Retroviral vectors elevate coexpressed protein levelsin trans through cap-dependent translationYongqiang Gou1, Hyewon Byun1, Adam E. Zook, Gurvani B. Singh, Andrea K. Nash, Mary M. Lozano,and Jaquelin P. Dudley2

Department of Molecular Biosciences, Center for Infectious Disease, and Institute for Cellular and Molecular Biology, The University of Texas at Austin,Austin, TX 78712-1191

Edited by Peter K. Vogt, The Scripps Research Institute, La Jolla, CA, and approved February 3, 2015 (received for review October 24, 2014)

Retroviruses cause immunodeficiency and cancer but also are usedas vectors for the expression of heterologous genes. Nevertheless,optimal translation of introduced genes often is not achieved.Here we show that transfection into mammalian cells of lentiviralor gammaretroviral vectors, including those with specific shRNAs,increased expression of a cotransfected gene relative to standardplasmid vectors. Levels of most endogenous cellular proteins wereunchanged. Transfer of lentiviral vector sequences into a standardplasmid conferred the ability to give increased expression of co-transfected genes (superinduction). Superinduction by the retro-viral vector was not dependent on the cell type or species, the typeof reporter gene, or the method of transfection. No differenceswere detected in the IFN, unfolded protein, or stress responses inthe presence of retroviral vectors. RT-PCRs revealed that RNAlevels of cotransfected genes were unchanged during superinduc-tion, yet Western blotting, pulse labeling, and the use of bicistronicvectors showed increased cap-dependent translation of cointro-duced genes. Expression of the mammalian target of rapamycin(mTOR) kinase target 4E-BP1, but not the mTOR inhibitor Torin 1,preferentially inhibited superinduction relative to basal proteinexpression. Furthermore, transcription of lentiviral vector sequencesfrom a doxycycline-inducible promoter eliminated superinduction,consistent with a DNA-triggered event. Thus, retroviral DNA in-creased translation of cointroduced genes in trans by an mTOR-independent signaling mechanism. Our experiments have broadapplications for the design of retroviral vectors for transfections,DNA vaccines, and gene therapy.

superinduction | retroviruses | translation | transfection

Translational control is critical for mammalian cells and theviruses that infect them. For example, picornaviruses and

some flaviviruses have an internal ribosomal entry site (IRES)(1) that allows cap-independent translation of viral RNAs andprovides preferential translation over most host cell mRNAs (2).Structural modifications at the 5′ ends of viral RNAs, such asdifferences in cap methylation or absence of a cap, often dis-tinguish viral mRNAs from their cellular counterparts (3).Multiple cellular proteins recognize foreign RNAs or DNAs,which trigger specific signaling pathways that lead to generaltranslational arrest and/or selective viral RNA degradation (4).Many viral RNAs trigger the IFN signaling pathway and trans-lational inhibition, but viruses encode various proteins or RNAsthat mute this response (5). Recognition of foreign nucleic acidsby cellular surface or cytosolic receptors also leads to signalingevents that provide an innate antiviral response (6).Retroviruses are positive-sense RNA viruses that have been

used extensively for introduction of genes or small hairpin RNAs(shRNAs) into cells, both in culture and for gene therapy. Unlikemost RNA viruses, retroviruses replicate through a DNA in-termediate and use RNA polymerase II to produce mRNAs withstructures that are very similar to those of host mRNAs, thusavoiding some cytosolic RNA sensors (7). Nevertheless, theunspliced and partially spliced RNAs are required for the syn-thesis of viral structural proteins. These RNAs require a highly

structured cis-acting element to facilitate export from the nucleusto the cytoplasm. These cis-acting sequences include a constitutivetransport element (CTE) or a Rev-like response element (RRE)that requires binding of a protein adapter (8). Furthermore, ret-roviruses specify several other highly structured RNA elements,including the packaging sequence Psi, plus-strand priming sites,and splice acceptor sites (9).We have made the unique observation that transfection of

retrovirus-based vectors leads to increased levels (superinduc-tion) of proteins encoded by cotransfected plasmids. Both len-tiviral and gammaretroviral vectors lacking viral protein-codingpotential, but not plasmid-based vectors, elevated translation ofproteins expressed from cotransfected plasmids, such as GFP,but not most endogenous proteins. Increased translation of ex-ogenous proteins was cap-dependent and did not lead to addi-tional mammalian target of rapamycin complex 1 (mTORC1)signaling. Retroviral sequences did not require transcriptionto facilitate cap-dependent translation of cotransfected genes.These results indicate that retrovirus-based vectors can be usedfor improved gene expression during transfection and DNA vacci-nation in multiple cell types without additional cloning.

