Domain-specific Interactions of Human HP1-type Chromodomain Proteins and Inner Nuclear Membrane Protein LBR*

HP1-type chromodomain proteins self-associate as well as interact with the inner nuclear membrane protein LBR (lamin B receptor) and transcriptional coactivators TIF1 (cid:97) and TIF1 (cid:98) . The domains of these proteins that mediate their various interactions have not been entirely defined. HP1-type proteins are predicted by hydrophobic cluster analysis to consist of two homologous but distinct globular domains, corresponding to the chromodomain and chromo shadow domain, separated by a hinge region. We show here that the chromo shadow domain mediates the self-associations of HP1- type proteins and is also necessary for binding to LBR both in vitro and in the yeast two-hybrid assay. Hydro- phobic cluster analysis also predicts that the nucleoplasmic amino-terminal portion of LBR contains two globular domains separated by a hinge region. The interactions of the LBR domains with an HP1-type protein were also analyzed by the yeast two-hybrid and in vitro binding assays, which showed that a portion of the sec- ond globular HP1 Hs (cid:97) values were obtained for HP1 Hs (cid:98) and HP1 Hs (cid:103)

HP1 is a heterochromatin protein originally identified in Drosophila melanogaster, where it functions as a suppressor of position effect variegation (1,2). HP1 and homologous proteins in other species are included in the chromo superfamily of proteins (3,4). The identifying feature of this superfamily is the chromodomain, which was first shown to be common to HP1 and Polycomb, a Drosophila protein involved in the downregulation of homeotic selector genes during development (5).
HP1-type proteins, which are included in Class A of the chromo superfamily, contain a second domain, homologous to but distinct from the chromodomain, which has been termed the chromo shadow domain (3).
In contrast to Drosophila, the functions of the highly conserved mammalian HP1-type proteins remain mostly unknown. In humans, three HP1-type proteins that arise from different genes, termed HP1 Hs␣ , HP1 Hs␤ , and HP1 Hs␥ , have been described (6 -8). In higher eukaryotic cells, a portion of the transcriptionally inactive heterochromatin is associated with the nuclear envelope (9,10), and we have demonstrated that human HP1-type proteins bind to the inner nuclear membrane protein LBR (8). LBR (lamin B receptor), which also binds to B-type lamins, has a nucleoplasmic amino-terminal domain of ϳ210 amino acids and a hydrophobic domain with eight putative transmembrane segments (11)(12)(13)(14). Besides possibly playing a role in the attachment of the nuclear lamina and the heterochromatin to the nuclear envelope in interphase, LBR may also function in targeting inner nuclear membrane vesicles to chromatin at the end of mitosis (15,16). Mouse HP1-type polypeptides have also been shown to bind to the transcriptional coactivators TIF1␣ and TIF1␤ (17). TIF transcriptional coactivators interact with the ligand-binding domains of retinoic acid and other nuclear receptors and may mediate their activation functions (18). In the yeast two-hybrid assay, mouse HP1-type proteins have also been shown to selfassociate (17).
The structure of HP1-type proteins suggests that different domains may mediate their different interactions. In this study, we have used hydrophobic cluster analysis (HCA) 1 (19 -21) to predict the secondary structures of human HP1-type proteins and LBR. We then used the structural information obtained by using HCA to identify the domains of these proteins that mediate some of their protein-protein interactions.

EXPERIMENTAL PROCEDURES
Protein Sequence Analysis Using HCA-Guidelines for the use of this method have been published previously (19 -21). HCA allows comparisons not only of amino acid sequences, but also the protein secondary structures statistically centered on hydrophobic clusters and their distributions (20). The effectiveness of HCA in predicting protein secondary structure and identifying low levels of sequence homology has been widely demonstrated (for some examples, see Refs. [22][23][24][25][26][27]. For sequence alignments, the accuracy was assessed by computing identity, similarity, and HCA scores as well as the corresponding Zscores (27). Z-scores represent the differences between the alignment score under consideration and the mean score of a distribution com-puted from the alignment of sequence 1 versus a large number of randomly shuffled versions of sequence 2. Differences are then expressed relative to the standard deviation of the random distribution. Scores that are 3.0 standard deviations or greater above the scrambled mean suggest authentic relationships (27).
