A Comprehensive RNA Sequencing Analysis of the Adeno-Associated Virus (AAV) Type 2 Transcriptome Reveals Novel AAV Transcripts, Splice Variants, and Derived Proteins

ABSTRACT

Adeno-associated virus (AAV) is recognized for its bipartite life cycle with productive replication dependent on coinfection with adenovirus (Ad) and AAV latency being established in the absence of a helper virus. The shift from latent to Ad-dependent AAV replication is mostly regulated at the transcriptional level. The current AAV transcription map displays highly expressed transcripts as found upon coinfection with Ad. So far, AAV transcripts have only been characterized on the plus strand of the AAV single-stranded DNA genome. The AAV minus strand is assumed not to be transcribed. Here, we apply Illumina-based RNA sequencing (RNA-Seq) to characterize the entire AAV2 transcriptome in the absence or presence of Ad. We find known and identify novel AAV transcripts, including additional splice variants, the most abundant of which leads to expression of a novel 18-kDa Rep/VP fusion protein. Furthermore, we identify for the first time transcription on the AAV minus strand with clustered reads upstream of the p5 promoter, confirmed by 5ˈ rapid amplification of cDNA ends and RNase protection assays. The p5 promoter displays considerable activity in both directions, a finding indicative of divergent transcription. Upon infection with AAV alone, low-level transcription of both AAV strands is detectable and is strongly stimulated upon coinfection with Ad.

IMPORTANCE Next-generation sequencing (NGS) allows unbiased genome-wide analyses of transcription profiles, used here for an in depth analysis of the AAV2 transcriptome during latency and productive infection. RNA-Seq analysis led to the discovery of novel AAV transcripts and splice variants, including a derived, novel 18-kDa Rep/VP fusion protein. Unexpectedly, transcription from the AAV minus strand was discovered, indicative of divergent transcription from the p5 promoter. This finding opens the door for novel concepts of the switch between AAV latency and productive replication. In the absence of a suitable animal model to study AAV in vivo, combined in cellulae and in silico studies will help to forward the understanding of the unique, bipartite AAV life cycle.

INTRODUCTION

Adeno-associated viruses (AAV) are helper-dependent members of the parvovirus group that require coinfection with an unrelated helper virus, particularly adenovirus (Ad) for productive replication (1). Despite of several identified AAV serotypes most research has been done with prototype AAV type 2. The AAV2 genome consists of a linear, single-stranded DNA of 4.7 kb. Both ends carry identical inverted terminal repeats (ITR) of 145 bp (2), which flank the two major open reading frames (ORF), called rep and cap. The rep ORF codes for four nonstructural proteins, Rep78 and a C-terminally spliced variant, Rep68. In addition, N-terminally truncated versions thereof are expressed, called Rep52 and Rep40, respectively. The Rep proteins are required as regulators for various steps of the AAV life cycle. AAV cap harbors the ORF for the capsid proteins (VP1 to -3) and a separate ORF for the assembly-activating protein (AAP) (3).

The current knowledge of AAV transcription dates back to early research in the 1980s when the three AAV2 promoters, p5, p19, or p40 and major transcripts derived thereof were characterized (4, 5). All AAV transcripts identified since then map to only one DNA plus strand (6), which led to the assumption that the complementary AAV minus strand was not transcribed. Furthermore, spliced AAV transcripts were identified displaying a single, shared intron located in the center of the AAV2 genome (4, 7). Initially, a single splice donor site at AAV nucleotide 1906 and a single splice acceptor site located at nucleotide (nt) 2228 were described. Later, an alternative splice acceptor site at nt 2201 was detected (8), allowing the expression of VP1 protein from the same promoter (p40) as VP2 and VP3. Compared to the major splice acceptor site (nt 2228), the so-called minor splice acceptor site was more divergent from the ideal consensus sequence, likely explaining its less-frequent use for splicing. During AAV replication the majority of p5- or p19-initiated transcripts are not spliced, whereas transcripts initiated from p40 are almost exclusively spliced (9, 10). All AAV transcripts share a 3ˈ end (7), coinciding with the polyadenylation signal identified between AAV2 nt 4424 and 4429 (2). The only other AAV serotype whose transcription pattern has been analyzed is AAV5, the genetically most distant member of the AAV family. Several variations in comparison to the AAV2 transcription pattern were described. Most importantly, AAV5 rep transcripts terminate at a second, internal polyadenylation signal located in the AAV5 intron (11).

Early analysis of AAV transcription relied on S1 nuclease mapping and reverse transcriptase-mediated primer extension (4, 7, 12, 13). These methods allowed the mapping and sequence analysis of defined and highly expressed transcripts. Subsequently, quantitative RNase protection was used to complement the currently accepted AAV transcription map (10). Using the described methods, only AAV RNAs in regions explicitly probed for could be detected. Recent NGS-based technologies allow high-throughput analysis of entire cellular transcriptomes, leading to unbiased identification and parallel quantification of all existing cellular and viral transcripts. With RNA sequencing (RNA-Seq), novel transcripts and splicing variants in human cells and of viral transcripts have been discovered (14, 15).

In this report we used Illumina-based RNA-Seq to provide a comprehensive picture of AAV2 transcription in the presence or absence of helper adenovirus. We were able to identify known and novel AAV2 transcripts, splicing variants, and a new protein derived therefrom. In addition, previously unrecognized RNAs transcribed from the AAV minus strand were detected.

MATERIALS AND METHODS

Cell culture and viruses.

HEK 293- and HeLa cells were cultivated as described previously (16). Human adenovirus type 2 (Ad2) was propagated and assayed in 293 cells.

Cloning and mutagenesis.

For the generation of different luciferase constructs, the corresponding AAV2 wild-type promoter sequences were cloned upstream of the firefly luciferase gene in a sense orientation, AAV(+)190-320-luc, or in a reverse complementary orientation, AAV(–)586-155-luc, AAV(–)320-180-luc, and AAV(–)320-256-luc, respectively. The constructs were verified by restriction analysis. For the generation of AAV mutants AAV-D1, AAV-D2, and AAV-A1, the corresponding mutations were introduced by the QuikChange site-directed mutagenesis protocol (Stratagene) into the AAV2 rep or cap genes of pTAV2-0 (17). The mutations were verified by DNA sequencing and restriction enzyme analysis. Similarly, the nucleotide exchange G4395T and the 9 amino acid C-terminal Flag-tag insertion (DYKDDDDK*) between AAV2 nucleotide positions 4411 and 4412 led to the constructs pTAV2-0 G4395T Flag (pAAVwt-Flag), and pTAV2-0 G4137A G4395T Flag (pAAV-A1-Flag). For the primers, see Table 1.

