Threading Analysis of Prospero-type Homeodomains

Sharmila Banerjee-Basu, David Landsman¹ and Andreas D. Baxevanis^*

Edited by J. Fickett; received October 21, 1998; revised February 8, 1999; accepted February 9, 1999

ABSTRACT

The homeodomain is a common structural motif found in many transcription factors involved in cell fate determination during development. We have used threading analysis techniques to predict whether the atypical homeodomain of prospero (pros) family members could form the three-helical homeodomain structural motif, even though these proteins are not statistically similar to canonical homeodomains as assessed by BLAST searches. Amino acid sequences of these divergent homeodomain proteins were threaded through the X-ray coordinates of the Drosophila engrailed homeodomain protein [Kissinger et al., 1990]. The analysis confirms that the prospero class of homeodomain proteins is indeed capable of forming the homeodomain structure despite its low degree of sequence identity to the canonical homeodomain. Energy calculations indicate that the homeodomain structure is stabilized primarily by hydrophobic interactions between residues at the helical interfaces. Although the atypical prospero-type homeodomain shows very little sequence similarity when compared to other homeodomain proteins, the critical amino acids responsible for maintaining the three-dimensional structure are highly conserved. A number of other homeodomain proteins, such as PHO2p from Saccharomyces and Pax6 from human, were also included in the threading analysis and were shown to be able to form the engrailed structure, indicating that there are no rigid overall sequence requirements for the formation of the homeodomain structural motif. Based on the threading experiments and the subsequent structural alignment, a new amino acid signature that unambiguously identifies the prospero-type proteins was deduced.

Key words: homeodomain; prospero; protein threading; structure prediction

INTRODUCTION

The homeodomain is a highly conserved DNA-binding domain of approximately 60 amino acid residues that is found in many eukaryotic transcriptional regulatory proteins [ Laughon, 1991; Gehring et al., 1994b]. Homeodomain proteins play a fundamental role in diverse developmental processes including the specification of body plan, pattern formation, and the determination of cell fate [Gehring et al., 1994a]. The regulatory function of a homeodomain protein is based upon its specific interaction with the transcriptional control region of a target gene. The DNA-binding mode of homeodomains has been extensively studied through both structural and site-directed mutagenesis experiments [Qian et al., 1989; Kissinger et al., 1990; Liu et al., 1990; Wolberger et al., 1991; Ceska et al., 1993; Dekker et al., 1993; Endo et al., 1994; Gruschus et al., 1997; Qian et al., 1994]. X-ray crystallographic and NMR spectroscopic studies on several members of this family revealed that they contain three helical regions folded into a compact globular structure, with an N-terminal extension. Helices I and II lie parallel to each other and across from the third helix, also known as the recognition helix. The entire structure is held together by a number of hydrophobic residues buried inside the protein, between the three helices. Amino acid conservation is highest within the recognition helix. It has been shown that the third helix lies in the major groove of DNA, where it makes specific contacts with the DNA bases [ Kissinger et al., 1990; Wolberger et al., 1991; Gruschus et al., 1997]. The N-terminal arm also makes additional specific contacts to DNA bases in the adjacent minor groove. In addition, the conserved loop between the helices I and II establishes specific contacts with the phosphate backbone. The third helix, in conjunction with the N-terminal arm, confers the DNA-binding specificity of individual homeodomain proteins. The homeodomain has been identified in a broad spectrum of organisms, from yeast to Drosophila to humans [Bürglin, 1994].

Based on sequence comparisons, the homeodomain family has been sub-classified into smaller groups [Bürglin, 1994]. Within each group, homeodomain proteins from a wide range of species show very high sequence conservation (80-100%) as well as conserved genetic function. Although the absolute sequence similarity varies among different groups, two positions in helix I (Leu16 and Phe20) and five positions in helix III (Trp48, Phe49, Asn51, Arg54, and Lys/Arg55) are almost always conserved. A class of atypical homeodomain proteins have been identified whose primary sequence diverges considerably from the homeodomain consensus, having insertions or deletions within the 60-residue homeodomain motif. X-ray crystallographic studies of the atypical MAT2 homeodomain revealed that it essentially forms the canonical homeodomain structural motif but has three extra residues in the loop between the first and the second -helix [Wolberger et al., 1991]. Based on these and other observations, it would be expected that a wide range of sequence variation can be accomodated in the formation of the three-dimensional structure of homeodomains.