ResultsLentiviral Vector Cotransfection Causes Superinduction in Trans. Weused calcium-phosphate transfection of an shRNA-expressinglentiviral vector (pLKO.1, Sigma) (Fig. 1A) to measure the effect

Significance

Translation is a key process that is regulated by cellular healthand responses to the environment, including virus infection.We show here that introduction of lentivirus and gammaret-roviral vectors into cells by transfection increased translation(superinduction) of cotransfected genes but not most endog-enous proteins. Superinduction was independent of the un-folded protein, stress, and interferon responses and did notrequire retroviral vector transcription. Retroviral vectors ele-vated cap-dependent translation initiation without increasedmammalian target of rapamycin (mTOR) kinase activity. Thus,DNA sequences from HIV-1 and other retroviruses increasetranslation of cotransfected genes in trans by mTOR complex1-independent signaling. Our results suggest that retroviralDNA manipulates translation, which has practical implicationsfor protein expression and design of vectors for transfectionassays, DNA vaccines, and shRNA knockdown experiments.

Author contributions: Y.G., H.B., M.M.L., and J.P.D. designed research; Y.G., H.B., A.E.Z.,G.B.S., A.K.N., M.M.L., and J.P.D. performed research; Y.G., H.B., A.E.Z., G.B.S., A.K.N.,M.M.L., and J.P.D. analyzed data; and Y.G., H.B., M.M.L., and J.P.D. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Y.G. and H.B. contributed equally to this work.2To whom correspondence should be addressed. Email: jdudley@austin.utexas.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1420477112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1420477112 PNAS | March 17, 2015 | vol. 112 | no. 11 | 3505–3510

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

24,

202

0

of targeted knockdowns of specific cellular genes on the ac-tivity of mouse mammary tumor virus (MMTV) Rem, a Rev-likeRNA-binding protein (10, 11). Rem activity was measured bycotransfection of a reporter plasmid (pHMRluc) expressing Renillaluciferase in 293T cells. Increased luciferase levels require bindingof Rem signal peptide (Rem-SP) to the Rem-responsive element(RmRE) in the reporter transcript (12). As anticipated, Remexpression elevated Renilla expression by ∼12-fold (Fig. 1B, Top).Unexpectedly, cotransfection of pLKO.1 expressing a controlshRNA (pLKO.1c) elevated both basal and Rem-induced lu-ciferase expression in 293T cells (an additional 6.6-fold and 8.9-fold, respectively). Further, Western blotting indicated that Remlevels were increased in the presence of the lentiviral vector (Fig.1C). Effects of the lentivirus vector were not limited to Rem-responsive plasmids. A firefly luciferase reporter vector that lackedRem responsiveness, pGL3-C, showed ∼threefold higher levelsin the presence of the lentiviral vector (Fig. 1B, Bottom). Theseresults indicated that the lentivirus increased expression of twodifferent reporter genes (firefly or Renilla luciferase genes) fromeither the SV40 (pGL3-C) or CMV (pHMRluc) promoters.We then determined whether different shRNA inserts down-

stream of the U6 promoter affected superinduction. Cotrans-fection of lentivirus vectors expressing a control shRNA (pLKO.1c)or an shRNA designed to knock down the AAA ATPase, p97(LK-4250 or LK-4252), each elevated reporter expression (Fig.1B). The highest induction occurred using LK-4252, which alsogave the highest Rem levels by Western blotting (Fig. 1C), butno effect on p97 levels was observed. Because the same totalamount of DNA (6 μg) was used, different DNA levels were notresponsible for increased reporter activity. Thus, vectors expressingvarious shRNAs, including those with no known target, couldenhance superinduction. Similar results were obtained in trans-fections of XC rat fibroblast cells using a lipid-based methodor Jurkat human T cells using electroporation (Figs. S1 and S2),indicating that lentivirus superinduction was not cell-type-,species-, or method-specific. The effect was independent of thepresence of the RmRE and Rem (Fig. S3).To test the effect of pLKO.1c on other transgenes, we co-

transfected this lentivirus vector with either EGFP or ER-mCherryexpression vectors (13) and monitored it by fluorescence mi-croscopy (Fig. 1D). The data showed that both fluorescent pro-teins were induced by the presence of pLKO.1c in trans, consis-tent with previous findings (Fig. 1 B and C). Similar transfectionswere quantitated by FACS analysis and revealed 8- to 10-fold in-duction for EGFP, which diffuses throughout cells, or ER-mCherry,which is localized to the endoplasmic reticulum (13) (Fig. 1 Eand F). Thus, lentivirus cotransfection increased expression of dif-ferent exogenous proteins that were localized to distinct cel-lular compartments.

Determinants of Lentivirus-Mediated Superinduction. To investigatethe lentivirus sequences needed for superinduction, we deletedthe shRNA hairpin (pLKΔsh) (Fig. 2A) and repeated cotrans-fection into 293 cells with both Rem-expression and Rem-responsive vectors. Removal of the shRNA decreased Rem ex-pression, but the deleted vector still caused Rem induction asdetermined by reporter assays and Western blots (Fig. 2 B and Cand Fig. S4A). We also deleted sequences through the U6 pro-moter (pLKΔU6) or the retroviral packaging region (pLKΔPsi) inpLKO.1c (Fig. 2A). Cotransfections of the lentivirus vector lackingthe shRNA, U6 promoter, RRE, and Psi (pLKΔPsi) retainedsuperinduction capability of 2.6-fold. Additional transfections ofreporter and expression vectors with a different lentivirus vectorlacking an shRNA (pLL3.7) revealed two- to threefold inductionof both Renilla and firefly reporter vectors and ∼5- to 10-fold Reminduction (Fig. S5). Thus, although the shRNA was a major de-terminant of increased expression, lentiviral sequences also causedsuperinduction of exogenous proteins from cotransfected vectors.