TABLE 1.

Oligonucleotides for mutagenesis PCR, qPCR, RT-PCR, and PCRs for detection of AAV2 splice events, Northern blot hybridization, and 5′RACE

Oligonucleotide	Sequence (5′-3′)
AAV-D1-Fwd	CTGACGGAATGGCGCCGTCTTTCGAAGGCCCCGGAGGCCCTT
AAV-D1-Rev	AAGGGCCTCCGGGGCCTTCGAAAGACGGCGCCATTCCGTCAG
AAV-D2-Fwd	CTGACGGAATGGCGCCGTGTTTCGAAGGCCCCGGAGGCCCTT
AAV-D2-Rev	AAGGGCCTCCGGGGCCTTCGAAACACGGCGCCATTCCGTCAG
AAV-A1-Fwd	CTTAAACACCCTCCTCCACAAATTCTCATCAAGAACACC
AAV-A1-Rev	GGTGTTCTTGATGAGAATTTGTGGAGGAGGGTGTTTAAG
Flag-Fwd	CCTGACTCGTAATCTGTAATGACTACAAGGATGACGATGACAAGTGATGCTTGTTAATCAATAAACCG
Flag-Rev	CGGTTTATTGATTAACAAGCATCACTTGTCATCGTCATCCTTGTAGTCATTACAGATTACGAGTCAGG
Stop-Mut-Fwd	CCATTGGCACCAGATACCTTACTCGTAATCTGTAATGAC
Stop-Mut-Rev	GTCATTACAGATTACGAGTAAGGTATCTGGTGCCAATGG
Rep2-Fwd	AGAAGGAATGGGAGTTGCCG
Rep2-Rev	TCTGACTCAGGAAACGTCCC
RT-Primer	CAGATTACGAGTCAGGTAT-CTGGTGCCAATGGGGCG
SD527-Fwd	AGAAGGAATGGGAGTTGCCG
SD988-Fwd	GTAAACGGTTGGTGGCGCAG
SD1906-Fwd	GCCATCGACGTCAGACGCGGAAGCTTC
SD3184-Fwd	CTTCAACAGATTCCACTGCC
SA807-Rev	GCTCCGTGAGATTCAAACAG
SA985-Rev	AGGCCGCATTGAAGGAGATG
SA1437-Rev	TTGCTTCCTCCGAGAATGGC
SA1635-Rev	TTGGTGACCTTCCCAAAGTC
SA2228-Rev	CGTAGGCTTTGTCGTGCTCG
SA4138-Rev	TGAAGGTGGTCGAAGGATTCGCAGGT
SDA1	CCTCGAGCCAATACGGCGCCATTC
p5-RACE-GSP1	GGTGGAGTCGTGACGTGAATT
p5-RACE-GSP2	TCATAGGGTTAGGGAGGTCCT

AAV2 production, purification, and quantification.

For AAV2 production, HEK 293 cells seeded at 30% confluence were transfected 24 h later with AAV plasmids and pHelper as described previously (18). AAV2 was purified from benzonase-treated, cleared freeze-thaw supernatants by one-step AVB Sepharose affinity chromatography and quantified by qPCR as described previously (19) with primers for AAV2 rep (Table 1). AAV2 infectious titers were determined by endpoint dilutions on Ad2-infected HeLa cells (20).

Virus infection and DNA plasmid transfection.

HeLa cells seeded at a density of 30% were infected 20 h later with AAV2 wild-type or AAV-A1 mutant (multiplicity of infection [MOI] 250) and/or Ad2 (MOI 25). Transfection of HeLa cells with AAV plasmids was performed as described previously (21). Infection with Ad2 (MOI 25) was performed 16 h posttransfection and cells were harvested at the indicated time points postinfection.

RNA extraction.

Total RNA was isolated by TRIzol reagent (Ambion) according to the manufacturer's protocol. RNA samples were treated with RNase-free Turbo DNase (Ambion) to remove residual DNA, followed by phenol-chloroform extraction and precipitation. RNA quality and integrity was verified on 0.8% agarose gels and on bioanalyzer (Agilent). Total RNA samples of proven high quality were used for reverse transcription-PCR (RT-PCR) and for RNA library generation. Poly(A)⁺ selected RNA was purified from total RNA with the Oligotex mRNA Midi purification kit (Qiagen).

RT-PCR.

Newly identified AAV2 splice sites were validated using 1 μg of DNase-digested total RNA, reverse transcribed by SuperScript II reverse transcriptase (Invitrogen) according to the manufacturer's protocol with a primer that binds at the rear 3ˈ end of the AAV2 genome (Table 1). The cDNAs were PCR amplified with AAV2-specific primer pairs (Table 1) located upstream of the predicted splice donor and downstream of the predicted splice acceptor sites.

Library preparation and Illumina sequencing.

1 μg of each of the four different total RNA samples (cells, AAV2, AAV2+Ad2, and Ad2) were subjected to total RNA library generation using the TruSeq stranded total RNA sample preparation kit according to the manufacturer's protocol (Illumina). The libraries were pooled and sequenced on a HiSeq 2000 platform (multiplexed SR 1×101+7 high-output mode).

Read mapping.

Demultiplexed sequencing reads (length, 101 bp; single end) were mapped after trimming Illumina adapters from the 3ˈ ends on the human genome (hg19, GRCh37) and the human transcriptome by using the Bowtie 2 read mapping tool (22). The transcriptome was derived from the RefSeq gene annotation downloaded from the UCSC table browser (https://genome.ucsc.edu/cgi-bin/hgTables). All reads that could not be mapped to the human genome/transcriptome were mapped to the reference sequences of Ad2 (RefSeq AC_000007.1) and of both the flip and the flop variants of AAV2 (RefSeq NC_001401.2). In all cases, only read mappings with an alignment score (as given in the XS field of the SAM output) of at least −20 were accepted. A detailed analysis of the reads mapped on AAV2 revealed that the genome of the sequenced AAV2 only differs from the reference at a single nucleotide (a C instead of a G at nt 3284).

