-binding domain and was originally recognized in, and named after, the protein products of the lin-11, isl-1, and mec-3 genes (3,4). The gene products of lin-11 and mec-3 transcriptionally regulate genes involved in cell fate determination and differentiation in Caenorhabditis elegans, and the isl-1 gene encodes a rat insulin I gene enhancer-binding protein. LIM domains are found in 1-5 copies in many different proteins of diverse functions, either alone or associated with distinct domains of defined function like homeodomains or protein kinase domains (5)(6)(7). The LIM motif is basically composed of two zinc finger structures separated by a 2-amino acid spacer and conforms to the consensus sequence CX 2 CX 16 -23 HX 2 CX 2 CX 2 CX 16 -21 CX 2 (C/H/D) (5)(6)(7). Spectroscopic studies of LIM domains derived from different LIM proteins revealed that each double finger LIM domain specifically binds two zinc ions (8 -11). A distinct family of genes, the CSRP genes, encode a specific class of LIM proteins, termed cysteine-rich proteins (CRPs) (12). CRP proteins contain 192-194 amino acid residues and exhibit two LIM domains, termed LIM1 (amino-terminal) and LIM2 (carboxylterminal). CRP LIM1 and LIM2 domains invariably conform to the 52-amino acid consensus sequence CX 2 CX 17 HX 2 CX 2 CX 2 CX 17 CX 2 C and are separated from each other by 56 -59 amino acids (12). Each CRP LIM motif contains two tetrahedral Zn(II)-coordinating modules, an amino-terminal S 3 N 1 site of the CCHC type, and a carboxyl-terminal S 4 site of the CCCC type (8 -10).
The expression patterns of CSRP genes and the structural properties of their CRP protein products suggest that these genes may have important roles in the regulation of cell differentiation and proliferation. The CSRP1 gene was shown to have properties typical for a primary response gene (13,14) and its protein product, CRP1, was found to be associated with specific components of the cytoskeleton (15,16). The CSRP2 gene encoding the CRP2 protein was discovered on the basis of its strong suppression in avian fibroblasts transformed by retroviral oncogenes or chemical carcinogens (17). The suppression of CSRP2 gene expression directly correlates with the transformed phenotype of avian fibroblasts in a conditional transformation system (12) and with the proliferative state of rat arterial smooth muscle cells after mitogenic stimulation (18). The CSRP3 gene was isolated on the basis of its induced expression during rat skeletal muscle differentiation, and its protein product, CRP3 (or MLP for muscle LIM protein), was shown to be a positive regulator of myogenesis (19). In pairwise alignments, the avian homologs of the three members of the CRP family of LIM proteins share 63-76% identical residues in their amino acid sequences and hence represent related but distinct members of this protein family (12). The precise biochemical function of LIM domains in general and of CRP proteins in particular has not been defined yet. The solution structure of the carboxyl-terminal LIM domain of chicken CRP1 was determined by nuclear magnetic resonance spectroscopy, and the protein fold of the tetrathiolate CCCC module was shown to be strikingly similar to that reported for the DNA-interactive CCCC modules within the DNA binding domains of the erythroid transcription factor GATA-1 and of the glucocorticoid receptor (20). Despite this modular structural similarity to DNAbinding proteins, specific interaction of CRP proteins with nucleic acids has not yet been demonstrated. On the contrary, it has been inferred from protein affinity assays that CRP LIM domains are involved in specific protein-protein interactions (21)(22)(23).
So far, the solution structures of three LIM domains from unrelated LIM proteins have been determined by nuclear magnetic resonance spectroscopy: the carboxyl-terminal LIM domain (LIM2) from chicken CRP1 (20), the single LIM domain from the developmentally regulated rat cysteine-rich intestinal protein (CRIP) (24), and the amino-terminal CCHC Zn(II)binding module of the single LIM domain from the Lasp-1 protein encoded by a gene that was identified on the basis of its overexpression in human breast carcinoma (25). Here we present the solution structure of the carboxyl-terminal LIM domain (LIM2) from quail CRP2 and assess structural conservation and diversity between closely related but distinct members of the CRP family of LIM domain proteins that apparently fulfill diverse functions in cellular differentiation and growth control.

