Abstract
In nematodes, flies, trypanosomes, and planarians, introduction of double-stranded RNA results in sequence-specific inactivation of gene function, a process termed RNA interference (RNAi). We demonstrate that RNAi against the Caenorhabditis elegans gene lir-1, which is part of the lir-1/lin-26 operon, induced phenotypes very different from a newly isolated lir-1 null mutation. Specifically, lir-1(RNAi) induced embryonic lethality reminiscent of moderately strong lin-26 alleles, whereas the lir-1 null mutant was viable. We show that the lir-1(RNAi) phenotypes resulted from a severe loss of lin-26 gene expression. In addition, we found that RNAi directed against lir-1 or lin-26 introns induced similar phenotypes, so we conclude that lir-1(RNAi) targets the lir-1/lin-26 pre-mRNA. This provides direct evidence that RNA interference can prevent gene expression by targeting nuclear transcripts. Our results highlight that caution may be necessary when interpreting RNA interference without the benefit of mutant alleles.
IN the last year wide acceptance of the technique called double-stranded-mediated RNA interference (RNAi) has made the investigation of gene function much more accessible (Fireet al. 1998). However, the mechanism underlying RNAi remains largely unknown. The potential relevance of RNAi, which was first described in Caenorhabditis elegans, has recently expanded with the discovery that RNAi can specifically inactivate gene function in insects (Kennerdell and Carthew 1998; Misquitta and Paterson 1999), trypanosomes (Ngoet al. 1998), and planarians (Sánchez Alvarado and Newmark 1999). RNAi has also been compared to post-transcriptional or homology-dependent gene silencing (PTGS) in plants (Voinnetet al. 1998; Waterhouseet al. 1998), raising the possibility that RNAi is a mechanism present in all eukaryotes (for recent reviews see Montgomery and Fire 1998; Sharp 1999).
It was initially proposed that RNAi could target four different stages of gene expression: (i) the gene itself could be targeted by direct mutagenesis; (ii) transcription could be prevented; (iii) the transcript could be targeted for degradation; (iv) translation could be prevented. Mutagenesis of the target gene was excluded because no changes in DNA sequence were found in animals affected by RNAi (Montgomeryet al. 1998). Several different types of experiments demonstrated that initiation of transcription is not a target for RNAi (Fireet al. 1998; Korfet al. 1998; Montgomeryet al. 1998). In particular, Montgomery et al. (1998) have shown that in embryonic blastomeres the accumulation of transcripts from reporter constructs is completely prevented in the cytoplasm and partially prevented in the nucleus. This reinforces the possibility that RNA is the target for RNAi and that it is degraded.
C. elegans is unusual among eukaryotes for several reasons. Most transcripts are trans-spliced at their 5′ ends to a small sequence called a spliced leader (SL; Blumenthal 1998) and ~25% of genes are organized in transcriptional operons (Zorioet al. 1994; Blumenthal 1998). Most operons conform to three criteria: (i) the genes are only 100–400 bp apart; (ii) generation of the downstream transcript is achieved by coupling polyadenylation of the upstream transcript with trans-splicing of the downstream transcript; (iii) if the upstream transcript is trans-spliced it is trans-spliced to SL1 whereas downstream transcripts are trans-spliced to SL2 or its variants.
Recently, we have been investigating the complex genomic organization of the genes lin-26 (lin, lineage abnormal), lir-1, and lir-2 (lir, lin-26 related), which encode homologous putative transcription factors and define a new C2H2 motif related to TFIIIA zinc fingers (Labouesseet al. 1994; Dufourcqet al. 1999). These genes are organized in two overlapping operons, both of which conform to the classic operon described above (Dufourcqet al. 1999). Trans- and alternative splicing of lir-1 results in at least six isoforms that can be categorized into two groups, long and short. Long lir-1 isoforms are trans-spliced to SL2 at exon 1 and are organized in a transcriptional operon with lir-2. Short lir-1 isoforms are trans-spliced to SL1 at exons 2 or 3 and are organized in a transcriptional operon with lin-26 (Figure 1). The first and second lir-1 exons are separated by an unusually long intron, which contains promoter sequences for the second operon (den Boeret al. 1998). The lir-2/lir-1(long) operon is expressed in all cells, whereas the lir-1(short)/lin-26 operon is expressed in nonneuronal ectodermal cells (Dufourcqet al. 1999).
The lir-2, lir-1, and lin-26 operons. The genes lir-2, lir-1, and lin-26 form two overlapping operons. The first operon [lir-2/lir-1(long) operon] includes lir-2 and long lir-1 isoforms starting at the first lir-1 exon (lir-1A-C); the second operon [lir-1(short)/lin-26 operon] includes short lir-1 isoforms starting at or after lir-1 exon 2 (lir-1D-F) and lin-26. Unshaded boxes correspond to untranslated sequences, open triangles represent introns, and variously shaded boxes represent exons of lir-2, lir-1, or lin-26.
lin-26, the best characterized of these three genes, is required for nonneuronal ectodermal cells to maintain their normal fates (Labouesse et al. 1994, 1996). Strong and null lin-26 alleles lead to embryonic lethality due to degeneration of most hypodermal (epidermal-like) cells and glial-like cells (Labouesse et al. 1994, 1996). A weak loss-of-function mutation, lin-26(n156), causes a set of hypodermal precursors to adopt a neural fate resulting in a vulvaless phenotype (Fergusonet al. 1987; Labouesseet al. 1994). This viable mutation leads to early larval lethality when in trans to a lin-26 null allele, with defects in tissues and organs made of hypodermal cells (Labouesseet al. 1994). These include an abnormal tail, excretory system, and rectum, as well as a dumpy body shape (Dpy). Furthermore, loss of lin-26 expression in the somatic gonad epithelium and its precursors leads to sterility (den Boeret al. 1998).
