Solution NMR Structures of Homeodomains from Human Proteins ALX4, ZHX1, and CASP8AP2 Contribute to the Structural Coverage of the Human Cancer Protein Interaction Network

Abstract

High-quality solution NMR structures of three homeodomains from human proteins ALX4, ZHX1 and CASP8AP2 were solved. These domains were chosen as targets of a biomedical theme project pursued by the Northeast Structural Genomics Consortium. This project focuses on increasing the structural coverage of human proteins associated with cancer.

Introduction

The human homeodomains comprising (i) residues 209–280 of “homeobox protein aristaless-like 4” (ALX4) (UniProtKB accession number: Q9H161), (ii) residues 462–532 (‘homeodomain 2’) of the “zinc fingers and homeoboxes protein 1” (ZHX1) (Q9UKY1), and (iii) the C-terminal domain comprising residues 1916–1982 of “caspase 8-associated protein 2” (CASP8AP2) (Q9UKL3) belong to the SCOP [1] ‘homeodomain-like’ superfamily SSF46689. ALX4(209–280) and ZHX1(462–532) belong to the very large Pfam [2] protein domain family Homeobox PF00046 which currently contains 25,115 members from eukaryotic organisms participating in 290 unique domain organizations (architectures). In contrast, CASP8AP2(1916–1982) has not yet been assigned to a Pfam family.

ALX4, a transcription activator, contains a single homeodomain and belongs to the ‘paired-class’ of homeodomain proteins which bind to palindromic DNA target sequences as homo- or heterodimers [3]. It plays an important role in craniofacial development [4, 5] so that loss of ALX4 expression results in both craniofacial and epidermal defects [6, 7]: mutant proteins R218Q [8], Q246Stop [8], and R272P [5] either completely or partially abolish DNA binding ability and transcriptional activation, which cause a rare genetic disorder, parietal foramina 2, manifested by abnormal skull bone development. In particular, the premature termination of transcription of mutant R265Stop [6] leads to ‘frontonasal dysplasia 2’ which is also associated with mental retardation [8]. Furthermore, hyper methylation of ALX4 is correlated with lung [9], bladder [10], gastric [11], and colorectal cancers [12–14], and reduced expression of ALX4 has been suggested as a biomarker for breast cancer[15].

ZHX1, likewise a transcriptional repressor [16, 17], contains five homeodomains and is ubiquitously expressed [16]. ZHX1 interacts with nuclear factor Y subunit A (NFYA) and DNA methyl transferase 3B (DNMT3B) for its repression activity [18, 19]. Changes in expression profiles of rat ZHX1 ortholog have been associated with glomerular disease [20, 21]. In addition to the five homeodomains, ZHX1, which also contains of two N terminal C2H2 zincfingers [22], forms homodimers via homeodomain 1 [23] and can also form heterodimers with ZHX3 [24]. Thus far, the solution NMR structure of a polypeptide segment containing the two zinc fingers (PDB:2GHF) [25] as well as X-ray structures of homeodomains 3 (2ECB) and 4 (3NAR) [26] were solved. ZHX1(462–532), the construct chosen for the present study, contains homeodomain 2 which shares very low sequence identity (< 20 %) with the other four homeodomains.

CASP8AP2, which may function as either a transcriptional activator or repressor, contains one homeodomain, and is involved in apoptosis, cell cycle progression through S-phase and activation of histone expression [27–29]. CASP8AP2 is a prognostic marker for acute lymphoblastic leukemia (ALL), as expression levels of CASP8AP2 are correlated to cells undergoing apoptosis in ALL cells [30, 31]. CASP8AP2 interacts with a nuclear protein, ataxia-telangiectasia locus (NPAT), an activator of histone transcription [32] and nuclear receptor coactivator 2 (NCOA2), an activator in glucocorticoid receptor activation [33]. CASP8AP2 C-terminal deletion mutants (involving residues 1403–1962 and 1916–1962 for NPAT; 1709–1982 for NCOA2) do not interact with NPAT and NCOA2 [32, 33].

ALX4(209–280), ZHX1(462–532) and CASP8AP2(1916–1982) were selected by the Northeast Structural Genomics (NESG) Consortium (http://www.nesg.org; target IDs HR4490C, HR7907F, and HR8150A, respectively) for structure determination with the aim to provide extensive structural coverage for human cancer-associated proteins and protein complexes constituting the ‘Human Cancer Protein Interaction Network’ (HCPIN) [34]. Here we report the high-quality solution NMR structures of ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982).

