Gene 263 (2001) 103±112

www.elsevier.com/locate/gene

Ingestion of bacterially expressed dsRNAs can produce speci®c and

potent genetic interference in Caenorhabditis elegans
Lisa Timmons a, Donald L. Court b, Andrew Fire a,*
a
Department of Embryology, Carnegie Institution of Washington, 115 West University Parkway, Baltimore, MD 21210, USA
b
National Cancer Institute, Gene Regulation and Chromosome Biology Laboratory, ABL-Basic Research Program,
Frederick Cancer Research and Development Center, Frederick, MD 21702-1201, USA
Received 22 September 2000; received in revised form 11 November 2000; accepted 30 November 2000
Received by D.I. Friedman

Abstract
Genetic interference mediated by double-stranded RNA (RNAi) has been a valuable tool in the analysis of gene function in Caenorhabditis
elegans. Here we report an ef®cient induction of RNAi using bacteria to deliver double-stranded RNA. This method makes use of bacteria
that are de®cient in RNaseIII, an enzyme that normally degrades a majority of dsRNAs in the bacterial cell. Bacteria de®cient for RNaseIII
were engineered to produce high quantities of speci®c dsRNA segments. When fed to C. elegans, such engineered bacteria were found to
produce populations of RNAi-affected animals with phenotypes that were comparable in expressivity to the corresponding loss-of-function
mutants. We found the method to be most effective in inducing RNAi for non-neuronal tissue of late larval and adult hermaphrodites, with
decreased effectiveness in the nervous system, in early larval stages, and in males. Bacteria-induced RNAi phenotypes could be maintained
over the course of several generations with continuous feeding, allowing for convenient assessments of the biological consequences of
speci®c genetic interference and of continuous exposure to dsRNAs. q 2001 Elsevier Science B.V. All rights reserved.
Keywords: RNAi; HT115(DE3); dsRNA; T7 polymerase

1. Introduction The manner in which an organism responds to nucleic

acid encountered in food is of interest from biological and
A variety of species exhibit a defense response in which a technical perspectives. The predominant component of the
double stranded RNA (dsRNA) `trigger' produces a prema- C. elegans diet in the laboratory is bacteria; therefore, intro-
ture loss of endogenous RNAs with extended regions of duction of dsRNA into the C. elegans diet has involved
sequence identity to the trigger (see Fire, 1999 for a review). engineering bacteria to produce dsRNA (Timmons and
A striking feature of dsRNA-triggered genetic interference Fire, 1998). Although speci®c interference for several
processes (RNAi) in C. elegans is the ability of the inter- genes was evident in earlier studies with dsRNA-producing
ference to `spread' to cells that are some distance from the bacteria (Timmons and Fire, 1998), the observed pheno-
initial site of dsRNA delivery. The ®rst indication of this types were limited in penetrance and expressivity. While
spreading effect came from the observation that dsRNA demonstrating that ingested dsRNA is competent to trigger
injected into the body cavity produced interference pheno- interference, these initial experiments left unresolved the
types throughout the injected animal (Fire et al., 1998). A question of whether loss-of-function phenotypes could be
later and more dramatic demonstration of a spreading effect effectively produced by ingested dsRNA. To address this
was observed after feeding animals food that contained question, we sought to increase the effectiveness of the
dsRNA (Timmons and Fire, 1998) or after soaking animals bacterial feeding technique by several means, including
in dsRNA (Tabara et al., 1998). modi®cations that increase the concentration of dsRNA in
the bacterial cell. We have found that a bacterial strain
Abbreviations: dsRNA, double stranded RNA; EDTA, ethylenedinitrilo lacking the dsRNA-speci®c endonuclease RNaseIII can be
tetraacetic acid; GFP, green ¯uorescent protein; IPTG, isopropyl b-d-thio- cultured to produce high levels of speci®c dsRNA and that
galactopyranoside; LB, Luria-Bertani medium; NGM, Nematode Growth these bacteria can effectively trigger strong and gene-speci-
Medium; RNase, ribonuclease; RNAi, RNA-mediated interference ®c phenotypic responses when fed to C. elegans.
* Corresponding author. Tel.: 11-410-554-1234; fax: 11-410-243-6311.
E-mail address: ®[email protected] (A. Fire).

0378-1119/01/$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0378-111 9(00)00579-5
104 L. Timmons et al. / Gene 263 (2001) 103±112

