tRNA Binding, TPR Eddy Fold, and Subcellular Localization of the Human Interferon-induced Protein, IFIT5

Abstract

Interferon-induced proteins, including the largely uncharacterized interferon-induced tetratricopeptide repeat (IFIT) protein family, provide defense against pathogens. Differing from expectation for tetratricopeptide repeat (TPR) proteins and from human IFITs 1, 2, and 3, we show that human IFIT5 recognizes cellular RNA instead of protein partners. In vivo and in vitro, IFIT5 bound to endogenous 5’-phosphate-capped RNAs including transfer RNAs (tRNAs). The crystal structure of IFIT5 revealed a convoluted intramolecular packing of eight TPRs as a fold that we name the TPR eddy. Additional, non-TPR structural elements contribute to an RNA binding cleft. Instead of general cytoplasmic distribution, IFIT5 concentrated in actin-rich protrusions from the apical cell surface co-localized with the RNA-binding retinoic acid-inducible gene-I (RIG-I). These findings establish compartmentalized cellular RNA binding activity as a mechanism for IFIT5 function and reveal the TPR eddy as a scaffold for RNA recognition.

INTRODUCTION

Viral infection triggers a broad host response that cripples metabolism and limits viral replication. Secreted cytokines such as the type I interferons (IFNs) are potent inducers of innate immune response gene expression programs for antiviral defense (Schoggins and Rice, 2011; Diamond and Farzan, 2012). Among the most strongly type I IFN-induced gene products are several proteins predicted to contain multiple TPRs (Schoggins and Rice, 2011; Fensterl and Sen, 2011). These IFIT proteins are believed to restrict viral infection by binding in an inhibitory manner to virus-specific targets as well as host-cell eIF3, a protein complex essential for translation initiation (Fensterl and Sen, 2011; Sen and Fensterl, 2012; Diamond and Farzan, 2012). Human IFITs 1, 2, and 3 appear to have concerted functions as interacting homodimers, heterodimers, and/or oligomers (Stawowczyk et al., 2011; Pichlmair et al., 2011; Fensterl and Sen, 2011; Yang et al., 2012). Humans have a fourth IFIT family member, IFIT5, which unlike IFITs 1, 2, and 3 has no murine ortholog (Schoggins and Rice, 2011; Fensterl and Sen, 2011). Curiously, IFIT5 is excluded from association with IFITs 1, 2, and 3 and also lacks any other strongly interacting protein partner (Hogg and Collins, 2007a; Pichlmair et al., 2011).

In other multi-TPR proteins, the bi-helical repeats pack regularly to create an extended super-helical platform for protein-protein interactions (D'Andrea and Regan, 2003; Zhu et al., 2012; Zeytuni and Zarivach, 2012). The precedent of TPR-scaffolded assembly of protein complexes fits with the observed multimerization of human IFITs 1, 2, and 3 and their nucleation of a much larger IFIT1/2/3 protein interactome (Pichlmair et al., 2011). Here we show that IFIT5 does not fit this paradigm for TPR protein function. We demonstrate that a monomer of IFIT5 binds directly, specifically, and autonomously to endogenous human cell RNAs with 5’ phosphate ends, including tRNAs. RNA recognition requires a convoluted intramolecular fold of the IFIT5 TPRs, which scaffolds unique additional helices that form an RNA binding cleft. Remarkably, IFIT5 and the antiviral double-stranded RNA-binding protein RIG-I co-localize in actin-rich structures at the apical cell surface, suggesting a general significance for cell-peripheral compartmentalization of innate immune response proteins with RNA ligands.

