Spatial arrangement of an RNA zipcode identifies mRNAs under post-transcriptional control

Abstract

How RNA-binding proteins recognize specific sets of target mRNAs remains poorly understood because current approaches depend primarily on sequence information. In this study, we demonstrate that specific recognition of messenger RNAs (mRNAs) by RNA-binding proteins requires the correct spatial positioning of these sequences. We characterized both the cis-acting sequence elements and the spatial restraints that define the mode of RNA binding of the zipcode-binding protein 1 (ZBP1/IMP1/IGF2BP1) to the β-actin zipcode. The third and fourth KH (hnRNP K homology) domains of ZBP1 specifically recognize a bipartite RNA element comprised of a 5' element (CGGAC) followed by a variable 3' element (C/A-CA-C/U) that must be appropriately spaced. Remarkably, the orientation of these elements is interchangeable within target transcripts bound by ZBP1. The spatial relationship of this consensus binding site identified conserved transcripts that were verified to associate with ZBP1 in vivo. The dendritic localization of one of these transcripts, spinophilin, was found to be dependent on both ZBP1 and the RNA elements recognized by ZBP1 KH34.

Figure 1.

Sequence specificity of the 5RE. (A) Change in free energy of binding of point mutants as measured by fitting a fraction of RNA bound in EMSA to the Hill equation to measure the binding constant [K_d,app]. Error bars represent standard deviations of at least three replicates. (B) Representative EMSA experiments for wild-type zipcode RNA, C3G mutant RNA, and A6U mutant RNA. The filled triangle represents a 1:1 serial dilution of ZBP1 KH34. The ZBP1 KH34–RNA complex (*) and free RNA (**) are labeled.

Figure 2.

Sequence specificity of the 3RE. (A) The family of sequence motifs recovered after nine rounds of selection is shown. The range of nucleotide spacing between the nonrandomized zipcode residues and the selected motifs is indicated for each sequence. Copy numbers for each motif from a pool of 78 sequenced clones are listed. (B) Quantification and fit to the Hill equation of EMSA results for ZBP1 KH34 binding to the random 3′ zipcode library (solid black line), a representative clone from the library pool (green line), the round 9 selected population (dashed black line), and representative clones from the round 9 population containing the three most commonly selected sequence motifs (red, blue, and orange lines).

Figure 3.

RNA sequence elements recognized by the KH domains of ZBP1 KH34. (A) ¹⁵N-HSQC spectra of KH34 (black) in complex with RNAs containing either the 5RE (blue) or the 3RE (red). (B) Location of chemical shifts of amide resonances altered upon binding to the 3RE RNA (red) and 5RE (blue). Amino acids whose amide resonances were not assigned are indicated in orange (KH3) and violet (KH4).

Figure 4.

ZBP1 KH34 is capable of inducing bidirectional RNA looping. (A) Sequence of RNA in which the 5RE and 3RE were swapped. (B) Representative EMSA result for ZBP1 KH34 binding to RNA with the 5RE and 3RE swapped. The filled triangle represents a 1:1 serial dilution of ZBP1 KH34. The ZBP1 KH34–RNA complex (*) and free RNA (**) are labeled. (C) Model of bidirectional RNA looping to facilitate simultaneous contact of both consensus RNA elements with the RNA-binding surfaces of KH3 and KH4. Polarity of the RNA elements to their respective KH domains is conserved in both models.

Figure 5.

Consensus sequence for ZBP1 KH34 binding identifies novel mRNA ligands of ZBP1. (A) Human and mouse mRNA 3′ UTRs containing the ZBP1 KH34-binding consensus sequence. (B,C) Gene ontology analysis of conserved mRNA ligands containing the REs in the actin-like (B) and 5′-3′–swapped (C) orientation. (D) Enrichment of ZBP1 KH34 target mRNAs in ZBP1 immunoprecipitation compared with IgG control. Negative controls (B2M, GAPDH, and RPL13A) that do not contain the REs in their 3′ UTRs (open rectangle) and ZBP1 KH34 target mRNAs (filled rectangle) are shown. Error bars represent the SEM of at least two replicates using independent RIP experiments from littermate embryos.

Figure 6.

The ZBP1 KH34 REs are required for distal dendritic RNA transport. (A) FISH analysis of GFP-spinophilin 3′ UTR and GFP-spinophilin 3′ UTR ΔKH34 RE reporter mRNAs in cultured hippocampal neurons. DAPI (blue) was used to demarcate the nucleus. Bars, 10 μm. (B) Distribution (left) and average (right) distance (in microns) of reporter mRNAs from the neuron cell body. Error bars represent the SEM; P < 0.005.

Figure 7.

Dendritic localization of spinophilin mRNA is disrupted in ZBP1-KO mice. (A) FISH analysis of spinophilin transcripts in cultured hippocampal neurons of wild-type and ZBP1-KO mice. Cell-permeable CMAC dye (red) was used to demarcate cell boundaries. Bars, 10 μm. (B) Average number of spinophilin transcripts in dendrites detected by FISH in developing hippocampal neurons from ZBP1 wild-type and knockout mice; P < 0.0001. Quantity of mRNA detected in dendrites was normalized to dendrite length in microns. (C) Average distance (in microns) of spinophilin transcripts from the neuron cell body. Error bars in B and C represent the SEM; P > 0.4.