Three Arginine Residues within the RGG Box Are Crucial for ICP27 Binding to Herpes Simplex Virus 1 GC-Rich Sequences and for Efficient Viral RNA Export

Abstract

ICP27 is a multifunctional protein that is required for herpes simplex virus 1 mRNA export. ICP27 interacts with the mRNA export receptor TAP/NXF1 and binds RNA through an RGG box motif. Unlike other RGG box proteins, ICP27 does not bind G-quartet structures but instead binds GC-rich sequences that are flexible in structure. To determine the contribution of arginines within the RGG box, we performed in vitro binding assays with N-terminal proteins encoding amino acids 1 to 160 of wild-type ICP27 or arginine-to-lysine substitution mutants. The R138,148,150K triple mutant bound weakly to sequences that were bound by the wild-type protein and single and double mutants. Furthermore, during infection with the R138,148,150K mutant, poly(A)⁺ RNA and newly transcribed RNA accumulated in the nucleus, indicating that viral RNA export was impaired. To determine if structural changes had occurred, nuclear magnetic resonance (NMR) analysis was performed on N-terminal proteins consisting of amino acids 1 to 160 from wild-type ICP27 and the R138,148,150K mutant. This region of ICP27 was found to be highly flexible, and there were no apparent differences in the spectra seen with wild-type ICP27 and the R138,148,150K mutant. Furthermore, NMR analysis with the wild-type protein bound to GC-rich sequences did not show any discernible folding. We conclude that arginines at positions 138, 148, and 150 within the RGG box of ICP27 are required for binding to GC-rich sequences and that the N-terminal portion of ICP27 is highly flexible in structure, which may account for its preference for binding flexible sequences.

The herpes simplex virus 1 (HSV-1) protein ICP27 is a multifunctional regulatory protein that is required for productive viral infection. ICP27 interacts with a number of cellular proteins, and it binds RNA (35). One of the functions that ICP27 performs is to escort viral mRNAs from the nucleus to the cytoplasm for translation (2, 3, 5, 10, 13, 21, 34). ICP27 binds viral RNAs (5, 34) and interacts directly with the cellular mRNA export receptor TAP/NXF1 (2, 21), which is required for the export of HSV-1 mRNAs (20, 21). ICP27 also interacts with the export adaptor proteins Aly/REF (2, 3, 23) and UAP56 (L. A. Johnson, H. Swesey, and R. M. Sandri-Goldin, unpublished results), which form part of the TREX complex that binds to the 5′ end of mRNA through an interaction with CBP80 (26, 32, 41). Aly/REF does not appear to bind viral RNA directly (3), and it is not essential for HSV-1 RNA export based upon small interfering RNA (siRNA) knockdown studies (20), but it contributes to the efficiency of viral RNA export (3, 23). ICP27 also interacts with the SR splicing proteins SRp20 and 9G8 (11, 36), which have been shown to shuttle between the nucleus and the cytoplasm (1). SRp20 and 9G8 have also been shown to facilitate the export of some cellular RNAs (16, 17, 27) by binding RNA and interacting with TAP/NXF1 (14, 16, 18). The knockdown of SRp20 or 9G8 adversely affects HSV-1 replication and specifically results in a nuclear accumulation of newly transcribed RNA during infection (11). Thus, these SR proteins also contribute to the efficiency of viral RNA export. However, the overexpression of SRp20 was unable to rescue the defect in RNA export during infection with an ICP27 mutant that cannot bind RNA (11), suggesting that ICP27 is the major HSV-1 RNA export protein that links viral RNA to TAP/NXF1.

ICP27 was shown previously to bind RNA through an RGG box motif located at amino acids 138 to 152 within the 512-amino-acid protein (28, 34). Using electrophoretic mobility shift assays (EMSAs), we showed that the N-terminal portion of ICP27 from amino acids 1 to 160 bound specifically to viral oligonucleotides that are GC rich and that are flexible and relatively unstructured (5). Here we report the importance of three arginine residues within the RGG box for ICP27 binding to GC-rich sequences in vitro and for viral RNA export during infection. We also performed nuclear magnetic resonance (NMR) structural analysis of the N-terminal portion of ICP27 for both the wild-type protein and an ICP27 mutant in which three arginines were replaced with lysines. The NMR data showed that the N-terminal portion of ICP27 is relatively unstructured but compact, and NMR analysis in the presence of oligonucleotide substrates to which the N-terminal portion of ICP27 binds did not show any discernible alterations in this highly flexible structure, nor did the arginine-to-lysine substitutions.

MATERIALS AND METHODS

Cells, viruses, and recombinant plasmids.

HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% heat-inactivated newborn calf serum. Vero cells and ICP27-complementing 2-2 cells (37) were grown in DMEM supplemented with 8% fetal bovine serum and 4% donor calf serum. HSV-1 KOS and the ΔRGG, R138,150K, and R138,148,150K RGG box mutants were described previously (38, 39). For the expression of the N-terminal 160 amino acids of ICP27 in bacteria, the codon-optimized ICP27 gene, obtained from Verdezyne, Inc. (formerly CODA Genomics), was used as described previously (5). PCR was used to amplify the sequence corresponding to the ICP27 N-terminal 160 amino acids. The PCR product was cloned into the pET21b expression vector (Novagen) with a C-terminal 6×His tag, and positive clones were verified by sequencing. The ICP27 N-terminal R148K and R138,150K RGG box mutants were made by site-directed mutagenesis using the QuikChange mutagenesis kit (Stratagene) and the wild-type pET21b ICP27 N-terminal plasmid as a template. The R148K single mutant was made by using forward primer 5′-dCGTCGCCGTGGTAAAGGTCGTGGTGG-3′ and reverse primer 5′-dCCACCACGACCTTTACCACGGCGACG-3′. The R138,150K mutant was constructed by using forward primer 5′-dGCAAAAGGTGGCCGTCGTGGTCGTCGCCGTGGTCGCGGTAAAGG-3′ and reverse primer 5′-dCCTTTACCGCGACCACGGCGACGACCACGACGGCCACCTTTTGC-3′. The R138,148,150K mutant was constructed by using the pET21b ICP27 R138K mutant as a template and forward primer 5′-dGGTCGTCGCCGTGGTAAAGGTAAAGGTGGTCCG-3′ and reverse primer 5′-dCGGACCACCTTTACCTTTACCACGGCGACGACC-3′. All mutations were verified by sequencing.

Protein expression and purification.

pET21b plasmids expressing the wild-type, codon-optimized, His-tagged ICP27 N-terminal 160-amino-acid R148K, R138,150K, or R138,148,150K N-terminal mutant were transformed into the Escherichia coli BL21 Rosetta strain (Novagen). Cultures were grown in Luria-Bertani (LB) broth with 50 μg/ml ampicillin for proteins used in EMSAs or Neidhardt's minimal medium supplemented with 19 mM ¹⁵NH₄Cl and containing 50 μg/ml ampicillin for proteins used in NMR analyses. Cells were grown to an optical density at 600 nm (OD₆₀₀) of 0.8 to 1.0 and were induced with 100 mM IPTG (isopropyl-β-d-thiogalactopyranoside) (Sigma) for 3 h at 37°C. His-tagged ICP27 proteins were purified by using Ni-nitrilotriacetic acid (NTA) agarose (Qiagen) according to the manufacturer's recommendations for native protein purification. The ICP27 N-terminal peptide was dialyzed into 50 mM Tris buffer (pH 8) for EMSA or NMR buffer (50 mM Na₂HPO₄, 50 mM NaH₂PO₄, 100 mM NaCl, and 200 mM K₂SO₄ [pH 8.0]) for NMR analysis and concentrated with a Centriprep centrifugal filter device (Millipore). The Coomassie Plus Bradford assay kit (Pierce) was used to determine protein concentrations.

Electrophoretic mobility shift assay.

Oligonucleotides from the HSV-1 glycoprotein C (gC) gene that were shown previously to bind ICP27 in EMSAs (5) were used in the EMSAs shown in Fig. 1. The numbering and derivation of these gC sequences was described previously (5, 6). DNA oligonucleotides were used here because they are more stable, but similar results were found by EMSA with the corresponding RNA oligonucleotides (5) Five picomoles of gC oligonucleotides gC 1-30 (5′-dCGCCGACCCTCCGTTGTATTCTGCACCGG-3′), gC 11-40 (5′-dCCGTTGTATTCTGTCACCGGGCCGCTGCCG-3′), and gC 31-60 (5′-dGCCGCTGCCGACCCAGCGGCTGATTATCGG-3′) were radiolabeled with [γ-³²P]ATP with Optikinase (USB) and purified by using the Qiaquick nucleotide removal kit (Qiagen). Twenty femtomoles of each oligonucleotide was incubated with increasing concentrations from 2.5 to 62.5 μM wild-type ICP27 N-terminal peptide or the R148K, R138,150K, or R138,148,150K RGG box mutant in 1× binding buffer (20 mM Tris [pH 8], 150 mM KCl, 1 mM EDTA [pH 8], and 1 mM dithiothreitol [DTT]) with 10% glycerol and 300 μg/ml bovine serum albumin (BSA) for 30 min at 37°C. Samples were loaded onto a prerun 5% acrylamide-bisacrylamide gel with 2.5% (wt/vol) glycerol and 1× Tris-acetate buffer and were subjected to electrophoresis for 2 h at 35 mA. Gels were dried onto filter paper under a vacuum and exposed to film.

FIG. 1.

The R138,148,150K mutant binds poorly to GC-rich sequences bound by wild-type (WT) ICP27. (A) Schematic representation of ICP27 showing the position and amino acid sequence of the RGG box. Arginines at positions 138, 148, and 150 are shown in red. NLS, nuclear localization signal. (B) Twenty femtomoles of radiolabeled oligonucleotides gC 1-30 and gC 31-60 was incubated with no protein (−) or with increasing concentrations of 2.5, 12.5, or 62.5 μM wild-type ICP27 N-terminal protein or the R148K, R138,150K, and R138,148,150K RGG box mutants. Samples were fractionated on a prerun acrylamide gel, and dried gels were exposed to film. Arrows indicate the migration of free probe and the shift due to protein binding. (C) Twenty femtomoles of radiolabeled gC 11-40 oligonucleotide was incubated with no protein or with 2.5, 12.5, or 62.5 μM wild-type ICP27 N-terminal protein or the R148K, R138,150K, or R138,148,150K mutant.

