LIM-only protein 2, Lmo2, is a regulatory protein that is essential for hematopoietic development and inappropriate overexpression of Lmo2 in T-cells contributes to T-cell leukemia. It exerts its functions by mediating protein–protein interactions and nucleating multicomponent transcriptional complexes. Lmo2 interacts with LIM domain binding protein 1 (Ldb1) through the tandem LIM domains of Lmo2 and the LIM interaction domain (LID) of Ldb1. Here, we present the solution structure of the LIM2 domain of Lmo2 bound to Ldb1LID. The ordered regions of Ldb1 in this complex correspond well with binding hotspots previously defined by mutagenic studies. Comparisons of this Lmo2LIM2–Ldb1LID structure with previously determined structures of the Lmo2/Ldb1LID complexes lead to the conclusion that modular binding of tandem LIM domains in Lmo2 to tandem linear motifs in Ldb1 is accompanied by several disorder-to-order transitions and/or conformational changes in both proteins.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
LIM-only protein 2 (Lmo2) is a transcriptional regulator that is essential for normal blood cell development. Mice in which this gene has been disrupted die midway through development (E10.5) without any blood cells.1 Lmo2 is associated with the maintenance of hematopoietic stem cells,2 the differentiation of red blood cells,3, 4 and angiogenesis in both normal development and solid tumor formation.5, 6 The inappropriate overexpression of Lmo2 in T-cells, through either chromosomal translocations or retroviral insertion of gene therapy vectors, leads to the onset of T-cell acute lymphoblastic leukemia (reviewed in Ref. 7). Tumors with similar properties occur in mice that carry an lmo2 transgene.8 Inhibition of Lmo2, using a specific intracellular antibody or peptide targeted toward the protein, blocks both Lmo2-induced erythropoiesis and tumor formation in a mouse explant model of leukemia.9, 10
Lmo2 is a 158-residue protein that contains two closely spaced LIM domains and very little other sequence. LIM domains (named for the first three genes in which the motif was identified (Lin-11, Isl-1, and Mec-3) are zinc fingers that coordinate two zinc ions, and that function as protein–protein interaction motifs (reviewed in Ref. 11). Lmo2 has been found in large transcriptional complexes alongside proteins such as LIM domain binding protein 1 (Ldb1/NLI/CLIM2), Tal1/Scl, E2A proteins, and GATA-family proteins.3, 12–14 Of these partner proteins, Lmo2 is found most frequently in association with Ldb1, a widely expressed protein that has been shown to contact, through its LIM-interaction domain (LID), the LIM domains of many LIM-only and LIM-homeodomain proteins.15–17 Ldb1 has essential roles in numerous developmental processes (reviewed in Ref. 18). Homozygous Ldb1 knockout mice die mid gestation (E9.5–E10) with numerous defects, including a lack of blood cells; this phenomenon is thought to indicate an essential role for Ldb1 in Lmo2-containing transcriptional complexes in the yolk sac.19 Although the LIM1 domain from Lmo2 appears to be the primary determinant of Ldb1-binding activity, both LIM domains are required for high affinity binding.16, 20, 21
Recombinant forms of Lmo2 tend to be insoluble, but it is possible to overcome this problem by creating fusion proteins in which either or both of the LIM domains are tethered to Ldb1LID via a flexible linker.20, 21 Using this approach, we have previously determined the solution structure of Lmo2LIM1 in complex with Ldb1LID (Lmo2LIM1–Ldb1LID)22; the crystal structure of the tandem LIM domain construct (Lmo2LIM1+2–Ldb1LID) has also been determined using a similar strategy.23 Here we report the solution structure of Lmo2LIM2–Ldb1LID and compare this structure and the solution structure of Lmo2LIM1–Ldb1LID with the crystal structures of Lmo2LIM1+2–Ldb1LID to gain a better understanding of how modular binding between the tandem LIM domains of Lmo2 and tandem linear motifs in Ldb1 facilitate complex formation.