Plasmid Construction-Complementary DNAs were generated by the polymerase chain reaction (28) using the Gene Amp kit (Hoffmann-La Roche) and a PCR System 2400 thermocycler (Perkin-Elmer Corp.). To amplify DNA sequences, custom oligonucleotide primers (DNAgency, Malvern, PA) designed with restriction endonuclease sites at their 5Ј-ends were used with clones for human LBR (13) and HP1 Hs␣ (8) as templates. Amplified cDNAs were purified and cloned into the plasmids of choice by standard methods.
As our previously characterized cDNA clone for HP1 Hs␥ was lacking the first 17 putative amino acids based on sequence homology to other HP1-type proteins (8), we generated a full-length clone by performing the polymerase chain reaction with HeLa cell cDNA from a plasmid library (CLONTECH) as template. Oligonucleotide primers (DNAgency) based on the 3Ј-sequence of the partial HP1 Hs␥ cDNA clone (GenBank™ accession number U26312) and a putative 5Ј-sequence in GenBank™ (accession number Z15820) were used in the reaction. The amplified cDNA was cloned into plasmid pACT2 (CLONTECH) and sequenced (the entry with GenBank™ accession number U26313 has been updated to reflect the complete cDNA sequence of HP1 Hs␥ ). This full-length HP1 Hs␥ cDNA clone was used as a template for the polymerase chain reaction to generate HP1 Hs␥ cDNAs for cloning as described above.
Binding Assays Using in Vitro Translated Proteins-For in vitro transcription-translation, cDNAs were cloned into pBFT4 (supplied by J. Licht, Mount Sinai School of Medicine, New York). Proteins expressed using this plasmid contained a FLAG epitope at their amino termini. In vitro translation was performed using the TNT T7 coupled reticulocyte lysate system (Promega) with [ 35 S]methionine (Amersham Corp). SDS-polyacrylamide slab gel electrophoresis (PAGE) (29) followed by autoradiography was performed to confirm the synthesis of proteins. In binding experiments, 20 l of in vitro translation lysate were added to 200 l of binding buffer (150 mM NaCl, 20 mM NaHepes (pH 7.4), 10% glycerol, 0.5% bovine serum albumin, and 0.05% Nonidet P-40) with 20 l of glutathione-Sepharose (Pharmacia Biotech Inc.) coupled to 3-5 g of a glutathione S-transferase (GST) fusion protein.
Plasmids that expressed GST fusion proteins were constructed by clon-ing the desired cDNA into pGEX-2T (Pharmacia Biotech Inc.). Fusion proteins were expressed and purified as described (30). Suspensions containing lysates and GST fusion proteins were incubated at 4°C with rotation for ϳ2 h. After incubation, glutathione-Sepharose was washed five times with binding buffer, and bound proteins were then eluted with 4% SDS, separated by SDS-PAGE, and detected by autoradiography.
Yeast Two-hybrid Assay-To generate plasmids for use in the yeast two-hybrid assay, cDNAs were cloned into pACT2, which expresses fusion proteins with the GAL4 transcription activation domain, or pGBT9 (provided by S. Fields, University of Washington, Seattle), which expresses fusion proteins with the GAL4 DNA-binding domain. The desired plasmids were cotransformed into Saccharomyces cerevisiae strain Y190 (provided by S. J. Elledge, Baylor College of Medicine, Houston, TX). The two-hybrid assay and measurement of ␤-galactosidase activity were performed as described previously (31,32).

