Computationally designed adeno-associated virus(AAV) Rep 78 is efficiently maintained within anadenovirus vectorVarsha Sitaramana, Patrick Hearingb, Charles B. Wardc, Dmitri V. Gnatenkoa, Eckard Wimmerb, Steffen Muellerb,Steven Skienac, and Wadie F. Bahoua,1

Departments of aMedicine, bMolecular Genetics and Microbiology, and cComputer Sciences, Stony Brook University, Stony Brook, NY 11794

Edited by Thomas Shenk, Princeton University, Princeton, NJ, and approved July 20, 2011 (received for review February 21, 2011)

Adeno-associated virus (AAV) is a single-stranded parvovirus retain-ing the unique capacity for site-specific integration into a transcrip-tionally silent region of the human genome, a characteristicrequiring the functional properties of the Rep 78/68 polypeptide inconjunction with AAV terminal repeat integrating elements. Pre-vious strategies designed to assemble these genetic elements intoadenoviral (Ad) backbones have been limited by the general in-tolerability of AAV Rep sequences, prompting us to computationallyreengineer the Rep gene by using synonymous codon pair recoding.Rep mutants generated by using de novo genome synthesis main-tained the polypeptide sequence and endonuclease properties ofRep 78, while dramatically enhancing Ad replication and viral titeryields, characteristics indistinguishable from adenovirus lackingcoexpressed Rep. Parallel approaches using domain swaps encom-passing WT and recoded genomic segments, coupled with iterativecomputational algorithms, collectively established that 3′ cis-actingRep genetic elements (and not the Rep 78 polypeptide) retain dom-inant-acting sequences inhibiting Ad replication. These data provideinsights into the molecular relationships of AAV Rep and Ad replica-tion, while expanding the applicability of synonymous codon pairreengineering as a strategy to effect phenotypic endpoints.

codon pair bias | gene therapy | hybrid virus | systems biology

Adeno-associated virus (AAV) is a nonpathogenic single-stranded parvovirus that displays the unique capacity for site-specific integration into the transcriptionally silent AAVS1 regionof the human genome located on 19q13.42 (1, 2). The small 4.7-kbAAV genome encodes three structural capsid (VP1–VP3) andfour nonstructural replication (Rep) proteins translated from twoORFs, and transcriptionally regulated by p5 (Rep 78 and Rep 68),p19 (Rep 52 and Rep 40), and p40 (VP1–VP3) promoters(reviewed in ref. 3). Productive AAV infection requires helperfunctions generally supplied by adenovirus (Ad) or herpesvirus,and latency likely occurs by nonhomologous deletion/substitutionevents (4–6), resulting in head-to-tail stably integrated con-catemers (7, 8). AAV-mediated site-specific integration requiresAAV Rep 78/68 delivered in trans (9), a cis-acting Rep-bindingelement found within the flanking terminal repeats (TRs) (10),and a restricted 33-bp cellular sequence withinAAVS1 (5). Recentdata have implicated a 138-bp integration efficiency element (i.e.,p5IEE) within the p5 promoter as being sufficient and necessaryfor efficient Rep 78/68-mediated site-specific integration (11).The incorporation of these AAV integrating elements into

larger-capacity hybrid viruses represents a logical strategy forsite-specific genetic replacement therapies of large transgenes.Although the AAV integrating elements (i.e., TRs or p5IEE) arereadily incorporated into herpesvirus (12, 13) or Ad vectors (14–16), Rep 78/68 is poorly tolerated. Moderate success has beenachieved with the use of complex homologous recombinationstrategies (17) and helper-dependent (18) or tightly regulated(19) expression systems, although the latter two approaches areadditionally restricted by the helper-dependent nature of Rep

78-containing Ad. The mechanism of Rep 78/68-mediated Adinhibition remains incompletely elucidated (20, 21), althoughRep 78 is known to inhibit Ad replication (22), and colocalizes toAd replication centers to prevent their maturation (23). Fur-thermore, the complexity of these relationships is highlighted bythe lack of Rep 78/68-associated Ad inhibition when delivered intrans, such as in Rep 78/68-expressing cell lines (24).In this article, we have reengineered the AAV Rep gene by

modifying synonymous codon pairs to phenotypically affect thereplicative properties of Ad-expressing Rep 78. Although codonbias (i.e., the preferential use of synonymous codons duringtranslation) is well recognized, codon pair bias (CPB) representsa second, independent bias present at multiple phylogeneticlevels from microorganisms to humans (25–27). Similar to codonbias, synonymous codons can be paired in multiple ways to en-code two contiguous amino acids, with evidence for strong CPBas evidenced by disproportionate over- and under-representationof codon pairs (28). Previous strategies to modify CPBs havebeen developed as novel approaches to synthesize attenuatedpoliovirus (26) and influenza virus (27), although not for char-acterization and amelioration of cis-inhibitory signals relevant tocomplex viral interrelationships (e.g., between Ad and AAV).Two computationally recoded Repmutants differing in their CPBscores (but with preserved amino acid sequence) considerablyenhanced Ad replication and viral titer yields while preservingcritical Rep78 endonuclease (i.e., excision) capacity. Iterativecomputational algorithms coupled with genomic domain swapsspecifically established that a dominant, cis-acting genetic ele-ment(s) was localized to a 3′-Rep sequence, and that these in-hibitory effects could be ameliorated by genetically restructuringcodon pairs. These data provide a unique application of synon-ymous codon pair reengineering to modulate biological systems.

