Site-directed Mutations in the vnd/NK-2 Homeodomain

Secondary structures, DNA binding properties, and thermal denaturation behavior of six site-directed mutant homeodomains encoded by the vnd/NK-2 gene from Drosophila melanogaster are described. Three single site H52R, Y54M, and T56W mutations, two double site H52R/T56W and Y54M/T56W mutations, and one triple site H52R/Y54M/T56W mutation were investigated. These positions were chosen based on their variability across homeodomains displaying differences in secondary structure and DNA binding specificity. Multidimensional NMR, electrophoretic mobility shift assays, and circular dichroism spectropolarimetry studies were carried out on recombinant 80-amino acid residue proteins containing the homeodomain. Position 56, but more importantly position 56 in combination with position 52, plays an important role in determining the length of the recognition helix. The H52R mutation alone does not affect the length of this helix but does increase the thermal stability. Introduction of site mutations at positions 52 and 56 in vnd/NK-2 does not modify their high affinity binding to the 18-base pair DNA fragment containing the vnd/NK-2 consensus binding sequence, CAAGTG. Site mutations involving position 54 (Y54M, Y54M/T56W, and H52R/Y54M/T56W) all show a decrease of 1 order of magnitude in their binding affinity. The roles in structure and sequence specificity of individual atom-atom interactions are described.

Homeodomain-containing proteins act as regulators of transcription by binding to various target DNA sequences and they specify positional information and segmental identity in the commitment of embryonic cells to specific developmental pathways (1)(2)(3)(4). The detailed understanding of the molecular mechanisms by which the homeodomain regulates development is of considerable current interest (5)(6)(7)(8)(9)(10)(11). The ability of such a protein to target its DNA binding site and regulate gene expression is strongly dependent upon the specific amino acid residues in the homeodomain. Furthermore, there is a growing body of data relating mutations in the homeodomain to a wide variety of developmental abnormalities (12)(13)(14)(15)(16). Some anomalies in development have been shown to result from single site amino acid residue replacements (17)(18)(19). A primary function of the homeodomain is to bind to a specific site or set of similar sites in DNA (20 -23). Knowledge of the role of individual amino acid residues in tertiary structure formation and sequence specific DNA binding thus can provide important clues toward the understanding of the molecular mechanisms involved in the behavior of the homeodomain as a transcription regulator.
Members of the homeodomain family show a high degree of homology in their tertiary structures (2). This high degree of structural homology arises from similarities such as hydrophobicity or charge state maintained by amino acid residues at unique positions in the homeodomain that make specific interresidual contacts, thereby yielding a stable low energy structure (24). The homeodomain contains three helical segments, with helix II and helix III forming the helix-turn-helix DNA binding motif (25). Helix III, the recognition helix, binds in the major groove of the DNA, and the partially flexible N-terminal arm orients in the adjacent minor groove (26 -32). The principal contacts with the bases of the DNA involve amino acid residues usually in positions 2 or 3 and 5 of the N-terminal arm as well as residues at positions 46,47,50,51, and 54 of the recognition helix. These protein-DNA contacts represent the major interactions responsible for nucleotide sequence-specific binding. Of these amino acid residues that contact bases of the DNA, only residues at position 54 are found to vary significantly among the homeodomains (2,33). Questions concerning the role of sequence specificity of the homeodomain-DNA interaction in the regulation of gene expression naturally arise (34 -36). Is there sufficient variability in the homeodomain to explain functional specificities in development? How do fine distinctions in properties of the homeodomain-DNA complex such as small variations in binding affinity or subtle differences in the three-dimensional structure correlate with target specificity in transcription regulation?
The homeodomain of particular interest in this study comes from the vnd (ventral nervous system-defective)/NK-2 gene (14,37) of Drosophila melanogaster and is the parent member of the vnd/NK-2 class of homeodomains first described by Kim and Nirenberg (38,39). The vnd/NK-2 gene is predominantly a neural gene regulator and gives rise to part of the central nervous system of the embryo (40,41). The three-dimensional structure of the vnd/NK-2 homeodomain has been reported previously (42,43). The full structural integrity of helix III of vnd/NK-2 is conserved only through residue 52 (i.e. 11 amino acid residues in length), in contrast to that for Antennapedia (44) where helix III is maintained through residue 60 (i.e. 19 amino acid residues in length). The structural behavior of vnd/NK-2 residues in the C-terminal region beyond residue 53 is difficult to describe due to increased conformational flexibility (45). This variability in the secondary structure of the recognition helix has been suggested to be largely due to the nature of the amino acid residue in position 56 (45).