Construction of a pET Expression Plasmid Encoding Recombinant
Quail CRP2(LIM2)-A polymerase chain reaction was performed using DNA from the gt10 clone W15 containing quail CSRP2 (qCSRP2) cDNA (17) as a template and the oligonucleotides 5Ј-d(CTAACCATG-GACAGGGGAGAG)-3Ј (SW001) and 5Ј-d(CTTATGAGTATTTCTTC-CAGGGTA)-3Ј (gt10 reverse sequencing primer) as 5Ј and 3Ј primers, respectively. The SW001 primer corresponds to nucleotides 245-265 of the published qCSRP2 cDNA sequence (17) with nucleotide substitutions at its 5Ј end introducing a novel NcoI site. The polymerase chain reaction product was first digested with HindII cutting at a site in the 3Ј-untranslated region of CSRP2 cDNA and then partially digested with NcoI to cut at the site generated by primer SW001 but preserving an internal NcoI site. The 435-nucleotide fragment was ligated into expression plasmid pET3d (26), which had been cut by BamHI, filled in by Klenow DNA polymerase, and then digested by NcoI. To rule out polymerase chain reaction-induced mutations and to verify the integrity of the CSRP2 coding region, the total nucleotide sequence of the inserted polymerase chain reaction fragment was determined by the dideoxynucleotide chain termination method using the T7 sequencing kit (Pharmacia, Vienna, Austria) and pET-specific primers. The expression plasmid pET3d-qCRP2(LIM2) encodes a 113-amino acid peptide encompassing amino acids 82-194 of qCRP2 including the carboxylterminal LIM domain (LIM2) (12,17).
Purification of Recombinant Quail CRP2(LIM2)-For the expression of the qCRP2(LIM2) protein in bacteria, pET3d-qCRP2(LIM2) was transformed into Escherichia coli strain BL21(DE3)pLysS (26). Bacteria were grown at 37°C in NZCYM medium containing ampicillin (100 g/ml) and chloramphenicol (25 g/ml) to an optical density at 600 nm of 0.5. Cells were induced to express qCRP2(LIM2) by the addition of isopropyl-␤-D-thiogalactoside (Boehringer, Vienna, Austia) to a final concentration of 0.5 mM, and incubation was continued for 3 h at 37°C. The cells were collected by centrifugation and resuspended in 20 ml of ice-cold buffer A (50 mM sodium phosphate, pH 6.4, 10 mM NaCl, 0.1% (v/v) ␤-mercaptoethanol) per liter of the original bacterial culture. All subsequent steps were carried out at 4°C. Bacteria were lysed by a freeze (Ϫ80°C)-thaw cycle, and the cell lysate was cleared by centrifugation at 23,000 ϫ g for 35 min. The supernatant containing the soluble protein fraction was loaded onto a CM-52 cation exchanger (Whatman, Maidstone, United Kingdom) column equilibrated in buffer A. The column (bed volume, 35 ml) was washed with approximately 50 ml of buffer A until the eluting solution showed background absorbance at 260 and 280 nm. Elution of qCRP2(LIM2) was achieved with 30 ml of buffer B (50 mM sodium phosphate, pH 8.0, 10 mM NaCl, 0.1% (v/v) ␤-mercaptoethanol). Pooled protein-containing fractions of the eluate were analyzed by a photometric assay (27) and by SDS-polyacrylamide gel electrophoresis (15%, w/v) to determine protein concentration and purity, respectively. The final yield of homogeneous qCRP2(LIM2) was approximately 12 mg/liter of bacterial culture. The structural integrity and purity of the protein preparation was verified by amino-terminal sequencing, and the stoichiometry of bound zinc ions was analyzed by atomic absorption spectroscopy and electrospray ionization mass spectrometry.
