RNA interference (RNAi) regulates gene expression by sequence-specific destruction of RNA. It acts as a defense mechanism against viruses and represses the expression of transposable elements (TEs) and some endogenous genes. We report that mutations and transgene constructs that condition cell death suppress RNA interference in adjacent cells in Drosophila melanogaster. The reversal of RNAi is effective for both the white (w) eye color gene and green fluorescent protein (GFP), indicating the generality of the inhibition. Antiapoptotic transgenes that reverse cell death will also reverse the inhibition of RNAi. Using GFP and a low level of cell death produced by a heat shock-head involution defective (hs-hid) transgene, the inhibition appears to occur by blocking the conversion of double-stranded RNA (dsRNA) to short interfering RNA (siRNA). We also demonstrate that the mus308 gene and endogenous transposable elements, which are both regularly silenced by RNAi, are increased in expression and accompanied by a reduced level of siRNA, when cell death occurs. The finding that chronic ectopic cell death affects RNAi is critical for an understanding of the application of the technique in basic and applied studies. These results also suggest that developmental perturbations, disease states, or environmental insults that cause ectopic cell death would alter transposon and gene expression patterns in the organism by the inhibition of small RNA silencing processes.

RNA interference (RNAi) uses double-stranded RNA (dsRNA) to target the destruction of the homologous messenger RNAs via short interfering RNAs (siRNA) (Zamore and Haley 2005). In Drosophila, dsRNA can be generated by transcribing a sequence from both directions or from inverted repeats. The ribonuclease Dicer-2 processes the dsRNA into ∼21-nt siRNA. Single-stranded siRNA is then incorporated into the RNA-interference silencing complex (RISC) and the latter recognizes and cleaves a target RNA, which perfectly or nearly perfectly matches the siRNA. This targeted degradation mechanism has been demonstrated as an immunity defense against viruses that have dsRNA in their life cycle (Li and Ding 2005). Discovery of endogenous siRNAs homologous to transposable elements (TEs) also suggests this siRNA pathway plays a role in suppressing transposition (Chung et al. 2008; Czech et al. 2008; Ghildiyal et al. 2008; Kawamura et al. 2008; Tam et al. 2008; Watanabe et al. 2008) in addition to the PIWI-associated small RNA (piRNA) pathway (Aravin et al. 2007; Brennecke et al. 2007).

Cell death is a central part of the immune system of many multicellular organisms. Cells infected with pathogens can trigger the apoptotic pathway and cell death to prevent the pathogens from spreading (Postigo and Ferrer 2010). Such a mechanism is also used to remove damaged cells or extra cells unused in tissue formation. When diseased or damaged cells are removed, proliferation signals are generated from the dying cells to the adjacent cells to stimulate compensatory cell divisions (Huh et al. 2004; Pérez-Garijo et al. 2004; Ryoo et al. 2004).

Deep sequencing of endogenous siRNAs reveals that a significant proportion of them are derived from transcription of certain sequences from opposite directions, of hairpin-structured sequences, or of homologous sequences (e.g., genes and their pseudogenes) from which complementary sense and antisense sequences are generated (Chung et al. 2008; Czech et al. 2008; Ghildiyal et al. 2008; Kawamura et al. 2008; Okamura et al. 2008; Tam et al. 2008; Watanabe et al. 2008). These siRNAs have been demonstrated to match important genes and their expression is repressed in oocytes of mice and somatic cells of flies, thus implying a genome-wide regulation role by RNAi. Among these genes, the endogenous mus308 gene in Drosophila increases in expression when RNAi is compromised (Czech et al. 2008; Okamura et al. 2008).

Here we report that cell death in Drosophila, induced either by apoptotic genes or caused by developmental defects, inhibits RNAi. We also demonstrate that cell death suppresses the silencing of mus308 and transposable elements. The increased expression is accompanied by siRNA reduction and dsRNA accumulation, suggesting that the processing of dsRNA to siRNA is impaired.

Materials and Methods

Strains, genetic tests, and microscopy

All eye images were obtained using a dissecting microscope with ×4 magnification with an attached digital camera. Ten to 30 flies of the same genotype were observed and representative flies photographed.

The RNAi strains with homozygous GMR-wIR insertions on chromosomes X or 3 (Lee et al. 2004) were kindly provided by R. Carthew, Northwestern University, Evanston, IL. These strains were crossed to the multiple balancer strain Basc/Basc; In(2LR)SM1, al2 Cy cn2 sp2/In(2LR)bwV1, ds33K dpOV bwV1; In(3LR)Ubx130, Ubx130 es/In(3LR)C, Sb; svspa-pol. The eye phenotype was recorded in the F1 with Bar and in the F2 with svspa-pol/svspa-pol. Also a male with GMR-wIR and B was recovered. A GMR-wIR B strain was generated and tested to confirm that the X chromosome carried w+, B, and GMR-wIR by recombination with a regular w X chromosome. The Bar effect in the males was also observed by using another strain y/BsY; tra2ts2 bw/CyO (From B. Taylor at Oregon State University, Corvallis, OR), which carries the Bar mutation on the Y chromosome.

The GMR-wIR / B larvae were treated with acetamine as described (Fristrom 1972).

The RNAi stocks were crossed to the following strains Pr Dr/TM3; Sco/In(2L)Cyt, Cy amosRoi-1; P{w+mc.hs = GawB} elavc155 w* P{ry+t7.2 = neo FRT} 19A; Bc1 EgfrE1/CyO and bic L2/CyO. The eye phenotype of the F1 with Dr, Roi (amosRoi-1), Egfr or L combined with GMR-wIR was documented. The strain w+; Gla/SM6a; TM3, Ser/MKRS was crossed to a strain carrying GMR-wIR on the X to produce the strain w+ GMR-wIR; Gla/SM6a; TM3, Ser/MKRS. The wgGla-1 (subsequently referred to as Gla) effect on RNAi was examined in the F1. This new strain was crossed to the stocks of ro and hhbar3 and the F2 with heterozygous or homozygous mutations and wIR was examined with or without Gla.

The GMR promoting cell death transgenes including the five inducers and the two inhibitors (Table S1) (Xu et al. 2003) were kindly provided by B. Hay, California Institute of Technology, Pasadena, CA. These strains, GMR-diap1 on the X, GMR-p35 on the X, GMR-hid on 2, GMR-grim/TM3, Sb, GMR-rpr/TM3, Sb, GMR-strica/CyO, and GMR-ttk88-myc(19)/CyO, were crossed to the RNAi strains and the F1 phenotype was recorded (all the transgenes carry a w+mc marker).

