Cellular Restriction of Retrovirus Particle-Mediated mRNA Transfer

Abstract

Analyzing cellular restriction mechanisms provides insight into viral replication strategies, identifies targets for antiviral drug design, and is crucial for the development of novel tools for experimental or therapeutic delivery of genetic information. We have previously shown that retroviral vector mutants that are unable to initiate reverse transcription mediate a transient expression of any sequence which replaces the gag-pol transcription unit, a process we call retrovirus particle-mediated mRNA transfer (RMT). Here, we further examined the mechanism of RMT by testing its sensitivity to cellular restriction factors and short hairpin RNAs (shRNAs). We found that both human TRIM5α and, to a lesser extent, Fv1 effectively restrict RMT if the RNA is delivered by a restriction-sensitive capsid. While TRIM5α restriction of RMT led to reduced levels of retroviral mRNA in target cells, restriction by Fv1 did not. Treatment with the proteasome inhibitor MG132 partially relieved TRIM5α-mediated restriction of RMT. Finally, cells expressing shRNAs specifically targeting the retroviral mRNA inhibited RMT particles, but not reverse-transcribing particles. Retroviral mRNA may thus serve as a translation template if not used as a template for reverse transcription. Our data imply that retroviral nucleic acids become accessible to host factors, including ribosomes, as a result of particle remodeling during cytoplasmic trafficking.

Retroviruses enter cells in a receptor-mediated manner, following which a reverse transcription (RT) complex is formed in the cytoplasm to reverse transcribe the genomic mRNA into double-stranded DNA. During the completion of RT, a hybrid virus-cellular nucleoprotein structure known as the preintegration complex (PIC) is formed. Eventually, active transport of the PIC into the nucleus or dissolution of the nuclear membrane during mitosis allows the viral integrase to integrate the viral double-stranded DNA into chromosomal cellular DNA (35). We have previously shown that retroviral vector mutants that are unable to initiate RT of their capped, plus-stranded mRNA genomes mediate a transient expression of the sequences cloned into the gag-pol-equivalent position of the vector genome (8). Particles conferring this activity required the presence of the retroviral mRNA packaging signal within the vector sequence, as well as the expression of both Gag and Env in the vector packaging cell, but not reverse transcriptase. We refer to this previously unexplored aspect of the retrovirus life cycle as retrovirus particle-mediated mRNA transfer (RMT). Using replication-defective retroviral vectors in which the gene of interest is cloned in the position of the gag reading frame, RMT can be exploited as a novel approach for the transient expression of a gene of interest (8).

The analysis of cellular restriction factors that belong to the innate immune response against retroviruses may provide further insights into the mechanisms of RMT. The cellular restriction factor Fv1 (Friend virus susceptibility factor 1) has been shown to have an impact on the sensitivity of mice to murine leukemia viruses (MLV) (17). Further studies identified two major alleles of Fv1. Whereas the Fv1ⁿ allele confers resistance to B-tropic MLV (B-MLV) infection, but not N-tropic MLV (N-MLV) infection, in NIH mice, Fv1^b renders BALB/c mice resistant to N-MLV but susceptible to B-MLV infection (3, 9). The differences in N- and B-MLV infectivities are due to a single amino acid residue at position 110 (arginine and glutamic acid, respectively) in the retroviral capsid. The exact mechanism of action of Fv1 remains to be elucidated.

Another retroviral restriction factor is the cytoplasmic body component TRIM5α, a member of the tripartite motif (TRIM) family of proteins (10, 14, 24, 33, 40). As a defining feature of TRIM proteins, TRIM5α harbors an RBCC motif, which consists of a RING (“really interesting new gene”) domain at the N terminus followed by a B box-2 domain and a coiled-coil domain (25). The C-terminal domain of TRIM5α is a B30.2 or PRY/SPRY domain whose amino acid sequence confers the specificity of retroviral restriction (28, 32, 34, 41). Whereas rhesus monkey TRIM5α has the ability to restrict human immunodeficiency virus type 1, human TRIM5α (huTRIM5α) restricts N-MLV, but not B-MLV, infection (40). Interestingly, as for Fv1, the same amino acid residue at position 110 within the retroviral capsid controls susceptibility to TRIM5α restriction (24, 36). However, although Fv1 and huTRIM5α seem to interact with the retroviral capsid at an early postentry step, the mechanisms of restriction appear to differ. huTRIM5α usually acts before RT, whereas Fv1 allows RT but blocks subsequent steps, including integration into the host genome (2, 7, 12, 21, 36, 37).

In the present study, we examined the sensitivity of RMT to cellular restriction factors and short hairpin RNA (shRNA). We found that RMT is sensitive to restriction by both huTRIM5α and Fv1. The restriction of RMT by huTRIM5α could be partially relieved by the proteasome inhibitor MG132. Interestingly, the restriction of RMT by huTRIM5α, but not by Fv1, correlated with the degradation of the retroviral genomic RNA in the cytoplasm of infected cells. shRNAs specifically targeting the retroviral genomic RNA inhibited RMT particles but did not interfere with reverse-transcribing particles. These observations shed new light on the cytoplasmic fate of nucleic acids contained in retroviral particles.