HCA of Human HP1-type Protein Sequences-
We have used HCA to analyze the sequences of human HP1 Hs␣ , HP1 Hs␤ , and HP1 Hs␥ . This method is highly effective in detecting threedimensional similarities between proteins with limited sequence identities (typically 10 -30%), primarily by its ability to detect related secondary structure elements statistically centered on hydrophobic clusters. Similar HCA plots for two sequences are therefore indicative of putative three-dimensional relationships. HCA of human HP1 Hs␣ revealed two globular domains separated by a hinge region of ϳ70 amino acids ( Fig.  1). Similar results were obtained for HP1 Hs␤ and HP1 Hs␥ (data not shown). The first of the two globular domains in the HP1 polypeptides was originally identified by Paro and Hogness (5) as the chromodomain common to the Drosophila proteins Polycomb and HP1. Analysis of the second globular domain of HP1 Hs␣ suggested that it contained a three-dimensional fold similar to the chromodomain, probably as a result of an internal duplication (Fig. 1). The validity of this prediction was assessed by calculation of Z-scores from the comparison of the FIG. 1. General organization of HP1 Hs␣ as predicted from the analysis of its HCA plot. HCA readily detects two globular domains (boxed) of similar structure separated by a hinge region largely made up of hydrophilic amino acids. The first globular domain is the chromodomain (labeled A), and the second is the chromo shadow domain (labeled B). Protein sequences are shown on a duplicated ␣-helical net with amino acid positions indicated above. The contours of the hydrophobic residues are automatically drawn to form clusters that mainly correspond to the internal faces of regular secondary structures (20). Similarities between the distribution of clusters and their features (light shading) as well as sequence identities (dark shading) are readily observed between the chromodomain and chromo shadow domain, suggesting that they would adopt a similar fold. The symbols used in the plot are as follows: open square, threonine; square with dot inside, serine; diamond, glycine; and star, proline. alignment of the tandemly duplicated domains with 10,000 alignments performed after randomization of one of the compared sequences. The Z-scores allow for evaluation of how the signal emerges from the background, with values exceeding 3.0 expected to represent an authentic relationship. The Z-identity value of 4.75, the Z-similarity value of 6.30, and the Z-HCA score of 4.72 confirm an authentic relationship between the chromodomain and the second globular domain in human HP1 Hs␣ . Similar values were obtained for HP1 Hs␤ and HP1 Hs␥ . While this work was in progress, Koonin et al. (4) and Aasland and Stewart (3), who termed the second globular domain the chromo shadow domain, published similar results obtained using different sequence analysis methods.
Self-associations of HP1 Proteins-Mouse HP1-type proteins, which are 98 -100% identical to their human homologues, self-associate via their chromo shadow domains in the yeast two-hybrid assay (17). We further examined the interactions of HP1 Hs␣ and HP1 Hs␥ in an in vitro solution binding assay (Fig. 2). 35 S-Labeled HP1 Hs␣ produced by in vitro translation in reticulocyte lysates was precipitated by the GST-HP1 Hs␣ and GST-HP1 Hs␥ fusion proteins, but not by GST (Fig.  2, lanes 1-4). Similarly, 35 S-labeled HP1 Hs␥ was precipitated by the GST-HP1 fusion proteins, but not by GST (Fig. 2, lanes  5-8). These results show that HP1 Hs␣ and HP1 Hs␥ can selfassociate and associate with each other in vitro.
To identify the domain of HP1-type proteins responsible for these associations, we expressed various portions of 35 S-labeled HP1 Hs␣ by in vitro translation (Fig. 3A). GST-HP1 Hs␣ and GST-HP1 Hs␥ fusion proteins precipitated full-length HP1 Hs␣ (Fig. 3B, lanes 1 and 5) from reticulocyte lysates, but not the portion of HP1 Hs␣ that contained the chromodomain plus the hinge region (Fig. 3B, lanes 2 and 6). The GST fusion proteins also bound to the hinge region plus the chromo shadow domain of HP1 Hs␣ (Fig. 3B, lanes 3 and 7) and to the chromo shadow domain alone (Fig. 3B, lanes 4 and 8). Hence, the chromo shadow domain was necessary and sufficient for the self-association of HP1 Hs␣ and its binding to HP1 Hs␥ . These results are consistent with those previously reported using the yeast twohybrid assay, which showed that the chromo shadow domain of the mouse homologue of HP1 Hs␣ interacted with mouse HP1 Hs␣ and a portion of MOD1, the mouse homologue of HP1 Hs␥ (17).
