Molecular Insights into the Coding Region Determinant-binding Protein-RNA Interaction through Site-directed Mutagenesis in the Heterogeneous Nuclear Ribonucleoprotein-K-homology Domains

Background: Coding region determinant-binding protein (CRD-BP) interacts physically with oncogenic mRNAs.

Results: Point mutation in the K-homology (KH) domains of CRD-BP abolishes its RNA-binding ability.

Conclusion: Two KH domains of CRD-BP are required for efficient binding to oncogenic mRNAs and for granule formation in zebrafish embryos.

Significance: Learning how the KH domains interact with mRNAs is crucial for understanding oncogenic function of CRD-BP.

Abstract

The ability of its four heterogeneous nuclear RNP-K-homology (KH) domains to physically associate with oncogenic mRNAs is a major criterion for the function of the coding region determinant-binding protein (CRD-BP). However, the particular RNA-binding role of each of the KH domains remains largely unresolved. Here, we mutated the first glycine to an aspartate in the universally conserved GXXG motif of the KH domain as an approach to investigate their role. Our results show that mutation of a single GXXG motif generally had no effect on binding, but the mutation in any two KH domains, with the exception of the combination of KH3 and KH4 domains, completely abrogated RNA binding in vitro and significantly retarded granule formation in zebrafish embryos, suggesting that any combination of at least two KH domains cooperate in tandem to bind RNA efficiently. Interestingly, we found that any single point mutation in one of the four KH domains significantly impacted CRD-BP binding to mRNAs in HeLa cells, suggesting that the dynamics of the CRD-BP-mRNA interaction vary over time in vivo. Furthermore, our results suggest that different mRNAs bind preferentially to distinct CRD-BP KH domains. The novel insights revealed in this study have important implications on the understanding of the oncogenic mechanism of CRD-BP as well as in the future design of inhibitors against CRD-BP function.

Introduction

Coding region determinant-binding protein (CRD-BP⁴; mouse), also known as IMP1 (human), belongs to a highly conserved family of RNA-binding proteins called VICKZ (Vg1 RBP/Vera, IMP-1,2,3, CRD-BP, KOC, ZBP-1) (1, 2). Other orthologous or paralogous members include IMP2 (human), IMP3 (human), ZBP1 (chicken), and Vg1-RBP/Vera (Xenopus). All members of VICKZ have two N-terminal RNA recognition motifs (RRM) followed by four C-terminal heterogeneous nuclear RNP-K-homology (KH) domains and have been implicated in the post-transcriptional regulation of a number of different mRNAs (1, 2).

CRD-BP was first discovered due to its ability to physically associate with a specific coding region of c-myc mRNA called the coding region determinant (3, 4), and its ability to influence c-myc mRNA stability in a cell-free model system (5). Several subsequent studies have provided in vivo evidence for the role of CRD-BP in controlling the c-myc mRNA half-life (6–9). In addition to c-myc, CRD-BP also has high affinity for a number of mRNAs whose gene products have been implicated in cancer, which include mRNAs for βTrCP1 (8), CD44 (10), IGF-II (11), GLI1 (12), β-catenin (13), K-Ras (14), MAPK4 (15), MITF, and MDR1 (17). The oncogenic role of CRD-BP is further exemplified by the following two observations: (i) overexpression of CRD-BP in many types of human cancers (2), and (ii) mice genetically engineered to overproduce CRD-BP in mammary glands developed mammary adenocarcinoma (18). Although the exact oncogenic function of CRD-BP is still unclear, many studies using cell lines have demonstrated that its ability to physically interact with target mRNAs is at least one major contributing factor. For instance, it was shown that CRD-BP binds with high affinity to several regions at the 3′-untranslated region (UTR) of CD44 mRNA to stabilize it, leading to cell adhesion, cytoplasmic spreading, and invadopodia formation (10). CRD-BP binds to the coding region of βTrCP1 mRNA, and overexpression of CRD-BP led to the stabilization of βTrCP1 mRNA and elevation of βTrCP1 expression levels in colorectal cancer cells (8). CRD-BP has recently been shown to bind to the coding region of β-catenin (13) and GLI1 mRNAs (12) to stabilize the transcripts, providing further evidence for the role of CRD-BP in the Wnt/β-catenin signaling pathway. CRD-BP also binds to the coding region and 3′UTR of K-Ras mRNA, and its overexpression led to increases in K-Ras expression as well as colon cancer cell proliferation (14). Furthermore, the 3′UTR of MITF mRNA is also a binding site for CRD-BP, and such an interaction is shown to be critical for protecting the MITF transcript from degradation by miR-340, a mechanism believed to be important in melanocytes and malignant melanoma.

To date, most if not all of our understanding on the structure and function of CRD-BP has been derived from studies on its paralogs. Deletion analysis studies of the human (IMP1), frog (Vg1RBP), and chicken (ZBP1) paralogs of CRD-BP have all concluded that the two RRMs are not important for RNA binding (19–21). However, there are some conflicting results regarding the significance of each KH domain with respect to their interaction with RNA. For instance, although the KH3&4di-domain of ZBP1 appears to be sufficient to bind to its RNA substrate, the 54-nt zipcode β-actin RNA (21), this was not the case for IMP1 and Vg1RBP. Unlike the full-length and KH1–4 recombinant proteins, the KH3–4di-domain of IMP1 showed no affinity for the 173-nt H19 RNA (20). Similarly, the recombinant protein containing only the KH3&4di-domain of Vg1RBP exhibited more than a 10-fold decrease in affinity for VLE RNA when compared with the wild-type protein (19). The crystal structure of the KH3&4di-domain of IMP1 has recently been determined, and it has been proposed that RNA looping induced by IMP1 binding is the mechanism whereby IMP1 provides specific recognition for its RNA substrates (21). Using nuclear magnetic resonance spectroscopy, it was demonstrated that the C-terminal KH3 and KH4 domains can recognize a bipartite RNA sequence, with KH4 binding to a 5′-CGGAC element and KH3 binding to a 5′-(C/A)CA(C/U) element (22). In deciphering the role of the RNA-binding domains of CRD-BP paralogs, all the above-mentioned studies have utilized isolated domains or employed deletion variants, and hence it is still unclear to what extent each KH domain plays in the context of the entire CRD-BP protein in binding to RNA.

The minimal eukaryotic (type 1) KH domain is characterized by a β₁α₁α₂β₂ topology, with the maximum KH domain containing an extra β′α′ at its C terminus. Typically, the KH domain fold consists of three anti-parallel β-strands in the order β₁, β′, and β₂ creating the core of the domain. The core β-sheet is framed on three sides by the α-helices, forming the outer face of the domain. The RNA-binding platform is located on the outer solvent-exposed face of the α₁α₂ region, where the RNA substrate is positioned between two flexible loops (23). Essential to the KH domain, the RNA-binding platform is a strictly conserved GXXG motif, located within a flexible loop between α₁ and α₂ (23). This GXXG motif, in concert with a variable loop located between β₂ and β′, functions to clamp an RNA substrate in place upon binding. It is thought that the glycine residues in the GXXG motif are essential due to their conformational flexibility and small steric size (24). Correspondingly, the RNA binding function of the KH domains of NusA, GLD-1, Sam68, and heterogeneous nuclear RNP K were all severely impeded upon mutating the first glycine in the GXXG motif to an aspartate (24–27). In fact, a recent study showed the feasibility of double mutating the GXXG loop, namely from GXXG to GDDG, in the KH domain of the RNA-binding protein KSRP and the KH3 and KH4 domains of IMP1, as an effective tool for the investigation of the nucleic acid-binding function of individual KH domains (28).