Prior to this work, no specific lir-2 or lir-1 mutations had been identified and efforts to ascertain their function form the background to this study. We report the identification of a lir-1 null mutant and compare its phenotype to that of lir-1(RNAi) animals. We show that lir-1(RNAi) induces hypodermal defects reminiscent of lin-26 phenotypes although the lir-1 null mutation is viable. We test possible hypotheses that could explain this apparent discrepancy and conclude that lir-1(RNAi) targets the pre-mRNA of the lir-1(short)/lin-26 operon, thereby interfering with both lir-1 and lin-26 gene function.
MATERIALS AND METHODS
Strains and general methods: Methods for genetic analysis and the reference wild-type strain were as described in Brenner (1974). Other strains were: CB187, rol-6(e187) (Brenner 1974); ML581, lin-26(mc15) unc-4(e120)/mnC1 [dpy-10(e128) unc-52(e444)] (den Boeret al. 1998); ML335, dpy-2(e489) mcDf1 unc-4(e120)/mnC1 [dpy-10(e128) unc-52(e444)] (Chanal and Labouesse 1997).
RNA interference: The templates used for RNA synthesis were amplified by PCR with primers that have a T3 (ATTAACCCTCACTAAAGG, to generate the sense strand) or T7 (AA TACGACTCACTATAGG, to generate the antisense strand) promoter sequence at their 5′ ends. The size and purity of PCR products was checked by agarose gel electrophoresis, but they were not sequenced. The following list gives the position of the starting nucleotide for each sense and reverse primer used in PCR reactions with cDNA or cloned genomic DNA (for introns). The first set of numbers refers to the nt coordinates of the primers in cosmid F18A1 (GenBank accession no. U41535), the second to the length of the PCR product:
lir-2(exons4-6): 14701 and 16283; 550 bp
lir-2(exons4-7): 14701 and 16805; 1020 bp
lin-26(full length): 28166 and 29857; 1300 bp
lin-26(exon3,partial): 29102 and 29653; 190 bp
lir-1(full length): 17686 and 27907; 1000 bp
lir-1(exons3-5): 26945 and 27907; 675 bp
lir-1(exons1-3): 17720 and 27388; 570 bp
lir-1(exons1-4): 17720 and 27719; 850 bp
lir-1(exons1-2): 17686 and 26940; 315 bp
lir-1(exon3): 27165 and 27569; 405 bp
lir-1(exons4-5): 27622 and 27907; 205 bp
lir-1(exon1): 17720 and 17784; 50 bp
lin-26(intron3): 29443 and 29640; 197 bp
lir-1(intron2): 27002 and 27160; 159 bp
The following list gives the position of the starting nucleotide for each sense and reverse primer used in PCR reactions with genomic DNA:
lin-15A: 6351 and 7419 in cosmid ZK678 (GenBank accession no. Z79605)
lin-15B: 2994 and 3975 in cosmid ZK678
ppp-1: 6867 and 5531 in cosmid C15F1 (GenBank accession no. AC006608)
tra-2: 4784 and 3754 in cosmid C15F1
rol-6: 229 and 943 in cosmid T01B7 (GenBank accession no. Z66499)
Sense and antisense RNA strands were individually synthesized using the mMESSAGE mMACHINE kit (Ambion, Austin, TX) following the manufacturer's instructions. DNA templates were removed with a 15-min DNaseI treatment. RNAs were extracted with phenol/chloroform and chloroform, precipitated in isopropanol, resuspended in 10 mm Tris-HCl (pH 7.5), 1 mm EDTA (pH 8.0), and annealed. Double-stranded RNA (dsRNA) was microinjected into the syncytial gonad arms of rol-6(e187) animals together with dsRNA corresponding to the rol-6 locus. The rationale for using this control procedure is that rol6(e187) animals are Rol, whereas the null phenotype of rol-6 is wild type (Hodgkin 1997); hence only oocytes that have incorporated the injected dsRNA will develop as non-rollers.
After being injected, animals were allowed to recover for 4 hr before they were cloned and subsequently transferred to fresh plates at 8-hr intervals for 24 hr. Terminal phenotypes were identified and analyzed especially with respect to elongation, morphology, hypodermis integrity, organ morphology, sexual identity, and fertility.
lin-26::gfp construct: The lin-26::gfp construct (pML702) is a modification of pML301 (den Boeret al. 1998). The GFP (green fluorescent protein) coding sequence was PCR amplified, so that a stop codon was introduced at its 3′ end, and cloned, in frame, into the NheI site of exon 2 of lin-26B (Dufourcqet al. 1999). This fusion protein was targeted to the nucleus by cloning a nuclear localization signal just upstream of the GFP. Transgenic lines were established as described by Mello and Fire (1995). pML702 was coinjected at 10 ng/μl with plasmid pRF4 [rol-6(su1006)] at 100 ng/μl. One transgene was integrated (den Boeret al. 1998) to generate the allele mcIs17, which was subsequently outcrossed four times with N2 animals.
Isolation of a lir-1 null mutation: To isolate a lir-1 mutation, we used a protocol adapted from Jansen et al. (1997). Briefly, wild-type N2 animals were incubated with trimethyl-psoralen at 20 μg/ml for 15 min and subsequently exposed to a UV lamp (4 W/365-nm model VL-4L from Bioblock Scientific) for 30 sec at a distance of 10 cm (Yandellet al. 1994; see also Chanal and Labouesse 1997). Pools of ~400 F1 animals were placed on a single plate. F2 animals from six different plates were collected together and divided into two equal populations, one that was frozen at −80° and another from which DNA was prepared. We collected, in this way, the progeny from 1.1 × 106 F1 animals and assayed them for the presence of potential deletions by nested PCR using the primers 5′ ATCACGTGAAGTGTGAAGGTC (lir-1 intron 1) and 5′ GAGTTGGAGACTCCTCTACTT (lin-26B exon 4) followed by 5′ GCCGAAAATGGGTGTGCGCA (lir-1 intron 1) and 5′ GAATGGAATATGGAACACTCCATGC (lin-26B intron 3).
