While dsRNA feeding in C. elegans is a powerful tool to assess gene function, current protocols for large scale feeding screens result in variable knockdown efficiencies. We describe an improved protocol for performing large scale RNAi feeding screens that results in highly efficacious and reproducible knockdown of gene expression.
RNA interference by feeding worms bacteria expressing dsRNAs has been a useful tool to assess gene function in C. elegans. While this strategy works well when a small number of genes are targeted for knockdown, large scale feeding screens show variable knockdown efficiencies, which limits their utility. We have deconstructed previously published RNAi knockdown protocols and found that the primary source of the reduced knockdown can be attributed to the loss of dsRNA-encoding plasmids from the bacteria fed to the animals. Based on these observations, we have developed a dsRNA feeding protocol that greatly reduces or eliminates plasmid loss to achieve efficient, high throughput knockdown. We demonstrate that this protocol will produce robust, reproducible knock down of C. elegans genes in multiple tissue types, including neurons, and will permit efficient knockdown in large scale screens. This protocol uses a commercially available dsRNA feeding library and describes all steps needed to duplicate the library and perform dsRNA screens. The protocol does not require the use of any sophisticated equipment, and can therefore be performed by any C. elegans lab.
Historically, genetic screens in C. elegans have been performed by treating animals with a chemical mutagen such as ethyl methanesulfonate to create random mutations within DNA. Progeny or grandprogeny of mutagenized animals are then isolated that exhibit a desired abnormal phenotype. The mutation responsible for causing the phenotype is then identified through a labor-intensive mapping procedure or by sequencing the mutant worm’s entire genome. There are many advantages to performing such chemical mutagenesis screens as they create permanent lesions within the worm genome that can result in both gain-of-function and loss-of-function phenotypes, but the mapping procedure is slow and expensive.
Double-stranded RNA interference (dsRNAi) provides an alternative means to identify genes involved in important biological processes. In this method, dsRNA matching the coding sequence of an endogenous gene is introduced into the worm. The dsRNA is processed in vivo to generate small (21-22 nucleotide) RNAs that serve as guides to find and bind the matching endogenous mRNAs, targeting them for destruction 1. This procedure is relatively quick, but because dsRNA targets mRNA rather than DNA, only reduction-of-function phenotypes are observed using this approach.
Introduction of dsRNAs into an animal can be achieved by direct injection of dsRNA 2, by soaking animals in dsRNA 3, or by feeding animals bacteria that express dsRNA 4. Each of these delivery methods cause systemic knockdown effects as most cells of the animal express SID-1, a transmembrane channel capable of importing dsRNA 5. Originally used as a reverse genetics approach to knockdown the expression of individual genes one at a time, the development of two feeding libraries now permits large-scale genetic screens. The commercially-available dsRNA libraries contain bacterial clones representing either 55% or 87% of all C. elegans genes 6, 7.
Unfortunately, large scale dsRNAi feeding screens often result in variable knockdown efficiencies 8-10. While the absolute level of knockdown by RNAi is not easily measured, the reduced knockdown efficiency observed in large scale screens must be caused by residual gene-specific mRNAs that escape degradation by RNAi. This, in turn, could be a direct result of dsRNA plasmid loss from bacteria fed to animals in these screens. Supporting this idea, animals fed mixtures of bacteria, in which only 50% of the bacteria express dsRNAs against a targeted gene, show dramatically reduced (if any) knockdown effects when compared to control fed animals 11. The extent of gene knockdown becomes a critical concern when performing dsRNAi screens as most genes in C. elegans are haplosufficient and thus gene expression must be reduced by at least 50% to cause loss-of-function phenotypes 12.
We have used previously published feeding protocols to perform large-scale screens and found the same variable knockdown effects reported by others 8-10. In a search for possible causes, we found that nearly 60% of the bacteria fed to animals in the screen had lost the dsRNA plasmid. We have developed a library duplication and screening protocol that dramatically reduces or eliminates plasmid loss and greatly improves knockdown efficiency. This protocol is suitable for performing large scale screens in C. elegans and can be used to screen either one of the commercially available dsRNA libraries. Using this protocol, we find that essentially 100% of the bacteria fed to animals during the screen retain the dsRNA-encoding plasmid and cause robust and highly penetrant loss of function phenotypes in all tissues examined.
