Mutagenesis of the Runt Domain Defines Two Energetic Hot Spots for Heterodimerization with the Core Binding Factor β Subunit*

Core-binding factors (CBFs) are a small family of heterodimeric transcription factors that play critical roles in several developmental pathways and in human disease. Mutations in CBF genes are found in leukemias, bone disorders, and gastric cancers. CBFs consist of a DNA-binding CBFα subunit (Runx1, Runx2, or Runx3) and a non-DNA-binding CBFβ subunit. CBFα binds DNA in a sequence-specific manner, whereas CBFβ enhances DNA binding by CBFα. Both DNA binding and heterodimerization with CBFβ are mediated by a single domain in the CBFα subunits known as the “Runt domain.” We analyzed the energetic contribution of amino acids in the Runx1 Runt domain to heterodimerization with CBFβ. We identified two energetic “hot spots” that were also found in a similar analysis of CBFβ (Tang, Y.-Y., Shi, J., Zhang, L., Davis, A., Bravo, J., Warren, A. J., Speck, N. A., and Bushweller, J. H. (2000) J. Biol. Chem. 275, 39579–39588). The importance of the hot spot residues for Runx1 function was demonstrated in in vivo transient transfection assays. These data refine the structural analyses and further our understanding of the Runx1-CBFβ interface.

human leukemia (4 -9). RUNX2 and CBFB are required for bone development, and mutations in RUNX2 cause the human skeletal disorder cleidocranial dysplasia (10 -15). Disruption of the Runx3 gene in mice causes hyperplasia of gastric mucosa (16) and severe limb ataxia (17,18). Runx3 is thought to be required for the controlled proliferation and apoptosis of gastric epithelial cells, the survival of dorsal root ganglia proprioreceptive neurons, and for epigenetic silencing of the CD4 gene in cytotoxic T lymphocytes (16 -19). Hemizygous deletion and hypermethylation of RUNX3 is found in a significant number of primary gastric cancers (16).
The DNA binding domain in CBF␣ is known as the "Runt domain" (20). The Runt domain is responsible for DNA binding as well as for heterodimerizing with the CBF␤ subunit (2,20,21). The Runt domain of the CBF␣ subunits is an s-type immunoglobulin fold in the p53 family of DNA-binding transcription factors, whose other members include STAT3␤, p53, NF-B, NFAT, and the Brachyhury T box family of proteins (22)(23)(24). DNA binding by the Runt domain is mediated by loops and ␤-strands at one end of the immunoglobulin ␤-barrel (25,26). The DNA is bent toward the protein by ϳ20° (25)(26)(27). CBF␤ increases the DNA binding affinity of the Runt domain by ϳ7-10-fold (25,28). CBF␤ binds to a region of the Runt domain distinct from the DNA-binding sequences and contacts neither the DNA nor amino acids in the Runt domain that are directly involved in DNA binding (25,26,29,30).
Comparison of the Runt domain-DNA and Runt domain-CBF␤-DNA complexes (25,26,31) with the recently determined crystal structures of the Runt domain in the free state (30,32) suggests a model for allosteric regulation of DNA binding by CBF␤. The Runt domain contains several regions that are likely to equilibrate between at least two conformations in the free state, including the ␤C-D, ␤EЈ-F, and ␤G-GЈ loops. CBF␤ appears to stabilize a specific conformation of the Runt domain. The most dramatic change occurs in the ␤G-GЈ loop that assumes two conformations in the free and complexed state crystal structures and has been referred to as the "Sswitch" (32). The dramatic change in the ␤G-GЈ loop and the changes in the ␤C-D loop are in agreement with an earlier study by NMR chemical shift perturbation that showed dramatic changes in the C␣ chemical shift for residues in the ␤G-GЈ loop indicative of a significant conformational change (28).