Materials and methods

ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982) were cloned, expressed, and purified following protocols [35–37] established by the NESG (see Supplementary Material and http://www.nmr2.buffalo.edu/nesg.wiki for details). The corresponding pET expression vectors [NESG HR4490C-209–280-NHT.2, HR7907F-462–532-AV6HT.2 and HR8150A-1916–1982-AV6HT.3] have been deposited in the PSI Materials Repository (http://psimr.asu.edu/). Protein samples for ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982), were prepared at 0.9. 0.4 and 0.8 mM concentrations, respectively, in 90% H₂O/10% D₂O containing 10 mM Tris-HCl, 100 mM NaCl, 10 mM DTT, 50 M DSS, 0.02% NaN₃ at pH 6.5 for ALX4 (209–280), or 20 mM MES, 100 mM NaCl, 10 mM DTT, 5 mM CaCl₂, 50 M DSS, 0.02% NaN₃ at pH 7.5 for ZHX1(462–532) and CASP8AP2(1916–1982), respectively. Additional [5% ¹³C; U-¹⁵N]-labeled samples enabled stereospecific assignments of the methyl groups of Val and Leu residues [38]. Isotropic overall rotational correlation times of ~7, ~ 6, and ~5 ns for ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982), respectively, were inferred from average ¹⁵N spin relaxation times (see Supplementary Material and http://www.nmr2.buffalo.edu/nesg.wiki), indicating that all three proteins are monomeric in solution. This finding was confirmed with analytical gel-filtration (Agilent Technologies) and static light scattering (Wyatt Technology Co.) (Figs. S1 – S3).

NMR data were acquired at 25 °C on Varian INOVA 600 or 750 MHz spectrometers equipped with cryogenic ¹H{¹³C,¹⁵N} probes. The total measurement times for ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982) were ~9, ~8, and ~7 days, respectively. Nearly complete sequence-specific ¹H, ¹⁵N and ¹³C resonance assignments (Table I; Figs. S4, S5, and S6) were obtained using G-matrix Fourier transform (GFT) triple resonance experiments for targets ALX4(209–280) and CASP8AP2(1916–1982), and with conventional triple-resonance NMR experiments for ZHX1(462–532) (Supplementary Material) using the automated resonance assignment program AutoAssign 2.3.0 [39, 40], followed by manual assignment of side-chain resonances. Chemical shifts, NOESY peak lists, and time domain NMR data were deposited in the BioMagResBank [accession numbers 18805, 18714, and 18352 for ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982), respectively].

Table I.

Statistics for NMR Structures of ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982)

	ALX4 (209–280)		ZHX1 (462–532)		CASP8AP2 (1916–1982)
Completeness of resonance assignmentsb
Backbone (%)	74.0		98.9		94.3
Side chain (%)	70.8		93.2		93.2
Aromatic (%)	100		94.4		78.3
Stereospecific methyl (%)	100		100		100
Conformationally-restricting constraintsc
Distance constraints
Total	841		875		892
intra-residue (i = j)	188		278		180
sequential (|i−j| = 1)	216		204		245
medium range (1 < |i − j| < 5)	232		217		271
long range (|i − j| ≥5)	205		176		196
Dihedral angle constraints	80		62		64
Hydrogen bond constraints	0		0		0
No. of long range constraints per residue	3.7		2.7		3.1
Residual constraint violationsc
Average no. of distance violations per structure:
0.1 – 0.2 Å	2.7		2.7		5.8
0.2 – 0.5 Å	0.6		0.7		1.8
> 0.5 Å	0		0		0
Average no. of dihedral angle violations per structure:
1 – 10°	0.7		1.6		3.9
> 10°	0		0		0
Model Qualityc
RMSD backbone atoms (Å)d	0.5		0.6		0.6
RMSD heavy atoms (Å)d	1.0		1.1		1.1
RMS deviation for bond lengths (Å)	0.022		0.022		0.022
RMS deviation for bond angles (°)	1.1		1.1		1.1
MolProbity Ramachandran statisticsc,d
Most favored regions (%)	99.3		99.3		94.9
Allowed regions (%)	0.7		0.7		5.1
Disallowed regions (%)	0.0		0.00		0.0
Global quality scores (Raw / Z-score)c
Verify3D	0.21	−4.01	0.16	−4.82	0.15	−4.98
ProsaII	0.65	0.00	0.34	−1.28	0.50	−0.62
Procheck (phi-psi)d	0.37	1.77	0.48	2.20	0.18	1.02
Procheck (all)d	0.28	1.66	0.36	2.13	0.12	0.71
MolProbity clash score	5.87	0.52	7.06	0.31	9.98	−0.19
RPF Scorese
Recall / Precision	0.987	0.894	0.989	0.888	0.930	0.867
F-measure / DP-score	0.938	0.804	0.936	0.773	0.897	0.750
BMRB accession number		18805		18714		18352
PDB ID		2M0C		2LY9		2LR8