Note: This protocol can be used to identify genes required for a variety of biological processes in a variety of tissue types. As a necessary quality control during the screen, animals are fed both positive and negative control dsRNA-expressing bacteria to demonstrate RNAi effects in the targeted tissue.
Other published feeding protocols grow dsRNA-expressing bacteria in the presence of either 50 μg/ml ampicillin or 25 μg/ml carbenicillin 8-10. We have found that bacteria grown under either of these conditions lose the dsRNA plasmid at high frequency. Indeed, we have found that even when bacteria are grown overnight in very high concentrations (up to 2 mg/ml) of ampicillin more than 50% of the bacteria no longer contain plasmid. While growth in carbenicillin generally resulted in higher retention of plasmid, 25 μg/ml was not sufficient to prevent plasmid loss. Instead, we found that carbenicillin concentrations of 500 μg/ml were required to block plasmid loss during growth in liquid cultures and 2 mg/ml carbenicillin was required to block plasmid loss when bacteria were grown on solid agar substrates.
Because plasmid retention is critical to the success of knockdown, only carbenicillin is used when dsRNA bacteria are cultured in this protocol. We have carefully optimized the concentration of carbenicillin used in the bacterial growth cultures described in this protocol to ensure that bacteria that do not contain plasmid do not grow in these cultures. However, as a quality control measure, plasmid loss is determined at several steps of the protocol. To ensure efficient knockdown, more than 80% of the bacteria at each of these steps must contain dsRNA plasmid. If more than 20% of the bacteria at any step of the protocol have lost the dsRNA plasmid, check the quality of the carbenicillin used. Carbenicillin should be prepared fresh the day it is used. Prepare fresh carbenicillin and repeat the culture.
1. How to Determine Plasmid Loss
2. Making Duplicate Library Plates
Overview: The dsRNA library will arrive from the manufacturer (OpenBiosystems) as 10,566 bacterial clones in 96 well plates. Store the plates immediately in -80 °C freezer. The original library plates will be duplicated and all feeding screens will use the duplicate library plates as the source of bacteria. It is advisable to store the original and duplicate library plates in separate freezers. For ease of handling, duplicating only 10 library plates per day is recommended.
Note: We have noticed that a significant percentage of bacteria found in each well of the original library plates do not contain plasmid. However, because the library bacteria are a precious resource it is not recommended to test for plasmid loss in the original library cultures.
3. Making Temporary Copies of the Library on Omniplates
Overview: To begin to prepare food for a dsRNA screen, bacteria from duplicate library plates are transferred to solid agar omniplates and grown overnight. To prevent plasmid loss on solid media, omniplates contain carbenicillin at 2 mg/ml. The bacteria on these plates can be used for a period of one week, and show no plasmid loss. We have not tested for plasmid loss in bacteria grown on omniplates longer than one week.
4. Preparation of Bacteria for Feeding Worms
Overview: Bacteria for feeding worms are prepared as 1 ml liquid cultures (in a 96 well format) using bacteria from the omniplates. These liquid cultures are grown overnight to generate saturated nonlogarithmic bacterial cultures. Overnight growth ensures that all cultures will contain similar concentrations of bacteria and therefore all worms will be fed the same amount of food. No plasmid loss during this growth should be observed.
5. Generating L1 Arrested C. elegans Larvae
Overview: L1-arrested larvae are used to start synchronized worm cultures on the dsRNA feeding plates. These larvae are generated by bleaching populations of gravid adults to isolate healthy eggs and then allowing the eggs to hatch in S-complete media without food. In the absence of food the hatched L1 larvae will arrest development, generating a synchronous population of L1 larvae. During the screen, prepare fresh L1-arrested larvae every week as these starved animals become sick over time and must be discarded. The choice of genetic background of the animals used in a dsRNA feeding strain can vary depending on application, for neuronal screens eri-1; lin-15B mutants are recommended.
note: In steps 5.8-5.11 worms and eggs are exposed to a bleach solution. Bleach will destroy worms and leave eggs relatively unharmed. However, if eggs are left in bleach solution too long, they will also be destroyed. Set a timer when bleach is added in step 5.8 to be certain that eggs do not remain exposed to bleach for more than 10 min.