To test vector determinants of superinduction, we then inserteda portion of pLKO.1c containing the packaging signal Psi, theRev-responsive element (RRE), and the U6 promoter-shRNAinto the multiple cloning site of plasmid vector pcDNA3 (Invi-trogen). This construct, Psi-sh (Fig. 2A), was cotransfected into293 cells with both Rem-responsive and nonresponsive vec-tors in the presence and absence of Rem expression. Interest-ingly, cotransfection of Psi-sh increased Renilla or firefly lucif-erase reporter gene expression (pHMRluc or pGL3-C vectors,respectively) by two- to fivefold in the absence of Rem. Remexpression also was induced by Psi-sh as determined by reporterassays (Fig. 2D and Fig. S4B) or Western blotting of transfectedcell extracts (Fig. 2E), but was not as effective as pLKO.1c. Lackof dependence on Rem was confirmed by FACS analysis aftercotransfections of EGFP expression plasmids with Psi-sh (Fig.2F). Therefore, vectors expressing lentivirus sequences inducedincreased expression of cotransfected genes in trans.To determine if other retroviruses had an effect on cotrans-

fected sequences, we transfected the gammaretroviral vectorMigR1 (14) with Rem expression and reporter vectors. Althoughthe firefly reporter showed no induction by the lentivirus vectorin 293T cells, cotransfection with MigR1 gave a significant in-crease (∼2.7-fold) (Fig. S4C). Lentivirus or gammaretrovirustransfection resulted in similar increases in Renilla luciferaseactivity (Fig. 2G). Rem expression was induced by both pLKO.1cand MigR1 (Fig. 2H). Because the gammaretroviral vectorencodes the heterologous GFP protein from a highly structured

Fig. 1. Lentivirus vectors stimulate cotransfected gene expression. (A) Dia-gram of pLKO.1-based lentivirus vectors. (B) The pLKO.1c vector stimulatesactivity from both Renilla and firefly reporter vectors. Cells (293) werecotransfected with the empty vector pcDNA3 or 5 μg of pLKO.1c with acontrol shRNA or pLKO.1 with either one of two different shRNAs againstthe AAA ATPase p97 (pLK-4250 or 4252) in the presence or absence of GFP-tagged Rem (12.5 ng) as indicated. The numbers in parentheses give foldinduction by pLKO.1 vectors of the Renilla reporter vector in the presence ofRem. Significant differences were observed for Renilla luciferase comparinglane 1 to all other lanes and in comparing lane 5 to lanes 6–8. (C) Exogenousexpression increases in the presence of the pLKO.1c vector. Extracts from thetransfections in Fig. 1B were subjected to Western blotting. (D) Lentivirusvector cotransfection stimulates EGFP and mCherry expression. Fluorescencemicroscopy of cells transfected with an EGFP or mCherry expression vectorin the absence or presence of pLKO.1c. A representative field is shown.(E ) Quantitation of EGFP expression by FACS. The GFP fluorescence in-tensities and SDs of triplicate transfections were determined in the presenceor absence of 2.5 ng pEGFP vector and reported relative to the pcDNA3 con-trol. (F) Quantitation of mCherry expression by FACS. Values are reported asin Fig. 1E.

3506 | www.pnas.org/cgi/doi/10.1073/pnas.1420477112 Gou et al.

Dow

nloa

ded

by g

uest

on

June

24,

202

0

RNA region (IRES), we deleted these sequences. Cotransfectionof the deleted derivative MigΔGFP induced Renilla and fireflyreporter activity similar to MigR1 (Fig. 2G and Fig. S4C) as wellas either GFP or GFP-Rem-SP levels (Fig. 2 F and H). Thus,different types of retroviral vectors enhanced expression ofcotransfected genes.

Lentivirus Vector Does Not Induce IFN or Stress Responses. Double-stranded RNAs activate RNA-dependent protein kinase (PKR),which phosphorylates the translation initiation factor, eIF2α,resulting in reduced translation initiation (15). To determine ifpLKO.1c altered eIF2α phosphorylation, we repeated cotrans-fections with MMTV Rem expression vectors and extracts weretested by Western blotting (Fig. S6A). No difference in phos-phorylated eIF2α was observed in the presence and absence ofpLKO.1c. Further, BiP levels increase during the unfolded pro-tein response and/or viral infections (16). Western blotting oftransfected cell extracts revealed no differences in BiP, B23, orGAPDH (Fig. S6B). Cellular stress also was assessed by cotrans-fection of pLKO.1c with the NF-κB reporter plasmid, NF-κB-GL2,expressing firefly luciferase. No evidence of NF-κB inductionby lentivirus cotransfection was observed (Fig. S7). Stress gran-ules and levels of the stress granule protein G3BP (17) wereunchanged by the lentivirus vector (Figs. S6C and S8). Fur-thermore, levels of the IFN-induced protein RIG-G (18) werenot affected by introduction of pLKO.1c (Fig. S6D). There-fore, lentivirus cotransfection does not influence levels of ex-

ogenous proteins through IFN signaling, stress, or the unfoldedprotein response.