In situ hybridization.

Cells grown on coverslips in 24-well plates were fixed in 3.7% formaldehyde after infection and then overlaid with 70% ethyl alcohol (EtOH) and stored at 4°C to permeabilize the cells. Cells were rehydrated for 5 min at room temperature in 15% formamide in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and then overlaid with 40 μl hybridization solution [15% formamide, 10% dextran sulfate, 40 μg yeast tRNA, 0.02% BSA, 5 ng biotinylated oligo(dT) (Promega), RNasin, 0.5 M DTT, 2× SSC] and incubated at 37°C for 90 min. Cells were washed twice for 30 min each at 37°C in wash solution (15% formamide, 2× SSC, 0.1% NP-40) and then immunostained with a monoclonal antibody to ICP27 (P1119; Virusys), as described previously (21). Cells were viewed by fluorescence microscopy at a magnification of ×100 with a Zeiss Axiovert S100 microscope.

Bromouridine labeling.

HeLa cells were infected at a multiplicity of infection (MOI) of 10 with HSV-1 KOS or the ΔRGG, R138,150K, and R138,148,150K RGG box mutants. At 7 h after infection cells were labeled with 4 mM 5-bromouridine (BrU) (Sigma) for 30 min. Cells were washed twice with phosphate-buffered saline (PBS) and then incubated with complete medium without 5-BrU for 45 min before the cells were fixed with 3.7% formaldehyde. When indicated, 10 μg/ml actinomycin D (Sigma) was added with 5-BrU to stall transcription (21). Cells were fixed and immunostained as described above. Fixed cells were stained with anti-BrU antibody used at a 1:100 dilution (Calbiochem) followed by biotinylated goat anti-mouse IgG at 1:200 (Pierce) and streptavidin-Texas Red (GE Healthcare). ICP4 was detected with anti-ICP4 monoclonal antibody P1114 (Virusys) at a 1:500 dilution directly conjugated with Alexa 488 (Invitrogen) according to the manufacturer's instructions. DAPI (4′,6-diamidino-2-phenylindole) staining was used to mark nuclei. Cells were viewed by fluorescence microscopy at a magnification of ×100 with a Zeiss Axiovert S100 microscope.

Circular dichroism.

The secondary structure of ICP27 was analyzed by circular dichroism (CD) spectroscopy. The purified ICP27 N-terminal protein in 50 mM Tris (pH 8.0) was analyzed at 1.0-nm-wavelength intervals by using a Jasco model 720 CD spectropolarimeter (Jasco, Easton, MD) at a scan speed of 50 nm/min and an average response time of 5 s. A total of 10 consecutive scans were accumulated for analysis. Measurements were made at 25°C using a 1-mm-path-length cell (Helma).

NMR analysis.

All NMR protein samples were prepared in the same buffer (50 mM Na₂HPO₄, 50 mM NaH₂PO₄, 100 mM NaCl, 200 mM K₂SO₄ [pH 8], and 10% D₂O) at a final concentration of 0.3 mM. The gC oligonucleotides used in this study (gC 1-30, gC 11-40, gC 31-60, gC 71-100, and gC 221-250) were described previously (5). The gC oligonucleotides (Operon) were resuspended in a solution containing 50 mM Tris (pH 8), 150 mM KCl, and 1 mM EDTA (pH 8) at a final concentration of 10 mM. Chelex 100 resin (Bio-Rad) was added to each oligonucleotide solution and incubated with rotation for 1 h at room temperature to remove contaminating metals. The Chelex resin was removed by centrifugation. For NMR samples that contained the ICP27 N-terminal protein and gC sequences, the gC oligonucleotides were added at four times the protein molar concentration. NMR spectra were recorded with a Varian 800-MHz spectrometer at 25°C. The ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) pulse sequence (22) was used for the characterization of the wild-type ICP27 N terminus, ICP27 RGG box mutants, and ICP27 N-terminal proteins incubated with gC sequences. HSQC spectra were 1,024 points (¹H) and 60 points (¹⁵N), zero filled, and processed with shifted sine function. Nuclear Overhauser effect spectroscopy (NOESY) HSQC (22, 43) was used to characterize the ICP27 N terminus alone and the ICP27 N-terminal protein incubated with gC 1-30. NOESY HSQC spectra were 1,024 points (¹H) in the direct dimension and 56 points (¹H) and 28 points (¹⁵N) in the indirect dimension, zero filled, linear predicted, and processed with shifted sine function. Heteronuclear NOE analysis (12) was used to characterize the ICP27 N-terminal protein. A reference heteronuclear NOE spectrum was collected without a relaxation delay and compared with spectra collected with a 3-ms relaxation delay. Spectra were 1,024 points (¹H) and 64 points (¹⁵N), zero filled, processed with shifted sine function in both dimensions, and linear predicted in the ¹⁵N dimension. All NMR data were processed with NMRpipe (9).