Results and Discussion
We determined the solution structure of Lmo2LIM2–Ldb1LID using standard NMR methods (Table I; Fig. 1). The structured regions of the complex, as defined by angle order parameters of ϕ > 0.9 and ψ > 0.9, are made up of residues Lmo2L93–G156 and Ldb1D300–L309, with the exceptions of Lmo2A103–M106 and Lmo2C97/D98. Residues Ldb1E317–L321 are also ordered, but do not form part of the Lmo2LIM2/Ldb1LID complex. The root mean square deviation (R.M.S.D) of atoms in the structured regions of the 20 lowest energy models is 0.85 Å for all backbone atoms and 1.28 Å for all heavy atoms. Together these parameters indicate that the solution structure of Lmo2LIM2–Ldb1LID is of medium to high resolution.24
Table I. NMR and Refinement Statistics for Lmo2LIM2–Ldb1LID
R.M.S.D. was calculated among 20 refined structures using the structured regions of Lmo2 (93–102/107–155) and Ldb1 (300–309). The isolated helical region in Ldb1 (residues 317–321) gave rise to R.M.S.D. values of 1.10 ± 0.26 and 0.24 ± 0.19 Å for all heavy atoms and backbone atoms, respectively.
For the structured residues in Lmo2 (residues 93–102/107–155) and Ldb1 (residues 300–309/317–321).
Note that residues 1–72 in the construct correspond to Lmo2 (residues 83–155), and residues 84–123 in the construct correspond to Ldb1 (residues 300–339).
The structure of this tethered complex closely resembles that of other LIM–Ldb1LID complexes,22, 23, 25–27 [e.g. Fig. 2(A)]. Ldb1LID binds in a head-to-tail fashion (i.e., with the N-terminus of Lmo2LIM2 close to the C-terminus of the structured region of Ldb1LID and the C-terminus of Lmo2LIM2 to the N-terminus of Ldb1LID). The LIM domain fold comprises two closely packed GATA-like zinc-binding modules (referred throughout as Zn1 and Zn2), each of which contains two orthogonally arrayed β-hairpins followed by a segment of helical structure of variable length.11, 28 For Lmo2LIM2–Ldb1LID, only a few regions of regular secondary structure were well defined in all members of the NMR ensemble: the final β-hairpin (Lmo2R134–L137/I142–V143) and the C-terminal α-helix (Lmo2I148–N155). The N-terminal part of Ldb1LID (Ldb1D300–L309) packs against Lmo2LIM2. A short β-strand in Ldb1 (Ldb1M302–V303) augments the well-defined β-hairpin in Lmo2LIM2 (Lmo2R134–L137/I142–V143) to form a small three-stranded anti-parallel β-sheet [Fig. 2(A)]. On average, ∼1500 Å2 solvent accessible surface is buried through the interaction of Lmo2LIM2 and Ldb1LID.
Both angle order parameters and a low abundance of non-sequential NOEs indicate that, overall, Ldb1322–339 is not well ordered. This conclusion is supported by both TALOS+ predictions and the Random Coil Index prediction of protein flexibility, which is based on an analysis of backbone chemical shifts29 (Supporting Information data). Within this region, however, residues Ldb1E317–L321 are ordered, containing a single turn of helix (Ldb1D318–L321) [Fig. 2(B)]; order angle parameters >0.9, TALOS+, and random coil index (RCI) analysis all support structure in this region and chemical shift index analysis and the presence of non-sequential NOEs indicate helical structure (Supporting Information data). In related complexes that contain the LIM1 domains from Lmo2, Lmo4, or Lhx3 in complex with Ldb1, these same residues form extended or β structure22,23,25–27 [e.g., Fig. 2(A,C), orange boxes], indicating that for these residues in Ldb1 a conformational change is brought about through binding to the LIM1 domains of Lmo2 and related proteins.