MATERIALS AND METHODS

Retroviral vectors and plasmids.

Gammaretroviral vectors termed SF91 were derived from SF91.GFP (11, 31). The vector SF91aPBS.GFP.pre was generated by introducing the artificial primer binding site (aPBS) as an XbaI/ApaI fragment from SF91aPBS.GFP (8, 15) in SF91.GFP.pre (31). The retroviral vector used for engineering human Cre reporter cells to express Fv1ⁿ was derived from pCFCR (21). The red fluorescent protein of this construct was first excised by AgeI and NotI, and the Zeocin resistance gene was inserted as a blunt NcoI/SalI fragment from pT/Zeo (kindly provided by Z. Ivics, Max-Delbrück-Center, Berlin, Germany).

The basic lentiviral construct pRRL.PPT.SF.GFPpre has been previously described (30) and is a derivative of pRRL.PPT.PGK.GFPpre (kindly provided by L. Naldini, Milano, Italy). For the construction of a lentiviral shRNA construct, an shRNA cassette consisting of an H1 (polymerase III) promoter and an shRNA coding sequence directed against green fluorescent protein (GFP) were introduced into the 3′ dU3 region using a previously introduced unique SnaBI site. The shRNA sequence was created using primer 5′ GFP (5′-GATCCCCG CGGCAAGCTGACCCTGAAGTTCATTTCAAGAGAATGAACTTCAGGG TCAGCTTGCCGTTTTTGGAAA-3′) and 3′ GFP (5′-AGCTTTTCCAAAAA CGGCAAGCTGACCCTGAAGTTCATTCTCTTGAAATGAACTTCAGGG TCAGCTTGCCGCGGG-3′), self-annealed, and cloned as a BglII/HindIII fragment into pSuper (Oligoengine, Seattle, WA). From there, the H1 promoter plus shRNA was cloned as an SmaI/HincII fragment into the SnaBI site of the lentiviral vector (see above).

To create integration-defective lentiviral vectors, an integrase-deficient gag-pol construct (pcDNA3.gpD64V.4xCTE) harboring a D64V point mutation in integrase was used (kindly provided by M. Milsom, Cincinnati Children's Research Foundation, Cincinnati, OH).

Gammaretroviral and lentiviral particle production.

Gammaretroviral and lentiviral vector supernatants were produced in human 293T packaging cells using the calcium phosphate precipitation method (calcium phosphate transfection kit; Sigma Aldrich, Munich, Germany), assisted by 25 μM chloroquine (Sigma Aldrich). The day before transfection, 5 × 10⁶ 293T cells were seeded in a 10-cm dish. For gammaretrovirus production, the retroviral vector expression plasmid (5 μg) was cotransfected with expression plasmids for Moloney-MLV gag-pol (M57DAW, 15 μg) and either ecotropic (K73, 3 μg; kindly provided by T. Kitamura, Tokio, Japan) or glycoprotein from vesicular stomatitis virus (VSVg) (pMD.G, 2 μg) envelope. For the production of N- or B-tropic gammaretroviral particles, we used 5 μg of either pCIG3N or pCIG3B gag-pol expression plasmid (4). To ensure equal transfection efficiencies of gammaretroviral nucleus-localizing Cre (nlsCre) vectors, 1 μg of the pEGFP-C1 expression plasmid (BD Clontech, Heidelberg, Germany) was cotransfected. For lentivirus particle production, 5 μg of the lentiviral vector expression plasmid was cotransfected with 12 μg lentiviral gag-pol (pcDNA3.gp.4xCTE), 5 μg Rev (RSV-Rev; kindly provided by T. Hope, Northwestern University, Chicago, IL), and 2 μg VSVg envelope expression plasmid. Supernatants were harvested 36 h, 48 h, and 60 h posttransfection, filtered through a 0.22-μm filter (Millipore, Schwalbach, Germany), and stored at −80°C until use. For comparison of RMT particles with episomal lentiviral particles (Fig. 1), supernatants were concentrated via ultracentrifugation as previously described (30).

FIG. 1.

Comparison of RMT using integrase-deficient lentiviral particles (eLV) and iRV and iLV particles. (A) Gammaretroviral vectors SF91.GFP.pre and SF91aPBS.GFP.pre with long terminal repeats (U3, R, and U5), a functional PBS or an aPBS, splice donor (SD), packaging signal (ψ), splice acceptor (SA), EGFP, and the wPRE. The plasmid's 5′ enhancer-promoter is from the myeloproliferative sarcoma virus (MPSV), and the 3′ U3 region from SFFV. The lentiviral self-inactivating vector RRL.PPT.SF.GFP.pre contains the Rev-responsible element (RRE), the central polypurine tract (PPT), and SFFV as the internal promoter. (B) Vectors were packaged into integration-competent (SF91.GFP.pre and RRL.PPT.SF.GFP.pre) or RT-deficient (SF91aPBS.GFP), as well as integrase-deficient (RRL.PPT.SF.GFP.pre+D64V), VSVg-pseudotyped particles. The mean fluorescence intensities (MFI) of transduced SC-1 fibroblasts were monitored by FACS, starting 7 h and terminating 10.5 days posttransduction. Mock-transduced cells were negative controls. (C) Histogram plot overlays of the results for the transduced cells described for panel B (dark gray lines). Transduction efficiencies are displayed in comparison to those of mock-transduced cells (light gray) at different time points (7 h, 22 h, 4 days, and 9 days). Arrows point to cell populations transduced by integration-competent particles (iRV and iLV). The asterisk indicates putative residual integrations caused by eLV vectors.