The Drosophila melanogaster neuronal protein prospero (pros) defines a family of proteins containing atypical homeodomains. Homologues of prospero have been identified in human, mouse, chicken, Fugu rubripes and Caenorhabditis elegans [Vaessin et al., 1991; Oliver et al., 1993; Bürglin, 1994; Tomarev et al., 1996]. Sequence comparison of prospero-type homeodomains revealed only 15-20% identity over the 60-residue region of representative homeodomains. Because of this low level of absolute identity, prospero-class proteins are not detected when a canonical homeodomain protein is used as the query to a BLASTP search. The primary sequence homology between pros-family members and typical homeodomain proteins is highest in the third helical region of the homeodomain, where all consensus amino acid residues are conserved. Within the pros-family members, the homeodomain region is absolutely conserved between chicken, mouse and human; it has 65-67% identity with sequences from Drosophila and C. elegans [Tomarev et al., 1996]. Among the pros-family members, sequence homology extends beyond the homeodomain. The 100 residues located C-terminal to the homeodomain (the prospero domain) are also conserved in the prospero-type proteins.

In this study, the sequences of prospero-type homeodomains were analyzed through threading analysis techniques [Bryant and Lawrence, 1993]. Threading methods such as the one employed here can be used to assess whether a given query sequence has the potential to adopt a known three-dimensional structure. This computational technique identifies alignments between a query sequence and a given folding motif that most likely represent stable conformational states, based on implied pairwise and hydrophobic interactions of residues which are non-local in the sequence [Bryant and Lawrence, 1993; Fetrow and Bryant, 1993]. As a consequence, this predictive method is capable of revealing structural similarities which are otherwise not obvious through conventional sequence-based methods. Here, primary sequences corresponding to the homeodomain region of pros-family members were threaded through the X-ray coordinates of the Drosophila engrailed homeodomain protein [Kissinger et al., 1990]. In the case of the divergent pros-family, the threading analysis predicts that it essentially forms the three helical-bundle motif of typical homeodomains, despite the fact that the primary sequences themselves are not statistically similar. The threading-derived alignment of the pros-family members with the common structural core of typical homeodomains is consistent with the known biochemical properties and mutagenesis studies of the pros-family [Hassan et al., 1997]. The threading alignment has also provided the basis for understanding the phylogenetic relationships between these proteins, as well as for the development of a new sequence signature that better characterizes this group.

MATERIALS AND METHODS

Threading Analysis

Threading experiments were performed by the method of Bryant and Lawrence [Bryant and Lawrence, 1993], with detailed derivations and methodology provided therein. Each query sequence was threaded through the atomic coordinates of the X-ray structure of the homeodomain of Drosophila DNA-binding protein engrailed [Kissinger et al., 1990]. Three core segments were defined based on the X-ray structure: CS1 spanned from residues 8 to 22, CS2 from residues 28 to 40, and CS3 from residues 42 to 54, the numbering being based on the sequence of pdb:1HDD presented in Figure 1. Intervening loop length constrains were 5 residues for loop 1 and 0-6 residues for loop 2. For each possible alignment, individual pairwise residue interactions were determined based on chemical type and distance intervals, lookup tables for which are present in Bryant and Lawrence [Bryant and Lawrence, 1993]. Using these values, a conformational energy G_R|M, defined as the expected work for substitution of a specific sequence R for a random sequence with the same composition in the context of folding motif M, was then calculated for each alignment. Z-scores (Z_R|M) and chance occurrence probabilities (E_R|M) were calculated to compare conformational energies for different alignments. Chance occurrence probabilities give the odds that a random sequence of the same length and amino acid composition would yield a threading energy as low as the query sequence R. Calculations of energies and statistical significance were performed using C and S-PLUS subroutines [Becker et al., 1988]. Critical interactions are defined as those having a pairwise interaction energy <1 kcal/mol. All energy scaffold figures were generated using the GRASP software package [Nicholls et al., 1991].