Previous studies in our laboratory and others identified amino acid side chains in CBF␤ that contribute functionally to heterodimerization with the Runt domain (25,29,33). Residues in the Runt domain that mediate heterodimerization with CBF␤ were identified by NMR spectroscopy and x-ray crystal-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  lography (23,25,26,31,34). Some of those residues have been analyzed functionally (25,29); however, a comprehensive quantitative study of all amino acid side chains in the Runt domain implicated by structural analyses to form contacts to the CBF␤ subunit has not yet been reported. Here we determine the energetic contribution of the amino acids in the Runt domain that contact CBF␤ to heterodimer formation. In doing so we identify two energetic hot spots at the heterodimerization interface involving Runt domain residues Asn-109 and Thr-161, and we discuss the importance of these contacts in terms of the overall ternary complex structure.

EXPERIMENTAL PROCEDURES
Protein Purification-We introduced alanine substitutions into a cDNA encoding the murine Runx1 Runt domain and subcloned the cDNA sequences into the bacterial pET-3c (Novagen) expression vector as described by Li et al. (35). We labeled the wild type and mutated Runt domains (residues 41-214, numbered according to Bae et al. (36)) with 15 N and purified them as described previously (23,37).
We subcloned the CBF␤141 cDNA, which encodes the CBF␤ heterodimerization domain (amino acids 1-141), into the bacterial expression vector pET3c (Novagen) using NdeI and BamHI sites. We transformed the resulting plasmid pET3c-CBF␤141 into Rosetta (DE3) cells (Novagen) and overexpressed the CBF␤ (141) protein at 37°C. We harvested cells from a 1-liter culture and resuspended them in 10 ml of 50 mM Tris-Cl (pH 7.5), 10 mM EDTA, and 25% sucrose. We then sonicated the cells, collected the insoluble inclusion body by centrifugation, and denatured it with 7 M urea. We passed the denatured protein through a 50-ml DEAE-Sephacel (Amersham Biosciences) column in the presence of 7 M urea, collected the protein fractions, and refolded the CBF␤ (141) protein by serial dialysis in 25 mM Tris-Cl (pH 7.5), 200 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.05% Triton X-100, 10% ethylene glycol, 0.5 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride with 1, 0.5, and 0 M urea. We concentrated the proteins and loaded them onto a 550-ml S-100 (Amersham Biosciences) gel filtration column. We pooled and concentrated the fractions of interest, which contained a single protein band on SDS-PAGE gels as visualized by Coomassie Brilliant Blue staining (not shown).
NMR Spectroscopy-We performed all measurements on a Varian Inova 500-MHz NMR spectrometer equipped with an actively shielded triple resonance probe from Nalorac Corp. We prepared samples of Runt domain-DNA complexes and recorded 15 N-1 H HSQC spectra at 40°C as described previously (23).
Urea Denaturation Monitored by Fluorescence Spectroscopy-We detected the urea denaturation of the Runt domain by tryptophan fluorescence as described previously (37). We performed all fluorescence experiments using a protein concentration of 20 -40 M in a total volume of 0.8 ml of 25 mM Tris-HCl (pH 7.5), 50 M EDTA, and 1 mM dithiothreitol on a SPEX Fluoromax spectrofluorometer (Spex Industries, Edison, NJ). We incubated individual 0.8-ml solutions of the Runt domain with increasing concentrations of urea (0 -6.0 M) and equilibrated the samples for 2 h at 4°C prior to fluorescence measurement to follow the Runt domain denaturation. We excited the samples at a wavelength of 280 nm and detected the emission at 340 nm. We took background spectra of the buffer alone for later subtraction. We then recorded spectra in the range of 300 -370 nm with an increment of 4 nm and an integration time of 1 s, and averaged two scans to minimize error. We obtained the parameters for the unfolding curves by nonlinear, least squares fitting using two-state equations (Equation 1 and Equation 2) as described elsewhere (38 -40).
⌬G N-D is the free energy of unfolding in the absence of denaturant. We used Origin 5.0 software for nonlinear least square fitting. Equilibrium Binding Constant Measurements-We determined equilibrium binding constants for the wild type and mutated Runt domains by electrophoretic mobility shift assays (EMSA) using conditions described previously (6,28,33,(41)(42)(43). We fit all data points to a rearranged mass action equation, [PD]/[D t ] ϭ 1/(1 ϩ K d /[P]), using nonlinear least squares analyses (Kaleidagraph, Synergy Software).
