We used eleven natural C. elegans isolates ( Supplementary Table S1A ) isolated at two North German locations between 2002 and 2012 (Haber, 2005; Petersen et al., 2014) and the C. elegans laboratory strain N2. The natural isolates were chosen based on genotypic differences at the microsatellite level that translated into phenotypic variation (Petersen et al., 2015). All strains were maintained following standard procedures (Stiernagle, 2006).

Six bacterial strains were used as representatives for different genera of C. elegans native microbiota able to colonize the worm’s intestine ( Supplementary Table S1B ) (Dirksen et al., 2016; Dirksen et al., 2020; Zimmermann et al., 2020; Pees et al., 2021): the Gram-negative bacteria Acinetobacter guillouiae MYb10, Pseudomonas fluorescens MYb11, Ochrobactrum anthropi MYb49, O. vermis MYb71, and Ochrobactrum pseudogrignonense MYb237 and the Gram-positive bacterium Rhodococcus erythropolis MYb53 (Dirksen et al., 2016). Microbiota isolates were thawed freshly, cultured for 2 days on tryptic soy agar (TSA) at 25°C, and subsequently grown in tryptic soy broth (TSB) in a shaking incubator for 42 h at 28°C (MYb10, MYb49, MYb53, MYb71) or 16 h (MYb11). E. coli OP50 was cultured in TSB for 16 h at 37°C. For the experiments, all bacteria were adjusted to OD₆₀₀10 in their supernatant if not stated otherwise.

Bacterial lawns (OD₆₀₀10) were grown for 24 h at 25°C on 9-cm nematode growth medium (NGM) plates. The lawns were scraped off, pooled from five plates, and pelleted at 4,000 g at 4°C for 15 min, and 100 µl of bacterial pellet was resuspended in 1 ml of RNAmagic (Bio-Budget Technology GmbH, Krefeld, Germany). Aliquots of 800 µl were frozen in liquid nitrogen, thawed at 45°C and frozen in three repeated cycles, and stored at −80°C until RNA isolation. Total RNA was isolated using MACHEREY-NAGEL (Düren, Germany) NucleoSpin Kit using the protocol for bacterial cells.

Choice behavior was examined using either multiple-choice experiments offering five bacterial strains simultaneously or binary-choice experiments with only two bacterial strains. C. elegans strains were either unconditioned or preconditioned with a specific microbiota bacterium. Worms touching a bacterial spot were scored, and all other worms were not taken into account. The position of the bacteria across assay plates was randomized, and all experiments were performed without current knowledge of worm and bacterial strain identity to avoid any observer bias.

For the multiple-choice experiments, four microbiota bacteria and OP50 were adjusted to OD₆₀₀10 in phosphate-buffered saline (PBS), and 25 µl of each bacterium and a PBS control were pipetted equidistantly in a circle onto 9-cm peptone-free NGM (PFM) plates or NGM plates and left to dry for 2 h. Approximately 100 synchronized C. elegans in the fourth larval stage (L4) were pipetted to the center of the plate. The number of worms residing on each bacterium was determined after 2 or 24 h. We calculated the mean proportion of worms per bacterium = number of worms on a bacterium/total number of worms on all bacteria per plate.

C. elegans was grown on OP50 until the L4 stage, washed three times in sterile M9 buffer, and pipetted to preconditioning plates freshly inoculated with one bacterium in OD₆₀₀10. Preconditioning was performed with live and heat-killed (80°C for 40 min) MYb71 and OP50, with OP50 lawns spiked with 21 ng of total RNA of MYb71 or OP50, or with the odor of viable MYb71 or OP50. For the latter, L4 worms were placed on OP50 lawns with either 500 µl MYb71 or OP50 on a thin NGM layer in the lid of the plate. Lining the edge of the plate with 1% palmitic acid in ethanol kept the worms on the OP50 lawn. After 24 h, the preconditioned adults were washed repeatedly in M9 buffer and subsequently applied to binary-choice assay plates.