Isothermal titration calorimetry and competition ELISA data showed that Ldb1LID binds Lmo2LIM1 with KD ≈ 2 × 10−7M−1, and Lmo2LIM1+2 with KD ≈ 2 × 10−8M−1, respectively.21 That is, the binding affinity increases by ∼10-fold when the LIM2 domain is present. The availability of structures of both the individual LIM-LID modules (Lmo2LIM1–Ldb1LID and Lmo2LIM2–Ldb1LID, both determined by NMR spectroscopy), and the double LIM–LID complex (Lmo2LIM1+2–Ldb1LID) gives us an opportunity to consider how the tandem binding domains from both proteins synergistically facilitate interactions in this complex. In the structure of Lmo2LIM1–Ldb1LID (PDB accession code 12JO) Zn1 is well defined (backbone R.M.S.D. of 0.39 Å for the 20 conformers in the ensemble), but substantially more variation exists between conformers in Zn2 (backbone R.M.S.D. 1.24 Å).22 Lmo2LIM1-Zn1 and the relevant binding residues in Ldb1LID are essentially identical in both the single and double LIM constructs. Lmo2LIM1-Zn2 has the same topology in all structures, but whereas this module is poorly defined in the NMR structure of the LIM1 construct, it is well defined in the crystal structures of the double LIM construct; the six different molecules in the asymmetric units of the two X-ray structures (PDB accession codes 2XJY and 2XJZ) are identical in this region.
The difference in structure of Lmo2LIM1-Zn2 between the double LIM constructs in the X-ray structures and the single LIM construct in our earlier NMR structure likely stems from a difference in construct design. The LIM1 domain was truncated at L87 in the Lmo2LIM1–Ldb1LID construct, but examination of the crystal structures shows that the sidechain of Lmo2F88 [Fig. 2(A)], the final helical residue in Lmo2LIM1, makes contacts with numerous other residues in the helix and the adjacent β-hairpin. This phenylalanine residue is highly conserved in all other LIM-only and LIM-homeodomain proteins, and adopts the same conformation in all of the known structures from this group of LIM domain proteins (Lmo4, Lhx3, and Lhx4).22,25–27,30 Thus, the lack of Lmo2F88 likely destabilized Lmo2LIM1-Zn2 in the NMR structure. The solution structure of an Lmo4LIM1–Ldb1LID construct (PDB accession code 1M3V) that does contain Lmo4F81 (which is equivalent to Lmo2F88) is considerably more ordered than the Lmo2LIM1–Ldb1LID structure,22 supporting an important structural role for this aromatic sidechain.
However, further comparison with Lmo4LIM1–Ldb1 and Lmo4LIM1+2–Ldb1 structures suggests that, despite the effect on structure, the lack of F88 does not have a major impact on the interaction between Ldb1 and Lmo2 in the Lmo2LIM1–Ldb1 complex. Similar levels of order were seen in the Ldb1 portions of both LmoLIM1–Ldb1 constructs (residues Ldb1I322–Q328 in the Lmo2 and Ldb1L321–E326 in the Lmo4 complex, are ordered). These ordered regions from the single domain NMR structures correspond well with the results of mutagenic studies that indicated a single binding hotspot in Ldb1LID, centered on Ldb1I322, that drives binding to the LIM domains of Lmo2, Lmo4, Lhx3, and Isl1.21, 25, 27
The only significant difference in the structures of the LIM2 domain constructs lies at start of this domain. Residues Lmo2G92/L93 form a well defined short β-strand in the double LIM domain X-ray structures, but are disordered at the start of the single LIM domain NMR structure [Fig 2(A), blue box]; Lmo2A20–M23 is also poorly ordered in the NMR structure, Lmo2 (Fig 1, cyan).
Comparing the extent of interaction of Ldb1 in the different LIM interactions, the ordered region of Ldb1LID is slightly longer in the Lmo2LIM2–Ldb1LID structures compared with the Lmo2LIM1–Ldb1LID and Lmo4LIM1–Ldb1LID structures. Mutagenic studies indicate two binding hotspots that form an extended interface in Ldb1LID for the LIM2 domains from Lmo2, Lmo4, and Lhx3.21, 25, 27 For Lmo2LIM1+2/Ldb1 interactions, the number of residues from Ldb1 for which electron density is defined varies between each of the six molecules from the X-ray structures, but apart from a minor hinging (∼11°) at Ldb1E313,23 Ldb1 adopts the same structure in each conformer [Fig. 2(C)]. Thus, Ldb1M310–L321, which is disordered in the single LIM domain complexes, becomes ordered when both LIM domains are present. At this stage no measurements of binding affinity exist for the isolated Lmo2LIM2–Ldb1 interaction, but similar solvent accessible surface areas buried by Lmo2LIM1 and Lmo2LIM2 by interaction with Ldb1 in the NMR and X-ray structures suggests a much higher binding affinity than the 10-fold increase observed for the tandem LIM domains over the isolated LIM1 domain. It seems likely that the various disorder-to-order transitions associated with these different complexes cause the binding of the tandem LIM domains to be anti-cooperative (less than the sum of the free energies of the individual interactions). We and others have observed similar effects in the interactions of other modular binding interactions, including fibronectin-binding31 and single-stranded RNA-binding systems.32, 33
In summary, comparison of the structures of several LIM–Ldb1 complexes indicates that the modular binding between Ldb1 and Lmo2 and related proteins is accompanied by a series of disorder-to-order transitions, particularly in Ldb1. These transitions reduce the magnification of binding affinity that might otherwise be expected in the formation of a multi-domain complex. It was also observed that for the Ldb1–Lmo2 interaction, solution structures for each of single LIM–LID constructs provided a good indication of binding hotspots.