 

Figure 1: Multiple sequence alignment of homeodomains of selected proteins analyzed by homology model building. The sequences, in single letter amino acid code, are of Drosophila prospero, human Prox 1, C. elegans 65 KD pros-like protein, S. cerevisiae PHO2p, human PAX6, Drosophila sine oculis, human POU-homeodomain protein OCT1, rat thyroid transcription factor 1, and Drosophila engrailed. Amino acid residues showing absolute identity among these proteins are shown in white against a blue background; those positions with conservative substitutions are shown with a yellow background. The numbering scheme at the top of the figure refers to amino acid positions within the engrailed homeodomain. The positions of the three a-helices defined in the X-ray study of Drosophila engrailed homeodomain [Kissinger et al., 1990] are schematically represented in the bar below the alignment marked PDB. The core segments used in the threading analysis correspond to the boxed areas of the alignment. ALSCRIPT was used to format the alignment [Barton, 1993]. A comprehensive list of all homeodomain sequences is available at http://genome.nhgri.nih.gov/homeodomain/ [Banerjee-Basu and Baxevanis, 1999].

Sequence analysis

The structural alignment was also used as the basis for calculating a sequence signature representing the group. First-pass signature generation was done using PRATT 2.1 [Jonassen et al., 1995]; the signature was then fine-tuned manually. ScanProsite was used to test for the occurrence of the pattern against all SWISS-PROT entries (http://expasy.hcuge.ch/sprot/s cnpsit2.html).

RESULTS AND DISCUSSION

In this study, model structures were generated for prospero-type homedomains using homology model building [Bryant and Lawrence, 1993]. An automated fold prediction search using the UCLA-DOE Fold Recognition Server (http://www.doe-mbi.ucla.edu/) with each of the pros-family members identified the homeodomain protein engrailed as the candidate whose fold most likely best represents the entire pros-family. Based on this, engrailed was selected as the structural template through which each query sequence was threaded. The three-dimensional structure of Drosophila engrailed homeodomain bound with DNA (1HDD, chain C; [Kissinger et al., 1990]) was used to define the conserved structural elements (the core segments) within this motif. The core segments used in this threading experiment exclude most of the flexible regions; unstructured regions such as loops cannot be modeled using this technique. The intervening loop length limits were based on the sequence variations observed within the homeodomain sequences [ Bürglin, 1994]. All possible placements of the core segments along the query sequence given the constraints of sequence length, core segment length, and limits of loop length were considered. Threading contact energies were corrected for sequence composition bias by random shuffling of the aligned residues to generate composition corrected threading scores (Z_R|M). To evaluate the statistical significance of threading scores, 100 random permutations of the query sequence were generated and the alignment optimization procedure was repeated on these shuffled sequences. From the distribution of these threading scores, the probability (E_R|M) that the threading score for the query sequence would be observed by chance was calculated. A summary of the threading results is shown in Tab. 1. Threads with the most favorable conformational energies (i.e., those with the lowest G_R|M) were selected for further study.

Energy scaffolds generated in the threading experiments are shown in Fig. 2. For the pros-family members, a three-helix bundle held together by hydrophobic interactions (thick, magenta-colored cylinders) is evident. Several of the highly conserved, large hydrophobic residues are involved in maintaining interactions within the hydrophobic core. These interactions are primarily between the conserved residues in the first and third helix. Examination of the energy scaffolds indicates that the most favorable hydrophobic interactions observed in the threading model of engrailed involve Phe8, Leu13, Leu16, and Phe20 in core region I; Leu34, Leu38, Leu40 in core region II; and Ile45, Trp48, and Phe49 in core region III (Fig. 2a). The homeodomain consensus sequence includes all of these positions [Bürglin, 1994]. In the atypical prospero-type homeodomain model, all the corresponding positions except for position 16 are involved in highly favorable hydrophobic interactions. At position 16, a Leu-to-Ala substitution is present in prospero-type homeodomains. Although the highly favorable Leu16 interaction is missing from pros-family members, additional hydrophobic interactions involving Met19 in helix I, Leu31 and Phe35 in helix II, and Val/Ile46 and Phe52 in helix III help create an energy-favorable hydrophobic core (Figs. 2b, 2c and 2e).