Detection of splice junctions.

We generated a splice junction reference, i.e., a set of sequences containing fusions of all possible combination of canonical splicing donor sites (GT dinucleotides) and downstream splicing acceptor sites (AG dinucleotides) in AAV2. A split alignment of the sequencing read onto the AAV2 genome using the Segemehl mapping tool (23) revealed no evidence of splicing at noncanonical splicing sites (data not shown). The sequencing reads were mapped to the splice junction reference using Bowtie 2, where insertions and deletions were suppressed by setting the gap penalties to very high values. We then determined the number of uniquely mapped reads for each possible junction of donor and acceptor sites. Only reads that covered at least 20 bp upstream of the splice donor site and 20 bp downstream of the splice acceptor site were counted.

Detection of polyadenylation sites.

We searched the reverse complement sequencing reads for polyadenine repeats of a length of 8 (i.e., AAAAAAAA) and mapped the parts of the reads upstream of the polyadenine repeats to the AAV2 genome. Note, that AAV2 contains no polyadenine repeat longer than 6. Mappings with quality below 20 were discarded.

Promoter prediction and ORF finder.

Promoters on the AAV2 wild-type genome were predicted by the Neural Network Promoter Prediction platform (NNPP) of the BDGP (http://www.fruitfly.org/seq_tools/promoter.html) (24). For the translation of alternative splice variants into potential new proteins an open reading frame (ORF) finder (NCBI) was used (http://www.ncbi.nlm.nih.gov/gorf/gorf.html).

Luciferase assay and Western blot analysis.

Luciferase assays and Western blot analysis were performed as described previously (16). Mouse monoclonal antibodies (MAbs) and rabbit polyclonal antibodies used for immunoblot analysis were anti-Rep MAb 303.9 (1:10; Progen), anti-VP MAb B1 (1:10; Progen), anti-VP1 MAb A1 (1:10; Progen), anti-VP1/2 MAb A69 (1:10; Progen) or rabbit and mouse anti-Flag (1:1,000; Sigma), respectively, followed by reaction with horseradish peroxidase-conjugated secondary antibodies and ECL detection.

Extraction of viral DNA and Southern hybridization.

Extraction of viral DNA by a modified Hirt procedure and Southern blot analysis were performed as described previously (16). Samples of DpnI-digested Hirt DNA (6 μg) were analyzed on Southern blots for AAV replication with a 1.6-kb fragment from the cap region of pTAV2-0 as probe labeled either with [³²P]dCTP for radioactive detection or with Biotin-11-dUTP for nonradioactive detection.

Northern hybridization with oligonucleotide probes.

Northern blot hybridization of total or poly(A)⁺ selected RNA samples was performed with a 24-nt oligonucleotide spanning 12 nt upstream and downstream of the AAV2 527:2228 splice junction (SDA1, see Table 1) labeled with T4 polynucleotide kinase in the presence of [γ-³²P]ATP. Hybridization was performed for 4 h at 68°C in hybridization solution (0.5% [wt/vol] sodium dodecyl sulfate [SDS], 6× SSC, 50 mM HEPES [pH 7.8], 150 μg/ml salmon sperm DNA, and 5× Denhardt solution). The filters were washed four times in 6× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)–0.1% SDS at room temperature for 5 min and subsequently twice with 6× SSC–0.1% SDS at 68°C for 30 min.

RNase protection assay.

RNase protection assays were performed with the RPA III kit (Ambion) essentially as described by the supplier. Two AAV2 sense RNA probes (nt 170 to 250 and nt 170 to 340) were radiolabeled with [α-³²P]UTP by in vitro transcription with T7 polymerase (Riboprobe system; Promega) using XbaI-linearized pBluescript vector containing the corresponding AAV nucleotides as the templates. The in vitro-transcribed RPA probes were gel purified on an 8% PAA–8 M urea denaturing gel. A total of 3 × 10⁴ cpm of purified radiolabeled probe were hybridized with 20 μg of total RNA, followed by RNase A/RNase T₁ treatment according to the manufacturer's protocol with an additional final ethanol precipitation step. The protected RNA fragments were electrophoresed on an 8% PAA–8 M urea gel and subsequently fixed in a solution containing 20% ethanol and 5% acetic acid for 30 min before vacuum drying for 2 h at 80°C. Gels were exposed to a storage phosphor screen and scanned by a Fujifilm FLA-3000 imager.

RACE.

Rapid amplification of cDNA ends (RACE) was conducted using the 5ˈRACE system (version 2.0; Invitrogen) according to the manufacturer's protocol. First-strand cDNA synthesis was performed on total RNA from AAV2/Ad2-coinfected HeLa cells with a primer binding to AAV2 nt 151 to 171 (p5-RACE-GSP1), followed by treatment with RNase H. The cDNA was purified and poly(C)-tailed using terminal deoxynucleotidyltransferase (TdT). After TdT removal, PCR amplification was performed using the TrueStart Hot Start Taq DNA polymerase (Thermo Scientific) and a nested primer binding to AAV2 nt 175 to 195 (p5-RACE-GSP2) and the anchor primer supplied by the manufacturer. The PCR products were analyzed and excised from agarose gels and subsequently cloned into pBluescript II SK(+). Randomly picked clones were subjected to DNA sequencing (Eurofins Genomics). For the primer sequences, see Table 1.

RESULTS

RNA-Seq library preparation.

To identify the ideal time point for RNA-Seq analysis, time course experiments with total RNAs and proteins extracted from HeLa cells infected with AAV2 in presence or absence of adenovirus (Ad2) were performed. At 27 h postinfection (hpi), AAV2 transcripts were clearly detectable on Northern blots, as were Rep and VP on Western blots. This time point was chosen for total RNA extraction from cells infected with AAV2 in presence or absence of Ad2 and from uninfected and Ad2-infected cells that served as controls. All RNA samples were analyzed by bioanalyzer tests displaying RNA integrity numbers higher than 8.90, which are indicative of highly intact RNA. Total RNA was rRNA depleted and digested to 100 nucleotide fragments. RNA libraries were sequenced on an Illumina HiSeq 2000 platform.

