Two Isoforms of the RNA Binding Protein, Coding Region Determinant-binding Protein (CRD-BP/IGF2BP1), Are Expressed in Breast Epithelium and Support Clonogenic Growth of Breast Tumor Cells*

Background: The RNA binding protein, coding region determinant-binding protein (CRD-BP), is expressed by tumor cells and protects key mRNAs. Results: This study identifies a novel variant of CRD-BP and finds that CRD-BP is required for breast tumor cell clonogenicity. Conclusion: CRD-BP has tumorigenic activity and is ubiquitously expressed in breast epithelium. Significance: Under-reporting of CRD-BP isoforms suggests that published studies may be incomplete. CRD-BP/IGF2BP1 has been characterized as an “oncofetal” RNA binding protein typically highly expressed in embryonic tissues, suppressed in normal adult tissues, but induced in many tumor types. In this study, we show that adult breast tissues express ubiquitous but low levels of CRD-BP protein and mRNA. Although CRD-BP mRNA expression is induced in breast tumor cells, levels remain ∼1000-fold lower than in embryonic tissues. Despite low expression levels, CRD-BP is required for clonogenic growth of breast cancer cells. We reveal that because the most common protein isoform in normal adult breast and breast tumors has an N-terminal deletion (lacking two RNA recognition motif (RRM) domains) and is therefore missing antibody epitopes, CRD-BP expression has been under-reported by previous studies. We show that a CRD-BP mutant mouse strain retains expression of the shorter transcript (ΔN-CRD-BP), which originates in intron 2, suggesting that the impact of complete ablation of this gene in mice is not yet known. Either the full-length CRD-BP or the N-terminally truncated version can rescue the clonogenicity of CRD-BP knockdown breast cancer cells, suggesting that clonogenic function is served by either CRD-BP isoform. In summary, although CRD-BP expression levels are low in breast cancer cells, this protein is necessary for clonogenic activity.

The RNA binding protein, coding region determinant-binding protein (CRD-BP) 2 , was originally isolated as the protein responsible for stabilizing mRNA for the oncogene, c-Myc, by binding to a sequence in the coding region of the transcript (1). Since then, CRD-BP overexpression has been shown to be sufficient for mammary tumor induction in transgenic mice (2). Knockdown studies have shown that CRD-BP is essential for survival of many different types of tumor cells (3)(4)(5)(6), and CRD-BP is highly expressed in melanoma, breast, ovarian, and colorectal cancers (reviewed by Bell et al. (7), among others). By way of a molecular explanation, CRD-BP has been shown to regulate many mRNAs encoding cancer-associated genes, including Gli1, PTEN, ␤TrCP1, MAPK4, MDR1, IGF2, H19, c-myc, CD44, and ␤-catenin mRNAs (1, 8 -16). However, the remarkably universal requirement for CRD-BP expression by such disparate tumor types is not yet understood.
CRD-BP is also known as IGF2BP1, ZBP1, and IMP1. The variety of names ascribed to the same protein illustrates the fact that investigators from various fields have identified distinct activities for the same molecule. Furthermore, CRD-BP is a member of the highly conserved family of RNA binding proteins known as VICKZ proteins (17), which are structurally composed of two RNA recognition motifs at the N terminus and four K homology (KH) domains at the C terminus. Target mRNAs bind CRD-BP KH domains via combinatorial interactions through a looped tertiary structure with short consensus sequences. This interaction makes the mRNAs difficult to predict a priori (18,19), but experimental results using overexpressed CRD-BP suggest there may be as many as 300 -900 different mRNA species in CRD-BP-associated granules, which are ϳ100 -300 nm in diameter (20,21). Through its mRNA binding activity, CRD-BP has been shown to affect RNA stability (C-MYC and BTRC), translatability (IGF2), and/or localization (ACTB) (1,10,12,22) and thus regulate cell survival, migration, and chemoresistance.
Here we find that the typical oncofetal expression pattern reported for CRD-BP is not observed in breast tissue. Instead, all adult breast epithelial cells express CRD-BP. Interestingly, although CRD-BP expression is retained in breast tissue, it is not exclusively the canonical, full-length CRD-BP previously described. Instead, we identify and describe a protein variant (⌬N-CRD-BP) as a common and sometimes predominant isoform in breast epithelial cells. The mRNA transcript for this protein variant is missing N-terminal exons encoding two conserved RNA binding domains. Because some anti-CRD-BP antibodies are directed to N-terminal epitopes, previous studies may have under-reported CRD-BP protein expression. Furthermore, in adult tissues, CRD-BP mRNA is expressed at only 0.1-1% of the levels found in embryonic tissues, and this has also led to under-reporting of mRNA expression. Despite this relatively low expression level, we show that CRD-BP is required for clonogenic growth and that either the full-length or N-terminally truncated isoform of CRD-BP can rescue clonogenicity of CRD-BP knockdown breast cancer cells.