ResultsComputational Reengineering of AAV Rep Gene. We constructeda first-generation Ad carrying the AAV2 Rep78 coding sequenceunder a tightly regulated tetracycline-inducible promoter withinthe background of an E1/E3-deleted (ΔE1ΔE3) Ad5 virus. Thisvirus was additionally modified to replace the Ad5 fiber knobwith that of Ad35 (Ad5/35 chimer) as a strategy for efficientinfectivity of hematopoietic stem cells (29). Although the iden-tical tetracycline-inducible Rep expression cassette was pre-viously used for the successful construction of helper-dependent

Author contributions: V.S., P.H., C.B.W., E.W., S.M., S.S., andW.F.B. designed research; V.S.and P.H. performed research; C.B.W., E.W., S.M., and S.S. contributed new reagents/analytic tools; V.S., P.H., D.V.G., E.W., and W.F.B. analyzed data; and V.S., P.H., and W.F.B.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: wadie.bahou@stonybrook.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental.

14294–14299 | PNAS | August 23, 2011 | vol. 108 | no. 34 www.pnas.org/cgi/doi/10.1073/pnas.1102883108

Dow

nloa

ded

by g

uest

on

July

6, 2

021

mailto:wadie.bahou@stonybrook.eduhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1102883108

Ad (19), we noticed that the same construct was incapable ofreplication in the context of a ΔE1ΔE3 Ad, demonstrating nosigns of viral growth despite multiple passages in 293 cells overthe course of 50 d. This effect was noted not only with thisconstruct (which was designed to solely express Rep78, and notRep 68, Rep 52, or Rep 40) (19), but also seen using WT Rep. Itremained unclear why a first-generation ΔE1ΔE3 Ad expressingRep78/68 was not viable—although consistent with previousobservations (16, 17, 20)—and presumably explained by “leaky”or higher expression levels of Rep 78 in the context of a ΔE1ΔE3backbone compared with that of a helper-dependent Ad vector.As an alternative explanation, we hypothesized that coex-

pression of AAV Rep sequences within the Ad5 backbone nega-tively modulated Ad replication via dominant-acting inhibitorysequences. Accordingly, we recoded the 1,866-bp Rep genomicsegment to precisely preserve the amino acid sequence of the Rep78 polypeptide while disrupting any cis-acting sequences that couldputatively inhibit Ad function. Two distinct mutantRep genes weredesigned by changingCPB for synonymous recoding ofRep 78 (26,

27). A previously developed computational algorithm was appliedto generate a scrambled (s) Rep mutant by shuffling synonymouscodon pairs while maintaining codon use and the free energy offolded RNA to prevent large changes in secondary structure (26).In parallel, we designed a maximally codon-deoptimized (d) Repgene (by incorporating under-represented codon pairs) to assessthe effect of attenuated Rep 78 translation on Ad replication (Fig.1 A and B and Fig. S1). Both dRep and sRep were generated byusing de novo genome synthesis and precisely maintained the na-tive Rep 78 polypeptide sequence.

Recoded AAV Rep Is Sufficient for Replication and Generation ofΔE1ΔE3 Ad. Infectious plasmid (p) clones containing dRep,sRep, or WT Rep constructs within the fiber-modified ΔE1ΔE3Ad5/35 chimer were used for assessment of Ad replication.Following transfection of 293 cells, complete cytopathic effect(CPE) was observed with both pAd/sRep and pAd/dRep withinthe first passage and by 15 d after transfection, in sharp contrastto that seen using pAd/wtRep. These results paralleled those

5’ 3’-OH

Wild-type AAV genome

-0.4 -0.2 0.0 0.20

2000

4000

6000

8000

10000

dRep78 sRep78 wtRep78

Deoptimized

Scrambled

0.2 0.6 1.0 1.4 1.8

Per

cent

Iden

tity

100

50

50

100

Size (kb)

TR P5 P40 setisecilpS04/86peR91P

P5 P19 P40

2.2

Codon pair bias score

10-1

100

101

102

103

104

105

Day 2 Day 15

Ad/sRep Ad/wtRepAd/AAV/BDD Ad/dRep

Rep

A

B C

Cap

Fig. 1. Characterization of codon-modified AAV Rep. (A) Annotated schema of the 4,679-bp AAV genome is shown, delineating the predominant Rep andCap ORFs (47) (Upper), along with an expanded view (Lower) depicting scrambled and deoptimized homology alignments to WT Rep nucleotides 1 to 2,400.Sequence identity plots encompassing recoded base pairs 321 to 2,186 were generated by using Vista genomic tools (48); the dashed line is set at 75% identityto WT Rep and areas of white shading depict segments displaying less than 70% identity. TR, AAV inverted TRs; promoters (p5, p19, and p40) and Rep 68/40alternate splice sites are shown. Complete sequences and alignments are provided in Fig. S1. (B) The calculated CPB score for wtRep 78, sRep 78, and dRep 78are shown compared with those of fully annotated human genes found in the RefSeq release 22 database (N = 23,731). Each circle represents the CPB forindividual genes as a function of amino acid length; underrepresented codon pairs give negative scores, whereas a positive CPB indicates the predominantincorporation of overrepresented codon pairs. The peak distribution of the human gene set has a positive CPB of 0.07. (C) Ad replication of distinct constructswas quantified by qPCR by using Ad-specific primers and DpnI-digested nuclear DNA isolated at designated time points. Ad genome accumulation wascalculated from triplicate wells (45), and results are expressed as the mean ± SEM from three complete sets of experiments.

Sitaraman et al. PNAS | August 23, 2011 | vol. 108 | no. 34 | 14295

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

6, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental/pnas.201102883SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental/pnas.201102883SI.pdf?targetid=nameddest=SF1

obtained by quantitative PCR (qPCR; Fig. 1C), which demon-strated progressive Ad genome accumulation (for pAd/sRep andpAd/dRep only) nearly identical to that of the ΔE1ΔE3 Ad/AAV/BDD hybrid virus lacking the Rep gene (14, 30) (Fig. 2A).These initial observations collectively established that geneticallyrecoded Rep fundamentally altered replicative properties of Adwhen carried in cis. As confirmation, we generated two addi-tional constructs that substituted the strong constitutively actinghuman small nuclear RNA HU1-1 promoter (31) in place of thetetracycline-inducible promoter, again observing complete CPEwithin 15 d after transfection of 293 cells. Furthermore, pro-ductive viral titer concentrations for all dRep or sRep-expressingviruses were nearly identical, and comparable to those seen withAd/AAV/BDD (Table 1). These results contrast sharply with ourconsistent inability to generate Ad when expressed with wtRep.