Lentivirus Vector Affects eIF4E-Dependent Translation.To determinewhether superinduction increased transcription, we performedcotransfection experiments with Rem expression and reportervectors in the presence and absence of pLKO.1c. SemiquantitativeRT-PCR results indicated that pLKO.1c did not affect the steadystate level of exogenous rem or Rluc or endogenous gapdh mRNAs(Fig. S9A) when exogenous protein levels increased (Fig. S9B).Thus, the lentivirus sequences did not superinduce transcription.To test whether lentivirus sequences increased translation of

exogenous genes, cells were cotransfected with EGFP expressionvector in the presence of the lentivirus or pcDNA3 only and usedfor pulse labeling with [35S]methionine. Radioactive incorpora-tion with and without the lentivirus vector was not statisticallydifferent (Fig. S9C). Additional aliquots were used for GFP af-finity purification and analysis on denaturing polyacrylamidegels. Few differences in radioactive background incorporationinto endogenous proteins were observed (Fig. S9D), but GFPlevels greatly increased in the presence of pLKO.1c (Fig. S9E).These experiments indicated increased translation of exogenousgenes cointroduced with the lentivirus vector. To assess whethervector-induced translational effects were cap-dependent, wetested different bicistronic vectors that expressed Renilla lucif-erase and firefly luciferase separated by an IRES from the sametranscript (Fig. 3A). Experiments using vectors with either thehepatitis C virus (HCV) or cricket paralysis virus (CrPV) IRES

Fig. 2. Retrovirus vectors lacking shRNAs stimulated expression of cotransfected genes. (A) Diagram of vectors containing retroviral sequences. (B) Retroviralvectors stimulate cotransfected reporter genes. Significant differences were observed for Renilla luciferase comparing lane 1 to all other lanes and incomparing lane 6 to lanes 7–10. (C) The pLKO.1-based vectors stimulate GFP-Rem expression. Extracts from the transfection in Fig. 2B were used for Westernblots, scanned, and quantitated relative to actin. Results are expressed relative to GFP-Rem-SP in the absence of pLKO.1-based vectors (relative value of 1.0).(D) The Psi-sh construct stimulates cotransfected reporter gene expression. Significant differences were observed for Renilla luciferase comparing lane 1 to allother lanes and in comparing lane 4 to lanes 5 and 6. (E) The Psi-sh vector increases cotransfected GFP-Rem expression. Extracts from the experiment shown inFig. 2D were used for Western blots. (F) MigR1 and Psi-sh vectors elevate cotransfected EGFP expression. Cells were transfected as in Fig. 1E and subjected toFACS analysis after 48 h. The fluorescence intensities and SDs of triplicate transfections are shown relative to transfections containing pcDNA3.(G) Cotransfection with gammaretroviral constructs enhances reporter gene expression. Triplicate transfections in 293T cells were performed and reported asfor Fig. 1B. Significant differences were observed for Renilla luciferase comparing lane 1 to all other lanes and in comparing lane 5 to lanes 6–8. (H) Gam-maretroviral vectors increase cotransfected GFP-Rem levels. Extracts from transfections shown in Fig. 2G were used for Western blots.

Gou et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3507

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

24,

202

0

(19) showed that lentiviral induction was cap-dependent (Fig. 3B).Transfection of plasmids carrying retroviral sequences confirmedcap dependence of both lentiviral (pLKO.1c and Psi-sh) andgammaretroviral (MigΔGFP) vectors, yet cap dependence did notcompletely correspond to induction of other reporter proteins,suggesting different signaling mechanisms (20) (Fig. 3C).To determine signaling pathways affected by lentivirus vectors,

we cotransfected 293 cells with an EGFP expression vector in thepresence and absence of pLKO.1c. No change in total or phos-phorylated p38 MAPK (Fig. 3D) or in total or phosphorylatedERK1/2 (p44/p42) was detected (Fig. 3E). Differences in phos-phorylation of AKT or p70 S6K also were not observed (Fig. 3F).Chemical inhibitors for MNK1 and p38 MAPK had no specificeffect on pLKO.1c induction of reporter genes (Fig. S10).These data suggested that signaling pathways through AKT,ERK1/2, MAPK, MNK1, and mTORC1 were not responsiblefor superinduction.Cap-dependent translation is controlled through eIF4E (21).

Unexpectedly, we observed ∼twofold increases in total, but notphosphorylated, eIF4E in multiple transfection experimentswith pLKO.1c (Fig. 4 A and B). Unphosphorylated 4E-BP1 bindseIF4E to prevent assembly of the cap-binding complex (22,23), and 4E-BP1 phosphorylation by mTORC1 increases cap-dependent translation. To test whether phosphorylation of 4E-BP1was affected by the lentiviral vector, we performed additionaltransfections in the presence and absence of pLKO.1c. Asexpected, GFP was superinduced by pLKO.1c, yet no differencein total or phospho-4E-BP1 levels was detected (Fig. 4C), con-sistent with no lentivirus-induced change in mTOR kinase ac-tivity (Fig. 3).