Illumina RNA-Seq next-generation sequencing.

For each set of data, between 21.2 and 26.6 million reads were obtained by Illumina RNA-Seq (Table 2). The vast majority of reads (>95%) could be successfully mapped either to the human genome or to the viral genomes of AAV2 or Ad2. In AAV2-infected cells, in the absence of Ad2 ca. 97% of all RNA reads were assigned to the human genome, which is comparable to uninfected control cells. Only few AAV-specific transcripts (0.04%) could be detected at 27 hpi. In the presence of Ad, AAV-specific transcripts rose to 13.6% of the total read count (Table 2). The majority of AAV reads mapped to the coding plus strand of the AAV2 genome. However, a lower, but significant proportion of reads mapped to the AAV2 minus strand. Parallel to the Ad-induced increase of AAV-specific transcripts, a reduction of the percentage of RNAs mappable to the human genome was found. This was also seen for cells infected with Ad2 alone, suggestive of an Ad-induced effect (Table 2). AAV2/Ad2-coinfected cells showed a 5-fold reduction of Ad2-specific reads compared to cells infected with Ad2 alone. This finding likely reflects the established effect that AAV inhibits adenoviral replication (25–28).

TABLE 2.

RNA-Seq analysis: assignment of the reads to the species from which they originate

Data set	Total no. of reads	No. (%) of mappable reads	No. (%) of human reads	No. (%) of AAV2 reads			No. (%) of Ad2 reads
				Plus strand	Minus strand	Ambiguous
Cells	23,380,236	22,665,799 (96.94)	22,662,280 (96.93)	1,404 (0.01)	14 (0.00)	0 (0.00)	2,101 (0.01)
AAV2	21,200,968	20,601,318 (97.17)	20,591,869 (97.13)	7,616 (0.04)	241 (0.00)	1 (0.00)	1,591 (0.01)
AAV2 + Ad2	23,047,796	22,167,728 (96.18)	16,327,433 (70.84)	3,061,388 (13.28)	68,099 (0.30)	184 (0.00)	2,710,624 (11.76)
Ad2	26,564,586	25,351,109 (95.43)	10,920,188 (41.11)	1,614 (0.01)	16 (0.00)	0 (0.00)	14,429,291 (54.32)

RNA read mapping on the AAV2 plus strand.

The population of AAV2-specific RNA reads for each data set was mapped to the wild-type AAV2 genome. Based on this alignment, a coverage map of the AAV2 genome was created that displays the number of reads that map to a specific AAV2 genome position. The genome coverage of the data set from AAV2/Ad2-coinfected cells is based on approximately 3.0 million AAV2-specific reads (Fig. 1A). Steep increases of RNA-Seq read counts are indicative of either transcription initiation or of splicing. The reads assigned to the AAV2 plus strand largely confirmed the well-established transcription initiation at the p5, p19, and p40 promoters. Also in line with previous studies, transcripts initiated at the p40 promoter were more frequent compared to p5- or p19-initiated transcripts in AAV2/Ad2-coinfected cells, whereas in cells infected solely with AAV2 the p5-initiated transcripts were the most prevalent (Fig. 1B). In the absence of Ad, few AAV-specific transcripts were detectable. With the exception of the prevalent p5-initiated transcripts, the overall AAV transcription profile was very similar to that in the presence of Ad, although at a >200-fold reduced level. Additionally, occasional reads were detected upstream of the p5 promoter on the AAV plus strand in the absence or presence of Ad2 (Fig. 1A and B), a finding indicative of transcription initiated within the left ITR of AAV2, as described previously (11, 29, 30). A single polyadenylation site near the right end of the AAV2 genome was defined before (31). RNA-Seq shows that indeed the vast majority of Ad2-induced AAV2 transcripts terminate by polyadenylation at nt 4450 (Fig. 1C). No other hotspots for polyadenylation were detected.

FIG 1.

RNA-Seq: mapping and transcription analysis on both strands of the AAV2 genome. (A) Coverage map of AAV2-specific RNA-Seq reads to the AAV2 genome of AAV2/Ad2-coinfected cells at 27 hpi. The AAV2 genome scale is displayed in the center. Reads mapped to the plus strand are presented above, and reads mapped to the minus strand are presented below. (B) Display as in panel A. RNA-Seq reads of AAV2-infected cells at 27 hpi are presented. Note the different scales of the read counts in panels A and B. (C) Mapping of RNA-Seq reads ending with a poly(A) tail. The genome position of poly(A) addition is displayed relative to the totality of polyadenylated AAV reads in the data set of AAV2/Ad2-coinfected cells. (D) Promoter activity of the AAV2 p5 region in reverse direction. The upper part shows the AAV2 genome with the nucleotide positions of the transcription start sites of the p5, p19, and p40 promoters (black arrows). In addition, putative promoters (gray arrows) originating from a promoter prediction program with a cutoff score above 0.9 are indicated on the plus or minus strand of the AAV genome with the respective nucleotide positions of the TSS. An asterisk (*) highlights the established p19 promoter, detected with a cutoff score below 0.9. The lower part shows the relative luciferase activities of the indicated candidate reverse p5 promoter sequences on the AAV2 minus strand. AAV(+)190-320-luc represents the AAV2 p5 promoter in the forward orientation serving as a positive control and empty pluc serving as a negative control. The relative luciferase activities of three different experiments were determined in the presence of Ad2 and are depicted as means ± standard deviations (SD). (E) 5ˈRACE analysis of AAV minus strand transcription. Displayed are the TSSs (arrows) derived from DNA sequences of cloned 5ˈRACE products from the AAV minus strand (nt 330 to 170). Single TSSs are displayed as black squares above the genome. The p5 promoter TSS at nt 287 (gray arrow) on the AAV plus strand is shown for orientation. (F) An RNase protection assay was performed with in vitro-transcribed hybridization probes covering AAV2 nt 170 to 250 (left part) or AAV2 nt 170 to 340 (right part) in a sense orientation for the detection of AAV minus strand transcripts in total RNA from HeLa cells either mock infected, AAV2 infected, Ad2 infected, or AAV2/Ad2 coinfected for 27 h. The undigested probes (–RNase) and the probes digested in the presence of yeast t-RNA (+RNase) are shown as controls. Both probes contain additional non-AAV2 sequences from the pBluescript vector used for T7 polymerase-dependent in vitro transcription. The maximal protected AAV2 regions are 81 nt for probe RPA 170-250, and 171 nt for probe RPA 170-340, respectively.

