MACKAY & MATTHEWS LAB

Protein structure, function and engineering

Structure Gallery

Structure of a ZNF217-DNA complex

PDB file ] [ PubMed link ]

We have shown previously that a two-zinc finger unit found in the transcriptional coregulator ZNF217 recognizes DNA but with an affinity and specificity that is lower than other classical ZF arrays. To investigate the basis for these differences, we determined the structure of a ZNF217-DNA complex. We show that although the overall position of the ZFs on the DNA closely resembles that observed for other ZFs, the side-chain interaction pattern differs substantially from the canonical model. The structure also reveals the presence of two methyl-p interactions, each featuring a tyrosine contacting a thymine methyl group. To our knowledge, interactions of this type have not previously been described in classical ZF-DNA complexes. We speculate that relatively low affinity/specificity interactions of this type might be important for gene regulation.

The mechanism of H3K9Me3 recognition by CHD4

PDB file ] [ PubMed link ]

The Nucleosome Remodeling and Deacetylase (NuRD) complex is essential for the normal regulation of gene expression in a wide range of organisms (even plants!). We are slowly trying to build up a picture of how NuRD works at a mechanistic level. One of the NuRD components, CHD4, contains two PHD-type zinc-finger domains. In collaboration with Tanya Kutateladze from the University of Colorado at Denver, we have shown that one of these two PHD domains can recognize the N-terminal tail of histone H3 when specifically modified by trimethylation of K9 (a repressive mark). We have determined the structure of the PHD:H3 tail complex, showing how the methylated K9 is recognized (on the right of the pic).

The first PHD domain of CHD4

PDB file ] [ PubMed link ]

The Nucleosome Remodeling and Deacetylase (NuRD) complex is essential for the normal regulation of gene expression in a wide range of organisms (even plants!). We are slowly trying to build up a picture of how NuRD works at a mechanistic level. One of the NuRD components, CHD4, contains two PHD-type zinc-finger domains. This picture shows the structure of the first PHD domains, which we have shown in collaboration with Tanya Kutateladze from the University of Colorado at Denver can recognize the N-terminal tail of histone H3.

The FOG1-RbAp48 complex

PDB file ] [ PubMed link ]

FOG1 is a coregulator protein that assists the transcription factor GATA1 in the control of gene expression during erythroid development. Gerd Blobel, a collaborator of ours in Philadelphia, showed that at least part of the mechanism by which FOG1 regulates GATA1 expression is through its ability to recruit the multiprotein NuRD (Nucleosome Remodeling and Deacetylase) complex. We have determined the structure of a complex formed between the NuRD-binding region of FOG1 (residues 1-15) and the RbAp48 component of the NuRD complex. RbAp48 forms a beta-propellor and a basic sequence in the FOG1 peptide docks into a cavity at one end of the propellor.

A new class of ssRNA-binding domain

PDB file ] [ PubMed link ]

The two zinc fingers of ZRANB2 (formerly known as ZNF265) can bind to single-stranded RNA with high sequence specificity. We have determined the structure of one of these fingers bound to its RNA target site (AGGUAA – determined by SELEX). The structure reveals a new class of RNA-binding domain. This ZnF forms a unique guanine-Trp-guanine aromatic stack, and the core nucleotides (GGU) are recognized by an extensive network of protein side-chain hydrogen bonds. Also notable are the two-headed hydrogen bonds that are formed between arginine side-chains and the two guanines – these interactions appear to provide strong selection for guanine in those positions.

Data-driven model of the MED-1:DNA complex

PDB file ] [ PubMed link ]

MED-1 is a GATA-family transcription factor that is essential for development in Caenorhabditis elegans. It’s single GATA-type zinc finger was shown by Morris Maduro to bind to a long and slightly divergent DNA site (GTATACTTTT), compared to the well-known consensus for mammalian GATA-family zinc fingers (AGATAA). We used a combination of NMR spectroscopy, gel shifts and mutagenesis to demonstrate that the MED-1 zinc finger forms an additional C-terminal helix, which is induced only upon binding to DNA. This helix is inserted into the major groove of DNA to bind to the 3′ end of the recognition sequence. These data demonstrate the difficulties in predicting the structures of protein complexes from sequence data alone – the additional helix was not predicted by secondary structure prediction algorithms.