Transient Transfection of P19 Cells and Luciferase Activity Measurements-We performed site-directed mutagenesis of a cDNA encoding the full-length (451 amino acid) murine Runx1 protein in pBluescript SKϩ as described above. We prepared pcDNA/Runx1 expression vectors by subcloning 1714-bp Runx1 cDNA fragments from pBluescript SKϩ between the EcoRI and XhoI sites of pcDNA3.1(ϩ) (Invitrogen). We used the TCR␤-LUC plasmid, which contains the firefly luciferase gene driven by a nucleotide 617-735 fragment from the T cell receptor ␤ chain enhancer (44), and the TCR␤(core)-LUC plasmid which contains the same fragment with mutations in the three CBF-binding sites in the enhancer (45) to detect Runx1 activity. We constructed both reporter plasmids by isolating a BamHI fragment containing the TCR␤ enhancer fragments from the previously described pTCR␤CAT plasmids (45), filling in the overhangs with Klenow polymerase and ligating the fragment into the SmaI site of pT81-LUC (46).
We seeded P19 cells at 2 ϫ 10 6 cells per well in 6-well plates the day before transfection, and we performed transient transfections with Li-pofectAMINE 2000 (Invitrogen) following the manufacturer's instructions. Typically, we mixed 1 g of the TCR␤-LUC reporter plasmids, 2 g of the pcDNA/Runx1 expression vectors, and 2 ng of transfection efficiency control plasmid pRL-SV40 (Promega) with 250 l of Opti-MEM (Invitrogen) and then added LipofectAMINE 2000 (7.5 l) with 250 l of Opti-MEM and incubated for 5 min at room temperature. We then added the plasmid DNA mixture to the LipofectAMINE 2000 mixture and incubated for 30 -45 min at room temperature. We rinsed Asn-104 OD1, Asn-104 ND2 a Listed are atoms within the Runt domain and CBF␤ within 3.5 Å from one another. Molecules A (Runt domain) and B (CBF␤) from Protein Data Bank entry 1H9D were used for the analysis. Contacts were determined using the CNS module of Endscript 1.0.
P19 cells with 2 ml of serum-free Dulbecco's modified Eagle's media and added another 2 ml of the same media to each well. The DNA/Lipo-fectAMINE 2000 mixture was then added, and the cells were incubated for 5 h at 37°C, and then 350 l of Dulbecco's modified Eagle's medium with 70% fetal calf serum was added to each well. We prepared cell lysates 48 h post-transfection, and we measured their luciferase activity following instructions provided in the manual accompanying the Dual-Luciferase Reporter Assay System (Promega). We used pRL-SV40, which carries the Renilla luciferase gene, to monitor the transfection efficiency for each sample, and we used P19 cells that were transfected in the absence of DNA as blanks. We calculated the reporter activity as follows: reporter activity ϭ (firefly luciferase activity Ϫ Blank)/(Renilla luciferase activity Ϫ Blank).  Table I). Eight of these amino acids in the Runt domain (Asn-69, Met-106, Tyr-113, Ser-14, Thr-149, Pro-156, Pro-157, and Thr-161) make side chain contacts to CBF␤ that involve atoms beyond the CB carbon. To determine which of these Runt domain side chains are energetically important for the interaction with CBF␤, we substituted all but the two prolines with alanine ( Fig. 1) (35). Substitution with alanine should remove interactions involving atoms beyond the CB of an amino acid side chain (47). Alanine substitution, on the other hand, will not allow us to examine the energetic contribution of main chain atoms or the CB moieties to heterodimerization. We also substituted Asp-66 and Val-159 with alanines as controls. Asp-66 and Val-159 make backbone contacts to CBF␤, and Asp-66 also makes a CB contact. Alanine substitution of these residues should not affect heterodimerization unless the Runt domain structure is perturbed. We substituted Asn-109, which makes backbone O, C, and CA contacts with CBF␤ with alanine because Asn-109 maps to a previously defined energetic hot spot at the heterodimerization interface identified in a similar analysis of CBF␤ (33).