RESULTS AND DISCUSSION

To gain insight about IFIT5 function, we characterized its cellular binding partners. We expressed triple-FLAG-tagged (3xF) human IFIT5 with (Figure 1A–C) or without (Figure S1A–C) IFN-β treatment of transiently transfected HEK293T cells. No consistent IFIT5 co-purification of other proteins was detected (Figure S2). In contrast, denaturing gel electrophoresis and SYBR Gold staining of RNA in the input versus IFIT5-bound samples showed selective IFIT5 enrichment of RNAs that were approximately the size of tRNAs but distinct from the tRNA pool of the input cell extract (Figures 1B and S1B). Similar RNA co-enrichment was observed upon purification of IFIT5 from HeLa cells (Figure S1D–F). Compared to IFIT5, human IFITs 1, 2, and 3, and the mouse IFIT2 proposed to functionally parallel human IFIT5 (Daffis et al., 2010), did not co-purify a similar amount or profile of cellular RNA (Figure 1B and S1B), whether assembled with other IFIT proteins in cells treated with IFN-β (Figure 1A, lower panel) or purified alone (Figure S1A, lower panel).

Figure 1. IFIT5 RNA binding specificity.

(A–C) HEK293T cells were transfected to express a human IFIT protein or mouse IFIT2 with an N-terminal 3xF tag. Mock indicates no tagged protein expression. Cells were treated with IFN-β as indicated. Protein purification was detected by anti-FLAG immunoblot (A, upper panel) or silver staining (A, lower panel) with the tagged protein in each lane indicated by an open circle; note that IFIT3 stained light orange. Co-purified RNA was detected by SYBR Gold (B), and blot hybridization of the gel in (B) was used to detect the 5' end of specific tRNAs (C).

(D–F) HEK293 cells expressing a doxycycline (Dox)-inducible 3xF-IFIT5 transgene were used for denaturing affinity purification after in vivo cross-linking. Y701-phosphorylated Stat1 established IFN response induction, and tubulin was a normalization control (D). Co-enriched RNAs were detected by staining (E) or blot hybridization (F).

(G–I) Recombinant IFIT1 and IFIT5 were detected by Coomassie blue R-250 staining (G) and used for EMSA with in vitro transcribed and radiolabeled 64-nt iMet tRNA (H), with unlabeled competitor RNAs mixed with IFIT5 prior to addition of the radiolabeled 64-nt iMet tRNA (I).

(J) RNA bound to HEK293 transgene-expressed 3xF-IFIT5 from non-denaturing cell extract was used for parallel treatments with 5’ polyphosphatase (PPtase) and/or Terminator exonuclease (5’P Exo) in Terminator Buffer A or B prior to gel resolution and SYBR Gold staining.

To establish whether the tRNA-sized RNAs that co-purified with IFIT5 actually included tRNAs, we used Northern blot hybridization. Northern blots revealed the presence of IFIT5-bound initiator methionine (iMet) tRNA and, to a lesser extent, some of the additional tRNAs tested (Figures 1C and S1C,F). Curiously, a probe complementary to the iMet tRNA 5' end detected numerous sizes of co-enriched RNA, including a prominent fragment ~10 nucleotides (nt) shorter than full-length mature tRNA that was not detected with a probe for the iMet tRNA 3' end (Figure S1C,F). Other tRNA 5’-end probes also detected forms of IFIT5-bound tRNA different in size from mature tRNA (Figures 1C and S1C,F). The extent of iMet tRNA 3’ truncation was investigated by ligation of IFIT5-enriched RNA to an adaptor oligonucleotide, reverse transcription, and PCR with an iMet tRNA forward primer. Cloned sequences included 3'-truncated iMet tRNA with a mismatch to genomic sequence anticipated by tRNA base modification and an appended 3' polyuridine tract (Figure S3). Polyuridine tailing has been previously reported for RNA degradation intermediates (Norbury, 2010). Truncation and/or tailing of IFIT5-bound tRNAs would give rise to tRNA sizes different from mature tRNA.