Plasmids and Reagents-Expression constructs were created for mouse and human full-length CRD-BP and ⌬N-CRD-BP in the pHIV-MCS-IRES-dTomato backbone (Addgene plasmid 21374). Full-length mouse CRD-BP sequence was obtained by digesting pSPORT1-mIGF2BP1 (Open Biosystems, clone 30008106). Human full-length CRD-BP was obtained by digesting pcDNA3.1-hIGF2BP1 (25). Expression constructs for mouse and human ⌬N-CRD-BP were generated using PCR products from WT MEF and HEK293T cDNAs for mouse and human, respectively, using the following primers: SmaI-AGC-CCCGGGGGCAAGCCATCATGAAGCTGA, and BamHI-TTGGATCCCGTTGTTGTCCTGGACTCGAAG for human,  and SmaI-TAACCCGGGGCAGACCAGGCAAGCTATC, and BamHI-CCTGGATCCGATCTTGGAGCCAATGGGAC for mouse. PCR products were inserted into the pHIV-MCS-IRES-dTomato backbone. Mouse CRD-BP-specific and control shRNA constructs in the pGIPZ backbone were purchased from Open Biosystems (clones V2LMM_66224, V2LMM_ 64131, and V2LMM_77760). Human CRD-BP-specific and Scrambled shRNA constructs in the pSilencer TM 1.0-U6 backbone were described in previous publications (10).
Transient Transfections, Generation of Lentiviral Particles, and Viral Transductions-pGIPZ lentiviral mouse CRD-BPspecific and control shRNA vectors (Open Biosystems; clones V2LMM_66224, V2LMM_64131, and V2LMM_77760) were transfected along with vesicular stomatitis virus G-protein and viral polymerase/core protein constructs into HEK293T cells (ATCC) using Lipofectamine LTX with Plus reagent (Life Technologies) according to the manufacturer's instructions. Virus particle-containing medium was harvested at 48 and 72 h post-transfection, and filtered using 0.45-m filters (Thermo Fisher Scientific) prior to use in transduction in the presence of 6 g/ml Polybrene (Sigma-Aldrich).
Immunofluorescence Staining and Confocal Microscopy-Cells were plated in 4-well chamber slides (Thermo Fisher Scientific) and then fixed/permeabilized using ice-cold methanol for 4 min, followed by ice-cold acetone for 2 min. Cells were blocked in 10% normal goat serum (Jackson Immunoresearch) for 2 h, incubated with 1 o antibodies at 4°C overnight, washed three times for 5 min in TBS-Tween (0.05%), and incubated with 2 o antibodies for 2 h at room temperature. The 1 o antibodies and dilutions used were: anti-CRD-BP (Abcam catalog no. ab82968; Cell Signaling catalog no. 8482; VS (25)) all at 1:100. The 2 o antibodies and dilutions used were: anti-mouse-Alexa 488 (Life Technologies) 1:200; anti-rabbit-Alexa 488 (Life Technologies) 1:200. Imaging was performed using a confocal microscope (Bio-Rad MRC1024).

Reverse Transcription and (Quantitative) Real Time PCR-
Total RNA was isolated from cells using the RNeasy mini kit (Qiagen). Reverse transcription and real time PCR were performed as previously described (23). Analysis was performed on each sample in duplicate using an ABI 7900HT Fast Real-Time PCR System (Applied Biosystems). Relative transcript levels were calculated using the comparative C t method (27) and normalized to housekeeping genes HPRT1, Y-Whaz, or TBP. Mouse and human CRD-BP qRT-PCR primer sequences are listed in Table 1. Mouse and human CRD-BP end-point PCR primer sequences are listed in Table 2.
Colony Formation Assay-Cells were plated at 10 4 cells/well in 6-well plates 48 h post-transfection with shRNA constructs. 96 h to 1 week after plating, cells were washed with PBS, fixed in ice-cold methanol for 20 min, washed again with PBS, and incubated for 1 h with 1% crystal violet solution (dissolved in 50% methanol). Plates were rinsed four times in distilled H 2 O and allowed to dry overnight before imaging.
Tumor Microarray Processing, Imaging, and Analysis-Human breast tumor microarray slides (University of Wisconsin Carbone Cancer Center) were deparaffinized and rehydrated. VS anti-CRD-BP antibody was diluted 1:1000 in Da Vinci Green diluent (Biocare Medical) and applied to the slides for 1 h. The slides were then rinsed and incubated for 20 min with Mach2 mouse HRP polymer (Biocare Medical) followed by another rinse and incubation with betazoid diaminobenzidine chromagen (Biocare Medical) for 1 min (all at room temperature). Slides were counterstained with CAT hematoxylin (Biocare Medical) and automatically scanned using the Vectra imaging system (PerkinElmer Life Sciences). All subsequent analysis was performed using the InForm analysis software (PerkinElmer Life Sciences).
Exon Usage Bioinformatics Analysis-IGF2BP1 exon data (reads per kilobase per million mapped reads) and tumor classification (PAM50) was extracted from the Cancer Genome Atlas data set (TCGA_BRCA_exp_HiSeqV2_exon-2014-08-28) (29) using a custom Python script. The mean and standard error of the mean reads per kilobase per million mapped reads (RPKM) values for IGF2BP1 exons 1-15 were calculated for each PAM50 subtype. The data were plotted using Prism software.