Cell culture and transduction.

293T, SC-1, HT1080, and previously described human (HT1080 derived) and mouse (SC-1 derived) Cre reporter cells (8, 38) were grown in Dulbecco's modified Eagle's medium (Biochrom, Berlin, Germany) supplemented with 10% fetal calf serum. Human Cre reporter cells ectopically expressing Fv1ⁿ were cultured in the presence of 150 μg/ml Zeocin (Invitrogen, Karlsruhe, Germany). The day before transduction, 5 × 10⁴ cells were seeded. Serial dilutions of retroviral or lentiviral supernatants were applied to the cells either in the presence or in the absence of 0.5 μM MG132 (Calbiochem, Bad Soden, Germany). The transduction procedure was assisted by protamine sulfate (4 μg/ml; Sigma Aldrich) and centrifugation for 60 min at 400 x g and 32°C. After 14 h of incubation, the virus-containing medium was replaced with fresh medium. The percentage of enhanced-GFP (EGFP)-positive or recombined cells was determined by flow cytometry (fluorescence-activated cell sorting [FACS]) analysis at the indicated time points.

Western blotting.

Human or mouse Cre reporter cells were infected with B- or N-tropic nlsCre-encoding RMT particles (multiplicity of infection [MOI] of 1) either in the presence or in the absence of 0.5 μM MG132 (Calbiochem). Ninety minutes posttransduction, the supernatant-containing medium was removed, the cells were washed three times with phosphate-buffered saline, and fresh medium either with or without MG132 was added. At the indicated time points, cells were harvested and cell lysates prepared using proteinase inhibitors (Complete Mini; Roche, Mannheim, Germany) containing radioimmunoprecipitation assay buffer. Samples were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (12.5%), transferred to nitrocellulose membranes (Bio-Rad, Munich, Germany), and probed with goat anti-RLV p30 serum (final concentration, 2 μg/ml, kindly provided by S. K. Ruscetti, National Cancer Institute at Frederick [NCI-Frederick], Frederick, MD) in Tris-buffered saline with 0.05% Tween and 3% milk powder (TBST-3% dry milk). A donkey anti-goat horseradish peroxidase conjugate (Santa Cruz, Heidelberg, Germany) diluted 1:2,000 in TBST-3% dry milk served as the secondary antibody. Detection was carried out by chemiluminescence (ECL kit; Pierce, Bonn, Germany). For the detection of Erk protein, membranes were incubated with polyclonal rabbit anti-Erk 2 (1:2,000 in TBST-3% dry milk; Santa Cruz), followed by incubation with goat anti-rabbit horseradish peroxidase (Santa Cruz) diluted 1:2,000 in TBST-3% dry milk.

Real-time RT-PCR quantification.

On the day of transduction, 3 × 10⁶ murine or human Cre reporter cells or human Cre reporter cells ectopically expressing Fv1ⁿ were infected with B- or N-tropic nlsCre-encoding retroviral particles (supernatants were adjusted to the Cre activity determined on permissive cells) either in the presence or in the absence of 0.5 μM MG132 (Calbiochem). At 2, 4, 6, and 8 h posttransduction, cells were washed three times with phosphate-buffered saline and harvested, and their total RNA was prepared using the RNAzol extraction method (WAK Chemicals, Steinbach, Germany). Before RT-PCR, RNA samples were treated two times with RNase-free TURBO DNase (Ambion, Dresden, Germany) and purified (Qiagen RNeasy Mini kit) according to the manufacturer's protocol (Qiagen, Hilden, Germany). First-strand cDNA synthesis was performed with a QuantiTect RT kit (Qiagen) using oligo(dT) and random hexamer primers (MBI Fermentas, St.-Leon-Rot, Germany) in the same molecular ratio. Quantitative PCR was performed with an Applied Biosystems 7300 real-time PCR system (Foster City, CA) using a QuantiTect Sybr green kit (Qiagen). The amplification of the Cre DNA sequence was carried out by using oligonucleotides 5′-AACATTTGGGCCAGCTAAACA-3′ and 5′-AGAGCCTGTTTTGCACGTTCA-3′. The Cre-specific signal was normalized to the signal obtained by the amplification of mouse or human β-actin DNA with oligonucleotides 5′-CCTCCCTGGAGAAGAGCTA-3′ and 5′-TCCATGCCCAGGAAGGAAG-3′. The results were quantified by using the comparative threshold cycle method.

Characterization of RMT vector and wild-type retroviral supernatants.