The vnd/NK-2 homeodomain recognizes the unusual DNA consensus sequence 5Ј-CAAGTG-3Ј. 1 The canonical DNA consensus sequence recognized by many homeodomains contains 5Ј-TAATGG-3Ј in its core (2). Both helix III (the recognition helix) and the N-terminal arm make direct contact with the DNA. It was shown recently that the interaction in the recognition helix of tyrosine at position 54 (Tyr 54 ) with the DNA is the major determinant of this uncommon nucleotide binding specificity (33). In addition to residues directly involved in base-specific interactions, amino acid side chains in positions that do not contact any DNA base nonetheless are believed to influence the interaction of the homeodomain with DNA and thus may be important functionally. For example, nonspecific electrostatic interactions between positively charged amino acid side chains and the negatively charged phosphate backbone influence binding affinity (2). The recognition helix contains many conserved amino acid residues including those involved in basespecific contacts as well as ones with positively charged side chains, some of which make nonspecific electrostatic contacts with the DNA. Thus, any variability in this region of the homeodomain must be of particular significance. For example, residue 56, which is thought to be important in determining the length of the recognition helix in the unbound state, is highly variable, with threonine at this position in the vnd/NK-2 homeodomain. Questions concerning a possible role of residue 56 in DNA binding specificity, although it does not contact the DNA, have been raised (45). The amino acid residue found in position 52, which also is in the recognition helix and not in direct contact with the DNA, is usually arginine. In rare instances, such as in vnd/NK-2, histidine is found in this position. Recently, single site mutations in rough and eve, where arginine is replaced by histidine at position 52, were found to produce temperature-dependent developmental abnormalities in the resultant D. melanogaster embryos (12,17).
Structural and chemical information regarding the role of a particular residue can be obtained, in part, by mutational analysis, where given residues of a homeodomain of interest are replaced systematically. In this paper, we describe effects on structural, thermal denaturation and DNA recognition properties of the homeodomain when positions 52, 54, and 56 are mutated simultaneously, in pairs, or singly (46). These positions, within the recognition helix, were chosen based upon their variability across the homeodomains. The wild type homeodomain is vnd/NK-2, and the positional mutations are H52R, Y54M, and T56W and are made to correspond to the amino acid residues found in the Antennapedia (Antp) homeodomain. The choice of mutations of these residues in vnd/NK-2 to ones found in Antp is based upon the differences observed in DNA specificities, thermal stabilities (Table I), and lengths of the respective recognition helices of Antp and homeodomains in the vnd/NK-2 class.

EXPERIMENTAL PROCEDURES
Protein Expression-Restriction sites for NdeI and BamHI sites were engineered, via polymerase chain reaction, into a DNA fragment containing the sequence encoding the vnd/NK-2 homeodomain. This NdeI-BamHI fragment served as a template for introducing mutations by polymerase chain reaction using overlapping oligonucleotides that contained the appropriate nucleotide substitutions (47). Wild type and six of the seven possible mutant vnd/NK-2 DNA fragments were cloned into the bacterial expression plasmid pET-15b (Novagen) and sequenced to verify the mutations and check for polymerase chain reaction errors. Screening for a colony containing a DNA insert that would encode the H52R/Y54M mutation in the homeodomain was unsuccessful. The resulting constructs, transformed into Escherichia coli BL21 (DE3) plys S, produced fusion proteins containing six histidine residues, the thrombin cleavage site, and the sequence encoding the vnd/NK-2 homeodomain. The proteins of interest are 80-amino acid residue fragments that encompasses the 60-residue homeodomain. Because of our choice here of the pET-15b plasmid, this 80-amino acid residue protein differs slightly at the N-terminal end over the 77-residue protein used in our previous studies (33,42,43). 1 Fusion proteins were purified by nickel affinity chromatography. Biotinylated thrombin was used to cleave the homeodomain from the fusion protein and then removed with streptavidin immobilized onto agarose. Polyhistidine fragments and uncleaved fusion proteins then were removed by passing the mixture over a nickel affinity column. 15 N-Labeled H52R/T56W mutant protein was purified from bacteria grown in 15 N-enriched Bioexpress media (Cambridge Isotope Laboratories) as described above. All protein preparations were determined to be homogeneous by SDS-polyacrylamide gel electrophoresis, and their molecular weights and purities were verified by electrospray mass spectrometry.