Concentration of protein solutions for NMR analysis was achieved by dialysis against buffer C (20 mM potassium phosphate, pH 7.2, 50 mM KCl, 0.5 mM dithiothreitol) and centrifugation of the dialyzed solution through Centriprep 10 ultrafiltration filters (Amicon, Witten, Germany). The final protein concentrations of qCRP2(LIM2) solutions used for NMR analysis ranged from 1.2 to 2.2 mM (14.5-26.6 mg/ml). 15 N labeling of qCRP2(LIM2) was performed by growing bacteria in minimal medium (4.8 g of Na 2 HPO 4 , 3 g of KH 2 PO 4 , 0.5 g of NaCl, 1 g of 15 NH 4 Cl in 1 liter of water) supplemented with 20 ml of an 18% (w/v) glucose solution, 2 ml of 1 M MgSO 4 , 4 ml of 10 mM ZnSO 4 , and ampicillin and chloramphenicol to final concentrations of 100 g/ml and 25 g/ml, respectively. 15 NH 4 Cl (98% isotope purity) was purchased from CIL (Andover, MA). After reaching an optical density of 0.5 at 600 nm, cells were induced to express 15 N-labeled qCRP2(LIM2) by the addition of isopropyl-␤-D-thiogalactoside to a final concentration of 0.5 mM, and incubation was continued for 5 h at 37°C. The purification procedure was as described above. The final yield of purified 15 Nlabeled qCRP2(LIM2) was approximately 25 mg/liter of bacterial culture.
NMR Analyses-NMR experiments were performed on a Varian UNITYPlus 500-MHz spectrometer equipped with a pulse field gradient unit and triple resonance probes with actively shielded z gradients. The NMR sample contained 1-2 mM qCRP2(LIM2), 20 mM potassium phosphate, pH 7.2, 50 mM KCl, 0.5 mM dithiothreitol in 90% H 2 O, 10% 2 H 2 O. NMR spectra were processed and analyzed using Varian Vnmr and NMRPipe software systems (28). Spectra recorded for spin system identification and sequential assignment include SS NOESY (150, 200 ms) (29), TOCSY (45, 75 ms) (30), two-dimensional 15 N-filtered and 15 Nedited NOESY (150 ms) (31,32), sensitivity-enhanced two-dimensional 15 N HSQC (33), three-dimensional 15 N TOCSY-HSQC, and three-dimensional 15 N NOESY-HSQC (34). Spectra were recorded at 26 and 35°C in order to resolve the residual water signal from some H ␣ protons. The NMR properties of the protein did not change significantly over this temperature range. Signal assignment was carried out at 26°C. Two-dimensional homonuclear and heteronuclear experiments were processed with shifted gaussians in both dimensions. The TOCSY spectrum resulted in a 512 ϫ 1024 data matrix with 32 scans per t 1 value, using a WATERGATE (35) double echo sequence for water suppression and a DIPSI-2 (36) mixing sequence. A two-dimensional NOESY spectrum was collected using a z filter prior to acquisition and a 300-ms selective excitation pulse (29). The NOESY and the twodimensional 15 N-edited and 15 N-filtered NOESY spectra resulted from a 512 ϫ 1024 data matrix with 32 scans per t 1 value. The one-bond 1 H-15 N shift correlation (HSQC) spectrum (33) of qCRP2(LIM2) resulted from a 2 ϫ 64 ϫ 1024 data matrix size, with 16 scans per t 1 value and a delay time between scans of 1 s. Decoupling (during acquisition) was achieved with the use of the GARP decoupling sequence (37), using a 1.5-kHz radio frequency field. Shifted squared sine bell windows were used both in t 1 and t 2 . The three-dimensional TOCSY-HSQC and threedimensional NOESY-HSQC experiments were performed with water flip-back pulse (38) and PFG sensitivity enhancement (33). The data (64 ϫ 32 ϫ 1024) were doubled by linear prediction in both indirect dimensions, processed using 80°shifted squared sine bells, and zerofilled to 256 (t 1 ) and 128 (t 2 ) points, respectively. DIPSI-2 and NOESY mixing times were set to 50 and 120 ms, respectively. Qualitative 3 J (H N H ␣ ) scalar coupling information was obtained from one-dimensional traces through the standard two-dimensional PFG 15 N HMQC