^aStructural statistics computed for the ensemble of 20 deposited structures.

^bComputed using AVS software [56] from the expected number of resonances, excluding: highly exchangeable protons (N-terminal, Lys, and Arg amino groups, hydroxyls of Ser, Thr, Tyr), carboxyls of Asp and Glu, and non-protonated aromatic carbons. The comparatively low completeness of assignment for ALX4 is due to the missing H^N signals while aliphatic signals were detected at the flexible tails.

^cCalculated using PSVS 1.4 [57]. Average distance violations were calculated using the sum over r-⁶.

^dBased on ordered residue ranges [S(phi) + S(psi) > 1.8], Residues 222–269 for ALX4, 472–489 and 492–519 for ZHX1, and 1928–1977 for CASP8AP2. Three α helices for each protein: I (473–485), II (491–501), and III (505–518) for ZHX1(462–532), I (223–235), II (241–251), and III (255–269) for ALX4(209–280), and I (1931–1943), II (1948–1958), and III (1962–1977) for CASP8AP2(1916–1982).

^eRPF scores [41] reflecting the goodness-of-fit of the final ensemble of structures (including disordered residues) to the NOESY data and resonance assignments.

Structure calculations were performed using standardized methods of the NESG consortium [41, 42] (http://www.nmr2.buffalo.edu/nesg.wiki). The process included consensus analysis of automated NOESY cross peak assignments provided by the programs CYANA [43, 44] and AutoStructure 2.2.1 [45] based on ¹H-¹H NOE-derived upper limit distance constraints, and analysis of backbone dihedral angle constraints derived from chemical shifts using the program TALOS+ [46] for residues located in well-defined regular structure elements. Stereospecific assignments of methylene protons were performed with the GLOMSA module of CYANA and the final structure calculations were performed with CYANA followed by refinement of selected conformers in an “explicit water bath” using the program CNS [47]. Structure validation of the resulting 20 refined conformers was performed with the Protein Structure Validation Software (PSVS) server 1.4 [48] and the agreement of structures and NOESY peak lists was verified using the AutoStructure/RPF 2.2.1 package[41].

Results and discussion

High-quality NMR structures of ZHX1(462–532), ALX4(209–280), and CASP8AP2(1916–1982) were obtained (Fig. 1a, Fig. S7, Table I) and their coordinates were deposited in the Protein Data Bank (PDB) [49], respectively, on 09/14/2012 (accession code 2LY9), 10/24/2012 (2M0C), and 03/27/2012 (2LR8). The three structures exhibit the well-known [50] fold of homeodomains (Fig. 1a) consisting of three α-helices (Table I) and are quite similar: the root mean square deviations (RMSDs) calculated for the mean coordinates of the backbone heavy atoms N, C, and C’ of the three -helices are 0.9, 1.8, and 1.6 Å, respectively, for ALX4(209–280) and ZHX1(462–532), ALX4(209–280) and CASP8AP2(1916–1982), and ZHX1(462–532) and CASP8AP2(1916–1982).

Fig. 1.

a: Overlay of the ribbon representations of ALX4 (209–280) (green), ZHX1 (462–532) (red), and CASP8AP2 (1916–1982) (blue). b: Structure based sequence alignment of ALX4(209–280), Al(87–144), ZHX1(462–532) and CASP8AP2(1916–1982) Residues involved in protein-DNA interactions in the Al homeodomain DNA complex (PDB ID: 3LNQ) are indicated with the # symbol. c: Superposition of ALX4(209–280) (green) onto Al(87–144) (pink) bound to double-stranded DNA. Side chains involved in protein-DNA interactions are shown as stick representations. The corresponding interacting residues in ALX4 are also shown (except for those located in the termini). Mutations of ALX4 residues Q246 and R265 (shown in black) have been implicated in parietal foramina 2 and frontonasal dysplasia 2, respectively [8].