Identification and Characterization of the Novel ⌬N-CRD-BP
Isoform in Mouse and Human Cells-We identified a discrepancy between the results of different primer sets designed to assess CRD-BP mRNA expression. To investigate this further, we performed 5Ј-RACE and found three separate transcripts, with differing 5Ј ends, in the HC11-derived mouse mammary EN cell line. One of the transcripts includes the sequence for full-length CRD-BP, including all 15 exons (NCBI accession NM_009951.4). The sequences for the other two products include the annotated exon 3 but do not include exons 1 or 2; instead they include parts of intron 2 (21,609 -21,721 and 23,093-23,174 bp, relative to the start of the annotated intron 2). This suggests there are alternate, shorter transcripts (start sites labeled Fi2a and Fi2b; Fig. 1A). In these shorter transcripts, we identified a candidate ATG that could initiate translation in exon 6, which is predicted to encode a CRD-BP isoform missing the N-terminal domain (amino acids 1-136), including the first two RNA binding domains (Fig. 1A, compare full-length and ⌬N-CRD-BP structures). Interestingly, the 5Ј-RACE products we identified match CRD-BP alternative transcripts predicted by Gnomon gene prediction software (NCBI).
To further validate alternative CRD-BP mRNAs, RT-PCR primers were designed to test exon linkage for mRNA transcripts in several mouse and human cell lines. Amplification of products from PCRs with forward primers in exon 1 (F1), exon 2 (F2), or exon 3 (F3) (with a reverse primer in exon 15, R) showed that although MEFs and mammary epithelial cells expressed full-length CRD-BP, the mouse mammary cell lines (EP and EN cells) did not (Fig. 1B). Instead, these cells predominantly expressed an mRNA initiating in the ϳ25-kb second intron of CRD-BP (producing a ϳ1.3-kb PCR product).
The relative amounts of full-length and ⌬N-CRD-BP for MEFs, mammary epithelial cells, and the mammary EP and EN cell lines are shown in Fig. 1C (knowing that the efficiency of primer pairs was approximately equal; data not shown). Embryonic cells expressed ϳ100-fold more CRD-BP mRNA than cell lines from adult tissues.
To test the generality of this observation, we assayed a panel of human cell lines (Fig. 2). The human embryonic kidney epithelial cell line, 293T, has been a standard for investigating mRNA targets for CRD-BP (20,21). PCR-based exon linkage analysis showed that 293T cells have the "embryonic" pattern of full-length CRD-BP mRNA expression, whereas the mRNA species that predominate in most breast tumor cell lines encode the truncated protein product (Fig. 2, B and C). Human tissues showed a pattern broadly similar to that observed for mouse embryonic and adult tissues. Indeed, 293T cells expressed 1000-fold more CRD-BP mRNA than human breast cancer cell lines (as measured by quantitative PCR; Fig. 2C). The breast tumor cell lines examined included representatives of different breast cancer subtypes (Luminal type, MCF7; HER2 overexpressers, BT474 and SkBr3; basal type, HS578T; triple-negative breast cancer, MDA-MB-231). MCF10A cells, a nontumorigenic breast cell line, had low/undetectable levels of CRD-BP mRNA. When these breast tumor cell lines were compared with one another, the CRD-BP mRNA expression levels varied 40-fold (Fig. 2C).
The long (7-kb) 3Ј-UTR of CRD-BP mRNA contains six conserved miRNA binding sites for let7 family members (let7 miRNA binding sites are depicted as white circles in Figs. 1A and 2A) that regulate the stability and translation of CRD-BP mRNA in embryonic tissues (30 -32). The CRD-BP/IGF2BP1/ IMP1 mRNA was one of the examples used to demonstrate the general shortening of 3Ј-UTRs in mRNAs expressed by tumor cell lines (33). Given the complexity of mRNA isoforms revealed by our current analysis, we re-examined the 3Ј-UTR shortening phenomenon in our panel of human breast cancer

Analysis of CRD-BP transcripts shows that adult mouse breast epithelial cells express a CRD-BP transcript that lacks exons 1 and 2 (⌬N-CRD-BP).