experiment (39). Data size was 512 ϫ 1024, with 16 scans per t 1 value. For the measurements of the 15 N attenuation factors, two sets of spectra with and without presaturation of the water signal were recorded with parameters identical to the PFG sensitivity-enhanced two-dimensional 15 N HSQC experiment (33). The attenuation factor is given as the ratio of the signal intensities in these two experiments (40). Dynamic information was obtained by measuring 15 N T 1 and T 2 relaxation, as described by Farrow et al. (41) and analyzed according to Habazettl and Wagner (42). Experimental and processing parameters were identical to the PFG sensitivity-enhanced two-dimensional 15 N HSQC experiment. Relaxation delays were 0, 45, 90, 135, and 180 ms for the T 2 and 0, 150, 300, 450, and 600 ms for the T 1 measurements, respectively. Relaxation times were obtained by measuring peak heights using nonlinear least squares curve fitting (28) with three adjustable parameters.
Structure Calculation-Three-dimensional structures were generated using experimentally observed NOE constraints in a simulated annealing and energy minimization protocol (43) using the program X-PLOR (44) on SGI Crimson and Indigo2 workstations. In the first stage of calculation, 247 inter-and intra-NOE restraints were applied to a template structure with randomized and angles and extended side chains to generate a set of 100 structures. NOE constraints were classified as strong (1.8 -3.0 Å), medium (1.8 -4.0 Å), and weak (1.8 -5.0 Å). Initial structure calculations were performed without the zinc ions. However, the zinc coordination sites were defined by enforcing tetrahedral geometry of residues Cys 120 , Cys 123 , His 141 , and Cys 144 and of residues Cys 147 , Cys 150 , Cys 168 , and Cys 171 , respectively. Upper and lower distance limits for the zinc coordination site were (in Å) 3.30 Յ S ␥ -S ␥ Յ 3.50, 3.30 Յ S ␥ -N ␦1 Յ 3.50. 36 structures with minimal constraint violations were selected. In the next step, upper and lower bounds were further refined, accounting for the respective maximum and minimum distances from the 36 structures. During the final refinement, the number of distance constraints was increased to 393, as well as 17 dihedral angle constraints. The resulting 15 structures with minimal constraint violations were used for final refinement using a restrained Powell energy minimization with the CHARMM force field (45). The final refinement also included 19 hydrogen bonding restraints, based on measured attenuation factors (NH exchange rates), distance restraints defining the acceptor for hydrogen bonds, and force field parameters given by Lee et al. (46). In particular, zinc was covalently attached to S ␥ of Cys and to N ␦1 of His. Zinc-ligand bonds were assigned . The highly soluble recombinant protein was purified to homogeneity in a single step employing CM-52 cation exchange chromatography. The identity and purity of recombinant qCRP2(LIM2) was confirmed by aminoterminal amino acid sequencing, which revealed that a minor portion (Ͻ5%) of the protein preparation lacked the initiating methionine. Atomic absorption spectroscopy showed that purified recombinant qCRP2(LIM2) contained 1.8 Ϯ 0.1 mol of zinc/mol of protein. An alignment of the amino acid sequence of the qCRP2(LIM2) peptide used in this study with the sequence of the corresponding segment from chicken CRP1 is shown in Fig. 1A. The sequence identity within this region is 77.5%, while between the native chicken CRP1 and quail CRP2 proteins it is 76.6% (chicken and quail CRP1 proteins are identical, and chicken and quail CRP2 proteins differ by a single amino acid substitution) (12). The schematic structure of the quail CRP2 LIM2 domain with the CCHC and CCCC zinc-binding modules is shown in Fig. 1B.
