Here we describe a protocol that is an adaptable, whole host, high-content screening tool that can be utilized to study host-pathogen interactions and be used for drug discovery.
The number of new drugs identified by traditional, in vitro screens has waned, reducing the success of this approach in the search for new weapons to combat multiple drug resistance. This has led to the conclusion that researchers do not only need to find new drugs, but also need to develop new ways of finding them. Amongst the most promising candidate methods are whole-organism, in vivo assays that use high-throughput, phenotypic readouts and hosts that range from Caenorhabditis elegans to Danio rerio. These hosts have several powerful advantages, including dramatic reductions in false positive hits, as compounds that are toxic to the host and/or biounavailable are typically dropped in the initial screen, prior to costly follow up.
Here we show how our assay has been used to interrogate host variation in the well-documented C. elegans—Pseudomonas aeruginosa liquid killing pathosystem. We also demonstrate several extensions of this well-worked out technique. For example, we are able to carry out high-throughput genetic screens using RNAi in 24- or 96-well plate formats to query host factors in this host-pathogen interaction. Using this assay, whole genome screens can be completed in only a few months, which can dramatically simplify the task of identifying drug targets, potentially without the need for laborious biochemical purification approaches.
We also report here a variation of our method that substitutes the gram-positive bacterium Enterococcus faecalis for the gram-negative pathogen P. aeruginosa. Much as is the case for P. aeruginosa, killing by E. faecalis is time-dependent. Unlike previous C. elegans—E. faecalis assays, our assay for E. faecalis does not require preinfection, improving its safety profile and reducing the chances of contaminating liquid-handling equipment. The assay is highly robust, showing ~95% death rates 96 h post infection.
The identification and development of effective, broad-spectrum antibiotics, now almost a century ago, led to a watershed moment in public health where there was a wide-spread belief that infectious disease would be a scourge of the past. Within a few short decades, this optimism began to wane, as pathogen after pathogen developed resistance mechanisms that limited these once miraculous treatments. For some time, the arms race between drug discovery efforts and the pathogens seemed balanced. However, the misuse of antimicrobials has recently culminated in the emergence of pan-drug resistant strains of Klebsiella pneumoniae, Acinetobacter baumanii, Serratia marcescens, and P. aeruginosa1,2,3,4.
P. aeruginosa is an opportunistic, gram negative, multi-host pathogen that is a severe threat to patients with severe burns, those who are immunocompromised, or have cystic fibrosis. It is also increasingly identified as a causative agent in severe nosocomial infections, particularly due to its ongoing acquisition of antimicrobial resistance. To begin to address this threat, we have used the well-documented C. elegans–P. aeruginosa infection system5. Our lab has leveraged this system to develop a liquid-based, high-throughput, high-content screening platform to identify novel compounds that limit the ability of the pathogen to kill the host6. Intriguingly, these compounds seem to belong to at least three general categories, including antimicrobials7 and virulence inhibitors8. Other high-content drug discovery assays in C. elegans have been reported for Mycobacterium tuberculosum, Chlamydia trachomatis, Yersinia pestis, Listeria monocytogenes, Francisella tularensis, Staphylococcus aureus, Candida albicans, and Enterococcus faecalis, among others9,10,11,12,13,14,15,16. These types of assays have several well-recognized advantages, such as limiting false positive hits that may be toxic to both the host and the pathogen, increased likelihood of bioavailability compared to a chemical screen, and the ability to identify hits beyond simply limiting microbial growth, such as anti-virulents, immune stimulatory molecules, or compounds that otherwise tilt the balance of the host-pathogen interaction in favor of the former. Additionally, the compounds discovered in these screens are often effective in mammalian hosts.