Arrested L1 larvae should be made fresh every week during a screen. Once the first batch has been made, subsequent batches can be started by adding 500 arrested L1 larvae (from the previous week’s batch) to each of four standard, seeded 6 cm plates (containing 100 μl overnight culture of OP50), and incubated at 20 °C for 4 days. On the fourth day the plates should contain many eggs and gravid adults and can be bleached to prepare fresh eggs for more synchronized L1 larvae.
6. Transferring Arrested L1 Larvae onto Feeding Plates
Overview: To perform a dsRNA screen, 20-30 L1 arrested larvae are transferred onto dsRNA feeding plates. Early knockdown effects such as larval arrest or lethality will be observed after animals have fed on dsRNA-expressing bacteria for 2-3 days. Depending on the screen conducted, desired phenotypes may be observed either in the original animals placed on the feeding plate (P0 animals) or in their progeny (F1 animals). The presentation of phenotype will depend on a number of variables including the perdurance of the protein encoded by the gene whose expression is knocked down and the time during development in which the gene of interest is required. Starting dsRNA feeding cultures with only 20-30 animals should permit observation of both P0 and F1 phenotypes before animals exhaust the food source.
To easily transfer 20-30 animals to feeding plates, first determine the concentration of L1 arrested animals in the arrested L1 culture.
7. Begin the Feeding Screen
Overview: To begin the dsRNA screen, 20-30 arrested L1 larvae are added to each well of 12-well feeding plates containing a thin lawn of dsRNA-expressing bacteria. Feeding plates contain standard nematode growth media (NGM) and are supplemented with 500 μg/ml carbenicillin (to prevent plasmid loss) and 1 mM IPTG (to induce dsRNA expression). Animals are allowed to feed and develop for several days at 20 °C. Two typical feeding regimens for animals are 3 days when performing a P0 screen and 6-7 days when performing an F1 screen. The duration of feeding, however, can be varied depending on the specific phenotypes sought in the screen.
note: The bacteria that remain on the feeding plate after 7 days of worm growth have been tested for plasmid loss and nearly 100% retained the dsRNA plasmid.
P0 knockdown phenotypes will begin to appear after 2-3 days of feeding animals dsRNA expressing bacteria. Some early phenotypes include larval arrest and lethality and should be observed in 2-3% of all feeding wells. These same phenotypes will also be observed in the F1 generation animals. The presence of F1 eggs that fail to hatch is another common F1 phenotype, and together, these three phenotypes will be observed in nearly 10% of all wells in F1 screens.
We used the protocol described here to examine knockdown effects for several genes expressed in a variety of tissue types for both P0 and F1 phenotypes. Because we also wanted to test for knockdown effects in neuronal tissue, we used eri-1; lin-15B mutants in our screens. Mutations in eri-1 and lin-15B, along with mutations in rrf-3, were identified in screens for mutations that caused enhanced sensitivity to RNAi 8,13,14.
We tested four genes for knockdown: egl-30, dpy-17, pat-10, and unc-4. Two of these genes, egl-30 and unc-4, are expressed in the nervous system 15-17, one gene, pat-10, is expressed in body-wall muscle 18, and the fourth gene, dpy-17, is expressed in hypodermal cells 19. When we fed eri-1; lin-15B animals bacteria that contained the empty pL4440 plasmid but did not express a dsRNA, we did not observe any morphological or behavioral defects (Figure 2A, Table 1).
egl-30, encodes the C. elegans ortholog of the G protein subunit Gaq. egl-30 null mutants arrest during early larval development, but some null escapers and hypomorphic loss-of-function egl-30 mutants grow to become adult stage and are defective in locomotion and egg-laying behaviors 20. When we fed L1 stage eri-1; lin15B larvae bacteria expressing egl-30 dsRNA, we found that the larvae grew to become adults that showed clear defects in locomotion and egg-laying behavior. After three days, 100% of the animals fed dsRNA against egl-30 were paralyzed and unable to lay eggs (Figures 2B, Table 1). We did not observe the larval arrest phenotype in our dsRNA knockdown experiments that is observed in egl-30(ad810) null animals. This is presumably because the L1 animals we plated onto egl-30 dsRNA bacteria had already passed the early developmental requirement for EGL-30. This highlights an advantage of feeding screens: late-stage phenotypes can be observed for genes that also play vital roles during development.