In this study, we use site-directed mutagenesis to mutate the first Gly residue in each of the GXXG motifs in the KH domains of CRD-BP as an approach to understand to what degree each of the four domains are involved in binding to RNA substrates in vitro and in cells. We show that, with the exception of KH3–4, mutations in any two GXXG motifs in the four KH domains lead to a complete abrogation of CRD-BP binding to c-myc and CD44 RNAs in vitro. We also show that these in vitro RNA-binding profiles correlate with CRD-BP granule formation in intact zebrafish (Danio rerio) embryos. However, we find that even a single GXXG point mutation in each KH domain individually can lead to a significant decrease in the amount of c-myc and CD44 mRNAs associated with FLAG-CRD-BP in HeLa cells. Taken together, our results reveal important insights into how CRD-BP physically associates with its RNA-binding partners.

EXPERIMENTAL PROCEDURES

Plasmid Construction

The plasmids pET28b(+)-CRD-BP and pcDNA-CRD-BP-FLAG containing mouse CRD-BP cDNA were generous gifts from Dr. Jeffrey Ross, University of Wisconsin. The pET28b(+)-CRD-BP was used to generate recombinant WT CRD-BP, and the pcDNA-CRD-BP-FLAG was used to overexpress FLAG-CRD-BP in HeLa cells. For the generation of various KH point mutation variants, we used the PCR-based site-directed mutagenesis method employing pET28b(+)-CRD-BP and pcDNA-CRD-BP-FLAG as templates. To generate the single KH variants, the following primer pairs were used: KH1, forward primer 5′-GGCGCTATCATTGACAAGGAGGGTGCC-3′ and reverse primer 5′-GGCACCCTCCTTGTCAATGATAGCGCC-3′; KH2, forward primer 5′-GGGCGACTCATTGACAAGGAAGGGCGG-3′ and reverse primer 5′-CCGCCCTTCCTTGTCAATGAGTCGCCC-3′; KH3, forward primer 5′-GGCGCCATCATTGACAAGAAGGGCCAG-3′ and reverse primer 5′-CTGGCCCTTCTTGTCAATGATGGCGCC-3′; and KH4, forward primer 5′-CCGCGTCATCGACAAAGGCGGCAAAAC-3′ and reverse primer 5′-GTTTTGCCGCCTTTGTCGATGACGCGG-3′. Bases altered from the CRD-BP cDNA are shown in boldface with underline and the targeted codons are underlined. KH variants containing a point mutation at the GXXG motif in two KH domains were sequentially generated. For instance, to generate the KH1–2 variant, the KH1 plasmid was used as a template, and the primer sets to point mutate the GXXG motif in the KH2 domain were used in PCR-based site-directed mutagenesis.

For the generation of the WT CRD-BP-EGFP-pSp64TNE plasmid, the open reading frame of CRD-BP was amplified by PCR using the plasmid pcDNA-CRD-BP-FLAG as a template and the following primer pairs: CRD-BP-Kpn1-F (5′-ACCAGGTACCATGAACAAGCTTTACATCGGCA-3′) and CRD-BP-Age1-R (5′-ACCAACCGGTAACTTCCTCCGAGCCTGGGCCA-3′). The amplified fragment was then inserted into the KpnI and AgeI restriction sites of the EGFP-pSp64TNE plasmid. Similarly, for the generation of the EGFP-pSp64TNE KH point mutation variants, the open reading frames of the KH point mutation variants were amplified using the appropriate pcDNA-CRD-BP-FLAG KH variants as templates and CRD-BP-Kpn1-F and CRD-BP-Age1-R primers, and the PCR-amplified fragments were inserted into the same sites of EGFP-pSp64TNE as described above. For the generation of truncated CRD-BP in EGFP-pSp64TNE plasmids, the following primer pairs were used: RRM1&2 (forward primer CRD-BP-Kpn1-F, reverse primer 5′-ACCAACCGGTAACCCTGCTGCCACGGGCGACCCTT-3′); KH1 to 4 (CRD-BP-KH1-Kpn1-F 5′-ACCAGGTACCATGATCCCTCTCCGGCTCCTGGT-3′ and reverse primer CRD-BP-Age1-R); KH1 (CRD-BP-KH1-Kpn1-F, CRD-BP-KH1-Age1-R 5′-ACCAACCGGTAACAAGATCATCTTGCACGCGGA-3′); KH2 (CRD-BP-KH2-Kpn1-F 5′-CCAGGTACCATGCTGAAGATCCTGGCTCATAAC-3′, CRD-BP-KH2-Age1-R 5′-ACCAACCGGTAACTCTCGAACTTTCTTCATGAT-3′); KH3 (CRD-BP-KH3-Kpn1-F 5′-ACCAGGTACCATGGTACAAGTGTTCAT-3′, CRD-BP-KH3-Age1-R 5′-ACCAACCGGTGTAGCCTCTGGGGGTCCAGT-3′); KH4 (CRD-BP-KH4-Kpn1-F 5′-ACCAGGTACCATGAAGCTAGAGACCCACATACGG-3′, CRD-BP-Age1-R); KH1&2 (CRD-BP-KH1-Kpn1-F, CRD-BP-Age1-R); KH2&3 (CRD-BP-KH2-Kpn1-F, CRD-BP-KH3-Age1-R); KH3&4 (forward primer 5′-ACCAGGTACCATGGCTCCCTATAGCTCCTTCATGCA-3′, CRD-BP-Age1-R). Two subcloning steps were used to generate the di-KH domains KH1&3, KH1&4, and KH2&4 EGFP constructs. For generating KH1&3-EGFP-pSp64TNE, KH1 was amplified by PCR and subcloned into the KpnI site at the 5′end of KH3-EGFP-pSp64TNE construct. The same fragment was subcloned into the 5′end of KH4-EGFP-pSp64TNE construct to generate KH1&4-EGFP-pSp64TNE. For generating KH2&4-EGFP-pSp64TNE, KH2 was amplified by PCR and subcloned into the 5′end of KH4-EGFP-pSp64TNE. All of the above generated constructs were verified by DNA sequencing performed by Macrogen (Seoul, Korea).

Generation and Purification of Recombinant CRD-BP and Its Variants

Recombinant CRD-BP was purified from Escherichia coli BL21 (DE3) using a 1-ml bed volume of nickel-nitrilotriacetic acid (Qiagen) column under denaturing conditions. Proteins eluted from the column at pH 5.4 were subjected to three steps of dialysis. The first step was for 24 h in pH 7.4 buffer containing 200 mm NaCl, 20 mm Tris-HCl, 1 mm reduced glutathione, 0.1 mm oxidized glutathione, 10% (v/v) glycerol, 2 m urea, and 0.01% (v/v) Triton X-100. The protein was then dialyzed twice, each for 2 h in the same buffer as above but without urea and the glutathiones. Following dialysis, samples were centrifuged at 13,200 rpm for 30 min to remove any precipitated proteins. The purified protein solutions were quantified and analyzed for purity using Coomassie Brilliant Blue-stained 12% SDS-PAGE.

Radiolabeled in Vitro Transcription

Plasmid pUC19-CRDmyc-1705–1886 was used to amplify the DNA template for use in synthesizing internally radiolabeled 182 nts c-myc CRD RNA. The PCR primers used to amplify c-myc DNA corresponding to nts 1705–1886 are as follows: forward primer GGATCCTAATACGACTCACTATAGGACCAGATCCCGGAGTTGG; reverse primer, TAGCTGTTCAAGTTTGTG. The T7 RNA promoter sequences are underlined. The plasmid pCYPAC2-CD44, which contains the last CD44 exon, was a gift from Dr. Finn C. Nielsen (University of Copenhagen, Denmark) and was used as template for PCR amplification. The PCR primers that were used to amplify the CD44 DNA nts 2862–3055, corresponding to the 3′UTR of CD44 mRNA, are as follows: forward primer GGATCCTAATACGACTCACTATAGGAAATTAGGGCCCAATTAA and reverse primer AAATTTCCTCCCAGGGAC. PCR-amplified DNA templates were used directly for in vitro transcription by T7 RNA polymerase. One μg of DNA template was incubated for 1 h at 37 °C in a 20-μl reaction containing 1× transcription buffer (Promega, Madison, WI), 10 mm dithiothreitol, 1 unit of RNasin (Promega), 0.5 mm ATP, 0.5 mm CTP, 0.5 mm GTP, 12.5 μm UTP, 1.5 units of T7 RNA polymerase (Promega, Madison, WI), and 40 μCi of [α-³²P]UTP (3000 Ci/mmol). Following incubation, 3 units of RNase-free DNase I (Promega) were added, and the reaction was further incubated for 10 min at 37 °C. Upon addition of 10 μl of Stopping dye (9 m urea, 0.01% bromphenol blue, 0.01% xylene cyanol FF, 0.01% phenol), the entire sample was electrophoresed on an 8% polyacrylamide, 7 m urea gel, and the band containing internally radiolabeled RNA was excised and eluted with elution buffer (10 mm Tris-HCl, pH 7.5, 0.1 m NaCl, 1 mm EDTA, 0.01% SDS) at 45 °C for 6 h. The purified radiolabeled RNA was then phenol/chloroform-extracted followed by ethanol precipitation. Specific activity of the RNA was then determined by scintillation counting.