Preserved Endonuclease Function of Rep 78. Plasmid transfection in293 cells established that Rep 78 protein expression was nearlyidentical between wtRep and sRep with approximately 50% re-duction of dRep (Fig. 2B), patterns predicted based onCPB scores.Functional competence of WT and mutant proteins was estab-lished by an endonuclease assay that evaluates Rep 78’s ability tocleave the RBS within the folded AAV TRs (32). These assayswere performed by using Ad/AAV/BDD hybrid virus as substrate(30), and were initially documented by using plasmid transfections,and subsequently confirmed by using intact Ad coinfections (videinfra). Cleavage of the AAV TR by Rep 78 during Ad DNA rep-lication is expected to release approximately 14 kb dimeric andapproximately 7 kb monomeric excision products (16) (Fig. 2A).Transfection of 293 cells with shuttle plasmids expressing thevarious Rep genes, followed by Ad/AAV/BDD infection, demon-strated identical monomeric and dimeric excision products for allthree Rep genes, enhanced in the presence of doxycycline andcomparable to Rep 78 delivered in trans from an AAV2 plasmidcontaining Rep/Cap genes and endogenous promoters (Fig. 2C).AttenuatedRep 78 protein expression from the dRep gene product

was associated with diminished generation of monomeric and di-meric excisional products compared with wtRep and sRep, asexpected based on the protein expression patterns.We subsequently used cesium chloride-purified Ad/sRep and

Ad/dRep virus coinfections with Ad/AAV/BDD as second-tieredconfirmation for Rep 78 endonuclease function in the context ofproductive and stable (passage 4) Ad generation. Similar to theresults identified earlier, excision products of the expected sizewere evident by using both Ad/sRep and Ad/dRep (Fig. 2D), andidentical to those generated by using a HeLa-derived cell linestably expressing the AAV Rep 78 and Cap proteins (24). Asexpected, no excision products were seen in the absence of AAVRep 78. Comparable amounts of dimeric and monomeric excisionproducts using eitherAd/sReporAd/dRep (comparedwith plasmidtransfections; Fig. 2C) are presumably explained by logarithmicviral replication which compensates for deoptimized Rep 78polypeptide expression. Similarly, we demonstrated that a singlebackbone Ad incorporating genetically recoded Rep (dRep) andAAV integrating elements retained the capacity for stable self-excision over three sequential passages (Fig. S2). These data col-lectively established that replication competence ofRep-expressingAd could be rescued by using genetically recoded Rep sequences,and were most consistent with cis-acting inhibitory sequenceswithin the context of ΔE1ΔE3 Ad and unrelated to expressionlevels or functional properties of the Rep 78 polypeptide.

Delineation of Rep Inhibitory Sequences That Block ΔE1ΔE3 AdReplication. To more specifically localize the sequence-specificRep genetic segment(s) that inhibit Ad replication, we generatedfour Rep chimers encompassing distinct combinations of wtRep orsRep (Fig. 3A). SRep genetic sequences were specifically chosen(over dRep) to recapitulate growth characteristics unrelated toattenuated Rep 78 polypeptide expression. All constructs wereinserted downstream of the tetracycline-inducible promoter, andproductive Ad viral replication was subsequently evaluated in 293cells as delineated earlier. Of the four constructs evaluated, only

Ad/ReppCMV tTS

A

AdpPF4 BDD AAV TR Ad/AAV/BDDp5IEE

pPF4 BDD AAV TRp5IEE

pPF4 BDD AAV TRp5IEE

pPF4BDDp5IEE

C

dRep or sRep

AdpTK IRES-EGFPRep

D

M

Ad/AAV/BDD excisional monomer (M) Ad/AAV/BDD excisional dimer (D)

Dox + + + - - -

*

D*

M

D

TRE

~ 7 Kb ~ 14 Kb

Rep 78

GAPDH

B

Fig. 2. Functional analyses of genetically reco-ded Rep 78. (A) Schema of genetic constructs,viruses, and predicted monomeric and dimericforms resulting from productive Rep 78 cleavageof Ad/AAV/BDD [based on previously character-ized models (16, 30)]. ψ, Ad left end packagingsequence (0.4 kb); pCMV, CMV core promoter(0.8 kb); tTS, tetracycline (Tet)-controlled tran-scriptional silencer (1.3 kb); TRE, Tet responseelement (0.3 kb); pTK, thymidine kinase pro-moter (0.2 kb); IRES-EGFP, internal ribosomeentry site with enhanced GFP (2.0 kb); Ad, Adbase pairs 3,330 to 3,940; BDD, human B-domaindeleted factor VIII (4.6 kb) (30); pPF4, PF4 pro-moter (1.1 kb) (14); p5IEE (135 bp) and AAV TR(145 bp each plus G–C tail) are shown. (B)Immunodetection of Rep 78 was established bytransfection of pFLAG/wtRep, pFLAG/sRep, andpFLAG/dRep in 293 cells, followed by detectionby using 1:1,000 anti-FLAG (Rep 78) or 1:1,000anti-GAPDH MAbs as control (10 μg lysates wereloaded per lane). (C) Excision assays were per-formed by transfecting 293 cells with individualpAd/Rep plasmids, followed by Ad/AAV/BDD in-fection (MOI of 50), in the presence (+) or ab-sence (−) of doxycycline (1 μg/mL) for 24 h beforeevaluation of dimeric (D) or monomeric (M) Ad/AAV/BDD excision products, generated only with functional Rep 78 endonuclease cleavage at the right TR (16).Genomic blots were performed by using 1 μg DNA per lane, and detected by using the approximately 750-bp PF4/BDD junction fragment as probe (A). pIM45 isan AAV plasmid expressing WT Rep and Cap genes, used as positive control for excision. Faint, low-level excision in the absence of doxycycline is presumablya result of leaky Rep 78 expression. (D) Excision assays using viral coinfections (MOI of 50) were performed in 293 (Ad/sRep or Ad/dRep) or C12 cells with Ad/AAV/BDD in the presence of 1 μg/mL doxycycline for 24 h, and at 48 h, Hirt DNA was isolated for genomic analysis as previously. C12 are HeLa-derived stablecell lines that express Rep and Cap upon Ad coinfection, and are used as positive controls. In C and D, parent Ad/AAV/BDD virus is depicted by an asterisk.