It is worth noting that at least two other assays17,18 are available to carry out high-throughput screens in C. elegans in liquid. However, each of these assays is a modification that allows the prototypical intestinal-colonization assay, known as slow-killing, to be performed in liquid, increasing throughput and allowing compounds to be more readily screened. Careful characterization has conclusively demonstrated that the mechanisms of bacterial virulence are different between these assays and our liquid-based screen7. Since both types of virulence are observed in mammalian systems, it is important to consider which virulence determinant is most relevant for the experimenter's interests prior to assay selection.
Here we demonstrate an optimized version of the liquid-based C. elegans-P. aeruginosa assay. We also report the adaptation of our liquid-based assay method to accommodate the gram-positive bacterial pathogen Enterococcus faecalis. Like P. aeruginosa, E. faecalis is increasingly identified as a serious nosocomial threat with a growing armament of antimicrobial resistance pathways1. Although a previous method for high-throughput screening of E. faecalis exists14, it requires preinfection with the pathogen, which complicates the procedure and increases the likelihood of contaminating equipment like the COPAS FlowSort. Our protocol eliminates the need for pre-infection, improving the safety profile. Finally, we report a means by which either of these assays can be combined with feeding RNAi, allowing the user to search for host factors that play a role in the establishment of, or resistance to, infection.
Caution: P. aeruginosa and E. faecalis are Biosafety Level 2 pathogens, and proper safety precautions must be taken to prevent accidental infection and to minimize contamination of surfaces. All media and materials that come into contact with pathogens must be sterilized and/or discarded. Further guidelines are available from the CDC publication Biosafety in Microbiological and Biomedical Laboratories (BMBL), 5th edition.
1. Preparation and Maintenance of P. aeruginosa
2. Preparation of RNAi Bacteria
3. Maintenance and Preparation of C. elegans
Note: Before initiating experiments, generate a synchronized population of gravid, adult hermaphroditic worms as follows.
4. Liquid Killing Assay Setup (Basic Protocol)
Note: This is a protocol for one bacterial strain and one source of worms.
5. Adaptation for Screening Multiple C. elegans Strains or Knockdowns (RNAi Screen Setup)
Note: This is an RNAi screen described for a 24-well plate setup.
6. Adaptation for E. faecalis
Note: Only the differences from P. aeruginosa assay are described.
Important parameters for assay performance
A proper understanding of the biology underlying this assay is necessary for troubleshooting and optimizing the assay. To that end, we refer first to several key papers elucidating the mechanisms of pathogenesis of P. aeruginosa-mediated killing in liquid7,20. Provided that the steps outlined above are followed (see Figure 1 for a schematic of the assay protocol) a time-dependent killing of C. elegans will be observed only in the presence of the pathogen (Figure 2A). In contrast, in the absence of key nutritional supplements (e.g., if the peptone is left out of the media, and only S Basal is added), little to no killing will be observed (Figure 2B). Interestingly, the two-step incubation of P. aeruginosa (for 24 h at 37 °C, then 24 h at 25 °C) which is critical for conventional slow-killing assays, and was originally implemented in Liquid Killing21, is dispensable in this assay (Figure 2C). While it is possible to add P. aeruginosa straight from overnight LB culture, doing so significantly changes lethality kinetics and is not recommended.
Understanding assay biology also permits simplification of the assay, if appropriate and desirable. For example, the most important driver of host killing in this assay is the siderophore pyoverdine, and host toxicity caused by pyoverdine is contingent upon its ability to bind iron7. As such, the assay can be simplified by substituting pyoverdine-rich filtrate, purified pyoverdine, or even some synthetic iron-chelating chemicals (e.g., 1,10-phenanthroline) for live bacteria, as the transcriptional response of C. elegans to these treatments is very similar (Figure 3)22.
Several different lines of evidence speak to the robustness of the experimental setup. First, the setup tolerates a wide range of initial bacterial concentrations. Concentrations as low as OD600 = 0.0025 (approximately 10-fold lower than recommended) still exhibit time- and concentration-dependent killing, although the timing does shift (Figure 4A). Second, the reproducibility of the assay is such that as few as 4 wells is frequently sufficient to obtain statistically significant data (Figure 4B).