 

a  b  c  d  e  f  g  h

Figure 2: Energy scaffolds for the models containing the homeodomain. The -carbon backbone of the protein is depicted as a curving "worm". Within the backbone, segments of the homeodomain comprising the core folding motifs are shown in blue, while the intervening loop regions are shown in yellow. Helices and loop regions are as defined in Fig. 1. Pairwise residue interaction energies between core residues [Bryant and Lawrence, 1993] are shown by the width and coloring of the connected -carbon positions on the protein backbone. Indicated interactions are limited to those with pairwise interaction energies <­1 kcal/mol. Thick, magenta-colored cylinders indicate the most favorable interactions. Intermediate cylinder thicknesses represent interactions with lower pairwise energies. Scaffolds were generated using the graphics program GRASP [Nicholls et al., 1991]. Several examples of threaded homeodomain sequences are displayed: (a) Drosophila engrailed homeodomain [Kissinger et al., 1990] threaded through its own structure; (b) vertebrate Prox 1 homeodomain [Tomarev et al., 1996]; (c) C. elegans CEH-26 protein; (d) S. cerevisiae PHO2p homeodomain; (e) Drosophila prospero homeodomain [Chu-Lagraff et al., 1991]; and (f) Drosophila prospero mutation, Trp48 and Phe49 replaced by Ala substitution [Hassan et al., 1997]; (g) Human Pax6 protein [Ton et al., 1991]; (h) Drosophila sine oculis protein [Cheyette et al., 1994]. The amino acid numbering corresponds to that in the multiple sequence alignment presented in Fig. 1.

A majority of the typical homeodomains recognize and bind with high affinity to the DNA sequence motif 5'-TAATNN-3' [Gehring et al., 1994b; Billeter, 1996]. In the X-ray structure of the engrailed homeodomain-DNA complex, several residues in the exposed hydrophilic face of helix III establish specific contacts with the last four base pairs of the recognition sequence, whereas the residues in the N-terminal arm contact the first two base pairs. The invariant amino acid residue Asn51 in the third helix establishes contact with the adenine residue at position 3 in the core sequence motif, while the other invariant residue, Arg53, makes phosphate contacts in the complementary DNA strand. The major groove contacts by Asn51 and Arg53 have been observed in all DNA-bound homeodomain structures solved so far [Ades and Sauer, 1995; Billeter, 1996]. These two positions are conserved in prospero-type homeodoamins. Amino acid residues in positions 47, 50, and 54, located at DNA-binding side of the third helix, contribute to the selectivity of the DNA-binding site. Site-directed mutagenesis of Gln to Lys at the position 50 alters the DNA-binding specificity of engrailed homeodomain without altering the homeodomain structure [Tucker-Kellogg et al., 1997]. Compilation of base contacts indicates that the residue at position 50 is the principal determinant of binding specificity of homeodomain sub-classes. In prospero-type homeodomains, the occurrence of Ser50 is reminiscent of paired-type homeodomains, while Lys47 and Glu54 in the recognition helix diverge completely from the homeodomain consensus. The combinatorial effect of Lys47, Ser50, and Glu54, along with a divergent N-terminal arm, might mediate recognition of a non-canonical DNA-binding site, 5'-C[A/t][c/t]NNC[T/c]-3', by Drosophila prospero [Hassan et al., 1997].

A prospero mutant involving both of the conserved Trp48 and Phe49 residues in the third helix (WF AA) was also analyzed [Hassan et al., 1997]. This loss-of-function mutant, which cannot bind to DNA, showed marked reduction in its ability to transactivate a luciferase reporter construct containing multiple pros-binding sites [Hassan et al., 1997]. As shown in the energy scaffold of this mutant model (Fig. 2f), several of the highly favorable hydrophobic interactions involving Trp48 and Phe49 residues are missing in this construct. A consequent 9.6% increase in the G_R|M of this model formation is also seen (Tab. 1). Other highly conserved residues in the primary sequence of pros homeodomain were also replaced to examine their contribution to the structure of prospero. Among the conserved positions in the third helix, replacement of either Asn51 or Arg53 by alanine residues did not significantly change the threading scores (data not shown). This observation is consistent with the prediction that the residues at the positions 51 and 53 lie on the hydrophilic-side of the recognition helix and do not participate in the formation of the hydrophobic core. Replacement of Ile38 by Ala in the second helix dramatically changed the interaction between the top residues in helix I and helix II (Fig. 3b). This interaction is probably necessary for positioning of the helices as well as proper placement of the conserved loop between the helix I and II.

