Argonaute Proteins Affect siRNA Levels and Accumulation of a Novel Extrachromosomal DNA from the Dictyostelium Retrotransposon DIRS-1*

Background: Retroelements are frequently under stringent control by RNAi mechanisms. Results: Disruption of the Argonaut AgnA in Dictyostelium leads to loss of retroelement siRNAs, retroelement-encoded proteins, and accumulation of a cytoplasmic cDNA that is abolished with additional deletion of AgnB. Conclusion: Two Argonautes with different functions are involved in retroelement regulation. Significance: AgnA is required to minimize retroelement expression. The retrotransposon DIRS-1 is the most abundant retroelement in Dictyostelium discoideum and constitutes the pericentromeric heterochromatin of the six chromosomes in D. discoideum. The vast majority of cellular siRNAs is derived from DIRS-1, suggesting that the element is controlled by RNAi-related mechanisms. We investigated the role of two of the five Argonaute proteins of D. discoideum, AgnA and AgnB, in DIRS-1 silencing. Deletion of agnA resulted in the accumulation of DIRS-1 transcripts, the expression of DIRS-1-encoded proteins, and the loss of most DIRS-1-derived secondary siRNAs. Simultaneously, extrachromosomal single-stranded DIRS-1 DNA accumulated in the cytoplasm of agnA− strains. These DNA molecules appear to be products of reverse transcription and thus could represent intermediate structures before transposition. We further show that transitivity of endogenous siRNAs is impaired in agnA− strains. The deletion of agnB alone had no strong effect on DIRS-1 transposon regulation. However, in agnA−/agnB− double mutant strains strongly reduced accumulation of extrachromosomal DNA compared with the single agnA− strains was observed.

Argonaute proteins are key players in all known RNA-mediated gene silencing pathways and are highly conserved among eukaryotes. Small regulatory RNAs of 20 -30 nucleotides guide the Argonaute proteins to complementary RNA targets that lead typically to reduced expression levels of the respective genes. Silencing is achieved by mRNA degradation, translational inhibition, or by chromatin remodeling (1). The main classes of small regulatory RNAs are small interfering RNAs (siRNAs), microRNAs, and piwi-interacting RNAs (piRNAs) that differ in their biogenesis and the mode of target regulation.
The canonical triggers of siRNA-mediated gene silencing pathways are linear and perfectly base-paired dsRNA molecules that are processed by Dicer in the cytoplasm (2). Endogenous siRNAs arise for example from transposons, pericentromeric repeats, or other repetitive sequences, whereas exogenous siRNAs originate from viruses or transgenes (3).
The primary siRNA duplex is loaded into the RNA-induced silencing complex that contains an Argonaute protein at its center. After unwinding the duplex and removal of the passenger strand, the RNA-induced silencing complex assembly is completed (4). The siRNA usually guides the complex to perfectly complementary target RNAs that are cleaved by the Piwi domain and further degraded by exonucleases (5). Some siRNA classes induce transcriptional rather than posttranscriptional silencing. In plants for example, cis-acting siRNAs (casiRNAs) derived from transposons initiate heterochromatin formation at loci from which they originate (6,7). Some organisms like plants, Schizosaccharomyces pombe and Caenorhabditis elegans require RNA-dependent RNA polymerases (RdRPs) 3 for efficient RNA-mediated gene silencing on the transcriptional or posttranscriptional level.
In the case of transgenes and some viruses, plant RdRPs initiate an RNAi response by catalyzing the synthesis of doublestranded RNA molecules (8,9). This can lead to a Ͼ20-fold amplification of viral siRNAs and reduces virus accumulation efficiently (10). However, the molecular basis of RdRP target specificity is not yet well understood.
In contrast to plants, C. elegans RdRPs can directly synthesize short secondary siRNAs from an RNA template (11,12). These secondary siRNAs are of antisense polarity and loaded onto Argonaute proteins of the WAGO-clade (12). The complexes contribute to target degradation, transposon silencing, and chromosomal segregation (13). Primary siRNAs, which are produced by Dicer, and secondary siRNAs, which are individual RdRP products, differ in that the latter have a 5Ј-triphosphate (11,12). Primary siRNAs that associate with the Argonaute protein Rde-1 (14) constitute only a small amount of the siRNA pool in C. elegans and thus appear insufficient for robust gene silencing (15,16). Therefore, the main function of primary RNA-induced silencing complexes appears to guide the RdRP machinery to the target RNAs for siRNA amplification (15).
From an evolutionary point of view, the social amoeba Dictyostelium discoideum is an interesting model as it branched off from the tree of life after the plants but before the animal-fungi split (17). The amoebae are characterized by their ability to alternate between uni-and multicellularity; they grow as single cells, but upon starvation they form multicellular organisms in a developmental process that culminates in a fruiting body consisting predominantly of stalk and spore cells (18).
D. discoideum is suitable in studying RNA-mediated gene silencing pathways (19) as all RNAi key components, namely the Dicer homologues DrnA and DrnB, the RdRP homologues RrpA to RrpC, and five Argonaute proteins AgnA to AgnE, homologous to the Piwi-clade, are encoded in the genome (20). RrpC is known to be involved in the accumulation of DIRS-1 siRNAs and transposon silencing (21).
The Argonaute proteins are expressed at very different levels and at different times in development (Fig. 1A) (22). Developmental regulation is not surprising as retroelements and microRNAs, which are most likely regulated by Argonaute proteins, are differentially expressed in the developmental cycle (23,24), but so far no biological functions have been assigned to these proteins.
Transposons and repetitive elements add up to 10% of the D. discoideum genome (25). Transposable elements can be classified in long terminal repeat (LTR) retroelements, non-LTR retroelements, and DNA transposons. The most abundant transposable element in D. discoideum is the LTR retrotransposon DIRS-1 (26), which is present in 16 complete and up to 200 incomplete copies (27,28). DIRS-1 elements are only found in clusters at the subtelomeric centromeres of all six chromosomes (17,29).
Mobility of LTR retrotransposons is mediated by almost fulllength transcripts that are synthesized by RNA polymerase II in the nucleus using a promoter within the 5Ј-LTR. The LTRs flank an internal region where structural proteins and enzymes for replication and subsequent transposition are encoded. The gag ORF encodes a protein that forms cytoplasmic or viruslike particles within which reverse transcription takes place. The pol ORF encodes several enzymes, among them the reverse transcriptase (RT) and an integrase that mediates cDNA insertion into the genome (30).