Electrophoretic Mobility Shift Assays--Complementary oligonucleotides that contain the vnd/NK-2 DNA consensus binding site sequence were chemically synthesized. Both plus (5Ј-TGTGTCAAGTGGCTG-TAG-3Ј) and minus (5Ј-CTACAGCCACTTGACACA-3Ј) strands were radioactively labeled with [␥-32 P]ATP using T4 polynucleotide kinase. An equimolar mixture of the two strands was heated to 90°C for 5 min, and then the strands were allowed to anneal by slowly cooling the solution to 4°C. A double-stranded DNA fragment, BS2-18, containing the Antennapedia consensus binding site also was prepared from synthetic oligonucleotides (plus strand, 5Ј-GAGAAAAAGCCATTAGAG-3Ј; 1 L.-H. Wang and M. Nirenberg, manuscript in preparation.

TABLE I Comparison of vnd/NK-2 and Antennapedia homeodomains
The sequences for residues 42-60, including the recognition helix, are shown for vnd/NK-2, other representative members of the vnd/NK-2 class of homeodomains (39), and Antp. The residues in the vnd/NK-2 homeodomain that make contact with DNA bases in the major groove are in boldface type. Positions that were mutated in the vnd/NK-2 homeodomain to the corresponding residues in the Antp homeodomain are underlined. The length of the recognition helix for vnd/NK-2 and Antp is represented by bars above the sequence. Consensus DNA binding sequences and thermal denaturation temperatures are also compared. Tyrosine at position 54 is unique to the vnd/NK-2 class of homeodomains. Tyrosine 54 contacts the cytosine, which is base-paired to the guanidine at position 4 of the vnd/NK-2 DNA target sequence (33). All other residues in the recognition helix that make contact with DNA (shown in boldface type) are also completely conserved in the vnd/NK-2 class of homeodomains. minus strand, 5Ј-CTCTAATGGCTTTTTCTC-3Ј) using the method described above. Protein-DNA complexes were formed on ice for 30 min in 10-l reaction mixtures containing 20 mM HEPES (pH 7.9), 90 mM NaCl, 0.4 mM EDTA, 0.3 mg/ml bovine serum albumin, 10% glycerol, 10 pM labeled double-stranded DNA, and varying amounts of the homeodomain proteins. Reaction mixtures were loaded onto a 10% native polyacrylamide gel and subjected to electrophoresis at 300 V in 0.5ϫ TBE for 40 min at 4°C. The gels were dried, and free and protein-bound DNA bands were visualized by autoradiography. Apparent K d values were obtained by quantitating the intensities of the bands and plotting the percentage of the total DNA that is bound to protein against the concentration of protein.
Circular Dichroism-Ellipticities in the far uv region (180/190 -250 nm) were measured on a Jasco J600 spectropolarimeter using a 0.2-mm jacketed cuvette attached to a recirculating bath. Wild type and mutant protein samples (50 M) were dissolved in 20 mM sodium acetate (pH 4.5). Spectra were recorded at discrete temperature intervals, allowing a 20-min equilibration time, and the data were averaged over eight scans. Values for mean residue ellipticity were calculated for all proteins at 222 nm and plotted against temperature.
NMR Spectroscopy-All NMR spectra were obtained on a Bruker Instruments AMX600 spectrometer. Samples were dissolved in 90% H 2 O, 10% D 2 O to final concentrations of 4 mM for the T56W and H52R/T56W mutants and 5 mM for the H52R mutant. The pH values were adjusted to 4.5 by the addition of small amounts of HCl. Homonuclear two-dimensional total correlation spectroscopy and NOESY 2 experiments were carried out with mixing times of 100 ms. Mild presaturation was used for solvent suppression. For comparison purposes, spectra for the individual mutant proteins were collected at temperatures that corresponded to the analogous position in the order-disorder transition characteristic of each mutant as measured by circular dichroism spectropolarimetry (Fig. 2). The temperatures are, respectively, 12°C for the T56W single mutant, 19°C for the H52R single mutant, and 25°C for the H52R/T56W double mutant vnd/NK-2 homeodomains. For all three proteins, the upfield resonances in the range Ϫ1.4 to 0.6 ppm assigned to Leu 16 , Leu 26 , and Leu 40 that are indicative of the formation of the hydrophobic core of the globally folded wild type vnd/NK-2 are present at these temperatures for the respective mutants. This comparison thus verifies the use of the temperature dependence of the 222-nm band in the CD spectra as an approximate and preliminary assessment of degree of folding. Spectra run at each temperature were calibrated using known values of the proton resonance of H 2 O. As an internal calibration check, the position of the ␣-CH 3 resonance of methionine engineered at position Ϫ8 was verified to be at 2.13 ppm in all spectra.
