Each SICKO experiment begins by age-synchronizing a population of C. elegans using standard techniques and initially maintaining animals in group culture on petri plates under standard conditions (NGM seeded with E. coli food; Fig. 1a[i]) (Porta-de-la-Riva et al., 2012). At a user-specified age, animals are exposed to bacteria modified to constitutively express a fluorescent protein (Fig. 1a[ii]). The length of exposure should be sufficient to allow gut colonization in a subset of the population and must be optimized for each bacterial strain of interest (e.g., 7 days starting at the L4 larval stage for E. coli strain OP50; 2 days starting at day 5 of adulthood for P. aeruginosa strain PA14). Following the exposure period, we move worms to petri plates containing NGM seeded with non-fluorescent bacteria for at least 16 hours to allow non-adherent bacteria to be passed from the gut and external bacteria to detach from the cuticle (Fig. 1a[iii]). Step iii ensures that the only fluorescently labeled bacteria remaining is that which has colonized the animal, as signal from non-colonizing bacteria on the animal or on the plate will generate signal that will confound the quantification of fluorescent colonies adherent to the internal tissues of the worm (Fig 1b). Worms are then transferred to individual wells in a culture environment designed to isolate animals on NGM pads seeded with non-fluorescent bacteria (Fig. 1a[iv], c), the preparation of which we recently published (Espejo et al., 2022). Finally, fluorescent images of individual free-crawling worms are captured daily, and worms are manually scored each day for survival and fleeing (Fig. 1a[v]). Once image collection is complete for an entire experiment, images are analyzed to quantify the area and intensity of fluorescent bacteria in the gut at each time point (Fig. 1a[vi]).

SICKO results in a rich dataset capturing colony progression throughout life and lifespan for each animal. The single-worm culture environment isolates individual C. elegans on small NGM pads surrounded by an aversive barrier of palmitic acid to discourage animals from fleeing their well (Fig. 1c). This distinguishes SICKO from methods that use microfluidic systems to monitor bacterial colonization in individual animals (Viri et al., 2021), allowing a more direct comparison to the majority of previous studies that use C. elegans to examine host-microbe interactions on solid media. The culture environment is enclosed in a chamber humidified with water absorbing beads, allowing for long-term cultivation of animals with repeated sampling of colony data and measurement of lifespan (Fig. 1c). The output of SICKO provides detailed information about when a bacterial colony becomes large enough to be detectable, colony size, and colony growth progression for each individual animal (Fig. 1d), in addition to insights into population-level dynamics such as infectivity and the impact of bacterial colonization and growth on host health and survival. The infection life history data for a population can be summarized in a heat map that reflects daily infection area or intensity for each animal, survival, and censoring events (Fig. 1e). Because the animals are imaged while crawling freely, the quantitative longitudinal infection data generated by SICKO comes with the current limitation of identifying the precise location of the infection within the worm. Each single-worm culture environment houses 96 animals and can be scaled using additional environments. In addition to allowing observation of the dynamic growth of individual colonies and the direct correlation of these dynamics to survival and health in the same animal, isolating individual animals also increases efficiency by allowing sample size to be precisely tuned to the needs of each experiment.

To validate SICKO, we first monitored C. elegans deficient in the p38 MAPK ortholog pmk-1, a central regulator of the C. elegans innate immune response to pathogenic bacteria (Engelmann and Pujol, 2010; Pukkila-Worley and Ausubel, 2012; Shivers et al., 2010; Troemel et al., 2006), challenged with E. coli strain OP50. E. coli OP50 is the strain most commonly used as a C. elegans laboratory food source and is thought to be mildly pathogenic. Specifically, worms fed UV-inactivated E. coli OP50 are longer-lived (Garigan et al., 2002; Gems and Riddle, 2000; Win et al., 2013), while worms fed live E. coli OP50 grown on alternative media that promotes bacterial growth are shorter-lived (Garsin et al., 2001; Mylonakis et al., 2002), than worms fed live E. coli OP50 grown on NGM. pmk-1 mutants have reduced lifespan relative to wild type when cultured on live E. coli OP50 (Alper et al., 2010).