To examine the combination of cell death inducers and inhibitors, the multiple balancer stock with GMR-wIR described above was first mated to males of inducer strains. The F1 males carrying the respective inducer transgene and GMR-wIR were then crossed to virgins of the inhibitor strains. The phenotype of the female offspring of these crosses was then analyzed and documented.

The symmetrically transcribing w RNAi strains w1118; SympUAST-w#8 and w1118; SympUAST-w#23/CyO, and the long stem-loop ones w1118; pUAST-IRsp-w#32/CyO and w1118; pUAST-IRsp-w#41 were kindly provided by E. Giordano (Giordano et al. 2002), Università di Napoli, Naples, Italy. These strains were first crossed to w+; Gla/SM6a; TM3, Ser/MKRS to replace the w mutant gene on the X chromosome and to balance the transgenes on the second chromosome and double balance the third chromosome. To test whether B affects RNAi in these strains, virgins of the multibalancer strain Basc/Basc; In(2LR)SM1, al2 Cy cn2 sp2/In(2LR)bwV1, ds33K dpOV bwV1; In(3LR)Ubx130, Ubx130 es/In(3LR)C, Sb; spapol spapol were crossed to males of y w/y+ w; UAS-EGFP Tub-Gal4/TM6b, and then the Stubble F1 males with Tub-Gal4 and Basc were crossed to the balanced transgenic strains. To simplify the genetic tests, GMR- and act5c-Gal4 trangenes on the second chromosome were chosen to recombine together in one chromosome with the UAS w RNAi transgenes. The new strains were then crossed with the cell death strains and F1 phenotypes were assayed.

The UAS-wRNAiDS (on the third chromosome) strain (Kalidas and Smith 2002) and the GMR-Gal4 or act5c-Gal4 (on the second chromosome) strain were first crossed to the multibalancer strain w+; Gla/SM6a; TM3, Ser/MKRS. Then the F1 flies were crossed together to produce a strain containing both transgenes balanced on the second and third chromosomes. This strain was used to cross with the cell death strains and offspring phenotype was documented.

The EGFP RNAi strain w[1118]; P{w[+mC]=UAS-EGFP.dsRNA.R}142 (Roignant et al. 2003) was crossed with the GFP strain y w/w+; UAS-EGFP Tub-Gal4/TM6b, and the F1 females were then crossed to the multibalancer strain y w67c23; Gla/SM6a; TM3, Ser/MKRS. The flies with recombined transgenes together on the third chromosome were selected by additive eye color (the transgenes from both original strains are marked by mini-white) and weak body fluorescence (EGFP silenced). The strain was confirmed to contain the three transgenes by combining the transgene chromosome (the third) with a wild-type one in females and finding that the offspring segregated unrepressed GFP flies due to recombination. In parallel, a strain v; bw; TM3, Ser/MKRS was produced. This strain was crossed to the GFP RNAi strain to make the strain v; bw; Tub-Gal4 UAS-EGFP UAS-EGFPir/MKRS. To combine v, bw, and a cell death gene in a strain, the following approach was taken. For B, balancer FM7a was used because it contains both B and v; and for cell death genes on the second chromosome, they were recombined together with bw. We also balanced the third chromosome constructs with MKRS or TM3, Ser. Finally, the EGFP RNAi strain was crossed to the cell death strains, individually, and the phenotype of the offspring was digitally photographed using the GFP dissecting microscope in the University of Missouri (MU) cytology facility.

The transgenic hs-hid stock strains from Bloomington Stock Center (y1 w*; Bl1 L2/CyO, P{w[+mC]=hs-hid}4 on chromosome 2 and y1 w*; Pr1 Dr1/TM3, P{w[+mC]=hs-hid}14, Sb1 on chromosome 3) were crossed to the strain GMR-wIR to observe color restoration in the eyes. The EGFP RNAi strain v; bw; Tub-Gal4 UAS-EGFP UAS-EGFPir/MKRS was crossed to the line with the third chromosome insertion or first crossed to the multibalancer strain w+; Gla/SM6a; TM3, Ser/MKRS and then crossed to the second chromosome insertion to generate the hs-hid and EGFP RNAi strains. The flies were raised at 18°. The third instar larvae were collected for dissection. For midgut analysis, larvae were collected before emerging from the food to avoid degradation that occurs at later stages (Jiang et al. 1997). Larval GFP was observed with the GFP dissecting microscope in the MU cytology facility. Dissected tissues were immediately mounted on slides with fluorescence mounting medium without covering and photographed under a Zeiss Universal microscope with a MagnaFire cooled charge-coupled device camera.

Multiple GMR-wIR strains were produced as follows. Both the GMR-wIR B strain and the GMR-wIR (on chromosome 3) strain were mated to the strain w+ GMR-wIR; Gla/SM6a; TM3, Ser/MKRS and the offspring was mated with each other to obtain males with the genotype GMR-wIR B; GMR-wIR/TM3, Ser (or MKRS). These males were mated to females of GMR-wIR (on chromosome 3) and the phenotypes of 1 (from F1), 2 and 3 (from F2) copies of GMR-wIR with B were photographed.

For mitotic recombination in the eyes, strains were generated with a cell death inducing transgene (GMR-hid, -ttk, -grim, or -rpr) on the second or third chromosome recombined in the same chromosome with a corresponding neoFRT transgene (Xu and Rubin 1993). The original neoFRT strains were modified to maintain w+ on the X chromosome and ey > FLP1 on the other autosome. These strains were then crossed to female w+ GMR-wIR; Gla/SM6a; TM3, Ser/MKRS flies and male offspring were collected to mate with the neoFRT cell death strains. The phenotypes of the offspring were documented. For simplicity, GMR-ttk eyes with clear boundaries between GMR-ttk and wild-type cell clones were chosen for illustration.

Acridine orange staining, Northern blotting, and RT–PCR

The transgenic hs-hid strains from the Bloomington Stock Center (y w*; Bl L2/CyO, P{w+mC=hs-hid}4 and y w*; Pr Dr/TM3, P{w+mC=hs-hid}14, Sb) and Canton-S were used for acridine orange (AO) staining. The larvae were raised at 18° and collected between the second instar to early third instar. Late third instar larvae were avoided because of normal degradation of the midguts (Jiang et al. 1997). Midguts were dissected and stained with AO (Jiang et al. 1997) and observed for green fluorescence under a ×4 objective lens of a Zeiss Universal microscope with a MagnaFire cooled charge-coupled device camera. For positive controls, larvae at the same age were used after heat shock at 37° in a water bath for 30 min and then allowed to recover at 18° for >1 hr. More than 10 pairs of Canton-S and hs-hid larvae and three heat shocked hs-hid individuals were analyzed.

