Plasmids

RNA reporter plasmids were constructed using the pHIV LTR vector [25]. The MPMV CTE, corresponding to nucleotides 7388–7549 (Genbank Accession #NC_001550 and kindly provided by David Rekosh) was PCR amplified, introducing an AflII site at the 5′ end and an NheI site followed by a 28-nucleotide BIV TAR hairpin [26] and an SpeI site at the 3′end, and cloned to create pHIV LTR CTE BTAR GFP. A corresponding CAT reporter was created by PCR amplifying a fragment with BstXI and NotI ends and cloning in place of GFP. The pHIV LTR CTE BTAR vectors contain the CTE followed by the BTAR 28-nucleotide hairpin with an additional lower stem of HIV-1 TAR.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Mapping the interaction domains of hnRNP A1.

(A) Schematic diagram of the hnRNP A1 constructs (1–320, 1–194, 195–320, and 1–234) used for in vitro pull down or Tat-hybrid reporter experiments. (B) GST pull-down assays with hnRNP A1 and Tap. This experiment utilized GST-hnRNP A1 (1–320) or GST-Tap (1–619) and the corresponding [³⁵S]-labeled in vitro translated (IVT) proteins as indicated. The two input lanes (5 and 6) represent 1/10 of the in vitro translated proteins used for binding. GST alone was used as a negative binding control. (C) GST-Tap pull-down assays of full-length hnRNP A1 and two deletion mutants (1–194 and 1–234). The three input lanes (7–9) represent 1/10 of the in vitro translated proteins used for binding. GST alone was used as a negative binding control. (D) Activity and specificity of the hnRNP A1 truncations in the Tat-hybrid assay. Activities were determined using the same methods as in Fig. 2. Tap 1–372 and BIV Tat were included as positive controls and HIV Tat as a negative control.

https://doi.org/10.1371/journal.pone.0048194.g004

To express HIV Tat fusion proteins, we generated a mammalian codon-optimized pSV2-Tat vector with a NotI site (encoding Gly-Gly-Arg following Tat amino acid 72) at the 3′ end of the gene by cloning a synthetic HindIII-XhoI fragment into pSV2 Tat [26], [27]. We inserted an additional 2 kB stuffer fragment to facilitate purification of the doubly-digested NotI-XhoI vector for fusion protein cloning [25]. Amino acids 2–372 from a TAP cDNA (kindly provided by E. Izzuaralde), as well as hnRNP A1 variants with amino acids 1–194 or 195–320 or mutants using cDNA templates, were PCR amplified and cloned as in-frame fusions into the codon-optimized vector. Clones to express GST fusion proteins were generated by inserting PCR fragments encoding Tap 1–619 or hnRNP A1 1–320 into BamHI and EcoRI sites of pGEX2T. Plasmids for in vitro translation were cloned into BamHI and HindIII or AflII and BamHI sites in pcDNA3.1+ (Invitrogen).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Properties of the out-of-frame library clones.

https://doi.org/10.1371/journal.pone.0048194.t002

Cell lines

A HeLa cell line expressing the pcDNA3 HIV-1 LTR-CTE-BTAR-GFP reporter was generated to obtain a consistent background for library screens. The plasmid was tranfected into HeLa cells using Lipofectin (Invitrogen) and stable integrants were selected using neomycin (G418) (800 µg/ml) for 10 days. Resistant cells were transfected with the pSV2 BIV Tat plasmid to activate the GFP reporter and GFP-expressing cells were isolated by FACS into 96-well plates. Stable cell lines were evaluated by transfecting with pSV2 BIV Tat plasmid and monitoring GFP signal-to-noise by FACS scans to identify those with highest activation levels.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Analyses of non-coding library clones.

Activities and specificities of the out-of-frame clones in the Tat-hybrid assay were determined as in Fig. 2. All fusions were independently tested using the HIV TAR CAT reporter to confirm that expression levels and transfection efficiencies were similar (data not shown).

https://doi.org/10.1371/journal.pone.0048194.g005

cDNA library

PolyA+ mRNA prepared from HeLa cells grown in spinner flasks was reverse transcribed using a poly(T)₁₈ primer with an XhoI site at the 5′end and BRL SuperscriptII reverse transcriptase according to manufacturers instructions in the presence of 10 mM 5-methyl dCTP. Second strand synthesis was carried out using E. coli DNA ligase, E. coli DNA pol I, and RNase H according to manufacturer instructions. cDNA was size-fractionated on a 5–20% potassium acetate gradient. NotI adapters (5′GGCCGCGCTCTCAGTG, 5′CACTGAGAGCGC) were ligated to double-stranded cDNA, and cDNA was digested with XhoI and cloned into pSV2Tat containing the stuffer. 6.7×10⁶ independent library clones were obtained and >75% had an insert of average size ∼1.8 kB.

