Conformational flexibility of the oncogenic protein LMO2 primes the formation of the multi-protein transcription complex

Abstract

LMO2 was discovered via chromosomal translocations in T-cell leukaemia and shown normally to be essential for haematopoiesis. LMO2 is made up of two LIM only domains (thus it is a LIM-only protein) and forms a bridge in a multi-protein complex. We have studied the mechanism of formation of this complex using a single domain antibody fragment that inhibits LMO2 by sequestering it in a non-functional form. The crystal structure of LMO2 with this antibody fragment has been solved revealing a conformational difference in the positioning and angle between the two LIM domains compared with its normal binding. This contortion occurs by bending at a central helical region of LMO2. This is a unique mechanism for inhibiting an intracellular protein function and the structural contusion implies a model in which newly synthesized, intrinsically disordered LMO2 binds to a partner protein nucleating further interactions and suggests approaches for therapeutic targeting of LMO2.

Figure 1. Structure of the LMO2 in complex with the anti-LMO2 VH#576.

The crystal structure of the dimeric complex of LMO2 and anti-LMO2 VH is shown either in space filling (A) or ribbon form (B). In both, the LMO2 protein is shown in blue and the VH framework region in cyan with CDR regions one, two and three highlighted in salmon, orange and cream respectively. In panel B, the zinc atoms are shown as grey spheres and sticks are used to represent residues involved in inter-molecular hydrogen bonds with oxygen and nitrogen atoms coloured red and blue respectively. For VH#576, there is one residue of CDR one (His31) forming a hydrogen bond, one in CDR two (Ser57) and four in CDR3 (Ser103, Glu105, Thr107 and Trp110).

Figure 2. Probing the interaction surface of VH#576 with LMO2 using site directed mutagenesis.

The CDR residues of VH#576 were mutated to glycine or alanine and the mutant sequences cloned into pEF-VP16 vector (prey vector) for use in a mammalian two-hybrid luciferase reporter assay. These assays were performed by transient transfection into CHO cells by co-transfecting the prey vector with a vector expressing a Gal4-LMO2 fusion bait (pM1-δLMO2); luciferase signals in the histograms are the ratio of Firefly luciferase to Renilla luciferase where the latter was used as a transfection control. Controls were performed using VH#576 (positive control) and anti-RAS VH#6 (negative control). The expression levels of each mutant VH#576 protein were established by Western detection with an anti-VP16 antibody (shown at the bottom of panels A, B, C). Panel A shows data for mutations of CDR1 residues, panel B mutations of CDR2 residues and panel C mutations of CDR3 residues. Each bar represents an average of luciferase activity measured for two wells (replicates) and the bar extensions indicate the standard deviations. Panels D and E highlight the key residues in the VH#576 that emerged from the mutation analysis shown in A-C. The CDR1, CDR2 and CDR3 of VH#576 are shown in salmon, orange and cream respectively. In panel D, LMO2 is shown in yellow (space filling format) and the framework region of VH in cyan, with the indicated amino acids being from the VH. Red patches are areas of oxygen atoms, blue are areas of nitrogen atoms. In panel E the interactions of VH#576 CDR3 are highlighted. A majority of the interactions occur through the CDR3 region (cream) with a total of seven CDR3 residues identified as critical for the interaction. A series of interactions occur across the hinge region (Phe88) of LMO2 including a salt bridge between Glu105 and Arg86, and polar contacts, represented by dashed lines. Zinc atoms are shown as yellow spheres.

Figure 3. Comparison of the geometry of the LMO2:VH dimer with the LMO2:LDB1-LID protein.

Superimpositions are shown of LMO2 with the LDB1 LID domain or with anti-LMO2 VH#576 and were performed using the first LIM domain for alignment. Panel A depicts the structure of LMO2 in green (with the zinc atoms shown as green spheres) when bound to LID or in blue when bound to VH#576. Panel B shows the dimeric LMO2 structures with LDB1-LID shown in red bound to LMO2 (green) compared with the structure of the LMO2 (shown in dark blue) complexed with the VH (shown in cyan). The angle between LIM1 and LIM2 domains differs by 23° between the two LMO2 structures. LMO2 wraps around VH#576 and forms a strong hydrogen bonding network including the β-strand of VH#576 typical of protein-protein interactions. The β-strands at this interaction site are parallel and an analogous interaction occurs between β-strand 7 of LMO2 (see Supplementary fig. 2) and LDB1-LID, with LMO2 residues Thr107 and Arg109, being key residues in both interactions. Hydrophobic amino acids at the protein-protein interface are indicated in panel C, focused around LMO2 residues Met106, Met108, Leu117 and Phe120 with VH#576 residues Trp47 and Leu45 indicating the hyrdrophobic effect contributes to binding. Zinc atoms are shown as green or blue spheres respectively where LMO2 interacts with LDB1 or with VH#576.