Retroviral SF91.nlsCre and SF91aPBS.nlsCre supernatants were produced, harvested, and concentrated via ultracentrifugation (32). The obtained retrovirus pellets were resuspended in phosphate-buffered saline, aliquotted, and stored at −80°C. For determination of the RNA content, concentrated supernatants were pretreated with RNase-free Turbo DNase (Ambion). Retroviral RNA was extracted with an RNeasy Micro kit (Qiagen) according to the manufacturer's protocol, including an additional DNase treatment. First-strand cDNA synthesis and real-time PCR quantification were performed as described above. An in vitro-transcribed retroviral RNA derived from SF91.nlsCre served as the standard for the quantification of RNA. All samples were checked for plasmid DNA contamination. Western blot analysis for retroviral CA (p30) was performed as described above using denatured supernatants. The levels of reverse transcriptase activity of the retroviral supernatants were determined by using a RetroSys C-type RT activity kit (Innovagen, Lund, Sweden) according to the manufacturer's instructions.

Statistical analysis.

Data from the experiments are expressed as the means ± standard deviations. Student's paired t test was used for the comparison of differences between indicated groups. A P value of <0.05 was considered significant.

RESULTS

Kinetics of RMT vectors in comparison with those of nonintegrating lentiviral episomes.

As a first step to elucidating the mechanisms underlying RMT, we compared the kinetics of gene expression after RMT, the delivery of episomal lentiviral (eLV) DNA, or transduction with integrating gammaretroviral (iRV) or integrating lentiviral (iLV) vectors encoding EGFP. All vectors contained the woodchuck hepatitis virus posttranscriptional regulatory element (wPRE) for the optimization of titers and RNA processing (42). The RMT vector SF91aPBS.GFP.pre encodes EGFP downstream of the splice acceptor sequence (Fig. 1A). RT was blocked by the presence of an aPBS that does not correspond to any cellular tRNA (18). RT is therefore possible only if the corresponding tRNA is cotransfected into the packaging cells (8, 18). In the present study, we used a vector with a PBS for the tRNA^Gln as the integration-competent control (SF91.GFP.pre). We also produced a third-generation self-inactivating lentiviral vector (RRL.PPT.SF.GFP.pre) expressing EGFP under the control of the strong enhancer-promoter derived from the long terminal repeat of the MLV spleen focus-forming virus (SFFV). Additionally, integration-defective lentivirus particles, competent to form episomal DNA by using an integrase-deficient variant of the lentiviral gag-pol plasmid (integrase D64V), were produced (20, 26).

In this set of experiments, all vector particles were pseudotyped with VSVg. Using a high MOI, we transduced SC-1 fibroblasts with these four different vector preparations: SF91aPBS.GFP.pre for RMT, RRL.PPT.SF.GFP.pre+D64V for delivery of episomal lentiviral DNA, RRL.PPT.SF.GFP.pre packaged with intact lentiviral gag-pol for delivery of integrating lentiviral vectors, and SF91.GFP.pre for delivery of integrating gammaretroviral vectors. All integrating vectors were used at an MOI of 10. EGFP expression was monitored by flow cytometry at regular intervals, starting 7 h after transduction and ending after 10.5 days. Mock-transduced cells served as negative controls.

This side-by-side comparison of the kinetics of the expression of EGFP revealed that RMT particles, which are RT-deficient retrovirus mutants containing EGFP vector RNA (18), express EGFP for a relatively short duration and to a low level (Fig. 1B). The peak of EGFP expression was 1 order of magnitude above the background fluorescence level and occurred 24 h after transduction. The continuous decay to background levels until day 6.5 is consistent with the half-life of EGFP (6). There was no evidence for residual integration events following the use of aPBS vectors, consistent with earlier reports (8, 18). After transduction with the episomal or integration-competent vectors, the peak of expression occurred later (day 2) and reached much higher levels: 3 orders of magnitude above background with the integration-defective vector and saturating levels with the integrating vectors. While expression remained stable over the observation period with both integrating vectors, EGFP expressed from the integration-defective lentiviral vectors decayed within 8 days but did not return to background levels. The continued expression in more than 1% of the target cell population was suggestive of residual integration events, as previously described (Fig. 1C) (20). The residual integration of the D64V mutant may be circumvented by using a double or triple mutant at the DDE catalytic site.

These data show that two important features distinguish RMT from other forms of retroviral delivery of genetic information (episomal or integrated DNA): relatively low levels of expression and complete reversion to background levels.

Characterization of RMT particles and wild-type viral particles.