The Chromo Shadow Domain of HP1 Hs␣ Is Necessary for Binding to LBR-We used the yeast two-hybrid assay to identify the domain of HP1 Hs␣ that binds to the nucleoplasmic amino-terminal domain of LBR (Fig. 4). Full-length HP1 Hs␣ and the hinge region plus the chromo shadow domain interacted with LBR in the two-hybrid assay. The chromo shadow domain alone also gave a positive result. Neither the portion of HP1 Hs␣ that contained the chromodomain plus the hinge region nor the hinge region alone interacted with the aminoterminal domain of LBR. The chromo shadow domain was therefore necessary and sufficient for the interaction of HP1 Hs␣ with LBR in the yeast two-hybrid assay.
We also examined the binding of various domains of HP1 Hs␣ , synthesized by in vitro translation, to a GST fusion protein of the amino-terminal domain of LBR (Fig. 5). 35 S-Labeled fulllength HP1 Hs␣ and polypeptides containing the chromodomain plus the hinge region, the chromo shadow domain plus the hinge region, and the chromo shadow domain alone were synthesized by in vitro translation in reticulocyte lysates. Fulllength HP1 Hs␣ was precipitated by the GST-LBR fusion protein (Fig. 5, lane 1), as was the protein that contained the chromo FIG. 2. HP1 Hs␣ and HP1 Hs␥ self-associate and bind to each other. 35 S-Labeled HP1 Hs␣ (lanes 1-4) and HP1 Hs␥ (lanes 5-8) were synthesized in reticulocyte lysates by in vitro translation and incubated with GST (lanes 2 and 6), GST-HP1 Hs␣ (lanes 3 and 7), or GST-HP1 Hs␥ (lanes 4 and 8) coupled to glutathione-Sepharose. Ten percent of the amount of incubated 35 S-labeled protein used in each binding reaction is also shown (lanes 1 and 5). After washing, the proteins that remained bound to glutathione-Sepharose were eluted with SDS, separated by SDS-PAGE, and detected by autoradiography. Migration of molecular mass standards (in kilodaltons) is indicated on the left.  1-4) or GST-HP1 Hs␥ (lanes 5-8) coupled to glutathione-Sepharose. After washing, the proteins that remained bound to glutathione-Sepharose were eluted with SDS, separated by SDS-PAGE, and detected by autoradiography. Approximately 10 -15% of the input material bound specifically to the proteins, which generated positive results. Migration of molecular mass standards (in kilodaltons) is indicated on the left.
shadow domain plus the hinge region (lane 3). Deletion of the chromo shadow domain from HP1 Hs␣ abolished its binding to LBR (Fig. 5, lane 2). The chromo shadow domain alone, however, did not associate with LBR in this assay (Fig. 5, lane 4). Hence, the chromo shadow domain was necessary for the binding of HP1 Hs␣ to LBR. In contrast to the results obtained in the yeast two-hybrid assay, it was not sufficient and required additional amino acids of the hinge region at its amino-terminal side. These additional amino acids plus the chromodomain did not bind to LBR, suggesting that they alone did not mediate the interaction.
HCA of the Sequence of the Nucleoplasmic Domain of LBR-We have previously demonstrated that the nucleoplasmic amino-terminal domain of LBR binds to human HP1-type proteins (8). The sequence of this segment of LBR was analyzed by HCA (Fig. 6). This analysis suggested that the amino-terminal domain of LBR contained two globular domains separated by a hinge region of ϳ40 amino acids. The first globular domain of LBR is located between amino acids 1 and 60. The second globular domain is roughly located between amino acids 105 and 210. The hinge region between these two globular domains is highly charged between amino acids 70 and 100. The globular domains of LBR are distinct from the chromodomain and chromo shadow domains in HP1-type proteins.