Preparation of Alanine-substituted Runt Domain Proteins and Assessment of Their Structural Integrity-Crystal
Because it is possible that an alanine substitution will perturb the Runt domain fold, the structural integrity of the mutated Runt domains was examined by two methods. We performed 15 N-1 H heteronuclear single quantum correlation (HSQC) spectroscopy, which is a very sensitive measure of structural integrity, on all of the 15 N-labeled, alanine-substituted Runt domains. NMR spectra of the isolated Runt domain are poor due to signal broadening caused by conformational exchange (22,23,48). For this reason, and because the mutations at the heterodimerization interface did not substantially affect DNA binding (35), we recorded the 15 N-1 H HSQC spectra on Runt domain-DNA complexes containing 15 N-labeled Runt domain and unlabeled DNA, which yield superior spectra. RD(T149A) spectrum is also significantly perturbed, albeit not to the same extent as that of RD(N69A). Table II summarizes the 15 N-1 H HSQC spectroscopy results for all of the alaninesubstituted Runt domains. The spectrum of RD(M106A) was severely perturbed, comparable with that of RD(N69A), suggesting that the overall structure of the Runt domain had been compromised, at least at the temperature (40°C) at which the spectra were recorded. The 15 N-1 H HSQC spectra of RD(D66A), RD(Y113A), RD(S114A), and RD(T161A) were essentially the same as that of the wild type Runt domain, with chemical shift changes observed only for amino acids in spatial proximity to the substituted alanine. The 15 N-1 H HSQC spectra of RD(N109A), RD(T149A), and RD(V159A) were perturbed but not to the same extent as those of RD(N69A) and RD(M106A).
Although NMR spectra are very sensitive to the structural integrity of the protein, they do not provide a quantitative measure of protein stability. In addition, other factors such as protein aggregation can affect the quality of NMR spectra, and a perturbed spectrum may not reliably reflect an altered protein fold. Therefore, for those mutants that showed altered NMR spectra, we also carried out urea denaturation measure-ments as an independent and quantitative measure of the effect of alanine substitution on protein stability (Table II). Denaturation with urea was monitored by means of the change in the fluorescence of the single tryptophan residue in the Runt domain, as described previously (37). Representative data for the wild type Runt domain and three of the mutant proteins are shown (Fig. 3). We carried out all measurements at 4°C, which is approximately the same temperature at which we determined the equilibrium binding constants summarized in Tables III and IV. Table II shows the relevant thermodynamic parameters obtained from fitting the data to a two-state model. All of the alanine-substituted Runt domains that we examined were less thermodynamically stable than the wild type Runt domain, with RD(N109A) being by far the most dramatically perturbed.
Alanine Substitution of Runt Domain Asn-109 and Thr-161 Severely Impairs CBF␤ Binding-We measured the affinities of Runt domain-DNA complexes for CBF␤ by electrophoretic mobility shift assay (EMSA) (K 3 in Fig. 4A). K 3 values are presented in Table III, and a representative EMSA is shown in Fig. 4. K 3 is calculated from the fractional occupancy of the  Recorded on a Runt domain-DNA complex. Mutants with spectra that had a wild type appearance with a limited number of peaks showing significant chemical shift changes were classified as normal, whereas mutants displaying spectra with large numbers of shifted peaks or many missing peaks were characterized as perturbed.
b Equilibrium denaturation data of the Runt domain and listed point mutants at 4°C. Protein denaturation was effected by addition of urea and monitored via the fluorescence of the single Trp residue in the Runt domain. All measurements were carried out in duplicate, and the S.E. are indicated. m is the slope of the unfolding transition.
[Urea] 50% refers to the concentration of urea required to effect 50% denaturation. ⌬G D-N refers to the calculated free energy change in kcal/mol associated with denaturation of the protein. ND, not determined.
c These spectra were perturbed to a greater extent than others.