For the proton-deuterium exchange experiment, 15 N-H52R/T56W vnd/NK-2 homeodomain was first dissolved in water, the pH was adjusted to 4.5, and then the concentration of the homeodomain was brought to 1.5 mM in 90% H 2 O, 10% D 2 O. A two-dimensional hetero-nuclear multiple quantum correlation spectrum was recorded at 12°C. The sample was freeze-dried and then dissolved in D 2 O. Two-dimensional heteronuclear multiple quantum correlation spectra were recorded at 11 min, 58 min, 105 min, and 26 h after dissolving the homeodomain in D 2 O.
For studies of the 15 N-Y54M-DNA complex, a 16-mer doublestranded DNA (plus strand, 5Ј-TGTGTCAAGTGGCTGT-3Ј) was prepared as described previously (42). The 15 N-Y54M-DNA complex was formed by titrating a 0.3 mM solution of 15 N-Y54M into a 0.3 mM solution of the DNA until a 1:0.95 molar ratio of DNA to protein was obtained. This solution was concentrated to give a final concentration of 1.5 mM of the complex, and the pH was adjusted to 6.0.

RESULTS
Homeodomain-DNA Binding Affinity-The concentration dependence of the affinity of the purified wild type and six mutant vnd/NK-2 homeodomain proteins for the 18-base pair DNA duplex (containing 5Ј-CAAGTG-3Ј as a central segment of the DNA) electrophoresed on a native polyacrylamide gel is shown in Fig. 1A. This 18-base pair DNA fragment contains the same 16-base pair segment as the one used in our previous structural studies on the vnd/NK-2-DNA complex (33). An AG duplex pair of bases is added to the 3Ј-end of the 16-base pair DNA fragment for optimal binding. The mobility shift data demonstrate high affinity binding of the wild type vnd/NK-2 homeodomain to DNA in the nanomolar range in accord with previous results (42).
In a previous study of the three-dimensional structure of the vnd/NK-2 homeodomain-DNA complex, it was shown that Tyr 54 is the only highly variable amino acid residue in the recognition helix found to make a base-specific contact with the DNA (33). To assess the role of the interaction of Tyr 54 with the DNA, the binding of the homeodomain with mutations involving position 54 was examined. The single Y54M mutant, the double Y54M/T56W mutant, and the triple H52R/Y54M/T56W mutant of vnd/NK-2 all show, within experimental error, a decrease of roughly 1 order of magnitude from 0.5-1.5 nM to about 7-14 nM (Fig. 1A) in the binding affinity for the consensus DNA sequence. These ranges of binding affinities are the result of a minimum of four experiments carried out for each protein.
To examine the roles of positions 52 and 56, DNA binding affinities for mutations involving these positions were determined. vnd/NK-2 is unusual in that it contains histidine in position 52. The amino acid residue in position 56 (threonine for vnd/NK-2) in the homeodomain has been implicated in determining the length of the recognition helix (42,45). Muta- tions involving positions 52 and 56, namely the two single mutants H52R and T56W as well as the double mutant H52R/ T56W, show no significant alteration of their binding affinity for the consensus DNA sequence from the nanomolar binding affinity found for the wild type vnd/NK-2 homeodomain under the mobility shift assay conditions (Fig. 1A). At least three repeated assays were carried out for each of these latter binding constant determinations.