A total of 50–80 late third instar larvae were collected for TRIzol RNA isolation according to the manufacturer’s protocol (Gibco BRL Life Technologies). If the RNA was for siRNA preparation, the RNA precipitation was placed on dry ice for 20 min before centrifugation to ensure that the small RNA was precipitated. Isolated RNA was dissolved in formamide at 55–65° and stored at −20°. RNA concentration was measured by 10- to 100-fold water dilution with a NanoDrop ND-1000 spectrophotometer.

EGFP mRNA and dsRNA levels were analyzed by Northern blotting according to Auger et al. (2005). Twenty micrograms of total RNA of Tub-Gal4 UAS-EGFP/TM6b, Tb (EGFP) and Canton-S (CS) and 5, 10, and 20 μg of total RNA of Tub-Gal4 UAS-EGFP UAS-EGFPir/MKRS (EGFPir) and Tub-Gal4 UAS-EGFP UAS-EGFPir/TM3, hs-hid Sb (hs-hid; EGFPir) were loaded in the gel. EGFP coding sequence amplified by primers WXo28 (AAG GGC GAG GAG CTG TTC AC) and WXo29 (CCA TGT GAT CGC GCT TCT CG) was cloned into pGEM-T Easy Vector (Promega) and sense and antisense EGFP probes were transcribed from the plasmid DNA using MaxiScript kit (Ambion). α-Tubulin exon 2 sequence was amplified by primers WXo67 (TCT ATC CAT GTT GGT CAG GC) and WXo68 (GGT AGT TGA TGC CAA CCT TG) and cloned into pGEM-T Easy Vector. Antisense probe was then transcribed to detect α-tubulin mRNA on the striped membranes as input controls.

EGFP siRNA detection by Northern analysis was performed according to Pal-Bhadra et al. (2002). Ten micrograms of total small RNA of Tub-Gal4 UAS-EGFP/TM6b, Tb (EGFP) and 2.5, 5, and 10 μg of total small RNA of Tub-Gal4 UAS-EGFP UAS-EGFPir/MKRS (EGFPir) and Tub-Gal4 UAS-EGFP UAS-EGFPir/TM3, hs-hid Sb (hs-hid; EGFPir) were loaded in the gel. Sense and antisense EGFP probes were fragmented before hybridization. The 5S rRNA sequence was amplified by WXo89 [TAA TAC GAC TCA CTA TAG GGc caa caa cac gcg gtg t, T7 promoter attached (upper case)] and WXo90 (gcc aac gac cat acc acg c) from genomic DNA and then purified to transcribe an antisense RNA probe, which was used to detect 5S rRNA level as input controls. Locked nucleic acid (LNA) probes against miRNA and endogenous siRNAs were designed and synthesized by Exiqon with optimal Tm for Northern blots.

All the Northern blotting experiments were carried out at least three times using RNA samples isolated separately. To quantify the difference of the EGFP mRNA, dsRNA, and siRNA with or without hs-hid, the exposed X-ray films were scanned and analyzed by the software Image Gauge (v 3.3) or exposed and scanned by a phosphorimager (Fujifilm Global) and analyzed by Multi Gauge. The fold difference between the death and no death samples was calculated in the formula: (Sh/Lh)/(S/L), in which Sh and S are the detected strength from a Northern blot for hid and no hid samples, respectively; Lh and L are the corresponding loading controls. When a sample with the detected signal strength was significantly less than the double amount of the twofold diluted sample, it was discarded as overexposed. Samples of death and no death were paired for comparison with the same amount of RNA transferred on the same membrane.

Late third instar “hs-hid; EGFPir” larvae raised at 18° were collected and heat shocked at 37° in a water bath for 30 min and allowed to recover at 18° for 1 hr and then frozen at −80°. Total RNA isolated from “EGFPir” larvae and heat shocked and not shocked “hs-hid; EGFPir” larvae was first precipitated from the formamide solution (25–40 μg RNA in 5 μl plus 15 μl 100% ethanol and 0.5 μl 5 M NaCl). The RNA was dissolved in 100 μl of nuclease free water and the concentration was measured with a NanoDrop spectrophotometer. A total of 10 μg RNA was treated with RQ1 DNase (Promega). First-strand cDNA was synthesized with SuperScript III reverse transcriptase (Invitrogen) and poly(T) primer. Primer WXo93 (aac aag cgc agc tga aca ag) were designed matching the UTS flanking sequence subsequent to the hsp70 promoter in the hs-hid transgene (Grether et al. 1995). Primer WXo91 (ga tga act cga cgc tac gtc) matches the coding sequence of hid exon 1. This pair of primers was used to amplify specifically the hs-hid transgene cDNA. PCR was terminated after 25, 30, and 35 cycles, respectively. PCR products were analyzed after gel electrophoresis and ethidium bromide staining.

Semiquantitative and real-time PCR

To study whether the expression of endogenous genes and transposable elements was changed, RNAs and cDNAs were prepared as described above from the larvae of “EGFPir” and “hs-hid; EGFPir” or Canton-S and hs-hid strains from the Bloomington Drosophila Stock Center.

At least two pairs of primers were designed for each mRNA matching close to the 5′ end to avoid possible detection of degraded 3′ fragments with a poly(A) tail, which can be reverse transcribed by a poly(T) primer. For semiquantitative PCR, a series of dilutions of the cDNA (one-, two-, and fourfold) was used as templates for PCR to ensure that the stop point was within the range of exponential amplification. Consensus results were observed for different primers and for technical and biological replications.

To confirm the mus308 expression change, real-time PCR was performed in an ABI 7300 system using the manufactor’s supplied Syber Green reaction master mix and the results were analyzed by the provided software. Multiple primer pairs were tested for the amplification efficiency and the most efficient ones (5′-CTTAATGGCGCTGGAGAAAG-3′ and 5′-GGCCTGATTCACTTCGTAGG-3′) were used. A cDNA concentration range was used for detecting differences in a dilution series. Three biological replications were analyzed.