Runt domain (RD)-DNA complex by CBF␤ using the formula: (CBF␤-RD-DNA)/(RD-DNA) ϭ 1/(1 ϩ K 3 /[CBF␤]). Therefore, even if a significant proportion of the Runt domain molecules for those mutants with lower thermodynamic stability were incorrectly folded, this should not affect the K 3 values because the Runt domain concentration does not factor into the K 3 determination. We found that the N109A and T161A substitutions in the Runt domain caused the most severe perturbation of CBF␤ binding, increasing K 3 by ϳ60.7and 40.0-fold, respectively. The Y113A and S114A mutations had little or no effect on heterodimerization (Table III). The N69A mutation also did not affect binding to CBF␤, despite the fact that it perturbed the 15 N-1 H HSQC spectrum and decreased the thermodynamic stability of the Runt domain. The M106A and T149A substitutions decreased the affinity of the Runt domain-DNA complex for CBF␤ by ϳ10-fold. The controls for the analysis, RD(D66A) and RD(V159A) (Asp-66 and Val-159 contact CBF␤ via atoms that would not be replaced by an alanine substitution), had K 3 values equivalent to that of the wild type Runt domain. Dissociation constants for RD(D66A), RD(N109A), RD(Y113A), and RD(T149A) with CBF␤ in the absence of DNA (K 1 ) were reported by Tahirov et al. (25), and the relative differences compared with the wild type Runt domain are basically in agreement with the data reported here. In the companion paper (35), however, we showed that RD(D66A) had decreased activity in both a yeast one-hybrid assay designed to detect DNA binding and in a yeast two-hybrid assay that detects heterodimerization with CBF␤. We predicted based on the yeast analyses that the D66A mutation disrupts the Runt domain fold (35). However, this prediction was not borne out by the biochemical analyses here or reported elsewhere (25). We analyzed previously (33) the energetic contribution of amino acids in the CBF␤ subunit to heterodimerization. The amino acids examined in that study were chosen based on NMR chemical shift perturbation data, which detected atoms in CBF␤ that experienced a change in their local chemical environment in the presence of the Runt domain-DNA complex (49). Subsequent to that study, the structure of the Runt domain-CBF␤ complex was solved by x-ray crystallography, and five additional amino acids in CBF␤ at the heterodimerization interface that were not identified in the NMR experiments were described (Arg-33, Val-58, Asn-63, Ser-65, and Gly-105) (31). We substituted each of these amino acids in the CBF␤ subunit with alanine, purified the 15     and performed 15 N-1 H HSQC spectroscopy to assess the protein fold, as described previously (33). All alanine-substituted proteins with the exception of CBF␤(G105A) produced normal 15 N-1 H HSQC spectra (not shown). The affinities of CBF␤(R33A), CBF␤(V58A), CBF␤(N63A), and CBF␤(S65A) for a wild type Runt domain-DNA complex were determined by EMSA (Table IV). Alanine substitution of CBF␤ Val-58, Asn-63, and Ser-65 had little (Asn-63) or no (Val-58 and Ser-65) effect on heterodimerization. On the other hand the CBF␤(R33A) mutation increased K 3 by 20-fold.
