Interactions of Isl1, Lhx3 and Ldb1 in developing V2 interneurons and motor neurons. (A) Schematics of Isl1, Lhx3 and Ldb1 from mice showing the arrangement of key domains. In addition to the LIM and HD domains, Isl1 contains the Lhx3-binding domain (LBD; residues 262–291; identified herein) and the Isl-specific domain (ISD, identified on the basis of sequence conservation 296–310; Thor and Thomas, 1997). (B) Relative arrangement of Ldb1 and Lhx3 in developing V2 interneurons (left) and Ldb1, Lhx3 and Isl1 in postmitotic motor neurons (right). The interactions examined in this paper are indicated with dashed circles (1, 2, 3). (C) Far-UV CD spectra for Isl1_LBD and Ldb1_LID at concentrations of ∼ 20 μM. Raw baseline-corrected data are presented. (D) Y2H data comparing the interaction of Isl1 (residues 133–349) with Lhx3_LIM1+2 in the absence or presence of Ldb1_LID. The selection conditions used were -L-W-H, 1 mM 3-AT.

Comparison of Lhx3-bound Ldb1_LID and Isl1_LBD. Interaction maps of (A) Lhx3_LIM1+2 and Ldb1_LID and (B) Lhx3_LIM1+2 and Isl1_LBD. Residues from Ldb1_LID are shown as yellow boxes and those from Isl1_LBD as green boxes. Residues from Lhx3 that form contacts with the LIDs are classed as indicated. (C, D) Structural alignment over the backbone atoms of (C) Lhx3_LIM1 and (D) Lhx3_LIM2 showing the side-chain heavy atoms of Isl1_LBD (green) and Ldb1_LID (orange). The Lhx3 LIM domains from the Isl1_LBD structure are shown as a grey surface. (E) Structure-based sequence alignment of Isl1_LBD and Ldb1_LID. Asterisks show residues that occupy equivalent positions in the structures. Key Lhx3-binding residues in (F) Ldb1_LID and (G) Isl1_LBD. Lhx3_LIM1+2 in each case is shown as a surface model (grey) with Ldb1_LID (yellow) or Isl1_LBD (green). For Ldb1_LID, the side chains of key residues from alanine scanning mutagenesis screens are classed as having a strong (red), moderate (orange) or weak (yellow) effect, whereas for Isl1_LBD the indicated residues (green) all have a strong effect.

Comparison of LIM–Ldb1_LID complexes. (A) Structure-based sequence alignment of the LIM1 domains from murine LMO2, LMO4 and Lhx3 (Lhx3b numbering is used although the LIM domains are the same in different splice variants of Lhx3). Dots show conserved residues, dashes indicate gaps in the alignment and zinc-ligating residues are indicated by an asterisk. A consensus sequence for LIM domains from LIM-HD and LMO proteins from Deane et al (2004) is also shown. Fully conserved residues are indicated as capitals, + indicates R or K, − indicates D or E, and ϕ indicates bulky hydrophobic residues. Residues that are conserved at the interface of LIM–Ldb1_LID interactions are shaded (grey—hydrophobic; blue—backbone–backbone hydrogen bonds; yellow—other hydrogen bonds; red—electrostatic; many residues coloured blue, yellow or red also make hydrophobic interactions). Underlined residues in Lhx3 are at the interface of the Lhx3/Isl1_LBD complex. (B, C) Ldb1_LID from complexes containing LMO2 (1J2O, cyan), LMO4 (1M3V, violet; IRUT, magenta) and Lhx3 (yellow) shown in relation to the LIM domains from Lhx3 (grey surface representation) structures are overlaid over the backbone atoms of the LIM domains. (B) LIM1 domains/C-terminal half of Ldb1_LID as backbone traces. (C) LIM2 domains/side-chain heavy atoms of the N-terminal half of Ldb1_LID. (D) Summary of alanine mutagenic screening of Ldb1_LID against LMO2, LMO4, Lhx3 and Isl1. The sequence of Ldb1_LID is shown at the top; coloured boxes indicate a weak (yellow), moderate (orange) and strong (red) effect on binding when residues were individually mutated to alanine. Grey boxes indicate no differences. Where structures have been determined (LMO2_LIM1, LMO4, Lhx3), β-strands formed in Ldb1_LID are shown as black bars. The spacer region from the Lhx3–Ldb1_LID complex is shown as a green bar.