For comparison, representative mobility shift data depicting the binding of the wild type and mutant vnd/NK-2 homeodomains to an 18-base pair fragment of the recognition DNA site for Antp, BS2-18, are presented in Fig. 1B. The triple H52R/ Y54M/T56W mutation results in a homeodomain where the segment of the recognition helix from residues 46 -57, which represents the part of the protein that resides in the major groove of the DNA, is identical to that of Antp. From the data in Fig. 1B, it is seen that the wild type and mutant vnd/NK-2 homeodomain proteins including the triple mutant all bind the BS2-18 with similar 4 -14 nM affinities, where this range in affinity arises from scatter in measurements taken from three and seven experiments for each individual protein studied.
Circular Dichroism-To evaluate the secondary structural behavior of the wild type and six mutant vnd/NK-2 homeodomains, CD spectra were obtained in the 250 -190-nm range. The mean residue ellipticities as a function of temperature at 222 nm are presented in Fig. 2. All of these proteins undergo reversible thermal denaturation within the temperature range of temperatures from Ϫ5 to about 50°C, as evidenced by changes in helix content determined from the 222-nm band. The wild type homeodomain is the most thermally labile homeodomain characterized thus far (26,42,45). In all examples studied, H52R site mutations result in increased thermal stability against denaturation as well as an increase in the cooperativity of the reversible denaturation transition. Furthermore, the H52R mutant vnd/NK-2 homeodomain shows an increase in the magnitude of the mean residue ellipticity in the folded state over that found for wild type, which presumably implies an increase in the conformational stability of the helical residues in the homeodomain. The denaturation temperatures at pH 4.5 change from approximately 25°C for wild type to around 34°C for the H52R/T56W and H52R/Y54M/T56W mutants. Introduction of the mutation involving position 56, i.e. T56W, only results in an increase in the cooperativity of the renaturation-denaturation transition as well as in the mean residue ellipticity and thermal stability, provided that the H52R mutation is present. Careful examination of the CD spectra did not show any significant alteration of the CD behavior associated with any of the Y54M site mutations (Fig. 2).
Nuclear Magnetic Resonance Spectroscopy-One-dimensional and two-dimensional NOESY and total correlation spectroscopy NMR spectra of the 80-amino acid residue proteins that encompass the respective wild type, the H52R (19°C) and T56W (12°C) single mutants, and the H52R/T56W (25°C) double mutant vnd/NK-2 homeodomains were analyzed. NMR studies were limited to these mutant proteins, where significant changes in the CD spectra relative to the wild type vnd/ NK-2 were observed. The sequence-specific assignments of all of the backbone and many of the side chain proton resonances were obtained using the well accepted strategy for proteins in this molecular weight range (26,48). The assignment procedure was greatly facilitated by a comparison of the tabulated resonance frequencies for wild type vnd/NK-2 reported previously (43). It is interesting to note that tracing through the two-dimensional NOESY spectrum of the H52R/T56W double mutant was significantly easier than for the wild type or either of the single mutants. The spectra in this latter case show more intense and better resolved cross-peaks, presumably in part due to the higher temperature and in part due to increased conformational stability. The chemical shift differences between the backbone amide protons of wild type vnd/NK-2 and the H52R, T56W, and H52R/T56W mutants are summarized graphically in Fig. 3. Shift differences observed for the amide residues of positions 58 -60 are most likely due to an increased stabilization of helical structure in this C-terminal segment of the recognition helix.
After making many of the resonance assignments for the three mutant homeodomains, sequential NOEs involving the amide protons were analyzed to characterize secondary structure in the recognition helix region between residue Pro 42 and Glu 66 . A summary of these short range NOE connectivities is shown in Fig. 4 each of the three mutants between helix II and helix III that are characteristic of the helix-turn-helix motif. The magnitudes of the d NN(i, i ϩ 1) , d _N(i, i ϩ 1) , d _N(i, i ϩ 3) , and d _N(i, i ϩ4) serve to characterize the behavior of the C-terminal region of helix III in each of the three mutant vnd/NK-2 homeodomains. The behavior of these NOE contacts is summarized in the secondary structure diagrams shown in Fig. 4. For the T56W mutant, the relative intensities of the various cross-peaks suggest that helix III is maintained through position 56 with residues 57-60 being less well ordered. In contrast, wild type vnd/NK-2 and the H52R mutant are well ordered only through residue 52 and 53, respectively. In the case of the H52R/T56W double mutant, sequential NOE contacts indicate that helix III is maintained through position 60. We see evidence for NOE contacts from the aromatic ring of Trp 56 to the side chain methylene protons of Arg 24 for the H52R/T56W double mutant, which are neither present in the wild type nor in either of the single mutant vnd/NK-2 homeodomains. However, due to degeneracies associated with Arg 24 , Arg 53 , and Lys 57 , this evidence remains ambiguous.