Zinc finger 2 of ZNF265/ZRANB2

PDB file ]

The two zinc fingers of ZRANB2 (formerly known as ZNF265) can bind to single-stranded RNA with high sequence specificity. In addition to determining the structures of these two domains, we have used a combination of chemical shift mapping and mutagenesis to define the RNA-binding surface of these domains. Other zinc fingers in the same structural class (RanBP2-type zinc fingers) have been shown to mediate protein-protein interactions, another reminder of the versatility of small zinc-binding domains as recognition motifs.

A circular protein complex

PDB file ] [ PubMed link ]

We used intein technology to generate a circular protein complex between LMO4 and the LIM-binding domain of Ldb1 (Ldb1-LID). The proteins, which bind in a head-to-tail fashion are joined by a flexible linker at each end. The circular complex is more stable than a tethered complex where the C-terminus of LMO4 is linked to the N-terminus of Ldb1-LID, however, the crystal structure of this complex shows that the structures (in blue and gold), are identical. Curiously, when the order of a single tethering step is reversed (Ldb1-LID-to-LMO4) the resulting protein is just as stable as the circular complex.

The crystal structure of a complex between two LIM-homeodomain proteins

PDB file ] [ PubMed link ]

In developing motor neurons Isl1 displaces Lhx3 as the binding partner of the essential LIM-HD cofactor protein Ldb1. Isl1 provides Lhx3 with a decoy binding domain from a region in the C-terminus of Isl1 that, despite very low sequence identity, binds Lhx3 is essentially the same way as Ldb1. The LIM domains from Lhx3 are shown as a grey surface. The Lhx3-binding domain from Isl1 is shown in green and the LIM-interaction domain from Ldb1 from the Lhx3-Ldb1 structure above is shown in gold.

The structure of an Lhx3-Ldb1 complex

PDB file ] [ PubMed link ]

In developing V2 interneurons (and various other tissues) the LIM domains of Lhx3 must contact the LIM-interaction domain of Ldb1 as part of a transcription complex that specifies the cell type. This is how they do it. The Lhx3 is in blue and Ldb1-LID is in gold.

Model of the complex between zinc finger 5 of MyT1 and DNA

PDB file ] [ PubMed link ]

We have used a variety of NMR, mutagenesis and binding data to build a model of the interaction between zinc finger 5 from the neuronal transcription factor MyT1 and its cognate DNA site. The model was built using the HADDOCK data-driven docking software (from Alexandre Bonvin) and shows that the domain fits snugly into the DNA major groove, making a number of base specific hydrogen bonds and hydrophobic contacts.

Structure of the fifth zinc finger of MyT1

PDB file ] [ PubMed link ]

MyT1 is a zinc finger transcription factor that is involved in neuronal development, controlling genes that are important for myelin sheath formation. It contains 7 zinc fingers with an unusual consensus sequence, and these domains have been shown previously to be responsible for the DNA-binding properties of MyT1. We have determined the solution structure of one of these zinc fingers, as a preliminary step towards understanding how these domains recognize DNA. The structure is different from all other known classes of zinc fingers, and contains no elements of regular secondary structure. We have gone on to examine the binding of this domain to DNA (see below!).

The THAP domain of C. elegans CtBP

PDB file ] [ PubMed link ]

The THAP (Thanatos-associated protein) domain is a recently discovered zinc-binding domain found in proteins involved in transcriptional regulation, cell-cycle control, apoptosis and chromatin modification. It contains a single zinc atom ligated by cysteine and histidine residues within a Cys-X(2-4)-Cys-X(35-53)-Cys-X(2)-His consensus. We determined the NMR solution structure of the THAP domain from Caenorhabditis elegans C-terminal binding protein (CtBP) and show that it adopts a fold containing a treble clef motif, with some similarity to the zinc finger-associated domain (ZAD) from Drosophila Grauzone. We have also shown using gel-shift data that CtBP-THAP is able to bind DNA. Other THAP domains have been reported to be involved in mediating protein interactions, suggesting that THAP domains might exhibit a functional diversity similar to that observed for classical and GATA-type zinc fingers.