We recovered one mutation that deletes a fragment of 3276 nucleotides from lir-1 (position 24453 to 27728 in cosmid F18A1). This mutation, which was named lir-1(mc33), allows for only the synthesis of less than 42-amino-acid LIR-1 proteins and is thus likely to be a lir-1 null allele. We demonstrated that the corresponding lir-1 sequences were indeed deleted, rather than translocated somewhere else in the genome, by carrying out PCR reactions on homozygous lir-1(mc33) animals using various primer pairs internal to the mc33 deletion: in no case did we find a band that would indicate the presence of translocated lir-1 sequences (data not shown).
Characterization of lir-1(mc33) animals: The mutation lir-1(mc33) was outcrossed six times, marked with the mutation unc-4(e120) and balanced with mnC1. Genetic characterization of lir-1(mc33) showed that it is tightly linked to an ≈300-kb deficiency that we could not separate from mc33; we named it mcDf3 (see Figure 6A for a genetic map). We mapped the extent of mcDf3 by PCR (Williamset al. 1992) starting from dead eggs laid by heterozygous mc33 mcDf3 animals and primers derived from the C. elegans genome sequence (TheC. ELEGANS Consortium 1998; names of cosmids that have been tested are available upon request). We mapped in a similar way the right breakpoint of the deficiency mcDf1, which deletes ≈350 kb including lir-1 and sequences further to its left (Chanal and Labouesse 1997). We found that mcDf3 deletes the equivalent of 11 cosmids to the right of lin-26 (between B0495 and C08B11), starting ≈4 kb downstream of lin-26, and that mcDf1 breaks in the middle of C06A8 (see Figure 6A for a genetic map). Therefore mcDf1 and mcDf3 both delete B0495, B0228, and part of C06A8. Although both deficiencies are embryonic lethal when homozygous, in crosses between ML335 males and lir-1(mc33) mcDf3 unc-4(e120)/mnC1 hermaphrodites we observed 25% Unc larvae that failed to develop beyond the L1 stage. These presumptive mc33 mcDf3/mcDf1 larvae are expected to be missing the products encoded by lir-1 and the genes located within the cosmids B0495, B0228, and the beginning of C06A8. Larval lethality of these animals is probably due in part to the absence of the gene let-253, which can be rescued by a cosmid that overlaps with B0495 (M. Labouesse, unpublished results).
To determine the potential phenotype of a lir-1 null mutation, we generated mc33 mcDf3/mcDf1 heterozygous animals carrying the cosmids that are deleted in the region of overlap between mcDf1 and mcDf3. To this end, we first introduced the cosmids B0495, B0228, and C06A8 (each at 5 ng/μl) into lir-l(mc33) mcDf3 unc-4(e120)/mnC1 animals by germline transformation using the myo-3::gfp plasmid pPD93.97 (at 20 ng/μl), a body wall muscle marker (Fireet al. 1998). Eleven stable fluorescent lines were isolated and tested by PCR for the presence of cosmids. Similarly, in control injections using the cosmids F18A1 (to test for lir-1 rescue), B0495, B0228, and C06A8, we obtained 10 stable transgenic lines. Transgenic animals carrying either the cosmids B0495, B0228, and C06A8, or the cosmids F18A1, B0495, B0228, and C06A8 were crossed with ML335 males. After moving parents daily to a fresh plate, mating plates were inspected under the dissecting scope for the presence of Unc cross-progeny [expected genotype mc33 mcDf3 unc-4(e120)/mcDf1 unc-4(e120) heterozygous animals] and under a GFP scope for the presence of possible GFP+ Unc larvae (as a test for the presence of the transgene). On each mating plate there were wild-type heterozygous animals (both GFP+ and GFP−), paralyzed sterile animals (GFP+ and GFP− homozygous mnC1), Unc adults that were exclusively GFP+, arrested Unc larvae that were GFP− or rarely GFP+, and no or very few dead eggs (Table 1).
RESULTS
lir-1(RNAi) affects hypodermal cells: To investigate the possible functions of lir-1 and lir-2, prior to the identification of any mutations, we decided to use RNA interference. This technique was first described using single-stranded RNA (Guo and Kemphues 1995); however, it was subsequently recognized that double-stranded RNA is the active molecule (Fireet al. 1998); the term RNAi itself was coined by Rocheleau et al. (1997). As an initial control we found that RNAi using lin-26-specific dsRNA phenocopied lin-26 null alleles (Figure 2C; Labouesseet al. 1994). Specifically, 100% of embryos failed to elongate beyond the 1.5-fold stage and contained degenerating hypodermal cells (Figures 2E and 3A; the specific stages of embryonic development are referred to by their morphology: lima bean, comma, 1.5-fold, 2-fold, 3-fold, and finally pretzel).
Full-length lir-1(RNAi) gave a phenotype that was reminiscent of lin-26(mc2), a moderately strong mutation (Figure 2D; Labouesseet al. 1994). Specifically, almost all embryos arrested at or just beyond the twofold stage with cells and droplets floating within the egg shell, suggesting that defects in hypodermal cells have resulted in leakage of cells or cytoplasm (Figures 2F and 3C). This raised the possibility that lir-1 shares a common biological function with lin-26, which would not be unexpected since they both code for homologous proteins and are organized in a transcriptional operon.
lir-1(RNAi) and strong lin-26 mutations induce similar embryonic phenotypes. Nomarski pictures show embryos at the 1.5-fold stage (A) or at the end of embryogenesis (B–F); anterior is to the left and dorsal is up in A, C, and E; all pictures are lateral views, except E, which is central. (A and B) Wild-type embryo. (C) lin-26(mc15) embryo; mc15 is a lin-26 null allele. Notice that it does not elongate beyond the 1.5-fold stage and shows signs of degenerating hypodermal cells (arrow). (D) lin-26(mc2) embryo; mc2 is a moderately strong lin-26 allele. This embryo could elongate slightly beyond the 2-fold stage but, due to hypodermal defects, cells or cytoplasm leaked through the hypodermis (arrows). (E) lin-26(RNAi) embryo. The phenotype is similar to that of lin-26 null embryos. (F) lir-1(RNAi) embryo. The phenotype is similar to that of lin-26(mc2) embryos (see leaking cytoplasm; arrow). Bar, 10 μm.