Amino Acids in the Heterodimerization Interface Contribute to Runx1 Function in Vivo-We assessed the in vivo activity of full-length Runx1 proteins containing several of the alanine substitutions in a transient co-transfection assay, using the enhancer from the T cell receptor ␤ (TCR␤) chain gene to detect Runx1 activity. A minimal "core" enhancer from the TCR␤ chain contains three Runx1-binding sites, and transactivation by Runx1 in vivo is dependent on the integrity of these sites (45,50,51). The transfection assays were performed in P19 embryonic carcinoma cells, which appear to contain CBF␤ activity but do not express Runx1 (52,53). Runx1 stimulated transcription from the TCR␤ enhancer by ϳ15-fold, and as previously shown this activation was dependent on the Runx1binding sites (Fig. 5) (45). The Runx1 proteins that we tested, Runx1:N109A, Runx1:Y113A, Runx1:T149A, and Runx1: T161A, contained mutations that spanned the range of in vitro CBF␤ binding activity, from the Y113A mutation that decreased the affinity of the Runt domain for CBF␤ by a relatively modest 4.9-fold, to the T161A and N109A mutations that impaired heterodimerization 40.0-and 60.7-fold, respectively. We found that although Tyr-113 makes extensive contacts with Lys-28, Arg-33, and Asn-63 in CBF␤, expression of Runx1 (Y113A) induced activity from the reporter gene by 14.6-fold, indicating that the Tyr-113 side chain is not essential for the in vivo activity of Runx1 in this assay. The Runx1:T149A protein (the T149A mutation decreased the affinity of the Runt domain for CBF␤ by ϳ13-fold) transactivated the TCR␤ reporter gene by a significantly smaller 9.3-fold. Runx1:T161A transactivated slightly less well than Runx1:T149A (7.8-fold), consistent with its lower affinity for CBF␤. The N109A mutation, which reduced the affinity of the Runt domain for CBF␤ 60.7-fold, severely impaired the ability of Runx1 to activate transcription of the reporter gene. We were unable to detect the wild type or mutated Runx1 proteins in transiently transfected P19 cells by Western blot analysis; therefore, we cannot exclude the possibility that some of the variances in transactivation were due to differences in Runx1 protein concentration and/or stability. This is most likely to be a factor for the Runx1 proteins with the T149A and N109A mutations, as these mutations in the context of the isolated Runt domain caused perturbations in 15 N-1 H HSQC spectra recorded at 40°C, which is close to the in vivo temperature. The N109A mutation also caused the largest decrease in thermodynamic stability that we observed. The fold activation was calculated as the luciferase activity induced by transfection of the expression vector encoding the wild type or alanine-substituted Runx1 proteins relative to that induced by the empty expression vector. The results from three independent pools of transfected cells are averaged, and representative results from one of three independent experiments are shown. pTCR␤(core)-LUC was the reporter construct pTCR␤-LUC with mutations in all three CBF-binding sites (45). The difference in activity from the wild type Runx1 protein was significant (p Յ 0.01) for Runx1:N109A, Runx1:T149A, and Runx1:T161A.

DISCUSSION
We carried out a structure-based alanine-scanning mutagenesis study of the CBF␤-binding interface on the Runx1 Runt domain to identify those residues that are energetically important for heterodimerization. As has been shown for a number of systems, there are a rather limited number of residues, referred to as "hot spots," that contribute the bulk of the binding energy at protein-protein interfaces (54 -57). A hot spot is generally defined as an alanine substitution that reduces the binding energy by Ն2 kcal/mol (56). We identified one mutation in the Runt domain, T161A, that did not disrupt the overall Runt domain structure but did have a significant impact on binding to CBF␤ (2.00 kcal/mol). In addition, the N109A mutation that significantly destabilized the Runt domain fold also caused a large perturbation in CBF␤ binding (2.22 kcal/mol). A proportional effect on in vivo function was also demonstrated for the T161A and N109A mutations in the context of full-length Runx1.
Asn-109 is located at the C-terminal end of strand ␤C in the Runt domain and makes van der Waals contacts to CBF␤ residue Gly-61, a residue that we demonstrated previously (33) is also very sensitive to alanine substitution. Although the N109A substitution should not affect the contact to CBF␤ per se, the perturbations seen in the NMR and urea denaturation data indicate that the N109A mutation causes an alteration in the Runt domain structure that contributes to the observed change in affinity. In the ternary complex the Runt domain Asn-109 ND2 forms a hydrogen bond to the main chain O of Ser-140 in the ␤EЈ-F loop, whereas OD1 interacts with the backbone nitrogens of Glu-111 and Asn-112 in the ␤C-D loop (Fig. 6C). Thus removal of the Asn-109 side chain may destabilize both loops, which could affect not only the Asn-109 O contact to the CA of CBF␤ Gly-61, but also other Runt domain residues that contact CBF␤ such as Tyr-113 and Ser-114. Consistent with this hypothesis, the dissociation constant (K 3 ) of the wild type Runt domain for CBF␤(G61A) is 3.5 ϫ 10 Ϫ7 M, whereas the K 3 for the RD(N109A) with CBF␤(G61A) is 13-fold higher (4.6 ϫ 10 Ϫ6 M (not shown)). The non-equivalence of these two K 3 values indicates that the RD(N109A) mutation is more deleterious than the CBF␤(G61A) mutation, and therefore by affecting the Runt domain conformation the N109A mutation disrupts more than the contact to CBF␤ Gly-61. The CBF␤(G61A) mutation did not affect the CBF␤ fold (33). We predicted that substitution of CBF␤ Gly-61 with alanine would move the main chain oxygen of Asn-109 in the Runt domain away from the heterodimer interface, thereby destabilizing the interaction.