2. Materials and methods and gene sequences, respectively, of the green ¯uorescent
protein from Aequorea victoria.
2.1. C. elegans strains The L4440 (pPD129.36) cloning vector contains two
convergent T7 polymerase promoters in opposite orienta-
All C. elegans strains were derived from the wild-type N2 tion separated by a multicloning site and was made using
strain of Brenner (Brenner, 1974). Three transgenic gfp standard cloning techniques (Timmons and Fire, 1998).
strains were used in experiments for which interference L4417 (pPD128.110) has 0.7 kb of GFP coding sequence
with gfp was tested: PD4251 (ccIs4251) is homozygous inserted between two convergent T7 promoters.
for a myo-3::gfp transgene insertion (Fire et al., 1998); L4431 (pPD129.17) contains two copies of the 0.8 kb GFP
VH41 is homozygous for an unc-119::gfp transgene inser- coding region in tail-to-tail inverted orientation, separated
tion (rhIs13) (a gift from Dr H. Hutter, Max Planck); and by 0.8 kb of unc-22 gene. A T7 promoter and T7 transcrip-
PD8047 harbors an extrachromosomal array with a let- tional terminator ¯ank the gfp inverted repeat.
858::gfp reporter (ccEx8047) (Kelly et al., 1997). him pLT61 contains 0.8 kb of unc-22 inserted into the L4440
strains used to produce males were: him-1(e879) I, him- vector.
3(e1256) IV, him-5(e1467) V, him-5(e1490) V, him- pLT76 contains two copies of a 0.6 kb unc-22 sequence in
6(e1423) IV, him-8(ay29) IV, him-10(e1511) III, and him- tail-to-tail orientation separated by 0.9 kb of gfp and vector
14(it44) II (Hodgkin et al., 1979; Kemphues et al., 1988). sequences. A T7 promoter and a T7 transcriptional termi-
nator ¯ank the unc-22 inverted repeat.
2.2. Bacterial strains L4366 (pPD128.18) contains 0.8 kb of unc-22 and a short
linker (from gfp) ¯anked by T3 and T7 promoters.
Bacterial strains used were: L4400 (pPD128.117) contains two copies of a 0.6 kb unc-
BL21: ompT hsdSB (rB2mB2) gal dcm (Studier et al., 1990; 22 sequence in head-to-head orientation separated by 0.8 kb
Tabor and Richardson, 1985; Wood, 1966). of gfp and vector sequences. A T7 promoter and a T3
W3110: IN(rrnD-rrnE)1 (Dasgupta et al., 1998; Takiff et promoter ¯ank the unc-22 inverted repeat.
al., 1989). pLT62 contains 0.8 kb of the unc-54 gene inserted into the
SDF204: W3110, rnc 1 TD1-17::DTn10(Dasgupta et al., L4440 vector.
1998; Takiff et al., 1989). pLT63 contains 0.8 kb of the fem-1 gene inserted into the
HT115: W3110, rnc14::DTn10(Dasgupta et al., 1998; L4440 vector.
Takiff et al., 1989). pLT64 contains 1.2 kb of the hlh-1 cDNA inserted into the
DH5aF 0 : f80dlacZDm15D(lacZYA-argF)U169 recA1 L4440 vector.
endA1 hsdR17(rK2mK1) deoR supE44 thi-1 gyrA96 Plasmids were transformed into bacterial hosts using
relA1{F 0 } (Life Technologies). standard CaCl2 transformation protocols and plated on anti-
HB101: mcrB mrr hsdS20 (rB2mB2) leuB6 supE44 ara14 biotic-containing LB-agar plates. All these plasmids
galK2 lacY1 proA2 rpsL20(Sm r) xyl-5 mtl-1 recA14) contained the b-lactamase gene. 75 mg/ml of ampicillin
(Boyer and Roulland-Dussoix, 1969). was included in plates and in liquid media. For XL1-Blue
TG1: supE thi-1 D(lac-proAB) D(mcrB-hsdSM)5 (rK2mK2) MRF 0 , SDF204, and HT115 strains, tetracycline was used at
{F 0 traD36 proAB 1 lacI qZDM15} (Stratagene). 12.5 mg/ml in both media and in plates.
SCS110: rpsL (Str 0 ) thr leu endA thi-1 lacY galK galT ara
tonA tsx dam dcm supE44 D(lac-proAB) {F 0 traD36 proAB 2.4. dsRNA feeding conditions
lacI qZDM15} (Stratagene).
XL1-Blue MRF 0 : D(mcrA)183 D(mcrCB-hsdSMR- A 2 ml 2 £ YT culture containing appropriate antibiotics
mrr)173 endA1 supE44 thi-1 recA1 gyrA96 relA1 lac {F 0 (see above) was inoculated with a single colony of host cell
proAB lacI qZDM15 Tn10 (Tet r)} d (Stratagene). 1 plasmid and cultured overnight at 378C. The culture was
Tn10 confers tetracycline resistance to the SDF204, diluted more than 100 fold and allowed to grow to
HT115, and XL1-Blue MRF 0 strains. Strains were modi®ed OD600 ˆ 0.4. Isopropyl-b-d-thiogalactopyranoside (IPTG)
to express T7 RNA polymerase from an IPTG-inducible was added to a ®nal concentration of 0.4 mM, and the
promoter by site-speci®c integration into the chromosomal culture was incubated with shaking for 2±4 h at 378C.
attB site of a lDE3 derivative carrying an IPTG-inducible After supplementing with additional ampicillin, tetracy-
copy of the phage gene 1 encoding T7 polymerase (Studier cline, and IPTG, the cells were either directly applied onto
and Moffatt, 1986). agar plates or were concentrated by centrifugation and then
applied. Unless otherwise noted, plates were composed of
2.3. Plasmids standard NGM/agar media (Brenner, 1974) supplemented
with 50±100 mg/ml ampicillin, 12.5 mg/ml tetracycline (if
Plasmids were constructed using standard cloning tech- appropriate), and 0.4 mM IPTG. Worms were added to
niques. Intermediate cloning steps were performed in plates individually; alternatively, agar chunks containing
DH5a 0 . GFP and gfp abbreviations refer to the protein many worms were added. RNAi phenotypes were observ-
 L. Timmons et al. / Gene 263 (2001) 103±112 105

able within 1±5 days, depending on the target gene. Pheno- (Timmons and Fire, 1998). Descriptions of plasmids and
types were monitored in the F1 progeny of the animals bacterial genotypes are provided in Section 2. To simplify
placed on the plate. Plates contained suf®cient quantities the descriptions of the bacterial strains used in the feeding
of bacteria to support the nematode growth during the protocols, we have employed the following terminology for
course of the experiments, which lasted from one to ®ve this paper: HT115±an RNaseIII mutant bacterial host;
days. Freshly seeded plates were used in all the experiments HT115(DE3)±host strain harboring a lDE3 lysogen (source
described here; however, seeded plates that were stored for of T7 polymerase); HT115(DE3)(L4417)±lysogenic strain
as long as two weeks at 48C also produced RNAi pheno- harboring a plasmid capable of expressing dsRNA.
types. For long-term maintenance of animals on dsRNA Two plasmid con®gurations utilizing the T7 promoter
expressing food, worms from successive generations were have proven useful in generating RNAi phenotypes
transferred to fresh plates. At no time during dsRNA admin- (Timmons and Fire, 1998). (i) A bi-directional promoter
istration were the animals depleted of food. con®guration consists of two T7 promoters ¯anking a single