dpy-17 encodes a collagen protein required for cuticle formation in C. elegans 19. dpy-17 null mutants are shorter and fatter than wild-type animals. We did not observe any morphological defects in P0 animals fed dpy-17 dsRNA. However, 98% of the F1 animals showed the Dpy phenotype (Figure 2C, Table 1). The lack of short and fat P0 animals indicates that dpy-17 gene function is required at early larval stages for proper body morphology. While dpy-17 has not been targeted in other feeding screens, it was knocked down by dsRNA injection 21. Interestingly, only 30% of the progeny of dsRNA injected animals showed the Dpy phenotype, suggesting that the feeding protocol described here is more efficient than the more labor-intensive injection approach, at least for some genes.
pat-10 encodes body wall muscle troponin C, a protein containing four EF Hand motifs that bind calcium 18. We found that 100% of eri-1; lin-15B animals fed pat-10 dsRNA became paralyzed within 3 days and laid no eggs leading to a lethal phenotype (Figure 2D, Table 1).
unc-4 encodes a homeodomain protein expressed in ventral cord motor neurons and is required for proper synaptic input choice 22,23. We chose unc-4 as a test gene because it has proven to be resistant to previous dsRNA feeding strategies 8,24. unc-4 mutants show a defect in locomotion behavior. As a result of defects in synaptic connectivity, unc-4 mutants are unable to back 22,23. Compared to wild-type animals that back freely in a sinusoidal motion when prodded on the head, unc-4 null mutants do not back and instead contract their midbody tightly, causing a dorsal flexure which often becomes so extreme that the head and tail of the animal touch, placing the body in a coiled position. We found that 68% of animals fed unc-4 dsRNA-expressing bacteria from our library preparation showed defects in backing. This is a dramatic improvement in knockdown efficiency when compared to the 0% backing defect observed when rrf-3 animals are fed unc-4 dsRNA using previously published protocols 8 (Table 1).
Figure 1. Flow chart for library duplication and large scale screen. Each 96 or 12 well plate generated during the screening protocol is indicated, along with the method used to transfer bacteria between plates. Asterisks indicate the steps at which bacteria should be tested for plasmid loss. Click here to view larger figure.
Figure 2. dsRNA knockdown phenotypes of C. elegans genes expressed in various tissue types. A) Control animals fed bacteria containing pL4440 empty vector. Black arrow indicates the position of a young adult animal. White arrow indicates the position of a group of laid eggs. Photograph shows freely moving animals as evidenced by their normal body posture. B) Animals fed dsRNA against egl-30. Note that all animals have adopted a rigid appearance typical of paralyzed animals. Also note the complete absence of laid eggs on the plate. C) Animals fed dsRNA against dpy-17. Note that adult animals in the field (one marked by arrow) are shorter and fatter than the adult animal shown in panel A. D) Animals fed dsRNA against pat-10. Note that all animals are paralyzed but can still move their head muscles to feed as indicated by the clearing of bacteria near the head, marked by arrow. Scale bar 1 mm.
dsRNA plasmid or gene targeted by feeding | Tissue expression of targeted gene | Percent animals with terminal phenotype |
pL4440 (negative control) | wildtype for all behaviors | |
egl-30 | nervous system | 100% Egl and Paralyzed |
dpy-17 | hypodermis | 98% Dpy |
pat-10 | body wall muscle | 100% Paralyzed |
unc-4 | nervous system | 68% Backing Defect |
Table 1. Terminal phenotype of animals fed dsRNA against genes expressed in various C. elegans tissues. n = 100 for all genes.
We have described a protocol that uses a commercially available dsRNA feeding library to perform large-scale screens in C. elegans. We provide all instructions to duplicate the library and to use it effectively. We demonstrate that this approach is able to identify genes involved in different biological processes in different tissues of the animal. While the phenotypes we describe here are easily observable (Unc, Egl, and Dpy), this protocol can be adapted to search for genes required for nearly any biological process. For example, we have used this protocol to identify genes required for neurotransmission in the worm 25. Once animals have fed on bacteria expressing dsRNA (typically 3 days for P0 screens, 6-7 days for F1 screens), they can be removed from the feeding plate and tested for any behavioral phenotype.