A, schematic diagrams show the gene structure for CRD-BP, with exons numbered as per NCBI annotation (accession NM_009951.4), together with the proteins predicted from the differential exon linkage products. A proposed novel start site (dotted circle) for transcription of a short form mRNA(s) (new exons identified by 5Ј-RACE depicted as red rectangles) is indicated along with the proposed start codon (ATG) in exon VI. Primer binding sites are indicated for canonical exons 1, 2, and 3 (F1, F2, and F3, respectively), and a reverse primer in exon 15 (R), along with forward primers that recognize intron 2 sequences (Fi2a and Fi2b). The annotated 7-kb 3Ј-UTR is shown, together with alternative polyadenylation sites (crosses) and let7 miRNA binding sites (white circles). The detailed mRNA sequences deduced from 5Ј-RACE analysis are also listed. The alternative transcription start site in intron 2 is predicted to produce different proteins, designated full-length (FL) and ⌬N (N-terminally deleted). Functional domains of the CRD-BP protein are shown. RRM, RNA recognition motif. B, RT-PCR analysis was performed using the primers described in A to analyze RNA isolated from wild-type MEFs and adult mouse mammary cell lines (EP and EN cells). C, RNA samples from MEFs, primary adult mouse mammary epithelial cells (MECs), and adult mouse mammary cell lines (EP and EN cells) were also analyzed using qRT-PCR to determine total CRD-BP mRNA levels (compared with MEFs) and the relative distribution of the FL and ⌬N CRD-BP isoforms.
Exon i2: [15,596] CCATCGTCATCAGGGGGCACTAAGGACCCCGAGGAGCACGGGGATGTCTGCCGGGTCCTTCTCATG GTGCGGTGGGAAGCTGCCGAAGACAGATTGCTTTGGAGGCTGGTCTGGTTTCCTCTCCTGGCTCCT GAGAGGGCCACCTCGGCTGTTCCCTCTGGAGGACAGAGCCCGACTCTGTCTGTTTTCAAG [15,787] Human CRD-BP RNA (NM_006546.3) cell lines. We designed three sets of qRT-PCR primers that span a region in the 3Ј-UTR lying between the two sets of three let7 miRNA binding sites (primer sequences listed under "Experimental Procedures"). Two of these primer sets, one of which corresponds to the sequence bound by the Agilent array probe, gave no signal for any of the adult tissues examined, suggesting that results of published assays based on the 3Ј-UTR could be compromised in general. Interestingly, however, data from one primer pair (with an amplicon at ϳ3.5-3.7 kb and the highest efficiency) showed that the proportion of CRD-BP mRNAs that include the 7-kb 3Ј-UTR varied between 0 and 60% for any given breast cancer cell line (Fig. 2D).
To test the exon usage of CRD-BP mRNAs in breast tumors, we analyzed the RNASeq data deposited in the Cancer Genome Atlas (29). The application of RNASeq analysis for this purpose has been reviewed recently (34). Our analysis confirmed that CRD-BP mRNA levels were elevated in breast tumors compared with normal breast tissue (Fig. 2E). Furthermore, we found that the expression of exons 1 and 2 was relatively depleted compared with exons 3-15 in breast tumors. Quantitation suggests that only 40% of the total transcripts (from all breast tumors) are likely to be full-length, and the remainder of mRNAs include exons 3-15. Another set of peaks implies that other transcripts are produced at this locus, although if these exons are linked together, they are predicted to be noncoding (exons 8, 9, and 12-14).
Each breast tumor subtype showed a characteristic expression level and exon usage for CRD-BP mRNA isoforms. Of interest, Luminal B type tumors (an estrogen receptor-positive but aggressive, treatment-resistant tumor type) showed higher CRD-BP mRNA expression compared with Luminal A type tumors. Basal and Her2-positive tumor subtypes showed highest expression. Basal, Her2 and Luminal B subtypes showed highly depleted exon1/2 expression, suggesting that the majority of CRD-BP protein expressed in these tumor subtypes in vivo in patients could be the ⌬N-CRD-BP isoform.
A Mouse Strain with a Mutant CRD-BP Allele Retains Expression of the ⌬N-CRD-BP Isoform-Hansen et al. (35) showed that mice with a mutant CRD-BP allele display dwarfism (ϳ30% decrease in size compared with wild-type littermates) and impaired gut development. It was perhaps surprising that the phenotype was so mild. A ␤geo gene trap strategy was used to create the mutant allele (Fig. 3A). However, the gene trap insertion site in intron 2 lies upstream of the putative transcription start site for ⌬N-CRD-BP. Indeed, we noticed that ⌬N-CRD-BP mRNA was still expressed in MEFs from these mice, but total mRNA expression was reduced by 80% (Fig. 3B). We conclude that these mice harbor a hypomorphic allele of CRD-BP rather than a null allele.
We utilized MEFs from the CRD-BP hypomorphic mouse strain to test the specificity of various antibodies used to characterize CRD-BP expression and subcellular distribution. The antibodies tested include two N-terminal epitope antibodies from Abcam and Cell Signaling, which we showed were highly specific for full-length CRD-BP by immunofluorescent staining of CRD-BP hypo MEFs (Fig. 3C). A monoclonal antibody to the C terminus of CRD-BP (VS) showed decreased staining in CRD-BP hypo MEFs compared with WT MEFs. We used this antibody for semiquantitative assessment of CRD-BP expression levels.