2.5. Quantitation of dsRNA produced in bacteria

Cell culture: Competent cells were prepared using stan-

dard CaCl2 methodology and were transformed with plas-
mid pLT61 or pLT76. Single colonies were inoculated into
2 £ YT media containing appropriate antibiotic and cultured
overnight. The bacterial cultures were then diluted 1:500 in
2 £ YT 1 antibiotic(s) and were incubated until all samples
reached OD600 ˆ 0.4. (Faster growing cultures were periodi-
cally diluted so that all cultures were equally turbid at time
of induction as quantitated by OD600.) T7 polymerase was
induced by the addition of 0.4 mM IPTG and the cells were
cultured for an additional 4 h. Final cell numbers for each
sample were calculated based on OD600.
Analysis of dsRNA produced in bacterial strains: Total
nucleic acids were extracted by lysing cell pellets (above) in
1 M ammonium acetate/10 mM EDTA plus phenol:chloro-
form at 658C; precipitating nucleic acids using ethanol;
removing DNA with DNAse I; and removing single-
stranded RNA with bovine pancreatic RNase in 25 mM
Tris, pH 7.5/50 mM NaCl/5 mM MgCl2/0.5 mM dithioery-
thritol. The nucleic acid pellets were resuspended in 10 mM
Tris/1 mM EDTA, pH 7.5 and loaded onto a 1% agarose/ Fig. 1. Quanti®cation of dsRNA produced in different bacterial strains.
TAE gel, stained with ethidium bromide, and photographed. Bacteria of the indicated genotypes were lysogenized with lDE3, trans-
Sample volumes were normalized with respect to cell formed with plasmids designed to express unc-22 dsRNA, grown in liquid
number after IPTG induction (calculated from OD600 read- media, induced with IPTG, and processed for total nucleic acid (see Section
ings of induced cells in culture). Approximately 10 9 cell 2). Following treatment with RNaseA (to remove single stranded RNA),
samples were resolved by electrophoresis on agarose, and nucleic acid was
equivalents were loaded onto each lane of the gel. visualized by staining with ethidium bromide. Top: Diagram showing two
distinct plasmid con®gurations. Lanes 1±4: cells harbored pLT61, which
contains an 887 base segment (composed of unc-22 and polylinker
3. Results and discussion sequences) ¯anked on each side by convergently oriented T7 promoters.
Lanes 5±8: cells harbored pLT76, which contains a single T7 promoter
3.1. Assays for bacteria-induced RNA interference driving a hairpin segment consisting of a 624 base stem (composed of
unc-22 sequences), an 864 base loop (composed of gfp and unc-22
In each experiment, an engineered bacterial strain was sequences), and a transcriptional terminator sequence `TER' from T7
gene 10B (Studier et al., 1990). Sample volumes were normalized with
provided as the sole source of food for C. elegans. The
respect to cell number. (Similar results were obtained with samples normal-
strains had been transformed with a single plasmid designed ized with respect to induction volume.) The rnc alleles used are: BL21:
to express a speci®c segment of dsRNA. The bacterial food rnc 1; W3110: rnc 1; SDF204: rnc 1; HT115: rnc 2. Substantially greater
was provided to worms on standard nematode culture plates levels of dsRNA were produced from the HT115(DE3) (rnc 2) host. Arrows
(1 additives) (Brenner, 1974). Interference phenotypes indicate the major RNase/DNase resistant bands (presumed dsRNA). DNA
size markers are shown at the left (slight differences in migration between
were monitored in the F1 progeny of animals placed on
equivalently sized dsDNA and dsRNA on these gels have been observed;
the modi®ed food. In general, the levels of interference LT and AF, unpublished). The dsRNA isolated from HT115(DE3) cells was
obtained with modi®ed bacterial strains were compared to also biologically active for RNAi; i.e. injection of this dsRNA into N2
the levels seen using the original BL21(DE3) strain worms produced a twitching phenotype (data not shown).
106 L. Timmons et al. / Gene 263 (2001) 103±112

copy of a segment from the target coding sequence, and (ii) 1975; Hughes et al., 1987; Gottesman et al., 1982; Guar-
A hairpin con®guration consists of a single T7 promoter neros et al., 1982; Houlberg et al., 1983; Hyman and Honig-
driving an inverted repeat structure that has been interrupted man, 1986). In addition, RNaseIII has positive and negative
by a non-homologous spacer DNA. These con®gurations are effects on the translational ef®ciency of speci®c mRNAs (a)
summarized in Fig. 1 (top). by inducing conformational alterations through site-speci®c
cleavages, (b) by altering the stability of certain mRNAs,
3.2. Improved effectiveness of bacteria-induced RNAi using and (c) by removing sequences that stimulate or inhibit
modi®ed bacterial strains ribosome binding (Portier et al., 1987; Takata et al., 1987;
Dunn and Studier, 1975; Court, 1993). Despite the relative
Escherichia coli strain BL21(DE3), used in the initial importance of the reactions catalyzed by RNaseIII, null
development of bacteria-induced RNAi (Timmons and mutations in rnc are compatible with growth (Takiff et al.,
Fire, 1998), was originally chosen to take advantage of an 1989).
IPTG-inducible T7 RNA polymerase gene contained within We tested whether bacterial RNaseIII activity affects
a stable insertion of a modi®ed lambda prophage lDE3 bacteria-induced RNAi by using a congenic set of bacterial
(Studier et al., 1990; Boyer and Roulland-Dussoix, 1969). strains, each with a different mutation in rnc: HT115 carries
The BL21 host had been optimized for over-production of an rnc null mutation, and SDF204 is a control strain contain-
recombinant proteins (Studier et al., 1990; Tabor and ing a wild-type rnc allele (Dasgupta et al., 1998; Takiff et
Richardson, 1985; Wood, 1966). For example BL21 is de®- al., 1989). These strains were derived from W3110. We
cient in at least two bacterial proteases: Lon and OmpT made each strain lysogenic for lDE3, and these lysogenic
(Wood, 1966; Phillips et al., 1984; Grodberg and Dunn, host strains were then transformed with plasmids designed
1988). We thus had no a priori reason to assume that to produce dsRNA corresponding to gfp or to endogenous C.
BL21 might be optimal for RNA production. Furthermore, elegans genes. The dsRNA was produced from both plasmid
we considered that the protease de®ciency, and perhaps con®gurations described previously (Section 3.1): double
other properties of this strain, might render it sub-optimal promoter and hairpin con®gurations.
for over-production of dsRNA. We therefore tested several The dsRNA produced in these bacteria was analyzed by
readily available E. coli strains not de®cient in the Lon and preparation of total RNA followed by treatment with
OmpT proteases for interference (DH5a 0 , TG1, SCS110, pancreatic RNaseA to remove single-stranded material.
XL1-Blue MRF 0 , W3110, and HB101). In each case, we We found that the rnc 2 cells [HT115(DE3)(pLT61) or
made the strain lysogenic for lDE3, and then transformed HT115(DE3)(pLT76)] were most effective in accumulating
with a plasmid (L4417) designed to produce gfp dsRNA. extended dsRNA duplexes, while rnc 1 strains produced
Although these strains yielded comparable degrees of gfp variable amounts of lower molecular weight dsRNAs (Fig.
interference to that seen with BL21(DE3)(L4417), we 1).
observed slightly improved interference in some experi- Fig. 2 shows the relative effectiveness of each strain in
ments with the bacterial strain W3110(DE3)(L4417). mediating RNAi against a gfp transgene. We de®ne full inter-
Subsequent optimizations have made use of strains derived ference as elimination of GFP ¯uorescence from .95% of
from W3110. cells in gfp transgenic animals. (This level of interference is
We reasoned that the yield of interfering RNA might be observed in progeny animals after injection of gfp dsRNA
improved if degradation of dsRNA in the bacterial host into myo-3::gfp adults (Fire et al., 1998)). When myo-3::gfp
could be minimized. We further reasoned that this might transgenic animals were raised on BL21(DE3)(L4417),
be accomplished by removing (by genetic mutation) BL21(DE3)(L4431), or W3110(DE3)(L4417) bacteria,
double-strand-speci®c RNases from the bacterial cell. A fewer than 15% of the adults in the population exhibited
dsRNA-speci®c endonuclease (RNaseIII) is encoded by full interference (Fig. 2A,G,C). In contrast, .98% of
the bacterial rnc gene, and mutations in this gene are avail- the adult progeny of this transgenic worm strain showed
able. RNaseIII plays a prominent role in the processing and full interference when reared on rnc 2 strains
degradation of a subset of bacterial RNAs (Dasgupta et al., HT115(DE3)(L4417) and HT115(DE3)(L4431) (Fig. 2E,I).
1998; Takiff et al., 1989; Crouch, 1974; Gegenheimer and This RNAi phenotype was dependent on T7 polymerase
Apirion, 1981). A major role of bacterial RNaseIII is to expression in the bacterial cell, as animals fed
initiate the maturation of rRNA from the 30S precursor HT115(L4417) cells lacking the lDE3 lysogen showed no
RNA (Bram et al., 1980; Sirdeshmukh and Schlessinger, loss of GFP ¯uorescence (Fig. 2B). Bacterial strains
1985). The enzyme has also been implicated in the post- SDF204(DE3)(L4417) and SDF204(DE3)(4431) were also
transcriptional processing of approximately 10% of endo- less effective in the feeding protocol: most of the animals did
genous cellular mRNAs (Dennis, 1984; King et al., 1986; not have a reduction in overall ¯uorescence intensity, and
Portier et al., 1987; Takata et al., 1987; Talkad et al., 1978), only a few animals had some muscle nuclei with no GFP
of transgenic over-expressed mRNAs (Gitelman and Apir- expression (Fig. 2D,H).
ion, 1980), and of transcripts from bacteriophages T7, T3, The relative effectiveness of the modi®ed feeding proto-
and lambda (Dunn and Studier, 1973; Dunn and Studier, col was quanti®ed by counting the residual, unaffected,
 L. Timmons et al. / Gene 263 (2001) 103±112 107