To determine whether vector-mediated superinduction iscontrolled by 4E-BP1, we cotransfected 293 cells with 4E-BP1and EGFP expression vectors in the presence or absence ofpLKO.1c. Superinduction was greatly reduced by excess 4E-BP1,but protein overexpression had little effect on GFP levels in theabsence of pLKO.1c (Fig. 4D). Similarly, excess 4E-BP1 partiallyinhibited cap-dependent translation using a bicistronic vector inthe presence of pLKO.1c (Fig. S11). The effect of exogenous4E-BP1 expression on superinduction was demonstrable by pulselabeling, consistent with a reduced level of cap-dependent GFPtranslation (Fig. S12). Further, we coexpressed EGFP and theF114A mutant of 4E-BP1 with or without pLKO.1c. This mutantdoes not bind to Raptor in the mTORC1 complex (24), leadingto increased levels of unphosphorylated 4E-BP1 and inhibitionof cap-dependent translation initiation (25). Although the F114Amutant greatly reduced both basal and lentivector-induced GFPexpression to a similar extent, normalization showed that expres-sion of wild-type 4E-BP1, which has both phosphorylated and

Fig. 3. Superinduction affects cap-dependent translation of exogenousproteins but not mTOR activity. (A) Structure of bicistronic vectors used tomeasure cap-dependent translation. (B) pLKO.1c increases cap-dependenttranslation. Cells (293) were cotransfected in triplicate with either pcDNA3or pLKO.1c in the presence of a bicistronic vector. The ratio of the two lu-ciferase values ± SD is shown. The differences in the presence and absence ofpLKO.1c were highly significant. (C) Lentiviral and gammaretroviral vectorcotransfection induces cap-dependent translation of reporter genes. Cells(293) were cotransfected with the CrPV bicistronic vector as in B. (D) Totaland phosphorylated p38 MAPK were unchanged after pLKO.1c transfection.Cells (293) were transfected as in Fig. 1B in all panels, and lysates weresubjected to Western blotting. (E) Total and phosphorylated p44/p42 ERKare unaltered after pLKO.1c transfection. Western blots were performed ontransfected cell lysates. (F) Phosphorylated AKT and p70 S6K are notchanged by pLKO.1c transfection. Western blots of transfected cell lysatesare shown.

Fig. 4. Lentivirus vectors selectively affect eIF4E-dependent translation incotransfection experiments. (A) Total levels of eIF4E increase after pLKO.1ctransfection. Western blots of transfected cell lysates are shown. (B) Quan-titation of eIF4E levels after pLKO.1c transfection. (C) Wild-type 4E-BP1 se-lectively inhibits pLKO.1c superinduction but not basal GFP levels. Cells weretransfected with the indicated plasmids, except that equivalent levels ofpcDNA3 were substituted for pLKO.1c. Transfected cell extracts were sub-jected to Western blotting. (D) FACS quantitation of wild-type and mutant4E-BP1 effects on pLKO.1c-mediated superinduction of GFP expression. Thefluorescence intensities and SDs of triplicate transfections in the absence ofpLKO.1c has been shown normalized to 1.0 for each condition (pcDNA3 onlyor in the presence of wild-type or mutant 4E-BP1). Expression of 4E-BP1significantly reduced pLKO.1c-mediated superinduction (compare lanes 4and 5; P < 0.005), whereas F114A expression had no effect (compare lanes4 and 6; P = 0.75). (E) The mTOR inhibitor Torin 1 similarly reduced bothbasal and pLKO.1c-induced GFP expression. Two exposures of the Westernblot with GFP-specific antibody are shown. (F) FACS quantitation of Torin 1effects on pLKO.1c-mediated superinduction of GFP. Although 4E-BP1 ex-pression significantly inhibited the effect of pLKO.1c (compare lanes 4 and 5;P < 0.001), lentivirus-induced GFP expression was not significantly differentwhen Torin 1 was added (compare lanes 4 and 6; P = 0.14).

3508 | www.pnas.org/cgi/doi/10.1073/pnas.1420477112 Gou et al.

Dow

nloa

ded

by g

uest

on

June

24,

202

0

unphosphorylated forms, had a selective effect on superinduction(Fig. 4D). Cotransfection of pLKO.1c with the EGFP expressionplasmid in the presence of the mTORC1 inhibitor, Torin 1, sim-ilarly decreased levels of both basal and lentivirus-induced GFPexpression (Fig. 4 E and F). Thus, although mTORC1 activityaffects superinduction, cap-dependent translation of GFP is moresensitive to 4E-BP1 inhibition in the presence of pLKO.1c than inits absence. Our experiments suggest that retroviral sequencesinduce an mTORC1-independent signaling event that increasescap-dependent translation of exogenous genes.