Paralleling a previous strategy to mimic an endogenously induced level of IFIT protein expression (Pichlmair et al., 2011), we expressed IFIT5 from a doxycycline-regulated transgene stably integrated in HEK293 cells. Also, to eliminate any reassortment of complexes after cell lysis, we purified the transgene-expressed IFIT5 under stringently denaturing conditions following in vivo cross-linking with formaldehyde (Niranjanakumari et al., 2002). IFIT5 specifically co-purified cross-linked tRNA-sized RNAs including iMet tRNA of precisely the mature tRNA size (Figure 1D–F). The profile of IFIT5-bound RNA was not affected by treatment of cells with IFN-β (Figure 1D–F). Because IFIT5 consistently co-purified forms of iMet tRNA longer and shorter than the mature tRNA that were not abundant in the input cell extract, and considering the substantial in vivo accumulation of ~30 and ~45 nt RNA fragments induced by high IFIT5 over-expression (Figure S1B), we speculate that IFIT5-bound tRNAs are preferentially targeted for degradation. The tRNA fragments co-purified with IFIT5 could represent intermediates in tRNA degradation that were stabilized by IFIT5 binding, tRNA secondary structure, and/or tRNA base modification.

To test for direct binding of iMet tRNA by IFIT5, we reconstituted the protein-RNA interaction using bacterially expressed and purified IFIT5 (Figure 1G) and in vitro transcribed iMet tRNA lacking the 3’ ~10 nt not required for robust IFIT5 co-purification of iMet tRNA from cell extract (Figures 1C,F and S1C,F and S3). The 5’ monophosphate present on mature tRNAs in vivo was conferred by radiolabeling of a precisely defined RNA sequence excised from a longer in vitro transcript (Fechter et al., 1998). By native electrophoretic gel mobility shift assay (EMSA), IFIT5 bound to iMet tRNA in the absence or presence of heparin, a standard competitor of non-specific RNA interactions (Ryder et al., 2008). Notably, IFIT5 produced a single RNA mobility shift at all but the highest, micromolar concentration of recombinant protein (Figure 1H, lanes 11–20). In comparison, the human IFIT1 reported to bind 5'-triphosphate-capped RNAs (Pichlmair et al., 2011) did not bind the 5'-monophosphate-capped iMet tRNA in the presence of heparin (Figure 1H, lanes 1–5). Without heparin, IFIT1 gave heterogeneous mobility shifts requiring higher protein concentration than RNA binding by IFIT5 (Figure 1H, lanes 6–10).

Bacterially expressed IFIT5 bound iMet tRNA dependent on the 5’ monophosphate (Figure 1I). IFIT5 preferentially bound 3’-truncated tRNA, but the full-length iMet tRNA and other full-length tRNAs could compete for IFIT5 binding as long as they contained a 5’ monophosphate (Figure S4A). A series of iMet tRNA 5’ fragments with successively larger 3' truncations retained competition for IFIT5 binding, but the isolated anticodon stem-loop did not (Figure 1I). Direct binding assays confirmed that IFIT5 preferentially bound the iMet tRNA fragments with a truncated 3’ end compared to the full-length tRNA or anticodon stem-loop, both of which would be predicted to have a base-paired rather than single-stranded 5’ end (Figure S4B). In vitro transcribed RNAs with a 5’ triphosphate were not more effective competitors for IFIT5 binding than a 5' monophosphate RNA, yet both were strong competitors relative to the corresponding RNA without any 5' phosphate group (Figure S4C). Purified IFIT5-bound cellular RNAs were assayed for degradation by a 5’-monophosphate-dependent exonuclease, with or without pre-treatment by a polyphosphatase to convert any 5’ polyphosphate ends to 5’ monophosphate. Much of the bound cellular RNA was degraded even without polyphosphatase treatment, consistent with the presence of a 5’ monophosphate, and almost all of the remaining RNA was degraded after 5’ polyphosphate conversion to 5’ monophosphate (Figure 1J). Together these results demonstrate IFIT5 binding to endogenous cellular RNAs with at least one 5’ phosphate group. Mature tRNAs are an abundant pool of monophosphate-capped cellular RNAs that, in their cycle between aminoacyl synthetase and ribosome interactions, would be highly accessible for IFIT5 binding. Because the mature iMet tRNA forms a relatively weak A-U base-pair at its 5’ end required for eukaryotic initiator tRNA function (Farruggio et al., 1996), it may be particularly prone to acceptor stem fraying and IFIT5 association in vivo.