Fig. 2. Effectiveness of bacteria-induced RNAi on a gfp transgene. myo-3::gfp transgenic animals (PD4251) were placed as L3 larvae on the indicated bacteria
and cultured at 208C until F1 progeny reached adulthood. A±F: ¯uorescence and Nomarski images of animals fed bacteria harboring plasmid L4417 in the
con®guration depicted. G±I: ¯uorescence and Nomarski images of animals fed bacteria harboring plasmid L4431 in the con®guration depicted. Five different
bacterial hosts were used and are listed above each set of photos. Note that young larvae in this experiment are less affected by dsRNA introduced using this
method than are their older siblings (F, magni®ed portion of panel E). A dramatic reduction in GFP ¯uorescence was observed in animals fed rnc 2 strains.

GFP 1 cells in myo-3::gfp animals that had been fed either

rnc 1 or rnc 2 strains (Fig. 3). In this experiment, none of the
animals raised on BL21(DE3)(L4417) bacteria were fully
affected. This rnc 1 strain produced no more than a 60%
reduction in the number of GFP 1 cells for any animal exam-
ined, and no animal had lost GFP ¯uorescence in all cells. In
contrast, all of the animals fed the rnc 2 strain were severely
affected: six percent of the HT115(DE3)(L4417)-fed
animals had no GFP 1 cells; 80% had ®ve or fewer GFP 1
cells per animal; and the remaining 14% had between ®ve
and 17 unaffected, GFP 1 cells per animal. The ¯uorescence
Fig. 3. Quantitation of the effectiveness of bacteria-induced RNAi. intensity in these residual GFP 1 cells of animals fed
myo-3::gfp transgenic animals (PD4251) were placed as L3 larvae on the HT115(DE3)(L4417) was much reduced in comparison to
indicated bacterial hosts expressing gfp dsRNA from plasmid L4417 and that seen in BL21(DE3)(L4417)-fed animals. Interestingly,
cultured at 208C until F1 progeny reached adulthood. (OP50 is the normal the GFP 1 cells in HT115(DE3)(L4417)-fed animals were
food source for C. elegans and does not contain a T7 polymerase.) Forty
almost exclusively found in the extreme anterior or posterior
animals from each experiment were mounted for ¯uorescence microscopy,
and all GFP 1 nuclei in each animal were scored. A dramatic reduction in ends of the animals.
both the number and ¯uorescence intensity of GFP 1 nuclei was observed in A marked improvement in the RNAi response using an
animals fed HT115(DE3)(L4417). rnc 2 host strain was also seen with endogenous C. elegans
108 L. Timmons et al. / Gene 263 (2001) 103±112