A structure based sequence alignment using the DALI server [51] reveals that ALX4(209–280) shares 26% sequence identity (over 61 aligned residues) and 19% sequence identity (over 58 residues) with ZHX1(462–532) and CASP8AP2(1916–1982), respectively. ZHX1(462–532) shares 11% with CASP8AP2(1916–1982) (over 56 residues) (Fig.1b). A search of the PDB for structurally similar proteins using the DALI server [51] yielded, as expected for structures of homeodomains [50], a very large number of highly significant hits with Z-scores > 4.0 [i.e. 244, 229, and 468 for ALX4(209–280), ZHX1(462–532), and CASP8AP2(1916–1982), respectively]. Remarkably, however, the highest scoring hit (Z-score 9.0) for ALX4(209–280) was a structure comprising residues 87 to 144 of the aristaless homeodomain protein [52] (Q06453), Al(87–144), from Drosophila melanogaster bound to DNA (PDB ID: 3LNQ: RMSD of 1.9 Å for the C atoms of 58 structurally aligned residues). The corresponding structure based sequence alignment reveals a high sequence identity of 78% for the two domains (Fig. 1b). Like ALX4, Al belongs to the ‘paired-class’ of homeodomain proteins which bind to palindromic DNA target sequences as homoor heterodimers [3]. Specifically, ALX4(209–280) binds to the palindromic repeat target sequence comprising 5'- TAAT-3' [53] preferably with three intervening nucleotides (5' –TAAT NNN ATTA-3'), while Al binds to a consensus DNA sequence of 5’-(T/C)TAATTAA(T/A)(T/A)G-3’[54]. Since the sequence alignment (Fig. 1b) shows that all residues interacting with the DNA duplex in the structure of the Al(87–144)-DNA complex [52] are conserved in ALX4(209–280), it appears quite likely that ALX4(209–280) binds to DNA in a very similar manner [52] (Fig. 1c). This finding suggests that this particular homeodomain-DNA interaction motif evolved before mammals evolved. Consistently, the sequence of human ALX4(209–280) is entirely conserved in all known mammalian genomes (see Supporting Information). Even though highly significant DALI hits were obtained also for ZHX1(462–532) [likewise Al(87–144) with Z-score = 8.8, RMSD – 1.5 Å, sequence identity 28%] , and CASP8AP2(1916–1982) [e.g., mouse 2610100B20RIK gene product homologous to Myb DNA binding protein; 1UG2; Z-score = 6.9; sequence identity: 43%], similar insights into the corresponding DNA interactions could not be derived because atomic resolution structures of DNA-complexes of homologues are not available.

Structural insights into Al(87–144)-DNA interactions (Fig. 1c) allow one to hypothesize that the ALX4 mutation R218Q [8] results in a malignant phenotype because the ALX4-DNA interaction is weakened and corresponding transcriptional activation impeded: the side chain of Arg 218 is expected to form hydrogen bonds with bases A and T in the minor groove which are removed in mutant protein. In contrast, mutant protein R272P [5] might result in a malignant phenotype because ALX4 homo- and/or heterodimer formation is affected. Arg 272 is located in the C-terminal flexible tail which is not expected to interact with DNA. However, this polypeptide segment plays a role for the formation of ALX4 dimer formation [3, 55] and the prolyl residue might restrict the conformational flexibility of this segment possibly required for protein-protein complex formation. Note that in the Q246Stop mutant the homeodomain is not formed, which abolishes DNA binding [8].

Abbreviations

ALX4

Aristaless-like 4

ZHX1

Zinc fingers and homeoboxes protein 1

CASP8AP2

Caspase 8 associated protein 2

Al

Aristaless

HCPIN

Human Cancer Pathway Interaction Network

DSS

4,4-dimethyl-4-silapentane-1-sulfonate sodium salt

DTT

Dithiothreitol

NESG

Northeast Structural Genomics Consortium

NOE

Nuclear Overhauser effect

PDB

Protein Data Bank

RMSD

Root mean square deviation