RESULTS

Three arginine residues within the RGG box of ICP27 are required for binding HSV-1 glycoprotein C sequences.

Recently, we reported that several GC-rich sequences from the glycoprotein C (gC) gene, which were derived as 30-mers from a 300-nucleotide clone that was positive for ICP27 binding in a yeast three-hybrid screen (5), bound the N-terminal portion of ICP27 from amino acids 1 to 160 in EMSAs (5, 6). This portion of ICP27 includes the RGG box from residues 138 to 152 (Fig. 1A). The importance of the RGG box for the binding of these gC sequences was tested by EMSA using several RGG box point mutants that were described previously (38, 39). Specifically, arginines at positions 138, 148, and 150 were replaced with lysine residues either singly or in pairs, or all three arginines were replaced. These arginines were chosen because our previous studies showed that these arginines are methylated during infection and that the methylation of these arginines regulates ICP27 export to the cytoplasm and its interaction with two cellular proteins (38, 39). Recently, it was shown that the arginine methylation of Aly/REF promotes the efficient handover of mRNA to TAP/NXF1, suggesting that arginine methylation can regulate the RNA-binding activity of Aly/REF (19). The proteins encoding the N terminus of wild-type ICP27 or RGG box point mutants were expressed in bacteria, and purified proteins were incubated with individual radiolabeled gC oligonucleotides at 37°C and resolved by nondenaturing acrylamide gel electrophoresis (see Fig. S1 in the supplemental material). The numbering of the gC oligonucleotide sequences within the 300-nucleotide yeast three-hybrid clone was described previously (5). In the previous study, the N-terminal portion of ICP27 (residues 1 to 160) was shown to shift or not shift specific sequences in EMSAs (5). The gC oligonucleotide gC 1-30 was shifted well by the wild-type ICP27 N-terminal protein (Fig. 1B). The substitution of lysine for arginine at residue 148 (R148K) or at residues 138 and 150 (R138,150K) did not affect the ability of the ICP27 N terminus to shift the gC 1-30 sequence. An arginine-to-lysine substitution at residue 138 or 150 also did not affect the ability of the ICP27 N terminus to shift gC 1-30 (data not shown). However, triple arginine-to-lysine substitutions at residues 138, 148, and 150 showed reduced binding because the mutant protein did not efficiently shift gC 1-30. These same arginine substitution mutants were also tested for their ability to shift other gC sequences. The gC 31-60 sequence, which was not shifted well by the wild-type ICP27 N-terminal protein, was also not shifted by the triple arginine mutant (Fig. 1B) or single or double arginine mutants (data not shown). This finding suggests that binding specificity was not altered in the RGG box mutants but that the ability to bind was affected in the triple mutant. The wild-type ICP27 N-terminal protein and the R148K and R138,150K RGG box mutants shifted oligonucleotide gC 11-40 very well, but the R138,148,150K RGG box triple mutant shifted gC 11-40 much less efficiently, indicating reduced binding (Fig. 1C). Quantification of the shifted bands by densitometry revealed that the intensity of the bands shifted by the R138,148,150K mutant for gC 1-30 and gC 11-40 was in the range of 7- to 15-fold lower than that of the wild-type protein or the single and double mutants (see Fig. S2 in the supplemental material). These results suggest that three arginines at positions 138, 148, and 150 within the RGG box motif are all required to efficiently shift gC sequences.

Poly(A)⁺ RNA and newly transcribed RNA accumulate in the nucleus of cells infected with the R138,148,150K mutant.

We showed previously that during infection with the RGG box deletion mutant ΔRGG, poly(A)⁺ RNA and newly transcribed RNA accumulated in the nucleus, indicating that ICP27 RNA binding through the RGG box is required for efficient viral RNA export (21). To determine if arginine residues 138, 148, and 150 within the RGG box were also required for efficient RNA export by ICP27 during infection, we looked at the localization of poly(A)⁺ RNA at 8 h after infection, when viral transcription is much more active than cellular transcription (40). Cells were infected with wild-type HSV-1 KOS and the ΔRGG, R138,150K, and R138,148,150K mutants (Fig. 2). Poly(A)⁺ was visualized by hybridization with an oligo(dT) probe. ICP27 is actively shuttling at 8 h after infection and was seen to be predominantly cytoplasmic in both wild-type and mutant infections (Fig. 2). Poly(A)⁺ was seen to be distributed in the cytoplasm of KOS-infected cells, as expected, whereas poly(A)⁺ RNA was confined to the nucleus in ΔRGG-infected cells, as seen previously. Poly(A)⁺ RNA was seen to be present in the cytoplasm of cells infected with the R138,150K mutant, but in contrast, poly(A)⁺ RNA was nuclear in cells infected with the R138,148,150K mutant (Fig. 2). Furthermore, exclusively nuclear fluorescence was seen for more than 90% of the cells visualized for the R138,148,150K mutant. This result was similar to what was seen for the EMSA (Fig. 1) in that the double mutant was able to bind to gC sequences bound by the wild-type protein, but the triple mutant did not.

FIG. 2.