Library screening

Protoplast fusion was carried out as previously described [8]. Briefly, the cDNA library was transformed into E. coli DH-5α and grown to an O.D. 600 of 1.0 at 37°C in LB media containing 100 µg/ml ampicillin. The plasmids were then amplified overnight with 250 µg/ml chloramphenicol. Protoplasts were prepared as described [28] and diluted into DMEM containing 10% sucrose and 10 mM MgCl₂. CTE HeLa cells were grown in 75 cc flasks and washed with serum free DMEM. A protoplast suspension was added to the flask and centrifuged at 1,650 g for 10 minutes at room temperature. The supernatant was removed and 4 ml of 50% (w/v) PEG 1500 diluted in DMEM was added. After 2 minutes, the supernatant was removed and cells were washed twice with serum-free DMEM. The cells were resuspended in DMEM containing 10% fetal bovine serum and penicillin/streptomycin (100 Units/100 µg/ml), and incubated for 48 h.

Fluorescence activated cell sorting (FACS)

Protoplast-fused cells were harvested with cell dissociation buffer (CDB, Invitrogen), washed with PBS, and resuspended in PBS containing 5% CDB, 0.3% fetal bovine serum, and 1 µg/ml propidium iodide. In the initial round of cell sorting, a sorting window was drawn manually to include approximately 0.1% background cells. Subsequent sorting windows were drawn to enrich for more highly expressing GFP-positive cells. Samples were analyzed using a FACScan flow cytometer (Becton Dickinson) or sorted using a FACSvantage (Becton Dickinson) cell sorter. Cells were excited at 488 nm and a 530+/−10 nm band pass filter was used to detect GFP emission. Flow cytometry was performed at the Howard Hughes Medical Institute (University of California, San Francisco) and FACS data were analyzed using CELLQuest softward (Becton Dickinson).

Plasmid recovery

Sorted cells were mixed with 10–20,000 carrier cells, spun down, and resuspended in 10 µl of TE buffer (10 mM Tris, pH 8.0, 1 mM EDTA) containing 0.2 mg/ml tRNA. Plasmids were prepared by alkaline lysis [29]. Electrocompetent cells were prepared by growing 500 ml cultures of E. coli DH-5α to O.D. 600 of 0.7, centrifuging to remove the media, and washing with 2 L of ice cold water through a 0.22 µM sterile filter. Cells were resuspended in 10% glycerol, and 50 µl aliquots were used for electroporation. High efficiencies of transformation were achieved using these methods (approximately 5×10¹¹ colonies/µg of DNA).

Transfections, CAT assays, and FACS scans

HeLa cells or the HeLa CTE cell line were cultured in DMEM (Gibco-BRL) with 10% fetal bovine serum, penicillin/streptomycin, and grown in 12-well plates for transfection experiments. For CAT assays, 50 ng of reporter plasmid (pHIV-LTR-CTE-BTAR-CAT, pHIV-LTR-BTAR-CAT, and pHIV-LTR-CAT) and 5–25 ng of Tat-fusion plasmid (pSV2tat72-based) were mixed with carrier Bluescript plasmid (adjusted to 1 µg total DNA) in Optimem low serum medium (Gibco-BRL). DNA was transfected into cells using Lipofectin (Gibco-BRL). Cells were grown for 48 h post-transfection, harvested, and subject to CAT assays as described previously [26]. CAT activity from different reporters was normalized to the activity of BIV Tat when appropriate. Library clones were tested on the pHIV-LTR-CAT reporter to control for transfection efficiency. For FACS scans, CTE cells were transfected with the library clone, or protoplast fused with E. coli containing the library plasmid, and analyzed by FACS scan after 48 h.