Sequencing and analysis of small RNAs

Total RNA from “EGFPir” and “EGFPir hs-hid” third instar larvae was first size separated by polyacrymide gel electrophoresis and 15–40 nt molecules were collected for constructing a cDNA library and then sequenced by a Solexa system in the MU DNA core facility. Primers (5′-CGACAGGTTCAGAGTTCTACAGTCCGACGATC (N)n TCGTATGCCGTCTTCTGCTTG-3′) from both ends in the raw reads were removed and only the sequences with sizes between 19 and 31 were selected. Trimming was performed at either or both ends of the read for all or part of the primer/adapter sequence above. If the 3′ end of the read matched all or part of the left end of the right primer then that matched part was removed from the read. Similarly, if the 5′ end of the read matched all or part of the right end of the left primer then that matched part was removed from the read. No mismatches were allowed in any of the trimming.

The reads for each sample are mapped to Drosophila known miRNAs (from miRBase, http://microrna.sanger.ac.uk/sequences/index.shtml), Drosophila transposons (from flybase, http://flybase.org/) using BLAT (James 2002) with which the parameters used allow maximum sensitivity. After BLAT search, however, more stringent criteria were used to parse the results to retain only the highly confident alignments. For miRNAs, only exact matches (same query/target sequence length and 100% sequence alignment identity) were counted for normalization. For si/piRNAs mapped to transposons, only fully mapped sequences with 100% identity were counted. To further verify the si/piRNAs, annotated pi/siRNAs were downloaded from GEO datasets (GPL4738 and GPL6664) (Brennecke et al. 2007; Ghildiyal et al. 2008) and all cleaned sequences were mapped on them with BLAT (only the exact matches were counted).


Eye mutations producing cell death restore eye color otherwise reduced by white RNAi

The Drosophila compound eye develops from an eye-antennal imaginal disc, which originates from a cluster of ectodermal cells invaginated during embryogenesis. The cells are proliferated through the first, second, and early third larval instars, but remain undifferentiated. Morphogenesis and pattern formation begin with the morphogenetic furrow (MF) sweeping from the posterior to the anterior of the eye disc in the mid- and late third instar and continuing through the late pupal stage (Reifegerste and Moses 1999). The morphogenetic furrow defines a clear boundary at which the cells undergo differentiation (Heberlein et al. 1993). Undifferentiated sections in mutants with impaired MF movement undergo cell death at the late third instar and are not included in the adult eye (Fristrom 1969; Heberlein et al. 1993; Mozer 2001). Cell death also regularly occurs from 24 to 36 hr after pupal formation to remove the extra accessory cells, namely, the secondary and tertiary pigment cells (Brachmann and Cagan 2003).

An RNAi construct (GMR-wIR) (supporting information, Figure S1) repressing the expression of the white (w) gene reduces the eye color from brick red to light yellow (Figure 1A) (Lee et al. 2004). In the process of manipulation of this construct, we crossed the RNAi stock to laboratory strains carrying dominant or recessive marker genes causing eye morphological abnormalities. Surprisingly, the normal red eye color was partially restored in Bar (B) and Drop (Dr) flies, and weakly restored in the presence of the Gla and sparkling-poliert (svspa-pol) mutations (Figure 1A and Table S1). Because the eye color change occurs in living cells and is adjacent to the missing portions in the B mutant eyes, we hypothesized that the inhibition of RNAi is caused by adjacent tissue that has undergone cell death.

Figure 1

Cell death-dependent inhibition of RNAi silencing of the white gene. (A) At top left, a wild-type (WT) eye with w+ is red. The yellow eye (second from top left) indicates that w is silenced by RNAi when the transgene GMR-wIR is present in the genome. Heterozygotes of the noted dominant mutations and the homozygote for the recessive genes were combined with a copy of GMR-wIR. Silencing is partially reversed in eye mutations B, Dr, Gla, ro, and hhbar3, but not in the mutation Roi (amosRoi-1). Acetamine treatment of the Bar mutant [B (acetamine)] reduces cell death in the eyes accompanied by reduced color restoration. (B) Induced cell death in the eyes by ectopic expression of grim, hid, rpr, strica, and ttk (all driven by the GMR promoter) caused reversal of RNAi. The first row shows the phenotype of the noted cell death transgenes with wild type w+. The second row shows the transgenes combined with GMR-wIR. The red color indicates inhibition of RNAi. The third and fourth rows show the cell death transgenes and GMR-wIR combined with two different inhibitors of cell death as noted. To the extent that cell death is reversed, RNAi is restored. (C) EGFP RNAi is inhibited by cell death. Cell death induced by ectopically expressed grim, strica, and ttk restored the fluorescence to most regions, if not the whole eye. In each comparison, control eyes with GFP silenced but without induced cell death are shown on the upper left. Bars, 0.1 mm.

Bar and Drop are gain-of-function mutants, ectopically expressing homeobox genes BarH1/BarH2 or muscle segment homeobox (msh), respectively, in the eye disc. Bar prematurely stops the MF movement beginning from the dorsal–ventral center and spreading laterally (Heberlein et al. 1993). Cell death occurs anterior to the arrested MF (Fristrom 1969) causing the adult phenotype. In the w RNAi and B flies, the restored eye color gradually diminishes posteriorly and laterally (Figure 1A), suggesting it may be caused by cell death in the adjacent region.

When B mutant larvae are raised on food containing 1–2% acetamine, the mutant effect can be largely (but not completely) reversed by prolonging development, which significantly reduces cell death in the late third instar (Fristrom 1972). When female animals with B and GMR-wIR on the X chromosomes were treated with acetamine, reversal of RNAi is diminished as expected (Figure 1A).

The hedgehog mutation bar3 (hhbar3), which also arrests the MF and causes a similar eye phenotype (Heberlein et al. 1993), likewise caused inhibition of w RNAi and restored the color around the indented anterior portion of the eye but appearing more focused than with B (Figure 1A). In the Drop mutants, the MF is initiated but defective in movement (Heberlein et al. 1993). Cells in the MF undergo normal development and grow to a few rows of regular ommatidia. The remaining cells in the anterior region are maintained in an undifferentiated state and degraded later by cell death (Mozer 2001). In the w RNAi strain, the differentiated ommatidia showed restored red color (Figure 1A). As detailed in Table S1, the Gla, ro (rough), and svspa-pol mutations also cause cell death and inhibit RNAi weakly, whereas the Egfr (Epidermal growth factor receptor), and Roi (amosRoi-1) mutations, which do not cause cell death, do not reverse RNAi (Figure 1A). Thus, the RNAi inhibition is only associated with mutations that condition cell death and is not a secondary consequence of eye morphological changes. The L2 (Lobe) mutant induces abnormal cell death before the initiation of differentiation (Singh et al. 2006) and does not affect w silencing, suggesting the duration of the signal from cell death is temporally limited.