Evaluation of CRD-BP antibody specificity by Western blotting confirmed the immunocytochemical stains (Fig. 3D). Fulllength CRD-BP, visualized as a 70kDa band, was undetectable in Western blots of lysates from CRD-BP hypo MEFs. Bands at 70 and 50 kDa were detected using an antibody directed against a KH domain-derived epitope ( Fig. 3D; pAb Sigma; epitope is indicated on protein scheme), which is predicted to be present in both full-length and ⌬N-CRD-BP species (50 kDa is the predicted molecular mass of ⌬N-CRD-BP). Expression of the shorter species was conserved in CRD-BP hypo MEFs. An antibody to a distinct KH domain-derived epitope (pAb; kind gift from David Herrick) also detected 70-and 50-kDa species in human 293T cells, with amounts that correlate with the relative mRNA abundance (Fig. 2C). Similar 50-kDa species have been observed in various cells and tissues by other studies, although those studies suggested these species could be degradation products (13, 36 -38). The C-terminally reactive VS mAb to CRD-BP used for immunostaining did not reliably detect endogenous levels of CRD-BP by Western blotting.
Homologous Transcription Start Sites in Intron 2 of Mouse and Human CRD-BP Genes-We searched for metadata to support the presence of a conserved start site for distinct CRD-BP mRNA species. We discovered highly conserved elements in intron 2 of both human and mouse CRD-BP gene loci, ϳ12 kb downstream of exon 2 (Fig. 4A). These elements are rich in CpG islands (often found near gene promoters) and in epigenetic marks associated with active transcription (H3K4 methylation).
Furthermore, the ENCODE ChIP-Seq data set (reported by the University of California, Santa Cruz cancer genome browser) showed extensive transcription factor binding/occupancy at the intron 2 candidate promoter, including occupancy by transcription factors such as POLR2A. Importantly, a pro-  Fig. 1A (based on NCBI accession NM_006546.3). The novel exon located in intron 2 is labeled Exon i2, and the location of the additional forward primer (Fi2) is shown. B, RNA was isolated from the human cell lines indicated, and RT-PCR analysis (as described for Fig. 1B) was performed using primers to an intronic region (orthologous to the mouse sequence) in the human CRD-BP gene (see also Fig. 4A). C, isoform-specific expression patterns of CRD-BP mRNAs were assayed by qRT-PCR analysis as described for Fig. 1C, and relative expression was compared with MCF7 breast cancer cell line. D, to evaluate whether the CRD-BP mRNA species detected express the long let7 miRNA regulated 3Ј-UTR, RNA preparations from the indicated cells/tissues were analyzed by qRT-PCR analysis. Results are shown comparing signals from a primer set amplifying a region in the 7-kb-long 3Ј-UTR to signals from a primer set amplifying total CRD-BP (amplicon located in exon 15). gram designed to identify transcription start sites (Augustus (39)) independently predicted a CRD-BP transcript initiating in intron 2.
Of two in-frame start codons that could serve as the translation initiation site for ⌬N-CRD-BP, only one had a high consensus, proximal Kozak sequence based on a prediction tool,

CRD-BP Is Required for Mammary Cell Clonogenicity
ATGpr (40). Translation of ⌬N-CRD-BP, predicted to start at Met-137 of full-length CRD-BP, generates a protein comprising 441 amino acids (compared with the 577 amino acids for the full-length protein; Fig. 4B).

CRD-BP Is Highly Expressed and Ubiquitous in Human Breast Tumors-Transgenic expression of CRD-BP has been
demonstrated to be oncogenic in the mouse mammary gland (2), and studies have suggested links between CRD-BP locus amplification or increased CRD-BP mRNA expression levels and mammary tumorigenesis (41,42). Given the previously reported functional links between CRD-BP expression and tumorigenicity for many tumor types, we evaluated the expression of CRD-BP in human breast tumors. A human breast tumor tissue microarray (243 distinct breast cancer samples, together with four normal breast samples) was stained with the C-terminally reactive VS mAb to CRD-BP, to evaluate total CRD-BP expression (including both the full-length and ⌬N isoforms). Breast tumor samples were predominantly (96%) obtained from Caucasian women with a mean age of 55 years. 80% of tumors on the microarray were estrogen receptor/progesterone receptor-positive, 12% were HER2-positive, and 8% were triple negative.
Interestingly, CRD-BP was ubiquitously expressed in normal breast tissues (although enriched in luminal breast epithelial cells) and also expressed in all breast tumors (examples of the stains are shown for various tumor types and normal tissues in Fig. 5A). There was no obvious heterogeneity of protein expression in tumors when evaluated for distinct cell types, proximity to hypoxic zones, stromalized areas, margins, or tumor subtype (data not shown).
Thus, both RNASeq and histological data suggest that the expression of CRD-BP isoforms is ubiquitous in breast epithelial cells. These sources were corroborated by mRNA expression data from Agilent microarrays, published by the Cancer Genome Atlas database. This analysis confirmed that CRD-BP mRNA was expressed by all tumor types and that the HER2 cohort of breast tumors showed the highest expression levels of CRD-BP (predicted based on co-amplification of the 17q21 amplicon containing the HER2 gene) (Fig. 5B).
