(Received for publication, March 17, 1997, and in revised form, April 1, 1997)
From the Departments of Medicine and of Anatomy and
Cell Biology, College of Physicians and Surgeons, Columbia University,
New York, New York 10032, the ¶ Systèmes Moléculaires
and Biologie Structurale, Laboratoire de Mineralogie-Cristallographie
Paris, (LMCP), CNRS URA09, Université Paris 6 and 7, 4 Place
Jussieu, 75252 Paris Cedex 05, France, and the
Departement de
Biologie Cellulaire, Institut Jacques Monod, CNRS, Université
Paris 7, 2 Place Jussieu, 75251 Paris Cedex 05, France
HP1-type chromodomain proteins self-associate as
well as interact with the inner nuclear membrane protein LBR
(lamin B receptor) and
transcriptional coactivators TIF1 and TIF1
. 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. Hydrophobic 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 second globular domain is necessary for binding. The
modular domain organization of HP1-type proteins and LBR can explain
some of the diverse protein-protein interactions at the
chromatin-lamina-membrane interface of the nuclear envelope.
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 down-regulation 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 HP1Hs, HP1Hs
, and HP1Hs
,
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
~210 amino acids and a hydrophobic domain with eight putative
transmembrane segments (11-14). 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 self-associate (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.
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. 22-27).
For sequence alignments, the accuracy was assessed by computing identity, similarity, and HCA scores as well as the corresponding Z-scores (27). Z-scores represent the differences between the alignment score under consideration and the mean score of a distribution computed 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 ConstructionComplementary 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
5-ends were used with clones for human LBR (13) and
HP1Hs
(8) as templates. Amplified cDNAs were
purified and cloned into the plasmids of choice by standard
methods.
As our previously characterized cDNA clone for HP1Hs
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 3
-sequence
of the partial HP1Hs
cDNA clone (GenBankTM accession
number U26312[GenBank]) and a putative 5
-sequence in GenBankTM (accession
number Z15820[GenBank]) were used in the reaction. The amplified cDNA was
cloned into plasmid pACT2 (CLONTECH) and sequenced (the entry with
GenBankTM accession number U26313[GenBank] has been updated to reflect the complete cDNA sequence of HP1Hs
). This full-length
HP1Hs
cDNA clone was used as a template for the
polymerase chain reaction to generate HP1Hs
cDNAs
for cloning as described above.
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 [35S]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 µl of in vitro translation lysate were added to 200 µl 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 µl of glutathione-Sepharose (Pharmacia Biotech Inc.) coupled to 3-5 µg of a glutathione S-transferase (GST) fusion protein. Plasmids that expressed GST fusion proteins were constructed by cloning 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 ~2 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 AssayTo 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).
We have used HCA to
analyze the sequences of human HP1Hs,
HP1Hs
, and HP1Hs
. This method is highly
effective in detecting three-dimensional 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 HP1Hs
revealed two globular domains
separated by a hinge region of ~70 amino acids (Fig.
1). Similar results were obtained for
HP1Hs
and HP1Hs
(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 HP1Hs
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 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 HP1Hs
. Similar values were obtained for
HP1Hs
and HP1Hs
. 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 HP1Hs and
HP1Hs
in an in vitro solution binding assay
(Fig. 2). 35S-Labeled HP1Hs
produced by in vitro translation in reticulocyte lysates was precipitated by the GST-HP1Hs
and
GST-HP1Hs
fusion proteins, but not by GST (Fig. 2,
lanes 1-4). Similarly, 35S-labeled
HP1Hs
was precipitated by the GST-HP1 fusion proteins,
but not by GST (Fig. 2, lanes 5-8). These results show that
HP1Hs
and HP1Hs
can self-associate and
associate with each other in vitro.
To identify the domain of HP1-type proteins responsible for these
associations, we expressed various portions of 35S-labeled
HP1Hs by in vitro translation (Fig.
3A). GST-HP1Hs
and
GST-HP1Hs
fusion proteins precipitated full-length
HP1Hs
(Fig. 3B, lanes 1 and
5) from reticulocyte lysates, but not the portion of
HP1Hs
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 HP1Hs
(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
HP1Hs
and its binding to HP1Hs
. These
results are consistent with those previously reported using the yeast
two-hybrid assay, which showed that the chromo shadow domain of the
mouse homologue of HP1Hs
interacted with mouse
HP1Hs
and a portion of MOD1, the mouse homologue of
HP1Hs
(17).
The Chromo Shadow Domain of HP1Hs
We used the yeast two-hybrid assay to identify the
domain of HP1Hs that binds to the nucleoplasmic
amino-terminal domain of LBR (Fig. 4). Full-length
HP1Hs
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
HP1Hs
that contained the chromodomain plus the hinge
region nor the hinge region alone interacted with the amino-terminal
domain of LBR. The chromo shadow domain was therefore necessary and
sufficient for the interaction of HP1Hs
with LBR in the
yeast two-hybrid assay.
We also examined the binding of various domains of
HP1Hs, synthesized by in vitro translation,
to a GST fusion protein of the amino-terminal domain of LBR (Fig.
5). 35S-Labeled full-length
HP1Hs
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. Full-length HP1Hs
was precipitated by the GST-LBR fusion protein (Fig. 5, lane 1), as was the protein that contained the chromo shadow domain plus the hinge region (lane 3). Deletion of the chromo
shadow domain from HP1Hs
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
HP1Hs
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 ~40 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 HP1Hs
We examined the interactions of various
portions of the amino-terminal domain of LBR with HP1Hs
in the yeast two-hybrid assay (Fig. 7). Full-length LBR
interacted with full-length HP1Hs
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
HP1Hs
. The portion of LBR from amino acids 97 to 208 that contained the second globular domain interacted with
HP1Hs
. LBR from amino acids 1 to 174 also interacted
with HP1Hs
, 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 HP1Hs
. 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 HP1Hs
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-HP1Hs fusion protein
(Fig. 8). The 35S-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-HP1Hs
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 HP1Hs
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
HP1Hs
fusion protein, but the region of LBR between
amino acids 124 and 208 was not precipitated (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.
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
HP1Hs and for its binding to HP1Hs
. In
the yeast two-hybrid assay, the chromo shadow domain of
HP1Hs
was also necessary and sufficient for its
interaction with LBR. However, in the in vitro binding
assay, ~30 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 HP1Hs
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 amino-terminal 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 HP1-type 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 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, chromodomain-containing 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 LBRLBR was originally identified by its in vitro binding to B-type nuclear lamins (11), which has been confirmed in several subsequent studies (13, 35-37). 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 protein-nucleic 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 p34cdc2-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 G1 interphase (33) may activate the binding capacity, allowing for the targeting of membrane vesicles to chromatin early in nuclear envelope reassembly (15, 16).
We thank P. Hossenlopp for invaluable suggestions; J.-P. Mornon for discussions on protein structure analysis; and S. J. Elledge, S. Fields, and J. Licht for reagents.