We exposed animals to E. coli OP50 constitutively expressing green fluorescent protein (GFP) under a trc promoter for 7 days starting at the L4 larval stage and monitored infection. SICKO detected significantly higher rates of colonization in pmk-1 mutants relative to wild type (Fig. 2a, S1a). Colonization significantly correlated with death for both strains (Fig. 2a). To compare the SICKO output to cross-sectional data generated in previous studies, we simulated a cross-sectional study design by only analyzing colonization on day 1 following challenge (16 hours after wash; Fig. 1a[iii]). The proportion of each population with a detectable colony (Fig. 2b), integrated infection intensity within each animal (Fig. S1b), and area of infection within each animal (Fig. 2c) were not significantly different for pmk-1 mutants relative to wild type using this cross-sectional approach. The cross-sectional design missed ∼29% of colonies that were present but below the detection threshold one day after challenge, but later grew into detectable colonies (Fig. 2d). Using SICKO we were able to capture late-emerging colonies through longitudinal imaging, and pmk-1 animals ultimately displayed a significantly higher total colonization rate (Fig. 2d). The day when a colony first becomes detectable significantly correlated with the day of death following challenge in both the wild-type and pmk-1 mutants (Fig. 2e, f), and time between detection of a colony and the death of an animal was slightly but significantly longer in pmk-1 relative to wild type animals (Fig. 2g). In summary, pmk-1 mutant animals display a higher rate of colonization than wild type animals, though this is only apparent after allowing colonies initially below the detection threshold to progress for several days. SICKO is capable of capturing this pattern, while a single examination of cross-section colonization rates using standard timing did not.

In addition to capturing differences in dynamics of host-microbe interaction between C. elegans with different genotypes, SICKO can capture differences in colonization dynamics between different microbes. To validate this capability, we compared C. elegans challenged with the E. coli OP50 expressing GFP described above to C. elegans challenged with P. aeruginosa strain PA14, a commonly used gram-negative bacterium that is strongly pathogenic to C. elegans, constitutively expressing mScarlet under a trc promoter. As expected, animals challenged with P. aeruginosa developed colonies at a higher rate and with increased severity relative to animals challenged with E. coli (Fig. 3a). Colonization by both bacteria was highly associated with death by 9 days post challenge (Fig 3a). Unlike our comparison between wild type and pmk-1 worms challenged with E. coli, the cross-sectional equivalent of observing the animals 16 hours after wash did show a significantly higher number of animals with a colony (Fig. 3b) and larger average colony area (Fig. 3c) for P. aeruginosa relative to E. coli. We detected the majority of colonies for both bacteria 16 hours following wash, and SICKO continued to detect new colonies throughout the imaging period in both populations (Fig 3d). Not surprisingly, the day when a colony becomes detectable in both E. coli and P. aeruginosa predicted the day when each animals died (Fig. 3e, f). Surprisingly, we observed a longer period between detection of the colony and death for the strongly pathogenic P. aeruginosa than the weakly pathogenic E. coli (Fig 3g). We speculate that this may be a consequence of the ability of P. aeruginosa to colonize healthier animals while E. coli acts more as an opportunistic pathogen that is only able to colonize less healthy animals.

E. coli OP50 is one of the most utilized laboratory C. elegans food sources and its pathogenicity is thought to minimally impact the survival or health of animals with a normal immune system. One application of the SICKO system is to differentiate colonized vs. non-colonized animals within each population. Surprisingly, the survival of both pmk-1 mutants and wild-type animals was dramatically lower in animals with active OP50 colonies relative to animals without detectable colonies (Fig. 4a, b). Additionally, pmk-1 mutants are thought to be deficient in several aspects of health even in the absence of a pathogen, contributing to the shortened lifespan (Alper et al., 2010). We observed that wild type and pmk-1 animals with and without detectable OP50 colonization have similar survival (Fig. 4c). The SICKO data paints a more complete picture of the health deficits faced by pmk-1 mutants. It appears that animals lacking pmk-1 have similar health characteristics to wild type in both colonized and uncolonized subpopulations, but are more likely to be colonized, resulting in a lower survival on average in the combined population.