Fig. 6. Improved effectiveness of feeding protocol by inclusion of antibio-

tics. Wild-type worms were fed HT115(DE3) bacteria harboring either
Fig. 4. Effectiveness of bacteria-induced RNAi on an endogenous gene. unc-22 dsRNA expressing plasmids pLT61 (®rst set of experiments) or
Wild-type (N2) animals from recently starved populations were placed onto pLT76 (bottom set of experiments). The cells were cultured, induced,
plates seeded with bacteria expressing unc-22 dsRNA derived from a plas- and plated in the presence or absence of the additives listed in the top
mid in the con®guration indicated (left column). Bacterial strains used are line of the ®gure. Additives were included in the media and culture plates
indicated at the top of the ®gure. Twitching phenotypes were scored in L4 at the concentrations described in Section 2. Twitching phenotypes were
and adult progeny in the absence of levamisole and are reported as a scored in the absence of levamisole in adult and L4 progeny. n ˆ 200 for
percentage of the total number of animals scored. As previously described Trial 1; n ˆ 100 for Trials 2 and 3.
for BL21(DE3) (Timmons and Fire, 1998), a weaker phenotype (twitching
in the presence of levamisole) was observed in many of the remaining
`unaffected' animals. Plasmids used in the experiments are indicated in absence of levamisole (Timmons and Fire, 1998; Fig. 4).
the second column of the ®gure. The T7 gene 10B transcriptional termi- With W3110(DE3)(pLT61) or W3110(DE3)(pLT76)
nator is indicated by `TER'. Plasmid promoters are indicated by arrows: strains, we saw strong twitching phenotypes in no more
triangular arrow ˆ bacteriophage T7 promoter; ¯at arrow ˆ bacteriophage
than 12% of animals (Fig. 4). The rnc 2 bacterial strains
T3 promoter (note that these bacteria have no source of T3 RNA polymer-
ase, thus no activity from the T3 promoter would be expected in these HT115(DE3)(pLT61) and HT115(DE3)(pLT76) were
cells). At least 200 animals were tallied in each experiment. most effective at inducing RNAi, producing strong pheno-
types in 100% of progeny (Fig. 4). Again the SDF204(DE3)
strain was less effective in inducing twitching with any of
genes (Figs. 4, 5, and 6). The gene unc-22 provides a parti- the plasmid con®gurations tested: fewer animals were
cularly useful assay for the effectiveness of genetic inter- affected and the number of affected animals varied from
ference, since phenotypes resulting from partial and severe experiment to experiment. The Unc-22 phenotype was
reductions in function can easily be distinguished. A severe dependent upon a source of T7 polymerase in the bacterial
reduction in function results in a strong twitching phenotype cell, as the HT115(pLT61), HT115(pLT76), and
visible under standard conditions, whereas a partial reduc- HT115(4400) strains with no lDE3 lysogen and the
tion in function [1.5±3-fold reduction in gene activity] HT115(DE3)(L4366) strain with a T7 polymerase source
results in a weak twitching phenotype visible only in the but only one T7 promoter site in the plasmid produced
presence of the acetylcholine agonist levamisole (Fire et al., few affected animals.
1998; Brenner, 1974; Moerman and Baillie, 1979). Using rnc 2 mutant bacteria also provided increased effective-
the rnc 1 strain BL21(DE3)(pLT61), we had obtained only a ness of RNAi in the feeding protocol for the target genes
partial loss-of-function phenotype±85% of adult animals unc-54, hlh-1, and fem-1 (Fig. 5). In the case of unc-54, no
twitched in levamisole and no animals twitched in the phenotype was ever observed when rnc 1 host cells were
employed. In contrast, use of the rnc 2 strain
HT115(DE3)(pLT62) in the feeding protocol gave reprodu-
cible RNAi phenotypes. The Unc-54 phenotypes produced
can be divided into three categories: strong (phenotypes
similar to the null unc-54 phenotype±complete paralysis
and egg-laying defects (Brenner, 1974)); intermediate
(phenotypes similar to strong hypomorphic unc-54
mutants±incomplete paralysis with severe movement
defects (Wills et al., 1983)); and no phenotype (wild-type
movement). Approximately half the affected progeny fed
Fig. 5. Effectiveness of bacteria-induced RNAi in generating Unc-54, Fem- HT115(DE3)(pLT62) exhibited a strong phenotype as
1, and Hlh-1 phenotypes. Wild-type N2 animals from a recently starved adults, while a similar number of adults exhibited an inter-
plate were fed bacterial cells harboring a dsRNA-expressing plasmid. (The mediate phenotype (Fig. 5).
bacterial strains used are indicated in the top row; the plasmid con®guration
hlh-1 is a gene whose function is required during late
and C. elegans gene insertions are indicated in the left column.) Plasmids
used in the experiments are indicated in the second column. At least 200 embyronic stages (Chen et al., 1994). Injection of dsRNA
animals were scored in each experiment. The phenotypes were scored in into adults produces a strong loss-of-function phenotype in a
both adult and L4 progeny of the initially transferred animals. majority of progeny (Fire et al., 1998). No phenotypically
 L. Timmons et al. / Gene 263 (2001) 103±112 109