DIRS-1 belongs to the DIRS group of LTR retrotransposons that have some unusual structural features compared with typical LTR elements: the three overlapping reading frames encode 1) a putative Gag protein (ORF1), 2) a tyrosine recombinase (ORF2) instead of an integrase (31), and 3) a RT, an RNase H, and a methyltransferase domain in ORF3 ( Fig. 2A) (28). The flanking LTRs are inverted (ITR), they are not absolutely identical in their sequence, and a 27-bp extension is present at the edge of the right ITR (27,31). In addition, the element harbors an internal segment close to the right ITR, which is referred to as the internal complementary region, because it contains the reverse complement of the outer nucleotides of  (22). B, relative expression of agnA and agnB mRNAs during development was analyzed by qRT-PCR in the Ax2 wild-type strain. cinD mRNA was used as a housekeeping gene for quantification. The expression level for each gene is given relative to its expression in vegetative cells (0 h ϭ 1). The data (black line) are plotted in comparison to the deep sequencing data (gray line) (22). n ϭ 2 (biological replicates). Error bars: mean with S.D. C, schematic representation of PAZ and PIWI domains of the different Argonaute proteins.
both ITRs (31). Based on these specific features, models for replication and transposition have been proposed (31, 32) but without any experimental evidence so far. According to the model, the internal complementary region is suggested to serve as a template to restore the ITR sequences during cDNA synthesis. Self-ligation of the cDNA would generate a singlestranded circular DNA. After conversion to a double-stranded circle, site-specific recombination could occur. This is supported by the fact that DIRS-1 targets pre-existing DIRS-1 elements for integration without generating a target site duplication (31).
Transcription of the 4.8-kb DIRS-1 element starts within the left ITR and terminates close to the extended sequence of the right ITR (33). DIRS-1 generates a population of several polyadenylated RNAs that accumulate during development and upon heat shock (26). Recently, a long antisense transcript driven by the right ITR has been identified in the D. discoideum wild type strain and in several mutants (21). Bidirectional transcription and Dicer cleavage of the resulting double-stranded RNAs could, therefore, be the source of primary siRNAs. Hinas et al. (23) showed that some of the DIRS-1-derived siRNAs are Dicer and some of them are RdRP products.
To elucidate the function of Argonaute proteins in general and specifically in DIRS-1 silencing, we focused on AgnA and AgnB as AgnC, -D, and -E knockouts have no obvious effect on DIRS-1 regulation. 4 We find that AgnA is involved in generation of siRNAs and in posttranscriptional regulation of DIRS-1. In contrast, AgnB appears to be an unusual positive regulator of DIRS-1 in that its knock-out abolishes accumulation of DIRS-1 cDNA in an AgnA-strain.

EXPERIMENTAL PROCEDURES
Growth and Transformants-All D. discoideum strains were grown axenically in HL5ϩ medium (Formedia) supplemented with Blasticidin S and/or Geneticin at concentrations of 10 g/ml when required. Transformation into the axenic strain Ax2 or derivates was done by electroporation as described previously (34).
For agnA gene (accession number DDB_G0276299) disruption a 2531-bp-long fragment (Ϫ222 bp upstream and 2309 bp downstream of the ATG) of the agnA gene was amplified by PCR with primers BB049 and BB050 and ligated into pGEM-4 F. Soederbom and T. Winckler, personal communication. Arrows indicate primers used to verify successful gene disruption. B, the agnA gene disruption was identified by PCR on genomic DNA using a primer set (BB059/BB060) that binds outside of the targeting fragment. Expected fragment sizes are 2645 bp for the Ax2 wild type and 3899 bp for the agnA gene disruption. C, the agnB gene disruption was identified by PCR using a primer set (Bsr G1-5A/BB048) where the forward primer binds in the Bsr cassette and the reverse primer binds downstream of the targeted fragment in the agnB terminator region. In the case of the agnB gene disruption, a fragment of 1935 bp is expected, whereas the Ax2 wild type gives no PCR product. D, removal of the Bsr cassette from the agnB knock-out strains was confirmed by PCR using a primer set (BB046/BB047) that flanks the original insertion site of the resistance cassette. After successfully removing of the Bsr cassette, the PCR reaction results in a shorter product of 2198 bp for agnBϪ rox strains compared with 3606 bp for the original agnBϪ strains. E, the agnA gene disruption in the agnB knock-out background was identified by PCR using a primer set (BB059/BB060) that binds outside of the targeted fragment. Expected fragment sizes are 2645 bp for single agnBϪ rox strains and 3899 bp for agnAϪ/BϪ strains.
T-Easy. The Blasticidin resistance (Bsr) cassette (35) was cloned in the agnA gene fragment via the endogenous BglII site. The vector for the agnB gene (accession number DDB_G0290377) disruption was generated in the following way; the whole 2963bp-long agnB gene was amplified by PCR using primers BB184 and BB185. A Bsr cassette flanked by LoxP sites was isolated from the pLPBLP vector (36) and cloned into the endogenous EcoRV site of the agnB gene. Gene disruption plasmids were linearized before transformation. Mutants were analyzed by PCR. For the agnAϪ/agnBϪ double mutant, the Bsr cassette from the agnBϪ strain was removed by transient expression of the Cre recombinase (36). Single, Blasticidin-sensitive clones were examined by PCR and then supertransformed with the agnA gene disruption construct. Single clones of the double mutant were analyzed by PCR. For the analysis of molecular phenotypes, two independent clones each were used.
Overexpressing constructs to monitor 3Ј-spreading activity in the DIRS-1 element were published previously (21). To analyze 5Ј-spreading activity, the DIRS-1 ORFs were cut out from the pJET.1.2/blunt cloning vectors (21) with BamHI and SpeI and ligated in the extrachromosomal vector pDM317 (37) that was previously linearized by BglII and SpeI. The ORFs were thus fused to an N-terminal GFP tag. Plasmids were trans-formed in the Ax2 wild type and the agnAϪ strain. Two independent transformation populations were used for transitivity analysis. Plasmids were transformed into the Ax2 wild type and the agnAϪ strain. Two independent transformation populations were used for transitivity analysis.