These data highlight the ability of SICKO to characterize biological phenotypes in relationship to host-microbe interaction. In both wild type and pmk-1 worms, the time between a colony being detected and death trended lower for animals with late-detected colonies vs. early-detected colonies, but did not reach significance (Fig. 4d, e). Our earlier observation indicates that death typically occurs an average of 3 days after the day when a colony first becomes detectable (Fig. 2g). Finally, SICKO allows size of a colony on each day to be isolated and independently compared to remaining survival. We observe a correlation between colony area (Fig. 4f, g) or intensity (Fig. S2a, b) and remaining lifespan in both wild type and pmk-1 animals. This provides a framework for predicting the remaining lifespan of an animal at any point in life based on colony size.

When we made these same comparisons between different bacteria, animals with intestinal colonies were again shorter lived than animals without colonies for both E. coli (Fig. 5a) and P. aeruginosa (Fig. 5b). Animals challenged with P. aeruginosa were shorter-lived than animals challenged with E. coli in both the colonized and uncolonized sub-groups (Fig. 5c), likely reflecting the fact that part P. aeruginosa exert toxicity to their host animals through both colonization and production of excreted toxins that can affect host independent of colonization (Liu, 1974). We also observed that the time between first detection of a colony and animal death was longer for P. aeruginosa than E. coli for both early- and late-detected colonies (Fig. 5e, f). Speculatively, this may again be a consequence of E. coli being more of an opportunistic pathogen, selectively colonizing animals with a lower health status, while P. aeruginosa initially colonizes healthier animals resulting in a longer period of active colonization before the animals dies. Finally, unlike our first set of experiments, we did not find a significant association between colony area and remaining lifespan for E. coli (Fig. 5f), though we did find an association for P. aeruginosa infected animals (Fig. 5g). Both the number of animals with colonies and the variance of colony area within this group was very low across E. coli challenged animals in this set of experiments relative to pmk-1 experiments despite using the same timing, which may provide a technical explanation for this disparity between experiments.

A further application of SICKO is to characterize colony dynamics in individual animals across the population. We asked whether progression of colony growth is linked to survival. We used the slope from a linear regression of colony area in individual animals over time as a first order estimate of the average colony growth rate for that animal. In both the wild type (Fig. 6a) and pmk-1 (Fig. 6b) animals, we found that colony growth significantly and negatively correlated with survival among colonized individuals. Surprisingly, colony growth rate was similar between wild type and immune deficient pmk-1 mutants (Fig. 6c). Colony growth rate also negatively correlated with survival for animals challenged with both E. coli (Fig. 6d) and P. aeruginosa (Fig. 6e). In the case of P. aeruginosa, we see that colony growth rate is higher, on average, than E. coli (Fig. 6f). This provides a more complete characterization of the colonization dynamics of P. aeruginosa. It not only colonizes animals at a higher rate than E. coli, but also grows faster once the colony is formed.

The primary limitation of the SICKO system arises from the need to transfer animals away from the fluorescently labeled bacteria following exposure. This allows colonies established in the gut to be distinguished from the bacteria outside of the animal on the plate. During our validation studies, we found that the pmk-1 and P. aeruginosa challenged animals were much more likely than wild type to die during the wash and transfer steps (Fig. 1a[iii], Table 1), introducing a high likelihood for selection bias. We speculate that the lost pmk-1 and P. aeruginosa animals were those with early severe colonies and that this may account, at least in part, for some of the unexpected differences observed in wild type vs. pmk-1 animals, particularly in terms of the survival of animals with detectable infections (Fig. 4c). A related limitation is that no single metric provides a straightforward quantification of infection severity for pathogenic bacteria in a population at given point in time; in part because once an animal dies there is no longer a colony to quantify in terms of area or integrated intensity. To address these issues, we developed metric that reflects infection severity within a population, which we term the SICKO score, that modifies either the colony area or integrated intensity measurement for each animal with a weighting that reflects death prior to observation time (including animals lost during the wash and transfer steps), death during the observation time, and the proportion of the population afflicted by pathogen infection (see Methods for details). In essence, the SICKO score is an infection index that considers a bacterial colony of a given size more “severe” for bacteria if it causes more deaths and/or a higher colonization rate within the measured population. The pmk-1 mutant animals showed a consistently higher SICKO score, calculated based on area of infection, throughout their lifespan relative to wild type animals (Fig. 6g). We observed a similar pattern when the SICKO score is derived from integrated infection intensity (Fig. S2c). The SICKO score was dramatically higher for animals challenged with P. aeruginosa vs. E. coli, reflecting the drastic difference pathogenicity between the two types of bacteria (Fig. 6h). In summary, the SICKO scores provide a composite metric of infection severity within a population for pathogenic bacteria.