DNA Causes Superinduction. To assess whether the transfected DNAsor their resulting transcripts increase expression of cotransfectedgenes, we cloned the Psi to shRNA sequences from pLKO.1cinto the vector pTRE3G-BI-mCherry, which contains a bi-directional doxycycline (DOX)-inducible promoter (Fig. 5A).DOX addition induces a conformational change in the trans-activator, resulting in promoter binding and up-regulation of geneexpression. The derivative vector (pTRE-Psi-sh) was cotransfectedwith an EGFP expression vector into 293 cells in the presence andabsence of the transactivator (Fig. 5B). As expected in the pres-ence of DOX and the transactivator, mCherry was induced in boththe parental and derivative vectors. Furthermore, the inclusionof the Psi-sh sequences increased mCherry expression ∼3.5-foldin the absence of the transactivator compared with pTRE3G-BI-mCherry, indicating that superinduction occurs in cis when thepromoter is largely inactive.To determine whether expression of the lentiviral sequences

was required for superinduction in trans, we cotransfected eitherpTRE3G or pTRE-Psi-sh and the EGFP expression vector. GFPexpression was slightly increased by the presence of pTRE3G,whereas sixfold superinduction was observed with pTRE-Psi-shin the absence of the transactivator (Fig. 5C). Promoter activa-tion by DOX and the transactivator greatly suppressed EGFPexpression in the presence of either the parental or pTRE-Psi-sh vector, suggesting that superinduction was eliminated when

transcription was activated. This result was corroborated bycotransfection experiments using the bicistronic reporter (Fig. 5D)instead of EGFP. Cap-dependent translation of firefly luciferaseonly was favored when transcription was inactivated. Finally, de-letion of the 5′ LTR promoter from pLKO.1c did not affect su-perinduction of cotransfected EGFP, nor did linearization of theplasmid DNA with either SphI or EcoRI as measured by FACSor Western blotting (Fig. 5 E and F). In contrast, cleavage ofpLKO.1c into fragments of ∼1 kb or less with HinfI eliminatedsuperinduction of EGFP. These experiments are consistentwith superinduction of cap-dependent translation by circularor linear DNA.

DiscussionUsing lentiviral vectors for RNA interference experiments, weobserved vector-induced expression of cotransfected genes (super-induction). Although vectors lacking retroviral sequences did notincrease expression from cotransfected plasmids, transfer of the5′ end of the lentivirus vector, including the shRNA sequences,caused superinduction of at least five different exogenous genes(mCherry, EGFP, Rem, firefly luciferase, and Renilla luciferase).Cotransfected plasmids showed no difference in the steady statelevel of RNA expressed, whereas Western blotting or fluorescencemicroscopy of transfected cells expressing retroviral sequencesrevealed ∼5- to 10-fold increases in protein expression. Dele-tions of pLKO.1c suggested that retroviral and shRNA sequenceeffects are additive. An unrelated gammaretroviral vector, MigR1,also gave superinduction (Fig. 2). Superinduction was observedwith different transfection methods (electroporation, calciumphosphate, and lipid reagents) and independent cell lines. Thus,the retroviral effect occurs in trans, is not dependent on specificreporters or transfection protocol, and is not limited to specificcell types or species.Experiments with a vector containing the pLKO.1c Psi-shRNA

region downstream of an inducible promoter indicated thatvector transcription inhibited superinduction of cotransfected

Fig. 5. Transcription of lentivirus sequences abolishes superinduction of cotransfected genes. (A) Diagram of bidirectional promoter vectors. (B) Insertion ofthe pLKO.1c Psi to shRNA sequence into pTRE3G-BI-mCherry superinduces mCherry in cis. The mean and SDs of triplicates are reported relative to pTRE3G-BI-mCherry in the absence of a transactivator. All values (lanes 2, 3, and 4) were highly statistically significant (P < 0.0001) relative to values from lane 1, as waslane 3 relative to 4. (C) Insertion of the pLKO.1c Psi to shRNA sequence into pTRE3G-BI-mCherry superinduces EGFP in trans. Without the transactivator,pTRE-Psi-sh elevated GFP expression ∼10-fold compared with pcDNA3 (P < 0.0001; lanes 5 and 1) and ∼sixfold compared with pTRE-3G-BI-mCherry (P < 0.0001;lanes 5 and 3). (D) Cap-dependent translation increases in the presence of Psi-sh sequences from pLKO.1c. In the absence of DOX, pTRE-Psi-sh elevated cap-dependent translation ∼13-fold relative to pcDNA3 (lanes 5 and 1; P < 0.005) and ∼sixfold relative to pTRE3G-BI-mCherry (lanes 5 and 3; P < 0.005). (E) Loss ofthe 5′ LTR and linearization of pLKO.1c does not affect superinduction. Cells (293) were transfected with 5 μg of the inducer plasmid and 2.5 ng of EGFPexpression plasmid and, after 48 h, analyzed by FACS as described in B. Asterisks indicate statistical significance with control samples containing uncleavedpcDNA3 (P < 0.05). (F) Western blotting confirms that loss of the 5′ LTR and linearization of pLKO.1c does not affect superinduction. A second transfectionexperiment was performed as described in E, but triplicate samples were pooled for Western blotting and quantitated by LI-COR methodology.