There are several changes that we do not recommend. For example, using initial bacteria inocula with concentrations much higher than those listed in the protocol; doing so results in thick, biofilm-like material that is difficult or impossible to effectively wash away, which complicates scoring lethality. Furthermore, high-concentration bacterial inocula also can compete for oxygen and trigger non-specific host killing7.
Another important note is to limit the period of time between stopping the assay and imaging the dead worms. Note that this time frame must include staining, so it is critical to be efficient. After death, the biological material within worms begins to extrude and/or be consumed by bacteria. Within a matter of 24 – 32 h, only the cuticle remains. The cuticle stains very poorly and is very light, making it easy to lose during washes. It is important to note that strains or strain/RNAi conditions with very different kinetics of killing can be complicated by this phenomenon (i.e., some worms may still be alive while others have already lost their content and are impossible to image). Figure 4A shows this phenomenon, as worms at the left side of the plate, inoculated with very low initial bacterial concentrations, are still largely alive while worms on the right side of the plate, which were exposed to much higher concentrations of bacteria, have been washed away.
A very important determinant of assay success is the method used to collect and analyze the data. In our lab, we have used at least three methods for collecting data from these assays: spectrophotometry, automated microscopy, and flow vermimetry (i.e., the adaptation of flow cytometry techniques to C. elegans; literally, the measurement of worms in a flowing solution). The first of these methods uses a spectophotometer to read the fluorescence of the dye or reporter of the worms in question. This method has the advantage of using a fairly ubiquitous piece of equipment (a standard spectrophotometer with a microplate reader) and is the fastest method to acquire data. However, it lacks the informational content available from automated microscopy and the statistical power of flow vermimetry.
Automated microscopy is another viable option. A number of microplate imaging systems are currently on the market, ranging in prices that are affordable to a single investigator (e.g., a Bio-Tek Cytation5) to larger, more expensive and higher-quality machines that are more commonly found in core facilities (e.g., a Molecular Devices ImageXpress Microscope). In practice, most of these solutions are amenable for scoring most screens and assays. Most automated imaging platforms can also be coupled to downstream software for image processing (e.g., MetaMorph, ImageJ, Gene5, or Cell Profiler) to further the yield of information. Examples of the utility of processing are shown in Figure 5 and Figure 6. When the signal-to-noise ratio is high (as in Figure 5),analysis is simple and discriminating between positive and negative conditions is trivial. In these cases, even weak hits can be readily identified. Many assays, however, have a weaker signal-to-noise ratio. For example, treatment of PINK-1::GFP22 worms (with constitutive mCherry expression in their pharynges) with 1,10-phenathroline increases the level of PINK-1::GFP. For data analysis, the inducible GFP reporter is normalized to the constitutive mCherry signal (Figure 6). In this case, the reporter has weaker expression and/or is not activated in the majority of the worms. Therefore, implementing additional image-processing tools, like Cell Profiler, that can reduce background and amplify signal may be advantageous.
Finally, if a COPAS FlowPilot is available, it can be used to acquire data via flow vermimetry. Much like the more familiar concept of flow cytometry, this instrument can be used to measure the fluorescence of C. elegans in at least two or three channels at a time. This method is very amenable to the acquisition of data with high statistical significance, but the throughput is very much diminished compared to automated microscopy. The most significant advantage of flow vermimetry is in the ability to analyze whole worm populations as a single group for a replicate, rather than splitting them into multiple wells and analyzing the average of the wells. Analyzing large groups in this fashion dramatically improves the statistical power and allows the detection of even small effects.
Modifications of the Liquid Killing Assay
Recent developments of the assay have extended its usefulness by including assaying activation of fluorescent reporters (Figure 6), using medium- to high-throughput RNAi techniques (Figure 3), and even the substitution of the original pathogen.