The aPBS within RMT vectors was designed not to match any naturally occurring tRNA molecule (18); therefore, the retroviral genomic RNA is packaged without the primer for the initiation of RT. To address the possibility that these modifications could affect the biochemistry of this type of viral particle, we compared viral RNA content, reverse transcriptase activity, capsid (p30) load, and the biological titers of RMT versus wild-type (iRV) retroviral supernatants (Fig. 2; Table 1). For this analysis, we packaged SF91.nlsCre (iRV) and SF91aPBS.nlsCre (RMT) vectors, which are similar to the EGFP vectors described above (Fig. 1A) but harbor the nlsCre cDNA instead and do not contain the wPRE. For each vector type (iRV and RMT vector), we analyzed four different retroviral preparations (B- and N-tropic, VSVg, and ecotropic pseudotypes). After harvest, the supernatants were concentrated via ultracentrifugation and the obtained pellets resuspended in phosphate-buffered saline for further analysis. The retroviral genomic RNA contents of the supernatants were determined via real-time PCR (Table 1). In all samples, DNA contamination was excluded. Table 1 shows a clear correlation between the biological titer of a retroviral supernatant and its retroviral genomic RNA content (R² = 0.99 for RMT supernatant and R² = 0.89 for iRV supernatant). In the case of RMT vector, the strict correlation of retroviral genomic RNA content and titer suggests that the biological activity is entirely mediated by packaged RNA, whereas in the case of iRV, subsequent steps of RT, integration, and de novo transcription contribute to the biological activity.

FIG. 2.

Capsid Western blot of viral supernatants of iRV and RMT vectors. Denatured ecotropic (Eco) and VSVg-pseudotyped particle supernatants were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and blotted, and retroviral capsid protein was detected with anticapsid antibody and chemiluminescence. Sizes for p65 (Gag precursor) and p30 (processed CA) are given on the right. Ponceau stain served as a loading control.

TABLE 1.

Characterization of RMT and iRV particles^a

Vector	Type of:			RNA content (fmol/ml)	Reverse transcriptase activity (U/ml)	Biological titer (107 t.U/ml)
	Vector	gag-pol tropism	Envelope
SF91.nlsCre	iRV	B	VSVg	1.2	9.2	0.8
SF91.nlsCre	iRV	N	VSVg	1.0	7.8	0.9
SF91aPBS.nlsCre	RMT	B	VSVg	6.1	8.6	4.5
SF91aPBS.nlsCre	RMT	N	VSVg	8.3	9.8	5.4
SF91.nlsCre	iRV	B	Ecotropic	193.0	93.6	92.4
SF91.nlsCre	iRV	N	Ecotropic	291.0	94.6	85.5
SF91aPBS.nlsCre	RMT	B	Ecotropic	26.6	95.6	7.6
SF91aPBS.nlsCre	RMT	N	Ecotropic	199.0	90.2	48.3

^aThe results for eight different virus preparations are shown. t.U, transducing units.

Importantly, comparison of the same pseudotypes (ecotropic and VSVg) revealed that the reverse transcriptase activities (Table 1; Fig. 2) were similar for iRV and RMT particles but did not correlate as nicely as the retroviral genomic RNA content with the biological titer. Furthermore, Western blot analysis of the retroviral supernatants used for the experiments whose results are shown in Table 1 revealed comparable capsid (p30) and Gag precursor protein (p65) levels for RMT and iRV particles (Fig. 2). We conclude that RMT particles have a composition similar to that of iRV particles, a conclusion which is supported by the results of previous studies demonstrating that MLV particle assembly occurs independently of the packaged retroviral RNA (16, 19).

RMT is restricted by inhibitory RNA expressed in target cells.

The above results and our previously published experimental results provide indirect evidence for the mechanisms underlying RMT (8). Further supporting the hypothesis that genomic RNA containing the packaging signal mediates the biological activity of RMT, we found that RMT activity depends on the amount of unspliced RNA expressed in packaging cells, whereas high expression of spliced, subgenomic RNA did not contribute to RMT activity (data not shown). To directly address whether mRNA delivered by retroviral particles is the cause of RMT, we tested whether it is sensitive to inhibition by shRNA as a form of an engineered restriction. We mixed unmodified control cells (shRNA negative) with cells coexpressing shRNA (either shGFP, directed against EGFP sequences on the vector mRNA, or the scrambled shControl) and DsRed fluorescent protein as a marker from the same lentiviral construct. DsRed expression therefore indicates cells expressing the shRNA. We transduced the two mixed cell populations (shRNA negative plus shGFP and shRNA negative plus shControl) using RT-proficient (iRV) and RT-deficient (RMT) GFP-encoding vectors and analyzed them 36 h posttransduction. shRNA directed against EGFP specifically and significantly (P < 0.0001; n = 9) inhibited the RMT-mediated expression of this protein, whereas the expression of the scrambled control shRNA had no effect (Fig. 3).

FIG. 3.

RMT is susceptible to ectopically expressed shRNAs. (A) SC-1 mouse fibroblasts were engineered to stably express either an shRNA directed against EGFP (shGFP) or a scrambled shRNA (shControl), marked by coexpression of the DsRed fluorescent protein. Before transduction, shRNA-expressing cells (either shGFP or shControl) were mixed with control cells and transduced with either SF91.eGFP (iRV; RT competent) or SF91aPBS.eGFP (RMT; RT deficient) ecotropic particles. After 36 h, transduction efficiencies and mean fluorescence intensities (MFI) were determined via FACS. Percentages of EGFP-positive cells are shown for the DsRed-positive (shRNA-expressing cells) and the DsRed-negative population (control cells not expressing shRNAs). (B) Same experimental setting as described for panel A, but the results of several experiments (n = 9; P < 0.0001) are displayed in the bar chart. To calculate the relative factor of downregulation by shRNAs, the MFI of cells not expressing shRNAs was divided by the MFI of cells expressing either shControl (dark gray) or shGFP (light gray). The black bar and the white bar indicate the background MFI levels of untransduced (mock) mixed cell populations: shControl and SC-1 cells or shGFP and SC-1 cells, respectively. Error bars show standard errors of the mean.