Poly(A)⁺ RNA accumulates in the nucleus of cells infected with the R138,148,150K mutant. HeLa cells were infected with wild-type HSV-1 KOS or with the ΔRGG, R138,150K, or R138,148,150K RGG box mutant, as indicated, at an MOI of 10. For treatment with AdOx, cells were mock infected or infected with KOS in the presence of AdOx (20 μM), which was added 2 h after infection. At 8 h after infection, cells were fixed, and in situ hybridization was performed by using a biotinylated oligo(dT) probe to detect poly(A)⁺ RNA. Poly(A)⁺ RNA was visualized by using a streptavidin-Texas Red secondary antibody. ICP27 was detected by staining with antibody P1119 (Virusys).

In previous studies we showed that arginines 138, 148, and 150 were methylated during infection and that arginine methylation regulates ICP27 export and its interaction with two cellular proteins, SRPK1 and Aly/REF, which interact with ICP27 through the RGG box region (38, 39). To determine whether methylation may affect the ability of the RGG box to bind RNA, we added the methylation inhibitor adenosine dialdehyde (AdOx) to KOS-infected cells and looked at the localization of poly(A)⁺ RNA (Fig. 2). RNA was efficiently exported to the cytoplasm in the presence of AdOx, which causes a hypomethylation of ICP27 (38). This indicates that arginine methylation is not required for the ICP27 RGG box to bind RNA. However, we cannot discount the possibility that arginine methylation may regulate the disassociation of ICP27 from RNA in the cytoplasm, perhaps in the handover to the translation machinery.

To look at newly transcribed RNA, we labeled mock-, KOS-, ΔRGG-, R138,150K-, and R138,148,150K-infected cells with 5-bromouridine (BrU), which becomes incorporated into newly synthesized RNA (24). Following a 30-min pulse at 7 h after infection, a chase was performed by incubating the cells in label-free medium for 45 min, at which time cells were fixed and immunostained. Labeled RNA was detected with an antibody specific to BrU. In mock- and KOS-infected cells, BrU-labeled RNA was clearly seen in the cytoplasm (Fig. 3), indicating the efficient export of newly synthesized RNA. Similarly, cytoplasmic BrU staining was seen in R138,150K-infected cells. In contrast, BrU-labeled RNA accumulated in the nucleus of cells infected with the ΔRGG and R138,148,150K mutants (Fig. 3). This result was found for the majority (more than 80%) of R138,148,150K-infected cells that incorporated the BrU label. We reported previously that replication compartment formation was delayed until about 12 h after infection in cells infected with the ΔRGG, R138,150K, and R138,148,150K mutants (38). In Fig. 3, at 8 h after infection, full-blown replication compartments were seen for wild-type-infected cells, but replication compartments were restricted to prereplication sites for the mutant infections. The finding that BrU was specifically incorporated into newly transcribed RNA was verified by treating mock- and KOS-infected cells with the transcriptional elongation inhibitor actinomycin D (Fig. 4). No label was incorporated in the presence of actinomycin D. We conclude from the results shown in Fig. 2 and 3 that arginine residues 138, 148, and 150 within the RGG box are essential for ICP27 binding and the export of RNA during infection. We previously showed that the R138,148,150K triple mutant was defective in viral replication (38), and therefore, we suggest that the defect in viral RNA export that we report here contributes to the defect in replication during infection with this mutant.

FIG. 3.

Newly transcribed RNA accumulates in the nucleus in R138,148,150K mutant-infected cells. HeLa cells were mock infected or were infected with HSV-1 KOS or the ΔRGG, R138,150K, or R138,148,150K mutant at an MOI of 10. At 7 h after infection, 4 mM 5-BrU was added for 30 min. Cells were washed twice in PBS, and label-free medium was added for 45 min, at which time the cells were fixed and immunostained with antibody specific for BrU and antibody specific for ICP4 to visualize replication compartments. DAPI staining was also performed.

FIG. 4.

Bromouridine is not incorporated in the presence of the transcriptional elongation inhibitor actinomycin D. Cells were mock infected or were infected with HSV-1 KOS. At 7.5 h after infection, cells were treated with BrU alone or BrU and actinomycin D (ActD) (10 μg/ml) and incubated for 30 min at 37°C. Cells were fixed and stained with anti-BrU antibody (red), anti-ICP4 antibody (green), and DAPI (blue).

Structural analysis of the N-terminal 160 amino acids of ICP27.

We previously presented evidence from NMR analysis that the N-terminal 160 amino acids of ICP27 are in a flexible configuration (6). To gain a better understanding of the structure of the N terminus of ICP27, circular dichroism (CD) was performed on the N-terminal 160 amino acids, which includes the RGG box. CD spectra allow the qualitative analysis of protein secondary structure and yield distinct spectra for alpha helix, beta sheet, and random-coil proteins. The CD spectra of the ICP27 N-terminal 160 amino acids (Fig. 5A) were consistent with a protein in a random-coil conformation due to the shape of the CD curve (31).

FIG. 5.