Mutations in a hot spot residue at the analogous position of Runt domain Asn-109 in the p53 tumor suppressor protein (Arg-175), which makes similar bridging interactions between the same two loops, also perturb the protein fold (39, 58). As Runt domain loops ␤C-D and ␤EЈ-F and the analogous L2 and L3 loops in p53 have little regular secondary structure, the lack of extensive hydrogen bonding interactions may be compensated by side chain to side chain and side chain to backbone interactions between the loops. By contrast with the Runt domain-CBF␤ complex, a bound zinc atom in addition to salt bridge interactions links the analogous loops in p53.
Thr-161, the second energetic hot spot residue, is located in the ␤G strand in the Runt domain. Comparison of binary and ternary complex structures with those of the free Runt domain reveals Thr-161 makes main chain and side chain hydrogen bonding interactions to CBF␤ residue Asn-104 (Fig. 6B). CBF␤ Asn-104 was identified as making one of the most critical contacts for heterodimerization, and Thr-161 also appears to play the same role in the Runt domain. Thr-161 is in a region of the Runt domain (amino acids 159 -166) that was shown by NMR to undergo a significant conformational change upon CBF␤ binding (28). In addition, recent crystal structures of the isolated Runt domain also showed this region of the protein undergoes a substantial conformational change upon CBF␤ binding (30,32).
We also examined the effects of four additional alanine substitutions in CBF␤ identified in later structural studies: Arg-33, Val-58, Asn-63, and Ser-65. Only R33A showed a significant effect, decreasing the binding affinity for the Runt domain-DNA complex by 20-fold. In the Runx1 Runt domain-CBF␤ complex, the Arg-33 guanidino group interacts with the aromatic ring of Tyr-113 via a hydrogen bond interaction with the hydroxyl group and perhaps also a cation-type of interaction that has been shown to be quite energetically favorable (59). Alanine substitution of Tyr-113 in the Runt domain resulted in a more modest (4.9-fold) increase in K 3 (Table III). The reason for the larger increase in K 3 seen with the Runt domain R33A mutant is not clear.
If hot spots are, as has been suggested, good binding sites in a general sense, then identifying them may prove useful for targeted drug design (57,60). Virtual screening of compounds against the hot spot residues at the Runt domain-CBF␤ interface may assist in the identification of small molecules or peptides that will interfere with heterodimerization. Coupling inhibitors of heterodimerization with those targeted to protein interfaces specific to the oncogenic Runx1 or CBF␤ fusion proteins may result in compounds with potent activities against the oncogenic proteins. For example, the inv(16)(p13;q22) associated with acute myelogenous leukemia fuses the N-terminal 165 amino acids of CBF␤ to the coiled-coil tail region of a smooth muscle myosin heavy chain (61). The resulting CBF␤-SMMHC protein dominantly inhibits Runx1-CBF␤ function (62)(63)(64)(65)(66). CBF␤-SMMHC binds to the Runx1 Runt domain with an affinity Ն10-fold higher than does CBF␤, and this high affinity binding may contribute to the potent dominant negative activity of CBF␤-SMMHC (67). A small molecule targeted to a hot spot at the Runt domain-CBF␤ interface covalently attached to another molecule that binds to the high affinity Runx1-binding site in CBF␤-SMMHC may prove useful in counteracting the dominant negative activity of this fusion protein.