C. elegans strain KU25 (pmk-1(km25) IV) was obtained from the Caenorhabditis Genetics Center (CGC), which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). Wild type (N2) worms were originally obtained from Dr. Matt Kaeberlein (University of Washington, Seattle, WA, USA). C. elegans were cultivated on 60 mm culture plates containing nematode growth media (NGM) solidified with agar as previously described (Sutphin and Kaeberlein, 2009). Briefly, NGM agar plates were spotted with 300 µl of the E. coli strain OP50 grown overnight in lysogeny broth (LB) in a 37 °C orbital shaker and allowed to dry for a minimum of 16 hrs. Worms were maintained with abundant food for a minimum of three generations following recovery from frozen stock or from dauer larvae prior to running experiments. Animals were incubated at 20 °C and passed every 3-4 days to fresh NGM plates containing food.

E. coli containing the pMF230 plasmid, which includes both an ampicillin resistance cassette and eGFP driven by a constitutively active trc promoter, was obtained from Addgene (catalog #62546) and cultivated on LB agar plates containing 100 μM ampicillin to select the plasmid. A gentamycin resistance cassette from plasmid pFGM1 (Addgene catalog #64949) was and cloned into pMF230 plasmid, creating pMFG-eGFP, a constitutively active eGFP plasmid with 10 mg/ml gentamycin resistance. Plasmids were isolated with a QIAprep Miniprep Kit (Qiagen) and quantified using 260 nm/280 nm absorbance on a Biotek Synergy H1 plate reader. OP50 was made competent and transfected with pMFG-eGFP resistance plasmid as previously described (Choi et al., 2006). Bacteria colonies displaying green fluorescence were selected by plating LB supplemented with 10 mg/ml gentamycin. We constructed plasmid pMFG-mScarlet by cloning mScarlet into plasmid pMFG-eGFP to replace eGFP. P. aeruginosa strain PA14 was made competent and transfected with plasmid pMFG-mScarlet as previously described (Choi et al., 2006).

In this study we report two sets of experiments. The first set compares wild type (strain N2) worms vs. pmk-1(km25) loss of function mutants (strain KU245) challenged with GFP-labeled E. coli OP50. The second set compares wild type worms challenged with GFP-labeled E. coli OP50 to mScarlet-labeled P. aeruginosa PA14. Each set consisted of three independent biological replicates. All figures present the pooled data across all three replicates within the indicated experiment set. The number of animals lost during the wash step are provided in Table 1, and the number of animals examined during the observation step are provided in Table 2 for each replicate and for the pooled dataset.

Single worm culture environments were prepared as previously described (Espejo et al., 2022). Briefly, for each environment a Terasaki tray is mounted inside a single-well tray using a custom printed 3D adapter. The plastic interior surface of the Terasaki tray surrounding the wells is coated with an aversive barrier by applying a solution of 10 mg/mL palmitic acid (to prevent fleeing) 40 units/mL nystatin (to prevent fungal contamination) in 30% Tween-20 35% ethanol and allowing the liquid to fully evaporate. Each well of the Terasaki tray is filled with NGM solidified with low-melt agarose in place of agar (lmNGM), supplemented with 1 mg/mL gentamycin to select fluorescent plasmids, and seeded with E. coli OP50 grown in LB overnight in a 37 °C shaker, pelleted, and resuspended at a 10-fold concentration in 85 mM NaCl solution. The space surrounding the Terasaki tray in the single-well tray is filled with saturated water-absorbing crystals to provide humidity. The single-well tray is closed and sealed with parafilm until worms are ready to load (no more than 24 hours).