Developmental Time Course-The Ax2 wt strain was grown in shaking culture to a density of 2,5 ϫ 10 6 cells/ml. 5 ϫ 10 8 cells were chilled on ice for 10 min before centrifugation at 390 rcf/ 4°C for 3 min. Cells were washed 3 times in cold Sorensen phosphate buffer (2 mM Na 2 HPO 4 , 15 mM KH 2 PO 4 , pH 6.5). Cell density was adjusted to 6.8 ϫ 10 7 cells/ml with Sorensen buffer. 750 l of cells each were distributed on autoclaved nitrocellulose filters (47-mm diameter; Sartorius). Cells were incubated at 22°C in a moist chamber. After 0, 4, 8, and 12 h, cells from one filter and after 16, 20, and 24 h cells from two filters were harvested in 10 ml of phosphate buffer and centrifuged at 390 rcf and 4°C for 5 min, and RNA was extracted as described previously (38).
Oligonucleotides-DNA oligonucleotides (Invitrogen) used in this study are listed in Table 1.
Analysis of RNA-RNA was extracted from 2 ϫ 10 7 cells of axenically grown D. discoideum cells as previously described (38). For detection of small RNAs, 10 g of total RNA were TAGAGGCATCATTATTATTAACAGT PCR primer #6 Skipper A1 fw GACTCTTCGTCCCACTGAAGTCGATG Skipper amplicon A1 Skipper A1 rev GATGCAGCACCATTGTCCATAATGG separated by gel electrophoresis on 11% polyacrylamide gels (20 mM MOPS, pH 7.0, 7 M urea). RNA was transferred by semidry blot to a nylon membrane (Amersham Biosciences Hybond TM -NX) for 10 min at 20 V. It was then chemically cross-linked to the membrane as described by Pall and Hamilton (39). Blots were probed with 5Ј-radio-labeled DNA oligonucleotides as described by Hinas et al. (23). Hybridization with an oligonucleotide complementary to small nucleolar RNA (snoRNA6; DdR-6) was used as a loading control. Probes are summarized in Table 1.
For the detection of DIRS-1 transcripts, 10 g of total RNA were separated by gel electrophoresis on a 1.6% agarose gel in 1ϫ MOPS (20 mM MOPS, 5 mM NaAc, 1 mM EDTA, pH 7.0). RNA samples (2 g/l) were denatured in 2 l of 10ϫ MOPS buffer, 2 l of 37% formaldehyde, and 10 l of formamide by heating for 5 min at 70°C, 6ϫ loading buffer was added, and samples were run on an agarose gel. After capillary transfer to a nylon membrane (Amersham Biosciences Hybond TM -NX), RNA was cross-linked by UV irradiation (312 nm, 0.560 Joule/ cm 2 ). Prehybridization and hybridization were carried out in Church buffer (40) (250 mM sodium phosphate, pH 7.2, 1 mM EDTA, 7% w/v SDS, 1% w/v BSA) using T4-PNK (Thermo Scientific) 32 P-labeled oligonucleotides or random-primed-labeled PCR products using Klenow Fragment (Thermo Scientific) (41). All blots were washed as described in Hinas et al. (23). Hybridization signals were detected after exposure to an image plate and read-out on phosphorimaging (FLA-7000, Fujifilm).
Deep Sequencing of Small RNAs-Cells from axenically grown D. discoideum Ax2 and rrpCϪ and agnAϪ strains were collected by centrifugation for 3 min at 390 rcf at 4°C and washed once with 10 ml of Sorensen buffer. Small RNAs (Ͻ200 nt) were extracted from 1 ϫ 10 7 cells using the NucleoSpin miRNA Kit (Macherey-Nagel). RNA isolation was carried out according to the manufacturer's instructions. cDNA libraries of these small RNA fractions were sequenced on an Illumina platform.
Specifically, cDNA synthesis, Illumina sequencing, and read mapping were performed as follows; RNA samples were poly-A-tailed, and the 5Ј-triphosphate was converted into 5Ј-P before an RNA adaptor was ligated. A poly-T-oligo was then used as primer for cDNA synthesis, and the resulting cDNAs were PCR-amplified. The cDNA libraries were sequenced using an Illumina HiSeq 2000 machine, resulting in around 5.3, 4.9, and 5.8 million reads with sufficient quality for the Ax2, the agnAϪ, and the rrpCϪ strain, respectively. The processed (clipped and quality trimmed) reads were mapped against the fulllength DIRS-1 element SB41 (accession number M11339.1) by the short read mapper segemehl (42) and the Dictyostelium genome (DDB0169550, DDB0215151, DDB0232428, DDB0232430, DDB0232432, DDB0237465, DDB0215018, DDB0220052, DDB0232429, DDB0232431, DDB0232433). Overall, 34,152 reads of the Ax2 library, 8,350 reads of the agnAϪ library, and 3,909 reads of the rrpCϪ library were successfully aligned to the DIRS-1 element. Based on these alignments coverage plots in wiggle format were generated. For reads that map equally well to different places in the DIRS-1 sequence, proportional fractions were added to the coverage values. The data have been deposited in NCBI Gene Expression Omnibus (43) and are accessible through GEO Series accession number GSE56111 (www.ncbi.nlm.nih.gov).
Quantitative Reverse Transcription (qRT) PCR-For quantitative mRNA expression analysis, qRT-PCR was performed in an Eppendorf realplex light cycler using the SensiFAST TM SYBR No-ROX One-Step kit (Bioline) by following the manufacturer's instructions. Primer sequences are listed in Table 1.
All measurements were carried out in triplicate and averaged. Differential gene expression was calculated by the ⌬⌬CT method (44) using cinD (accession number DDB_G0273061) as the reference gene.
DNA Isolation and Southern Blot Analysis-Preparation of genomic DNA from isolated nuclei and Southern blot analysis were carried out as described before (45).