Materials and Methods
Expression and purification
Lmo2 refers to the mouse protein (UniProtKB/Swiss-Prot P25801) and Ldb1 refers to the mouse protein (P70662-3). Lmo2LIM2–Ldb1LID comprises residues 84–155 from Lmo2 followed by a GGSGGSGGSGG linker and residues 300–339 from Ldb1; this construct was generated as described previously.21 The protein was expressed with an N-terminal glutathione-S-transferase (GST) fusion protein in Escherichia coli BL21(DE3) cells and purified by glutathione affinity chromatography, treated with thrombin to remove the GST tag and finally subjected to size-exclusion chromatography as described previously.21–23, 34
Protein samples (0.3–0.8 mM) were prepared in 20 mM 2-(N-morpholino)ethanesulfonic acid, 100 mM NaCl, 1 mM dithiothreitol at pH 6.5 in 90% H2O and 10% D2O or 100% D2O where appropriate. All NMR spectra were recorded at 298 K on Bruker AvanceIII 600 and 800 MHz spectrometers, both equipped with 5-mm triple resonance TCI cryoprobes. Sequential assignments of the backbone and sidechain resonances were reported previously.34 Approximate interproton distances were obtained from 1H-NOESY, 15N-edited nuclear overhauser effect spectroscopy (NOESY)-heteronuclear single quantum coherence (HSQC), and 13C-edited NOESY-HSQC spectra. An HNHA spectrum was used for the calculation of ϕ angles based on the method of Bax and Grzesiek.35 All spectra were processed in Topspin (Bruker Biospin) and were analyzed using Sparky (T. D. Goddard and D. G. Kneller, University of California at San Francisco).
Nuclear overhauser effect (NOE) peaks from 2D and 3D NOESY spectra were picked manually and integrated using Sparky. Unambiguous peaks were assigned manually. All picked peaks were calibrated and used for structure calculations in CYANA 3. Backbone angle restraints (107 ϕ and ψ angles) were obtained from the program TALOS.36, 37 HNHA-derived three-bond J-coupling constants38 were also used to derive an additional 45 ϕ angle dihedral restraints. Structures were calculated using a simulated annealing approach with 10,000 torsion angle dynamic steps per conformer. In order to introduce zinc ions into the structure in a manner compatible with the experimental constraints, the lowest energy structure from the CYANA 3 output was used as an input in CNS (anneal.inp script) and 20 new conformers were calculated using standard MD protocols with the distance and angle restraints from the previous step; zinc atoms and restraints to keep zinc–ligand bond lengths and angles in an appropriate range (based on LMO4 and LMO2-containing crystal structures, PDB codes 1RUT and 2XYJ) were added in this step. The 20 conformers were used to represent the solution structure of Lmo2LIM2–Ldb1LID. Analysis of geometrical properties were performed using the Protein Structure Validation Suite (available from the BioMagResBank, BMRB) and PROCHECK_NMR.39 The structure coordinates and the chemical shift data have been deposited in the Protein Data Bank (PDB code: 2LXD) and the BMRB database (BMRB code: 16779), respectively. Solvent accessible surfaces areas were calculated using MolMol.40 Images of protein structures were generated using PyMOL.