The most obvious reason to use RNAi set up is to test for host defense pathways against P. aeruginosa (or other pathogens). Examples of RNAi knockdown of genes relevant for survival in the assay are shown in Figure 3. Consistent with previous observations20,23,24,25, daf-2(RNAi) enhanced C. elegans survival during exposure to P. aeruginosa, while daf-16(RNAi), zip-2(RNAi) and cebp-1(RNAi) shortened it. As noted, RNAi is most simply included in the assay growing worms on multiwell plates where each well contains a different RNAi clone. Due to the high viscosity of the agar and small volumes involved, it is difficult to achieve uniform, bubble-free filling of 96-well plates without specialized equipment. In addition, drying the bacterial strains carrying the RNAi can also be an issue, as the outer wells dry faster than inner wells, leading to non-uniformity and significant potential for agar cracking. As noted above, this encourages worms to burrow, reducing the number of animals available for screening and risks clogging liquid-handling machinery. For both of these reasons, 24-well plates are easier to work with than 96-well plates. Although this reduces assay throughput, it improves reliability, and is suggested, particularly for initial screens.
Finally, the assay has been modified the assay to replace P. aeruginosa with E. faecalis. (In principle, substitution with other bacterial species should not provide undue difficulties, provided that appropriate culturing and exposure conditions are identified.) The most important factor to consider is the media requirements of the pathogen, both for growth and development and for the induction of virulence. For example, the Gram positive, opportunistic pathogen E. faecalis requires nutrient-rich media (like BHI) and incubation at higher temperatures compared to P. aeruginosa14,26. By taking these factors into account, adaptation of the the assay to use E. faecalis was straightforward. As noted for P. aeruginosa, we saw time-dependent killing (Figure 7). Interestingly, the timing of death was similar to previously published assays that used pre-infection on agar plates26, despite the absence of this step, which may suggest that the mechanisms involved in pathogenesis vary.
Figure 1: Scheme for liquid-based C. elegans – P. aeruginosa infection assay. Steps involving nematode propagation and preparation are at left side, bacterial preparation are at right. Steps involving both are centered. Please click here to view a larger version of this figure.
Figure 2: Killing of C. elegans by P. aeruginosa is specific and requires proper bacterial nutrition. (A) C. elegans death in the presence of E. coli and P. aeruginosa. (B) C. elegans death in the presence of P. aeruginosa with (right) and without (left) Slow Kill media added to mix (i.e., S basal only). S Basal is a minimal media and does not contain the nutrients required for bacterial growth and precludes the production of pyoverdine. (C) C. elegans death in the presence of differently prepared P. aeruginosa. *p <0.01, based on Student's t-test. Error bars represent S.E.M. 10 wells were used per biological replicate per condition. Please click here to view a larger version of this figure.
Figure 3: C. elegans resistance to P. aeruginosa depends on host immune pathways. A panel of selected RNAi knockdowns was treated with either P. aeruginosa (A) or 1,10-phenanthroline (a chemical chelator that mimics exposure to P. aeruginosa in liquid) (B). *p <0.01, #p <0.05, based on Student's t-test. Error bars represent S.E.M. 10 wells were used per biological replicate per condition. Please click here to view a larger version of this figure.
Figure 4: C. elegans killing by P. aeruginosa is robust and depends on bacterial concentration. (A-B) Representative images (A) and quantification (B) of C. elegans death after exposure to varying concentrations of P. aeruginosa for 72 h. p-values were calculated based on Student's t-test. Scale bar in (A) is 1 mm. 16 wells were used per biological replicate per condition. BF: Bright Field, FL: Fluorescence. Please click here to view a larger version of this figure.