Extrachromosomal DIRS-1 DNA from 2 ϫ 10 7 axenically grown D. discoideum cells was isolated by the GeneJet TM Plasmid Miniprep kit (Thermo Scientific) according to the manufacturer's instructions. Total DNA was isolated from 2 ϫ 10 7 cells using the GeneJET Genomic DNA Purification kit (Thermo Scientific). For Southern blots under semidenaturing conditions, DNA was denatured in 2 l of 10ϫ MOPS buffer, 2 l of 37% formaldehyde, and 10 l of formamide, heated at 70°C for 5 min, and chilled on ice before loading on a 1.4% 1 x MOPS (20 mM MOPS, 5 mM NaAc, 1 mM EDTA, pH 7.0) agarose gel. The following steps were carried out as described before (45).
Isolation of DNA after Cell Fractionation-A total of 5 ϫ 10 7 cells were harvested from HL5ϩ medium and washed in icecold Sorensen buffer. Cells were collected at 390 rcf and 4°C for 5 min. The cell pellet was lysed in 5 ml of nuclear lysis buffer (50 mM HEPES, pH 7.5, 40 mM MgCl 2 , 20 mM KCl, 5% saccharose, and 1% Nonidet P-40). Nuclei were collected by centrifugation at 2680 rcf at 4°C for 10 min. The supernatant, referred to as the cytoplasmic fraction, was transferred into a new tube. Nuclei were resuspended in 5 ml of nuclear lysis buffer without Nonidet P-40. SDS was added to both fractions to a final concentration of 0.7%. After proteinase K and RNase A treatment, proteins were removed by phenol-chloroform extraction. DNA of both fractions was precipitated and resolved in 50 l of double distilled H 2 O. For further fractionation 1 ml of cytosolic supernatant was centrifuged at 5,000 ϫ g, 10,000 ϫ g, 15,000 ϫ g, and 20,000 ϫ g. DNA was prepared from pellets, and the supernatants were treated were treated as described above and analyzed by Southern blot.
Nuclease Treatment of Extrachromosomal DIRS-1 DNA-10 l of cytoplasmic DNA from the agnAϪ strain were subjected to DNase I, exonuclease III, exonuclease I, S1 nuclease and micrococcal nuclease (Thermo Fisher) in a final volume of 20 l according to the manufacturer's instructions. After heat inactivation of the respective enzyme, 10 l of each digest were used for agarose gel electrophoresis and Southern blot analysis. Extrachromosomal DIRS-1 DNA was detected by 32 P-labeled probes as indicated.
Isolation of Proteins after Cell Fractionation-1 ϫ 10 8 cells were lysed, and nuclei were pelleted as described above except that 1 tablet of protease inhibitor (complete, EDTA-free (Roche Applied Science)) was added to 10 ml of lysis buffer. The cytoplasmic supernatant was removed and precipitated by the addition of TCA to a final concentration of 10%. Proteins were col-lected by centrifugation at 15,000 rcf at 4°C for 45 min. The protein pellet was washed 4 times with 80% cold acetone. Finally the dry pellet was resuspended in 20 l of PBS and 20 l of 2ϫ Laemmli buffer (46). Nuclei were resuspended in 50 l of PBS and 50 l of 2ϫ Laemmli buffer.

RESULTS
Expression Analysis of agnA and agnB mRNA in Development-We analyzed the expression of agnA and agnB mRNAs in the wild-type strain Ax2 by qRT-PCR and compared the data with the deep sequencing-based results in dictyExpress (22) (Fig. 1, A-C). The relative expression of agnA mRNA coincided well with the published data for the Ax4 wild type. However, the developmental regulation of agnB mRNA appeared to be different, as expression did not increase after 8 h of development.
AgnA Prevents Accumulation of Transposon Transcripts-We constructed single knock-out strains of the agnA (agnAϪ) and the agnB gene (agnBϪ) as well as double knock-out strains (agnAϪ/agnBϪ) (Fig. 2). Multiple independent transformants did not show any significant mutant phenotype in growth and development under laboratory conditions (data not shown). When we analyzed DIRS-1 sense transcripts by Northern blot, we observed an accumulation of full-length ϳ4.5-kb transcripts and of shorter variants in the agnAϪ strain compared with the Ax2 wild type (Fig. 3, A and B). The agnBϪ strain showed no obvious difference in transcript levels whereas the double knock-out strain was more similar to the agnAϪ strain. The data revealed that AgnA reduces the accumulation of all DIRS-1 transcripts, including full-length in the cell whereas AgnB had no major effect.

Extrachromosomal DIRS-1 cDNA in agnA؊ Strains
It should be noted that the qRT-PCR targets several of the DIRS-1 transcripts seen in the Northern blot (Fig. 3B). Because these processing and/or degradation products have different abundance we cannot determine the absolute number of fulllength DIRS-1 RNAs.
DIRS-1 derived siRNAs Are Reduced in agnAϪ Cells-To elucidate the role of the AgnA and AgnB in DIRS-1 regulation, we analyzed the expression of DIRS-1-derived siRNAs in the respective knock-out strains. Northern blot analysis revealed a strong, up to 90% reduction in the agnAϪ and in the agnAϪ/ agnBϪ strains as compared with the Ax2 wild type. This was observed for sense and antisense DIRS-1 siRNAs as well as for different positions within the element (Fig. 4, A-C). The data indicated that AgnA is involved in the generation of siRNAs, and the reduced siRNA level upon agnA deletion could explain the increase of DIRS-1 mRNA shown in Fig. 3.
Localization of DIRS-1 siRNAs-To elucidate whether DIRS-1 siRNAs accumulate in the cytoplasm or in the nucleus, cells were fractionated, and siRNAs were analyzed in nuclear and cytoplasmic fractions. A Northern blot showed that DIRS-1 siRNAs localized predominantly in the cytoplasm (Fig.  5), suggesting a posttranscriptional mechanism of DIRS-1 silencing.