14296 | www.pnas.org/cgi/doi/10.1073/pnas.1102883108 Sitaraman et al.

Dow

nloa

ded

by g

uest

on

July

6, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental/pnas.201102883SI.pdf?targetid=nameddest=SF2www.pnas.org/cgi/doi/10.1073/pnas.1102883108

Ad/s(wt3) Rep encompassing the approximately 550 bp 3′-se-quence fromWTRep (nucleotides 1,623–2,186) failed to replicateas long as 50 d after transfection. The other three constructscontaining 3′-scrambled sequences [Ad/s(wt1) Rep, Ad/s(wt2)Rep, and Ad/s(wt1,2) Rep] demonstrated complete CPE, and viralyields for all three replicating Ad chimers were nearly identicaland comparable to those of Ad/dRep, Ad/sRep, and Ad/AAV/BDD (Table 1). These data provided strong presumptive evidencethat dominant-acting inhibitory sequences were contained withina restricted genomic segment encompassing 3′ WT Rep sequen-ces, and that genetically recoded Rep sequences encompassingthis region sufficiently relieved this effect.

Delimitation of Rep Inhibitory Sequences Using Combinatorial GroupTesting. In a final approach, we applied combinatorial group testing

and balanced Gray codes to both confirm and further delimit thesequence-specific Rep inhibitory signal in wtRep (Fig. 3B). Giventhe binary nature of the phenotypic endpoint (i.e., Ad replication/no replication), we predicted that the replicative characteristics ofa limited number of WT/scrambled Rep chimers could be appliedto further delineate distinctRep inhibitory sequences. Accordingly,we synthesized four chimers, each with 24 − 2 (n = 14) combina-tions of WT or scrambled Rep sequences encompassing equal-sized (∼135 bp) segments (Fig. S3). These 14 interwoven segmentsof sRep and wtRep thereby provided unique signatures whosegrowth characteristics would more precisely delimit Rep inhibitorysequences (note that two homogeneous sRep and wtRep sequenceswere omitted as experimental controls). The replicative charac-teristics of these four constructs (Ad/Rep 1, Ad/Rep II, Ad/Rep III,and Ad/Rep IV) were subsequently studied in ΔE1ΔE3 Ad. Ad/Rep II,Ad/Rep III, andAd/Rep IV replicated efficiently in 293 cells,althoughAd/Rep I showed no signs of viral replication as long as 50d after transfection (Fig. 3 C and D and Table 1). These growthpatterns were entirely concordant with (and served as independentvalidation of) domain swaps delineated earlier. Furthermore,growth patterns specifically delimited the 3′-terminal Rep in-hibitory signals to a discrete 135-bp genomic segment spanning bp1,782 to 1,916, a region encompassing both the p40 promoter andRep 68/Rep 40 splice site, and previously identified as critical forRep-dependent p40 promoter activity (33). We then generatedcomplementation chimers encompassingWT135-bp sequences onthe background of sRep (wt135/sRep) or 135-bp sRep on thebackground of WT (s135/wtRep). Both viruses were capable ofgrowth with slightly attenuated Ad/s135/wtRep viral titers (Table1), reaffirming the computational predictions and confirming thata discrete, recoded 135-bp genomic segment was sufficient in re-lieving Rep-mediated Ad inhibition.

DiscussionWe have genetically recoded the AAV Rep gene by using syn-onymous codon pair reengineering to overcome Rep’s inhibitoryeffects on Ad replication. Two computationally redesignedmutants with distinct CPB scores dramatically enhanced Adreplication and viral titer yields to levels nearly identical to thoseof Rep-negative Ad. Distinct complementary approaches applied

Wild-type Scrambled

Scrambled

ScrambledWild-type

Wild-typeScrambled

Wild-type Scrambled

321 AfeI 2186

S (wt1) Rep

S (wt2) Rep

S (wt3) Rep

S (wt1,2) Rep

A Viability+

+

+

-

BstBI

B Viability-+

+

+

Wild-type sequence

Ad/Rep IIAd/Rep I

Ad/Rep IIIAd/Rep IV

Scrambled Sequence

*

C2 2 2210 10 1010

Ad/Rep II

KbDay

** ****

p

2.0

3.0

1.0

Ad/Rep IAd/Rep IV Ad/Rep III

D

10-1

100

101

102

103

104

105

Ad/sRep Ad/AAV/BDD Ad/Rep IIAd/Rep IVAd/Rep IIIAd/Rep IAd/wtRep

Day 2 Day 5 Day 10 Day 15

Aden

oviru

s D

NA

( ng/

ngAc

tin)