Structures of Lhx3_LIM1+2–LID complexes. (A) Ribbon representation of the lowest energy structure of Lhx3_LIM1+2–Ldb1_LID in solution. Only the structured regions of Lhx3 (blue) and Ldb1_LID (yellow) are shown. The zinc ions (grey spheres) and the side chains of the zinc-ligating residues (orange) are shown. The position of the unstructured linker is indicated. (B, C) Backbone traces of the 20 lowest energy structures of the Lhx3_LIM1+2–Ldb1_LID complex. Overlays over the backbone atoms of (B) LIM1 (Lhx3_28–89 plus Ldb1_316–327) and (C) LIM2 (Lhx3_91–151 and Ldb1_301–311) are shown. (D) The structured residues of monomeric Lhx3_LIM1+2–Isl1_LBD complex. Lhx3 is shown in blue and Isl1_LBD is shown in green. (E) The symmetry-related dimer Lhx3_LIM1+2–Isl1_LBD. Half-dimers are coloured green and blue, respectively, with Lhx3 in the darker and Isl1 in the lighter colour. (F) Overlay of dimeric (Lhx3 (green)/Isl1_LBD (turquoise)) and monomeric (Lhx3 (yellow)/Isl1_LBD (magenta)) forms of the Lhx3_LIM1+2–Ldb1_LID complex overlaid over the backbone atoms of Lhx3_LIM2. Only residues 262–273 of Isl1_LBD from the dimer are shown for clarity. Figures were prepared in MolMol (Koradi et al, 1996).

Binding affinities and modelling of complex formation. (A) ELISAs showing the ability of FLAG-tagged Ldb1_LID to bind to the cut Isl1/Ldb1_LID and Lhx3/Ldb1_LID complexes. Data were corrected for background binding to GST, and the error bars show the range of values obtained from triplicates of a single experiment. The inset shows a schematic of tethered intramolecular constructs of the Isl1–Ldb1_LID, where the linker contains a Factor Xa protease site. After treatment with Factor Xa, the linker is cut to form an intermolecular complex. (B) GdnHCl denaturation curves for Lhx3–Ldb1_LID and Lhx3–Isl1_LBD constructs as indicated. (C–E) Trends for the formation of protein and protein:DNA complexes containing Ldb1, Lhx3 and Isl1. Concentrations of all components were set at 1, 10 and 100 nM. (C) Distribution of Ldb1/LIM-HD complexes using the binary model (no interaction between Isl1 and Lhx3, and no DNA). (D) Distribution of protein-only complexes using the ternary model, which also considers binding of Isl1 and Lhx3. (E) Distribution of key protein:DNA complexes relative to Ldb1/Lhx3/DNA. The I series shows the results of using the independent model, the SA series shows the results of the single-molecule DNA-binding advantage model where the binding of two and four HDs to a single molecule of DNA shows increased binding affinities over single HD binding.

Putative mechanism for cofactor exchange between Lhx3 and Isl1. Only Ldb1_LID is shown for simplicity. (A) Lhx3 and Ldb1 form a complex where Ldb1_LID extends across both LIM domains of Lhx3. (B) Lhx3_LIM2 binds more tightly than Lhx3_LIM1 and for some of the time Lhx3_LIM1 is exposed while Lhx3_LIM2 remains bound. (C) Isl1_LIM1 can bind the C-terminal half of Ldb1_LID. (D) The C-terminal half of Isl1_LBD is brought into close proximity to Lhx3 and binds Lhx3_LIM1 (E, F) The N-terminal half of Ldb1_LID dissociates from Lhx3_LIM2 to bind Isl1_LIM2, allowing the N-terminal half of Isl1_LBD to bind Lhx3_LIM2. Complexes in (C–F) are stabilised through the simultaneous binding to both Lhx3 and Isl1 recognition sites on target genes.