Plasmid loss:
We found that the reduced knockdown efficiencies observed during large-scale dsRNA screens could be directly attributed to the loss of dsRNA plasmid from the bacteria fed to the worms. Plasmid loss occurs in bacterial growth cultures because β-lactamase, encoded on the dsRNA library plasmid, acts to degrade both ampicillin and carbenicillin, reducing the concentration of these antibiotics over time. When the concentration of antibiotic becomes sufficiently low, bacteria that do not contain the plasmid can survive and populate the culture. The use of such mixed bacterial cultures in dsRNA screens must be avoided, as they show reduced knockdown activity and often do not cause loss-of-function phenotypes 11. Surprisingly, we found that high levels of plasmid loss occur even in bacterial cultures grown in very high concentrations of ampicillin (up to 2 mg/ml). While carbenicillin is more resistant to degradation by β-lactamase, we still found it necessary to use concentrations of carbenicillin that were 10-40 fold higher than those used in previously-published dsRNA feeding screens 8-10,26 to ensure that all bacteria in the culture retained the dsRNA plasmid.
Because most genes in C. elegans are haplosufficient, it is critically important that all bacteria fed to animals express dsRNA. This ensures highly efficient gene knockdown and increases the probability of observing loss-of-function phenotypes in large-scale screens.
Positive and negative controls:
It is critical to include both positive and negative control bacteria in each feeding plate when performing screens. The negative control is particularly important if behavioral tests will be used to identify genes of interest. For a negative control we use bacteria containing the empty vector pL4440. For a positive control, it is best to use bacteria that express dsRNA against a gene that causes the desired knockdown phenotype. However, if no such genes are known, a positive control for knockdown should be used that causes an easily observable phenotype. When choosing such a gene, preference should be given to genes that are expressed in the tissue type targeted in the screen. For example, in dsRNA feeding screens for neuronal phenotypes, a neuronal gene should be chosen as a positive control. An observable knockdown effect in animals feeding on the positive control bacteria will indicate that the dsRNA interference is working in the desired tissue type.
It is best if the person scoring the feeding plates for phenotypes is blind to the location of the positive control wells. The screener should be able to easily identify the positive well in each 96-well plate. For behavioral assays it is helpful to score the negative well first before scanning the remaining wells.
Throughput:
Using this protocol, one person should be able to set up and screen 16, 12-well plates per day. To move efficiently through the library, we generally test each well of each library plate only once during a screen. Any wells that scored as positive are recorded and are retested in triplicate. We require that the positive wells retest in all three retests. The plasmid is then purified from the bacteria and the insert is sequenced to positively identify the gene. If null mutants are available for the gene, we will order them and test the null animals for the behavioral defect identified in the screen.
The authors have nothing to disclose.
This work was supported by funds from the National Institutes of Health MH097163. Some strains were provided by the CGC, which is funded by the NIH Office of Research Infrastructure Programs (P40-OD010440).
Reagent/Material | |||
96-well Falcon flat bottom plate | Fisher Scientific | 353072 | sterile |
adhesive foil | USA Scientific | 2923-0100 | |
Nunc omniplate | Fisher Scientific | 267060 | |
2.0 ml deep 96-well polypropylene Masterblock | USA Scientific | 5678-0285 | autoclavable |
Lids for deep well plates | USA Scientific | 5665-6101 | |
50 ml combitips | USA Scientific | 4796-5000 | individually wrapped |
Reservoir | USA Scientific | 2977-8500 | autoclavable |
12-well plates | USA Scientific | CC7682-7512 | individually wrapped |
dsRNA feeding library | OpenBiosystems | RCE1181 | store -80 °C |
Ampicillin | Fisher Scientific | BP1760-5 | |
Tetracycline | Fisher Scientific | BP912-100 | |
Carbenicillin | allow 2-4 weeks for delivery www.biochemicaldirect.com | ||
Bacteriolgical Agar Ultrapure grade A | Affymetrix | 10906 | for worm plates |
Equipment | |||
Brayer roller | USA Scientific | 9127-2940 | |
Boekel 96-pin replicator | Fisher Scientific | 05-450-9 | autoclavable |
microshaker | Thomas Scientific | 1231A93 | 3 mm orbit |
12 channel multichannel pipet (0.5-10 μl) | USA Scientific | 7112-0510 | |
12 channel multichannel pipet (30-300 μl) | USA Scientific | 7112-3300 | |
Repeater Plus manual pipettor | USA Scientific | 4026-0201 |