Fig. 3. Delineation of Rep inhibitory sequences. (A)Chimeric Rep genes were assembled by poly-nucleotide domain swaps encompassing discretesegments of WT or scrambled sequences, and viabil-ity established by Ad replication and titer determi-nations (16) (Table 1 shows detailed viral titers). (B)Four distinct Rep chimers each containing 14 discrete(132–135 bp) segments of WT or scrambled sequen-ces were synthesized, and Ad replication (i.e., viabil-ity) was studied in HEK 293 cells (complete nucleotidesequences are provided in Fig. S3). Note that thecolumns can be permuted in any of 14! (∼8.7 × 1010)combinations with equivalent ability to identify crit-ical domains, provided the sequence is encompassedwithin one of 14 segments. To minimize the effect ofsignals on boundaries, columns were ordered tominimize transitions, in effect creating a balancedGray (binary) code whose distinct genetic signaturesand phenotypic growth patterns can be applied fordelineation of the critical Rep inhibitory segment(delineated by the asterisk, and encompassing WTRep sequences 1,782–1,916). (C) Southern blot anal-ysis was performed using DpnI/SbfI double-digestedHirt DNA isolated at day 2 or day 10 from 293 cellstransfected with pAd/Rep I, pAd/Rep II, pAd/Rep III,or pAd/Rep IV, and detected using Ad base pairs 1–194 as probe. Note the diminution of DpnI-sensitiveinput plasmid (p) with concomitant appearance of DpnI-resistant replicated Ad (asterisk) at D10 for all constructs except Ad/Rep I. (D) Ad genome accumulationwas determined by qPCR by using viruses isolated at distinct time points; results are the mean ± SEM from three complete sets of experiments.

Table 1. Viral titers

Virus Titer* (pfu/mL × 108)

Ad/sRep 9.50 ± 0.50Ad/dRep 9.50 ± 0.25Ad/wtRep No growthAd/HU1-1/sRep 7.50 ± 0.75Ad/HU1-1/dRep 8.00 ± 0.50Ad/HU1-1/wtRep No growthAd/s(wt1) Rep 8.13 ± 0.12Ad/s(wt2) Rep 8.75 ± 0.25Ad/s(wt1,2) Rep 8.13 ± 0.12Ad/s(wt3) Rep No growthAd/Rep I No growthAd/Rep II 11.5 ± 0.75Ad/Rep III 10.4 ± 0.37Ad/Rep IV 9.38 ± 0.12Ad/s135/wtRep 2.98 ± 0.50Ad/wt135/sRep 8.00 ± 1.00Ad/AAV/BDD 10.5 ± 0.50

*Titers were calculated by serial dilution and plaque assay in HEK 293 pack-aging cells, and are reported in pfu/mL as the mean ± SEM from two distinctdeterminations.

Sitaraman et al. PNAS | August 23, 2011 | vol. 108 | no. 34 | 14297

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

6, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental/pnas.201102883SI.pdf?targetid=nameddest=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental/pnas.201102883SI.pdf?targetid=nameddest=SF3

to further delineate Rep inhibitory sequences included: (i) do-main swaps encompassing WT and scrambled genomic segments,(ii) the application of a combinatorial sequence algorithm spe-cifically designed to sublocalize discrete genomic signals basedon Ad replicative growth characteristics, and (iii) final validationwith complementation chimers. These collective strategies pro-duced concordant conclusions, establishing that 3′-terminal Repsequences (restricted to discrete genetic elements encompassingbase pairs 1,782–1,916) retain cis-acting inhibitory signals whoseeffects can be relieved by genetically reengineering codon pairs.Strategies to computationally redesign large genetic elements

as a means of effecting specific biological functions have maturedin parallel with technological advances that provide efficientsynthesis of sizable DNAs (34, 35). We have previously usedcodon-pair deoptimization methodologies to synthesize attenu-ated poliovirus (26) and influenza virus (27), and report here (forthe first time to our knowledge) the application of this strategyfor identification and strategic amelioration of cis-inhibitorysignal sequences relevant to complex viral functions. Surpris-ingly, there is a paucity of experimental evidence focusing on themechanism(s) or the evolutionary pressures for selective CPB.Our collective experiences clearly demonstrate that geneticrecoding strategies designed to usurp this evolutionary processhave broad applications to modulate diverse biological systems.What is the mechanism whereby genetically modified Rep

ameliorates Ad replication? Productive AAV infection in thepresence of Ad causes coordinate induction of p5, p19, and p40promoters mediated by Rep gene products in trans. To date, itremains unestablished if these Rep 78 binding interactions func-tion in concert with other cellular transcription factors. Optimalinduction of p19 and p40 are dependent on the presence of mul-tiple cis-acting elements acting in concert (33), and in the case ofp40, it is intriguing that these elements have beenmapped (by usingdeletion mutagenesis) to a 90-bp sequence that overlaps with the135-bp genetic element identified by our data (33). Multipletranscriptional consensus sequences (i.e., AP1, SP1, GGT) arefound in the p40 promoter overlap region (Fig. S1), althoughprevious data are more consistent with complex DNA-proteinjuxtapositioning of various Rep genetic elements collectively in-volving p5, p19, and p40 (33). Our data with the use of positive andnegative complementation chimers further dissected the criticalfunction of the discrete 135-bp segment in modulating Ad repli-cation, suggesting the presence of a complex model involvingnondiscrete, juxtapositioned Rep elements (33). Alternative ex-planations involving DNA secondary structure, poorly delineatedconsequences of CPB bias, or the presence of cryptic AAV-encoded miRNAs remain plausible (36). Given the complexity ofthese interactions, we speculate that the identification of thesegenetic elements would not have been elucidated by using routinemutagenesis strategies involving limited Rep segments, but ratherrequired the broad-based computational recoding strategy thatextends beyond mutagenesis of candidate genomic regions.Insertional mutagenesis remains a fundamental concern for

long-term gene replacement strategies (37, 38) and for geneticreprogramming of pluripotent stem cells (39). AAV site-specificintegration is a unique example of an evolutionarily developedeukaryotic system that is capable of minimizing adverse eventsassociated with insertional mutagenesis by targeting a transcrip-tionally silent region of the human genome. This effect is clearlycomplicated, and involves not just the Rep 78–Rep-bindingelement–AAVS1 trimolecular complex, but includes incompletelycharacterized host cell recombination proteins (40). Furthermore,AAVS1 is located within a gene-dense region of the genome, withevidence that the majority of viral/cellular junctions are foundwithin the contiguous MBS85 gene (41, 42). An explanation forthe relatively benign nature of the disrupted integration site hasbeen proposed that incorporates partial duplication of the targetlocus, presumably resulting in a preserved functional copy of