Analysis of transcription on the AAV minus strand.

As stated above, a significant proportion of reads mapped to the AAV2 minus strand in the absence, as well as in the presence of Ad coinfection (Fig. 1A and B). These new transcripts, detected here for the first time, were largely restricted to the AAV region between the left ITR and the p5 promoter. Their abundance was about one-third compared to p5-initiated transcripts on the AAV plus strand (Fig. 1A). To obtain hints, where transcription on the AAV2 minus strand was initiated we first searched for candidate upstream promoters. In silico promoter prediction (NNPP) (24) under stringent conditions with a cutoff score above 0.9 was applied to both AAV strands. While the well-characterized AAV2 p5 and p40 promoters could be readily detected with cutoff scores of 0.95 and 1.0, respectively, the p19 promoter was only detected when lowering the cutoff to a score of 0.53. On the AAV2 minus strand three candidate promoters were detected with cutoff scores above 0.9. Two of these are located near the left ITR, with putative transcription start sites at nt 230 (score 0.93) and nt 305 (score 0.94), respectively (Fig. 1D). These sites are in agreement with the assumed start sites for AAV2 minus strand transcription (Fig. 1A). The activity of the candidate promoters was tested using reporter gene constructs. Various AAV2 fragments comprising the p5 promoter region were cloned in reverse orientation upstream of the luciferase gene. Ad2-infected HeLa cells were transfected with the constructs, as depicted in Fig. 1D. The luciferase activity measured at 48 h posttransfection showed strong promoter activity of AAV(–)320-180-luc, the reverse complement of the established AAV p5 promoter AAV(+)190-320-luc, which was 7-fold more active. Deletion of nt 180 to 255 in the AAV(–)320-180-luc sequence led to an >30-fold drop in promoter activity, whereas an extended version of the p5 reverse complement, AAV(–)586-155-luc, was marginally stronger compared to AAV(–)320-180-luc. Relative promoter activities in sense and antisense orientation correlated well with the relative proportions of transcripts detected on the AAV plus and minus strands (Fig. 1A).

To more closely map transcription start sites on the AAV minus strand, we used 5ˈRACE with the gene-specific primer used for first-strand cDNA synthesis binding to AAV nt 151 to 171. The PCR products were cloned and individual representative clones sequenced to identify the respective transcription start site (TSS). The identified TSSs mapped to the region between AAV2 nt 227 to 320 (Fig. 1E). AAV minus strand transcription was further confirmed by RNase protection analysis (RPA) with probes covering AAV2 nt 170 to 250 or nt 170 to 340, respectively (Fig. 1F). With both probes, a rather heterogeneous population of AAV-specific protected RNA fragments in the size range of about 40 to 80 nt was detected specifically in AAV2/Ad2-coinfected cells (Fig. 1F) but not in mock- or Ad-infected cells.

Analysis of AAV2 splicing.

Spliced AAV transcripts were identified by RNA-Seq analysis as AAV reads joined to distant, defined parts of the AAV genome. Bioinformatic tools with stringent filters were utilized to exclude background noise. Only splice events above a critical abundance were investigated further. Reads covering the known AAV2 splice donor site at position 1906 were highly abundant, with >35,000 reads in the data set for AAV2/Ad2-coinfected cells. In ca. 72% of those reads, no splice event was detectable (Table 3). In about 23% of the reads, the splice donor site was joined to the described major splice acceptor site (nt 2228), and 3.5% of the reads were fused to the minor splice acceptor site (nt 2201), as expected from previous studies (8). Less frequently used splice donor sites were identified at nt 527, 988, and 3184. In addition, occasional, alternative AAV splice acceptor sites were detected mostly at nt 4138 (Table 3).

TABLE 3.

Summary of AAV2 splice events by RNA-Seq analysis: AAV2 + Ad2

Splice donor	Read count	% not spliced	% spliced with acceptora
			807	985	1437	1635	2201	2228	4138	Others
527	34,472	93.64	0.53	0.33	0.07	0.11	0.09	0.50	4.21	0.52
988	26,404	95.05			0.17	0.56	0.23	0.89	2.45	0.65
1906	35,746	72.25					3.57	23.42	0.36	0.40
3184	50,575	99.30							0.51	0.19

^aThe displayed values give the percentages for the splice donors joined to the various splice acceptors. Values in boldface are splice variants that were validated by RT-PCR.

The existence of alternative splice products was verified by RT-PCR with sets of primers whose binding sites were located upstream of the respective splice donor and downstream of the corresponding splice acceptor sites (Fig. 2). Additionally, AAV mutants were generated with an inactivated splice donor site at nt 527 (Fig. 2B, D1 and D2 mutants) or splice acceptor site at nt 4138 (Fig. 2D, A1 mutant), respectively. The described splice products 1906:2201/2228 were detected by RT-PCR, as shown for AAV wild-type and mutant A1 after coinfection with Ad2 (Fig. 2A). The 527:2228 splice variant was clearly detected with the AAV wild type but, as expected, not with the D1 and D2 mutants (Fig. 2B). Northern blots with total and with poly(A)⁺-selected RNA from AAV2/Ad2-coinfected HeLa cells were probed with an oligonucleotide spanning the 527:2228 splice junction. A single band of about 2.4 kb was detected (Fig. 2C) corresponding to a p5-initiated spliced mRNA that extends to the common poly(A)⁺ site (theoretical size, 2,436 nt). The splice events involving the splice acceptor site at nt 4138 leading to the splice products 527:4138, 988:4138, and 3184:4138 were detected by RT-PCR for the AAV2 wild type but, as expected, not for AAV mutant A1 (Fig. 2D to F). Notably, splice product 527:4138 had already been faintly detected in the absence of Ad2 infection, but none of the other splice products had (Fig. 2D). Further splice variants with lower frequencies in the RNA-Seq screen were also confirmed by RT-PCR (Table 3, values in boldface).