Strikingly, lir-1(RNAi) using smaller stretches of lir-1 exonic sequence resulted in a varied array of phenotypes. lir-1(RNAi) with sequences corresponding to exons 3–5 [lir-1(exons3-5); this nomenclature will be used throughout] also resulted in a highly penetrant embryonic arrest phenotype reminiscent of lin-26(mc2) (Figure 3D). However, smaller stretches of sequence that included exons 1–2, 3, or 4–5 resulted in predominantly larval and/or adult phenotypes (Figure 3, G–K, and Figure 4). Most of the larvae died as L1/L2 larvae (there are four larval stages designated L1 through L4). Further analysis revealed that they had a variable range of hypodermal defects, such as a dumpy body shape, abnormal tails, excretory organs, and/or rectums (Figure 4B). Although these phenotypes have been associated with partial loss of lin-26 function in hypodermal cells (Labouesseet al. 1994), they could also reflect lir-1 function.
Animals that survived the first two larval stages sometimes displayed a molting problem resulting in a failure to shed the old cuticle. When this unshed cuticle blocked the mouth it usually led to lethality by the L3 stage; otherwise it formed a waist-like constriction that maintained its smaller diameter as the rest of the body grew (Figure 4, C and E). Since molting requires complete lysis of the matrix that attaches the cuticle to the hypodermis, this molting phenotype is consistent with defects in hypodermis function. Indeed, the only known mutation to affect the physical mechanism of molting alters a gene that acts in the hypodermis (Yochemet al. 1999). In addition, it has been shown that RNAi against the nuclear hormone receptor nhr-23, which is expressed in the epidermis among other tissues, creates a phenotype similar to that of weak lir-1(RNAi) animals (Kostrouchovaet al. 1998).
Most surviving adults were Dpy with abnormal tails and had a vulval phenotype (Figure 4, E and F): they usually had a protruding vulva (Pvl), were sometimes multi-vulva (Muv), or occasionally vulvaless (Vul). In addition, these animals were often egg-laying defective (Egl) and infrequently sterile due to oocytes being absent or abnormal. These adult phenotypes are again reminiscent of known lin-26 phenotypes: for instance, of the two viable lin-26 alleles, ga91 is Pvl (D. Eisenmann, personal communication) and n156 is Vul (Ferguson and Horvitz 1985; Labouesseet al. 1994). Furthermore, specific loss of lin-26 expression in the somatic gonad gives rise to Pvl and sterile animals (den Boeret al. 1998).
It is noticeable that variations in phenotypic severity seen in lir-1(RNAi) correlate with the time postinjection and the size of the dsRNA injected. As the injected dsRNA becomes smaller and as the time postinjection becomes greater, the percentage of embryonic lethality decreases (compare Figure 3, G–J). Furthermore, if two smaller dsRNAs are injected together, there is an increase in phenotypic severity and duration (for instance, compare the incidence of embryonic lethality between Figure 3, I and J, vs. K). Although this suggests that the overall length of sequence homology affects efficiency of interference, RNAi remains most efficient when the interfering dsRNA is present as a single molecule rather than two separate pieces (compare Figure 3, D and K).
To summarize, RNAi using dsRNA corresponding to various regions and sizes of lir-1 sequence results in embryonic and larval phenotypes that are attributable to hypodermal cell defects and can be classified as lin-26-like.
LIN-26 expression is severely reduced in lir-1(RNAi) arrested embryos: Since lir-1(RNAi) results in a lin-26-like phenotype, we examined whether or not lin-26 expression was normal in lir-1(RNAi)-arrested embryos. Using LIN-26 antiserum (Labouesseet al. 1996) we showed that LIN-26 is severely downregulated in lir-1(full length) embryos (Figure 5B). To directly confirm this observation in live animals, we examined expression of an integrated lin-26::gfp construct in lir-1(exons3-5) embryos. This construct contains genomic sequences encompassing both lir-1 and lin-26 (except for the very first nucleotides of lir-1(long) isoforms; Dufourcqet al. 1999). We found that lir-1(exons3-5)-arrested embryos have an almost total loss of GFP expression (Figure 5F). These results demonstrate that loss of lin-26 expression accounts for most if not all lin-26-like phenotypes obtained with lir-1(RNAi).
Graphs illustrating the range of phenotypes obtained with RNAi. Progeny (N) were counted every 8 hr between 4 and 28 hr postinjection (see materials and methods). The lengths of injected dsRNA species are indicated under their names. The key for shading in the graphs is given at the bottom of the figure. (A and B) lin-26(RNAi) results in almost 100% arrested embryos throughout the 24 hr. (C and D) lir-1(RNAi) using full-length sequence or sequence corresponding to exons 3–5. These dsRNA molecules were highly potent, resulting in >92% arrested embryos throughout the time points. (E) lir-2(RNAi) did not induce a significant phenotype. (F) rol-6 dsRNA injected alone results in >98% wild-type progeny. (G–J) Experiments in which dsRNAs corresponding to different regions of lir-1 were injected. (K) Combined injection of the dsRNAs used in I and J. (L) lir-1(exon1) failed to induce a phenotype.
It has been reported that cross-interference between homologous genes occurs when RNAi is carried out using dsRNA containing regions of high similarity (Fireet al. 1998). Since lin-26 and lir-1 define with lir-2 and lir-3 a new gene family (Dufourcqet al. 1999), there was a possibility that the lir-1(RNAi) phenotypes represented cross-interference with another member of this family. The nucleotide sequence similarity between these four genes is low except in the region coding for the zinc fingers, where it is ~50% identity. Yet, RNAi against full-length lir-2 or a fragment coding for its zinc fingers (Figure 3E), or full-length lir-3 (data not shown), failed to give a significant phenotype. This indicates that the lir-1(RNAi) phenotypes are not due to homologous sequences interfering with lin-26.
lir-1 null animals are viable: Two alternative explanations could explain why lir-1(RNAi) causes loss of lin-26 expression: (1) LIR-1 is directly or indirectly required for the positive regulation of lin-26; or (2) since lir-1(short) and lin-26 are organized in an operon, RNAi against the upstream gene (lir-1) could also interfere with expression of the downstream gene (lin-26).