C. elegans were age synchronized by hypochlorite treatment (Porta-de-la-Riva et al., 2012). and plated on to NGM seeded with E. coli strain OP50. For worms challenged with E. coli, L4 larval stage worms were transferred to NGM plates supplemented with 500 μM floxuridine (FUdR) to prevent reproduction and 1 mg/ml gentamycin for plasmid selection and spotted with GFP labeled E. coli OP50 bacteria and incubated at 20 °C for 7 days. For worms challenged with P. aeruginosa, L4 larval stage worms were first transferred to NGM plates supplemented with 500 μM floxuridine (FUdR) to prevent reproduction seeded with unlabeled E. coli OP50 and incubated for 5 days. These day 5 adult animals were then transferred to NGM plates supplemented with 500 μM floxuridine (FUdR) to prevent reproduction and 1 mg/ml gentamycin for plasmid selection and spotted with mScarlet labeled P. aeruginosa PA14 bacteria and incubated at 20 °C for 2 days.

In both cases, worms were transferred following challenge (day 9 of adulthood) were transferred to fresh NGM plates supplemented with 500 μM FUdR and seeded with unlabeled E. coli OP50. Worms were placed on the edge of the plate outside of the bacterial spot prompting them to crawl toward the food. The animals are incubated for at least 16 hours at 20 °C to allow non-adherent GFP-expressing bacteria to fully pass out of their gut and be shed from the external surface of their body. Following this process, the only remaining GFP-expressing bacteria are in adherent colonies in the C. elegans gut. At the end of the 16-hour incubation, randomly selected animals are transferred to individual wells of prepared single-worm culture environments. The number of animals transferred, the number remaining alive on intermediate plate, and the number that died on the plate are all recorded for calculation of SICKO coefficient later.

In optimizing SICKO, we examined multiple exposure windows for wild type C. elegans to GFP-expressing E. coli OP50 and mScarlet-expressing P. aeruginosa to optimize the challenge step. Wild type animals challenged for with E. coli OP50 for 7 days starting at the L4 larval stage, or with P. aeruginosa for 2 days starting at day 5 of adulthood, result in detectable gut infections in approximately 30-50% of animals over the subsequent observation period in test experiments. Exposure to either strain during development or for fewer days starting from the L4 stage results in colonies only rarely (data not shown).

Starting on the day the animals are loaded onto the single-worm culture environments, three images of each animal are captured daily using the GFP channel (excitation: 425/60 nm, emission: 480 nm) of a Leica M205 FCA Fluorescent Stereo Microscope equipped with a Leica K6 sCMOS monochrome camera using 2.5x zoom and saved as TIF files in the folder structure specified by the SICKO software documentation (Fig. S3) (Freitas, 2024). Capturing multiple images per animal ensures the signal is not impacted by worm movement (i.e., a non-blurry image can be selected for each animal). Worms are exposed to blue light to stimulate movement and recorded as dead if no movement observed, or fled if the animal is no longer present in the well.

Version March.13.2024.v1 of the SICKO software, built in MATLAB version 2022b, was used in this study (Freitas, 2024). SICKO takes as input raw fluorescent images each containing a single worm infected with fluorescently labeled bacteria, removes background, identifies the infection area, and quantifies the area and integrated fluorescent intensity of the infection. The SICKO image processing GUI allows users to manually select image regions that contain artifacts (usually fluorescently labeled bacteria outside of the worm or light pollution) that are removed from analysis. The SICKO analysis software can be downloaded along with complete documentation at: https://github.com/Sam-Freitas/SICKO (Freitas, 2024). A step-by-step protocol for using the software is provided. Generation of figures and statistics using the SICKO output files was conducted in RStudio version 2023.12.1 Build 402 running R version 4.3.2. The scripts used to generate these figures are available with instruction for use at: https://github.com/lespejo1990/SICKO_Analysis (Espejo, 2023).