Using our purpose-built, isoform-specific primers, we performed quantitative PCR analysis to determine the expression of full-length and ⌬N-CRD-BP mRNAs in a limited cohort of breast tumors and normal breast tissue (Fig. 5C). The results also showed that relative CRD-BP mRNA expression was elevated in tumors and demonstrated that whereas some tumors expressed only one variant (either full-length CRD-BP or ⌬N-CRD-BP), other tumors expressed both. Finally, we found that CRD-BP transcripts could be spliced to either include or exclude the let7-regulated long 3Ј-UTR. Thus, tumor mRNAs that include the long 7-kb 3Ј-UTR comprised 20 -96% of total CRD-BP transcripts detected (Fig. 5D).

CRD-BP Is Required for Clonogenic Growth of Mouse and Human Breast Epithelial Cells-Previous studies have shown
that not only is CRD-BP overexpressed in cancer cells, but also that CRD-BP is important for tumor cell survival, as summarized by Bell et al. (7). However, a previous study suggested that CRD-BP/IMP1 had no effect on breast cancer cell growth; instead it was ascribed tumor suppressor functions, because CRD-BP knockdown resulted in increased growth of metastatic cells and increased cell migration (43). To test the functionality of CRD-BP in breast tumor cells, we knocked down expression in both mouse (by stable transduction of shRNA constructs) and human (by transient transfection of shRNA constructs) mammary cell lines and found that although CRD-BP was not required for growth in mass culture, it was essential for the clonogenic activity of mammary cell lines. Thus, 50 -90% knockdown of CRD-BP mRNA (Fig. 6, A and C) resulted in a corresponding decrease in clonogenic growth for both human and mouse breast tumor and nontumorigenic cells (Fig. 6,  B-D). Most of these cell lines express majority ⌬N-CRD-BP mRNA. For a nontumorigenic mouse epithelial cell line (EP cells), expressing very low amounts of CRD-BP (only 0.1ϫ expression compared with tumorigenic EN cells; Fig. 1C), 60 -90% knockdown did not affect clonogenicity (Fig. 6D).
Re-expression of Either Full-length or ⌬N-CRD-BP Rescues Clonogenicity-To the best of our knowledge, all previous studies investigating the effect of CRD-BP expression on clonogenicity and cell survival have only tested the function of fulllength CRD-BP (for summary see review by Yisraeli (17)). We have shown that CRD-BP is expressed in at least two distinct forms, where the ⌬N-CRD-BP variant can predominate in breast tumor cells in vitro and in vivo. For example, this is the main mRNA species expressed by MCF7 cells (Fig. 2C). To compare the functionality of ⌬N-CRD-BP compared with fulllength CRD-BP, we knocked down expression of CRD-BP in MCF7 human breast cancer cells using a human CRD-BP-specific shRNA (or a scrambled shRNA control) and rescued expression with mouse full-length or ⌬N-CRD-BP. Knockdown and overexpression were assessed by immunofluorescent staining, Western blotting, and qRT-PCR (Fig. 7, A-C). Colony formation assays performed on MCF7 cells showed that either full-length or ⌬N-CRD-BP mouse proteins, when expressed at comparable levels, partly rescued the decrease in clonogenicity observed following CRD-BP knockdown (Fig. 7D). We conclude that either full-length or ⌬N-CRD-BP show similar activity with respect to their support of breast tumor cell clonogenicity.

Discussion
Our investigation of discrepancies arising between analyses of CRD-BP mRNA and protein expression has revealed that transcriptional regulation of the CRD-BP gene is more complex  Fig. 1A). B, RNA isolated from wild-type and CRD-BP hypo MEFs was analyzed by RT-PCR analysis using the primers indicated (left panel). qRT-PCR was used to assess total CRD-BP mRNA levels (compared with wild-type cells) and to measure the relative expression of the ⌬N-CRD-BP isoform (right panel). C, wild-type and CRD-BP hypo MEFs were fixed and stained with anti-CRD-BP antibodies as indicated (detailed under "Experimental Procedures"; nuclei were counterstained with DAPI). Epitope locations for antibodies used in this study are indicated in the CRD-BP schema depicted above the images. Scale bars, 50 m. D, Western blots of protein lysates from wild-type and CRD-BP hypo MEFs were probed with the CRD-BP antibodies indicated. E, Western blots of protein lysate from 293T cells probed with the C-terminally reactive antibody (kind gift from David Herrick). than previously understood. An internal promoter drives expression of a short isoform (⌬N) lacking the N terminus in adult normal and tumor tissues. Although the full-length protein is characteristic of embryonic tissues (and is the target of several commercially available antibodies), the short form is widely expressed. Indeed, we observe CRD-BP staining (using an antibody directed to the C-terminal domain predicted to react with all known isoforms) in adult tongue, liver, lung, and colon tissue sections (Fig. 5 and data not shown). Importantly, the antibody stains we report are validated by the use of cells with a hypomorphic CRD-BP locus, as well as cells transfected with CRD-BP shRNA in vitro. The subcellular distribution of CRD-BP staining is specifically cytoplasmic, unless the protein is overexpressed (Fig. 7B). This immunocytochemical staining pattern is supported by a published biochemical study that showed, using Nycodenz gradient centrifugation, that overexpressed and endogenous CRD-BP showed distinct fractionation patterns (20). As this study noted, altered localization of overexpressed CRD-BP may lead to the acquisition of novel functions.