Gou et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3509

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

24,

202

0

genes and cap-dependent translation (Fig. 5). Furthermore,elimination of the 5′ LTR within pLKO.1c had little or noeffect on cotransfected gene expression. Linearization of pLKO.1cwithin the 5′ enhancer (SphI) or downstream of the shRNA(EcoRI), which would reduce transcript levels by differentmeans, did not affect superinduction, but multiple cleavages byHinfI did. Together, these data strongly suggest that DNA is re-sponsible for superinduction. Because transfection of multiplesequences that have little sequence identity, such as HIV-1,shRNA, and MuLV, increased cotransfected gene expression,our results are consistent with superinduction by a distinctDNA structure.Several pieces of evidence argue that retroviral sequences

alter the translation of coexpressed genes. First, no difference inRNA levels of cotransfected genes was detected. Second, pulselabeling experiments showed increased incorporation of radio-active amino acids into GFP expressed from a plasmid cotrans-fected with pLKO.1c. Third, transfection of bicistronic vectorstogether with pLKO.1c increased cap-dependent expressionpreferentially compared with IRES-dependent reporter expres-sion from the same transcript. Therefore, introduction of retro-viral vectors has the opposite effect of stress or IFN induction.Increased translation of cotransfected genes indicates that

retroviral sequences trigger cell signaling. Many signals affectingtranslation alter mTOR kinase activity (26). Nonetheless, nodifferences were detected in the phosphorylation of targets up-stream or downstream of mTORC1, such as AKT, p70 S6K, and4E-BP1, in the presence or absence of pLKO.1c. The mTORC1kinase inhibitor Torin 1 (27) showed no selective effect on len-tivirus-mediated superinduction versus basal expression. The4E-BP1 mutant, F114A, which cannot interact with Raptor in themTORC1 complex and cannot be phosphorylated and releasedfrom eIF4E (24), inhibited basal and lentivirus-induced trans-lation to similar extents (Fig. 4D). In contrast, expression of wild-type 4E-BP1 selectively inhibited pLKO.1c-mediated GFP, but

not basal GFP levels (Fig. 4 C and D). Increased expression of4E-BP1 dramatically increased phosphorylated forms (Fig. 4C),but how this would inhibit translation initiation of cappedmRNAs is unclear. We detected twofold increased levels ofeIF4E in the presence of pLKO.1c, suggesting signaling thatelevates eIF4E and its ability to stimulate eIF4A helicase activity(28). Many cancer cells have elevated eIF4E levels, resulting inincreased translational efficiency (29). Nevertheless, eIF4E over-expression did not reproduce retroviral superinduction.DNA elements in both lentiviral and gammaretroviral vectors

increased cap-dependent translation of cotransfected genes intrans and in cis (Fig. 5). Because modified vectors would elimi-nate the need for recloning, addition of these vectors duringtransfection will save both time and expense while improvingprotein production. Our results reveal unforeseen effects ofshRNA expression vectors. Expression of an shRNA, even thosewith no known targets, by transient transfection may dramaticallyincrease levels of protein products from cotransfected expressionplasmids. In addition, we found that cap-dependent translationcan be controlled by mechanisms that are independent of phos-phorylation by the mTOR kinase. These experiments suggest thatretroviral DNA controls the translation of coexpressed genes, thusaffecting subgenomic mRNA expression. Understanding the activityof these critical retroviral sequences will improve our knowledge oftranslational control as well as aid the design of vectors for genetherapy, DNA vaccines, and molecular biology.

Materials and MethodsConditions for growth of cell lines and transfections have been previouslyreported (30). Detailed methods and reagents are given in SI Materialsand Methods.

ACKNOWLEDGMENTS. We acknowledge helpful discussions with membersof the J.P.D. laboratory and Drs. Jon Huibregtse, Arlen Johnson, and ChrisSullivan. This work was supported by Grant R01CA167053 from the NationalInstitutes of Health.

1. Lu HH, Wimmer E (1996) Poliovirus chimeras replicating under the translationalcontrol of genetic elements of hepatitis C virus reveal unusual properties of the in-ternal ribosomal entry site of hepatitis C virus. Proc Natl Acad Sci USA 93(4):1412–1417.

2. Sarnow P, Cevallos RC, Jan E (2005) Takeover of host ribosomes by divergent IRESelements. Biochem Soc Trans 33(Pt 6):1479–1482.

3. Goubau D, et al. (2014) Antiviral immunity via RIG-I-mediated recognition of RNAbearing 5′-diphosphates. Nature 514(7522):372–375.

4. Reikine S, Nguyen JB, Modis Y (2014) Pattern recognition and signaling mechanismsof RIG-I and MDA5. Front Immunol 5:342.

5. Swiecki M, Omattage NS, Brett TJ (2013) BST-2/tetherin: Structural biology, viral an-tagonism, and immunobiology of a potent host antiviral factor. Mol Immunol 54(2):132–139.