 

a	Figure 3: Energy scaffolds for prospero mutation I38A. Details of energy scaffold representation are the same as for Figure 2. (a) Drosophila prospero homeodomain [Chu-Lagraff et al., 1991]; (b) Drosophila prospero Ile38Ala substitution.
b

Table 1: Statistics for optimal threads of homeodomain-containing proteins

	Accession Number	deltaG(R|M)	Z-score	P-value [Z(R|M)]	Log odds	Percent Identity
Prospero	P29617	-53.043	3.815	0.005	0.711	20
Prox 1	Q92786	-50.812	3.664	0.026	0.069	20
CEH-26	P34522	-60.224	4.495	<0.0001	3.152	20
Prospero WF -> AA	P29617	-47.907	3.418	0.035	-0.059	20
PHO2	P07269	-51.694	3.988	0.007	0.606	37
Pax6	P26367	-49.153	3.903	0.012	0.384	37
DSO (sine oculis)	Q27350	-56.340	5.277	<0.0001	3.170	25
Oct-1	P14859	-62.427	4.415	0.001	1.290	26
TTF-1	P23441	-44.897	3.669	0.024	0.107	40
Engrailed	S03667	-45.229	3.757	0.030	0.000	100

The Saccharomyces cerevisiae homeodomain protein PHO2p, which contains all of the homeodomain consensus elements, was also included in this threading analysis. The PHO2p sequence shows a significant threading score with a high chance occurrence probability of homeodomain structure formation (Tab. 1). The energy scaffold for PHO2p homeodomain model (Fig. 2d) preserves all the favorable hydrophobic interactions that were observed for the engrailed self-thread. In addition, Val15 and Leu37 are also involved in favorable interhelical interactions. The human Pax6 protein, which is implicated in aniridia, congenital cataracts, and other clinical disorders [Prosser and van Heyningen, 1998], also gives good threading scores with the parameters used in this study (Tab. 1). Threading of the Drosophila sine oculuis sequence, which has low similarity with the homeodomain consensus (25% identity), yields the best probability for the homeodomain motif formation (Tab. 1). In the energy scaffold for the sine oculis homeodomain model, additional favorable hydrophobic interactions are conferred by Val15 and Trp19 in helix I (Fig. 2h). To test the validity of the core segments defined for engrailed (see Materials and Methods), the primary sequences of Oct-1 and TTF-1 homeodomain proteins, whose three-dimensional structures are known, were also threaded in this analysis. As expected, both Oct-1 and TTF-1 yield good ZR|M and ER|M scores (Tab. 1). The threading scores show little correlation with the degree of sequence identity (Tab. 1), indicating that there are no rigid sequence requirements for the formation of homeodomain structural motif.

Based on this threading study, residues within the homeodomain can be classified into two groups: the larger group is involved in forming and maintaining the three helix homeodomain motif as well as in establishing contacts with the DNA phosphate backbone, while the smaller group is involved in determining the specificity of DNA binding. Although the atypical prospero-type homeodomain shows very low sequence conservation within the homeodomain, prospero-type homeodomains not being detected as homeodomains in a standard BLASTP search, the critical amino acids responsible for maintaining the three-dimensional structure are almost all conserved.

Surface potential of Drosophila engrailed homeodomain

Inspection of the electrostatic potential distribution on the homeodomain surface reveals that the majority of the positive charges are concentrated on the DNA-binding face of the homeodomain (Fig. 4). Indeed, a positively charged DNA-binding groove is formed by the basic residues in both the third helix and the N-terminal arm, with the amino acid residues conferring target DNA-binding site specificity being located within this groove. Positive charges are primarily contributed by Arg3 and Arg5 in the N-terminal arm, as well as Lys55 and Lys58 in helix III. The opposite face of the engrailed homeodomain (away from the DNA-binding side) is predominantly hydrophobic. Based on sequence comparisons, it was evident that the homeodomains of the pros family members contain significantly fewer basic amino acid residues. None of the positions in the engrailed homeodomain that contribute to the formation of the positively charged DNA-binding face of the homeodomain are conserved in prospero-type homeodomains. However, two other positions in the helix III of engrailed, Ile47 and Ala54, are replaced by charged residues (Lys47 and Glu54) in prospero-type homeodomains. The divergence in charged amino acid distribution on the homeodomain surface indicates a very different mode of DNA-protein interactions for pros family members as compared to engrailed.