To rule out the possibility that cell death affects w expression as the basis of the above results, we crossed Dr or B to hypomorphic w mutants including point mutations wa2 and wa3 and insertion mutations, w-blood (wbl) and w-apricot (wa) (Zachar and Bingham 1982) and found no impact (Figure S2). These results indicate that modulation of the white gene itself is not the basis for the suppression. Another GMR-wIR transgene inserted in chromosome 3, was tested with B and showed similar inhibition of RNAi (data not shown), indicating that the location of the RNAi transgene also does not affect the outcome.

Proapoptotic transgenes inhibit w RNAi depending on cell death

To test further the role of cell death signaling for the inhibition of RNAi, cell death was induced by overexpression of apoptosis promoting genes. The proapoptotic genes grim, head involution defective (hid, or Wrinkled, W) and reaper (rpr), the caspase gene strica (or dream) and the transcription factor tramtrack (ttk) have been shown to cause extensive cell death and irregular eye phenotypes when ectopically expressed by the GMR promoter (Figure 1B) (for example, Xu et al. 2003). When w RNAi was also introduced into these strains, the expression of the w gene can be assayed in the remaining cells and was partially restored, as indicated by dispersed red spots or a centered dorsal–ventral strip of increased color in female ttk flies, without exception, when cell death occurred (Figure 1B, second row).

The double homozygous mutations of vermilion (v) and brown (bw) are epistatic to w+ and show pale-yellow eyes, thus resembling the w RNAi phenotype. When the cell death inducer transgenes were combined with these mutations, no pigment accumulation was found (Figure S3). This control indicates that the increase in pigment with the cell death transgenes has white RNAi as a target rather than being a consequence of altered eye morphology (e.g., closer arrangement of pigment cells).

The possibility that the antiapoptotic transgenes restore RNAi directly can be excluded by the following experiments. The cell death phenotypes can be variably suppressed, except for the overexpression of ttk, by simultaneous expression of the apoptosis inhibitors Drosophila inhibitor of apoptosis 1 (diap1 or thread, th) or baculovirus p35 (Figure 1B) (Xu et al. 2003; Hay et al. 2004). When cell death is reversed fully (as indicated by the eye shape and ommatidia arrangement returning to normal), RNAi of w is restored; when the cell death was reversed partially, inhibition still occurs (Figure 1B, hid+diap1). The transgenes diap1 and p35 do not suppress the ttk induced cell death; in this combination, RNAi is still suppressed, excluding the possibility that these antiapoptotic genes themselves restore the RNAi silencing. Also, crosses that introduce only the antiapoptotic transgenes show no impact on w RNAi.

It is known that cells overexpressing hid will produce a proliferation signal to neighboring cells, a process not affected by the suppression of cell death by p35 (Huh et al. 2004; Pérez-Garijo et al. 2004; Ryoo et al. 2004). Indeed, the expression of p35 suppressed both cell death and the RNAi inhibition, indicating that this proliferation signal is independent of any signal affecting RNAi. In contrast, another proliferation signal produced by dying cells is affected by p35 and diap1 (Fan and Bergmann 2008), and thus it cannot be ruled out as coincident with a potential RNAi suppression signal.

Cell death inhibits w RNAi induced by diverse constructs

Different RNAi constructs were used to generate dsRNAs of the w gene (Giordano et al. 2002; Kalidas and Smith 2002) (Figure S1). These transgenes utilize an upstream activating sequence (UAS) for expression and silence the w gene to varying degrees when the transcription factor Gal4, which targets the UAS, is expressed by the promoters GMR, actin5c (act5c), or α-tubulin (Tub) (Giordano et al. 2002; Kalidas and Smith 2002) (Figure S4). To confirm that the RNAi inhibition is not limited to certain constructs, we crossed these strains to the eye mutations B, Dr, and Gla, as well as to cell death strains overexpressing grim, hid, rpr, strica, and ttk. Eye color restoration was observed in all the tested strains with strica, and in all but one construct with Gla, Dr, and grim (Figure S4 and Table S2). However, in some cell death and RNAi combinations, no clear restoration was detected, especially with the constructs that alone show stronger silencing. Weaker pigment restoration was also observed when the silencing is achieved by a stronger promoter for Gal4, and different degrees of restoration were observed with different transgenic strains from the same RNAi construct (Figure S4 and Table S2). These results suggest that the strength of the silencing plays a role in the magnitude of RNAi reversal. Moreover, because different promoters were used in these experiments, they demonstrate that downregulation of the GMR promoter of GMR-wIR by cell death signaling is not the basis of the restored eye color in the original observations.

Cell death suppresses GFP RNAi

To test the generality of the cell death inhibition, RNAi of Green Fluorescent Protein (GFP) was examined within the cell death strains. GFP fluorescence can be detected in the eye only in the absence of the normal pigment (Plautz et al. 1996; Berghammer et al. 1999). To overcome this obstacle, GFP was combined with the brown and vermilion mutations, which together eliminate the pigments. Under these circumstances, we observed strong green fluorescence of the EGFP transgene driven by Tub-Gal4 in the eyes. Weaker fluorescence is found after EGFP is silenced by the transgene EGFPir (Figure S1 and Figure S3) as previously noted in larvae (Roignant et al. 2003). In this EGFP RNAi background, the cell death mutations B, Dr, and transgenes GMR-grim, -hid, -rpr, -strica or -ttk were introduced. Stronger and global restoration of GFP was observed in the eyes of grim, strica, and ttk (Figure 1C and Figure S5). Compared to the cell death control, which did not cause detectable autofluorescence in this assay (Figure S5), weaker and area-restricted restoration occurred in the Dr, hid, and rpr eyes.

The results of cell death induced inhibition of w and EGFP RNAi are summarized in Table S2. Greater cell death is caused by overexpression of hid and rpr than that of grim, strica, and ttk (as shown by the eye size); yet there is less RNAi reversal for both w and EGFP. The strica construct restored EGFP expression to the strongest level as was also the case with restoration of w expression. Thus, in general, the relative strength of reversal of RNAi is related to the mode of cell death but is independent of the genes (w or EGFP) involved.

