Structural basis for the recognition of ldb1 by the N-terminal LIM domains of LMO2 and LMO4

Abstract

LMO2 and LMO4 are members of a small family of nuclear transcriptional regulators that are important for both normal development and disease processes. LMO2 is essential for hemopoiesis and angiogenesis, and inappropriate overexpression of this protein leads to T-cell leukemias. LMO4 is developmentally regulated in the mammary gland and has been implicated in breast oncogenesis. Both proteins comprise two tandemly repeated LIM domains. LMO2 and LMO4 interact with the ubiquitous nuclear adaptor protein ldb1/NLI/CLIM2, which associates with the LIM domains of LMO and LIM homeodomain proteins via its LIM interaction domain (ldb1-LID). We report the solution structures of two LMO:ldb1 complexes (PDB: 1M3V and 1J2O) and show that ldb1-LID binds to the N-terminal LIM domain (LIM1) of LMO2 and LMO4 in an extended conformation, contributing a third strand to a beta-hairpin in LIM1 domains. These findings constitute the first molecular definition of LIM-mediated protein-protein interactions and suggest a mechanism by which ldb1 can bind a variety of LIM domains that share low sequence homology.

Fig. 1. Interaction between LMOs and ldb1 is mediated by a single LIM domain. (A) Schematic diagram of LMO2, LMO4 and ldb1. (B) Immunoprecipitation experiment. Human embryonal kidney (293T) cells were transfected with expression constructs encoding Flag-tagged derivatives of LMO4 representing either the single LIM domains [F-LMO4-LIM1 or F-LMO4-LIM2 (lanes 1 and 2, respectively)] or full-length protein (F-LMO4, lane 3), together with plasmid encoding HA-tagged ldb1. Whole-cell lysates were prepared and proteins were immunoprecipitated with pre-immune or anti-ldb1 antisera. Immuno blotting was performed with anti-Flag monoclonal antibody. Arrows indicate F-LMO4-LIM1 and F-LMO4. Western blot analysis of lysates confirmed expression of HA-ldb1 and Flag-LMO proteins (lower panels). The arrow depicts full-length LMO protein (17 kDa), while the lower bands correspond to individual LIM domains. (C) Size-exclusion chromatography. Isolated protein at ∼1 µM (bottom panels) or solutions containing equimolar concentrations of MBP–LMO2-LIM1 or MBP– LMO4-LIM1 and MBP–ldb1-LID at the indicated concentrations were applied to a Superose12™ size-exclusion column (Amersham Biosciences). The elution volumes of uncomplexed MBP–LMO-LIM1 proteins are indicated by an asterisk in each panel. Uncomplexed MBP–ldb1-LID is shown as a dotted chromatogram in the bottom panels. The elution volumes of the complexes are indicated by broken lines.

Fig. 2. The solution structure of FLIN4. (A) Amino acid sequences of FLIN4 and FLIN2. Dashes indicate gaps in the sequence; dots are identical residues. Residues 1–71 of FLIN4 correspond to residues 16–86 of LMO4 (DDBJ/EMBL/GenBank accession No. XM_030627) with the mutations C37S/C49S, residues 72–82 constitute the linker, residues 83–122 correspond to residues 300–339 of ldb1 (DDBJ/EMBL/GenBank accession No. NM_003893). The same numbering is used for FLIN2; however, ³Gly comes from the polylinker region of the pGEX-2T plasmid, while residues 4–65 correspond to residues 26–87 of LMO2 (DDBJ/EMBL/GenBank accession No. BC035607). Residues derived from LMO proteins are in blue, residues from ldb1 are in red and the linker is in black. (B) Stereo view of the structured regions of FLIN4 with the 20 lowest-energy structures overlaid on the backbone atoms of the LMO4-LIM1 domain. LMO4-LIM1 is shown in blue, while ldb1-LID is in red. (C) Ribbon diagrams showing a comparison of FLIN4 (left) with LIM domains from qCRP2-LIM1 (Kontaxis et al., 1998) (right, upper, 1A71) and PINCH1-LIM1 (Velyvis et al., 2001) (right, lower; 1G47). Zn₁ is shown in cyan, Zn₂ in magenta and ldb1 in yellow. Zinc-ligating side chains are shown in orange. (D) Comparison of CCCC [mGATA-1 N-finger, dark blue, 1GNF (Kowalski et al., 1999); cGATA-1 C-terminal finger, mid-blue, 1GAT (Omichinski et al., 1993); Zn₂ from qCRP2-LIM1, cyan] and CCCD (Zn₂ from FLIN4, magenta) modules. Only the zinc ion from FLIN4 is shown for clarity. Backbone r.m.s.d. over residues 34–38 and 50–65 (or equivalent) is 0.96 Å. (E) Comparison of CCCD (Zn₂ from FLIN4, magenta) and CCCH (PINCH1-LIM1, yellow) overlaid over same residues.