Figure 4. Intracellular binding of anti-LMO2 VH#576 to the LMO2 protein.

A mammalian luciferase two hybrid assay was used to assess interactions of LMO2 with partners (panel A) or VH (panel B). COS7 cells were transfected with GAL4-DBD bait plasmids expressing GAL4-LMO2 or GAL4-LMO2:LDB1-LID and VP16 activation domain prey plasmids expressing TAL1/SCL-VP16, VH#576-VP16 or VH#6-VP16. Where indicated, a plasmid expressing the E47 bHLH was co-transfected. Triplicate samples were analysed. p values of <0.005 were found for the difference between the interaction of LMO2 and TAL1/SCL with E47 or LMO2:LDB1-LID and TAL1/SCL with E47; <0.001 for the difference between the interaction of LMO2 and TAL1/SCL or LMO2:LDB1-LID and TAL1/SCL (both indicated by *) and 0.0001 for the difference between the interaction of TAL1/SCL and LMO2/LMO2:LDB1-LID with and without the presence of E47. The antibody fragment (VH#576) that binds LMO2 with high affinity reflected in the high luciferase signal in the two-hybrid reporter assay (panel B). However, the same VH has a low binding with the LMO2-LID fusion protein that comprises the LMO2 LIM1 and LIM2 domains fused to the LDB1 LIM-interacting domain (almost basal level compared to luciferase signal with a non-relevant control, anti-RAS VH, VH#6, panel B). (C) Western blotting of pull-down proteins. COS-7 cells were co-transfected with LMO2, LDB1, TAL1, E47 and GATA-1 and Flag-tagged VH#576 or Flag-tagged VH#6 (a non-relevant VH). Protein complexes were isolated with anti-Flag antibody beads and purified proteins separated on SDS-PAGE. The presence of LMO2, LDB1, GATA-1 or TAL1 was detected by Western blotting. (D). The effect on protein stability of binding VH#576 to LMO2 was measured using MEL585 cells stably expressing VH#576 or VH#6 (anti-RAS). Protein extracts from these or untransfected MEL585 (-) was separated by SDS-PAGE and transferred to membrane for Western analysis with anti-flag mouse monoclonal antibody (a); anti-tubulin as a protein loading control (b); or anti-LMO2 monoclonal antibody (AbD Serotec) (c) and anti-tubulin as a protein loading control (d). The Western blots have been cropped for clarity; please see Supplementary figures 5 and 6 for full length blots.

Figure 5. The sequestration model for building of LMO2 protein complexes.

In the structural data of this paper and previous publications, it seems that the newly synthesized LMO2 protein (shown in green) is intrinsically unstable, with few structured regions. One structured region is the central short α-helix between LIM1 and LIM2 domains, giving the LMO2 protein options for interacting with partner proteins. When it binds to a natural partner such as LDB1 (shown in red), a structural constraint is imposed on LMO2 and the heterodimer can nucleate subsequent protein complex formation, depicted here the complex found in erythroid cells including GATA1. The inhibitory effect of the VH#576 (shown in blue) exploits the LMO2 hinge region and the unstable properties by sequestering LMO2 (shown in green) into a complex that has poor interacting properties with natural partners. TAL1/SCL is shown in yellow, E47 in orange and GATA1 in blue. The diagrammatic model of the pentameric complex of LMO2 (green), LDB1 (red), E47 (orange), TAL1 (yellow) and GATA-1 (blue) was generated by in silico modeling. The structure of a guide DNA including the binding motifs for E47-TAL1 (CAGGTG) and GATA-1 (GATA) transcription factors was generated using Nucleic Acid Builder of AMBER tools. Subsequently, the structures of GATA-1, E47 and TAL1 bound DNA were derived by homology modeling using GATA-1 (PDB codes 1gat and 1gnf72) and heterodimer E47/NeuroD1 transcription factor (PDB code 2ql273) structures as templates for GATA-1 and E47-TAL1 respectively. The orientation of LMO2-LDB1 was manually adjusted by maximizing the overlap of interactions regions described in the literature: LMO2-E47-TAL1 and LMO2-GATA-1.