Nomarski pictures showing postembryonic defects associated with lir-1(RNAi). (A) Wild-type L1 larva. (B) lir-1(exon3+exon4-5)-arrested Dpy larva with numerous vacuoles in the hypodermis (arrowheads). (C) lir-1(exon3) larva showing remnants of a previous molt, which include the tail (arrowheads) still attached to the body causing a constriction (arrow). (D) Wild-type anterior gonad. Note the oocytes (white arrowheads), the sperm (white arrow), the eggs (black arrowheads), and the nonprotruding vulva (black arrow). (E) lir-1(exons4-5) adult with a protruding vulva or Pvl (white arrow) and a severe molting defect constriction that closed off the gut (black arrow; the arrowhead points to remnants of the cuticle). (F) lir-1(exons1-2) adult with a Muv phenotype (arrows). Bars, 30 μm.
Using reverse genetics (Jansenet al. 1997), we isolated the mutation lir-l(mc33), which is probably a lir-1 null allele since it removes 3276 nucleotides between the end of the first lir-1(long) intron and the end of the last lir-1 intron. Further genetic characterization of the mutation lir-1(mc33) revealed that it is tightly linked to a chromosomal deficiency, which we called mcDf3 (see materials and methods and Figure 6A). Embryos homozygous for lir-1(mc33) mcDf3 failed to elongate and died during embryogenesis (Figure 5G). When we stained these embryos with LIN-26 antiserum, we observed that lin-26 expression was not severely downregulated as in lir-1(RNAi) embryos (compare Figure 5, B and H), suggesting that lir-1 is not essential for lin-26 expression and that sequences deleted by mc33 are not essential for lin-26 expression. However, since lir-1(mc33) mcDf3 embryonic arrest occurs prior to lir-1(RNAi) embryonic arrest, the presence of the deficiency mcDf3 could potentially mask a requirement for lir-1 function during late embryogenesis or early larval development.
To assess the phenotype of lir-1(mc33) independently of the deficiency mcDf3, we took advantage of the fact that mcDf3 overlaps with another deficiency, mcDf1 (Chanal and Labouesse 1997). This deficiency starts to the right of lin-26, extending in the opposite direction to mcDf3, so that they overlap by three cosmids (B0495, B0228, C06A8; see Figure 6A). We generated heterozygous animals carrying mcDf1 on one chromosome, lir-1(mc33) mcDf3 on the other, and a transgene containing the three cosmids that are deleted by mcDf1 and mcDf3 (see materials and methods). The resulting animals, which we refer to as “lir-1(mc33) hemizygous animals,” have no functional lir-1 but should have at least one copy of the remaining genes deleted by mcDf1 and mcDf3. In principle, their phenotype should reflect the phenotype of lir-1 null animals. We found that lir-1(mc33) hemizygous animals were completely viable and could reach adulthood, thereby ruling out that lir-1 is zygotically necessary for normal lin-26 expression (Figure 6; Table 1).
lir-1(RNAi), but not lir-1(mc33), affects lin-26 expression. Confocal projections (A, B, and H) show the top third of each embryo costained with a LIN-26 antiserum (green) and the monoclonal antibody MH27 (red), which recognizes an adherens junction component (Francis and Waterston 1991), Nomarski pictures (C, E, and G) or epifluorescence images (D and F) show expression of the integrated lin-26::gfp transgene mcIs17. (A) Wild-type 1.5-fold stage embryo. (B) lir-1(RNAi)-arrested embryo, which had elongated until the 2-fold stage (the head is above the body). LIN-26 is barely detectable (arrows); for comparison, the inset shows a wild-type pretzel embryo of the same age that had been processed on the same slide and analyzed by confocal microscopy within the same field and with the same settings. (C and D) Control lin-26::gfp 1.5-fold stage embryo, lateral focal plane. (E and F) lin-26::gfp; lir-1(exon3-5) 1.5-fold stage embryo, lateral focal plane [the ventral bulge, arrow, was often seen in lir-1(RNAi) and lin-26(null) embryos]. (F) Only three cells weakly express the lin-26::gfp construct (arrows). (G and H) lir-1(mc33) mcDf3-arrested embryo. The Nomarski phenotype (G) is much more severe than that of lir-1(RNAi) embryos due to the deficiency mcDf3, but LIN-26 (H) was still detected. lir-1(mc33) mcDf3 embryos, like embryos homozygous for the overlapping deficiency mnDf106 (Chanal and Labouesse 1997), had ~50% additional cells compared to wild-type embryos (data not shown). Bar, 10 μm.