 

Figure 4: Molecular surface of Drosophila engrailed homeodomain. Regions of the surface greater than +10 kT, equal to 0 kT, and less than -10 kT are colored blue, white and red, respectively. Colors are linearly interpolated for the intermediate values. The surface potentials were calculated using GRASP software package [Nicholls et al., 1991]. The majority of the positively charged surfaces (blue) are concentrated at the DNA-binding face of the homeodomain. The amino acid residues that contribute to the formation of the basic surface are labelled as R3, R5, K18, K55, and K58 using single letter amino acid abbreviations.

Homeodomain domain signature

Based on the threading experiments and the subsequent structural alignment, a new amino acid signature was deduced for the homeodomain domain. Using the standard PROSITE notation [Bairoch et al., 1997], the signature

[FY]-x(19)-[LMF]-x(4,7)-[LIV]-x(2)-W-F-x-N-x-R

unambiguously identifies the threaded homeodomain proteins. As with other signatures previously put forth for the homeodomain and homeodomain-related motifs (PROSITE PDOC00027, PDOC00032, and PDOC00033), most of the invariant and conserved positions used to identify the domain correspond to helix III, within the DNA-binding region. In contrast to these other signatures, though, the first position of the new signature is at position 20, within helix I. This position is occupied by a phenylalanine in all the homeodomains examined in this threading analysis except for the Drosophila sine oculis homeodomain. All sine oculis family members contain a conserved tyrosine residue at position 20 within the homeodomain. In each of the energy scaffolds generated in this threading analysis (Fig. 2), position 20, located at the C-terminal end of helix I, is involved in at least one major interaction with residues along helix III. This interaction is most likely important in maintaining the position of both helix I and loop 1 with respect to the rest of the homeodomain motif. In addition to the highly conserved residues in the DNA-binding face of helix III (W-F-X-N-X-R), the residues at position 40 (LMF) and position 45 (LIV) included in this signature are involved in maintaining interactions at the helical interfaces. The variable spacer accommodates the divergence observed in loop lengths between the helix II and III among the homeodomain family members considered here. As alluded to above, the homeodomain domain signature (PDOC00027) encompasses only the C-terminal portion of helix II and all of helix III. That pattern does not match the structural alignment put forth herein, due to the variability in the loop between helices II and III and the newly-assigned positions of the individual residues based on structural considerations.

There is an important distinction to be made between the two signatures in terms of what sequences they are able to identify. As previously stated, the prospero class of proteins is considered to be atypical, and indeed, PDOC00027 does not recognize the human, chicken, mouse, or Drosophila prospero sequences. (Recall also that BLASTP searches do not detect prospero-class proteins when a canonical homeodomain protein is used as the query.) Since the role of prospero as a sequence-specific DNA-binding protein involved in neuronal differentiation [Hassan et al., 1997] argues that these proteins do indeed function as homeodomain proteins, it appears that the new signature is then simply identifying a discrete subset of homeodomain proteins. The new signature is more inclusive (or specific) for this particular homeodomain class. Based on this, it can be concluded that there are alternative structural classes of homeodomain proteins, all of which share certain (but not global) common sequence characteristics. While all of these proteins bind to DNA, the subtle differences within the homeodomain classes are most likely responsible for their differences in cellular function and localization. This case provides an interesting example of how computational experiments at the structural level can influence analysis at the sequence level, providing the basis for further refinement of sequence-based definitions of this and other protein families.

Future Perspectives

In this study, we have used threading algorithms developed by Bryant et al. [Bryant and Lawrence, 1993; Bryant and Altschul, 1995; Bryant, 1996] to predict the three-dimensional structural model of prospero-type homeodomains. Although the pros-family members show low sequence identity with the homeodomain consensus, the threading analysis predicts that the homeodomain of this sub-class essentially forms the three helical-bundle motif of typical homeodomains. This is consistent with the idea that structure is indeed conserved to a greater extent than sequence, and there is only a limited repertoire of backbone structural motifs [Chothia, 1992].