NMR Analyses-The chemical shift dispersion in a sensitivity-enhanced PFG two-dimensional 15 N-1 H HSQC spectrum (Fig. 2) indicates that qCRP2(LIM2) adopts a well folded structure in aqueous solution; however, only a fraction of all NH signals are visible, presumably due to fast exchange with bulk water at this pH (7.2) and conformational exchange contributions. A total of 61 secondary backbone amides are visible. Examination of the spectrum indicated that only one structural form of the protein was present in solution, since there is one cross-peak for the amide of each non-proline residue in the protein. Signal assignment followed the well established strategy (48). In the three-dimensional TOCSY-HSQC, spectral identification of the spin systems was straightforward. The sequential assignment was achieved by combining information from two-dimensional homonuclear NOESY and three-dimensional 15 N NOESY-HSQC experiments. To further disentangle NOESY peaks in the downfield region of the proton spectra, isotope filter techniques were applied to selectively monitor NOEs involving either amide or aromatic protons. Both, 15 Nfiltered (aromatic-aliphatic NOEs, Fig. 3) and 15 N-edited (amide-aromatic/aliphatic NOEs) two-dimensional NOESY spectra were recorded. In some cases, side chain protons of residues with longer side chains could not be assigned unambiguously due to signal degeneracies and overlap in the upfield region of the TOCSY data.
Secondary structure elements were identified primarily on the basis of cross-peak patterns observed in homonuclear twodimensional NOESY, 15 N-filtered, 15   c Pairwise r.m.s. deviations relative to mean (͗DG-SA͘) calculated from the superposition of backbone heavy atoms CЈ, C ␣ , and N of the 15 structures generated by distance geometry-simulating annealing (DG-SA) methods. d Pairwise r.m.s. deviations relative to mean (͗DG-SA͘) calculated from the superposition of all non-hydrogen atoms for the given range of residues of the 15 DG-SA structures.
e Numbering of residues is as in Fig. 1 Fig. 4A. There is fairly good agreement between the secondary shift and the definition of secondary structure elements from NOE data. In particular, ␤-strands ␤II, ␤III, ␤IV, ␤VII, and ␤VIII exhibit diagnostic secondary shifts, in contrast to ␤-strands ␤I, ␤V, and ␤VI. 3 J(H N H ␣ ) scalar coupling constants taken from traces in the 15 N HMQC spectrum additionally corroborate the secondary structure assignment. Significant couplings were found for residues Cys 123 , Val 127 , Tyr 128 , Glu 131 , Ile 134 , Lys 138 , Trp 140 , Cys 144 , Lys 149 , Cys 150 , Leu 154 , Thr 160 , Glu 161 , Lys 162 , Glu 165 , Ile 166 , and Tyr 167 and could be correlated with NOESY information indicating ␤-sheet structures. Cys 123 is located in a turn region. Amide proton attenuation factors (i.e. the retardation of intermolecular exchange of amide protons with bulk water) were used to monitor hydrogen bonding effects. Fig. 4B shows the measured attenuation factors as a function of residue position. It is evident that there is a significant decrease in amide proton attenuation factors for residues located in loop regions of qCRP2(LIM2). This indicates greater solvent exposure and facilitated solvent accessibility of these protons compared with residues found in secondary structure elements. Of particular interest are the significantly higher attenuation factors for amide protons of residues Cys 120 , Val 127 , Val 133 , Ile 134 , Phe 145 , Lys 162 , Glu 163 , Glu 165 , Ile 166 , Cys 171 , Tyr 172 , and Ala 173 . They correspond to well defined hydrogen bonds within secondary structure elements, both ␤-sheet structures and the carboxylterminal ␣-helix. 15 N magnetic relaxation data were interpreted in terms of the simplified method proposed by Habazettl and Wagner (42) and helped identify secondary structure elements. 