Hemizygous lir-1(mc33) animals reach adulthood. (A) The top line shows part of linkage group II. The middle part shows the extent of four chromosomal deficiencies and four cosmids that map to this area. The deficiencies mcDf1 and mcDf3, which is very tightly linked to lir-1(mc33), overlap with most of B0495, all of B0228, and part of C06A8. Notice that neither lir-1(mc33) nor mcDf3 deletes lin-26. At the bottom is an enlargement of the lir-1/lin-26 region. The extent of lir-1 that is deleted by the mutation mc33 is symbolized by the double-headed arrow. (B) PCR was performed on single animals using primers located on both sides of lir-1(mc33) to test for the presence or absence of a wild-type copy of lir-1. Lane L, 1-kb ladder (GIBCO-BRL, Gaithersburg, MD); lane 1, wild-type animal; lane 2, lir-1(mc33) mcDf3 unc-4(e120)/mnC1 animal; lane 3, DNA purified from a lir1(mc33) mcDf3 unc-4(e120)/mnC1 population; lane 4, GFP−-arrested lir-1(mc33) mcDf3 unc-4(e120)/mcDf1 unc-4(e120) Unc L1 larva; lane 5, GFP+ viable lir-1(mc33) mcDf3 unc-4(e120)/mcDf1 unc-4(e120) Unc adult transgenic for the cosmids B0495 + B0228 + C06A8; lane 6, GFP+ viable lir-1(mc33) mcDf3 unc-4(e120)/mcDf1 unc-4(e120) Unc adult transgenic for the cosmids F18A1 + B0495 + B0228 + C06A8. (C, E, and G) Nomarski images and (D, F, and H) matching epifluorescence images (in the same focal planes) showing expression of the myo-3::gfp construct used as a cotransformation marker. (C and D) Four-day-old lir-1(mc33) mcDf3 unc-4(e120)/mcDf1 unc-4(e120) Unc L1 larva. Although the cause of death of these larvae was not analyzed in detail the hypodermis appeared normal, unlike hypodermis in lir-1(RNAi) larvae. Almost all larvae that died at the L1 stage were GFP−; this exceptional larva was highly mosaic and inherited the transgene in very few cells (arrows). (E and F) Viable lir-1(mc33) mcDf3 unc-4(e120)/mcDf1 unc-4(e120) Unc adult transgenic for the cosmids B0495 + B0228 + C06A8 and the myo-3::gfp marker (see F). This animal, like most animals of this genotype (see Table 1), was Pvl (arrow) and sterile with no visible gametes (compare to Figure 4D). (G and H) Viable lir-1(mc33) mcDf3 unc-4(e120)/mcDf1 unc-4(e120) Unc adult transgenic for the cosmids F18A1 + B0495 + B0228 + C06A8 and the myo-3::gfp marker (see H). This animal was Pvl (arrow) and sterile (see Table 1) with abnormally small eggs (arrowheads) and vacuoles (asterisk). Bars, 20 μm.
“mc33 hemizygous animals” develop into sterile adults
Although lir-1(mc33) hemizygous animals were viable, they were completely sterile and generally had a protruding vulva (Figure 6C; Table 1). The most likely cause for sterility is a germline differentiation defect as gametes were either absent or abnormal (Figure 6E). We are not certain whether the sterility and Pvl phenotypes are attributable to the lack of lir-1 function, for the following reason: In control experiments in which we introduced the cosmid F18A1, which should complement lir-1, in addition to the three cosmids that are deleted by mcDf1 and mcDf3, most animals were still partially sterile and often had a protruding vulva (Figure 6G; Table 1). Possible explanations for the sterility phenotype will be discussed later.
In summary, the fact that lir-1(mc33) hemizygous animals are viable strongly argues that the phenotypes of lir-1(RNAi) animals are primarily, if not entirely, due to their organization in an operon, so that interference of the upstream gene also induces loss of downstream gene expression.
RNAi and operons: To test whether or not this is generally the case we selected two other operons (ppp-1/tra-2 and lin-15B/lin-15A) for which the null phenotypes of the downstream genes had been identified. Null mutations in tra-2 result in XX animals that have the soma and germline of males (Kuwabaraet al. 1992). The null phenotype of ppp-1, which encodes a pyrophosphorylase, is unknown, but it has not been linked to sex determination (Cline and Meyer 1996). Null mutations affecting lin-15B or lin-15A do not result in a mutant phenotype, but those affecting both lead to a Muv phenotype (Clarket al. 1994; Huanget al. 1994). In both cases, we found that injection of dsRNA corresponding to the upstream gene in the operon does not interfere with expression of the downstream gene. Although XX ppp-1(RNAi) animals were Dpy, sterile, and barely viable, they had the body and gonad morphology of hermaphrodites whereas XX tra-2(RNAi) animals looked like males. In agreement with Montgomery et al. (1998) we also found that lin-15A(RNAi) and lin-15B(RNAi) animals had a normal vulva whereas injection of lin-15A dsRNA together with lin-15B dsRNA caused a Muv phenotype (data not shown; see also Tabaraet al. 1998).
Clearly, it is not a general feature of all operons that RNAi directed against the upstream gene will also induce loss of expression of the downstream gene as happens in the lir-1(short)/lin-26 operon.
RNAi with intronic sequences from lir-1 and lin-26 induces a phenotype: It is generally accepted that, for most genes, transcription is coupled to pre-mRNA processing so that as the pre-mRNA is being synthesized it is also being modified to produce the mature transcript (Neugebauer and Roth 1997). For this reason the steady-state level of pre-mRNA varies between genes. A possible explanation for our results is that in the lir-1(short)/lin-26 operon transcript maturation occurs less efficiently than in the ppp-1/tra-2 and lin-15B/lin-15A operons, thereby allowing the pre-mRNA to be targeted by RNAi.
RNAi against lir-1 and lin-26 introns resulted in specific phenotypes. (A and B) These graphs illustrate the range of phenotypes seen with lir-1(intron2) and lin-26(intron3). The stronger and most persistent phenotypes are obtained with lin-26(intron3), which is the largest intron at 197 bp. (C) A lin-26(intron3)-arrested embryo that had elongated to three-fold but had leaking cells within the egg shell (arrows). (D) A lir-1(intron2) early L1 larvae that was Dpy with many vacuoles. Bar, 10 μm.
It has been reported that RNAi using intronic or promoter sequences does not result in detectable interference (Fireet al. 1998), but if the lir-1(short)/lin-26 pre-mRNA accumulates, then dsRNA specific for intronic sequences should induce a phenotype. We tested this with dsRNA corresponding to three introns: lir-1(intron2), which is 158 bp; lin-26(intron2), which is 147 bp; and lin-26(intron3), which is 197 bp. The smaller dsRNA molecules, lir-1(intron2) and lin-26(intron2), induced a lin-26-like embryonic arrest phenotype between 4 and 12 hr postinjection (34 and 21%, respectively) but no significant phenotype at later developmental stages or time points. lin-26(intron3) produced a stronger phenotype in both severity and endurance but this is probably because it is 35% larger than the other two (Figure 7).