Figure 5: Data acquisition and image processing of an assay with a strong signal. (A) Representative images of live (negative control) and dead (positive control) C. elegans. (B) Statistical analysis of negative and positive controls based on data acquisition method, processing mode, and number of wells analyzed. (C) Fluorescence was analyzed for worms from four individual wells of a 96-well plate (two positive and two negative control wells). Each well contained ~100 worms. Each dot on the graph represents a single worm, showing the time-of-flight (which correlates with but, due to the coiling of living worms, imperfectly represents size) and the measured fluorescence for worms that are stained with Sytox Orange. Positive control worms were pre-killed by heat exposure. Due to changes in tension and shape of dead worms (as can be seen in (A), dead worms appear more rod-like and have few body bends) compared to their living counterparts, dead worms have a longer TOF. p-values were calculated based on Student's t-test. Scale bar in (A) is 1 mm. Please click here to view a larger version of this figure.
Figure 6: Data acquisition and image processing of an assay with a moderate signal. (A) Representative images of C. elegans carrying an inducible PINK-1::GFP reporter and an mCherry constitutive marker (to normalize expression of genes from the extrachromosomal array). C. elegans exposed to DMSO (negative control) or 1,10-phenanthroline (positive control). (B) Statistical analysis of negative and positive controls based on data acquisition method, processing mode, and number of wells analyzed. (C) Four individual wells of a 96-well plate (two positive and two negative control wells) were analyzed using flow vermimetry. p-values were calculated based on Student's t-test. Scale bar in (A) is 1 mm. Please click here to view a larger version of this figure.
Figure 7: C. elegans killing by E. faecalis is robust and time dependent. (A-B) Representative images (A) and quantification (B) of C. elegans death after exposure to E. faecalis. *p <0.01, based on Student's t-test. 10 wells were used per biological replicate per condition. Error bars represent S.E.M. Scale bar in (A) is 1 mm. Please click here to view a larger version of this figure.
This assay (or similar assays where other pathogens are substituted for P. aeruginosa or E. faecalis) is useful for a variety of purposes, including drug discovery. It is also useful for addressing fundamental biological questions, such as identifying virulence factors, the elucidation of host defense pathways, and determining the regulatory machinery involved in the host-pathogen interaction.
Although the P. aeruginosa Liquid Killing assay is robust, there are several points where careful attention should be paid to minimize the chance of failure. First, as noted, the bacterial source plate for P. aeruginosa inoculations must remain fresh, lest the bacteria lose virulence, despite overnight growth in LB. Second, it is key to make sure that calcium and magnesium are added to P. aeruginosa killing assays. Without it, virulence will be significantly compromised. Finally, it is worth noting that worms reared on HT115 (for RNAi-based assays) will die significantly later than worms reared on OP50 (N. Kirienko, personal communication). For this reason, if switching to RNAi-based studies, it is important to empirically determine the best time point for the assay.
For killing assays with either P. aeruginosa or E. faecalis, it is crucial to ensure that there is adequate food for the development of the worms from the L1 stage until they are ready to be used in the assay. Starvation, even in the short-term, is known to activate a number of host defense pathways, such as the DAF-2/DAF-16 insulin/IGF signaling network27. Our data suggest that DAF-16/FOXO is already slightly activated by being cultured in liquid20, and further activation is strongly undesirable, since DAF-16/FOXO promotes broad-spectrum pathogen resistance23.
Finally, it is crucial to use completely sterile worms in any assay that relies on mortality as a readout. If worms are fertile, being in liquid causes difficulties in egg laying. As a result, some of the embryos hatch within the parent, eventually causing its death. For this reason, we generally use a genetic lesion, such as glp-4(bn2)28, that demonstrates a completely penetrant sterile phenotype for these assays. The use of chemicals that prevent embryonic development but still allow egg laying, such as 5-fluorodeoxyuridine (FUdR), should be avoided as they may also interfere with bacterial growth and development.
In general, we have found it quite helpful to plan several time points for analysis. Although the assay is robust and the timing is generally consistent, stochastic and imperceptible changes can shift the timing of death in the assay within 2 – 4 hours. When troubleshooting is necessary, it is often helpful to consider first the simplest of answers; i.e., prepare fresh media, discard the most recent bacterial cultures, and re-streak bacterial strains from frozen stocks, etc. Only very rarely do these measures not restore the assay to functionality.