RNAi inhibition is quantitatively related to the level of RNAi

As noted above, a difference in expression of an RNAi transgene affects the amount of restored w expression. We also noticed a weaker restoration observable with lesser expression of Gal4 (e.g., GMR-Gal4) driving an RNAi construct but not with more strongly expressed cases (e.g., Tub-Gal4). This finding suggests that a stronger RNAi effect is above the threshold of suppression. To test this possibility, we crossed multiple GMR-wIR transgenes with B and Gla. B and Gla show an additive effect on RNAi suppression. The results indicate that the restored color is negatively related to the copy number of the GMR-wIR transgene (Figure S6).

The RNAi inhibition can be nonautonomous

The reversal of white RNAi in cells adjacent to the cell death sectors of Bar, bar3, and Dr (Figure 1A) provided an indication of the nonautonomy of RNAi inhibition. As noted above, all these mutants cause a defect of the movement of the MF (Heberlein et al. 1993). The MF determines the fate of the cells: those cells encompassed by it will differentiate and remain alive in the adult eye and the remaining cells will be removed by cell death and will not be present in the adult eye. The color reversal observed in the adult eyes must be located in living cells, which did not undergo cell death. Therefore, the inhibition of w RNAi in adult eyes must have been induced nonautonomously by the death of adjacent cells.

To examine this feature of inhibition by a second method, we produced mosaic cell death in the eyes. Mitotic recombination by FLP–FRT (Golic 1991; Xu and Rubin 1993; Stowers and Schwarz 1999) was conducted in the eye imaginal discs to resolve the zygotic heterozygous genotype into homozygous sectors of proapoptotic transgenes and normal cells using a promoter for FLP that acts before cell death, which was induced by the GMR-driven expression of the death inducer genes in the late third instar (see Materials and Methods). This technique produces mosaic eyes composed largely of two kinds of homozygous lineages: those for the death inducer construct and those for wild-type chromosomes (Stowers and Schwarz 1999). After the adults eclosed, cell death sectors in the eyes can be observed as an irregular arrangement of ommatidia. The most useful involves the overexpression of ttk, which has a dominant phenotype and causes a smooth eye surface due to cell death of specific cell types within the ommatidia. Areas with smooth surfaces were observed in the mosaic eyes, which allow sectors homozygous or occasionally heterozygous for the ttk construct to be identified definitively. The adjacent regions are homozygous for normal chromosomes; otherwise they would show a smooth eye surface. The ttk sectors were accompanied by color restoration spreading from two to several lines of ommatidia, depending on the size of the cell death containing areas (Figure 2A and Figure S7). Thus, the inhibition of RNAi is occurring in the homozygous normal cells, consistent with the pattern of color restoration in the B, hhBar3, and Dr eyes. The color intensity in the normal sectors progressively decreases with the distance from the death sectors. Collectively, these data suggest the existence of a signal to suppress RNAi initiates from the region of cell death and travels a limited distance into regions with normal cells.

Figure 2

Cell death caused RNAi inhibition in neighboring cells. Mosaic cell death of GMR-ttk and GMR-hid generated by mitotic recombination. (A). The color is restored by the GMR-ttk cells. Selected examples of sectors of ttk cell groups are circled by black lines, which were indicated by the smooth surfaced areas. Because the GMR-ttk/+ genotype has a dominant phenotype (Figure 1B), the normal sectors are those in which mitotic recombination during development have resolved into +/+ sectors. These wild-type cells show reversal of RNAi when adjacent to GMR-ttk sectors. Surrounding two examples of the death areas, the color that is restored in normal cells is circled by white lines. The strength of inhibition diminishes with the distance from the ttk sectors. (B) Death in homozygous GMR-hid cells did not cause inhibition of w RNAi silencing. Mosaic cell death of GMR-grim (C) and GMR-rpr was generated by mitotic recombination (D). In these cases, w RNAi was occasionally reversed as indicated by the red sectors in the eyes. Bars, 0.1 mm.

In contrast, cell death induced by the hid transgene in an analogous mitotic recombination experiment did not invoke RNAi inhibition in most cases, potentially indicating cell death does not inhibit RNAi under certain conditions (Figure 2B). Given that the same transgene is able to inhibit RNAi when heterozygous (Figure 1B) and the mitotic recombination is effective at homozygosis at >90% (Stowers and Schwarz 1999), we suggest this difference is due to the rapid elimination of the dying cells in the case of the homozygous transgene. Thus, there is possibly a required threshold of duration of the dying cells to provoke substantial inhibition of RNAi.

Mosaics of rpr or grim produced by the FLP–FRT method exhibit colored sectors with the size usually not over an ommatidium in all eyes examined (Figure 2, C and D). In this case, the eyes from experiments involving homozygosis of rpr and grim transgenes do exhibit RNAi inhibition, possibly because rpr and grim transgenes cause less severe cell death effects than hid.

Cell death inhibits RNAi in different tissues

To investigate the possibility of cell death inhibition of RNAi in tissues other than the eye, we tested strains with a heat-shock promoter transgene hs-hid in the chromosomal balancers CyO or TM3, respectively. When the flies are grown at 18° or room temperature without heat shock, w RNAi is slightly reversed, showing red spots in the eye (Figure 3A). This result indicates that the constitutive expression of the transgenic hid is strong enough to cause some RNAi inhibition. AO staining, which is used to detect cell death (Jiang et al. 1997), is also observed in the transgenic larvae (Figure S8A). The expression of the hid transgene was confirmed by RT–PCR (Figure S8B). A further confirmation that the restored color results from a low level of cell death conditioned by the hs-hid transgene was that suppression of RNAi was reversed by adding the p35 transgene to the genotype (Figure S8C). Collectively, these results demonstrate that the transgene was expressed and caused sporadic cell death.