To confirm that lir-1(RNAi) induces lin-26-like phenotypes by targeting the pre-mRNA, we performed an additional RNAi experiment, which capitalizes on the fact that lir-1(mc33) mcDf3 and the lin-26 null allele mc15 complement each other for lethality. In lir-1(mc33) mcDf3/lin-26(mc15) animals, the functional copy of lin-26 is linked in cis to the mutation lir-1(mc33). If the lin-26-like phenotypes are the result of the pre-mRNA being targeted, then these animals should be immune to injection of dsRNA corresponding to lir-1 sequences deleted by mc33 (Figure 6A). lir-1(mc33) mcDf3/lin-26(mc15) animals subjected to lir-1(exon1-4) RNAi had the same level of lethality and fertility as uninjected animals, whereas injected control animals had nearly 100% lethality (Table 2). This result also demonstrates that the lir-1 maternal contribution does not account for the lir-1(RNAi) phenotypes described previously.
These results provide direct evidence that RNAi can target the pre-mRNA. We conclude that lir-1(RNAi) induces loss of lin-26 expression because the pre-mRNA for this operon is targeted by the dsRNA, thereby resulting in loss of expression of both genes.
DISCUSSION
We have found that RNA interference against lir-1 leads to severe lin-26-like hypodermal defects that result in embryonic or larval lethality. We have excluded the possibility that this is due to cross-interference between homologous sequences because injection of lir-2 dsRNA, which is as similar to lin-26 as lir-1 is, fails to give any phenotype. We have also demonstrated that the lir-1(RNAi) phenotypes are not due to a maternal lir-1 contribution by showing that these phenotypes are entirely dependent on the presence of lir-1 sequences in cis to lin-26. We account for the lin-26-like phenotypes by showing that lin-26 expression is severely downregulated in lir-1(RNAi)-arrested embryos. Since a null lir-1 mutant is viable, we exclude the possibility that the lethal phenotypes are attributable to lir-1 zygotic function. Finally, we have shown that injection of lir-1 and lin-26 intron sequences leads to phenotypes similar to those resulting from injection of exon sequences. We conclude that lir-1(RNAi) specifically interferes with lin-26 expression because lir-1 and lin-26 are organized in an operon (referred to as the lir-1(short)/lin-26 operon) for which the common lir-1/lin-26 pre-mRNA is available for targeting.
lir-1 function: This work started because we were trying to obtain information about the function of lir-1 and lir-2. We presently derive our conclusions about lir-1 function from two complex genetic backgrounds: hemizygous lir-1(mc33) animals (null for lir-1; see results) and lir-1(mc33) mcDf3/lin-15(mc15) animals in which lir-1 dsRNA had been injected. We believe that lir-1(mc33) represents a true null allele because we have demonstrated absence of the sequences corresponding to wild-type lir-1 and because mc33 deletes lir-1 exons 2–4, which include the putative zinc-finger domains (see Figure 6A). Since hemizygous lir-1(mc33) animals are viable, we can conclude that, unlike lin-26, lir-1 is not essential for hypodermal development. Although sterility is observed in hemizygous lir-1(mc33) animals, lir-1(RNAi) did not affect lir-1(mc33) mcDf3/lin-15(mc15) fertility, so we suggest that lir-1 is not essential for development of the germline. There is an alternative explanation for the sterility observed in hemizygous lir-1(mc33) animals. In previous deletion mapping experiments, we have shown that animals heterozygous for mcDf1 and mnDf106, a deletion that breaks in B0228 (see Figure 6A), are viable but sterile (Chanal and Labouesse 1997), which suggests the existence of a locus necessary for germline development in B0228 and/or in C06A8. Complete rescue of a deletion of this locus [as occurs in hemizygous lir-1(mc33) animals] is expected to be difficult as transgenes are generally poorly expressed in the germline (Kellyet al. 1997). Since we observed that lir-1(mc33) mcDf3/lin-26(mc15) and lir-1(RNAi) animals are predominantly Pvl (protruding vulva), it is possible that lir-1 has a late function in vulval and/or uterine development (normal uterine development is necessary for proper eversion of the vulva; Seydouxet al. 1993). We do not think that mc33 deletes a cis-acting sequence essential for normal lin-26 expression during vulval or uterine development. Indeed if we introduce the deletion mc33 on a lin-26 transgene that normally completely rescues the null phenotype of lin-26(mc15), we still observe full rescue (S. Quintin and M. Labouesse, unpublished results). However, reduced gene dosage of lir-1 and lin-26 may reveal synthetic effects or more subtle requirements for a cis-acting sequence. A definitive assessment of the role of lir-1 in vulval morphogenesis awaits the isolation of another lir-1 mutation that leads to a less complex genetic background.
lir-1(RNAi) induces a phenotype only if lir-1 is in cis to lin-26
RNAi and operons: As summarized before, we have determined that targeting the upstream gene with RNAi also interferes with downstream gene expression for the lir-1(short)/lin-26 operon. However, for two other operons (lin-15B/lin-15A and ppp-1/tra-2), we found that targeting the upstream gene has no effect on the downstream gene and vice versa. This is in agreement with results reported for lin-15B/lin-15A (Montgomeryet al. 1998) and mes-6/cks-1 (Korfet al. 1998). We have two possible explanations, which are not mutually exclusive, to account for the RNAi response of different operons: (i) the processing of lin-15B/lin-15A and ppp-1/tra-2 pre-mRNA is more efficient so that a single target for RNAi does not exist; (ii) in all operons a significant proportion of pre-mRNA is targeted by RNAi, but lin-26 is much more sensitive to gene dosage. To clarify, in other operons there remains enough expression of the downstream gene to ensure wild-type function, whereas reduction of lin-26 expression by lir-1(RNAi) is fatal.