2/T 2 Ϫ 1/T 1 values are sensitive to slow motions on millisecond to microsecond time scales. Fig. 4C shows the distribution of 2/T 2 Ϫ 1/T 1 along the backbone of qCRP2(LIM2). There is a good correlation between backbone 15 N dynamics (Fig. 4C), secondary structure, and hydrogen exchange (Fig. 4B). Residues within loop regions exhibit significantly higher 2/T 2 Ϫ 1/T 1 values compared with residues located in secondary structure elements. Since hydrogen exchange depends on structural isomerization processes (i.e. the breakage of a blocking hydrogen bond), these values indicate conformationally flexible sites that transiently form open, unprotected states in which the exchangeable amide protons are accessible to the hydrogen exchange catalyst. In structured regions of the COOH-terminal CCCC module, the 2/T 2 Ϫ 1/T 1 values are quite uniform, the average value being 8.3 Hz. The CCHC module generally exhibits higher 2/T 2 Ϫ 1/T 1 values (average of 9.2 Hz), suggesting a less rigid structure for the CCHC module compared with that of the CCCC module. However, to more specifically define the motional characteristic of qCRP2(LIM2) (e.g. fast motion), heteronuclear 1 H-15 N NOE data and a more elaborate model including motional anisotropy must be used.
Structural information in the form of distance constraints was derived from two-dimensional NOESY and three-dimensional NOESY-HSQC spectra. From the two-dimensional 15  filtered NOESY experiment (Fig. 3) additional distance constraints involving aromatic and aliphatic protons in the CCHC module could be obtained. Structure calculations were performed in two consecutive steps (see "Experimental Procedures"). The final structures have no NOE violations greater than 1.0 Å (Table I). A superposition of the backbone coordinates from the final 15 X-PLOR structures is shown in Fig. 5. The average r.m.s. deviation from the mean structure for ordered backbone atoms (N, C ␣ , and CЈ), including residues 119 -173 of the qCRP2(LIM2) domain is 0.98 Ϯ 0.29 Å (Table I), which is comparable with a so-called "second generation structure" (51). Better convergence is obtained when the aminoterminal CCHC module and the carboxyl-terminal CCCC module are compared independently (residues 119 -144, 0.87 Ϯ 0.22 Å; residues 145-173, 0.81 Ϯ 0.17 Å).
Tertiary Structure of qCRP2(LIM2) and Relation to Chicken CRP1(LIM2)-A schematic ribbon drawing of the carboxylterminal domain of qCRP2(LIM2) (residues 118 -174) is presented in Fig. 6, A and B. The domain starts out with an anti-parallel ␤-sheet (residues Lys 119 -Tyr 128 ), connected via a rubredoxin type turn ("Rd knuckle") (52), with characteristic hydrogen bonding between S ␥ Cys 120 and H N Arg 122 as well as a hydrogen bond between S ␥ Cys 123 and H N Asp 125 . It is followed by a second ␤-sheet, which is oriented perpendicular to the first one (Lys 132 -His 141 ). The linker regions between the two ␤-sheets as well as that between the two antiparallel ␤-strands ␤III and ␤IV are flexible and appear to have no noticeable hydrogen bonding. The residues Lys 142 -Cys 144 form a turn and thus complete the amino-terminal CCHC module. NMR spectral data (low attenuation factor, high 2/T 2 Ϫ 1/T 1 value) for Asn 143 and Cys 144 indicate conformational flexibility for these residues, although Cys 144 is coordinated to the zinc ion. Similar observations were made for the residues involved in zinc binding within the zinc finger DNA binding domain of Xfin (53) and some of the ligand binding cysteines of E. coli Ada (54). Within the CCCC module, residues Phe 145 -Leu 154 comprise a third anti-parallel ␤-sheet, again containing a rubredoxin type turn, with a similar hydrogen bonding pattern between H N Lys 149 and S ␥ Cys 147 and between H N Lys 152 and S ␥ Cys 150 . Following a conformationally flexible loop from Glu 155 to Thr 157 , a final antiparallel ␤-sheet is formed by residues Thr 160 -Cys 168 . This sheet is the most regular ␤-sheet structure in both modules, as can be seen not only by the NOE connectivities, but also by the chemical shift index and scalar coupling constants (48). A short ␣-helix starts out at residue Gly 170 , although only reasonably well defined within residues Cys 171 -Lys 174 . Residues 82-118 and 175-194 of qCRP2(LIM2) (cf. Fig.  1) were not visible in the 15 N HSQC spectra and thus could not be analyzed.   