Amide proton exchange rates for the 15 N-H52R/T56W double mutant protein were investigated and compared with the rates of the wild type protein previously reported (42). Both experiments were carried out at 12°C at pH 4.5 and at protein concentrations of 1.5 mM (Fig. 5). The patterns of slow exchanging protons were very similar in both proteins, although the double mutant protein displayed greater conformational stability. After 26 h, all amide protons of the wild type protein were exchanged for deuterium, whereas amide protons of 20 residues, from the three helices as well as Leu 40 in the turn, remained in the H52R/T56W mutant. Another significant difference between the two proteins is the presence of slow exchanging amide protons at positions 51-53 in the double mutant but absent in the wild type homeodomain, thereby supporting our evidence for the longer recognition helix in the double mutant.
To investigate the structural basis of the observed loss of binding affinity in the Y54M mutant for its target DNA, we studied a complex of 15 N-labeled Y54M mutant with the 16base DNA fragment containing the vnd/NK-2 target sequence. 15 N-edited three-dimensional NOESY-heteronuclear multiple quantum correlation spectra of the complex revealed several cross-peaks between the side chains of Arg 5 , Gln 50 , and Asn 51 with the DNA, analogous to those seen for the wild type protein-DNA complex (33), although subtle differences in intensities and chemical shifts were found for Gln 50 (data not shown). These observations suggest that the mutant is binding to the DNA in the same manner as the wild type and that the only structural differences are localized around position 54. Thus, only the modified DNA interactions involving Met 54 , and perhaps Gln 50 , are responsible for the 10-fold loss of binding affinity. A complete spectral analysis and determination of the solution structure of the Y54M mutant vnd/NK-2-DNA complex is in progress in our laboratory. DISCUSSION The effects of mutations in positions 52, 54, and 56 of the vnd/NK-2 on secondary structure, the ability to bind the vnd/ NK-2 consensus and the BS2-18 (Antennapedia) DNA sequences, and the thermal stability of the homeodomain have been described. These three positions constitute the most variable residues in the region of the recognition helix that interacts with the DNA and thus are believed to be important in terms of both DNA binding specificity and function. We have identified specific interactions that are modified by site-directed mutagenesis. Residues in positions 52 and 56 influence primarily the thermal stability of the homeodomain against denaturation and the structure of the homeodomain in the free state, respectively. Residues in position 54 influence DNA binding specificity.
The length of the recognition helix and its elongation upon binding to DNA might be important in determining the strength of binding. In the case of Antennapedia, helix III has a well ordered helical N-terminal region (residues 42-52) and a more loosely ordered, but nonetheless defined helical C-terminal region (residues 53-59) (44). In vnd/NK-2 and fushi tarazu (45) no clear helical structure is observed beyond position 52 in the free form of the homeodomains. Since Antennapedia and fushi tarazu are highly homologous and differ by only one residue in helix III, at position 56, Qian et al. (45) proposed that the tryptophan at this position in Antennapedia is important for maintaining helical stability of the C-terminal region of helix III, due, perhaps, to the interaction of Trp 56 with the side chain of Arg 24 . In this study, we show that the residue at position 56 is indeed a determinant of helix stability in the C-terminal region of helix III. However, mutations at both positions 52 (histidine to arginine) and 56 (threonine to tryptophan) were required to increase the stability of helix III out to position 60, implicating arginine at position 52 also as a requirement for helix stability in the C-terminal region of the homeodomain. It is curious to note that no increase in binding affinity is found for the H52R/T56W mutant for DNA, although the recognition helix is elongated. One would expect the binding affinity of the wild type homeodomain to be lower than the double mutant because of a larger negative entropic contribution to the binding free energy associated with helix elongation. This observed anomaly is analogous to that first described by Qian et al. (45) in comparing binding affinities of fushi tarazu and Antennapedia.