Figure 3

Cell death-related inhibition of RNAi occurs in different tissues and organs and the increased expression of the marker gene EGFP is caused by impaired processing of dsRNA into siRNA. (A–F). The default expression of the transgene hs-hid inhibits w silencing in the eye (A) and EGFP silencing in larvae (B), wing disc (C), eye-antennal disc (D), midgut (E), and salivary glands (F). In each image except the eye, controls without hs-hid but with EGFP RNAi are located to the right. A salivary gland showing restored fluorescence and a control pair of glands is shown in the lower right. Bars, 0.1 mm. (G–I). Northern blotting detects altered levels of EGFP mRNA, dsRNA (G), and siRNA (H) in the hs-hid strain. A dilution series (1X, 2X, and 4X) of total RNA or total small RNA was loaded and indicated by the numbers of micrograms. Levels of α-tubulin (Tub) mRNA (G), and 5S rRNA (H) were probed, respectively, as input controls. The patterns of siRNA were not different when probed with either sense or antisense EGFP probes (not shown). Statistical analysis of the EGFP mRNA, dsRNA, and siRNA levels in “hs-hid; EGFPir” larvae is shown in I. Northern results were scanned and analyzed by Image Gauge. The fold difference compared to the samples without hid (“EGFPir”) was normalized by the loading control. The heights of solid bars with error bars indicate the fold difference; *P < 0.05 (no cell death), **P < 0.01, ***P < 0.001.

To analyze the transgenic hid effect on EGFP RNAi, third instar larvae were directly examined for green fluorescence or following dissection into tissues and organs. Stronger EGFP expression was obvious in larvae over all and in midguts, salivary glands, brains, and imaginal discs (Figure 3, B–F). Similar results were observed when hs-hid was located on either chromosome 2 or 3 balancers (Figure S9). This result indicates that the phenotype is indeed caused by the hs-hid gene and not by other factors on the balancer chromosomes, which without the transgene have no effect on RNAi. Thus, we conclude that cell death inhibition of RNAi operates in different tissues and developmental stages.

RNAi is inhibited by blocking the processing of dsRNA into siRNA

The expression change of EGFP in multiple larval tissues provides an opportunity to investigate mRNA, dsRNA, and siRNA changes during suppression of RNAi. As expected, the EGFP mRNA level in hs-hid transgenic animals increases about twofold (Figure 3, G and I). This change probably reflects a mixture of affected and unaffected cells, so the change in the affected cells is likely of much greater magnitude. Interestingly, we detected more than a threefold increase of the dsRNA in the hid transgenic strain (Figure 3, G and I). This accumulation is likely caused by impairment of EGFP dsRNA processing. Consequently, we expected to see an siRNA decrease. Indeed, EGFP siRNA is reduced ∼40% in the transgenic strain (Figure 3, H and I). The results suggest the target step of RNAi suppression is the conversion of dsRNA to siRNA.

Adenosine deaminase acting on RNA (ADAR), an RNA editing enzyme, has been shown to compete for dsRNA as a substrate and thus inhibit RNAi (Scadden and Smith 2001; Wang et al. 2005; Heale et al. 2009). ADAR edits the dsRNA (Carpenter et al. 2009) and makes it unfavorable as a substrate for Dicer-2, the enzyme component of RNAi that processes dsRNA (Scadden and Smith 2001). Thus, we examined the possibility that ADAR might be induced by cell death signaling and compete with RNAi for dsRNA substrates. However, this possibility was excluded because there is no increase in expression of ADAR accompanying cell death, and when ADAR expression is reduced by RNAi, cell death-mediated RNAi inhibition was not affected (Figure S10).

Cell death inhibition of RNAi regulates endogenous gene and transposable element expression

Recent work shows that siRNAs exist in somatic cells and match selected endogenous genes and transposable elements. Among those genes, mus308, encoding an enzyme with DNA polymerase and helicase function, has been demonstrated to be repressed by RNAi (Czech et al. 2008; Okamura et al. 2008). To investigate whether cell death upregulates mus308 by its inhibition of RNAi, RT–PCR was performed to detect the expression level of this gene in the hs-hid transgenic larvae. A doubling of the expression was observed by semiquantitative PCR, which was confirmed by real-time PCR (Figure 4A). This result is similar to the EGFP upregulation and suggests that endogenous genes under RNAi control can be affected in the same manner.

Figure 4

Cell death upregulates the expression of the endogenous gene mus308 and transposable elements and reduces the siRNA level. (A) mus308 mRNA amount was doubled in the hs-hid strain, as detected by real-time PCR and semiquantitative PCR. The fold difference compared to the samples without hid (“EGFPir”) was normalized by β-tubulin. As noted in the text, the magnitude in affected cells is undoubtedly greater. Levels of statistical significance are designated as described in Figure 2. (B) Expression of transposons 1731, mdg1, 297, BEL, DOC, and S elements in “hs-hid EGFPir” larvae. Semiquantitative RT–PCR results were analyzed by Image Gauge. The fold difference compared to the samples without hid (“EGFPir”) was normalized by the loading control. (C) Northern results show let-7 level remained unchanged but the endogenous esiRNA-sl-1 was reduced to ∼45% in the hs-hid strain. The statistical results for triplicate experiments are shown below. *P < 0.05 (no cell death), **P < 0.01, ***P < 0.001. (D) Distribution of small RNAs matching TEs from deep sequencing data shows a >50% reduction of the 21-nt length in the cell death strain (EGFPir hs-hid) compared to the control. The read number for each sample was normalized by the miRNA counts. We also used 2S rRNA, which is 30 nt long and included in our sequencing data, to normalize the counts with similar results. Biological replications are indicated by #1 and #2.

When RNAi is compromised, previous studies have shown that a severalfold increase of expression for some TEs was detected in S2 cells (Czech et al. 2008; Ghildiyal et al. 2008; Kawamura et al. 2008) and in adult flies (Chung et al. 2008). We examined five of them, 1731, mdg1, 297, BEL, DOC, and S element to analyze the possible effect from cell death. It was demonstrated that dicer-2 knockout did not significantly affect 1731 and mdg1 expression in adult flies, whereas 297 and BEL mRNAs are increased, although both mdg1 and 297 siRNA levels are reduced (Chung et al. 2008). Our results also show BEL and 297 expression increased approximately twofold (P < 0.05) in the presence of hs-hid–caused cell death, whereas mdg1 and 1731 expression is not significantly changed, and DOC and S may be slightly upregulated (P < 0.1) (Figure 4B). These results are consistent with the hypothesis that cell death reduces siRNA synthesis and thus impairs RNAi.

To further confirm the impact of hs-hid on siRNAs and miRNA, LNA probes were synthesized (Exiqon) for detecting let-7, miR-1, and esiRNA-sl-1 (the siRNA molecule complementary to mus308 mRNA; Kawamura et al. 2008) by Northern blotting (Chung et al. 2008). A change in the amount of a control miRNA was not detected (let-7 in Figure 4C). However, esiRNA-sl-1 was reduced to ∼45% of the control as also found for the EGFP siRNA (Figure 4C).