Protein purification and gel shift assays

E. coli cells (strain BL21) expressing GST-fusion proteins were inoculated in LB-Amp and grown to O.D. 600 of 0.7–1.0. Plasmids were induced with 100 µM IPTG for 6–8 hours of induction with 100 µg/ml IPTG, then centrifuged, resuspended in column buffer (50 mM Tris, pH 8.0/150 mM NaCl/2 mM DTT/10% glycerol/1 mM PMSF/0.5% Triton X-100) and sonicated. The lysate was centrifuged at 15,000 g and the supernatant was applied to a column containing 1 ml glutathione agarose. The beads were washed with several volumes of column buffer, and the GST proteins were eluted with 5 mM glutathione in column buffer. For RNA transcription, the CTE was PCR amplified and cloned into the SacI and KpnI sites of a pBluescript SK vector. ³²P labelled RNA for gel shifts was transcribed using T7 polymerase and purified by SDS-PAGE. Binding reactions for gel shift assays were carried out at 4°C and contained 10 mM HEPES/50 mM KCl/2.5 mM MgCl₂/500 µM DTT/10% glycerol/0.025% NP40 and 1000–5000 cpm RNA, 0.02 mg/ml Poly (C) RNA, GST-fusion protein, and RNAsin (Promega). After binding, 0.02 mg/ml heparin was added and samples were run on 6% acrylamide Tris-Glycine native gels and analyzed using Phosphorimager software.

GST pull-down assays

E. coli cells (strain BL21) expressing GST-fusion proteins were inoculated in LB-Amp and grown to O.D. 600 of 0.7–1.0. Plasmids were induced with 100 µM IPTG and grown for 2–8 h at 30°C. Cells were centrifuged to remove the media, and resuspended in lysis buffer (20 mM Tris-HCl pH 8.0/200 mM NaCl/1 mM EDTA/0.5% NP40) with 25 µg/ml PMSF and 0.2% protease inhibitor cocktail (Sigma) [30]. The cell suspension was sonicated, centrifuged at 16,000 g, and the supernatant transferred to a fresh tube. Glutathione agarose beads (Sigma) were prepared by swelling in lysis buffer, and 50 µl of cell lysate was mixed with 50 µl of bead slurry and 2 µl of in vitro translated protein in each pull-down reaction. In vitro translated proteins were prepared using TnT T7 Quick Coupled Transcription/Translation System (Promega) according to the manufacturer's protocol with [³⁵S]-Methionine (Amersham). GST binding reactions were incubated at 4°C for 1 h. After binding, the beads were centrifuged at 16,000 g and the supernatant was removed. Beads were washed 4 times with lysis buffer, and resuspended in SDS-PAGE loading buffer. Samples were run on 10% SDS-PAGE gels and analyzed by Phosphorimager (Molecular Dynamics) using Imagequant software (Molecular Dynamics).

In vitro methylation

Methylation assays were performed essentially as described [31]. A plasmid encoding GST-Prmt1 was a kind gift of H. Herschman. 23 µl reactions containing 4 µM GST-fusion protein, 3 µl [³H]-Ado-Met (NEN Life Science, 0.55 mCi/ml), and 0.2 µg of GST-Prmt1 were incubated at 37°C for 1.5 h. The reactions were mixed with 7 µl of SDS sample buffer and samples were run on a 10% SDS-PAGE gel, transferred to a nitrocellulose filter, and analyzed using Phosphorimager and Imagequant software.

A library screen to identify CTE-interacting proteins

We have developed a screening platform for identifying proteins that interact with RNA in the context of mammalian cell lines. The platform makes use of Tat-mediated HIV transcription activation assays [8], [25] in which HIV-1 Tat enhances elongation from the HIV-1 LTR by binding to the TAR RNA site located at the 5′ end of the nascent viral mRNAs. For this screen, TAR was replaced by the CTE (Fig. 1A), and a Tat-fusion library was generated with cDNAs fused to HIV-1 Tat, expecting that fused CTE-binding proteins would bind the RNA and activate GFP expression. The GFP reporter also encoded the BIV TAR element downstream of the CTE, providing an independent binding site for BIV Tat, but not HIV Tat [26]. In practice, BIV TAR served three critical roles: first, as a positive control for reporter activity; second to quantify the efficiency of protoplast fusion delivery; and third to allow use of the BIV Tat-TAR interaction as a calibration standard to score the specificities of positive library clones. The reporter plasmid was stably integrated into HeLa cells and a cell line was selected that displayed low constitutive GFP expression but high expression when transfected with BIV Tat.