Deep Sequencing of Small RNAs Confirmed Depletion of DIRS-1 siRNAs in the agnAϪ Strain-To confirm the Northern blot data and to obtain further insight into the distribution of siRNAs along the DIRS-1 element, we performed RNA-seq of size-fractionated samples from the wild type, the agnAϪ strain, and a rrpCϪ strain, which has previously been shown to display reduced levels of DIRS-1 siRNAs (21). In the agnAϪ strain we observed an ϳ70% reduction of siRNAs along the DIRS-1 element compared with Ax2. For the rrpCϪ we found an ϳ90% reduction ( Table 2) that was in the range of the previously published data. Size-sorted reads of the small RNAs peak at 21 nt and are thus defined as siRNAs rather than random degradation products (Fig. 6A). Fig. 6B shows that siRNAs were unequally distributed along the sequence with some predominant peaks. The distribution P4 is a 32 P-labeled 21-nt oligonucleotide that detects a subset of DIRS-1 antisense siRNAs. T1 and T2 are strand-specific 32 P-labeled riboprobes, detecting sense and antisense siRNAs in a distinct 300-bp region of the DIRS-1 element. Oligonucleotide probes (P) and riboprobes (T) shown below the element are directed against sequences in antisense orientation, and probes shown above the element are directed against sense transcripts. B, representative Northern blot of DIRS-1 antisense siRNAs detected by P4. Equal loading was verified by rehybridization of the membrane with a 32 P-labeled probe directed against the snoRNA DdR-6. Numbers in brackets indicate clone numbers of independent biological replicates. Quantification of DIRS-1 siRNAs (right panel) is given relative to the DdR-6 loading control and normalized to the Ax2 wild type (ϭ100%). n ϭ 8. Error bars: mean with S.D., paired t test: p ϭ Ͻ 0.0001 (***) (Ax2 wt: agnAϪ), p ϭ Ͻ 0.0001 (***) (Ax2 wt: agnBϪ), p ϭ Ͻ 0.0001 (***) (Ax2 wt: agnAϪ/agnBϪ). C, quantification of DIRS-1 siRNAs from Northern blots using strand-specific riboprobes T1 and T2 relative to DdR-6. Error bars: mean with S.D. T1: n ϭ 4, paired t test: p ϭ 0.0025 (**) (Ax2 wt: agnAϪ). T2: n ϭ 4, paired t test: p ϭ 0.0016 (**) (Ax2 wt: agnAϪ).

FIGURE 5. Subcellular localization of DIRS-1 siRNAs.
Fractions of cytoplasmic (cyt) and nucleoplasmic (nuc) RNA were isolated from the Ax2 wild type and the agnAϪ strain and analyzed by Northern blot. The membrane was probed with the 32 P-labeled oligonucleotide P4 (see Fig. 3A) to detect DIRS-1 specific siRNAs in antisense orientation. The enrichment for nuclear and cytoplasmic RNA was shown by hybridization with a 32 P-labeled antisense oligonucleotide against snoRNA DdR-6 and against tRNA Asp . The tRNA is almost exclusively in the cytoplasm, whereas DdR-6 is enriched in the nucleus. of sense (red) and antisense (gray) siRNAs was mostly asymmetric, suggesting either different stabilities of the two strands or preferential synthesis of one strand by the RdRP. The ITR sequences and most of the gag gene (ORF1) showed almost no matching siRNAs except for a peak close to the gag 3Ј end. For the agnAϪ strain, the pattern of siRNA distribution was very similar to the wild type except that the total number was lower, thereby confirming the results of the Northern blots. Remarkably, new siRNAs that were barely or not detectable in the wild type, appeared in the 5Ј half of ORF1 and the right ITR (Fig. 6C). As previously observed by Hammann and co-workers (21), the rrpCϪ strain also displayed strongly reduced siRNA levels. The distribution of the residual DIRS-1 siRNAs was, however, dis-tinct from what we found in agnAϪ cells. Similarly, the distribution of new siRNAs in the rrpCϪ strain was most pronounced in the ends of the retroelement but different from those in the agnAϪ strain. It should be noted that independent deep sequencing experiments are difficult to compare, but our overall results for the Ax2 and the rrpCϪ match the data by Hammann and co-workers (21) very well.
AgnA Is Necessary for DIRS-1 siRNA Amplification in 3Ј and 5Ј Direction and for Posttranscriptional Silencing-In C. elegans, fungi, and plants an amplification system based on RdRPs produces double-stranded RNAs on transcripts and thus expands the pool of siRNAs (13). This phenomenon is termed transitivity. Primed synthesis by RdRPs will proceed in the 5Ј direction of a given RNA. A different mechanism of RdRP guiding by siRNAs or by proteins may also amplify siRNAs in 3Ј direction (9,15,49).
In C. elegans the Argonaute protein rde-1 has an important function in the recruitment of the amplification machinery (15). To examine if AgnA in D. discoideum acts in a similar way, we analyzed 3Ј and 5Ј transitivity in the knock-out strain and in the Ax2 wild type. If AgnA was important to recruit the amplification machinery one would expect no transitivity at all in the  respective knock-out strain. We used a set of six extrachromosomal plasmids (37) containing one of the three DIRS-1 ORFs, each which were fused to a C-terminal and to an N-terminal GFP tag (21). The constructs were transformed into the agnAϪ and into the Ax2 wild type strains. If the endogenous DIRS-1 siRNAs or any protein factor could guide an RdRP to the transgene, secondary, transitive siRNAs homologous to GFP would be expected. In fact, we detected GFP siRNAs in the Ax2 wild type with each construct (Fig. 7A) (21). Transitivity was strongest with ORF3, intermediate with ORF2, and very weak with ORF1. Effective spreading occurred in the 5Ј and 3Ј direction, and in both cases sense and antisense siRNAs were detected (Fig. 7, A and B). Unexpectedly, transitivity did not correlate with the amount of siRNAs found in the 3Ј or 5Ј parts of the different DIRS-1 ORFs (compare Fig. 6). In the agnAϪ strain, no GFP siRNAs were found (Fig. 7A), thus demonstrating the requirement of the protein for the spreading process.