Approximately 70 naïve L4 worms or preconditioned adults were pipetted centrally between two opposing 30-µl bacterial spots on 6-cm NGM plates. The number of worms residing on each bacterium was determined after 2 and/or 24 h. We calculated a choice index = (number of worms on the bacterium opposing the microbiota bacterium − number of worms on the microbiota bacterium)/(total number of worms on both bacteria). To test for bacterial signals produced upon contact with worms, the initially placed worms were removed after 24 h, and naïve adult worms (from OP50) were pipetted centrally onto the same plates. The choice index for the newly introduced worms was determined after 2 h (26 h total time).

Synchronized C. elegans MY2079 L4 worms were preconditioned on MYb71 or OP50 for 24 h. Adults were washed repeatedly in M9 buffer, and approximately 50 adults were pipetted onto NGM plates inoculated with 50 µl of either MYb71 or OP50. The number of worms on and outside the bacterial lawn was scored after 2 h. We calculated the proportion of escaped worms = number of worms outside the lawn/total number of worms on the plate.

NGM and PFM plates were inoculated daily with 50 µl MYb71 or OP50 at OD₆₀₀10 adjusted in PBS and stored at 20°C for 24 h. Initially, one L4 worm was picked onto the plate and transferred daily to a new plate. We scored the hatched offspring of each worm until the worm stopped laying eggs. Data of worms that died or escaped from the plate before the end of the egg-laying period were excluded.

Statistical calculations were performed with R studio software (version 1.3.1093) and can be found in Supplementary Table S2 . Graphs were produced with R Studio (version 1.3.1093) and edited with Inkscape (version 1.0.1).

N2 can be considered as a laboratory animal, domesticated to a monoxenic environment with only E. coli OP50 as a food source. Natural C. elegans isolates can differ in their behavioral response to bacterial cues (Petersen et al., 2015; Pradhan et al., 2019). Since C. elegans microbiota was originally extracted from natural C. elegans isolates, we selected eleven genotypically divergent natural C. elegans isolates (Petersen et al., 2015) and the laboratory strain N2 to screen their choice behavior toward four microbiota isolates from different genera and OP50 as control ( Figure 1A ).

We defined choice behavior as present if worms preferred at least one of the offered bacteria over the others. In an initial screen, we identified general significant choice behavior after 24 h in one natural C. elegans isolate, MY2079 (p = 0.022; Supplementary Figure S1 ), and a trend in eight natural C. elegans isolates (p < 0.063). We confirmed the results of the screen for MY2079 with an additional multiple-choice experiment using a higher replication (n = 15), in which not only MY2079 (p = 0.012) but this time also N2 (p = 0.008) showed choice behavior after 24 h. Yet the choice behavior between MY2079 and N2 differed significantly (p < 0.001; chi-squared test) ( Figure 1B ).

In its ephemeral natural habitats, C. elegans lives a boom-and-bust life cycle (Frézal and Félix, 2015), because as the availability of plant material varies, so do the growth conditions and the availability of food organisms. Hence, we wondered whether the availability of nutrients for bacterial growth influences the choice behavior toward microbiota. We examined the effect of bacterial proliferation on the choice behavior of C. elegans N2 and the natural isolate MY2079 in a multiple-choice experiment on NGM, which supports bacterial proliferation. C. elegans MY2079 showed choice behavior after 24 h (p < 0.001) but not after 2 h, whereas C. elegans N2 showed generally no choice behavior on NGM ( Figure 1C ). Further, the choice behavior between C. elegans MY2079 and N2 on NGM differed significantly after 2 h (p < 0.001; chi-squared test) and 24 h (p < 0.001; chi-squared test) ( Figure 1C ).

These results indicate that the choice behavior of N2 and MY2079 differs and that C. elegans MY2079 shows choice behavior particularly in the presence of proliferating microbiota.