We constructed a cDNA library derived from HeLa cells that contained an average insert of ∼1.8 kb fused to amino acids 1–72 of Tat. This portion of Tat encodes both the activation domain (AD, amino acids 1–48) and arginine-rich motif (ARM) RNA-binding domain (amino acids 49–57). The ARM was included because it also serves as an NLS [32], [33] and helps generate higher GFP signals, particularly for weak RNA binding proteins, while not appreciably increasing activation levels through nonspecific RNA binding (see Figure 2, HIV Tat). Because Tat 1–72 fusions also bind HIV TAR, we derived the additional benefit of conducting parallel assays with an HIV TAR reporter to quantify functional expression of each fusion protein [10]. The library was delivered into HeLa reporter cells by protoplast fusion such that, on average, less than one bacterial cell is fused to each HeLa cell and thus delivery of a single library plasmid is nearly clonal [8], [34]. Protoplast delivery also has the advantage that GFP expression levels are relatively consistent among the fused cell population, as demonstrated by FACS analyses with the CTE-binding protein, Tap (amino acids 1–372), fused to Tat (Fig. 1B). Transient transfection, by contrast, generates a more heterogeneous population [8] and makes the choice of a sorting window for screening somewhat more arbitrary. Furthermore, recovering plasmids from GFP-positive cells sorted by FACS is more efficient following protoplast fusion, as many copies of a single plasmid are delivered per HeLa cell and can be captured by electroporation of highly competent E. coli cells [8].

To calibrate a FACS sorting window to identify positive clones and to determine an enrichment factor per round of plasmid delivery, FACS, and recovery, we conducted a mock screen by adding Tat-fused Tap (1–372) into the cDNA library at a 1∶100,000 ratio and performing two rounds of screening (Fig. 1C). In round one, we analyzed ∼18.5×10⁶ HeLa cells and sorted ∼41,000 GFP positive cells using a conservative window derived from the activity of Tat-fused Tap (Fig. 1C, center panel). Following electroporation into E. coli, we recovered 20,500 transformants and generated a plasmid pool. In a second round we used a higher sorting window to enrich for strong GFP activators (Fig. 1C, right panel), analyzed 1.85×10⁶ HeLa cells, and recovered 2300 GFP-positive cells and 900 transformants. PCR analyses of 700 transformants identified four Tap clones, representing an enrichment factor of 570 over the two rounds.

We next conducted the library screen targeting the CTE, beginning with 72×10⁶ HeLa cells and resulting in 16,500 sorted cells and 5,000 transformants after the two rounds (Fig. 1D). We conducted a third round which yielded a higher fraction of positives but chose to analyze individual clones following the second round to preserve diversity and include potentially weaker CTE-binding proteins. DNA was prepared from 1,500 clones and digested with restriction enzymes to reveal the size of the insert. About half of the clones had small or no inserts or abnormalities in the plasmid backbone and were discarded. The several hundred remaining clones were retested individually for activity on the CTE reporter and those showing moderate to high GFP levels were sequenced.

Identities and activities of CTE-interacting clones

To more quantitatively evaluate the activities of the positive Tat-fused clones, we measured the activation levels of the 17 most active clones utilizing a CTE CAT reporter instead of GFP (Fig. 2, Top). In addition, we tested the clones on an unrelated BIV TAR reporter to assess the specificity of each clone for the CTE (Fig. 2, Bottom). Figure 2 shows that several of the clones conferred strong activity on the CTE reporter and low activity on the BTAR reporter, similar to the Tat-Tap control. The largest group of positive clones encoded fusions to known RNA-binding proteins, most containing RGG repeats, and in general showed the highest activation levels (Table 1, Fig. 2). Several of the strongest and most specific clones were hnRNP proteins, including hnRNP A1 which was previously implicated in RNA export [35]. Sam68, which also showed good specificity in the Tat-hybrid assay, has been reported to enhance the activities of RRE- and CTE-dependent reporters and appears to work synergistically with Tap [36], [37]. Other identified RNA-binding proteins, such as FUS/TLS, EWS, and nucleolin, also have reported roles in export in addition to their activities in transcription, splicing, and ribosomal RNA transcription and biogenesis (reviewed in [38], [39]. Properties of the most active proteins are annotated in Table 1.