As expected, mRNA of the ORF 2 and 3 transgenes was strongly reduced in the wild type but accumulated in the agnAϪ strain. No silencing effect was seen with the ORF1 transgene (Fig. 7C). The silencing effect on the mRNA level by siRNAs was further confirmed by Western blots (Fig. 7D). Although ORF1 (Gag) was equally expressed in the mutant and the wild type, ORF2 (tyrosine recombinase) and ORF3 (RT ϩ RNase H ϩ methyltransferase) fusion proteins were only detected in the agnAϪ strain. Complete silencing was observed in the wild type for the latter two even though the constructs were expressed from a strong promoter and were present in multiple copies. The agnAϪ/BϪ strain behaved exactly like the agnAϪ strain, whereas the single agnBϪ strain showed DIRS protein expression like the Ax2 wild-type strain (data not FIGURE 7. Transitivity in the Ax2 and in the agnA؊ strain. A, Northern Blots to detect GFP siRNAs generated by 3Ј and 5Ј transitivity. Radioactively labeled oligonucleotide P5 was used to detect GFP-derived sense siRNAs. Equal loading was verified by rehybridization of the membrane with a 32 P-labeled probe directed against snoRNA DdR-6. Total RNA from the Ax2 wild type and from an Ax2 strain transformed with the extrachromosomal plasmid pDM317 containing an expressed GFP tag only were loaded as controls. B, Northern blot to detect GFP sense and antisense siRNAs generated by 3Ј and 5Ј transitivity in the Ax2. The radioactively labeled probes detected either sense or antisense GFP siRNAs and are complementary to each other. Equal loading was verified by rehybridization of the membrane with a 32 P-labeled probe directed against snoRNA DdR-6. C, Northern blot analysis of DIRS-1 ORF-GFP fusion mRNAs. The membrane was probed with a 32 P-labeled oligonucleotide probe (P5). As a loading control, the membrane was rehybridized with a 32 P-labeled oligonucleotide probe directed against actin mRNA. D, cell lysates of different strains were analyzed by Western blot to detect the GFP fusions proteins. DIRS-1 ORF2 and ORF3 GFP accumulate in the agnAϪ but not in the Ax2 wild type, whereas DIRS-1 ORF1 GFP was expressed in both strains. An anti-GFP antibody was used to detect the GFP fusion proteins. To show equal loading, blots were probed for the endogenous protein Coronin with an anti-Coronin antibody.
shown). In addition to AgnA-mediated spreading, the experiment indicates that endogenous siRNAs can target a transgene in trans on the mRNA level.
Extrachromosomal DIRS-1 cDNA Is Accumulated in the agnAϪ and Reduced in the agnAϪ/agnBϪ Strains-Assuming that DIRS-1 deregulation may be sufficient for mobilization of the retroelement, we performed restriction analysis followed by Southern hybridization to detect new DIRS-1 integrations. We did not find any indication of new inserts but rather unexpectedly found additional identical hybridizing fragments in several independent strains (data not shown).
We then hybridized undigested DNA from the different mutants and the Ax2 and detected two hybridization signals, one corresponding to the genomic high molecular weight DNA and one of an apparent molecular weight of 2.8 -4 kb that was strong in agnAϪ cells, much less intense in the agnAϪ/agnBϪ strain, and not visible in the wild type and in agnBϪ cells (Fig. 8,  A and B). Because both Argonaute proteins are developmentally regulated, we examined if changes in accumulation of new extrachromosomal DIRS-1 DNA could be observed at high expression of agnB in development (8 h). But DIRS-1 extrachromosomal DNA levels did not change either in the agnAϪ or the agnA/BϪ strains (data not shown).
As determined by strand-specific probes, the extrachromosomal DNA was single-stranded and corresponded to the almost complete antisense strand of the retrotransposon (Fig.  8B and data not shown). The extrachromosomal property of the additional DIRS-1 DNA was further supported by the fact that it could be readily isolated by a kit designed for plasmid purification (data not shown).
Cell fractionation demonstrated that the extrachromosomal DNA was accumulated or exclusively present in the cytoplasm Either a mixture of oligonucleotides as indicated or P14 alone (binding to the right ITR) was used. Oligonucleotide probes (P) shown below the element are directed against antisense sequences of the DIRS-1 mRNA. RH, RNase H; MT, methyltransferase. B, DNA was isolated with the GeneJET Genomic DNA Purification kit (Thermo Scientific). The gel with undigested DNA was run under native conditions. The Southern blot was hybridized with probes P6 -P13 that detect DIRS-1 copies in the chromosome as well as extrachromosomal DIRS-1 cDNA. Extrachromosomal cDNA that corresponds to the minus strand of the transposon is found in the agnAϪ strains and is barely detectable in the agnAϪ/agnBϪ strain. C, DNA was isolated from the cytoplasm and from the nucleoplasm. Southern Blots were performed under native (upper panel) and under semi-denaturing (lower panel) conditions. The apparent molecular weight of the dominant extrachromosomal DIRS-1 DNA in the cytoplasmic fraction depended on running conditions and varied between 2800 and 4500 nt. P6 to P13 were used as probes in the left panels. The same blots were stripped and hybridized with P14 (right panels). Below the dominant almost full-length extrachromosomal DIRS-1 cDNA, shorter fragments of 400 to 500 nt became detectable. DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51 (Fig. 8C). Especially under denaturing conditions we detected an additional smear or group of predominant bands migrating at ϳ500 nt (Fig. 8C). Using various radiolabeled probes, the smaller bands were found to contain sequences of the right ITR (Fig. 8, A and C, lower panel). We do not know how the smaller molecules associate or hybridize to the long single-stranded DNA. Variations in the apparent molecular weight of the large and the small extrachromosomal DNA fragments are probably due to secondary structures, which are more or less resolved depending on subtle changes in running conditions and treatment of the DNA sample. The DNA ladder only serves as a crude ruler, not as a precise measure of molecular weight.