Identification of the Domain of LBR That Interacts with HP1 Hs␣ -We examined the interactions of various portions of the amino-terminal domain of LBR with HP1 Hs␣ in the yeast two-hybrid assay (Fig. 7). Full-length LBR interacted with full-length HP1 Hs␣ in this assay. The first 100 amino acids of LBR that contained the first globular domain detected by HCA plus the hinge region did not interact with HP1 Hs␣ . The portion of LBR from amino acids 97 to 208 that contained the second globular domain interacted with HP1 Hs␣ . LBR from amino acids 1 to 174 also interacted with HP1 Hs␣ , as did the stretch between amino acids 97 and 174, which contained only the first portion of the second globular domain. The region of LBR from amino acids 124 to 208 did not interact with HP1 Hs␣ . These results show that the first portion of the second globular domain identified by HCA in the amino-terminal domain of LBR is responsible for its binding to HP1 Hs␣ in the yeast two-hybrid assay.
To confirm the results obtained in the yeast two-hybrid assay, we examined the binding of domains of LBR synthesized by in vitro translation to a GST-HP1 Hs␣ fusion protein (Fig. 8). The 35 S-labeled LBR amino-terminal domain and several portions were synthesized by in vitro translation in reticulocyte lysates (Fig. 8A). The LBR amino-terminal domain was precipitated by the full-length GST-HP1 Hs␣ fusion protein (Fig. 8B,  lane 1), as was the portion from amino acids 97 to 208 that contained the second globular domain (Fig. 8B, lane 2). A polypeptide that contained the first globular domain plus the hinge region from amino acids 1 to 100 of LBR was not precipitated from reticulocyte lysates by the HP1 Hs␣ fusion protein (Fig. 8B, lane 3). LBR from amino acids 1 to 174 and from amino acids 97 to 174 (Fig. 8B, lanes 4 and 5) were also precipitated from lysates by the HP1 Hs␣ fusion protein, but the region of LBR between amino acids 124 and 208 was not pre- cipitated (Fig. 8B, lane 6). These findings are consistent with those obtained in the yeast two-hybrid assay and show that the first portion of the second globular domain identified by HCA in the nucleoplasmic domain of LBR mediates its binding to human HP1-type proteins.

Domain-specific Interactions of HP1-type Chromodomain
Proteins-We have used HCA to analyze the structures of human HP1-type proteins and have shown that the chromo shadow domain mediates the self-associations of these proteins as well as their binding to the inner nuclear membrane protein LBR. In both the yeast two-hybrid and in vitro binding assays, the chromo shadow domain was necessary and sufficient for the self-association of HP1 Hs␣ and for its binding to HP1 Hs␥ . In the yeast two-hybrid assay, the chromo shadow domain of HP1 Hs␣ was also necessary and sufficient for its interaction with LBR. However, in the in vitro binding assay, ϳ30 additional amino acids in the hinge region at the amino-terminal side were required for the chromo shadow domain to bind to LBR. The portion of HP1 Hs␣ containing the chromodomain and these same 30 amino acids of the hinge region did not bind to LBR in either the yeast two-hybrid or in vitro binding assay. These additional amino acids may have been necessary for the initial portion of the chromo shadow domain to achieve a proper conformation. In the yeast two-hybrid assay, the chromodomain also contained a portion of yeast GAL4 at its aminoterminal side.
The chromo superfamily of proteins is generally divided into several classes with distinct structural features (3,4). Members of all classes in this superfamily contain at least one classical chromodomain, which was first shown to be common to the Drosophila proteins HP1 and Polycomb (5). The HP1type proteins compose one class (Class A) of the chromo superfamily. Our finding that the HP1-type proteins contain a second globular domain that is similar to the classical chromodomain was also reported by others while this work was in progress (3,4). This second globular domain, called the chromo shadow domain, is not present in other classes of the chromo superfamily. The Polycomb-type proteins contain only one classical chromodomain, and some other superfamily members, such as CHD-1, contain two classical chromodomains. Proteins with one or two classical chromodomains and no chromo shadow domain can be considered as members of separate classes (Classes B and C, respectively). Using HCA, we have recently demonstrated that Tetrahymena Pdd1p contains three chromo-like domains and that it may represent a fourth class in the chromo superfamily. 2 The chromodomain and chromo shadow domain demonstrate their greatest sequence divergence in their carboxyl-terminal regions (3,4). 2 As aromatic amino acids play important roles in protein-protein interactions (34), differences in such residues FIG. 6. General organization of the nucleoplasmic amino-terminal domain of LBR as predicted from the analysis of its HCA plot. HCA detects two globular domains (boxed) separated by a hinge region that is highly charged between amino acids 70 and 100. The first putative transmembrane segment of LBR that follows the nucleoplasmic domain stretches from approximately amino acids 210 to 230 (shaded). Details on the presentation of the data and the symbols used can be found in the legend to Fig. 1. may explain the specific interactions of the chromo shadow domains with themselves and with proteins such as LBR (8) and the TIF transcriptional coactivators (17). In particular, Phe-167 and Tyr-168 are highly conserved in chromo shadow domains, but not in chromodomains of different proteins (3,4). 2 These residues may therefore be critical in the specificity of protein-protein interactions of the chromo shadow domain. On the other hand, Trp-41 of the chromodomain, which corresponds to Trp-142 of the chromo shadow domain, is conserved in almost all chromo-type domains and could be significant in their general interacting properties. The chromodomain, and various other domains in different members of the chromo superfamily of proteins, may mediate protein-protein (and possibly protein-nucleic acid) interactions different from those mediated by the chromo shadow domain. Hence, chromodomaincontaining proteins may function in determining overall chromatin organization by mediating a series of hierarchical modular interactions with many other proteins and possibly DNA.