affected animals were obtained following growth on conferred a selective growth disadvantage. It would thus be
BL21(DE3)(pLT64) cells, while phenotypes were observed expected that procedures which allow extended periods of
using HT115(DE3)(pLT64) cells (Fig. 5). The phenotypes T7 polymerase induction would result in selection for cells
induced by feeding can be divided into three categories: that have lost the capacity to over-express dsRNA. Such
strong (Lumpy/Dumpy appearance in early larvae±indica- selection could result in the loss of T7 RNA polymerase,
tive of loss-of-function for hlh-1); intermediate (Lumpy or reversion of the rnc mutation through loss of Tn10, or loss
Dumpy appearance in larvae or adults±indicative of partial of the dsRNA-producing plasmid. These elements can be
loss-of-function of hlh-1); and no phenotypes (wild-type selected for by incorporation of antibiotics into the medium:
appearance). A modest number of animals with strong and tetracycline selects against revertant rnc 1 host cells, while
intermediate phenotypes and a large number of unaffected ampicillin selects against cells that have lost the dsRNA-
animals were obtained (Fig. 5). Affected progeny of animals expressing plasmid. The addition of either antibiotic alone
fed HT115(DE3)(pLT64) arose at a much lower frequency to the plates resulted only in slight improvements in the
(17%) than has been observed from standard genetic protocol. The combination of tetracycline 1 ampicillin
mutants or from injection of dsRNA into the parental germ- consistently and markedly improved the results obtained
line (essentially 100%). (Fig. 6). The inclusion of antibiotics to the assay also serves
In contrast to the somatically active genes described to prevent contamination from extraneous bacteria±a factor
above, the fem-1 gene is required for proper germline devel- that can negatively affect the outcome of the feeding experi-
opment (Doniach and Hodgkin, 1984). Current models for ments.
sex determination suggest that this gene is required auton-
omously in the germline (Kuwabara et al., 1998; Mehra et 3.4. Reversibility of bacteria-mediated RNAi
al., 1999). Loss-of-function fem-1 animals produce no
sperm; therefore, the animals lack fertilized embryos and We have observed that the RNAi phenotype depends upon
produce only oocytes. As with the somatically expressed the continuous presence of dsRNA-expressing bacteria on
genes described above, greater numbers of RNAi-affected the plate. Animals that twitch due to ingestion of unc-22
animals were obtained in feeding experiments using the dsRNA-expressing bacteria produce progeny with normal
rnc 2 bacteria expressing fem-1 dsRNA (Fig. 5). A fraction movement when switched to normal food. Furthermore,
of Fem-1 phenocopies were observed from feeding animals switched to normal food as young larvae fail to
BL21(DE3)(pLT63) (45%); however, the numbers of maintain the twitching phenotype. Similar results (re-expres-
affected animals increased signi®cantly when the sion of silenced genes when transferred to normal food) have
HT115(DE3)(pLT63) strain was used (88%) (Fig. 5). The been seen in animals transgenic for gfp that were fed gfp
fact that a signi®cant fraction of animals is affected in the dsRNA-expressing food. In addition, if all the unc-22
rnc 1 strain is probably a re¯ection of the sensitivity of this dsRNA- or gfp dsRNA-expressing food is depleted, most
gene to RNAi; however, we cannot rule out the possibility of the animals begin to lose the RNAi phenotypes.
that this and perhaps other dsRNAs might be more resistant
to bacterial RNaseIII. 3.5. Limitations of RNAi and comparisons of dsRNA
As shown in Figs. 2 and 4, increased interference in rnc- delivery methods
backgrounds was observed using both double promoter and
hairpin con®gurations. Although the two con®gurations For most genes, a reduction in mRNA levels to a critical
were equally effective in these examples after several days threshold is required in order for a phenotype to be
of feeding, hairpin con®gurations often induced RNAi observed. The quantitative effectiveness of phenocopy
phenotypes more quickly than a bi-directional con®guration production by RNAi for a given target gene, tissue, and
(data not shown). One interpretation of this observation is developmental stage depends on a complex set of para-
that double-strandedness may be more easily achieved when meters that determine whether or not this critical threshold
sense and anti-sense strands are covalently connected, as in is achieved. Some of these parameters are likely to depend
the inverted repeat. on the balance between dsRNA-mediated mRNA decay and
de novo mRNA synthesis. First, degradation may be
3.3. Effective feeding-based RNAi is dependent upon affected by tissue- and stage-speci®c aspects of dsRNA
conditions for bacterial growth and dsRNA induction uptake and metabolism. Second, degradation may be
affected by the availability of the RNAi machinery in a
We have tested for interference using a variety of differ- given tissue or at a particular developmental stage. Third,
ent conditions for the growth and induction of dsRNA- the natural stability of a target RNA has a signi®cant effect
producing bacteria. Although the optimal conditions are on the potency of RNAi: highly stable RNAs provide a more
likely to depend on the interfering construct and the target susceptible target for gene-speci®c mRNA decay, whereas
gene, several relevant parameters are noted below. an mRNA that is intrinsically unstable will be less affected
We have found that continuous induction of T7 polymer- (Fire, 1999). Finally, it is conceivable (although not yet
ase in the presence of plasmids capable of producing dsRNA demonstrated) that some mRNAs may be resistant to
110 L. Timmons et al. / Gene 263 (2001) 103±112

RNAi by virtue of sequence composition, secondary struc- tions of dsRNA. However, the embryonic period may be
ture, or sequestration within the cell. temporarily isolated from dsRNA delivery because the
Although the feeding protocol can be very effective in eggshell is impermeable to dsRNA or because the animal
producing a decrease in gene activity in adults and early is not feeding at this stage, and this could provide a respite
embryos, we have in general observed a less severe effec- from the effects of dsRNA. We hypothesize that lack of
tiveness in young larvae, even those larvae whose ancestors feeding-mediated interference at early larval stages is a
have been raised on dsRNA-expressing food for several consequence of the relative impermeability of the nematode
generations. Plates of GFP-expressing animals (PD4251) eggshell.
grown on HT115(DE3)(L4417) bacteria contained large Consistent with these hypotheses are two other instances
numbers of unaffected (GFP 1) L1 and L2 larvae (Fig. 2F). in which interference in older worms is more readily
Likewise, a stock of animals grown continuously on observed from dsRNA introduced by feeding than from
HT115(DE3)(pLT62) bacteria (expressing unc-54 dsRNA) dsRNA introduced by injection. In the PD4251 myo-3::gfp
contained L1 and L2 larvae with near-normal movement. strain, vulval muscle gfp expression is partially resistant to
These observations are the converse of what is seen with RNAi following parental injection (Fire et al., 1998);
injection of these dsRNAs into the parent: for both target however, gfp expression in this tissue is not resistant to
genes, effective interference is observed in the resulting gfp dsRNA when introduced by feeding (Fig. 2E,F,I; Fig.
progeny at early larval stages, while adult progeny of 3, data not shown). Likewise, egg-laying defects are not
injected animals show partial recovery of the targeted generally observed in animals injected with unc-54
genes (Fire et al., 1998). dsRNA; however, after feeding animals unc-54 dsRNA,
Delivery of dsRNA by injection provides a bolus of progeny animals exhibit this phenotype.
dsRNA to the pre-fertilized egg, some of which is distrib- We have observed that all tissues are not equally sensitive
uted throughout the embryo. This deposited dsRNA perse- to dsRNA delivered by feeding. In particular, several groups
veres through embryogenesis to produce interference at making use of RNAi have observed partial or complete
later larval stages±stages when the embryo is effectively resistance of certain genes expressed in the mature nervous
isolated from the environment by the eggshell which is system (M. Nonet, pers. Comm.; J. Fleenor, M. Hsu and A.
deposited after fertilization. We hypothesize that the loss Fire, unpublished). This resistance to RNAi is independent
of RNAi in adult progeny of injected animals is due to of the method of dsRNA delivery, as delivery of dsRNA by
simple dilution of the dsRNA trigger. (Consistent with injection, soaking, or feeding produces the same effects in
this hypothesis, interference is generally lost in the second the nervous system (Fig. 7). The fact that dsRNA can be
generation after injection.) Delivery of dsRNA by feeding delivered continuously to the worm through feeding argues
provides a more continuous delivery of smaller concentra- that the resistance seen in the nervous system is not stage-