Extrachromosomal DIRS-1 cDNA in agnA؊ Strains
To further characterize the composition of the extrachromosomal DIRS-1 DNA, PCR analysis and digestions with different nucleases were performed (Fig. 9). Amplicons covering the predicted transcription start in the left LTR as well as the complete right LTR could generate PCR products (Fig. 9, A and B). However, oligonucleotides located in the left LTR upstream of the PCRs were performed on DNA isolated from the cytoplasmic (cyt) fraction of Ax2 and the agnAϪ strain. As a control, all PCRs were performed on genomic DNA (gDNA) as template. PCRs were performed with 28 cycles (1). PCR was carried out to analyze the presence of the left LTR. The forward primer (#1) starts at the predicted DIRS-1 transcription start. Another forward primer that binds 20 nucleotides upstream in the left LTR gave no PCR product (2). PCR was carried out to analyze the presence of the right LTR. Cytoplasmic DNA from the agnAϪ strain but not from the Ax2 wild type gave a PCR signal indicating that the complete right LTR is part of the extrachromosomal cDNA (3 and 4). PCRs were carried out to test for circularized extrachromosomal cDNA (with or without complete left LTR). No signals of the expected sizes (indicated by a star) were detectable for the cytoplasmic DNA fraction of agnAϪ. PCR products on genomic DNA are most likely derived from incomplete fragments in the DIRS-1 clusters (5). Control PCR on the skipper retrotransposon is shown. Weak signals indicate low levels of contaminating chromosomal DNA in the cytoplasmic fractions. C, left panel, cytoplasmic DNA (left) and nuclear DNA (right) from the agnAϪ strain was digested with RNase A alone or in the addition with DNase I, S1 nuclease (S1), exonuclease III (ExoIII), and micrococcal nuclease (MNase). DNase I and MNase are endonucleases that cleave single and double-stranded DNA. S1 nuclease is an endonuclease that is specific for single-stranded DNA. Exonuclease III only cleaves double-stranded DNA. Right panel, cytoplasmic and genomic DNA from the agnAϪ strain was digested with Exonuclease I (ExoI). which degrades single-stranded DNA in the 3Ј-5Ј direction. Both panels, Southern blots were performed under native conditions. Extrachromosomal DIRS-1 DNA was detected by probes P6 to P13 (see Fig. 8A).
transcription start did not give any PCR products (data not shown). Thus, the extrachromosomal cDNA covers the complete DIRS-1 element except for the left LTR sequences upstream of the predicted transcription start site. PCR reactions across the junction of the left and the right LTR that would indicate a circular DIRS-1 extrachromosomal DNA, failed (Fig. 9B).
Using various nucleases, the extrachromosomal DNA was further confirmed to be single-stranded (Fig. 9C, left panel). In particular, the single strand-specific nuclease S1 digested the extrachromosomal DNA but not the nuclear DNA, whereas the double strand-specific nuclease exonuclease III digested the chromosomal DNA but not the cytoplasmic copies. Unexpectedly, exonuclease I, which degrades single-stranded DNA in 3Ј to 5Ј direction, did not digest the extrachromosomal DNA (Fig.  9C, right panel). Although this may argue for a circular molecule, we rather assume that unusual secondary structures or modifications prevented the attack of the enzyme on the 3Ј end of a linear cDNA.
AgnA and AgnB Are Present in the Cytoplasm as Well as in the Nucleus-The observation that DIRS-1 extrachromosomal DNA accumulated to a lesser extent in strains where agnA and agnB were disrupted (compared with the single agnAϪ strain, Fig. 8B) was in agreement with the assumption that AgnB somehow promotes expression or amplification of the DIRS-1 retrotransposon. A direct effect of AgnB in DIRS-1 transcription or cDNA synthesis is feasible because the protein is localized in the nucleus as well as in the cytoplasm (Fig. 10).

DISCUSSION
We have analyzed the role of the two Argonaute proteins AgnA and AgnB in regulation of siRNA-mediated transposon silencing in D. discoideum. AgnA and AgnB are the most closely related (66.9% consensus positions) Argonautes among the five family members, and both are present in the cytoplasm and in the nucleoplasm. The strongest expression of agnA mRNA was found in vegetative cells, whereas expression of agnB peaks at 4 h in development. Differences to the previously published data could be due to (a) strain differences, (b) differences in the method (deep sequencing versus qRT-PCR), or (c) growth conditions (bacterial lawn versus axenic media) (Fig. 1). Although at the first glance a knock-out of agnB had no obvious effect on regulation of the DIRS-1 retroelement, a knock-out of agnA showed a strong reduction of DIRS-1-derived siRNAs and, consistently, a moderate accumulation of transposon transcripts. Using the DIRS-1-derived ORFs as transgenes (see below), we have shown that the silencing mechanism works in trans and that the DIRS-1-encoded proteins RT and tyrosine recombinase are not detectable in the wild type but are detectable in the agnAϪ strain. Furthermore, we could detect extrachromosomal single-stranded DIRS-1 DNA in the cytoplasm of the knock-out strain. AgnA thus plays an essential role in transposon silencing, probably by the generation of secondary siRNAs and ultimately by inhibition of RT and tyrosine recombinase expression.

Distribution of Endogenous DIRS-1 siRNAs in the Wild Type and in Different
Mutants-Analysis of the DIRS-1-derived siRNAs by deep sequencing revealed that the small RNAs are unevenly distributed along the retroelement and that only very few siRNAs match the ITRs and the 5Ј half of ORF1. The distribution of sense and antisense siRNAs is asymmetric, suggesting either differential stability of both strands or asymmetric synthesis of secondary siRNAs by RdRPs. This agrees with a previous observation by Wiegand et al. (21).
Similar to the rrpCϪ strain, siRNAs were not completely lost in the agnAϪ strain. The pattern of residual siRNA distribution was more similar to the wild type in the agnAϪ than in the rrpCϪ, but it appears that the antisense siRNAs are overall more reduced. For both mutants some new siRNAs that were not detected in the wild type appeared in the ITRs and the 5Ј half of ORF1. Again, the distribution pattern differed between the two mutants, suggesting that siRNAs have different origins and that the two proteins contribute to partially different pathways of siRNA generation. Our hypothesis is that AgnA is involved in the synthesis of secondary siRNAs by recruiting RdRPs to DIRS-1 transcripts, which would allow for 3Ј and 5Ј transitivity. Good experimental evidence for a similar model exists in C. elegans. There, transitivity occurs in a 3Ј and in a 5Ј direction, and a lack of the Argonaute protein rde-1 results in the loss of secondary siRNAs (15,50).