HOP – a corepressor comprising a single homeodomain

PDB file ] [ PubMed link ]

Homeodomain-only protein (HOP) is an 8-kDa transcriptional corepressor that is essential for the normal development of the mammalian heart. A combination of sequence comparison and our structural data revealed that HOP consists entirely of a homeodomain, and it is the only human protein to have this topology. We have also shown that, unlike other classic homeodomain proteins, HOP does not appear to interact with DNA, and it appears that it instead functions as a bridge in the formation of HDAC-type repressive complexes on DNA. However, the mechanism by which this repression occurs is still only partially resolved. Our results demonstrate that the homeodomain fold has been co-opted during evolution for functions other than sequence-specific DNA binding.

p22HBP – a new heme-binding protein in red blood cells

PDB file ] [ PubMed link ]

p22HBP is a 22-kDa mammalian protein that is highly upregulated during erythroid development, and appears to be a target gene of GATA-1. Its function is currently unknown, although it has been reported to bind to a range of different porphyrins, suggesting a role in heme biosynthesis. We have determined the structure of p22HBP and used HSQC titration data to map its porphyrin binding site. Interestingly, our structure reveals that p22HBP has structural (but not sequence) homology to a bacterial multi-drug resistance protein BmrR that functions by binding to a variety of small hydrophobic drug molecules.

EAS: a fungal hydrophobin

PDB file ] [ PubMed link ]

Hydrophobins are small fungal proteins that form polymeric fibrils, known as rodlets. These rodlets form a matted coating on the surface of aerial structures like air-dispersed spores. The coating is extremely amphipathic, with an outward facing hydrophobic surface that “water-proofs” the spores. The coating is also extremely robust and is being considered for material science applications. Our structure of the monomeric hydrophobin from Neurospora crassa shows that it is made up entirely of beta structure (top right), forming a 4-stranded beta-barrel. The structure contains two very flexible loops (seen top left), which may play a role in fibril formation. The electrostatic surface properties of the EAS structure are also consistent with the amphipathic nature of the rodlets (bottom), and have allowed us to create a model for fibril formation (see below).

Designed-for-function CHANCE 7 (DFF7)

PDB file ] [ PubMed link ]

This CHANCE peptide was one of two (+DFF5) designed to mimic the GATA-binding surface of the transcriptional repressor U-shaped. NMR structural work showed both that DFF7 was well-folded (right) and that the grafted residues occupied positions that were similar (although not identical) to those found in the native USF1 protein (left). Both DFF5 and DFF7 displayed measureable binding to GATA-1, as judged by NMR titration experiments, although the nature of the binding may not be as specific as intended.

Designed-for-function CHANCE 5 (DFF5)

PDB file ] [ PubMed link ]

This CHANCE peptide was designed to mimic the GATA-binding surface of the transcriptional repressor U-shaped. We have previously defined the surface involved in this interaction, and we attempted to transplant this surface onto the minimal CHANCE domain. NMR structural work showed both that DFF5 was well-folded and that the grafted residues occupied positions that were similar (although not identical) to those found in the native USF1 protein. In the picture, the grafted residues are shown overlayed with the same residues from USF1.

Designed-for-function CHANCE 2 (DFF2)

PDB file ] [ PubMed link ]

This is one of the first CHANCE peptides that was designed to emulate the binding function of another protein. The N-terminal nucleocapsid domain (NUC1, left) of the HIV-1 nucleocapsid protein is able to bind to a short oligonucleotide from the virus’ genome. We identified the binding residues from NUC1 and ‘grafted’ them onto our CHANCE domain, in an arrangement that appeared to mimic their spatial orientation in the NUC1 protein. The structure of DFF2 shows that the grafted residues do indeed form a surface that bears a reasonable resemblance to the template protein. Unfortunately, NMR titration data indicated that DFF2 was not actually able to specifically recognize the NUC1 target oligonucleotide. Back to the drawing board!

Stripped down CHANCE domain #2

PDB file ] [ PubMed link ]

This is a second minimalised version of the CHANCE domain. This time, 13 of the 25 residues have been mutated (mostly to alanine), and the fold is retained. These two stripped down domains retain only the zinc-binding residues and the residues that make significant numbers of contacts.

Stripped down CHANCE domain #1

PDB file ] [ PubMed link ]

This is a minimalised version of the CHANCE domain. 12 of the 25 residues have been mutated (mostly to alanine), but the fold is retained. This ‘stripping down’ of the surface of the domain is the first step in re-designing the domain to have new binding functionality. This structure contains ~50% Ala and has a very similar fold to the wild type sequence, indicating that zinc-ligands and a few key hydrophobic residues are all that is required for structure.