FIG 2.

Validation of RNA-Seq-derived AAV2 splice events by RT-PCR and Northern blot analysis. (A) The AAV2 genome with unspliced/spliced transcripts is displayed at the top. The primer for reverse transcription is shown with its binding position. The forward primer (nt 1862 to 1889) and reverse primer (nt 2440 to 2421) are displayed for PCR analysis of 1906:2201/2228 splicing. Below, the results of RT-PCR for AAV wild-type- and mutant A1-infected cells are displayed, separated on an agarose gel. PCR products are indicated by arrows. (B) RT-PCR of splice event 527:2228, performed as in panel A with cells transfected with plasmids for AAV wild-type or mutants D1 or D2. Recoded nucleotide sequences of AAV-D1 and AAV-D2 are displayed below the AAV genome. (C) Northern blot analysis of total and poly(A)⁺ selected RNA of cells coinfected with AAV2 and Ad2 using a 24-nt oligomer as a probe that spans the AAV2 splice junction 527:2228, as depicted below the AAV genome. The arrow points to the spliced transcripts. (D to F) RT-PCR analysis of additional AAV RNA splice events, performed as in panel A. (D) Forward primer (nt 415 to 434) and reverse primer (nt 4186 to 4161) used for the analysis of 527:4138 splicing. The corresponding mutation of AAV-A1 is displayed below the AAV genome. The asterisk (*) marks a faint band of the splice product in the absence of Ad. (E) Forward primer (nt 874 to 893) and reverse primer (nt 4186 to 4161) for the analysis of 988:4138 splicing. (F) Forward primer (nt 3051 to 3070) and reverse primer (nt 4186 to 4161) for the analysis of 3184:4138 splicing.

Involvement of the novel splice variants in the AAV replication cycle.

To address the functional significance of the newly identified splice variants for the AAV2 replication cycle, the AAV mutants for the splice donor site at nt 527 (D1 and D2) and the splice acceptor site at nt 4138 (A1) were assayed for Rep and Cap expression and for AAV DNA replication. In the D1 mutant the absolutely conserved G nucleotide at the 5ˈ end of the intron had been abolished, at the expense of a conservative exchange of valine to isoleucine in Rep78/68, whereas the D2 mutant maintains the amino acid sequence of Rep78/68 (Fig. 2B). On Western blots, mutant D1 showed strongly reduced Rep78 and capsid protein expression and some reduction of Rep52, whereas mutant D2 showed a less-pronounced effect on Rep, and VP expression was indistinguishable from that of AAV wild type (Fig. 3A). Also, AAV DNA replication was reduced more extensively for mutant D1 than D2 (Fig. 3B). In line with these findings, the titers of infectious AAV were reduced by 3 to 4 logs for mutant D1 and by 1 log for mutant D2 compared to AAV wild type (data not shown).

FIG 3.

Influence of alternative splicing on AAV replication. (A) Analysis of AAV2 Rep and VP expression in AAV wild type-, splice mutant AAV-D1-, or AAV-D2 (see Fig. 2B)-transfected HeLa cells at the indicated time points after Ad infection. AAV Rep and VP proteins are detected by MAb 303.9 or MAb B1, respectively. (B) Southern blot analysis of Hirt extracts of AAV wild type-, AAV-D1-, or AAV-D2-transfected and Ad-infected HeLa cells. AAV-specific monomeric (RF_M) and dimeric (RF_D) replicative forms, detected by probing with a labeled 1.6-kb cap-derived DNA sequence, are indicated by arrows. Input refers to undigested input DNA. (C) Western blot analysis as described in panel A with mutant AAV-A1 (see Fig. 2D). Mock- and Ad-infected cells served as controls. (D) Southern blot analysis as performed for panel B with mutant AAV-A1.

The AAV-A1 mutant, with inactivated splice acceptor site at nt 4138 and maintained VP amino acid sequence (Fig. 2D) showed markedly reduced Rep and Cap expression early (24 hpi), but not later (48 hpi) after infection (Fig. 3C). Furthermore, in contrast to AAV2 wild type, Rep52 was undetectable for mutant A1 in the absence of Ad2 (Fig. 3C). In line with these findings, AAV DNA replication was reduced, as evidenced by diminished monomeric and dimeric AAV2 replicative forms (Fig. 3D). In summary, the newly described, alternative splice products appear to play some role in the regulation of Ad-induced AAV replication, especially in the early stages of AAV replication.

Prediction and identification of novel AAV2 proteins.

Some of the alternatively spliced AAV transcripts could possibly lead to the translation of novel proteins, as predicted in silico. Candidate proteins of at least 50 amino acids, initiated by an ATG start codon or alternatively by ACG or CTG, were considered for further analysis. The most abundant splice variants and the predicted proteins are depicted in Fig. 4A. The splice variant 527:2228 could lead to a small protein of 9.1 kDa consisting of the N terminus of Rep78 fused to a short C terminus translated in a different reading frame as the Rep and VP proteins (Fig. 4A). Unfortunately for this predicted protein, no antibody was available, nor was an antibody available for the splice variant 988:2228-derived 26-kDa protein that lacks the central portion of Rep68. However, the presumed translation product of splice variant 527:4138 could be assayed by MAb B1, which is directed against AAV2 VP1-3. Indeed, an 18-kDa protein was detected in the low-molecular-weight range in cells infected with AAV wild type but not with mutant A1 (Fig. 4B). The 15-kDa bands visible in all extracts infected with AAV most likely represent autolytic capsid cleavage products, as described previously (32). The newly identified 18-kDa protein consists of 159 amino acids (Fig. 4C), of which the first 69 amino acids correspond to the N-terminal part of Rep78 and the C-terminal 90 amino acids correspond to VP. As described above, the novel 18-kDa Rep/VP fusion protein might have a function on the regulation of early AAV gene expression (see above).

FIG 4.