To define the structural basis for IFIT5 RNA recognition, we determined the crystal structure of IFIT5 at 2.2-Å resolution (Table 1 and Figure S5). The IFIT5 secondary structure is composed entirely of helices, including eight TPRs (Figure 2A). Consistent with the IFIT5 monomer evident by gel filtration (Figure S6), the TPRs pack with exclusively intramolecular contacts to form a fold that we term the TPR eddy (Figure 2A). The N-terminal 46 amino acids of IFIT5 constitute an extended buttress framing the first three TPRs, which at first reach toward the protein N-terminus (TPRs 1 and 2) and then double back (TPR 3). A two-helix insertion positions the subsequent TPRs 4 and 5, followed by an abrupt turn introduced by another helical hairpin (residues 283–333) that projects an arm between TPRs 5 and 6. Reminiscent of the tRNA recognition arm of seryl-tRNA synthetase (Cusack et al., 1996), the arm of IFIT5 contributes one side of a deep cleft lined with basic side chains (Figure 2A,B) that is completed by packing of TPRs 6 and 7 and a non-TPR connecting helix (residues 413–427). TPR 8 and a C-terminal helix (residues 469–478) cap the end of the TPR eddy (Figure S7), creating a prominent saddle between the N- and C-terminal lobes of the protein (Figure 2A,B).

Table 1.

Crystallographic Data Collection and Refinement Statistics

	Native	SeMet
Data Collection
Space group	P212121	P212121
Cell dimensions
a, b, c (Å)	63.83, 71.79, 116.92	64.07, 71.73, 116.45
α, β, γ (°)	90, 90, 90	90, 90, 90	90, 90, 90
		Peak	Remote
Wavelength	1.1111	0.97965	0.96817
Resolution (Å)	50-2.2 (2.24-2.20)	50-2.85 (2.90-2.85)	50-2.85 (2.90-2.85)
Rsym or Rmerge	0.114 (0.699)	0.137 (0.616)	0.135 (0.627)
I / σI	14.6 (2.0)	20.8 (4.3)	20.6 (4.3)
Completeness (%)	99.9 (100)	100 (100)	100 (100)
Redundancy	4.0 (4.0)	8.2 (8.4)	8.2 (8.5)
Refinement
Resolution (Å)	47.70-2.20
No. reflections	27,860
Rwork / Rfree	0.1778/0.2398
No. atoms
Protein	3913
Ligand/ion	7
Water	232
B-factors
Protein	26.66
Ligand/ion	46.84
Water	28.87
R.m.s deviations
Bond lengths (Å)	0.008
Bond angles (°)	1.121
Ramachandran statistics
Most favored regions	98.3%
Allowed regions	1.7%
Disallowed regions	0.0%

Values in parentheses are for the highest resolution shell.

Figure 2. The TPR eddy of IFIT5.

(A) Secondary structure of IFIT5 is depicted with cylinders representing helices. TPRs (separately colored, and the first helix of each numbered) are interspersed with non-TPR elements (gray). N- and C-termini are indicated.

(B) Electrostatic potential surface representation of IFIT5. Views are related by 180° rotation, with the view on the left similar to that in (A). Blue and red contours begin at +3 and −3 kT, respectively.

(C) Ribbon diagram of IFIT5 (yellow) superimposed on a monomer from the domain-swapped IFIT2 dimer (pink). N- and C-termini are indicated.

(D) Sequence conservation among IFIT5 proteins plotted from magenta (conserved) to cyan (variable) on the protein surface, based on primary sequence alignment (Figure S9). Views are related by 180° rotation, with the view on the left similar to that in (A).