Electrophoretic Mobility Shift Assay

The electrophoretic mobility shift assay (EMSA) binding buffer (5 mm Tris-Cl, pH 7.4, 2.5 mm EDTA, pH 8.0, 2 mm DTT, 5% glycerol, 0.1 mg/ml bovine serum albumin, 0.5 mg/ml yeast tRNA, 5 units of RNasin) (17) was prepared on ice prior to each experiment. To facilitate RNA denaturation and renaturation, the [³²P]RNA sample was heated to 55 °C for 5 min and cooled to room temperature before adding the EMSA binding buffer and the appropriate amount of purified recombinant CRD-BP to a final volume of 20 μl. Reactions were incubated at 37 °C for 10 min and transferred to ice for 5 min. The reaction was again incubated at 37 °C for 10 min and transferred to ice for an additional 5 min. A total of 2 μl of EMSA loading dye (250 mm Tris-Cl, pH 7.4, 0.2% bromphenol blue, 0.2% xylene cyanol, 40% sucrose) was added to each reaction, and 15 μl of the EMSA reaction was loaded onto a 4% native polyacrylamide gel and resolved at 25 mA for 60 min. Following electrophoresis, the gel was exposed overnight at −80 °C and subjected to autoradiography using the Cyclone PhosphorImager and OptiQuant software.

EMSA saturation binding experiments were carried out as described above, and the dissociation constant (K_d) for the CRD-BP-RNA interaction was determined using the Hill equation. The saturation binding data were analyzed by densitometry of the autoradiograph using the Cyclone Storage Phosphor-System and OptiQuant software. For each reaction, the total activity in each lane was determined; this involved summing the total activity in bound complexes with the total activity present in the unbound fraction. The percentage of bound RNA and the protein concentration (in nanomolars) were inserted into the Hill equation, and the results were expressed graphically.

Cell Culture, Transfection, and Immunoprecipitation

HeLa human cervical cancer cells purchased from ATCC were cultured in minimum essential media supplemented with 10% fetal bovine serum at 37 °C in 5% CO₂. The day before transfection, 10 × 10⁴ cells/ml were plated onto 100-mm dishes. Transient transfection of 10 μg of pcDNA-CRD-BP-FLAG plasmids was carried out using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's instructions. As a negative control, cells were transfected with the plasmid vector pcDNA-FLAG. Forty eight hours after transfection, cells were lysed with 1 ml of Total Cell Lysis (TCL) buffer (50 mm Tris-Cl, pH 7.4, 150 mm NaCl, 1 mm EDTA, 0.1% Triton X-100) supplemented with 1 mm vanadyl ribonucleoside, 0.5 mm DTT, 0.05 units of RNasin, and protease inhibitor tablet (Roche Diagnostics). After 5 min on ice, lysed cells were aspirated using a 25.5-gauge needle five times to break the nuclei. After a 30-min incubation on ice, the cell lysate was centrifuged at 14,000 rpm for 10 min. The collected supernatant was subjected to preclearing by incubating with 50 μl of equilibrated protein-G-agarose beads (50% slurry) at 4 °C for at least 1 h. The resin was spun down at 3000 × g for 1 min, and the lysate was collected. The pre-clearing step was repeated once after which the pre-cleared lysate was added to 5 μl of anti-FLAG antibody (F1804, Sigma) and mixed overnight at 4 °C. Protein G-agarose beads were then added to the antibody/lysate mixture and mixed for 4 h at 4 °C to capture the FLAG antibody. The agarose beads were then washed four times with TCL buffer followed by five washes with TCL buffer containing 1 m urea. Following the fourth wash with TCL buffer containing urea, 50% of the agarose beads were collected and spun down at 3000 × g for 1 min followed by resuspension in 16 μl of water for Western blot analysis. The remaining 50% of the agarose beads was subjected to a final wash followed by resuspension in 100 μl of TCL buffer containing 0.3 mg/ml proteinase K and 0.1% SDS. The sample was then incubated at 50 °C for 30 min. Following the incubation, RNAs physically associated with the resin were extracted by the phenol/chloroform/isoamyl alcohol method and quantified using a Nanodrop spectrometer (Wilmington, DE). One μg of RNA from each treatment group was treated with DNase (DNA-free^TM kit, Ambion) before subjecting the sample for cDNA synthesis and quantitative real time PCR as described below. We also used an equal volume of RNA samples for normalization purposes and found no differences in the quantitative real time PCR results.

Western Blot Analysis

Protein samples were separated using a 12.5% polyacrylamide/SDS Laemmli gel system, transferred to a nitrocellulose membrane, and incubated against anti-FLAG antibody (200472-21, Stratagene, La Jolla, CA). PageRuler Plus Prestained Protein Ladder (Thermo Fisher Scientific, Rockford, IL) was used to identify the molecular weight of FLAG-CRD-BP bands. The FLAG-CRD-BP bands were detected using the standard chemiluminescent technique and visualized using Alpha Innotech FluorChem 5500. The AlphaEaseFC software program was used to assign an Integrated Density Value to each of the FLAG-CRD-BP bands.

Quantitative Real Time PCR

The first strand cDNA synthesis was performed using iScript cDNA synthesis kit (QBio-Rad) on 1 μg of total RNA or an equi-volume of RNA samples, and the qPCR was performed using iQ SYBR Green Supermix (Bio-Rad) on an iQ5 Multicolor real time PCR detection system (Bio-Rad). The PCR primers synthesized by IDT Inc. were as follows: CD44 forward primer, 5′-CAT CAG TCA CAG ACC TGC CCA ATG C-3′, and CD44 reverse primer, 5′-ATG TAA CCT CCT GAA GTG CTG CTC C-3′; c-myc forward primer, 5′-ACG AAA CTT TGC CCA TAG CA-3′, and c-myc reverse primer, 5′ GCA AGG AGA GCC TTT CAG AG-3′; β-actin forward primer, 5′-TTG CCG ACA GGA TGC AGA AGG A-3′, and β-actin reverse primer, 5′-AGG TGG ACA GCG AGG CCA GGA T-3′. The cycling protocol was at 95 °C for 3 min with 40 cycles of denaturation at 95 °C for 10 s and annealing at 52 °C for 30 s. To confirm amplification specificity, we performed a melting curve analysis at the end of each cycle. Each sample was run in triplicate, and the data were analyzed using iQ5 optical system software. Serial dilutions were carried out for each total RNA sample and reverse-transcribed under the above-mentioned conditions for each primer set to ensure amplification with efficiencies near 100%. The C_T values for target genes (CD44, c-myc, and β-actin) were then used in the comparative C_T method or commonly known as the 2^−ΔΔCT method (29) to determine the expression level of the target gene in pcDNA CRD-BP-FLAG-transfected cells (WT and variants) and pcDNA-FLAG-transfected cells.

To ensure that there was no genomic DNA contamination in the cDNA samples used for the immunoprecipitation qPCR experiments, no-reverse transcriptase (no-RT) controls were used in each of the cDNA synthesis reactions. These samples were subjected to the same RT-qPCRs as the test samples, and genomic DNA contamination was assessed.