MBS85 (6). We did not alter the Rep 78 polypeptide or itsfunctional properties, thereby preserving the DNA binding (43),ATPase (32), helicase (44), and endonuclease (32) activities es-sential for targeted AAV site-specific integration strategies.Identification of genetically modified Rep that is readily toler-

ated within the Ad genome has implications for gene therapystrategies at two levels: (i) facilitated assembly of a single-back-bone delivery system retaining the requisite genetic elementsnecessary for site-specific integration, and (ii) generation of one-step packaging systems for recombinant AAV (rAAV) viral pro-duction (16). Indeed, our initial feasibility studies confirmed theviability of such a hybrid virus that was stable and retained self-excision capacity over three passages. rAAV is widely used asa gene transfer vector retaining the capacity for long-term extra-chromosomal persistence and transgene expression from non-integrated genomes (rAAVs are unable to accommodate Repbecause of size and toxicity constraints) (31). Nonetheless, cur-rent strategies for rAAV generation and packaging remain cum-bersome, requiring multiple plasmids and/or helper viruses forproduction of clinical grade material. Given its high titer and in-fectivity, Ad carrying an AAV TR-flanked transgene on the samebackbone with sRep (or dRep) and Cap genes provides the req-uisite elements for one-step rAAV production.

Materials and MethodsReagents, Ad Substrates, and Cell Lines. HEK 293 cells, HeLa cells, and HeLa-derived C12 cells expressing AAV Rep and Cap genes (24) were maintained asadherent monolayers in DMEM containing 10% FCS. Ad/AAV/BDD was gener-ated from the parent Ad/AAV hybrid virus expressing human B-domain–deletedfactor VIII [FVIIIΔ761–1639 (BDD)] (14, 30), and was specifically modified to in-corporate the platelet factor 4 (PF4) promoter upstream of BDD (14), and tan-dem 135-bp p5IEEs integrating elements upstream of the 145-bp AAV TR (11).

Molecular Genetic Studies. Codon pair deoptimized and scrambled AAV Repsequences were designed by using computational algorithms (vide infra) (26,27), and synthesized de novo (GenScript). All Rep constructs were designedwith flanking (i.e., unique) SbfI and SwaI sites, and a uniquely designed AfeIsite (base pair 981) was incorporated to facilitate genetic cloning and ma-nipulation. Constructs were generated by using standard molecular techni-ques, and fully sequenced to ensure proper assembly. Epitope-tagged Rep 78polypeptide was generated by PCR amplification of sRep, dRep, and wtReplacking the stop codon, and expressed in-frame with a C-terminal (3×) FLAGpolypeptide (DYKDDDK) within pCMV-3Tag-3A (Stratagene) for cellulartransfection and immunodetection (vide infra). Viral characterization andreplication was established by qPCR or by genomic blot analysis using DpnI/SbfI-restricted Hirt DNA and alkaline phosphatase-labeled Ad base pairs 1 to194 as probe. As DpnI requires dam methylated substrates, it specificallydigests transfected DNA of bacterial origin, to the exclusion of hemi/unme-thylated replicating viral DNA. qPCR was performed on nuclear DNA using Ad-specific primers (Table S1); quantifications were determined from triplicatewells (45), and standardized to β-actin to ensure cross-sample comparisons.

Ad Assembly and Characterization. The tetracycline-inducible Rep78 expres-sion cassette was provided by F. Mavilio (Italian Institute of Technology, Unitof Molecular Neuroscience, Istituto Scientifico H. San Raffaele, Milan, Italy)and A. Recchia (Department of Biomedical Sciences, University of Modenaand Reggio Emilia, Modena, Italy). All plasmid (p) constructs used for Adgeneration were assembled in shuttle vectors derived from pAd/AAV-EGFP-Neo (16), and were generated by homologous recombination in BJ5183 cellsbetween the shuttle vectors and pTG3602 ΔE3-F5/35, which contains theintact WT Ad 5 genome (46), specifically modified to replace the Ad5 fiberknob with that of Ad35 (29). E1/E3-deleted (ΔE1ΔE3) Ad was generated andtitered in 293 cells as previously described (30).

Rep 78-mediated excision assays were performed in 293 cells with plasmidtransfections or Ad coinfections, by using the TR within Ad/AAV/BDD as thesubstrate for endonuclease function. Plasmid transfections by using pAd/Repshuttle vectors were performed in 293 cells supplemented (or not) with 1 μg/mL doxycycline, followed at 24 h with Ad/AAV/BDD superinfection (50 pfu/cell); excision assays with viral coinfections were demonstrated by usingidentical multiplicities of infection (MOIs; 50 pfu/cell). In both situations,Hirt DNA was isolated 48 h after infection for genomic analyses by usingdigoxigenin-labeled probes (16, 30).

14298 | www.pnas.org/cgi/doi/10.1073/pnas.1102883108 Sitaraman et al.

Dow

nloa

ded

by g

uest

on

July

6, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental/pnas.201102883SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102883108/-/DCSupplemental/pnas.201102883SI.pdf?targetid=nameddest=ST1www.pnas.org/cgi/doi/10.1073/pnas.1102883108

Immunoblot Analysis. Rep 78 immunodetection was established in 293 cellstransfected with FLAG-tagged Rep plasmids (pFLAG/wtRep, pFLAG/dRep, orpFLAG/sRep), and at 48 h, protein-solubilized lysates were prepared for 4% to15% SDS/PAGE and immunoblotting as previously described (14). Immunode-tection was performed by enhanced chemiluminescence by using anti-FLAG M2(1:1,000; Sigma)andanti-GAPDHMAB374 (1:1,000;Millipore) as a loading control.