To determine whether any global change occurred for TE siRNAs, we sequenced the small RNAs (19–40 nt) from the hs-hid transgenic and control strains. The read numbers of sequenced RNAs matching TEs were first normalized by total miRNA reads, then plotted on the basis of their size (Figure 4D, original data see Table S3). Peaks for siRNA (21 nt) and for piRNA (23–29 nt) were clearly evident. The siRNA peak is reduced to ∼43% (averaged) in cell death samples. When the 21-mer set (excluding miRNAs) was compared for homology to the published siRNA/miRNA dataset from fly heads (Ghildiyal et al. 2008), we observed ∼40% reduction of matched siRNAs with cell death (620 to 372 reads). The 21-mer siRNA molecules derived from gene CG4068, which were shown to silence mus308 (Okamura et al. 2008), are also reduced (35 to 13 reads) (see Table S3 for total siRNA number derived from CG4068), matching the Northern results. There is also possibly a weaker reduction in total piRNA amount (Figure 4D and Table S3). However, we were not able to detect a significant reduction of EGFP siRNA level in the hs-hid strain by this method (Table S3) for unknown reasons, although this might be the case if the high expression of EGFP is at saturating amounts. Overall, our results indicate that the expression change of mus308 and TEs is caused by a genome-wide reduction of siRNAs.


We demonstrate in this study that cells undergoing death inhibit RNAi. This process occurs in different stages of the life cycle, in different tissues, and for selected genes silenced by RNAi. Various constructs that generate dsRNA were all found to be effective, suggesting this step is not affected. Indeed, we observed accumulation of dsRNA and reduction of siRNA, indicating that the processing of dsRNA to siRNA is the step impaired. Either modification of dsRNA or Dicer-2 and/or its cofactors may be the target of the cell death signaling. However, we were not able to detect a change of Dicer-2 expression at either the transcriptional or protein level (Figure S11). The phenotype of the eye color suggests this is not an all-or-none regulation, but rather the restored color in the eye is quantitatively reversed relative to the silencing strength of w RNAi.

We demonstrate that cell death inhibits RNAi nonautonomously in neighboring normal cells. The inhibition is conditionally triggered in that when strong cell death was induced, little or no RNAi inhibition was observed. We assume that a threshold of duration of the cells undergoing cell death is required for the effect. This time-lag hypothesis is consistent with the fact that processing of dsRNA to siRNA is reduced and thus caused the RNAi inhibition. Therefore, the lag time may be related to the generation and perdurance of the siRNA molecules. This hypothesis also explains three other observations. First, we did not observe GFP RNAi inhibition during normal embryo development in which programmed cell death occurs. Second, when the hs-hid transgenic larvae were heat shocked, they died within 1 day. During this period of time, we were not able to detect any GFP expression difference in the GFP RNAi animals compared to the controls that were not heat shocked. Third, we also failed to detect a GFP expression change in the eye discs with EGFP RNAi in the late third instar when proapoptotic genes were expressed by the GMR promoter, that is, immediately after ectopic cell death was induced. However, it is important to note that when reversal of RNAi occurred, it did so in normal cells as evidenced from the results from eye mutations and the somatic recombination mosaic analysis.

One possibility for the existence of cell death induced inhibition of RNAi might be that RNAi acts as a first line of defense against viruses (Li and Ding 2005). If this fails and cell death results from viral proliferation, signaling might occur to neighboring cells to induce processes that modify dsRNA as a second line of defense, making them inaccessible for processing to siRNA, so that they cannot enter the RNAi pathway leaving the integrity of the target mRNA intact, analogous to the interferon response in mammals (Kawai and Akira 2007; Randall and Goodbourn 2008). This hypothesis can account for the collective data that indicate a quantitative negative relationship between the cell death signal and the effectiveness of RNAi. Such a second line of defense would be selected if RNAi regularly failed to stop viral infection, which is obviously the case, and would provide a different means of attacking dsRNA.

Apoptosis is known to play a role in immunity to viruses in insects in that the apoptotic response can severely limit viral replication and thus many viruses encode antiapoptotic genes (reviewed in Clarke and Clem 2003; Clem 2005). Recent data on viral infection in Drosophila show that long dsRNAs but not the siRNAs stimulate whole-body immunity to homologous virus infection (Saleh et al. 2009). Thus, the existence of cell death induced inhibition of RNAi causing local maintenance of dsRNA could possibly be beneficial for viral defense by RNAi throughout the body if localized pathogen-induced cell death occurs.

Knocking down of Dicer-2 or Argonaute-2, the core components of the siRNA pathway, has been shown to increase the expression of many transposable elements, indicating that this pathway plays a role in repressing TEs in addition to the piRNA pathway (Chung et al. 2008; Czech et al. 2008; Ghildiyal et al. 2008; Kawamura et al. 2008; Okamura et al. 2008). Our findings suggest that chronic ectopic cell death inhibition of RNAi also elevates TE expression. This finding suggests that chronic diseases or long-term exposure to pathogens, which might cause cell death in the stressed tissue, may cause activation of endogenous TEs in the adjacent normal cells. It has been documented in yeast cells under stress, that the retrotransposon Ty5 will randomize its integration, which otherwise mostly targets heterochromatin, and thus become mutagenic (Dai et al. 2007). Similarly, the activation of TEs in cells surrounding cell death may increase the somatic mutation frequency. Therefore, the inhibition of RNAi induced by chronic ectopic cell death may provide insight for the genesis of certain types of diseases, such as some cancers, in which persistent inflammation and somatic mutations are implicated (Cooper 1995; Coussens and Werb 2002; Copeland and Jenkins 2009).

Lastly, RNAi is widely used in genetic analysis to eliminate specific gene function and many practical applications of this tool have been suggested (Aigner 2007). The finding that chronic ectopic cell death alters the pattern of RNAi is critical for an understanding of the application of the technique in basic and applied studies.


We thank R. Carthew, B. Hay, E. Giordano, C. Antoniewski, T. Zars, and C. Tan and the Bloomington Drosophila Stock Center for providing fly strains. We thank Ryan Donohue for performing some PCR experiments, Sean Blake for sequencing small RNAs, and William G. Spollen for generating and trimming the sequence data. We also thank Kyungju Chin, Robert Gaeta, and Louis Mega for advice.


  • Received March 5, 2011.
  • Accepted May 15, 2011.

Literature Cited

View Abstract