Due to the relative simplicity of the method, a wide variety of modifications can easily be performed. Changing the worms from a uniform glp-4(bn2) background for high-throughput RNAi screens, such as those we describe herein, is useful for the identification of host response pathways for a wide range of xenobiotics and toxic chemicals. For example, by adding the same chemical or toxin into each well of RNAi plates (e.g., Cry5B toxin, phenanthroline, etc.), pathways that alter the sensitivity to these toxins can be readily identified. These types of screens can also be used to identify defense networks against any of the panoply of pathogens that infect C. elegans.
The authors have nothing to disclose.
This study was supported by the Cancer Prevention and Research Institute of Texas (CPRIT) Award RR150044, Welch Foundation Research Grant C-1930, and by the National Institutes of Health K22 AI110552 awarded to NVK. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
COPAS FP BioSorter | Union Biometrica | Large object flow cytometer/worm sorter | |
Cytation 5 | BioTek | ||
EL406 Washer Dispenser | BioTek | ||
Multitron Pro | Infors HT | ||
24 Deep-Well RB Block | Thermo Fisher Scientific | CS15124 | |
384-Well plate | Greiner Bio-One | MPG-781091 | |
Nematode Growth Media (NGM) | Amount per liter: 18 grams agar, 3 grams NaCl, 2.5 grams Peptone, 1 mL CaCl2 (1 M), 1 mL MgSO4 (1 M), 25 mL Phospate buffer, and 973 mL of milli-Q water | ||
Slow Killing (SK) plates | Amount per liter: 18 grams agar, 3 grams NaCl, 3.5 grams Peptone, 1 mL CaCl2 (1 M), 1 mL MgSO4 (1 M), 25 mL Phospate buffer, and 973 mL of milli-Q water | ||
Slow Killing (SK) media | Amount per liter: 3 grams NaCl, 3.5 grams Peptone, 1 mL CaCl2 (1 M), 1 mL MgSO4 (1 M), 25 mL Phosphate buffer, and 973 mL of milli-Q water | ||
Lysogeny Broth (LB) | USBiological Life Sciences | L1520 | |
Brian Heart Infusion broth (BHI) | Research Products International Corp | 50-488-526 | |
Worm Bleach Solution | Amount per 100 mL: 10 mL of 5 M NaOH solution, 20 mL of 5% Sodium Hypochlorite Solution, and 70 mL of sterile water | ||
S Basal | Amount per liter: 5.85 grams NaCl, 6 grams KH2PO4, 1 gram K2HPO4, and 1 Liter of milli-Q water | ||
Agar | USBiological Life Sciences | A0930 | |
NaCl | USBiological Life Sciences | S5000 | |
Peptone | USBiological Life Sciences | P3300 | |
CaCl2 | USBiological Life Sciences | ||
MgSO4 | Fisher Scientific | M63-500 | |
Phospate buffer | amount per liter: 132 mL of K2HPO4 (1M) and 868 mL of KH2PO4 (1M) | ||
KH2PO4 | Acros Organics | 7778-77-0 | |
K2HPO4 | USBiological Life Sciences | P5100 | |
5% Sodium Hypochlorite Solution | BICCA | 7495.5-32 | |
NaOH solution | Fisher Scientific | SS255-1 | |
Breathe-easy | Diversified Biotech | BEM-1 | |
SYTOX Orange Nucleic Acid Stain | Fisher Scientific | S11368 | |
Bacterial Strains | |||
P. aeruginosa (PA14) | |||
E. faecalis(OG1RF) | |||
E. coli superfood (OP50) | |||
E. coli RNAi expressing bacteria (HT115) | |||
Worm Strains | |||
glp-4(bn2) (Beanan and Strome, 1992, PMID: 1289064) | |||
PINK-1::GFP reporter (Kang et al., 2018, PMID: 29532717) |