The contorted TPR eddy of IFIT5 differs from the extended super-helical arrangement of TPRs in other multi-TPR proteins (D'Andrea and Regan, 2003; Zeytuni and Zarivach, 2012). The IFIT5 TPR eddy has overall similarity but local and global differences with the recently reported structure of a human IFIT2/ISG54 dimer (Yang et al., 2012). IFIT2 dimerization occurs by domain-swapping of a 3-helix bundle containing TPR 3, as well as other cross-subunit appositions enabled by a hinged closing of the cleft and opening of the saddle (Figures 2C and S8). The IFIT5 cleft has sequence conservation across IFIT5 proteins (Figure 2D) and unique to IFIT5 proteins (Figure S9) that can underlie structural differences between IFIT5 and IFIT2 (Figure S8), IFIT5 exclusion from the homo- and hetero-dimerizations observed for IFITs 1, 2, and 3 (Sen and Fensterl, 2012), and the unique IFIT5 RNA binding specificity (Figures 1B and S1B). However, mapping the IFIT5 structural elements (Figure 2A) on to a primary sequence alignment of human IFIT family members (Figure S9) suggests that all IFIT family members could form a TPR eddy.

We investigated IFIT5 structural requirements for RNA binding using C-terminal truncations from the beginning of TPR 7 or 8 (Figure 3A). IFIT5 co-purification of cellular RNAs including iMet tRNA was reduced by C-terminal truncation from TPR 8 (Δ8) and eliminated by truncation from TPR 7 (Δ7–8) (Figure 3B,C). Likewise, the bacterially expressed and purified TPR Δ8 showed partially reduced iMet tRNA binding activity, and TPR Δ7–8 retained no RNA binding activity (Figure 3D,E). The truncation from TPR 7 would eliminate one side of the cleft (Figure 2A). As expected for an RNA binding site, the cleft has the greatest positive surface potential in the structure (Figure 2B), and side chains project into the cleft with conformational flexibility (Figure S5).

Figure 3. An IFIT5 RNA binding cleft.

(A–C) C-terminally truncated proteins were expressed by transfection of HEK293T cells, with relative purified protein amounts determined by immunoblot (A), co-purification of RNA visualized by SYBR Gold (B), and co-purification of iMet tRNA tested by Northern blot (C).

(D–E) C-terminally truncated proteins purified from E. coli were detected by Coomassie blue R-250 staining (D) and used for EMSA with the 64 nt iMet tRNA (E).

(F) Space-filling representation of IFIT5 side chains individually targeted for mutational analysis, highlighted on the ribbon diagram of IFIT5.

(G,H) IFIT5 sequence variants were purified from transiently transfected HEK293T cells and detected by immunoblot (G), with input (left) and co-purified (right) RNAs stained by SYBR Gold (top) and used for blot hybridization (bottom) to detect iMet tRNA (H).

(I,J) IFIT5 sequence variants purified from E. coli were detected by Coomassie blue R-250 staining (I) and used for EMSA with the 64 nt iMet tRNA (J).

We substituted individual residues in the cleft with alanine (Figure 3F), targeting solvent-accessible side chains with sequence conservation among IFIT5 orthologs (Figure S9). IFIT5 variants with a L291A, K302A, R307A, K309A, F388A/H389A, K415A, K422A, or K426A substitution were expressed to comparable levels in HEK293T cells and purified comparably (Figure 3G). All of the cleft side-chain substitutions except L291A, which was deepest in the cleft (Figure 3F), reduced IFIT5 cellular RNA binding activity (Figure 3H). We compared the direct iMet tRNA binding activity of bacterially expressed, purified proteins across a range of in vivo RNA binding deficiency using the L291A, R307A, K415A, and K426A sequence variants (Figure 3I). Paralleling the results of cellular IFIT5 expression, the purified IFIT5 proteins demonstrated an in vitro iMet tRNA binding activity that was near normal for the L291A protein, 5-fold reduced for the K426A protein, or more than 25-fold reduced for the R307A and K415A proteins (Figure 3J). These results strongly implicate the IFIT5 arm and cleft as an RNA binding site.