Fig. 3. The linker region of FLIN4 is unstructured and does not affect the LMO4-LIM1:ldb1-LID complex. (A) Heteronuclear ¹⁵N–¹H NOE data for FLIN4. NOE values were calculated as the ratio of peak intensities with and without proton saturation and plotted as a function of residue number. Data were only calculated from those residues that had well-resolved amide peaks. (B) Overlay of part of the ¹⁵N HSQC spectrum of FLIN4 containing a Factor Xa protease site (FLIN4-Xa, line) with that of the same protein following treatment with Factor Xa (dashes). The chemical shifts of peaks from FLIN4-Xa are identical to those obtained for FLIN4. Over 90% proteolysis was achieved according to analysis by SDS–PAGE.

Fig. 4. The ldb1-LID forms an additional β-strand. (A) Schematic diagram showing observed NOEs between the β-strand in ldb1-LID and the second β-hairpin in LMO4-LIM1. Observed NOEs are shown as bold lines; NOEs that may be present but obscured by spectral overlap are shown as broken lines. Only backbone atoms are shown; side chains have been omitted for simplicity. (B) Hα–Hα NOEs from a 2D homonuclear NOESY spectrum of FLIN4.

Fig. 5. The solution structure of FLIN2. (A) Stereo view of the structured regions of FLIN2 with the 20 lowest-energy structures overlaid over the backbone atoms of the LMO2-LIM1 domain (residues 6–40 and 49–65). LMO2-LIM is shown in blue, ldb1 is shown in red and residues 41–48 are shown in cyan. (B) The same structures overlaid on the backbone atoms of the first zinc-binding module of LMO2-LIM1. (C) The same structures overlaid on the backbone atoms of the second zinc-binding module of LMO2-LIM1. Residues 41–48 are shown in cyan. (D) Ribbon diagram showing FLIN2. Zn₁ is shown in cyan, Zn₂ in magenta and ldb1 in yellow. Zinc-ligating side chains are shown in orange. (E) Ribbon diagrams showing a comparison of FLIN2 and FLIN4, FLIN2 is in magenta and orange, and FLIN4 is in cyan and yellow. The structures are overlaid using the backbone atoms of the structured regions of the LIM domains.

Fig. 6. Binding of ldb1-LID to LMO-LIM1. (A) Surface potential figure of LMO4-LIM1 (left) and LMO2-LIM1 (blue positive, red negative) drawn using MOLMOL (Koradi et al., 1996); ldb1 is in yellow. Residues from ldb1 are labeled in red, and residues from LMO4 and LMO2 are labeled in blue. (B) Sequence comparison of Group I LIM1 domains. Well-defined regions of secondary structure in FLIN4 are indicated with arrows for β-strands and a coil for the α-helix. Hydrophobic residues and positively charged residues that appear to mediate the LMO4-LIM1:ldb1 interaction are boxed. Zinc-ligating residues are highlighted in yellow, while other fully conserved residues are highlighted in cyan. In a consensus sequence of LMO and LHX LIM1 domains, well-conserved residues are indicated in lower-case letters. (C) Specificity of ldb1 binding. SDS–PAGE analysis shows the relative binding of MBP–ldb1-LID (arrow) to immobilized GST, GST-LMO4-LIM1 and GST-LMO4mut. Components of lanes 1–6 are indicated, lane 7 carried molecular weight markers and lane 8 shows 5% input of MBP–ldb1-LID.