6. Ribbon drawing of qCRP2(LIM2). A and B, orthogonal views of a selected representative structure from the calculated set for qCRP2(LIM2) (residues 118 -174), including the zinc-binding cysteine (yellow) and histidine (blue) side chains. The zinc ions are represented by spheres (magenta). In B, side chains of the hydrophobic core residues (purple) and of the interacting Arg 122 (green) and Glu 155 (red) are also shown. hydrogen bond acceptor was not a backbone carbonyl but a side chain functional group, most likely a carboxyl group. Further inspection revealed a spatial proximity of Glu 155 and Arg 122 side chain functional groups, and thus it was concluded that H N⑀ Arg 122 and/or H N Arg 122 were forming a hydrogen bond and/or salt bridge to the side chain carboxyl group of Glu 155 . Significantly, these two residues are absolutely conserved within the CRP family of LIM proteins (12), suggesting that they are important determinants for the relative orientation of the two zinc finger modules in the CRP LIM2 domains. In contrast, in the CRIP LIM domain the corresponding amino acid positions are Lys and Thr, and the orientation of the two modules is different from that in CRP LIM2 domains (24). This may indicate that not only hydrophobic interactions in the core region but also salt bridges or hydrogen bonds are important elements contributing to the global fold of the CRP LIM2 domain. There is also evidence (strong NOEs between H ␦2 His 141 and H ␤,␥ Glu 131 ) for hydrogen bonding between H N⑀2 His 141 and the carboxyl group of Glu 131 . This was also found for chicken CRP1 (20) and CRIP (24), and it was suggested that this is an important interaction for defining the conformation of the CCHC module.
Similar to the CCCC modules of chicken CRP1(LIM2) and CRIP (20,24), the CCCC module of qCRP2(LIM2) shows striking structural similarities to the DNA-interacting CCCC modules of the glucocorticoid receptor and GATA-1 DNA-binding domains (56,57) and hence may also form a DNA-contacting structure possibly involved in CRP-nucleic acid interactions. Conformational flexibility in the Ser 156 -Thr 158 loop segment of the qCRP2(LIM2) CCCC module is intriguing, given that it is highly conserved in all CRP proteins and partially conserved between CRP and the DNA-binding GATA-1 and steroid hormone receptor proteins. Furthermore, the loop segment Ala 129 -Glu 131 connecting the ␤II and ␤III strands in the CCHC module of qCRP2(LIM2) exhibits conformational flexibility, and this segment is, again, absolutely conserved between CRP proteins (12). These two conserved loop segments in the CCHC and CCCC modules, exhibiting conformational disorder, are located at the same side of the qCRP2(LIM2) molecule (Fig. 6, A and  B), and it is tempting to suggest that their conformational flexibilities may be relevant for the fine tuning of intermolecular interactions with a putative DNA target and optimization of the binding interface. Further biochemical and structural analyses of CRP proteins, including analyses of the aminoterminal LIM1 domain and of putative functional cooperativity between LIM1 and LIM2, will be important to aid in the unequivocal identification of the cellular targets for these proteins and to elucidate the molecular basis for their diverse physiological functions.