In contrast, the transduction efficiency of iRV particles was not significantly altered by shRNAs targeting the retroviral RNA genome, although we found a clear reduction of de novo-synthesized RNA in cells transduced with iRV (Fig. 3A, lower dot blot of iRV, compare upper-right quadrant showing cells expressing the shRNA and lower-right quadrant showing unmodified cells). The DsRed-negative, shGFP-negative cell population served as an internal positive control and expressed high levels of EGFP after transduction. The degrees of shRNA-mediated inhibition of EGFP expression were similar for RMT and integrating vectors (∼2.4-fold) (Fig. 3B). Together with the data shown in Table 1 and our previous findings (8), these data imply that retroviral particles can deliver unspliced retroviral RNA containing a packaging signal into target cells for immediate ribosomal translation.

RMT is sensitive to TRIM5α.

To consider the role of the retroviral particle in RMT, we analyzed the sensitivity to cellular restriction factors targeting the capsid protein. If RMT is mediated by specifically packaged mRNA contained in retroviral particles, then those formed by the N-tropic MLV capsid should be sensitive to restriction by huTRIM5α. To address this question, we packaged RT-proficient (iRV) and RT-deficient (RMT) forms of the Cre vector into B-tropic or N-tropic VSVg-pseudotyped MLV virions and transduced our previously described human (HT1080) and mouse (SC-1) Cre reporter cells (38). To avoid saturation of huTRIM5α restriction, retroviral supernatants were used at MOIs lower than 1 (MOI 0.05 to 1). In permissive murine Cre reporter cells, the potencies of the two virus supernatants to express Cre were comparable (Fig. 4A). However, when transducing human Cre reporter cells (which endogenously express huTRIM5α), the N-tropic vector particles were strongly restricted, independently of their ability to reverse transcribe. Both N-tropic vector particles were inhibited by huTRIM5α. SF91aPBS.nlsCre (RMT) was inhibited by around fourfold (Fig. 4D and E) and SF91.nlsCre (which can initiate RT and form integrating retroviral DNA) by around 10-fold compared to the inhibition of their B-tropic counterparts (Fig. 4C and E). Figure 4F shows the potencies (Cre activity in mouse Cre reporter cells) of all retroviral supernatants used for the experiments whose results are illustrated in Fig. 4C to E. Together, these experiments reveal that RMT is sensitive to restriction by TRIM5α and that the restriction occurs independently of RT.

FIG. 4.

RMT is susceptible to TRIM5α restriction but can be rescued by the inhibition of proteasomes. All experiments were performed with Cre reporter cells (38). (A) Comparable efficacies of all supernatants to recombine mouse Cre reporter cells. RT-competent (SF91.nlsCre) and RT-deficient (SF91aPBS.nlsCre) vectors encoding Cre were packaged into N- or B-tropic VSVg particles. (B) huTRIM5α restricts N-tropic particles. Supernatants tested in experiments whose results are shown in panel A were used for the infection of human Cre reporter cells. (C) Human Cre reporter cells were transduced with the indicated amounts of N- or B-tropic RT-competent supernatants (SF91.nlsCre) either in the presence or in the absence of MG132. The graph displays the recombination efficiencies from five independent experiments.All supernatants used in this experiment were checked for comparable Cre activity on mouse Cre reporter cells (Fig. 4F). (D) Experimental setup was as described for panel C, but RT-deficient particles were used (SF91aPBS.nlsCre). Absolute recombination efficiencies from five independent experiments are shown. (E) Relative display of the data sets shown in panels C and D. The proteasomal inhibitor MG132 efficiently antagonizes restriction of RMT by huTRIM5α. Human Cre reporter cells were transduced with N- or B-tropic supernatants (RT competent as well as RT deficient) either in the presence (light-gray bars) or in the absence of MG132 (dark-gray bars). The bar chart reflects the relative increase of recombined cells from MG132 treatment (n = 12; P < 0.001) in relation to nonrestricted B-tropic particles. Error bars show standard errors of the mean. (F) Recombination efficacies of the supernatants used for the experiments whose results are shown in panels C to E. The bar chart displays the Cre-transducing units per ml of the indicated supernatants for five independent experiments, determined on permissive mouse Cre reporter cells.