Data collected from four biological replicates were pooled, and one-way ANOVA statistical analysis was performed to compare mRNA levels in the variants to the wild-type CRD-BP.

Circular Dichroism Analysis

Circular dichroism (CD) spectra of all CRD-BP variants were attained using a Jasco J-815 CD spectrometer. Temperature during scanning was maintained at 25 °C using a Peltier temperature-control device, and proteins were sampled in a 0.1-cm path-length quartz cell. CD measurements were recorded using wavelength scans of 195–240 nm, with all protein sample concentrations between 1 and 2 μm resuspended in 20 mm Tris buffer, pH 7.4, 10% glycerol, 200 mm NaCl. Measurement parameters were standard resolution 100 millidegrees, bandwidth 1 nm, data integration time 2 s, scan rate 100 nm/min, and accumulation of eight scans. For data analysis, the secondary structure estimates were de-convoluted using the K2D3 neural networking algorithm (30).

Embryo Collection

The AB strain of zebrafish (D. rerio) (from the Zebrafish International Resource Centre, University of Oregon, Eugene) were maintained according to standard laboratory practices, and all experiments were approved by the HKUST Institutional Animal Care and Use Committee. Fishes were maintained on a 14-h light/10-h dark cycle to stimulate spawning, and their fertilized eggs were collected as described previously (31). Embryos were maintained in 30% Danieau's solution (17.4 mm NaCl, 0.21 mm KCl, 0.18 mm Ca(NO₃)₂, 0.12 mm MgSO₄·7H₂O, 1.5 mm Hepes, pH 7.2) at ∼28.5 °C throughout development and during all experiments.

Transient Expression of EGFP-CRD-BP, Microinjection, Microscopy, and Image Acquisition and Analysis of Zebrafish Embryos

For mRNA production, the WT CRD-BP-EGFP and various mutant CRD-BP-EGFP plasmids were linearized and transcribed using an mMESSAGE mMACHINE^TM (Ambion) in vitro transcription kit. For the transient expression of EGFP-CRD-BP fusion proteins, zebrafish embryos were injected with mRNA at the one-cell stage, unless otherwise stated. For acquisition of live images, embryos were immobilized in grooves made in a 1% agarose gel and bathed in 30% Danieau's solution (31). Most images were acquired using a Nikon C1 confocal system mounted on a Nikon 90i upright microscope using Nikon Fluor ×40/0.80-watt or ×60/1.00-watt water immersion objectives and a pinhole size of 60 μm. Alternatively, some images were acquired using a Carl Zeiss LSM 780 system configured on an inverted Observer Z1 microscope and a ×40 oil immersion objective. All images shown are single confocal sections of 512 × 512 pixels. Subsequent color and contrast enhancement was done using Corel PHOTOPAINT 11. Granules were quantified in all images using the “Analyze Particles” function in ImageJ software with a filter to detect granule size of 6–200 pixels² and circularity of 0.40–1.00. The number of granule is expressed per cell, and the number of cells in each image was counted manually.

RESULTS

Point Mutations in the GXXG Motif of Two KH Domains Generally Abolish the Ability of CRD-BP to Bind c-Myc and CD44 RNAs in Vitro

To assess the importance of each KH domain in binding to RNA in the context of the full-length protein, we generated single point mutations within the GXXG motif of each individual KH domain. We mutated the first glycine of each GXXG motif to an aspartate as a way to remove the function of each individual KH domain (Fig. 1A). For simplicity, we have used the following abbreviation to name the point mutations as KH variants. For instance, KH1 variant has the point mutation in the GXXG motif within the KH1 domain, whereas the KH2 variant has the point mutation in the GXXG motif within the KH2 domain. The KH1–2 variant has the point mutations at the GXXG motifs of both the KH1 and KH2 domains, whereas the KH3–4 variant has the point mutations in the GXXG motifs of the KH3 and KH4 domains, and so on (Fig. 1A). As negative controls, we also generated four CRD-BP variants with point mutations that are not expected to affect the RNA-binding ability. CRD-BP variants with point mutations in the RRM1 (Y5A) or RRM2 domain (Q84A) were selected (Fig. 1A). The E445D and D526E variants containing a point mutation in the variable loop located between β₂ and β₃ in KH3 and KH4 domains (21), respectively, were also generated (Fig. 1A). The recombinant mouse wild-type (WT) CRD-BP and mutant proteins were purified to ∼95% homogeneity (Fig. 1B).

FIGURE 1.

Generation of wild-type CRD-BP and CRD-BP point mutation variants. A, schematic representation of the mouse CRD-BP and its variants used in the gel shift experiments described in Figs. 2 and 3. The two RRM and four KH domains are shown. As shown, the first Gly of the GXXG motif of the KH domain was mutated to an aspartate. Single (KH1, KH2, KH3, and KH4) or double (KH1–2, KH1–3, KH1–4, KH2–3, KH2–4, and KH3–4) point mutation was performed to generate the KH CRD-BP variants. Single point mutation was performed to generate the Y5A, Q84A, E445D, and D526E variants. B, purification of recombinant His-tagged wild-type CRD-BP and its point mutation variants. Two μg of purified recombinant wild-type (WT) CRD-BP and its variants were run on a 12% SDS-PAGE followed by staining with Coomassie Brilliant Blue.

Using electrophoretic mobility gel shift assays, we compared the binding profiles of the WT CRD-BP and its variants on two high affinity target RNAs as follows: the 182-nt ³²P-labeled c-myc RNA (nts 1705–1886) (Fig. 2), and the 194-nt ³²P-labeled CD44 RNA (nts 2862–3055) (Fig. 3) (10, 32). Fig. 2A shows representative results from gel shift assays for binding the 182-nt c-myc RNA. It is clear that the KH1–2, KH1–3, KH1–4, KH2–3, and KH2–4 variants had completely abolished the ability to bind the c-myc RNA. In contrast, KH3–4 and the single point mutation variants KH1, KH2, KH3, and KH4, all exhibited binding to c-myc RNA (Fig. 2A). Results from three biological replicates were pooled, and the saturation binding data were fitted to the Hill equation and expressed graphically to determine the dissociation constant (K_d) and Hill coefficient. A summary, showing K_d for the interaction of each of the recombinant protein variants with c-myc RNA is shown in Table 1, and the results are plotted for direct comparison in Fig. 2B. As shown in the quantitative data, KH3–4 and KH1 displayed slightly higher K_d values as compared with the WT CRD-BP, whereas KH2, KH3, KH4, Y5A, Q84A, D526E, and E445D all displayed binding profiles that are comparable with the WT CRD-BP. However, KH1–2, KH1–3, KH1–4, KH2–3, and KH2–4 displayed significantly reduced or no binding to c-myc RNA based on data from three biological replicates (Fig. 2B).

FIGURE 2.

Binding profile of CRD-BP and its point mutation variants on the coding region determinant of c-myc RNA. A, electrophoretic mobility gel shift assay on the binding of purified recombinant WT CRD-BP and its point mutation variants to ³²P-labeled c-myc CRD RNA nts 1705–1886. Various concentrations of proteins, as indicated, were incubated with 40 nm of the radiolabeled c-myc RNA. The positions of protein-RNA complexes (Bound) and unbound RNA (Unbound) are indicated. Samples within each panel indicate the same experiment. B, summary of dissociation constants (K_d) of the WT CD-BP and its variants. The K_d values were taken from saturation binding curves (n = 4). The asterisk indicates that the p value is less than 0.05 based on Student's t test in comparing with the K_d value of WT CRD-BP. For KH3–4 the p = 0.0321 and for KH1 the p = 0.0122.

FIGURE 3.