Circular dichroism and ¹H-¹⁵N HSQC spectra for the ICP27 N-terminal region are consistent with a protein in a random-coil secondary structure. (A) Circular dichroism spectra for the ICP27 N-terminal protein from amino acids 1 to 160 were collected between 195 nm and 260 nm as described in Materials and Methods. mdeg, millidegrees. (B) ¹H-¹⁵N HSQC spectra for the ICP27 N-terminal protein alone performed at 25°C are shown. Panels are set to include all peaks detected expect for tryptophan. (C) ¹H-¹⁵N HSQC spectra for the ICP27 N-terminal protein alone are shown as blue spectra, with the protein spectra collected in the presence of gC 1-30 sequences overlaid as pink spectra. (D) ¹H-¹⁵N HSQC spectra for the ICP27 N-terminal protein alone are shown as blue spectra, with the protein spectra collected in the presence of gC 11-40 overlaid as pink spectra. (E) ¹H-¹⁵N HSQC spectra for the ICP27 N-terminal protein alone (blue spectra) with the protein spectra collected in the presence of gC 31-60 (pink spectra).

Next, we performed NMR analysis to further investigate the structure of the ICP27 N terminus. Because our previous results showed that the N-terminal 160 amino acids of ICP27 are highly flexible (6), we sought to determine if ICP27 bound to a gC oligonucleotide would confer more rigidity or folding on the N terminus. This was seen in NMR analyses of the cellular protein fragile X mental retardation protein (FMRP) when the RGG box was bound to G-quartet RNA substrates (30). The protein encoding the N-terminal 160 amino acids of ICP27 was expressed in E. coli Rosetta cells in minimal medium containing ¹⁵NH₄Cl to isotopically label the expressed proteins. Purified protein was subjected to a heteronuclear single quantum coherence (HSQC) analysis to evaluate the nitrogen and hydrogen spectra of the protein backbone. ¹H-¹⁵N HSQC spectra for the ICP27 N-terminal protein alone showed peaks that were not well dispersed in the proton dimension and that resonated between 7.5 and 9.0 ppm (Fig. 5B). Proton HSQC chemical shifts in this range are typical of shifts found in proteins with a random-coil conformation (42). Each HSQC signal represents one N-H group in the protein, and there should be at least 1 peak per amino acid of the protein. Approximately 70 clearly defined HSQC peaks were observed in the ICP27 N-terminal HSQC spectra, which is less than half of the expected 160 HSQC peaks. There was also a high degree of signal overlap in the region between 8.0 and 9.0 ppm in the proton dimension and in the region between 120 and 124 ppm in the nitrogen dimension, which prevents the resolution of all 160 HSQC peaks. The lack of peak dispersion and the high degree of peak overlap in the ICP27 N-terminal HSQC spectrum suggests that the protein is not in a single conformation, nor is it rigidly folded.

To determine the effect of binding on conformation, the gC 1-30, gC 11-40, and gC 31-60 sequences were used in NMR experiments with the ¹⁵N-labeled ICP27 N-terminal protein (Fig. 5C to E). The ICP27 N-terminal protein was incubated with two different gC sequences that were shifted well (gC 1-30 and gC 11-40) and one sequence (gC 31-60) that was not shifted in EMSA experiments (5). ¹H-¹⁵N HSQC spectra were collected (Fig. 5C to E). The HSQC spectra collected on the ICP27 N terminus alone are shown in blue, and the spectra collected on the ICP27 N terminus incubated with gC 1-30, which binds ICP27, are overlaid in pink (Fig. 5C). The HSQC spectra for gC 11-40, which also binds ICP27, are shown in Fig. 5D, with the protein alone in blue and the spectra gathered in the presence of gC 11-40 in pink. The protein peak position is dependent on the chemical environment of the amide N-H group, and chemical shifts for the N-H groups involved in the interaction would change. There were no detectable shifts in the peak position when gC sequences that bind specifically to ICP27 were added (Fig. 5B to D). A similar spectrum was observed when gC 31-60, which was not bound by ICP27 (5), was mixed with the ICP27 N-terminal protein (Fig. 5E). This finding suggests that regions of the protein involved in the binding of gC sequences are not visible in the two-dimensional HSQC spectra and either are in the overlapping region between 7.5 and 9.0 ppm in the hydrogen dimension or are dynamic and flexible residues not able to give well-defined peaks. These results indicate that the binding of the RGG box to oligonucleotide sequences does not confer more structure in this region.

Heteronuclear nuclear Overhauser effect (NOE) analysis was performed to further investigate the structural dynamics of the ICP27 N terminus. NOEs are interactions between the nuclear dipoles. The NOE from hydrogen to nitrogen for the ICP27 N-terminal protein is shown in Fig. 6. The efficiency of this process is dependent on dynamics. Compact regions usually give positive NOE signals (blue), while disordered or random-coil regions give negative NOE signals (pink). Rapid motion for the side chains of glutamine and asparagine are evident by the negative peaks, which are contoured pink, and are located between 112 and 113 ppm in the nitrogen dimension and at 7.6 ppm in the hydrogen dimension (Fig. 6). However, the NOE signals for the bulk of the ICP27 N-terminal protein were positive NOE signals, which is consistent with a compact state and not a random-coil or fully unfolded state. These data indicate that although the protein signals resonate in the random-coil region, the molecule is somewhat folded.

FIG. 6.