Evidence for an interaction between the CTE, hnRNP A1, and Tap

We chose to focus follow-up experiments on hnRNP A1 for several reasons: 1) it was by far the dominant clone isolated from our library screen (three different clones found 21 times); 2) it showed high activation levels and strong specificity for the CTE; 3) it contains four RGG motifs, which is representative of many of the clones isolated; and 4) it has been implicated in RNA export and thus was a plausible candidate for a CTE-interacting protein.

We tested whether hnRNP A1 interacts directly with the CTE using RNA gel shift assays. A titration of GST hnRNP A1 alone (Fig. 3, lanes 2–6) showed a ladder of discrete hnRNP A1 species bound to the 185-nt CTE RNA, indicating multiple binding sites. Similar results were observed with an antisense version of the CTE (data not shown), suggesting that these complexes are relatively non-specific. In contrast, a titration of GST Tap showed the formation of one main complex at µM concentrations with the CTE RNA (Fig. 3, lanes 8–13), similar to previously reported results [19], [40]. Because the Tat-hybrid system can assemble larger multi-protein complexes with endogenous proteins in vivo [10], we reasoned that our observation of specificity in the reporter system might reflect the ability of Tat-fused hnRNP A1 to assemble more selectively on the CTE via an interaction with endogenous Tap. Indeed, a gel shift experiment using 2.5 µM Tap and titrating hnRNP A1 (Fig. 3, lanes 14–18) showed the formation of a supershifted band on the CTE RNA (lane 18), providing evidence that hnRNP A1 and Tap can assemble into a complex with the CTE RNA in vitro.

Interacting domains of hnRNP A1

Given that hnRNP A1 is able to form a complex with CTE/Tap, we next evaluated a possible pairwise interaction between hnRNP A1 and Tap. hnRNP A1 is 320 amino acids long and contains two RRM domains (amino acids 1–184), an RGG domain (amino acids 195–234), and a C-terminal region that includes the M9 NLS/NES (amino acids 268–305) (Fig. 4A). GST-fused hnRNP A1, but not GST alone, was able to pull down in vitro translated [³⁵S]-labeled Tap (full-length 1–619) (Fig. 4B, lanes 2 and 1). Similarly, GST-Tap but not GST pulled down in vitro translated hnRNP A1 (Fig. 4B lanes 4 and 3).

To assess which domains of hnRNP A1 mediate the interaction with Tap, we performed pull-down assays with GST-tagged Tap and a set of in vitro translated hnRNP A1 deletion proteins based on the protein domain structure and endpoints of hnRNP A1. A deletion protein containing the RGG domain (1–234) interacted as well as the full-length hnRNP A1 while the RRMs alone (1–194) did not interact with Tap (Fig. 4C, Lanes 4, 5, and 6). No binding was seen to GST alone. Although we were unable to express the RGG domain (195–234) alone using these methods, the overlap between 1–234 and 1–194 constructs suggested that the RGG domain (195–234) mediates the interaction with Tap.

Deletion analysis was also used to identify regions of hnRNP A1 important for CTE interactions in the Tat-hybrid context. Deletion constructs containing either the RRMs alone (1–194) or only the RGG and M9 domains (195–320) were engineered and tested in the CTE binding assay. Both constructs conferred moderate activity and specificity (Fig. 4D) suggesting that both the RRM and RGG/M9 domains of the protein contribute to the interaction between hnRNP A1 the CTE.

Library clones were enriched for RGG domains

It was readily apparent that the list of highly activating clones was enriched for proteins that contained RGG domains, including four of the most active clones EWS, LSM14A/RAP55, nucleolin, and hnRNP R (Table 1). Intriguingly, we also identified 25 out-of-frame or non-coding clones that conferred activation of the CTE reporter and were enriched in Tat-fused RGG tri-peptide and/or RG di-peptide sequences (Table 2). The frequency of RGG sequences among ∼32,000 non-redundant human genes is 0.00027 per amino acid and was enriched about 30-fold in our screen. RG and RGG motifs often are observed in RNA-binding proteins (reviewed in [41], [42]) and are common sites of arginine methylation [24].

We tested the activity of three representative non-coding clones using the Tat-hybrid assay and these clones demonstrated modest activity and reasonable specificity for the CTE (Fig. 5). This result is consistent with the hypothesis that RGG domains are a determinant for complex formation on the CTE, as is apparently the case for hnRNP A1. In addition, several of the non-coding clones, as well as hnRNP A1, were substrates for arginine methylation in vitro (Fig. S1). While these modification data are provocative, they do not indicate whether methylation directly contributes to the CTE interaction.