Binding profile of CRD-BP and its point mutation variants on a specific 3′-untranslated region of CD44 RNA. A, electrophoretic mobility gel shift assay on the binding of purified recombinant WT CRD-BP and its point mutation variants to ³²P-labeled CD44 RNA nts 2862–3055. Various concentrations of proteins, as indicated, were incubated with 40 nm of the radiolabeled CD44 RNA. The positions of protein-RNA complexes (Bound) and unbound RNA (Unbound) are indicated. Samples within each panel indicate the same experiment with the exception of the panel showing WT and KH3–4. B, summary of dissociation constants (K_d) of the WT CD-BP and its variants. The K_d values were taken from saturation binding curves (n = 4). The asterisk indicates that the p value is less than 0.05 based on Student's t test in comparing with the K_d value of WT CRD-BP. p = 0.0371 for Y5A, p = 0.005 for D526E, p = 0.0121 for KH1, p = 0.0091 for KH2, and p = 0.0143 for KH3.

TABLE 1.

A summary of the dissociation constants (K_d) of WT CRD-BP in comparison with its KH variants for binding to c-myc CRD and CD44 RNAs

NA indicates not applicable because significantly reduced or no binding occurred and hence the binding curve cannot be plotted. ND indicates not determined. Data are means ± S.E.

CRD-BP variant	c-myc RNA nts 1705–1886		CD44 RNA nts 2862–3055
	Dissociation constant (Kd)	Hill coefficient	Dissociation constant (Kd)	Hill coefficient
WT	398.0 ± 52.79	1.768 ± 0.011	149.32 ± 8.42	1.916 ± 0.0080
KH1	723.1 ± 75.13	1.539 ± 0.004	211.18 ± 34.94	1.824 ± 0.0173
KH2	360.3 ± 70.41	1.680 ± 0.013	250.27 ± 26.72	1.756 ± 0.0089
KH3	320.8 ± 42.75	1.790 ± 0.012	219.32 ± 19.35	1.826 ± 0.0090
KH4	320.4 ± 59.08	1.787 ± 0.017	NA	NA
KH1–2	NA	NA	NA	NA
KH1–3	NA	NA	NA	NA
KH1–4	NA	NA	NA	NA
KH2–3	NA	NA	NA	NA
KH2–4	NA	NA	NA	NA
KH3–4	586.9 ± 42.89	1.728 ± 0.004	133.33 ± 11.04	1.849 ± 0.0112
Y5A	467.6 ± 45.68	1.755 ± 0.007	100.11 ± 16.41	1.978 ± 0.0297
D526E	402.4 ± 37.25	1.767 ± 0.007	90.65 ± 1.64	1.986 ± 0.0034
Q84A	490.6 ± 20.01	1.748 ± 0.003	ND	ND
E445D	459.5 ± 42.18	1.751 ± 0.006	ND	ND

Fig. 3A shows representative results from gel shift assays for binding of the CRD-BP variants to the 194-nt CD44 RNA. From the results, it is clear that the KH1–2, KH1–3, KH1–4, KH2–3, and KH2–4 mutants are unable to bind CD44 RNA. In contrast, the Y5A, D526E, KH1, KH2, KH3, and KH3–4 mutants all displayed the ability to bind CD44 RNA. The slightly different shift in the complex for KH3–4 as compared with the WT was because the two groups of samples were electrophoresed on different gels. KH4 appeared to bind CD44 RNA very weakly (Fig. 3A), and hence the saturation binding curve as well as the K_d value could not be obtained. A summary of the K_d values for the WT and mutant CRD-BP proteins that were obtained from the saturation binding data is shown in Table 1 and plotted for direct comparison in Fig. 3B. Both the Y5A and D526E mutants displayed slightly lower K_d values indicating their higher affinity for CD44 RNA, whereas KH1, KH2, and KH3 had slightly higher K_d values indicating their lower affinity for CD44 RNA. Results from three biological replicates confirmed that the KH4, KH1–2, KH1–3, KH1–4, KH2–3, and KH2–4 mutants have significantly reduced or no binding to CRD-BP (Fig. 3B).

In summary, we found that with the exception of KH3–4, simultaneous point mutation of the GXXG motif in two KH domains generally abolished the ability of CRD-BP to bind c-myc and CD44 RNAs. We also found that although the KH4 variant has high affinity for c-myc RNA, it was very inefficient in binding CD44 RNA. Similarly, the KH2 and KH3 variants have higher K_d values than that of the WT CRD-BP in binding CD44 RNA but not for binding c-myc RNA.

No Global Secondary Structure Changes Occur in CRD-BP upon Point Mutation of the KH Domain GXXG Motif

To determine whether there were major global changes in the secondary structure upon point mutation at the GXXG motifs, we performed circular dichroism spectroscopy on the WT CRD-BP as well as the KH domain mutants in the wavelength range of 195–240 nm. The “w-shaped” spectrum with a clear trough close to 208 nm and a less distinctive trough near 222 nm is indicative of the presence of α-helical structures. The CD spectrum of the entire CRD-BP protein is remarkably similar to that generated for the recombinant KH3&4di-domain of ZBP1 (22). All of the point mutation variants exhibited CD spectra that are very similar, if not identical, to that of the WT CRD-BP (data not shown). We also estimated the percentages of α-helix and β-sheet present in the WT CRD-BP and the KH variants, and we found no significant differences (data not shown). Hence, we conclude that point mutations in the GXXG motifs of the KH domains do not alter the global secondary structure of CRD-BP.

CRD-BP Granule Formation in Zebrafish Embryos Is Dependent on Newly Transcribed Zygotic RNAs

Next, we were interested in determining whether our findings that the RNA-binding ability of CRD-BP is abrogated upon introducing the GXXG motif point mutations in any two KH domains were translatable to an in vivo cellular system. Because the ability of IMP1 to bind RNA has been shown to be a pre-requisite for the formation of IMP granules in NIH3T3 cells (20), we decided to use CRD-BP granule formation as our first cellular system. We turned to the zebrafish embryo as an in vivo model because the optical clarity of this system is microscopically superior and thus facilitates the rapid identification of CRD-BP granules. Furthermore, CRD-BP granule formation in zebrafish embryos is likely to be visible within a few hours upon translation, allowing visualization of the protein-RNA interaction at early time points, which is not possible using a regular mammalian transfection system.

We first performed characterization experiments to explore and understand mouse CRD-BP granule formation in intact zebrafish embryos. We found that during the first 3 h in the absence of zygotic bio-molecules, CRD-BP granules were absent (Fig. 4A, panel i). However, at 4 h post-fertilization (hpf), CRD-BP granules were clearly visible (Fig. 4A, panel ii), suggesting that CRD-BP granule formation strictly coincides with the mid-blastula transition that indicates a switch from maternal to zygotic transformation. We also assessed granule formation upon injection with the human IMP1-EGFP mRNA. As shown in Fig. 4A, panels ii and iii, the granule formation by the mouse CRD-BP is remarkably similar to the human IMP1, indicating that granule formation by the two orthologs is very similar in zebrafish embryos. To determine whether CRD-BP granule formation is dependent on newly transcribed RNA, we injected the transcriptional inhibitor actinomycin D immediately after CRD-BP-EGFP mRNA injection. We found that actinomycin D completely abolished the formation of CRD-BP granules (Fig. 4B, panel ii), confirming that expression of zygotic genes is required for CRD-BP granule formation. To determine whether zygotic RNAs or proteins are important for granule formation, we injected the protein synthesis inhibitor cycloheximide either immediately after injection of CRD-BP-EGFP mRNA to inhibit total protein synthesis or at the 16-cell stage to allow expression of CRD-BP-EGFP before inhibition of zygotic protein synthesis. We found that the expression of CRD-BP-EGFP was inhibited by cycloheximide treatment if given immediately after mRNA injection (Fig. 4B, panel iii). However, if cycloheximide was given at a later stage to block translation of the earliest zygotic transcripts, CRD-BP granules were still present (Fig. 4B, panel iv), suggesting that zygotic RNAs rather than proteins are critical for the formation of CRD-BP granules.

FIGURE 4.