Detection of novel AAV proteins derived from alternative splicing. (A) The AAV2 genome and current transcription/translation map is represented at the top. The different shading patterns indicate alternative reading frames. Below, the newly identified spliced transcripts are displayed with the deduced novel fusion ORFs. The molecular mass of the derived protein candidates is indicated to the right in kilodaltons. (B) Analysis of the predicted protein (18 kDa) expressed from splice variant 527:4138 on a Western blot reacted with MAb B1. AAV-A1 represents the AAV2 mutant at nt 4138 (see Fig. 2D). An asterisk (*) marks the previously described 15- to 17-kDa band, which is indicative of degraded input capsids. (C) Amino acid sequence of the novel 18-kDa protein. The first 69 amino acids correspond to the N-terminal part of Rep78; the last 90 amino acids correspond to the C terminus of the VP proteins, including the epitope for MAb B1. (D) Detection of AAVwt-Flag and AAV-A1-Flag constructs on a Western blot reacted with rabbit anti-Flag upon extensive exposure. The arrow marks a faint band of ∼18 kDa consistent with the previously proposed X protein (38). (E) Displayed is the proposed X-gene-derived C terminus of the resulting protein (dark gray) before and after insertion of the Flag tag in comparison to the confirmed capsid (VP) protein sequence (light gray). The asterisk (*) denotes the stop codon in the wild-type sequence.

The predicted protein of 43-kDa translated from splice variant 3184:4138, although containing the epitopes of MAb A1 and of MAb A69 could not be identified with either antibody. To verify expression of other predicted proteins, especially from splice variant 988:4138, an in-frame Flag tag was introduced in the AAV wild-type and mutant A1 DNA sequences (Fig. 4E). Neither of the predicted 35- or 13-kDa proteins could be detected on Western blots with two alternative anti-Flag antibodies, which successfully detected the positive control (data not shown). Furthermore, the Flag tag allowed to search for the proposed X protein (18 kDa) (33) derived from the same ORF. Indeed, on very long exposures faint bands of approximately 18 kDa were detected in AAV wild-type and in AAV-A1 transfected cells, a finding consistent with the existence of the X protein (Fig. 4D).

Comparison of the splice sites to other AAV serotypes.

The biological relevance of the new splice variants was further evaluated by comparing their conservation in the genomes of alternative AAV serotypes. The genomes of AAV serotypes 1 to 9 were aligned, and the splice sites were compared to the consensus sequence of mammalian splice donor and acceptor sites (34). The well-documented splice donor site at AAV2 nt 1906 is highly conserved among all AAV serotypes but shows minor variations compared to the consensus sequence (Fig. 5A). The splice donor site at nt 527 is similarly conserved among AAV serotypes and is therefore assumed to be functional in these serotypes as well. The splice donor sites at nt 988 and 3184 are more divergent from the consensus sequence. Although the splice donor site nt 988 may be also functional in AAV3 and AAV4, the splice donor site at nt 3184 exists only in AAV2.

FIG 5.

Alignment of splice donor and acceptor sites for AAV serotypes 1 to 9. (A) Splice donor sites at nt positions 1906, −527, −988, and −3184 of AAV2 were aligned to the corresponding sites of AAV serotypes 1 to 9 and compared to the mammalian consensus sequence of splice donor sites (34). Exon-intron boundaries are indicated by dashed lines. Below the consensus sequence the frequency (%) of a particular nucleotide in average splice donor sites is displayed. White areas indicate aberrations to the consensus sequence in the given AAV serotype. (B) Splice acceptor sites at nt positions 2228, −2201, −807, and −4138 of AAV2 were aligned to the corresponding sites of AAV serotypes 1 to 9 and compared to the mammalian consensus sequence of splice acceptor sites. The display is as described for panel A. Py, pyrimidine.

The described major splice acceptor site at nt 2228 of AAV2 is highly conserved among all AAV serotypes (Fig. 5B), as is the minor splice acceptor site at nt 2201, although with reduced pyrimidine levels indicative of lower utilization rates. The newly identified splice acceptor site at nt 4138 is also conserved, with the exception of AAV3, AAV4, and AAV5. Another splice acceptor site at nt 807 validated by PCR for splice variant 527:807 is conserved among the AAV serotypes (Fig. 5B). Unfortunately, the predicted protein of 60 kDa is very similar to the molecular mass of Rep68, and we were therefore unable to differentiate it on Western blots.

DISCUSSION

We present here the first NGS analysis of the AAV2 transcriptome during latency and productive AAV replication. Each of the four data sets analyzed by Illumina RNA-Seq gave rise to more than 21 million reads, more than 95% of which could be successfully mapped to the human or viral genomes. The high quality of the raw data allowed a comprehensive and comparative analysis of the transcription profiles of AAV2 with or without Ad2 infection. The high sensitivity of RNA-Seq combined with unbiased RNA detection uncovered previously unrecognized transcripts, alternative splice variants, and a new 18-kDa protein that appears to play a role in early AAV replication. Most importantly, we are the first to document transcription from the AAV minus strand, leading to an update of the current concept of AAV gene expression.

Analysis of the AAV2 transcription profile.

As a single-stranded DNA virus AAV is dependent on the synthesis of the complementary strand to generate a double-stranded template for transcription initiation. In the absence of a helper virus, only limited Rep78/68 expression occurs from the p5 promoter (35). This is reflected in the RNA-Seq analysis of AAV2-infected cells in the absence of Ad2, where few AAV transcripts (0.04%) are found, only ∼4-fold above background level. Of these, p5 promoter-initiated transcripts are the most abundant RNAs, but transcripts originating from p19 and p40 are also detected. The findings are in line with AAV2 biology, where Rep78/68 was shown to repress AAV2 promoter activity, which is viewed as a mechanism to establish and maintain the latent state (35, 36).

In the presence of Ad2, AAV2-specific reads are increased by >200-fold, with a shift of transcription toward p40 expressed genes, which is consistent with previous reports (10). RNA-Seq also detects more-abundant transcription from the p5 promoter than from the p19 promoter, a finding that is at variance with the described 1:3 p5/p19 ratio found by RNase protection-based quantification late in AAV infection (10) and is likely explained by experimental differences. Furthermore, RNA-Seq identified reads mapping to the AAV plus strand upstream of the p5 promoter. These transcripts appear to be derived from the described transcription initiation site (Inr) located within the ITR, observed before for AAV5 and for recombinant AAV2-based vectors (11, 29, 30). In addition, RNA-Seq precisely mapped the previously defined, unique AAV2 polyadenylation site (31). Mammalian mRNAs typically cleave the RNA molecule 10 to 30 nt behind the polyadenylation signal (37). In the case of AAV2, we show that cleavage occurs exactly 21 nt behind the AAUAAA polyadenylation signal at position 4450 (2).