Is AgnA Involved in Secondary siRNA Synthesis?-To substantiate AgnA involvement in secondary siRNA synthesis, we designed an experiment to shed light on the question if endog- FIGURE 10. Subcellular localization of AgnA-3xHA and AgnB-3xHA fusion proteins. Protein was prepared from total cells (input) and from fractionated cytoplasm (cyt) and nuclei (nuc). Sources were agnBϪ cells overexpressing AgnB-3xHA and agnAϪcells overexpressing AgnA-3xHA. Proteins were separated by SDS-PAGE and blotted. HA-tagged proteins were detected with a specific antibody. Cell equivalents were loaded, and fractions were not normalized for total protein content. As a cytoplasmic marker, ␣-actinin was used. The blot indicates that both tagged proteins can be detected in the cytoplasm as well as in the nucleus. enous siRNAs can act in trans on transgenic DIRS-1 gene fusions. This was obviously the case as ORF2 and ORF3 transcripts were barely detectable in the wild type but clearly accumulated in the agnAϪ strain. Secondary siRNAs can be unambiguously identified by spreading or transitivity into adjacent regions (12). In fact, we detected siRNAs matching the GFP gene, which was fused to the C terminus and the N terminus of the ORFs. Surprisingly theses siRNAs were found in sense and antisense orientation, thus excluding a simple copying mechanism by RdRP from the mRNA. This is different to C. elegans where secondary siRNAs are only of antisense polarity with respect to the original trigger (15,50). There is so far no precedence for this observation, and the current model of 3Ј spreading has to be adjusted. Silencing of the fusion genes was even more obvious on the protein level than on the RNA level. Consistent with the hypothesis that AgnA was involved in secondary siRNA generation, no GFP-derived siRNAs and no transgene silencing were observed in the agnAϪ strain. However, we cannot clearly distinguish if this was due to the overall reduction of siRNAs in the mutant or to its inability to recruit the amplification machinery. One may assume that the strength of transitivity depended on the amount of endogenous siRNAs at the ends of the ORFs as these siRNAs could guide an RdRP close to the fusion point and generate secondary siRNAs of the 3Ј and 5Ј tag. But there was no correlation between transitivity and the number of 3Ј or 5Ј siRNAs. Remarkably, ORF1 GFP showed essentially no transitivity even though there was a substantial amount of siRNAs close to the 3Ј end. These findings and the fact that we could not see any transitivity in the agnAϪ strain even though 30% of DIRS-1 siRNAs were still present suggested that AgnA recruits RrpC for siRNA amplification and that the residual siRNAs are not sufficient. Based on the transgene experiments, one could assume that the ORF1-encoded Gag protein is also expressed in wild type cells but insufficient to assemble functional virus like particles in which reverse transcription is thought to take place (30). Gag may either be insufficient to form virus-like particles or, even if virus-like particles were formed, no RT is expressed in wt cells to promote reverse transcription.
Is DIRS-1 on the Way to Transpose?-Expression of the reverse transcriptase and the tyrosine recombinase was strongly enhanced in the agnAϪ strain (Fig. 6D). Even though this could only be shown with the respective transgenes, we believe that the endogenous RT and tyrosine recombinase proteins are also up-regulated in the mutant. With the disruption of agnA, proliferation of the retroelement proceeds further than in the wild type. However, we could not show enhanced transposition in the mutant. The accumulated single-stranded DNA in the cytoplasm is most likely a product of the DIRS-1encoded reverse transcriptase and an intermediate on the way to transposition, but there appears to be another safeguard mechanism that prevents completion of this pathway.
According to hybridization and PCR analysis with different oligonucleotides, this long DNA is not complete but is only an almost full-length product. It does not contain the left ITR up to the previously suggested transcription start site at approximately nucleotide 300. The cDNA thus covers only ϳ4500 nt of the 4800 nt. These cytoplasmic molecules have not been observed before in D. discoideum. Notably, a second, less distinct DNA family of apparent 500 nt is associated with the almost full-length cDNA. The origin of these molecules is unknown; they could be abortive reverse transcripts or processing products. Southern blot experiments showed that they harbor sequences of the right LTR and were also single-stranded. Because no sequence complementarity between the small and the long cDNAs is obvious; protein-aided association of partially complementary regions may be possible to support the generation of dsDNA of a complete DIRS-1 element. The theoretical model for the amplification of the DIRS-1 retroelement proposed by Cappello et al. (31) does not include such short molecules. Poulter and Goodwin (32) proposed a model where a second, possibly smaller DNA serves as an annealing platform for reconstruction of the complete element and for synthesis of dsDNA. However this model is only suggested for retroelements with direct repeats.
Accumulation of DIRS-1 cDNA in the agnAϪ strain may be due to two interrelated reasons; DIRS-1-derived full-length RNA and processing or degradation products accumulated 4 -8-fold in the agnAϪ strain compared with the wild type, and relative accumulation levels depended on the position of the amplicon that was used for qRT-PCR. This rather moderate increase in transcripts was by far surpassed by the accumulation of the single-stranded cDNA molecule in the agnAϪ mutant. A more dramatic effect was also seen on protein expression in the transitivity approach (Fig. 7) where ORF 2 and ORF 3 proteins were only detectable in the agnAϪ strain, whereas the ORF 1 protein was well translated in the wild type. In the end it is most likely the expression of the DIRS-1 proteins that mediates the accumulation of DIRS-1 cDNA. In the wild type, protein expression is inhibited or reduced by low levels of mRNA due to high levels of secondary siRNAs. However, we cannot exclude that siRNAs have an additional effect on translation.
The Role of AgnB in DIRS-1 Regulation-The single agnB disruption had only minor effects on DIRS-1 regulation. Upon closer inspection, a reduction in DIRS-1 transcripts and DIRS-1-derived siRNAs was observed, but these effects were relatively mild. DIRS-1 transcripts were reduced in comparison to the single agnAϪ strain but still significantly higher than in the wild type. However, there was a strong effect in the double mutant where the accumulation of extrachromosomal DNA was almost abolished. AgnB thus has an apparently positive effect on the retroelement, or it prevents further productive processing of the cDNA on the way to transposition. For retroviruses it has been shown that Argonaute proteins promote their amplification: human Ago2 may facilitate the assembly of retroviral particles, but detailed mechanisms are not well understood (51).
Both Argonaute proteins that we analyzed have substantial effects on silencing of the retroelement DIRS-1 and appear to function on different levels. AgnA has a more canonical role in the generation of siRNAs, probably in the cytoplasm where most of DIRS-1 siRNAs are present. AgnB, however, is to our knowledge the first Argonaute with an apparently positive effect on a retroelement.
Further investigations are required to unravel the role of AgnB in regulation. Similarly, the exact composition of the extrachromosomal single-stranded DNA that is accumulated in the agnAϪ strain requires further analysis. Because we did not observe a substantial mobilization of the retroelement, it will be of interest to elucidate the additional safeguard mechanism(s) that inhibits full activation.