Computational Algorithms. The CPB score for the 1,866-bp Rep sequence wascalculated as the arithmetic mean of individual codon pair scores (26),updated to incorporate the most recent annotation of the National Centerfor Biotechnology Information RefSeq data set (version 22; March 5, 2007).We applied our computational algorithm to manipulate the CPB of the1,866-bp Rep 78 coding region spanning base pairs 321 to 2,186 withoutmodifying the initiator methionine. This recoded segment was designed toprecisely maintain the polypeptide sequence and the codon use (i.e., thefrequency of use of each existing codon), thereby “shuffling” existingcodons to manipulate the CPB. The algorithm uses a mathematical formulafor simulated annealing suitable for full-length optimization, and isdesigned to prevent manipulation of regions with large secondary structuressuch as hairpin or stem loops (26).

We developed a sequence design procedure using concepts from discretecombinatorial group testing to locate Rep inhibitory sequences that manifesteither inhibitory or facilitatory phenotypic functions related to Ad replica-tion [i.e., a balanced Gray (binary) code]. In this model, four different se-quence designs (each of which contained discrete 132–135-bp segments ofWT or scrambled Rep nucleotide sequences) allow for 24 combinations of WTor scrambled sequences, each with a distinct interwoven pattern of chimericgenomic segments. These 16 combinations (minus the two homogenous WTand scrambled combinations) generate columns that can be permuted in 14!orderings, each of which maintains the structure to locate critical sequences.We reduced the effect of signals on boundaries by ordering the columns tominimize the number of transitions, thereby creating a series of orderedmatrices in which neighboring regions differ in exactly one of the fourcomputationally generated mathematical designs.

ACKNOWLEDGMENTS. WethankDr. Jizu Zhi for generation ofVista plots, andDrs. Nicholas Muzyczka and Paul Freimuth for helpful discussions. This workwas supported by New York State Stem Cell Board Grants C024317 and N09S-006 (to W.F.B.); National Institutes of Health Grants AI075219 (to E.W.) andAI41636 (to P.H.); and National Science Foundation Grant IIS-1017181 and In-telligence Community Postdoctoral Fellowship HM1582-07-BAA-0005 (to S.S.).

1. Samulski RJ, et al. (1991) Targeted integration of adeno-associated virus (AAV) intohuman chromosome 19. EMBO J 10:3941–3950.

2. Kotin RM, et al. (1990) Site-specific integration by adeno-associated virus. Proc NatlAcad Sci USA 87:2211–2215.

3. Berns KI, Giraud C (1996) Biology of adeno-associated virus. Curr Top Microbiol Im-munol 218:1–23.

4. McCarty DM, Young SM, Jr., Samulski RJ (2004) Integration of adeno-associated virus(AAV) and recombinant AAV vectors. Annu Rev Genet 38:819–845.

5. Linden RM, Winocour E, Berns KI (1996) The recombination signals for adeno-asso-ciated virus site-specific integration. Proc Natl Acad Sci USA 93:7966–7972.

6. Henckaerts E, et al. (2009) Site-specific integration of adeno-associated virus involvespartial duplication of the target locus. Proc Natl Acad Sci USA 106:7571–7576.

7. Duan D, et al. (1998) Circular intermediates of recombinant adeno-associated virushave defined structural characteristics responsible for long-term episomal persistencein muscle tissue. J Virol 72:8568–8577.

8. Duan D, Yan Z, Yue Y, Engelhardt JF (1999) Structural analysis of adeno-associatedvirus transduction circular intermediates. Virology 261:8–14.

9. Surosky RT, et al. (1997) Adeno-associated virus Rep proteins target DNA sequences toa unique locus in the human genome. J Virol 71:7951–7959.

10. McCarty DM, et al. (1994) Identification of linear DNA sequences that specifically bindthe adeno-associated virus Rep protein. J Virol 68:4988–4997.

11. Philpott NJ, Gomos J, Berns KI, Falck-Pedersen E (2002) A p5 integration efficiencyelement mediates Rep-dependent integration into AAVS1 at chromosome 19. ProcNatl Acad Sci USA 99:12381–12385.

12. Johnston KM, et al. (1997) HSV/AAV hybrid amplicon vectors extend transgene ex-pression in human glioma cells. Hum Gene Ther 8:359–370.

13. Wang Y, et al. (2002) Herpes simplex virus type 1/adeno-associated virus rep(+) hybridamplicon vector improves the stability of transgene expression in human cells by site-specific integration. J Virol 76:7150–7162.

14. Damon AL, et al. (2008) Altered bioavailability of platelet-derived factor VIII duringthrombocytosis reverses phenotypic efficacy in haemophilic mice. Thromb Haemost100:1111–1122.

15. Fisher KJ, Kelley WM, Burda JF, Wilson JM (1996) A novel adenovirus-adeno-associ-ated virus hybrid vector that displays efficient rescue and delivery of the AAV ge-nome. Hum Gene Ther 7:2079–2087.

16. Sandalon Z, Gnatenko DV, Bahou WF, Hearing P (2000) Adeno-associated virus (AAV)Rep protein enhances the generation of a recombinant mini-adenovirus (Ad) utilizingan Ad/AAV hybrid virus. J Virol 74:10381–10389.

17. Carlson CA, Shayakhmetov DM, Lieber A (2002) An adenoviral expression system forAAV rep78 using homologous recombination. Mol Ther 6:91–98.

18. Wang H, Lieber A (2006) A helper-dependent capsid-modified adenovirus vectorexpressing adeno-associated virus rep78 mediates site-specific integration of a 27-kilobase transgene cassette. J Virol 80:11699–11709.