Previously, a candidate promoter (p81) was described on the AAV plus strand near genome position 3800 (38). We were unable to detect this promoter in silico under the stringent conditions applied. Furthermore, RNA-Seq did not detect an increase of AAV-specific reads around nt 3800, as would have been expected downstream of an active promoter. However, on Western blots with Flag-tagged constructs a faint 18-kDa protein became visible consistent with the proposed X protein (33). Its very weak expression likely explains the difficulties in confirming its existence by alternative methods but also poses questions as to its in vivo significance.

Transcription on the AAV2 minus strand.

All AAV2 transcripts described in previous reports were derived from a single DNA strand, referred to as the AAV plus strand. Therefore, the complementary AAV2 minus strand has not been considered transcribed until now (6, 7, 12). RNA-Seq analysis, however, showed that a small proportion of AAV-specific reads from AAV2/Ad2-coinfected cells (0.30%) map to the AAV minus strand. AAV minus strand reads map to a region between the left ITR and the p5 promoter and have an abundance of approximately one-third of the transcripts initiated at the p5 promoter of the AAV2 plus strand. In silico, candidate promoters are predicted on the AAV2 minus strand with transcription start sites at nt 230 and 305, respectively. Reporter gene assays revealed that the AAV2 sequence from nt 320 to 180 was sufficient to initiate transcription in minus strand direction. An extended sequence from nt 586 to 155 only marginally enhanced transcriptional activity. Obviously, the p5 promoter region is also active in opposite direction, which likely explains transcription of the AAV2 minus strand. This finding is unique for p5; transcripts in the opposite direction were not detected upstream of p19 or p40.

The finding is reminiscent of a mechanism called divergent transcription (39–42) that is related to transcription initiation in opposite directions from the same promoter. The rather short transcripts initiated at nonuniform start sites that have been described for the upstream direction in divergent transcription are in line with the heterogeneous population of AAV2 minus strand transcripts that we identified both by 5ˈRACE and by RNase protection. One assumed role for divergent transcription is to facilitate DNA unwinding, so that transcription initiation rates and expression of the downstream gene are enhanced (43). Furthermore, the short antisense transcripts are believed to regulate the transcription of downstream genes by affecting the chromatin and thereby modulating promoter activity (44). An analogous mechanism may be advantageous for the regulation of AAV p5-initiated transcripts, since these transcripts are the first AAV2 genes to be transcribed both in the presence and in the absence of adenovirus. It will be interesting to see whether divergent transcription at p5 serves as a key for the switch between the latent and the lytic AAV replication cycles.

AAV splicing.

Splicing of AAV2 transcripts is essential for the generation of Rep68 and Rep40 and for the correct stoichiometry of VP1, VP2, and VP3. The corresponding splice sites have been known for nearly 2 decades and are located at AAV2 nt 1906 (the splice donor) and nt 2201 and 2228 (the splice acceptor) (7, 8). The majority of splice products detected by RNA-Seq correspond to these sites. The ratio of usage for the established minor or major splice acceptor site, respectively, joined to the splice donor site at nt 1906, was roughly 1:7, a finding consistent with previous reports (31). The proportion of unspliced versus spliced transcripts at these sites is higher than previously reported (10) and is likely explained by the use of total cellular RNA for RNA-Seq compared to cytoplasmic RNA in previous studies. In addition, splicing enhances the export of mRNAs (45). Therefore, the total cellular RNA used here may be enriched for unspliced transcripts still located in the nucleus.

The first indication of additional splice products came from the genome coverage of total AAV2-specific reads. The read count at the far right end of the AAV genome, downstream of nt 4000, was significantly increased, despite the absence of known or predicted promoters in this region. Of several newly detected splice donor and acceptor sites utilized at variable efficiencies, the splice products involving a splice donor site at nt 527 and a splice acceptor site at nt 4138 were particularly noteworthy. Reads covering the 527:4138 splice variant were even more abundant than the known splice event 1906:2201, allowing expression of the newly described 18-kDa Rep/VP protein. Mutants involving these particular splice donor (AAV2 nt 527) or splice acceptor (AAV2 nt 4138) sites showed reduced expression levels of Rep and Cap and also some decrease in DNA replication.

Unlike the AAVs, most parvovirus genera exhibit more than one intronic sequence to generate alternative transcripts. The autonomous parvovirus minute virus of mice (MVM) possesses a large intron on the left-hand side of the genome, in addition to a short intron analogous to the single intron of the AAVs (46–48). For MVM, splicing of the large intron generates a transcript, which is translated into the nonstructural protein NS2. NS2 is required for viral growth in its natural murine host, while it is dispensable in many nonmurine cells (49). Furthermore, the human parvovirus B19 contains two introns and the simian parvovirus (SPV) contains three introns, which leads to various spliced transcripts with partially unknown protein function (50, 51). The newly identified AAV2 splice donor site at nt 527 appears to be analogous to the described splice donor site of MVM (nt 514) and the erythroviruses B19 (nt 531) and SPV (nt 279). All of these are located close to the left ends of their genomes. Similarly, the newly identified splice acceptor site at AAV2 nt 4138 may have counterparts in B19 (nt 4883) and SPV (nt 4638) at the right ends of the respective genomes. For parvovirus B19, a multiply spliced transcript involving the above-described splice sites is translated to a small protein of 11 kDa, which appears to increase viral infectivity in vivo (52) and may play a role in apoptosis and cell signaling pathways (53, 54). In cell culture the novel 18-kDa Rep/VP protein only had a minor effect on AAV replication. However, analogous to NS2 of MVM, the 18-kDa protein may have a more pronounced role for the AAV2 life cycle in vivo. Unfortunately, in vivo models of wild-type AAV replication do not exist, and in humans and primates the site(s) and mechanisms of AAV latency and productive replication are largely unknown. It will be challenging to further unravel possible functions of the novel RNAs and derived proteins during the AAV life cycle in its natural host.