The structural model of prospero-type homeodomains presented here is the first step towards understanding the molecular basis of its non-canonical DNA-binding site recognition. The availability of this model will help in the intelligent design of site-directed mutagenesis experiments to determine the precise sites of interaction in homeodomain/DNA complex that ultimately give rise to the unique binding specificity of this class of homeodomains.

REFERENCES

---

• Ades, S. E. and Sauer, R. T. (1995). Specificity of minor-groove and major-groove interactions in a homeodomain-DNA complex. Biochemistry 34, 14601-8.

  ---

• Bairoch, A., Bucher, P. and Hofmann, K. (1997). The PROSITE database: its status in 1997. Nucleic Acids Res. 25, 217-221.

  ---

• Banerjee-Basu, S., Ferlanti, E. S., Ryan, J. F. and Baxevanis, A. (1999). The homeodomain resource: sequences, structures and genomic information. Nucleic Acids Res. 27, 336-337.

  ---

• Barton, G. J. (1993). ALSCRIPT: a tool to format multiple sequence alignments. Protein Eng. 10, 37-40.

  ---

• Becker, R., Chambers, J. and Wilks, A. (1988). The new S language-A programming environment for Data Analysis and Graphics, Wadsworth, Pacific Grove, CA, USA.

  ---

• Billeter, M. (1996). Homeodomain-type DNA recognition. Prog. Biophys. Mol. Biol. 66, 211-225.

  ---

• Bryant, S. H. (1996). Evaluation of threading specificity and accuracy. Proteins 26, 172-185.

  ---

• Bryant, S. H. and Altschul, S. F. (1995). Statistics of sequence-structure threading. Curr. Opin. Struct. Biol. 5, 236-244.

  ---

• Bryant, S. H. and Lawrence, C. E. (1993). An empirical energy function for threading protein sequence through folding motif. Proteins Struct. Func. Genet. 16, 92-112.

  ---

• Bürglin, T. R. (1994). Guidebook to the Homeobox Genes (Duboule, D., Ed.), Oxford University Press, Oxford.

  ---

• Ceska, T. A., Lamers, M., Monaci, P., Nicosia, A., Cortese, R. and Suck, D. (1993). The X-ray structure of an atypical homeodomain present in the rat liver transcription factor LFB1/HNF1 and implications for DNA binding. EMBO J. 12, 1805-1810.

  ---

• Cheyette, B. N., Green, P. J., Martin, K., Garren, H., Hartenstein, V. and Zipursky, S. L. (1994). The Drosophila sine oculis locus encodes homeodomain-containing protein required for the development of the entire visual system. Neuron 12, 977-996.

  ---

• Chothia, C. (1992). Proteins. One thousand families for the molecular biologist. Nature 357, 543-544.

  ---

• Chu-Lagraff, Q., Wright, D. M., McNeil, L. K. and Doe, C. Q. (1991). The prospero gene encodes a divergent homeodomain protein that controls neuronal identity in Drosophila. Development, Suppl. 2, 79-85.

  ---

• Dekker, N., Cox, M., Boelens, R., Verrijzer, C. P., van der Vliet, P. C. and Kaptein, R. (1993). Solution structure of the POU-specific DNA-binding domain of Oct-1. Nature 362, 852-855.

  ---

• Endo, T., Ohta, K., Saito, T., Haraguchi, K., Nakazato, M., Kogai, T. and Onaya, T. (1994). Structure of the rat thyroid transcription factor-1 (TTF-1) gene. Biochem. Biophys. Res. Commun. 204, 1358-1363.

  ---

• Fetrow, J. S. and Bryant, S. H. (1993). New programs for protein tertiary structure prediction. Biotechnology 11, 479-484.

  ---

• Gehring, W. J., Affolter, M. and Bürglin, T. (1994a). Homeodomain proteins. Annu. Rev. Biochem. 63, 487-526.

  ---

• Gehring, W. J., Qian, Y. Q., Billeter, M., Furukubo-Tokunaga, K., Schier, A. F., Resendez-Perez, D., Affolter, M., Otting, G. and Wüthrich, K. (1994b). Homeodomain-DNA recognition. Cell 78, 211-223.