To understand the cellular context for IFIT5 RNA binding, we investigated the localization of IFIT5 expressed by transient transfection (Figure 4) or by inducible expression from the integrated transgene (Figure S10). Both 3xF-IFIT5 (Figures 4A and S10) and IFIT5 tagged with mCherry (Figure 4B) localized to actin-rich protrusions of the apical cell surface. Similar microvillus-like projections are present in many cell types (Chhabra and Higgs, 2007; Lange, 2011), and correspondingly these projections were present independent of IFIT5 expression (Figure 4A, compare the cell at upper left versus lower right). IFIT5 subcellular localization was not overtly altered by treatment of cells with IFN-β (Figure S10). IFIT5 localization appeared to be at least partially independent of RNA binding, because at least some of the C-terminally truncated, RNA-binding defective TPR Δ7–8 protein co-localized with actin (Figure 4A, lower panels). In fibroblast-derived cells with well-resolved actin structures (Figure 4A), IFIT5 concentrated specifically in the microvillus-like structures at the apical cell surface. In hepatocyte-derived Huh7 cells (Figure 4B), IFIT5 also co-localized with the apical cell surface membrane ruffles that move rearward from a cell’s leading edge and coalesce to form macroscopic actin filament rings (Chhabra and Higgs, 2007).

Figure 4. IFIT5 localization to cell-surface actin structures.

Confocal microscopy was performed on fixed cells, focusing on the apical cell surface. Scale bars represent 10 microns.

(A) Fibroblast-derived WI-38 VA-13 cells were transfected to express 3xF-IFIT5 full-length (top panels) or a C-terminal truncation (bottom panels) and imaged for FLAG-tagged protein by indirect immunofluorescence (red). Cells were co-stained to detect filamentous actin (green) and DNA (blue).

(B) Hepatocyte-derived Huh7 cells were transfected to express mCherry-IFIT5 (red) and eGFP-RIG-I (green). Cells were co-stained to detect filamentous actin (blue). Different levels of focus show the apical cell surface (top panels) and cell edge (bottom panels) of separate cells.

The antiviral sensor RIG-I, which recognizes double-stranded 5’-triphosphate-capped RNA (Kolakofsky et al., 2012; Berke et al., 2012), has also been detected in actin-containing projections from the cell surface (Mukherjee et al., 2009). We found that mCherry-tagged IFIT5 and green fluorescent protein (eGFP)-tagged RIG-I concentrated in the same apical cell surface microvillus-like and membrane ruffle structures (Figure 4B). Consistent with the observed localization of IFIT5 or RIG-I to actin structures without co-over-expression of the other protein, we did not detect a biochemical association of IFIT5 and RIG-I that was stable to purification from cell extract. We speculate that each protein independently adopts the same subcellular compartmentalization as a mechanism for influencing RNA ligand recognition and/or degradation, for example to limit RNA binding or degradation to the cell periphery or to link these activities to actin depolymerization at sites of pathogen contact and entry (Delorme-Axford and Coyne, 2011).

IFIT proteins other than IFIT5 have been reported to homo- and/or hetero-dimerize and to interact with diverse additional proteins as mechanisms of biological function (Fensterl and Sen, 2011; Sen and Fensterl, 2012; Diamond and Farzan, 2012). In addition, IFIT1 has binding activity for synthetic 5’-triphosphate-capped RNA and a recombinant IFIT2 homodimer has binding activity for synthetic AU-rich duplex RNA (Pichlmair et al., 2011; Yang et al., 2012). It remains to be established how any RNA associations of IFIT1/2/3 complexes relate to their protein ligand interactions. Unlike the other human IFIT proteins, IFIT5 folds autonomously as a monomer without any apparent homo- or hetero-dimerization. The IFIT5 monomer harbors a deep cleft with positive surface potential that was not formed by an IFIT2 homodimer (Yang et al., 2012). Also unlike the other human IFIT proteins, IFIT5 readily recognizes cellular rather than only viral-hallmark RNA structures, accounting for the robust IFIT5 co-purification of endogenous human RNAs from uninfected cells. We conclude that while IFIT5 retains the IFN-induced expression of other human IFIT family members, IFIT5 has evolved independent biochemical function and thus distinct antiviral activity (Schoggins et al., 2011).

