Regulation of protein synthesis is fundamental for all those aspects of eukaryotic biology by controlling development, homeostasis, and stress responses1,2. c-Jun and BTG1, reveals that eIF3 employs different modes of RNA stem loop binding to exert either translational activation or repression. Our findings illuminate a new role for eIF3 in governing a specialized repertoire of gene expression and suggest that binding of eIF3 to specific Calcipotriol monohydrate mRNAs could be targeted to control carcinogenesis. Considerable genetic evidence implicates eIF3 in other functions in translation outside of its general role as a protein scaffold for formation of initiation complexes. Mutation or inactivation of eIF3 subunits results in developmental defects in and zebrafish6,7. Furthermore, analyses of human tumors reveals that overexpression of eIF3 is usually linked to diverse cancers, including breast, prostate, and esophageal malignancies4,8. The integral role of eIF3 during cellular differentiation, growth, and carcinogenesis, suggests eIF3 might drive specialized translation. Consistent with this hypothesis, translation of the hepatitis C computer virus RNA occurs through essential interactions between eIF3 and a structured Internal Ribosome Access Site (IRES) element in the viral genome, indicating the feasibility of translation regulation being driven by distinct cellular eIF3CmRNA contacts9. To identify candidate transcripts regulated through direct interactions with eIF3, we first used a genome-wide approach to determine the eIF3 RNA binding targets in human 293T cells. Because eIF3 is composed of 13 subunits (eIF3aCm), we adapted a 4-thiouridine photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP)5 approach to allow analysis of a large multimeric complex, with isolation of individual subunit-RNA libraries (Fig. 1a). As overexpression of single eIF3 subunits can alter complex assembly8, we optimized immunoprecipitation of the full endogenous eIF3 complex using an antibody that recognizes the eIF3b subunit (Fig. 1b). High salt washes were used to ensure removal of potentially contaminating translation factors, such as eIF4G or the small ribosomal subunit (Fig. 1c). After RNase digestion, separation of crosslinked eIF3CRNA complexes by denaturing gel electrophoresis demonstrated that four of the thirteen subunits crosslink directly to RNA (Fig. 1d), identified by mass spectrometry as eIF3a, b, d, and g (Extended Data Fig. 1). Figure 1 PAR-CLIP of the multi-protein translation initiation factor complex, eIF3 For each subunit, separate cDNA Calcipotriol monohydrate libraries were generated from the isolated crosslinked RNAs and Calcipotriol monohydrate deep sequenced using Illumina technology. Sequenced reads from three biological replicates were mapped to the genome and grouped into eIF3-binding sites by using the cluster-finding tool Paralyzer10. Read clusters were found in 479 unique genes, with eIF3a, b, d, and g crosslinking to 328, 264, 356, and 352 transcripts, respectively (Supplementary Table 1, 2). The limited number of interacting genes supports capture of specific eIF3CRNA contacts, as these targets compromise only 3% of total expressed transcripts (Extended CSP-B Data Figure 2). As Calcipotriol monohydrate a further control, we do not see crosslinking to highly abundant rRNAs, in agreement with biochemical and structural studies showing that eIF3 interacts primarily with the protein-rich face of the small ribosomal subunit11-14. The majority of RNAs contained a single eIF3-binding site, with a median cluster length of 25 nt (Fig. 2a, b). These RNAs interact with distinct combinations of eIF3a, b, d, and g subunits (Fig. 2c). To validate the RNAs identified by PAR-CLIP, we performed eIF3 immunoprecipitation in the absence of crosslinking. We detected Calcipotriol monohydrate eIF3CRNA interactions for five top candidate genes using RT-PCR; whereas a negative control mRNA, the PSMB6 transcript, was not immunoprecipitated (Fig. 2d). Figure 2 Analysis and validation of eIF3 PAR-CLIP-derived binding sites In eukaryotic protein synthesis, the 5 UTR of mRNA is thought to be the major site of translation regulation3. In agreement with identifying translation regulation roles of specific eIF3CmRNA interactions, the eIF3 binding sites predominantly mapped to the 5 UTR (70%) (Fig. 2e). To examine the impact of transcript-specific engagement of eIF3 on translational control, we focused on two genes with an eIF3 binding site in the 5 UTR, and (translation extracts with m7G cap analog inhibited translation of both c-Jun and BTG1 luciferase reporter mRNAs, demonstrating that eIF3-dependent translation regulation of these transcripts is cap-dependent and thus distinct from viral IRES-like mechanisms18 (Fig. 3f, g). These results demonstrate that eIF3 can act as both a translation activator and repressor of specific cellular mRNAs. Figure 3 eIF3 is a positive and negative transcript-specific translational regulator To understand how eIF3 binding to mRNA leads to opposing translation phenotypes, we next identified the full RNA elements for eIF3 recognition in the c-Jun and BTG1 mRNAs. While PAR-CLIP marks the localized vicinity of eIF3 in the 5 UTR, eIF3 interaction could occur either through recognition of a linear sequence or in the context of RNA secondary structure. Using selective 2 hydroxyl acylation analyzed by primer extension (SHAPE), we experimentally determined the secondary structure around the eIF3 binding sites (Fig. 4a, d). For both c-Jun.