Fig. 7. Partial resistance to RNAi of some genes in the nervous system. Animals harboring two GFP-expressing transgenes were used: a let-858::gfp reporter
transgene produces a nuclear-localized GFP-tagged protein that is expressed in all cells (Kelly et al., 1997), while an unc-119::gfp transgene produces a
cytoplasmic GFP whose expression is limited to the nervous system (personal communications from Dr M. Maduro, Dr D. Pilgrim, and Dr H. Hutter). Panels
A±D show animals carrying both transgenes. Panels E and F show animals carrying only the let-858::gfp transgene. Panel B shows progeny of animals injected
with gfp dsRNA. Panels C and F show progeny of animals raised at 208C on HT115(DE3)(L4417) bacteria expressing gfp dsRNA. Panel D shows animals
soaked as L1/L2 larvae in gfp dsRNA and subsequently analyzed as adults for GFP ¯uorescence (Tabara et al., 1998). Note the persistence of GFP in the ventral
nerve cord system in panels B, C, D, and F (arrows).
 L. Timmons et al. / Gene 263 (2001) 103±112 111

speci®c. Rather, the lack of effectiveness of dsRNA in the RNAi strategies. From a bioengineering perspective, it
nervous system may be due to limiting components of the may be possible to use a bioengineered dsRNA comestible
RNAi mechanism or to a factor destabilizing dsRNA in this to modify gene expression or block viral infection in any
tissue. organism which has a gene-speci®c dsRNA response similar
An additional difference in RNAi susceptibility by feed- to that of C. elegans. To respond to such an intervention, the
ing is observed with respect to gender. During the course of target organism would need to have both a dsRNA-response
our studies, we have noted that males appear to be less mechanism (RNAi) and a dsRNA uptake/spreading
affected by dsRNA delivered through feeding. Males are mechanism that would produce responses such as those
slower to exhibit the RNAi phenotype than hermaphrodites: observed in C. elegans.
at 208C, wild-type males exhibit the unc-22 phenotype 3±5 This work also raises the question of whether certain
days later than sibling hermaphrodites. In addition, several species might have evolved mechanisms that use stable
him strains were tested and we noted that these mutant dsRNAs to deter predatory attack. Further, such mechan-
males were also less affected than their hermaphrodite isms might conceivably have been used to speci®cally
siblings. The explanation for these results may be trivial: modulate gene expression in members of the same species
perhaps males simply eat less food. A more complicated that cohabit the same area, or to modulate gene expression
explanation may re¯ect a fundamental difference between in different species that cohabit an area.
the genders relative to components of the RNAi mechanism
or to dsRNA uptake.
These observations argue that bacteria-induced RNAi and Acknowledgements
microinjection of dsRNA will have distinctive applications
for production of mutant phenocopies. We thank N. Costantino, H. Ellis, J. Fleenor, M. Hsu, H.
Hutter, C. McGill, P. Newmark, M. Nonet, and J. Yanowitz
3.6. Consequences of long-term feeding
for their help and suggestions. This work was supported by
Wild-type N2 worms and worms expressing different PHS grants R01GM37706 (AF) and F32HD08353 (LT).
GFP-expressing transgenes (PD4251 and PD8047) have
been reared on dsRNA-expressing food for over 20 genera-
tions. No loss of viability or morphological phenotypes were References
observed (other than gene-speci®c RNAi). Thus we have not
observed any kind of response in worms that could be clas- Boyer, H.W., Roulland-Dussoix, D., 1969. A complementation analysis of
si®ed as a general `panic response' against dsRNA. the restriction and modi®cation of DNA in Escherichia coli. J. Mol.
Biol. 41, 459±472.
Bram, R.J., Young, R.A., Steitz, J.A., 1980. The ribonuclease III site ¯ank-
ing 23S sequences in the 30S ribosomal precursor RNA of E. coli. Cell
4. Conclusions
19, 393±401.
Brenner, S., 1974. The genetics of Caenorhabditis elegans. Genetics. 77,
We have shown that continuous feeding of C. elegans 71±94.
with bacteria that have been engineered to produce high Chen, L., Krause, M., Sepanski, M., Fire, A., 1994. The Caenorhabditis
levels of dsRNA can yield animals with loss-of-function elegans MYOD homolog HLH-1 is essential for proper muscle function
phenotypes for a variety of target genes. Our data suggests, and complete morphogenesis. Development 120, 1631±1641.
Court, D., 1993. RNA processing and degradation by RNaseIII. In: Belasco,
given several caveats, that an accurate indication of func- J.G., Brawerman, G.L. (Eds.), Control of mRNA Stability. Vol. 3.
tional roles of genes for which loss-of-function mutations Academic Press, New York, pp. 71±116.
are not available can be achieved by comparisons of RNAi Crouch, R.J., 1974. Ribonuclease III does not degrade deoxyribonucleic
phenotypes from dsRNA injection and feeding techniques. acid-ribonucleic acid hybrids. J. Biol. Chem. 249, 1314±1316.
We have observed speci®c limitations in RNA interference Dasgupta, S., Fernandez, L., Kameyama, L., Inada, T., Nakamura, Y.,
Pappas, A., Court, D.L., 1998. Genetic uncoupling of the dsRNA-bind-
protocols (Fire et al., 1998; this work). These include ing and RNA cleavage activities of the Escherichia coli endoribonu-
lowered sensitivity in older larvae and adults (seen in the clease RNaseIII-the effect of dsRNA binding on gene expression.
progeny of injected animals but not with continuous feed- Molec. Microbiol. 28, 629±640.
ing), lowered sensitivity in late embryonic and early larval Dennis, P.P., 1984. Site speci®c deletions of regulatory sequences in a
stages (seen with feeding but not injection), and lowered ribosomal protein-RNA polymerase operon in E. coli. J. Biol. Chem.
259, 3202±3209.
sensitivity in the mature nervous system (seen with RNA Doniach, T., Hodgkin, J.A., 1984. A sex-determining gene, fem-1, required
injection and feeding). The limitations in effectiveness of for both male and hermaphrodite development in C. elegans. Dev. Biol.
RNAi indicate substantial `room for improvement' in these 106, 223±235.
current protocols and reinforce the point that RNAi-based Dunn, J.J., Studier, F.W., 1973. T7 early RNAs and E. coli ribosomal RNAs
genetic interference can augment (but not substitute for) are cut from large precursor RNAs in vivo by ribonuclease III. Proc.
Natl. Acad. Sci. USA 70, 3296±3300.
standard loss-of-function based mutational analyses. Dunn, J.J., Studier, F.W., 1975. Effect of RNaseIII cleavage on translation
Beyond the obvious applications to functional genomics, of bacteriophage T7 messenger RNAs. J. Mol. Biol. 99, 487±499.
there are several implications for ef®cient feeding-based Fire, A., Xu, S., Montgomery, M.K., Kostas, S.A., Driver, S.E., Mello,
112 L. Timmons et al. / Gene 263 (2001) 103±112