19. Recchia A, Perani L, Sartori D, Olgiati C, Mavilio F (2004) Site-specific integration offunctional transgenes into the human genome by adeno/AAV hybrid vectors. MolTher 10:660–670.

20. Casto BC, Armstrong JA, Atchison RW, HammonWM (1967) Studies on the relationshipbetween adeno-associated virus type 1 (AAV-1) and adenoviruses. II. Inhibition ofadenovirus plaques by AAV; its nature and specificity. Virology 33:452–458.

21. Casto BC, Atchison RW, Hammon WM (1967) Studies on the relationship betweenadeno-associated virus type I (AAV-1) and adenoviruses. I. Replication of AAV-1 incertain cell cultures and its effect on helper adenovirus. Virology 32:52–59.

22. Timpe JM, Verrill KC, Trempe JP (2006) Effects of adeno-associated virus on adeno-virus replication and gene expression during coinfection. J Virol 80:7807–7815.

23. WeitzmanMD, Fisher KJ, Wilson JM (1996) Recruitment of wild-type and recombinantadeno-associated virus into adenovirus replication centers. J Virol 70:1845–1854.

24. Clark KR, Voulgaropoulou F, Johnson PR (1996) A stable cell line carrying adenovirus-inducible rep and cap genes allows for infectivity titration of adeno-associated virusvectors. Gene Ther 3:1124–1132.

25. Gutman GA, Hatfield GW (1989) Nonrandom utilization of codon pairs in Escherichiacoli. Proc Natl Acad Sci USA 86:3699–3703.

26. Coleman JR, et al. (2008) Virus attenuation by genome-scale changes in codon pairbias. Science 320:1784–1787.

27. Mueller S, et al. (2010) Live attenuated influenza virus vaccines by computer-aidedrational design. Nat Biotechnol 28:723–726.

28. Tats A, Tenson T, RemmM (2008) Preferred and avoided codon pairs in three domainsof life. BMC Genomics 9:463.

29. Shayakhmetov DM, Papayannopoulou T, Stamatoyannopoulos G, Lieber A (2000)Efficient gene transfer into human CD34(+) cells by a retargeted adenovirus vector.J Virol 74:2567–2583.

30. Gnatenko DV, et al. (2004) Expression of therapeutic levels of factor VIII in hemophiliaA mice using a novel adeno/adeno-associated hybrid virus. Thromb Haemost 92:317–327.

31. Gnatenko DV, et al. (1999) Human factor VIII can be packaged and functionally ex-pressed in an adeno-associated virus background: Applicability to haemophilia Agene therapy. Br J Haematol 104:27–36.

32. Im DS, Muzyczka N (1990) The AAV origin binding protein Rep68 is an ATP-dependent site-specific endonuclease with DNA helicase activity. Cell 61:447–457.

33. McCarty DM, Christensen M, Muzyczka N (1991) Sequences required for coordinateinduction of adeno-associated virus p19 and p40 promoters by Rep protein. J Virol 65:2936–2945.

34. Mueller S, Papamichail D, Coleman JR, Skiena S, Wimmer E (2006) Reduction of therate of poliovirus protein synthesis through large-scale codon deoptimization causesattenuation of viral virulence by lowering specific infectivity. J Virol 80:9687–9696.

35. Gibson DG, et al. (2010) Creation of a bacterial cell controlled by a chemically syn-thesized genome. Science 329:52–56.

36. Pfeffer S, et al. (2004) Identification of virus-encoded microRNAs. Science 304:734–736.

37. Hacein-Bey-Abina S, et al. (2003) LMO2-associated clonal T cell proliferation in twopatients after gene therapy for SCID-X1. Science 302:415–419.

38. Cavazzana-Calvo M, et al. (2010) Transfusion independence and HMGA2 activationafter gene therapy of human β-thalassaemia. Nature 467:318–322.

39. Stadtfeld M, Nagaya M, Utikal J, Weir G, Hochedlinger K (2008) Induced pluripotentstem cells generated without viral integration. Science 322:945–949.

40. Daya S, Cortez N, Berns KI (2009) Adeno-associated virus site-specific integration ismediated by proteins of the nonhomologous end-joining pathway. J Virol 83:11655–11664.

41. Dutheil N, Shi F, Dupressoir T, Linden RM (2000) Adeno-associated virus site-specifi-cally integrates into a muscle-specific DNA region. Proc Natl Acad Sci USA 97:4862–4866.

42. Tan I, Ng CH, Lim L, Leung T (2001) Phosphorylation of a novel myosin binding sub-unit of protein phosphatase 1 reveals a conserved mechanism in the regulation ofactin cytoskeleton. J Biol Chem 276:21209–21216.

43. Chiorini JA, et al. (1994) Sequence requirements for stable binding and function ofRep68 on the adeno-associated virus type 2 inverted terminal repeats. J Virol 68:7448–7457.

44. Wonderling RS, Kyöstiö SR, Owens RA (1995) A maltose-binding protein/adeno-associated virus Rep68 fusion protein has DNA-RNA helicase and ATPase activities.J Virol 69:3542–3548.

45. Gnatenko DV, et al. (2003) Transcript profiling of human platelets using microarrayand serial analysis of gene expression. Blood 101:2285–2293.

46. Chartier C, et al. (1996) Efficient generation of recombinant adenovirus vectors byhomologous recombination in Escherichia coli. J Virol 70:4805–4810.

47. Srivastava A, Lusby EW, Berns KI (1983) Nucleotide sequence and organization of theadeno-associated virus 2 genome. J Virol 45:555–564.

48. Mayor C, et al. (2000) VISTA: Visualizing global DNA sequence alignments of arbitrarylength. Bioinformatics 16:1046–1047.

Sitaraman et al. PNAS | August 23, 2011 | vol. 108 | no. 34 | 14299

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

6, 2

021