  ---

• Gruschus, J. M. , Tsao, D. H., Wang, L. H., Nirenberg, M. and Ferretti, J. A. (1997). Interactions of the vnd/NK-2 homeodomain with DNA by nuclear magnetic resonance spectroscopy: basis of binding specificity. Biochemistry 36, 5372-5380.

  ---

• Hassan, B., Li, L., Bremer, K. A., Chang, W., Pinsonneault, J. and Vaessin, H. (1997). Prospero is a panneural transcription factor that modulates homeodomain. Proc. Natl. Acad. Sci. USA 94, 10991-10996.

  ---

• Jonassen, I., Collins, J. F. and Higgins, D. G. (1995). Finding flexible patterns in unaligned protein sequences. Protein Science 4, 1587-1595.

  ---

• Kissinger, C. R., Liu, B. S., Martin-Blanco, E., Kornberg, T. B. and Pabo, C. O. (1990). Crystal structure of an engrailed homeodomain-DNA complex at 2.8 A resolution: a framework for understanding homeodomain-DNA interactions. Cell 63, 579-590.

  ---

• Laughon, A. (1991). DNA Binding Specificity of Homeodomains. Biochemistry 30, 11357-11367.

  ---

• Liu, B., Kissinger, C. R. and Pabo, C. O. (1990). Crystallization and preliminary X-ray diffraction studies of the engrailed homeodomain and of an engrailed homeodomain/DNA complex. Biochem. Biophys. Res. Commun. 171, 257-259.

  ---

• Nicholls, A., Sharp, K. A. and Honig, B. (1991). Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins 11, 281-296.

  ---

• Oliver, G., Sosa-Pineda, B., Geisendorf, S., Spana, E. P., Doe, C. Q. and Gruss, P. (1993). Prox 1, a prospero-related homeobox gene expressed during mouse development. Mech. Dev. 44, 3-16.

  ---

• Prosser, J. and van Heyningen, V. (1998). PAX6 mutations reviewed. Hum. Mutat. 11, 93-108.

  ---

• Qian, Y. Q., Billeter, M., Otting, G., Muller, M., Gehring, W. J. and Wüthrich, K. (1989) The structure of the Antennapedia homeodomain determined by NMR spectroscopy in solution: comparison with prokaryotic repressors. Cell 59, 573-580.

  ---

• Qian, Y. Q., Furukubo-Tokunaga, K., Resendez-Perez, D., Muller, M., Gehring, W. J. and Wüthrich, K. (1994). Nuclear magnetic resonance solution structure of the fushi tarazu homeodomain from Drosophila and comparison with the Antennapedia homeodomain. J. Mol. Biol. 238, 333-345.

  ---

• Tomarev, S., Sundin, O., Banerjee-Basu, S., Duncan, M., Yang, J. and Piatigorsky, J. (1996). Chicken homeobox gene Prox 1 related to Drosophila prospero is expressed in the developing lens and retina. Dev. Dyn. 206, 354-367.

  ---

• Ton, C. C., Hiroven, H., Miwa, H., Weil, M. M., Monaghan, P., Jordan, T., Heyningen, V., Hastie, N. D., Meijers-Heijboer, H., Drechsler, M. et al. (1991). Positional cloning and characterization of a paired box- and homeobox-containing gene from the aniridia region. Cell 67, 1059-1054.

  ---

• Tucker-Kellogg, L., Rould, M. A., Chambers, K. A., Ades, S. E., Sauer, R. T. and Pabo, C. O., (1997). Engrailed (Gln50->Lys) homeodomain-DNA complex at 1.9 Å resolution: structural basis for enhanced affinity and altered specificity. Structure 5, 1047-1054.

  ---

• Vaessin, H., Grell, E., Wolff, E., Bier, E., Jan, L. Y. and Jan, Y. N. (1991). prospero is expressed in neuronal precursors and encodes a nuclear protein that is involved in the control of axonal outgrowth in Drosophila. Cell 67, 941-953.

  ---

• Wolberger, C., Vershon, A. K., Liu, B., Johnson, A. D. and Pabo, C. O. (1991). Crystal structure of a MAT alpha 2 homeodomain-operator complex suggests a general model for homeodomain-DNA interactions. Cell 67, 517-528.