CRD-BP granule formation in zebrafish embryos. A, granule formation coincides with maternal zygotic transcription. Embryos were injected with mouse CRD-BP-EGFP mRNA at the one-cell stage and then imaged at 3 hpf (panel i) or 4 hpf (panel ii). For comparison, the human IMP1-EGFP mRNA was injected and imaged at 4 hpf (panel iii). B, effect of transcriptional inhibitor or protein synthesis inhibitor on CRD-BP granule formation. After injection with CRD-BP-EGFP mRNA at the one-cell stage, the embryos were injected with actinomycin D (Act D) (20 μg/ml) immediately (panel ii). Some embryos were injected with cycloheximide (CHX) (20 μg/ml) immediately (panel iii) or at 16-cell stage (panel iv). Embryos were then imaged at 4 h post-fertilization to examine the effect of drug treatment on CRD-BP granule formation. Data shown are representatives of three biological replicates. Scale, 10 μm.

KH3&4 Di-domain Is Sufficient for CRD-BP Granule Formation in Zebrafish Embryos

To determine whether the requirement for CRD-BP granule formation in zebrafish embryos is similar to that previously observed in mouse NIH3T3 embryo fibroblast cells (20), we first assessed the deletion variants of CRD-BP-EGFP. A schematic diagram showing all deletion variants is illustrated in Fig. 5A, and representative results of CRD-BP granule formation are shown in Fig. 5B. Injection of RRM1&2-EGFP or KH1&2-EGFP mRNAs failed to result in any granule formation (Fig. 5, B and C), which is consistent with previous results in NIH3T3 cells (20) and in several studies which concluded that the RRM domains and KH1&2 domains alone are insufficient for interaction with target RNAs (19–21). Fig. 5B shows that KH1 to 4-EGFP granule formation is comparable with that of the WT CRD-BP-EGFP, and this is further confirmed by quantifying the number of granules per cell as shown in Fig. 5C. This is in good agreement with the previous finding that KH domains 1–4 are sufficient to assemble granules (20) and bind target RNAs (19, 20). Injection of di-KH3&4-EGFP mRNA, but not KH4-EGFP mRNA, also resulted in granule formation (Fig. 5B). Surprisingly, we found a significantly higher number of granules per cell in di-KH3&4-EGFP-expressed embryos as compared with embryos expressing the WT CRD-BP-EGFP (Fig. 5C). This result is consistent with findings that KH domains 3–4 can bind target RNA efficiently (21). However, a previous study showed no granule formation in NIH3T3 cells upon expressing GFP-KH3–4, and KH3–4 of IMP1 could not bind H19 RNA (20). The deviation in granule formation results could be due to one or a combination of the following: (i) cellular differences between zebrafish embryos and NIH3T3 cells; (ii) time differences and hence differences in the dynamics of CRD-BP-RNA interaction; or (iii) granules are less distinguishable in NIH3T3 cells. The deviation in RNA-binding ability of KH3–4 is most likely due to differences in RNA targets used in the different studies, as we have observed the different contribution of each KH domain of CRD-BP in binding different RNA molecules (Figs. 2 and 3) (data not shown). There was an accumulation of di-KH3&4-EGFP in the nuclei of zebrafish embryos, which was not seen with KH1 to 4-EGFP (Fig. 5B), suggesting that there may be a nuclear export signal located within KH domains 1 and 2. In support of this hypothesis was the observation of a similar nuclear accumulation of KH4-EGFP (Fig. 5B). Indeed, Nielsen et al. (33) have reported that there are IMP1 nuclear export signals located in KH2 and KH4. We also examined the expression of other single KH domains, namely KH1-EGFP, KH2-EGFP, and KH3-EGFP, but again no granule formation in zebrafish embryos was observed (Fig. 5, B and C). We also noted that KH1-EGFP accumulated in the nuclei, although KH3-EGFP accumulated in the cytoplasm, and KH2-EGFP appeared to be evenly distributed throughout the nuclei and cytoplasm (Fig. 5B). However, we cannot rule out possible differences in cell cycle when these images were acquired. Finally, to determine whether the granule formation by a combination of two KH domains is unique to di-KH3&4-EGFP, we examined all other possible di-domain combinations. Fig. 5, B and C, show that di-KH1&3-EGFP, di-KH1&4-EGFP, di-KH2&3-EGFP, and di-KH2&4-EGFP did not show any distinct granule formation, although some of these di-domains appeared to be more patchy (Fig. 5B). Hence, we conclude that KH3&4di-domain is the minimal region of CRD-BP necessary for granule formation.

FIGURE 5.

Comparing truncated CRD-BP variants on granule formation in zebrafish embryos. A, schematic representation of the mouse CRD-BP and its truncated variants used in this study. B, embryos were injected with the WT CRD-BP-EGFP or various truncated variant mRNAs as shown at the one-cell stage. Embryos were then imaged at 4 hpf. Data shown are representatives of three biological replicates. Scale, 10 μm. C, number of granules per cell were counted as described under “Experimental Procedures” and expressed as shown in the bar graph. At least three images per sample were counted from two separate mRNA injections. One-way ANOVA was performed as statistical analysis. The asterisk indicates p < 0.01 when compared with the WT CRD-BP.

Point Mutations in the GXXG Motif in Two KH Domains Generally Reduces CRD-BP Granule Formation in Zebrafish Embryos

Having shown that CRD-BP granule formation in zebrafish embryos indeed correlated with the known RNA-binding ability among CRD-BP deletion variants, we next investigated the role of each KH domain in CRD-BP granule formation in zebrafish embryos in the context of the full-length mutant proteins. To do this, we generated pSp64TNE-CRD-BP plasmid constructs containing the point mutation in the GXXG motif of each of the individual KH domains of CRD-BP. The KH domain point mutation variants generated were those used in the EMSA studies described earlier (Figs. 2 and 3). As shown in Fig. 6A, granule formation for the single point mutants was indistinguishable from that of the WT CRD-BP. Results corresponding to CRD-BP granule formation for the KH domain double mutants are shown in Fig. 6A, and the quantitative data are shown in Fig. 6B. Interestingly, KH3–4 was the only double mutant that exhibited comparable granule formation to the WT CRD-BP (Fig. 6, A and B). Granule formation was significantly reduced when zebrafish embryos were injected with KH1–3-EGFP, KH1–4-EGFP, KH2–3-EGFP, or KH2–4-EGFP mRNAs (Fig. 6, A and B). Granule formation with KH1–2 was also reduced but did not reach statistically significant levels (Fig. 6, A and B). In summary, we observed that the CRD-BP granule formation in zebrafish embryos by the KH point mutation variants correlated remarkably with their ability to bind c-myc and CD44 RNA in vitro (Table 1).

FIGURE 6.

Comparing CRD-BP point mutation variants on granule formation in zebrafish embryos. A, embryos were injected with mRNA for the single KH point mutation CRD-BP, KH double point mutation CRD-BP, or WT CRD-BP-EGFP at the one-cell stage. Embryos were then imaged at 4 hpf. Data shown are representatives of three biological replicates. Scale, 10 μm. B, number of granules per cell were counted as described under “Experimental Procedures.” At least three images per sample were counted from three separate mRNA injections. One-way ANOVA was performed as statistical analysis. The asterisk indicates p < 0.05 when compared with the WT CRD-BP.

Point Mutation at the GXXG Motif in KH Domains Generally Reduces the Ability of CRD-BP to Associate with c-myc and CD44 mRNAs in HeLa Cells

Next, we were interested in determining whether the point mutations in the GXXG motif of the KH domains had any effect on the ability of CRD-BP to physically interact with CD44 and c-myc mRNAs in mammalian cells. We transfected HeLa cells with pcDNA-CRD-BP-FLAG plasmids containing the point mutations as described in Fig. 1A. The pcDNA-FLAG and pcDNA-CRD-BP-FLAG plasmids containing Y5A, D526E, Q84A, or E445D mutations were used as negative controls. No genomic DNA contamination was observed in the immunoprecipitated RNA samples. This was indicated by a lack of PCR product amplified in the no-RT qPCRs (data not shown).