We next addressed whether the restriction by TRIM5α can be overcome by treatment with the proteasome inhibitor MG132. Since MG132 increases the efficiency of lentiviral infection in a cell type-dependent manner (27), we examined the influence of MG132 on restricted particles in comparison to its influence on nonrestricted particles. To minimize unspecific toxicity, we used MG132 at a relatively low concentration (0.5 μmol/liter), which is at least four times lower than the concentration used in related studies on retroviral restriction (5, 29). Both the RT-proficient (iRV) (Fig. 4C and E) and RT-deficient (RMT) (Fig. 4D and E) N-tropic vector particles were partially rescued by MG132. Rescue occurred at all doses of virus tested, revealing that the doses used did not result in a saturation of either restriction or proteasomal degradation. Importantly, treatment with MG132 significantly increased the efficiency of Cre delivery by N-tropic particles, with similar values for RT-proficient and RT-deficient particles (4.8-fold and 4.5-fold, respectively) (Fig. 4E). This increase of infectivity mediated by MG132 was greater in the context of restricted particles (P < 0.001; n = 12) (Fig. 4E). MG132 only led to a slight increase in the infectivity of unrestricted RT-deficient particles (1.5-fold), whereas it even reduced the infectivity of RT-proficient particles, possibly due to a residual cytotoxic effect. Nevertheless, proteasome inhibition did not allow a complete rescue of RMT following restriction by TRIM5α.

The restriction factor Fv1 also inhibits RMT.

Previous work has demonstrated that restriction by the mouse gag-like restriction factor Fv1 occurs after RT (13). However, Fv1 can compete with TRIM5α for restricted virus, suggesting that it interacts with the virion at the same time as TRIM5α, before significant RT has occurred (21). To address whether restriction by Fv1 depends upon the initiation of RT, we packaged RT-proficient (iRV) and RT-deficient (RMT) Cre vectors with the B-tropic gag-pol and transduced human Cre reporter cells that were engineered to express Fv1ⁿ. Interestingly, Fv1ⁿ clearly reduced RMT (mediated by SF91aPBS.nlsCre), although this restriction was less profound than that observed with the RT-proficient vector (SF91.nlsCre) (Fig. 5). We thus found that Fv1ⁿ partially inhibits RMT, a process which, as shown above, requires all the retroviral proteins except reverse transcriptase and integrase. This reveals that restriction by Fv1ⁿ occurs irrespective of the initiation of RT, the formation of retroviral DNA, and the subsequent maturation of the PIC.

FIG. 5.

RT-competent and RMT particles are sensitive to Fv1 restriction. (A) Nonrestricting human Cre reporter cells or human Cre reporter cells ectopically expressing Fv1ⁿ were transduced with increasing amounts of RT-competent (SF91.nlsCre), B-tropic particles. The mean values and standard errors of the mean of the results of six independent experiments are shown. (B) Same experimental setup as described for panel A, but RT-deficient (SF91aPBS.nlsCre) instead of RT-competent particles were used.

Restriction by TRIM5α, but not Fv1, is associated with reduced retroviral genomic RNA levels.

Since RMT is mediated by packaged retroviral genomic RNA (Table 1; Fig. 3) and is restricted by TRIM5α (Fig. 4) and, to a lesser extent, by Fv1ⁿ (Fig. 5), we wanted to know whether the restriction is associated with destruction of the retroviral genomic mRNA. To address this point, we transduced human Cre reporter cells endogenously expressing huTRIM5α with restricted (N tropic) or nonrestricted (B tropic) RMT particles in the presence or absence of MG132. At 2, 4, 6, and 8 h postinfection, we harvested total RNA and performed quantitative real-time RT-PCR using primers targeting the retroviral genomic RNA. Real-time PCR analysis revealed that the restriction by TRIM5α was associated with the degradation of the retroviral genomic mRNA (Fig. 6A). Strikingly, at 8 h posttransduction, the RNA levels of the restricted N-tropic particles were 10.8 times lower than for the nonrestricted B-tropic particles. In contrast, we found no significant difference for N- and B-tropic particles in nonrestrictive mouse Cre reporter cells (data not shown). Interestingly, proteasome inhibition with MG132 allowed partial recovery of the retroviral genomic RNA (4.5-fold), suggesting that it is lost through recruitment to the proteasome by TRIM5α (Fig. 6A). The reduction of mRNA was predominantly detectable at the later time point (>2 h), suggesting that degradation does not immediately follow particle uptake. Furthermore, we monitored, in the same cell populations, the fate of the retroviral capsid (p30) in restrictive and nonrestrictive cells (Fig. 6C) up to 8 h after transduction. Similar to the RNA data, we saw reduced capsid levels for the restricted N-tropic particles, which could be compensated by the addition of MG132.

FIG. 6.