C.C., 1998. Potent and speci®c genetic interference by double-stranded Mehra, A., Gaudet, J., Heck, L., Kuwabara, P.E., Spence, A.M., 1999.
RNA in Caenorhabditis elegans. Nature 391, 806±811. Negative regulation of male development in Caenorhabditis elegans
Fire, A., 1999. RNA-triggered gene silencing. Trends Genet. 15, 358±363. by a protein-protein interaction between TRA-2A and FEM-3. Genes
Gegenheimer, P., Apirion, D., 1981. Processing of procaryotic ribonucleic Dev. 13, 1453±1463.
acid. Microbiol. Rev. 45, 502±541. Moerman, D., Baillie, D., 1979. Genetic organization in C. elegans: ®ne-
Gitelman, D.R., Apirion, D., 1980. The synthesis of some proteins is structure analysis of the unc-22 gene. Genetics 91, 95±104.
affected in RNA processing mutants of Escherichia coli. Biochem. Phillips, T.A., VanBogelen, R.A., Neidhardt, F.C., 1984. lon gene product
Biophys. Res. Commun. 96, 1063±1070. of Escherichia coli is a heat-shock protein. J. Bacteriol. 159, 283±287.
Gottesman, M., Oppenheim, A., Court, D., 1982. Retroregulation: control Portier, C., Dondon, L., Grunberg-Manago, M., Regnier, P., 1987. The ®rst
of gene expression from sites distal to the gene. Cell 29, 727±728. step in the functional inactivation of the E. coli polynucleotide phos-
Grodberg, J., Dunn, J.J., 1988. ompT encodes the Escherichia coli outer phorylase messenger is a ribonuclease III processing at the 5 0 end.
membrane protease that cleaves T7 RNA polymerase during puri®ca- EMBO J. 6, 2165±2170.
tion. J. Bacteriol. 170, 1245±1253. Sirdeshmukh, R., Schlessinger, D., 1985. Ordered processing of Escheri-
Guarneros, G., Montanez, C., Hernandez, T., Court, D., 1982. Posttran- chia coli 23S rRNA in vitro. Nucleic Acids Res. 13, 5041±5054.
scriptional control of bacteriophage lambda int gene expression from Studier, F.W., Moffatt, B.A., 1986. Use of bacteriophage T7 RNA poly-
a site distal to the gene. Proc. Natl. Acad. Sci. USA 79, 238±242. merase to direct selective high-level expression of cloned genes. J. Mol.
Hodgkin, J., Horvitz, H.R., Brenner, S., 1979. Nondisjunction mutants of Biol. 189, 113±130.
the nematode C. elegans. Genetics 91, 67±94. Studier, F.W., Rosenberg, A.H., Dunn, J.J., Dubenforff, J.W., 1990. Use of
Houlberg, U., Hove-Jensen, B., Jochimsen, B., Nygaard, P., 1983. Identi- T7 RNA Polymerase to direct expression of cloned genes. Methods
®cation of the enzymatic reactions encoded by the purG and purI genes Enzymol. 85, 60±89.
of E. coli. J. Bacteriol. 154, 1485±1488. Tabara, H., Grishok, A., Mello, C.C., 1998. RNAi in C. elegans: soaking in
Hughes, J.A., Brown, L.R., Ferro, A.J., 1987. Nucleotide sequence and the genome sequence. Science 282, 430±431.
analysis of the coliphage T3 S-adenosylmethionine hydrolase gene Tabor, S., Richardson, C.C., 1985. A bacteriophage T7 RNA polymerase/
and its surrounding ribonuclease III processing sites. Nucleic Acids promoter system for controlled exclusive expression of speci®c genes.
Res. 15, 717±729. Proc. Natl. Acad. Sci. USA 82, 1074±1078.
Hyman, H.C., Honigman, A., 1986. Transcription termination and proces- Takata, R., Mukae, T., Hori, K., 1987. RNA processing by RNaseIII is
sing sites in the bacteriophage lambda PL operon. J. Mol. Biol. 189, involved in the synthesis of E. coli polynucleotide phosphorylase.
131±141. Mol. Gen. Genet. 209, 28±32.
Kelly, W.G., Xu, S.-Q., Montgomery, M.K., Fire, A., 1997. Distinct Takiff, H.E., Chen, S-M., Court, D.L., 1989. Genetic analysis of the rnc
requirements for somatic and germline expression of a generally operon of E. coli. J. Bacteriol. 171, 2581±2590.
expressed Caenorhabditis elegans gene. Genetics 146, 227±238. Talkad, V., Achord, D., Kennel, D., 1978. Altered mRNA metabolism in
Kemphues, K.J., Kusch, M., Wolf, N., 1988. Maternal-effect lethal muta- ribonuclease III-de®cient strains of E. coli. J. Bacteriol. 135, 528±541.
tions on linkage group II of Caenorhabditis elegans. Genetics 120, 977± Timmons, L., Fire, A., 1998. Speci®c interference by ingested dsRNA.
986. Nature 395, 854.
King, T.C., Sirdeshmuck, R., Schlessinger, D., 1986. Nucleolytic proces- Wills, N., Gesteland, R.F., Karn, J., Barnett, L., Bolten, S., Waterston, R.H.,
sing of ribonucleic acid transcripts in procaryotes. Microbiol. Rev. 50, 1983. The genes sup-7 X and sup-5III of C. elegans suppress amber
428±451. nonsense mutations via altered transfer RNA. Cell 33, 575±583.
Kuwabara, P.E., Okkema, P.G., Kimble, J., 1998. Germ-line regulation of Wood, W.B., 1966. Host speci®city of DNA produced by Escherichia coli:
the Caenorhabditis elegans sex-determining gene tra-2. Dev. Biol. 204, bacterial mutations affecting the restriction and modi®cation of DNA. J.
251±262. Mol. Biol. 16, 118±133.