As shown in Fig. 7A, with the exception of E445D and Q84A variants, equal amounts of FLAG-CRD-BP were immunoprecipitated from cells transfected with different CRD-BP variants with single or double point mutations. All expressed FLAG-CRD-BP at approximately equal amounts based on similar volumes of cell lysate used for immunoprecipitation. The equal integrated density values obtained from quantification using the AlphaEaseFC software program confirmed this observation. The reason for the lack of FLAG-CRD-BP detected in the Q84A and E445D variants is unknown but could possibly be the result of a masked FLAG tag as a consequence of structural changes. Using equal amounts of immunoprecipitated proteins, we measured CD44, c-myc, and β-actin mRNAs that were physically associated with FLAG-CRD-BP (Fig. 7B). As shown, the level of all three transcripts associated with FLAG-CRD-BP in cells transfected with Y5A and D526E variants were similar to those in WT CRD-BP. In contrast, the level of CD44 mRNA associated with FLAG-CRD-BP was significantly reduced in cells transfected with any of the KH domain point mutation variants. Similarly, with the exception of the KH4 variant, the amount of c-myc mRNA associated with FLAG-CRD-BP was also significantly reduced in cells transfected with KH variants. On the contrary, the levels of β-actin mRNA physically associated with FLAG-CRD-BP in the WT CRD-BP- and KH mutant-transfected cells were very similar except for the KH3, KH1–4, and KH2–4 variants. In summary, we conclude that the GXXG mutants generally retarded the ability of CRD-BP to bind CD44 and c-myc mRNAs in HeLa cells. The results also suggested that, as observed in vitro, there is a differential contribution of the CRD-BP KH domains to binding its various mRNA targets in HeLa cells.

FIGURE 7.

Comparing mRNA levels that are physically associated with CRD-BP and its point mutation variants in HeLa cells. HeLa cells were transfected with plasmid vector pcDNA3-FLAG, pcDNA3-FLAG WT CRD-BP, or various pcDNA3-FLAG plasmids carrying CRD-BP variants as shown. Forty eight hours after transfection, lysates from cells were isolated and subjected to immunoprecipitation using anti-FLAG antibody as described under “Experimental Procedures.” A, 16 μl of immunoprecipitated FLAG-CRD-BP proteins were subjected to Western analysis using anti-FLAG antibody as shown. The FLAG-CRD-BP bands from each transfection were quantified as integrated density value (IDV) and expressed relative to the WT CRD-BP. B, RNAs that were physically associated with pulled down FLAG-CRD-BP were extracted and subjected to qRT-PCR for measurements of CD44, c-myc, and β-actin mRNAs. mRNA levels were expressed relative to the pcDNA3-FLAG vector. The data collected from four biological replicates were then pooled and expressed relative to the WT CRD-BP, which was taken as 1.0. One-way ANOVA statistical analysis was then performed to compare the single (middle panel) and double (right panel) KH variants to the WT CRD-BP as follows: CD44 mRNA/single KH variants, n = 5, F = 21.08, *, p < 0.0001; CD44 mRNA/double KH variants, n = 7, F = 22.43, *, p < 0.0001; c-myc mRNA/single KH variants, n = 5, F = 10.36, **, p = 0.0003; c-myc mRNA/double KH variants, n = 7, F = 13.01, *, p < 0.0001; β-actin mRNA/single KH variants, n = 5, F = 0.909, p = 0.4837; β-actin mRNA/double KH variants, n = 7, F = 1.005, p = 0.4579.

DISCUSSION

A number of animal studies have now confirmed the critical role played by CRD-BP during embryogenesis and tumorigenesis (18, 34–37), but unfortunately, the exact molecular mechanism of this onco-fetal protein is still unclear. However, cumulative evidence from studies in vitro and in cell lines suggests that the ability to physically interact with a subset of mRNAs plays a significant part in the function of CRD-BP and its orthologs. Hence, it is of the utmost importance to clearly understand how CRD-BP interacts with its RNA substrates. To date, in vitro and granule formation studies in mammalian cells using CRD-BP orthologs have shown that the KH domains, and not the RRM domains, are directly involved in binding RNA substrates (19, 20). However, the extent to which each of the KH domains of CRD-BP contributes to the physical interaction with RNA is less clear. This is partly because most of the studies were based on deletion analysis. The RNA binding function of the KH domains for a number proteins have been shown to be severely impeded upon mutating the first glycine in the GXXG motif to an aspartate (24–27). Using a similar approach, we investigated the role of each of the KH domains in the context of the full-length protein in binding c-myc and CD44 RNAs in vitro and in cells, in addition to granule formation in intact zebrafish embryos.

Using electrophoretic mobility gel shift assays, we found that a single point mutation in the GXXG motif at each of the four KH domains generally had no impact on the ability of CRD-BP to bind c-myc and CD44 RNAs (Figs. 2 and 3). The only exception to this was an inefficient binding of the KH4 variant to CD44 RNA. However, point mutations in the GXXG motif of any combination of two KH domains, with the exception of the simultaneous mutation at both KH3 and KH4 domains (KH3–4), resulted in complete abrogation of RNA binding (Figs. 2 and 3). CD spectral analysis showed no global secondary structural changes for the mutant proteins. This is also consistent with a recent finding that the GXXG-to-GDDG mutation in the four KH domains of KSRP as well as in the KH3 and KH4 domains of ZBP1 did not destabilize the KH domains (28). However, we do not rule out the possibility of local structural changes in the KH domains of the point mutants, which cannot be detected using CD spectroscopy. This may explain why the KH4 single mutant has significantly reduced binding, although the KH3–4 double mutant binds CD44 RNA efficiently. It is possible that for this specific double mutant, local structural changes at regions outside the GXXG motif within KH3 and KH4 facilitate binding to CD44 RNA. Further structural experiments that are beyond the scope of this study are required to resolve this apparent anomaly.

The ability of IMP1 to bind RNA has been considered a pre-requisite for the formation of IMP1 granules in mammalian cells (20). The IMP1 granules are a ribonucleoprotein complex that is believed to form leading to functional IMP1 involved in mRNA stability (9) and granule localization/transport (20). The granules collected from HEK293 human embryonic kidney cells 48 h after transfection with plasmid carrying FLAG-IMP1 are distinct from stress granules and P bodies, and they are enriched in mRNAs encoding proteins involved in endoplasmic reticulum and Golgi function (38). Therefore, we investigated whether our in vitro EMSA data with the KH point mutants correlate with CRD-BP granule formation by the variants. Here, we provide the first description of CRD-BP granule formation in intact zebrafish embryos and demonstrate that CRD-BP granule formation is dependent on zygotic mRNAs and not maternal mRNAs (Fig. 4). The zygotic mRNA dependence of CRD-BP granule formation is in-line with the onco-fetal role of this protein, which is overexpressed in neonatal tissues (39). Also, we find that the ability of the various mutant proteins to bind RNA in vitro is a good indication for CRD-BP granule formation as demonstrated using intact zebrafish embryos. Truncated CRD-BP-EGFP, such as RRM1&2 and KH1&2, which have been shown to have no ability to bind RNA (19, 20), could not form granules in zebrafish embryos. However, KH1–4 CRD-BP-EGFP, which binds RNA in vitro (19, 20), resulted in granule formation that was comparable with the WT CRD-BP (Fig. 5). We found a significantly greater number of granules per cell in embryos expressing the truncated KH3&4CRD-BP-EGFP as compared with the WT CRD-BP (Fig. 5C), which is consistent with its ability to bind RNA efficiently as reported previously (21). Such an effect is unique to the KH3&4 di-domains because all of the other possible KH di-domain combinations did not show granule formation (Fig. 5). On examining the KH point mutants for granule formation, we found a remarkable correlation between the ability to bind c-myc and CD44 RNAs in vitro and CRD-BP granule formation in zebrafish embryos. For instance, the single point mutants generally did not display reduced granule formation or RNA binding (Table 2). In contrast, with the exception of KH3–4, the KH domain double mutants displayed a significant reduction in both granule formation and RNA binding. Based on these observations, we propose that at least two KH domains of CRD-BP bind to RNA in tandem. This is consistent with a recent x-ray crystallography study, whereby the individual domains in the KH3&4 di-domain were shown to be arranged in an intramolecular anti-parallel pseudodimer conformation, which requires that the bound RNA must loop to contact both KH domains simultaneously (21). Also, in agreement with the previous study is the observation that the KH3&4 di-domain is important for binding RNA and that sites other than the GXXG motif within the KH3 and KH4 domains play a critical role in binding RNA substrates. This is because we found that a point mutation in the GXXG motif in both the KH3 and KH4 domains in combination had no effect on CD44 and c-myc RNA binding as well as granule formation.