We noticed in the multiple-choice experiment on proliferating bacteria that MY2079 changed its choice behavior toward MYb71 and OP50 over time and eventually avoided MYb71 in favor of OP50 ( Figure 1C ). Hence, we wondered whether the avoidance behavior toward MYb71 is due to repulsion from MYb71 or increased attraction toward one of the other bacteria. We performed a binary-choice experiment with MYb71 opposing one test bacterial strain at a time ( Figure 2A ). C. elegans N2 did not change its choice behavior over time and chose MYb71 and the opposing test bacteria equally. Similarly, C. elegans MY2079 chose MYb10, MYb11, MYb53, and OP50 equally to MYb71 at 2 h ( Figure 2B ). However, after 24 h, MY2079 preferred all other bacteria over MYb71 (p = 0.028). These results suggest that the change in choice behavior toward MYb71 over time is specific for C. elegans MY2079 and is caused by avoidance of proliferating MYb71.

What might be responsible for the change in the choice behavior of MY2079 toward MYb71 over time – a phenomenon that we as of now call “late microbiota avoidance response”? In the course of the choice experiment, two main factors changed: 1) bacterial proliferation led to an increase in cell density and therefore higher concentration of bacterial products after 24 compared with 2 h. 2) At the same time, the worms developed from the L4 stage to the adult stage. We therefore compared the attraction of C. elegans MY2079 toward MYb71 in OD₆₀₀10 and OD₆₀₀30 (directly adjusted from a liquid culture) and toward MYb71 lawns grown on a plate for 2 or 24 h. The latter was done for L4 larvae and adults to examine developmental influences. Both L4 larvae (p < 0.001) and adults (p < 0.001) chose the 24 h MYb71 lawn over the 2 h lawn ( Figure 2C ). Similarly, L4 larvae of N2 (p = 0.003) and MY2079 (p = 0.014) chose MYb71 at OD₆₀₀30 over OD₆₀₀10 ( Supplementary Figure S2 ).

These findings indicate that C. elegans prefers a higher MYb71 cell density in direct comparison with a lower MYb71 cell density independent of the developmental stage. However, in the presence of an alternative bacterial lawn, the MYb71 lawn after 24 h is less attractive ( Figure 2B ). Therefore, an increased cell density in the course of the experiment did not explain the late microbiota avoidance response toward MYb71 in the previous experiments.

Bacterial communities are able to react to their environment, including being grazed on by bacterivores. Polynucleobacter asymbioticus, for example, changes its transcriptional profile upon grazing by chrysophytes concerning transcriptional, translational, and stress responses (Beisser et al., 2019). These transcriptional differences in bacteria and resulting changes in the availability of bacterial products might be perceived by C. elegans and consequently influence its behavior. To test the potential production of worm-repelling compounds after worm exposure, we exposed MYb71 to C. elegans in a binary-choice experiment, removed the worms after 24 h, and exposed naïve worms to the same bacteria. As seen before, the initially placed C. elegans showed the late microbiota avoidance response toward MYb71 (p < 0.001; Figure 2D ). Interestingly, naïve worms did not avoid the 24-h-old, worm-exposed MYb71 but preferred this lawn over OP50 instead (p < 0.001).

These findings suggest that the production of a repellent bacterial product triggered by exposure to worms is rather unlikely.

As we did not observe a late microbiota avoidance response in naïve worms but in worms that have likely fed on MYb71, we hypothesized that worms had to be exposed to MYb71 to show the late microbiota avoidance response. Therefore, we preconditioned worms with MYb71 or OP50, i.e., exposed L4 worms for 24 h to either MYb71 or OP50, and subsequently performed a binary-choice experiment scoring behavior only after 2 h. Interestingly, we found the late microbiota avoidance response in preconditioned worms as in previous assays after 24 h, whereas worms preconditioned with OP50 did not show the late microbiota avoidance response (p < 0.001; Figure 3A and Supplementary Figures S3A, B ). This indicates that preconditioning with MYb71 is enough to induce the late microbiota avoidance response.