To investigate whether short form ⌬N-CRD-BP mRNA is expressed by human breast tissue in vivo, we interrogated RNASeq data from the Cancer Genome Atlas database. This analysis revealed that normal breast tissues express low levels of the full-length CRD-BP mRNA, whereas breast tumors express relatively elevated levels of CRD-BP transcripts, the majority of which are ⌬N-CRD-BP mRNA species (exons 3-15). This broadly corresponds with results of exon-specific qPCR analysis of a small number of biopsy samples from breast tumors (Fig.  5C) and the expression patterns of breast tumor cell lines in vitro (Fig. 2C). We predict therefore that the principle CRD-BP protein expressed by the most malignant breast tumors in vivo in patients is the ⌬N-CRD-BP short isoform that we have identified here.
Our immunohistochemical analysis of CRD-BP expression shows that this protein is ubiquitously expressed in normal breast epithelial cells and also in breast tumor cells in vivo. Although expression levels are lower in adult cells compared with embryonic sources, the CRD-BP gene locus cannot accurately be described as oncofetal. More accurately, the gene locus is regulated heterochronically in embryo and adult (32), with respect to both transcription start sites and overall expression levels. Also, the post-transcriptional regulation of CRD-BP/IMP1 mRNAs has been shown to depend at least in part upon binding of the let7 class of miRNAs (11,30). We confirmed that embryonic 293T cells express CRD-BP transcripts that include the long form let7-regulatable 3Ј-UTR CRD-BP mRNA ( Fig. 2D and Northern data not shown) but also show that breast cancer cell lines (Fig. 2D) and human tumor samples (Fig. 5D) contain significant amounts of mRNAs without this long 3Ј-UTR. This partly confirms data from Mayr and Bartel (33), who showed that alternative polyadenylation of mRNA species, leading to 3Ј-UTR shortening, was characteristic of a variety of cancer cell lines. RNA species with shorter 3Ј-UTRs were more stable and produced more cognate protein (10ϫ more than mRNAs with long 3Ј-UTRs associated with normal cells). Specifically for CRD-BP/IGF2BP1, they showed that the 3Ј truncated CRD-BP/IGF2BP1 mRNA was a more potent oncogene, leading to increased clonogenicity in NIH 3T3 cells overexpressing CRD-BP/IGF2BP1. Other studies have shown that CRD-BP is significantly expressed in adult tissues. Thus, using reagents targeted toward total CRD-BP (including both the full-length and ⌬N isoforms), Dimitriadis et al. (44) reported expression of CRD-BP in intestinal crypts and showed widely divergent expression of CRD-BP in different colorectal tumors. They also show that high CRD-BP expression levels correlated with poor patient outcomes. A recent study (26) employing an Agilent microarray platform showed that CRD-BP/IGF2BP1 is expressed in adult liver. This study utilized PCR primers directed to sequences in exon 11 to confirm the gene expression changes observed for hepatocellular tumors; we predict that this assay would capture the majority of mRNA isoforms (equivalent to our total CRD-BP assay with qRT-PCR primers targeting a region in exon 15). We confirmed the expression pattern reported in this study (Fig. 5A).
Overall, expression of the CRD-BP locus is likely to have been significantly underestimated in previous studies, for two reasons: 1) antibody-reactive epitopes are lacking from the newly identified ⌬N-CRD-BP isoform, and 2) there are a variety of mRNA isoforms that exist, not all of which are detected by PCR primers directed to the canonical GenBank TM sequence.
Our results showed that expression of CRD-BP mRNA isoforms, as well as total CRD-BP expression, varied more between breast cancer cell lines in vitro than for tumors in vivo. Also, the amount of CRD-BP protein per cell was not strictly related to mRNA expression (data not shown), implying considerable posttranscriptional regulation. For example, we expected to observe higher protein expression of CRD-BP in HER2-positive breast cancer cells and tumors, given the previously described co-amplification of HER2 and CRD-BP (41). In fact, there was surprisingly little difference between CRD-BP protein expression levels in the tissue samples from various breast tumor subtypes (Fig. 5).
We find that CRD-BP protein is predominantly cytoplasmic in breast epithelial cells and is distributed in a granular pattern. This distribution has also been reported in other studies (20,45). Thus, using a monoclonal antibody raised against fulllength recombinant CRD-BP, Gutschner et al. (26) showed a cytoplasmic stain of both normal hepatocytes and hepatocellular carcinoma tumor cells that resembles the subcellular distribution we report. Knockdown of CRD-BP/IGF2BP1 in hepatocellular carcinoma cell lines resulted in reduced growth rate and increased rate of apoptosis and correlated with decreased rates of growth of xenografts in vivo (26).