TABLE 2.

A summary of the electrophoretic mobility shift assay, zebrafish granule formation, and immunoprecipitation-coupled qPCR experiments

RNA	Type of KH variants	In vitro EMSAa (<1 h)	Granule formation in zebrafish embryosb (4 h)	mRNA associated with FLAG-CRD-BP in HeLa cellsc (48 h)
c-myc	Single KH point mutation	All bind	+++	All reduced except KH4
	Double KH point mutation	All no binding except KH3–4	All defect except KH3–4	All reduced
CD44	Single KH point mutation	All bind except KH4	+++	All reduced
	Double KH point mutation	All no binding except KH3–4	All defect except KH3–4	All reduced

^a Data were taken from Figs. 2 and 3.

^b Data were taken from Fig. 6. The granule formation does not focus on specific mRNA species and the row for c-myc and CD44 is repeated. +++ indicates normal granule formation compared with WT CRD-BP.

^c Data were taken from Fig. 7B. Reduced means mRNA species were found to have a reduced level associated with the corresponding FLAG-KH variants as compared with its association with the FLAG-WT CRD-BP.

However, based on our point mutation studies, we further observe that KH1 and KH2 also play a critical role in binding RNA substrates. Our study revealed that KH1 and KH2 domains also participate with at least another KH domain in binding RNA. The role for KH1 and KH2 in RNA binding is also supported by two earlier indirect observations as follows: (i) unlike the full-length protein, KH3&4 does not associate with RNA below 100 nm (11); and (ii) the KH1&2 di-domain modulates binding of IMP1 to β-actin and c-myc RNA in vitro (2). Furthermore, using a filter binding assay, significant reduction in RNA binding was only observed upon GXXG to GEEG mutation in all four KH domains of ZBP1 (40). Although there is generally a similar trend in binding c-myc and CD44 RNAs by the KH variants, there are a few notable deviations. Although the KH4 variant binds c-myc RNA with a K_d value similar to that of WT CRD-BP, it did not bind CD44 RNA efficiently, and hence no K_d value can be calculated. Similarly, the KH2 and KH3 variants bind to CD44 RNA less efficiently than to c-myc RNA. These results suggest that indeed CRD-BP utilizes distinct KH domains when binding to different RNA substrates.

The investigation into the physical interaction between CRD-BP mutants with c-myc, CD44, and β-actin mRNAs using immunoprecipitation coupled with the qPCR method yielded somewhat surprising results. Consistent with the in vitro EMSA and CRD-BP granule formation results, significant reduction in both c-myc and CD44 mRNAs, but not the β-actin mRNA, were found associated with the KH domain double mutants. Unexpectedly, we found that the KH3–4 variant also had a significantly reduced amount of CD44 and c-myc mRNAs physically associated with it. Similarly, a single point mutation in just one KH domain significantly reduced the level of CD44 and c-myc mRNAs binding to it, with the exception of the KH4 variant, which was still able to bind c-myc mRNA. We reasoned that the biochemical method employed in this study is valid because Y5A and D526E variants that bind RNA as effectively as the WT CRD-BP in vitro (Figs. 2 and 3) had levels of c-myc and CD44 mRNAs comparable with that of WT CRD-BP in the immunoprecipitation experiments. Furthermore, except for KH1–4, KH2–4, and KH3 variants, there was no significant reduction in the amount of β-actin mRNA associated with all the other KH variants. The surprising results with KH3–4 and the single point mutants that were different from the in vitro EMSA and granule formation results may be potentially explained by the following: CRD-BP is known to associate with various RNA-binding proteins and mRNAs in ribonucleoprotein complexes (9, 38). Hence, it is not unexpected to find differences in the dynamics of CRD-BP-RNA interactions 48 h after transfection as performed in the HeLa cell experiments as opposed to the direct in vitro EMSA and in the granule formation experiments in zebrafish, which occurred 4 h after injection of mRNAs (Table 2). Such a proposal is supported by an earlier report of the lack of granule formation in NIH3T3 cells 48 h after transfection with the KH3&4-GFP di-domain (20). The immunoprecipitation experiments also revealed the following novel findings: (i) different mRNAs bind to CRD-BP mutants variably in cells as exemplified by the interaction of c-myc mRNA with the KH4 variant and β-actin mRNA binding to the KH3, KH1–4, and KH2–4 variants, and (ii) the individual KH domains play an important role in the physical interaction between CRD-BP and specific mRNA in cells. Indeed, our data support an earlier proposal that individual KH domains may have distinct functions. This was based on multiple sequence alignment analyses of CRD-BP and its orthologs, which indicate a higher degree of amino acid sequence similarity between the equivalent KH domains from different proteins than between KH domains within the same protein (19).

Understanding the RNA sequences and/or structures that are recognized by CRD-BP is important to uncovering its cellular function. Consensus primary RNA sequences have been identified using various methods and were proposed to be targets of IMP1 or ZBP1. Using the SELEX method, the consensus sequence 5′-ACACCC-3′ was found in RNAs bound to the full-length ZBP1 as well as to the truncated KH3–4 domains of ZBP1 (41). Using the immunoprecipitation method coupled with microarray analysis, it was found that 307 transcripts associated with IMP1 granules were enriched in CCYHHCC (where Y = C or U and H = C, U, or A) motif (38). More recently, the PAR-CLIP method was used to assess RNA-binding protein recognition elements for FLAG/HA-tagged IMP1, IMP2, and IMP3 (42). An RNA recognition element CAUH (H = A, U, or C) was found in more than 75% of the top 1000 clusters from 100,000 sequence clusters recognized by the IMPs (43). Interestingly, our initial analysis of the secondary structure of the regions of c-myc (43) and CD44 (16) RNA that bind with high affinity to CRD-BP showed that sequences located in single-stranded regions have some potential overlap with the previously reported consensus sequences recognized by IMP1/ZBP1. For instance, 5′-AAACA-3′ at nts 20–24, 5′-ACAG-3′ at nts 63–66, and 5′-ACA-3′ at nts 70–72 are located in single-stranded regions of c-myc RNA and contain a partial match to the consensus sequence ACACCC. Similarly, 5′-CCCAAUU-3′ is located at nts 11–17 in a single-stranded region of CD44 RNA and is close to matching the consensus sequence CAUH. This shows that there is no strict “consensus” sequence for binding RNA and suggests that secondary and tertiary structural elements are likely key for interaction. Therefore, further in silico analysis of structural elements within target RNAs and crystallographic studies on the CRD-BP-RNA complex are required to fully elucidate the factors governing molecular recognition.

Using site-directed mutagenesis in the context of the entire protein, this study shows for the first time that at least two KH domains of CRD-BP act in tandem for efficient binding to two oncogenic mRNAs, c-myc and CD44, as well as for early RNP granule formation in zebrafish embryos. These results show that in vitro EMSA is indeed a valid method for assessing specific CRD-BP-RNA interactions that is translatable to an in vivo system, as shown for RNP granule formation in zebrafish embryos. This study supports the notion that mutating the GXXG motif within the KH domains (28) is a powerful tool for investigation into the function of individual KH domains in RNA-binding proteins. We show that mutating the first glycine to an aspartate in the GXXG motif is sufficient for such purposes. We also show for the first time that each KH domain is required for binding c-myc and CD44 mRNAs at the later stage of RNP formation as observed in the HeLa cells. Based on our overall data, we propose that the dynamics of the CRD-BP-mRNA interaction vary over time in vivo. This study has important implications for our understanding of the oncogenic mechanism of CRD-BP as well as in the future design of inhibitors targeted against CRD-BP function.