6. McClure R, Massari P (2014) TLR-dependent human mucosal epithelial cell responsesto microbial pathogens. Front Immunol 5:386.

7. Rustagi A, Gale M, Jr (2014) Innate antiviral immune signaling, viral evasion andmodulation by HIV-1. J Mol Biol 426(6):1161–1177.

8. Shida H (2012) Role of nucleocytoplasmic RNA transport during the life cycle of ret-roviruses. Front Microbiol 3:179.

9. Pollom E, et al. (2013) Comparison of SIV and HIV-1 genomic RNA structures revealsimpact of sequence evolution on conserved and non-conserved structural motifs. PLoSPathog 9(4):e1003294.

10. Byun H, et al. (2010) Retroviral Rem protein requires processing by signal peptidaseand retrotranslocation for nuclear function. Proc Natl Acad Sci USA 107(27):12287–12292.

11. Mertz JA, Simper MS, Lozano MM, Payne SM, Dudley JP (2005) Mouse mammarytumor virus encodes a self-regulatory RNA export protein and is a complex retrovirus.J Virol 79(23):14737–14747.

12. Mertz JA, Chadee AB, Byun H, Russell R, Dudley JP (2009) Mapping of the functionalboundaries and secondary structure of the mouse mammary tumor virus Rem-responsive element. J Biol Chem 284(38):25642–25652.

13. Hosokawa N, et al. (2008) Human XTP3-B forms an endoplasmic reticulum qualitycontrol scaffold with the HRD1-SEL1L ubiquitin ligase complex and BiP. J Biol Chem283(30):20914–20924.

14. Pear WS, et al. (1998) Efficient and rapid induction of a chronic myelogenous leu-kemia-like myeloproliferative disease in mice receiving P210 bcr/abl-transduced bonemarrow. Blood 92(10):3780–3792.

15. Donnelly N, Gorman AM, Gupta S, Samali A (2013) The eIF2α kinases: Their structuresand functions. Cell Mol Life Sci 70(19):3493–3511.

16. Buchkovich NJ, Yu Y, Pierciey FJ, Jr, Alwine JC (2010) Human cytomegalovirus inducesthe endoplasmic reticulum chaperone BiP through increased transcription and activationof translation by using the BiP internal ribosome entry site. J Virol 84(21):11479–11486.

17. White JP, Cardenas AM, Marissen WE, Lloyd RE (2007) Inhibition of cytoplasmic mRNAstress granule formation by a viral proteinase. Cell Host Microbe 2(5):295–305.

18. Yu M, et al. (1997) Cloning of a gene (RIG-G) associated with retinoic acid-induceddifferentiation of acute promyelocytic leukemia cells and representing a new mem-ber of a family of interferon-stimulated genes. Proc Natl Acad Sci USA 94(14):7406–7411.

19. Petersen CP, Bordeleau M-E, Pelletier J, Sharp PA (2006) Short RNAs repress trans-lation after initiation in mammalian cells. Mol Cell 21(4):533–542.

20. Roux PP, Topisirovic I (2012) Regulation of mRNA translation by signaling pathways.Cold Spring Harb Perspect Biol 4(11):1–23.

21. Shin S, et al. (2014) Glycogen synthase kinase-3β positively regulates protein synthesisand cell proliferation through the regulation of translation initiation factor 4E-bindingprotein 1. Oncogene 33(13):1690–1699.

22. Gingras AC, et al. (1999) Regulation of 4E-BP1 phosphorylation: A novel two-stepmechanism. Genes Dev 13(11):1422–1437.

23. Matsuo H, et al. (1997) Structure of translation factor eIF4E bound to m7GDP andinteraction with 4E-binding protein. Nat Struct Biol 4(9):717–724.

24. Dennis MD, Kimball SR, Jefferson LS (2013) Mechanistic target of rapamycin com-plex 1 (mTORC1)-mediated phosphorylation is governed by competition between sub-strates for interaction with raptor. J Biol Chem 288(1):10–19.

25. Gigoux M, et al. (2014) Inducible costimulator facilitates T-dependent B cell activationby augmenting IL-4 translation. Mol Immunol 59(1):46–54.

26. Thoreen CC (2013) Many roads frommTOR to eIF4F. Biochem Soc Trans 41(4):913–916.27. Thoreen CC, et al. (2012) A unifying model for mTORC1-mediated regulation of

mRNA translation. Nature 485(7396):109–113.28. Feoktistova K, Tuvshintogs E, Do A, Fraser CS (2013) Human eIF4E promotes mRNA

restructuring by stimulating eIF4A helicase activity. Proc Natl Acad Sci USA 110(33):13339–13344.

29. Ruggero D, et al. (2004) The translation factor eIF-4E promotes tumor formation andcooperates with c-Myc in lymphomagenesis. Nat Med 10(5):484–486.

30. Byun H, et al. (2012) Requirements for mouse mammary tumor virus Rem signalpeptide processing and function. J Virol 86(1):214–225.

3510 | www.pnas.org/cgi/doi/10.1073/pnas.1420477112 Gou et al.

Dow

nloa

ded

by g

uest

on

June

24,

202

0