HET-NOE spectrum of the ICP27 N-terminal 160 amino acids reveals a flexible yet compact conformation. A portion of the HET-NOE spectrum between 7.5 and 8.8 ¹H ppm and 107 and 129 ¹⁵N ppm collected with a 3-ms relaxation delay were collected for the ICP27 N-terminal protein alone at 25°C and compared to a reference spectrum collected with no relaxation delay. Blue peaks are amide N-H groups in the protein backbone with positive NOE signals. Pink peaks (marked with arrows) are N-H groups with negative NOE signals. The pink peak between 112 and 113 ppm on the ¹⁵N axis represents the flexible-side-chain N-H group for the amino acids glutamine and asparagine.

Next, three-dimensional NOESY-HSQC analysis was performed on the ICP27 N-terminal protein alone, which was incubated with the gC 1-30 sequence to resolve the overlapping peaks between 7.5 and 9.0 ppm in the proton dimension and aid in the identification of peaks shifted upon the binding of gC 1-30. Four representative nitrogen planes from the NOESY-HSQC spectra collected on the ICP27 N-terminal protein are shown in blue in Fig. 7. The corresponding spectra collected on the ICP27 N-terminal protein incubated with gC 1-30 are overlaid in pink. In the portion of the ¹⁵N planes shown, the first NOE observed resonated at 4.7 ppm, indicating a strong NOE between water and the amide proton in the protein backbone. This finding suggests that the amide protons in the protein backbone of the ICP27 N terminus are exchanging rapidly with the solvent. This was true whether the ICP27 N terminus was analyzed on its own or when it was incubated with the gC 1-30 oligonucleotide (pink peaks). Additional NOE peaks were observed vertically in the ¹H dimension above the 4.7-ppm peak, likely corresponding to protons that are components of the side-chain groups of the amino acid residues. No long-range NOE signals were observed above 0.5 ¹H ppm. This indicates a lack of long-range NOEs and intramolecular interactions, which are required to undertake a structure determination. The addition of gC-binding substrates did not alter the number of NOE peaks detected (Fig. 7).

FIG. 7.

NOESY-HSQC spectrum of the ICP27 N-terminal protein alone and incubated with gC 1-30 sequences. Panels represent portions of selected ¹⁵N frequencies from NOESY-HSQC analysis of the ICP27 N-terminal protein alone (blue) or incubated with gC 1-30 sequences (overlaid in pink) between 7.9 and 8.7 ¹H ppm on the x axis and 5.1 and 0.1 ¹H ppm on the y axis. (A) 108.79 ¹⁵N ppm; (B) 119.88 ¹⁵N ppm; (C) 122.92 ¹⁵N ppm; (D) 124.37 ¹⁵N ppm.

Arginine-to-lysine substitutions at positions 138, 148, and 150 have no discernible effect on structure.

Because the mutation of arginines 138, 148, and 150 to lysines in the ICP27 RGG box impaired the ability of the ICP27 N terminus to bind HSV-1 gC sequences in EMSAs (Fig. 1), the effect of these mutations on the structural conformation of the ICP27 N terminus was investigated by NMR. A ¹H-¹⁵N HSQC spectrum was collected for the wild-type ICP27 N-terminal protein (Fig. 8A) and the R138,148,150K RGG box mutant (Fig. 8B). The wild-type ICP27 spectrum in blue is overlaid with the RGG box mutant spectra in pink (Fig. 8C). There were no detectable differences observed in the HSQC spectrum of the RGG box mutant compared to the wild-type spectrum, even though a difference in binding gC sequences was observed (Fig. 1). There was also no difference compared to the wild-type protein for spectra collected on the single (R138K and R150K) or double (R138,150K) RGG box mutant (data not shown). Because no differences were observed in the NMR spectra between the wild type and the RGG box mutant, it appears that the region of the ICP27 N terminus that binds to RNA is not structured and visible in NMR spectra and that mutants in this region that disrupt the binding of gC sequences are also not structured. These data suggest that even though the RGG box of the ICP27 N terminus is responsible for the binding of gC sequences in vitro, it is a flexible portion of the protein, and the conformation of the protein was not altered when these arginines were mutated to lysines. However, RNA binding was impaired, indicating that arginine residues 138, 148, and 150 are required for ICP27 RNA binding through the RGG box both in vitro and in vivo.

FIG. 8.

HSQC spectra of the ICP27 R138,148,150K RGG box mutant. (A) The ¹H-¹⁵N HSQC spectrum for the wild-type ICP27 N-terminal protein is shown. Panels are set to include all peaks detected except for tryptophan. (B) ¹H-¹⁵N HSQC spectrum for the R138,148,150K mutant. (C) ¹H-¹⁵N HSQC spectrum for the wild-type ICP27 N terminus (blue spectra) with the R138,148,150K mutant spectrum overlaid in pink.