SICKO analysis relies on a threshold to distinguish infection from background. A user defined threshold determines a mask that captures the area in the image representing bacterial colony (the “colony mask”). To compensate for background differences across the field of view, background correction is performed radially of the colony mask, ensuring animal position within the field of view minimally impacts signal intensity and accounts for uneven background. The intensity of all pixels within the colony mask is integrated to calculate the integrated infection intensity for each animal. The number of pixels within the colony mask defines the colony area.

Examining the area or integrated intensity of a colony over time in live animals across a population provides information on the state of the colony size at a given time point but has limited utility in tracking colony progress over time because it cannot account for animals that died during the pathogen challenge, during the wash, or during an earlier observation time. Animals that died at earlier stages likely represent a more severe response to colonization, and thus simply quantifying infection area or integrated intensity at a given time point will tend to underestimate the pathogen severity by excluding data from dead animals. We developed an infection severity index for pathogenic bacteria called the SICKO score that provides a quantitative metric of pathogen colony (aka “infection”) progress within a population that systematically accounts for the number of worms that died before a given time, including during the post-challenge wash, worms that died at the time of observation, and fraction of animals harboring a colony within a population (a metric of infectivity). To determine the SICKO score, we first calculate a SICKO coefficient, which is then used to weight the colony area or integrated intensity for each animal at each time point. The SICKO coefficient is updated for each time point to reflect the fraction of animals that died with a colony up until that day.

First, on each day of observation, t, we calculate the cumulative number of animals that have died while harboring a bacterial colony during the observation period (Fig. 1a[v]) up until that day, D_CO(t), and the total number of animals that died while harboring a bacterial colony across the full observation period, : where D_COi is the number of worms the died while harboring a colony at time point i, and t_max is the last day of observation.

Next, we estimate the number of worms that died during the challenge Fig. 1a[ii] and wash Fig. 1a[iii] steps while harboring a bacterial colonization, D_CW: where D_W is the total number of worms that died during the challenge and wash steps, D_CO is the number of animals that died while harboring a colony during the full observation period (Fig. 1a[iv]), and D_O is the total number of worms that died during the observations period (Fig. 1a[iv]). We note that this cannot be measured directly, because while worms are on the challenge and wash plates, they will have fluorescent bacteria in their intestine and on their external surface that is not part of an established tissue colony but cannot be distinguished from bacteria in colonies. In our case, we make the assumption that the fraction of worms that die with a colony during the challenge and wash steps will be the same as the fraction that die with a colony during observation period. Other assumptions and approaches can be employed in this case, for example that all worms lost during challenge and wash harbor a colony (which may be warranted for highly pathogenic bacteria that are likely to result in early animal death).

From equation (3), we next estimated number of worms that ever had a detectable bacteria colony at any time during the experiment, P_C: where A_W is the number of animals alive after the challenge and wash step (including those not selected for observation, Fig. 1a[iii]), T_CO is the total number of non-censored animals that were every observed to have a colony in the observation group (Fig. 1a[v]), and T_O is the total number of non-censored animals in the observation group (Fig. 1a[v]).

Using the values from Equations (1), (3), and (4), we next calculate the SICKO coefficient, S_C which provides a population-level weighting that increases with increasing infectivity (fraction of animals the develop a colony) and toxicity (fraction of animals that die with a colony prior up until the current day, t):

In Equation (5), the term in the left parentheses represents a weighting that increases with the fraction of animals that develop a bacterial colony, while the term in the right brackets represents a weighting that increases with the number of animals that have died while harboring a colony up until the current day, t.

Finally, the SICKO coefficient in Equation 5 is used to as a weight to calculate at SICKO infection severity index based on either the measured colony area, S_I,area(t), or integrated intensity, S_I,intensity(t), for each worm, w, on each day, t: where C_area(w,t) and C_intensity(w,t) are the measured colony area and integrated intensity, respectively, for worm w on day t. The values for S_I,area and S_I,intensity are plotted in Fig. 6e,f and Fig. S2c, respectively.