The role of the H52R mutation in the thermal stability of the vnd/NK-2 homeodomain was examined in three site-directed mutants. Replacement of histidine by arginine in each of these three mutants results in a homeodomain that is more thermally stable against reversible denaturation, with a higher mean residue ellipticity at 222 nm. As first noted in the case of the eve homeodomain (31), the positively charged side chain of Arg 52 forms a salt bridge with the side chain of Glu 17 . Such a salt bridge involving position 17 cannot form when histidine is present in position 52 due to the shorter length of the side chain relative to that of arginine. Also, the charge state of histidine under physiological conditions would not favor formation of a salt bridge. In addition to increasing the thermal stability, this salt bridge would help constrain arginine in position 52, namely the backbone carbonyl of arginine, in a manner that would stabilize the longer recognition helix present in the H52R/T56W double mutant. Histidine in position 52 is rare, with arginine being the amino acid residue found most often in this position.
The tertiary structures of the various homeodomains determined thus far by x-ray or by NMR are homologous. Thus, the role of an individual amino acid residue in a specific position of the homeodomain is homologous in terms of structure and sequence-specific DNA recognition. Although information in the literature on functional abnormalities associated with single site amino acid residue replacements is limited, there are data available on position 52. Mutations in position 52 of eve (17) and rough (12), including single site mutations where arginine is replaced by histidine (i.e. the reverse of the site mutation described in this study), show temperature-dependent abnormal developmental phenomena. In mutants of both eve (17) and rough (12) normal development is observed at low temperatures (16 and 18°C, respectively), whereas abnormal behavior is observed if the embryos are allowed to develop at higher temperatures (30 and 25°C, respectively). It is quite tempting to conclude that these temperature-dependent functional abnormalities in development are caused by the slight thermal destabilization of the homeodomain introduced by the arginine to histidine mutation. Incorporation of the H52R mutation in Drosophila transgenic experiments now becomes significant.
We have investigated the structures, thermal denaturation and its temperature dependence, and DNA binding behavior of the various vnd/NK-2 mutations.
We showed, from structural studies on a 1:1 complex of the homeodomain with DNA, that the interaction of Tyr 54 with the DNA is the major determinant of the uncommon nucleotide sequence containing 5Ј-CAAGTG-3Ј (33). The side chain of Tyr 54 contacts the cytosine paired with the guanine in this consensus DNA sequence. The lowering of the binding affinity in site mutations containing Y54M by an order of magnitude thus is in qualitative accord with our earlier structural results. An analogous observation was made previously (36) using thyroid transcription factor 1 (Nkx2.1, previously termed TTF-1), although binding affinities (apparent binding constants obtained by studying the affinity over a range of homeodomain/ DNA concentration ratios) were not reported. The question of the conformation of the methionine side chain and its corresponding interactions with DNA becomes significant. Preliminary results from experiments on a complex between the Y54M mutant and DNA containing the vnd/NK-2 consensus sequence FIG. 5. Proton-deuterium exchange experiments. The rates of exchange of amide protons were determined for wild type protein (open circles) and the H52R/ T56W mutant (closed circles). Amide protons with no circles represent lifetimes of less than 11 min. One circle represents a proton with a lifetime of between 11 and 58 min, two circles represent a lifetime of between 58 and 105 min, three circles represent a lifetime of between 105 min and 26 h, and four circles represent a lifetime of greater than 26 h. show that the longer methionine side chain contacts the bases paired with TG at the 3Ј-end of the consensus sequence in a manner analogous to that observed for Antp (26,27) with no other significant structural perturbations. Interestingly, the binding affinity of the Y54M mutants to the Antp operator DNA BS2-18 that contains 5Ј-TAATGG-3Ј was determined and found to be similar to the vnd/NK-2 consensus DNA sequence. The affinity for these mutant proteins is comparable with the affinity of Antp to the same DNA sequence. One possibility is that the orientations of the Y54M mutant vnd/NK-2 and Antp homeodomains in the major groove of the BS2-18 DNA are analogous and that very little specificity is provided by the interaction of the N-terminal arm in the minor groove. In summary, our mutational results link subtle changes in structure (i.e. H52R) with developmental abnormalities and show that, to the extent that DNA target specificity of the homeodomain is a critical feature of transcriptional control in development, discrimination across homeodomains must be attributed to quite modest variations in binding affinity. Structural studies to provide unambiguous answers to questions raised here are in progress, and transgenic investigations to determine functional consequences of the Y54M mutation have been initiated.