DISCUSSION

In this study, we investigated which residues within the RGG box of ICP27 are required for RNA binding in ICP27's role as an export adaptor for HSV-1 mRNA. The replacement of arginine residues 138, 148, and 150 with lysine impaired ICP27 binding in vitro. RNA export during infection with the R138,148,150K mutant was also adversely affected. These studies were performed to determine how the RGG box confers specificity to ICP27 binding. Recently, we showed that ICP27 binds specifically to GC-rich sequences but that the N-terminal portion of ICP27 encoding the RGG box does not bind G-quartet structures (5). This finding was surprising because the RGG boxes of two other proteins that bind nucleic acids, FMRP and EBNA1, have been shown to bind to G-quadruplex or G-quartet structures (29, 30). G quartets are complex tertiary structures that can form readily in DNA or RNA containing three or more consecutive guanines (4, 8, 30). Not only does the N-terminal 160 amino acids of ICP27, containing the RGG box, not bind to G-quartet structures, but NMR analysis of the sequences to which it does bind showed that ICP27 also does not bind to highly structured sequences but binds only to sequences that are flexible (5). These findings have led us to propose a model by which ICP27 may distinguish between viral and cellular mRNA. The HSV-1 genome is around 68% GC rich, whereas the cellular genome is around 44% GC rich. Thus, ICP27 may be able to distinguish between viral and cellular mRNA based upon GC content. Because of its preference for flexible sequences, it would also be less prone to binding GC-rich cellular sequences such as rRNAs because these are highly structured. Thus, ICP27 may have a preference for flexible GC-rich regions within HSV-1 mRNAs.

We also investigated the structure of the N-terminal portion of ICP27 from amino acids 1 to 160 by NMR. In accordance with data from our previous study (6), we found that the N terminus of ICP27 is flexible and not well folded. NMR analysis of the RGG box region of FMRP also showed that this region was flexible and unstructured; however, when a G-quartet sequence was present, the FMRP RGG box region became more structured (30). In the case of ICP27, we did not discern any differences in the spectra of the N-terminal protein alone compared to those of the N-terminal protein in the presence of gC sequences to which it binds (Fig. 5). Furthermore, the spectrum of the R138,148,150K mutant was not altered compared to that of the wild type. This result is in contrast to NMR data collected for ICP27 N-terminal phosphorylation site mutants in which three serine residues at positions 16, 18, and 114 were mutated to alanine (S16,18,114A) or glutamic acid (S16,18,114E) (6). The ¹H-¹⁵N HSQC spectra for these ICP27 N-terminal mutants were significantly different from the wild-type spectra and were missing over 10 HSQC peaks usually detectable in wild-type spectra, indicating that the mutant proteins were even more flexible and unstructured (6). Interestingly, this disruption of the overall protein conformation did not affect the ability of the phosphorylation site mutant proteins to bind gC sequences (6).

The N-terminal region of ICP27 from amino acids 1 to 160 contains a number of functional domains. The leucine-rich nuclear export signal is at the N terminus, and the nuclear localization signal is adjacent to the RGG box (Fig. 1A). ICP27 was shown previously to interact with a number of proteins that require the N-terminal region, including RNA polymerase II (7), Hsc70 (25), TAP/NXF1 (2), Aly/REF (3), and SRPK1 (36), in addition to the requirement for the RGG box for RNA binding. Therefore, the flexibility of this region may be required for ICP27 to undergo successive interactions with proteins and RNA. Interestingly, the phosphorylation site mutants, which were found to be even more disordered than wild-type ICP27, were severely impaired in their ability to interact with proteins that interact with wild-type ICP27 (6, 33). Despite the effects on protein interactions, the phosphorylation site mutants were able to bind gC sequences with the same affinity and specificity as those of wild-type ICP27 (6). The arginine-to-lysine substitution mutants were also impaired in their ability to bind to two cellular proteins, Aly/REF and SRPK1, which interact with the region of ICP27 that spans the RGG box (39). However, this is likely due to hypomethylation rather than the change to lysine because no changes in structure were seen (Fig. 8), and the same effect on Aly/REF and SRPK1 interactions could be induced in KOS-infected cells by the addition of the methylation inhibitor AdOx (39). Arginine methylation does not contribute to RNA binding because the bacterially expressed ICP27 N-terminal protein used in EMSAs is unmethylated, and the addition of AdOx to KOS-infected cells had no effect on the export of poly(A)⁺ RNA (Fig. 2). Thus, flexibility in the N-terminal portion of ICP27 may be required for its binding to flexible sequences as well as for its interactions with other proteins.

We cannot discount the possibility that the N-terminal portion of the protein may have more structure conferred within the context of the full-length protein. We recently showed that the N and C termini of ICP27 undergo a head-to-tail intramolecular interaction using bimolecular fluorescence complementation (BiFC), in which the two halves of the Venus fluorescent protein were fused to the N and C termini of ICP27 (15). Because the renaturation of the Venus fluorescent protein results in a covalent bonding of the two halves of the Venus molecule to yield BiFC, the head-to-tail interaction of NC-Venus-ICP27 locks ICP27 in a closed configuration. This prevented ICP27 from interacting with RNA polymerase II and TAP/NXF1 (15). Thus, conferring rigidity on the flexible regions of ICP27 inhibits its interactions with several proteins. Therefore, the flexibility of the N-terminal portion of ICP27 may be the reason why ICP27 can interact with so many different proteins, as well as with itself in an intramolecular interaction, and may be why the ICP27 RGG box can recognize flexible RNA sequences instead of highly structured RNAs.