In the case of the lir-1(short)/lin-26 operon, Northern blots and RT-PCR experiments have confirmed that several pre-mRNA species do exist, which are at least 20-fold less abundant than the mature lir-1 and lin-26 transcripts. Specifically, we can detect three precursors at 3.3, 4.2, and 6.2 kb; the 3.3-kb precursor (the most abundant of them) starts, based on its size, at the beginning of lir-1 exon 3, while the other two presumably start in lir-1 intron 1 (data not shown; note that we have previously reported the existence of a lir-1 RT-PCR product beginning at lir-1 exon 3; Dufourcqet al. 1999). So the degree of susceptibility of lin-26 gene function to lir-1(RNAi) probably reflects the competition between RNA processing and RNAi-mediated degradation of the pre-mRNA. This is supported by the observation that for a given length intronic lin-26 dsRNA is less potent than exonic lin-26 dsRNA and that short lir-1 dsRNA molecules are less potent than longer ones, suggesting that a proportion of pre-mRNA escapes the RNAi effect. In addition, we believe that gene dosage is an important aspect of lin-26 biology in that its correct function depends on specific levels of LIN-26 activity. The precise requirement for LIN-26 in certain cells might explain the very particular phenotypes seen in viable lin-26 alleles (n156 and ga91). For instance, LIN-26 activity in the hypodermal cells of homozygous n156 animals allows them to reach adulthood but does not allow the Pn.p cells to adopt a hypodermal fate, resulting in a vulvaless phenotype (Fergusonet al. 1987). Furthermore, when n156 is in trans to lin-26 null alleles, LIN-26 activity is reduced to a level that results in larval lethality (Ferguson and Horvitz 1985; Labouesseet al. 1994).
On the basis of the relatively low abundance of the lir-1(short)/lin-26 pre-mRNA, we hypothesize that when the pre-mRNA of an operon exists it provides a target for multi-gene RNAi. Although many operons have been predicted from the physical map, only the few discussed in this work have been biochemically and genetically characterized. For this reason, further testing of our prediction with other operons is not feasible. The recently dissected operon, mes-6/cks-1, was shown to accumulate pre-mRNA. However, contrary to our hypothesis, RNAi against the upstream gene (mes-6) does not interfere with the downstream gene (cks-1; Korfet al. 1998). How can we explain this apparent contradiction? The mes-6/cks-1 operon is one of three operons recently classified as belonging to a new group of operons (Hengartner and Horvitz 1994) where the genes are only a few nucleotides apart, so that 3′-end formation of the upstream transcript and trans-splicing of the downstream transcript are competing processes (Williamset al. 1999). This competition results in a single molecule of pre-mRNA being processed to produce either the upstream transcript or the downstream transcript, but not both. We suggest that, since trans-splicing of the downstream transcript occurs preferentially to 3′-end formation of the upstream transcript, enough cks-1 mature transcript is processed to carry out cks-1 function.
Our results allow us to make several important predictions concerning the mechanism of RNA interference. First, showing that small dsRNA molecules are less potent than long molecules suggests that RNAi efficiency depends on the length of sequence homology between the dsRNA and the target RNA molecules. It is very likely that RNAi utilizes a number of cellular enzymes to ultimately degrade the transcripts that have been targeted in a sequence-specific manner by the injected dsRNA. Length dependence could reflect that as the dsRNA gets longer more cofactors (for instance, nucleases) are recruited to degrade the target RNA at multiple positions. Second, demonstrating that dsRNA is able to target pre-mRNA provides the first direct evidence that RNAi can target transcripts in the nucleus. Thus our results show that the cellular proteins involved in RNAi must be located in the nucleus, at least. This conclusion is consistent with the observations that the smg RNA surveillance system is not essential for RNAi in C. elegans (Montgomeryet al. 1998) and that PTGS in plants, a phenomenon possibly related to RNAi, does not involve ribosomes (Holtorfet al. 1999). Third, showing that, for a given length, intronic lin-26 dsRNA is less potent than exonic lin-26 dsRNA supports the hypothesis that a certain proportion of pre-mRNA escapes intron-mediated RNAi by RNA splicing, or that pre-mRNA is partially protected against RNAi (for instance, by splicing factors). An alternative explanation for the reduced potential of intronic sequences could be that RNAi is predominantly active in the cytoplasm. In other species, there is mounting evidence that RNAi is active mostly or only in the cytoplasm. On the basis of the observation that RNAi does not take place in Trypanosoma brucei if pre-mRNA processing is drug inhibited, Ngo et al. (1998) have suggested that RNAi can occur only in the cytoplasm or that the pre-mRNA is protected. In Drosophila, the successful targeting of maternal RNA, which is localized in the cytoplasm of the syncytial blastoderm, indicates that RNAi occurs in the cytoplasm (Kennerdell and Carthew 1998; Misquitta and Paterson 1999). In C. elegans, the experiments performed by Montgomery et al. (1998) show that RNAi completely prevents the accumulation of transcripts in the cytoplasm and, partially, in the nucleus. On the basis of our study and results from different species, we suggest that the cellular cofactors involved in RNA interference must exist both in the nucleus and in the cytoplasm, where they might be more active (or more abundant).
Now that the C. elegans genome has been sequenced (TheC. ELEGANS Consortium 1998), the easiest and fastest way to ascertain a gene's function is by RNAi. Since ~25% of C. elegans genes are organized in operons (Zorioet al. 1994), it is probable that there will be other operons in which targeting of the pre-mRNA will also produce phenotypes not specific for the gene being tested. We have so far referred to the fact that lir-1 is the upstream gene and lin-26 is the downstream gene, but there is no evidence to suggest that the same result would not occur if the genes were in the opposite orientation. Finally, we suggest that interpretation of RNAi results should be carefully considered and, where possible, corroborated before being accepted as fact, at least for genes organized in operons.
Acknowledgments
We are grateful to Andy Fire for the gift of plasmid pPD93.97. We thank Bernard Boulay for pictures, Grégoire Michaux, Sophie Quintin, Nick Skaer, James Stévenin, and Uwe Strähle for critical reading of the manuscript. J. M. Bosher was supported by an European Molecular Biology Organization fellowship. This work was supported by funds from the CNRS, INSERM, Hôpital Universitaire de Strasbourg, the Human Frontier Science Program Organization, the Association pour la Recherche sur le Cancer, the Groupement de Recherche et d'Etudes sur le Génome, the CNRS Genome Program, and the Ministère de la Recherche.
Footnotes
-
Communicating editor: R. K. Herman
- Received June 1, 1999.
- Accepted July 12, 1999.
- Copyright © 1999 by the Genetics Society of America