The structural basis for RNA selectivity by the IMP family of RNA-binding proteins

Abstract

The IGF2 mRNA-binding proteins (ZBP1/IMP1, IMP2, IMP3) are highly conserved post-transcriptional regulators of RNA stability, localization and translation. They play important roles in cell migration, neural development, metabolism and cancer cell survival. The knockout phenotypes of individual IMP proteins suggest that each family member regulates a unique pool of RNAs, yet evidence and an underlying mechanism for this is lacking. Here, we combine systematic evolution of ligands by exponential enrichment (SELEX) and NMR spectroscopy to demonstrate that the major RNA-binding domains of the two most distantly related IMPs (ZBP1 and IMP2) bind to different consensus sequences and regulate targets consistent with their knockout phenotypes and roles in disease. We find that the targeting specificity of each IMP is determined by few amino acids in their variable loops. As variable loops often differ amongst KH domain paralogs, we hypothesize that this is a general mechanism for evolving specificity and regulation of the transcriptome.

Fig. 1

ZBP1 and IMP2 overview and crystal structures. a Overall domain structure and sequence conservation of the IMP family members above. Below, sequence conservation across the IMP family members. b Ribbon diagrams overlaying the crystal structures of ZBP1KH34 in red (PDB: 3KRM) and IMP2KH34 in blue (this study, PDB:6ROL). c Representative EMSAs for wild-type C28 ß-actin zipcode RNA. The filled triangle represents a 1:1 serial dilution of ZBP1KH34 (top) and IMP2KH34 (bottom). The RBP–RNA complex (*) and free RNA (FREE) are labeled

Fig. 2

SELEX discovers targets that are IMP2 specific. a The sequences of RNAs after nine rounds of SELEX are shown. The range of nucleotide spacing between the nonrandomized IMP2 recognition elements is indicated for each sequence. Copy number and percentage of pool are listed. b Representative EMSAs for N30 SELEX library (top) and round 9 SELEX library (bottom). The filled triangle represents a 1:1 serial dilution of IMP2KH34. The RBP–RNA complex (*) and free RNA (FREE) are labeled. c Quantification and fit to the Hill equation of representative EMSA results for IMP2KH34 (solid red line) and ZBP1KH34 (solid blue line) binding to the round 9 SELEX library pool. d Quantification and fit to the Hill equation of representative EMSA results for IMP2KH34 (Round 7, solid red line) and ZBP1KH34 (round 7, solid cyan line, round 9, solid blue line) binding to different SELEX library pools. e Quantification of SELEX specificity for IMP2KH34 within each round SELEX library pool. The library specificity was calculated as the ratio between the K_d of ZBP1KH34 and the K_d of IMP2KH34 at a particular round of selection. f After sequencing each round of SELEX, the individual “GG” (left) and “CA” (right) motif occurrences were counted as a percentage of total SELEX sequences. The four most abundant motifs were plotted in terms of their relative abundance in each of the sequenced SELEX rounds. Source data are provided as a source data file

Fig. 3

IMP2 recognizes specific RNAs through interactions between the GXXG motifs and the variable loops. a ¹H ¹⁵N HSQC spectra of IMP2KH34 showing residues perturbed (black text) during separate titrations of the CA (left, blue) and GG (right, red) element containing RNAs. Sequence of the RNA used is depicted on bottom right of each spectrum. Enlarged spectra of the titrations are shown in Supplementary Fig. 3e. b Location of amide resonances altered upon binding to the CA element containing RNA (blue) and GG element containing RNA (red)

Fig. 4

Mutations of the KH3 variable loop are sufficient to determine RNA-binding. a Surface rendering of IMP2KH34 shows putative binding site of RNA with variable loops in blue and GXXG motifs in red. b Top: double loop replacement allows for IMP2 to now bind to the ß-actin zipcode. Middle: the KH4 loop swap by itself does not increase affinity to the ß-actin zipcode compared to WT IMP2KH34. Bottom: KH3 variable loop replacement is sufficient to swap the specificity of IMP2KH34 and allow it to bind to ß-actin zipcode. c Quantification and fit to the Hill equation of top and middle representative EMSA results (in b) for IMP2KH3 VL MUT (solid red line) and IMP2KH4VL MUT (solid blue line) binding to ß-actin zipcode. d Dissociation constants (K_d) of ZBP1KH34 WT (gel in Fig. 1c), IMP2KH34 WT (gel in Fig. 1c), and the IMP2 variable loop mutants (i.e., KH34 double VL MUT, KH3 VL MUT, and KH4 VL MUT) and ß-actin zipcode measured by EMSA

Fig. 5

Mutations validate the sequence preference of individual IMP2 RNA recognition elements. a K_d measurements of IMP2KH34 WT, ZBP1KH34 WT (from Patel et al.) and IMP2KH3 variable loop mutant. Point mutations were made to the ß-actin actin zipcode sequence (x-axis) and K_d was measured by EMSA (gels in Supplementary Fig. 5). b Summary of sequence preferences for WT and variable loop mutants. Single-nucleotide mutants of the ß-actin zipcode consensus elements were tested against IMP2 variable loop mutants and the preferred nucleotide at each position in written (gels in Supplementary Fig. 5). Data for ZBP1 consensus sequence preference was obtained from Patel et al.. c Quantification of representative EMSAs corresponding to B (gray box, gels in Supplementary Fig. 5). d Quantification of representative EMSAs corresponding to B (green box, gels in Supplementary Fig. 5) showing differences between mutant zipcode sequences that bind (red) and those that do not bind (blue)

Fig. 6

Genome wide search across 3′ UTRs for the IMP2 consensus sequence. a Human and mouse mRNA 3′ UTRs containing the IMP2KH34 binding consensus sequences. Conserved RNA targets were used for gene ontology analysis in b. b Gene ontology analysis of conserved mRNA ligands containing the bipartite IMP2KH34 RE. Gray box highlights gene symbols associated with diabetes. c Conserved targets for ZBP1KH34 (blue) and IMP2KH34 (orange) show little overlap. Box highlights gene symbols that are predicted to bind to both ZBP1 and IMP2