We further wondered whether the choice behavior changed due to general intestinal colonization with microbiota and examined the choice behavior after preconditioning with two other representatives of colonizing microbiota, P. lurida MYb11 and A. guillouiae MYb10. C. elegans MY2079 chose OP50 over MYb11 independent of the preconditioning (p < 0.001; Figure 3B and Supplementary Figures S3C, D ). However, worms preconditioned with MYb11 chose MYb11 slightly more compared with worms preconditioned with OP50 (p = 0.014). Preconditioning with MYb10 had no effect on choice behavior, and MYb10 and OP50 were always chosen equally ( Figure 3C and Supplementary Figures S3E, F ).

We used two additional Ochrobactrum microbiota isolates, O. pseudogrignonense MYb237 and O. anthropi MYb49, to test for a genus-specific preconditioning effect on choice behavior. C. elegans MY2079 chose OP50 over MYb237 independent of the preconditioning (p < 0.001), but OP50 was even more preferred in worms preconditioned with MYb237 (p < 0.001; Figure 3D and Supplementary Figure S3G ). Preconditioning with MYb49 led to the choice of OP50 over MYb49 (p = 0.014), whereas worms preconditioned with OP50 chose OP50 and MYb49 equally in two out of three runs ( Figure 3E and Supplementary Figures S3H, I ). Overall, preconditioning with MYb49 compared with preconditioning with OP50 had no effect on the choice behavior.

These results support that preconditioning with colonizing microbiota can influence C. elegans choice behavior. However, neither colonization with microbiota nor preconditioning with bacteria of the genus Ochrobactrum resulted in a late avoidance response of the microbiota, in which avoidance occurred only after preconditioning with the test bacterium. Hence, the late microbiota avoidance response appears to be specific for O. vermis MYb71.

Pathogenic bacteria, bacteria with insufficient nutritional value, or bacteria that negatively affect fitness in any other way can be avoided by C. elegans in order to reduce their uptake (Schulenburg and Müller, 2004; Meisel and Kim, 2014). Since we observed avoidance of MYb71, we evaluated whether the late microbiota avoidance response is linked to potential pathogenicity of MYb71. We found neither any lawn leaving behavior from MYb71 ( Supplementary Figure S4A ) nor a detrimental effect of MYb71 on the worm’s fitness ( Supplementary Figure S4B ). These observations suggest that MYb71 does not have a pathogenic effect on C. elegans.

Even without obvious pathogenicity of MYb71 to C. elegans, the question still remains whether C. elegans displays learned avoidance behavior mediated by bacterial products or whether gut colonization is necessary to elicit the worm’s late microbiota avoidance response.

Recently, the learned pathogen avoidance behavior of C. elegans against Pseudomonas aeruginosa PA14 has been demonstrated to depend on a single non-coding RNA derived from PA14 (Kaletsky et al., 2020; Moore et al., 2021). Hence, we tested the influence of preconditioning with total RNA of MYb71 or OP50 on choice behavior and found that worms from both treatments chose OP50 over MYb71 in a binary choice assay (p = 0.044, Figure 4A ).

The learned pathogen avoidance response against PA14 is further mediated by bacteria-derived odors (Zhang et al., 2005). Here, worms preconditioned with MYb71 odor chose MYb71 and OP50 equally independent of the bacteria used for odor preconditioning ( Figure 4B ).

To answer the question of whether C. elegans needs to be colonized by MYb71, we tested the late microbiota avoidance response of worms preconditioned with either heat-killed bacteria or viable bacteria. Both worms preconditioned with viable (p < 0.001) and heat-killed MYb71 (p = 0.011) chose OP50 over MYb71 ( Figure 4C ). However, worms preconditioned with viable MYb71 resided more often on OP50 lawns than worms preconditioned with heat-killed MYb71 (p = 0.022). Worms from OP50 chose MYb71 and OP50 equally, independent of the preconditioning.

These results indicate that the late microbiota avoidance response is not caused by total bacterial RNA nor bacteria-derived odors but primarily by the worm’s colonization with viable MYb71.