Of note, the phenotype of the CRD-BP/IMP1 knock-out mouse, as reported by Hansen et al. (35), requires re-evaluation. This strain was produced by insertion of a gene trap cassette in intron 2, which effectively knocks out transcription of the fulllength mRNA; however, the insertion point is ϳ10 kb upstream of the putative internal entry promoter, thereby maintaining transcription of ⌬N-CRD-BP (Fig. 3). These mice show progressive dwarfism that initiates during embryogenesis, after the peak of CRD-BP/IMP1 expression (from the annotated 5Ј upstream promoter) at embryonic day 12.5. Half of newborn hypomorphic pups die, and this is associated with intestinal malformation. The phenotype of the total CRD-BP knock-out is therefore likely to be more severe. The promoter that drives embryonic expression of CRD-BP is a hot spot for transcription factor binding (Fig. 4) and is preferentially methylated in breast cancer cells (43) and in human mesenchymal stem cells (the latter by TET1/2) (46). A more select group of transcription factors (identified by ChIP analysis from the ENCODE database) is loaded onto the novel, internal promoter we describe here for adult normal and tumor tissues.
We show that loss of CRD-BP has a dramatic effect on clonogenicity of breast cancer cell lines, while not significantly affecting the growth of cells in mass cultures (another study using reagents targeted toward total CRD-BP also showed the latter result for MCF7 cells (3)). We found that loss of clonogenicity increases with higher efficiency of knockdown. We have shown that the short form of CRD-BP, ⌬N-CRD-BP, can rescue clonogenic activity of breast cancer cells in vitro and propose therefore that the widely expressed ⌬N isoform is able to substitute for the full-length CRD-BP isoform at least in the context of clonogenic activity. ⌬N-CRD-BP lacks the two N-terminal RNA binding motifs. These motifs have been suggested to be a relatively recent evolutionary addition (47), and previous functional studies have suggested that the four KH domains dominate RNA binding activities of this protein (18,19). Although one structure function study of CRD-BP suggests that the first two RNA recognition motifs control the targeting of CRD-BP/ ZBP1-associated granules to the leading edge of fibroblasts (19), other results suggest that these motifs are largely dispensable for specific activities (48,49).
Functional studies in breast epithelial cells have shown that CRD-BP is an oncogene with 100% penetrance when expressed in transgenic mice (albeit with a long latency of almost 12 months depending on the amount of CRD-BP expressed (2)). However, studies by Singer and co-workers (43,50) have shown that CRD-BP/IGF2BP1 has tumor suppressor functions, such that loss of CRD-BP/IGF2BP1 increased growth and motility of breast cancer cells. This conclusion varies from our study and others. The functional aspects of the variant studies focused on the metastatic mammary tumor cell line, MTLn3, which we did not assay. Furthermore, some of the reagents described in that study, including the siRNA constructs used, were specific to full-length CRD-BP. To confirm the effects we observed, we performed the knockdown experiment using two different approaches, first using CRD-BP knockdown constructs together with puromycin selection (data not shown) and second without co-transfection of the selection marker or addition of puromycin (Fig. 6). Results obtained from both methods were concordant and showed a dramatic effect of CRD-BP knockdown on clonogenic growth for a wide variety of cell lines. We are currently investigating the molecular basis of the requirement of tumor cells for CRD-BP, knowing that clonogenicity has been related to various factors, including integrin activity, mitochondrial function, and growth factor production.
In summary, the description of CRD-BP as an oncofetal protein/gene glosses over crucial differences between embryonic and adult expression of the CRD-BP gene locus. The predominant mRNA transcribed in adult cells is likely to be induced by different transcription factors and epigenetic programs, and the predominant protein isoform expressed in adult tissues is different from that expressed in embryonic tissues. As far as we are aware, prior studies in vitro and in vivo have focused on the full-length canonical protein described prominently in the databases, and the current study is the first to identify and characterize expression of an N-terminally deleted short form of CRD-BP.
We do not eliminate the possibility that there are other splice variants expressed by this locus, perhaps several different polyadenylated versions of each mRNA, and potentially several distinct protein products. Indeed, there are precedents for such complexity. For example, the N-terminally truncated CRD-BP species resembles protein variants observed for another member of the IMP family, IMP2 (51), as well as another RNA binding protein, FMRP (fragile X mental retardation protein). Expression of the FMR1 gene locus shares other characteristics with CRD-BP. Thus, as many as 12 transcript isoforms are derived from the FMR1 locus, which vary in abundance at least 2 orders of magnitude between adult and embryonic mouse brain (52). These various isoforms have different functions and show specific regulation.
Despite low expression levels that make the protein(s) difficult to detect, we show that CRD-BP is important for clonogenic growth of breast cancer cells in vitro. The mRNA binding partners regulated by full-length and/or ⌬N-CRD-BP to maintain clonogenicity of breast cancer cells are not yet characterized but are likely to be important to the expression of fully malignant characteristics of breast cancer cells in vivo.