Retroviral genomic RNA and capsid levels of restricted (N) and nonrestricted (B) particles from 2 to 8 h postinfection. (A) Human Cre reporter cells were transduced with N- or B-tropic RMT particles either in the presence or in the absence of the proteasomal inhibitor MG132. Genomic retroviral RNA levels were determined via real-time RT-PCR at the indicated time points (2, 4, 6, and 8 h postinfection) and normalized to the (nonrestricted) B-tropic value (2 h, no MG132, 100%). RNA preparations used in this experiment were treated three times with DNase. (B) Nonrestrictive human Cre reporter cells or human Cre reporter cells ectopically expressing Fv1ⁿ were transduced with B-tropic RMT particles. The genomic retroviral RNA levels at the indicated time points are shown. Quantitative real-time RT-PCR was performed as described for panel A. (C) Capsid (p30) levels of retroviral particles in human Cre reporter cells endogenously expressing huTRIM5α (upper panel) and in nonrestrictive mouse Cre reporter cells (lower panel). Western blot of protein samples harvested 2 to 8 h posttransduction is shown. Presence (+) or absence (−) of MG132 is indicated. Normalized to cellular Erk protein, densitometry revealed 50% reduction of N-tropic p30 at 4 h in the absence of MG132 and restoration of N-tropic p30 to the level of B-tropic p30 by the addition of MG132.

Restriction by Fv1ⁿ did not alter the levels of retroviral genomic RNA (Fig. 6B). This is consistent with the observation that Fv1-restricted particles can still undergo RT but are blocked at a later step (13). Together, the results of these experiments reveal that RMT is mediated by retroviral particles and is thus dependent on the amount of mRNA made accessible to ribosomes in the target cells. Furthermore, these observations support the notion that Fv1 and TRIM5α interact with the particle independently of the initiation of RT and that restriction by TRIM5α possibly leads to proteasomal degradation of the particle.

DISCUSSION

Our experiments have established that RT-deficient retroviral particles are able to make their genome accessible for translation in the unspliced gag-pol reading frame. Retroviruses are thus capable of RMT, resulting in low-level, transient expression of virally encoded gene products in transduced cells. RMT may occur if retroviral particles have not packaged a tRNA primer, if reverse transcriptase is mutated, or, as used in our experimental approach, if retroviral vectors are generated that are unable to bind the tRNA primer. Previously, we have demonstrated that retroviral RMT depends upon the presence of the packaging motif in the transduced mRNA, gag, and env, but not reverse transcriptase. The passive transfer of protein and the contamination of retrovirus-conditioned medium with plasmid DNA have been excluded as underlying this phenomenon (8). Here, we have demonstrated that the efficiency of RMT correlates with the expression of packaged retroviral mRNA rather than the amount of protein encoded by the gag-pol reading frame in viral producer cells (Table 1). Finally, the sensitivity to shRNA expressed in target cells (Fig. 3) clearly shows that the viral mRNA is responsible for RMT.

We went on to show that RMT is restricted by the cytoplasmic restriction factors TRIM5α and Fv1, both of which are directed against the retroviral capsid. The side-by-side comparison of RT-deficient virus (i.e., RMT) with RT-competent integrating virus (iRV) reveals that TRIM5α restricts RMT particles to a lesser extent than iRV. This implies that RMT-competent virions are partially able to escape restriction and release their nucleic acids for translation. In other words, we hypothesize that the somewhat weaker restriction of RMT particles than of iRV particles by restriction factors targeting the capsid reflects the fact that iRV particles still have to complete a number of complicated steps in their life cycle (RT, formation of a PIC, and integration), whereas RMT particles only have to release their mRNA for subsequent translation. Importantly, data obtained in functional assays of biological activity mediated by RMT correlated well with RNA levels determined by real-time PCR.

Furthermore, restriction of RMT particles could be rescued more efficiently than restriction of RT-proficient particles by inhibition of the proteasome with MG132, suggesting that the restriction of RMT is more dependent on the proteasome. The RNA data correlated with the capsid levels determined by Western blot analysis (Fig. 6).

We also found that RMT is sensitive to restriction by Fv1 when delivered by the appropriately Fv1-sensitive capsid. Again, RMT vectors were less sensitive to this form of restriction than reverse-transcribing vectors. Strikingly, we found that restriction by TRIM5α, but not by Fv1, leads to clear destruction of the viral RNA (Fig. 6). These data are consistent with recent observations that inhibition of the proteasome during restriction by TRIM5α rescues RT and support the notion that the RT block is due to the destruction of the particle and the RNA by the proteasome (1, 39). While the proteasome might not degrade the RNA directly, we imagine that degradation of the virion protein would render the genome sensitive to degradation by cellular nucleases. These observations are inconsistent with an uncoating mechanism for TRIM5α, which might be expected to increase the release of the genome and RMT (5, 22, 23). Finally, our use of retroviral vectors which cannot reverse transcribe due to modification of the PBS demonstrates that sensitivity to TRIM5α and proteasomal degradation of the mRNA do not depend on the initiation of RT.

In summary, RMT suggests a potential evolutionary role of immediate early translation of retroviral nucleic acids. As shown here, this by-product of the retroviral life cycle can be exploited to study cytoplasmic restriction of retroviral particles, using both biological activity and biochemical parameters as readouts. We thus found that the sensitivity to TRIM5α does not depend on the initiation of RT and that the degradation of RNA and capsid is correlated with restriction mediated by TRIM5α. Our data also support a hypothesis that, in nonrestrictive cells, retroviral nucleic acids become accessible to host factors, including ribosomes, as a result of particle remodeling during cytoplasmic trafficking. Particle modifications that trigger mRNA release after entry are thus expected to further increase the efficiency of RMT.