Our determinations of IFIT5 structure and specific residues required for RNA binding provide knowledge about a previously recognized structural framework for protein-RNA interaction. With consistent in vivo and in vitro specificity, we find that IFIT5 binds tRNAs with a 5’ monophosphate group. This insight expands the scope of cellular tRNA fates, adding to the growing awareness of a new biology of tRNA roles, interaction partners, regulated modifications, processing reactions, and turnover pathways (Phizicky and Hopper, 2010; Sobala and Hutvagner, 2011; Geslain and Pan, 2011; Hasler and Meister, 2012). Recent studies provide a parallel discovery of tRNA binding by another IFN-induced protein, schlafen 11, which suppresses protein synthesis by human immunodeficiency virus 1 in a codon-usage dependent manner (van Weringh et al., 2011; Li et al., 2012). In addition to 5’-monophosphate-capped RNAs, IFIT5 also binds RNAs with a 5’ polyphosphate group and thus could have viral as well as cellular RNA ligands. To increase the cellular specificity of RNA interactions, we propose that IFIT5 is regulated by its localization to dynamic actin structures at the cell surface. The membrane-adjacent compartmentalization of IFIT5 and its tRNA binding activity suggest that IFIT5 could impose a localized translational inhibition. Overall our results expand the eukaryotic repertoire of protein folds in general and RNA binding motifs in particular, establish a long-sought activity for human IFIT5, define the structural principles by which IFIT5 gained unique function, and introduce another player in the increasingly complex cellular regulation of tRNA.

EXPREIMENTAL PROCEDURES

Cleared whole-cell extracts were adjusted to 0.2 M NaCl before affinity purification with M2 FLAG antibody resin (Sigma) and elution with triple-FLAG peptide (Sigma). In vivo cross-linking and purification under denaturing conditions were performed as described (Hogg and Collins, 2007b), with controls to ensure no background of RNA binding without cross-linking. N-terminal six-histidine tagged IFIT5 and IFIT1 were expressed in E. coli and purified using nickel-charged resin. The iMet tRNA for direct binding assays was transcribed by T7 RNA polymerase using a hammerhead ribozyme to produce the mature tRNA 5' end (Fechter et al., 1998), gel-purified, and radiolabeled with a 5' monophosphate using T4 polynucleotide kinase. To interrogate IFIT5-bound RNA 5’ ends, RNA was pre-treated with or without 5´ polyphosphatase (Epicentre) and digested with Terminator 5’-monophosphate-dependent exonuclease (Epicentre).

IFIT5 was crystallized by vapor diffusion from 100 mM BisTris, pH 6.5, 4% (w/v) PEG monomethyl ether 5K. The structure was determined from X-ray diffraction data using multi-wavelength anomalous dispersion of selenomethionine-labeled protein crystals. Additional details of the bacterial expression constructs, purifications, crystallization, and structure determination are provided in Supplemental Experimental Procedures. For microscopy, cells were fixed with 3.7% paraformaldehyde for 10 min, permeabilized with 0.1% Triton X-100 for 15 min, and processed as described (Wong et al., 2002). Actin was visualized with phallodin conjugated to AlexaFluor 647 or 546 (Invitrogen).

HIGHLIGHTS.

• Human IFIT5 specifically and directly binds cellular RNAs including transfer RNAs

• IFIT5 binding requires a 5’ monophosphate or polyphosphate cap

• IFIT5 folds as an intramolecular TPR eddy that scaffolds unique additional helices

• IFIT5 co-localizes with RIG-I to actin structures at the apical cell surface