Domain-specific Interactions of LBR-LBR was originally identified by its in vitro binding to B-type nuclear lamins (11), which has been confirmed in several subsequent studies (13,(35)(36)(37). The amino-terminal domain of LBR has also been shown to bind to double-stranded DNA in vitro (13) and more recently to HP1-type proteins (8). Using HCA, we have predicted that the amino-terminal domain of LBR can be divided into distinct subdomains. This suggests that distinct domains may mediate different protein-protein and possibly proteinnucleic acid interactions.
Indirect evidence suggests that the first 60 amino acids of LBR, which correspond to the first globular domain identified by HCA, may play an important role in its association with B-type lamins. LBR autoantibodies present in occasional patients with primary biliary cirrhosis have been shown to be anti-idiotypic to some B-type lamin autoantibodies, suggesting that they recognize a region of LBR involved in its binding to B-type lamins (35). We have mapped the minimal epitope of LBR recognized by these autoantibodies to the region of the protein between amino acids 1 and 60 (38), a region that corresponds to the first globular domain identified by HCA. Although the entire amino-terminal domain of LBR is necessary for binding to B-type lamins in vitro (13), the data obtained on the anti-idiotypic autoantibodies suggest that its first globular domain may be critical for this interaction.
We have also shown that the stretch of LBR from amino acids 70 to 100 can bind to double-stranded DNA in vitro. Sequence-specific DNA binding has not been identified, and the physiological significance of the in vitro protein-DNA interaction is not yet known. HCA shows that the DNA-binding domain of LBR is the hinge region of the amino-terminal domain. This region does not have predicted secondary structure, is highly basic, and contains many proline residues, features that would be expected in polypeptides that can bind to DNA.
In this study, we have shown that part of the second globular domain of LBR mediates its binding to HP1-type proteins. A threonine residue in this second globular domain of LBR is phosphorylated by p34 cdc2 -type protein kinase during mitosis (33). Although this residue is outside of the minimal binding domain, it is located within the globular domain, and phosphorylation may alter the structure of the entire domain, including the HP1-binding region. Hence, mitotic phosphorylation of LBR on a threonine residue in the second globular domain may disrupt its binding to HP1-type proteins at the start of mitosis when the inner nuclear membrane dissociates from chromatin. Dephosphorylation at the mitosis to G 1 interphase (33) may activate the binding capacity, allowing for the targeting of membrane vesicles to chromatin early in nuclear envelope reassembly (15,16).  35 S-labeled portions of LBR shown in A were synthesized in reticulocyte lysates by in vitro translation and incubated with GST-HP1 Hs␣ coupled to glutathione-Sepharose. After washing, the proteins that remained bound to glutathione-Sepharose were eluted with SDS, separated by SDS-PAGE, and detected by autoradiography. Approximately 10 -15% of the input material bound specifically to the proteins, which generated positive results. Migration of molecular mass standards (in kilodaltons) is indicated on the left.