We conceptualized feeding effort in C. elegans as the amount of pharyngeal muscle activity required to ingest a given quantity of food. The starting point of our experiments was to make bacteria more effortful to consume. Computer simulations of pharyngeal mechanics predict that large bacterial cells are swallowed less efficiently than small cells (Avery and Shtonda, 2003). Consistent with this result, worms develop more slowly when cultured on large bacteria, possibly because they are acquiring nutrients less efficiently (Avery and Shtonda, 2003). These findings led to the hypothesis that larger bacteria require more effort to eat, necessitating more pharyngeal pumps per unit of food consumed. To test this hypothesis, we artificially increased the size of an easily ingested bacterium, Comamonas spp., using the antibiotic cephalexin. In Gram-negative bacteria such as Comamonas, cephalexin interferes with bacterial cell septation during mitosis, resulting in elongated bacterial filaments that nevertheless remain edible (Millet et al., 2022; Figure 1—figure supplement 1).

Pharyngeal pumping rates in normal and elongated bacteria were measured in the reference strain N2 by placing individual worms in a microfluidic channel filled with a bacterial suspension and fitted with electrodes that record the electrical signals associated with each pharyngeal contraction (Lockery et al., 2012). Bacteria were washed with cephalexin-free buffer before being fed to worms. Pumping frequencies in normal and elongated bacteria were equivalent (Figure 1A; Table 1, row 1), indicating that worms expend approximately the same amount of effort per unit time feeding on normal and elongated food. To test the hypothesis that elongation interferes with food absorption, we measured nutrient uptake in worms cultured on normal or elongated bacteria, using fat stores as a proxy for carbohydrate and lipid consumption. This was done by staining fat stores with the specific label Oil-Red-O (Soukas et al., 2009) and quantifying the density of staining after a 2-day culture period extending from the L1 to the L4 stage. We found that fat levels were significantly lower in worms cultured on elongated food (Figure 1B2; Table 1, row 2). Given that the pumping rate was unchanged in the presence of normal versus elongated bacteria (Figure 1A), the lower fat stores observed in worms fed elongated bacteria suggest that the amount of nutrition incorporated per pump is lower in worms feeding on elongated versus normal bacteria. This supports our hypothesis that elongation impairs food transport, requiring longer feeding bouts and thus increased muscular activity to consume an equivalent amount of food.

Our effort-discounting paradigm assumes that the normal and elongated bacteria are equivalent except for the amount of effort required to consume them. For example, elongation could affect nutrient content or olfactory attractiveness of bacteria. C. elegans food preferences are sensitive to the nutrient content of bacteria (Feng et al., 2025). We therefore asked whether cephalexin treatment affects nutrient content. We focused on key macronutrients: carbohydrates, lipids, and proteins. For protein and carbohydrates, we used colorimetric assays (Materials and methods; Figure 1C, D). For lipids, we used phase separation to harvest and measure lipid content (Materials and methods; Figure 1E). In all three tests, we detected no differences between control and cephalexin-treated bacteria (Table 1, rows 3–5). We conclude that cephalexin treatment does not alter bacterial nutritional composition.

Worms are attracted to particular species of bacteria by the characteristic mixture of volatile organic compounds each species emits (Worthy et al., 2018). Thus, another concern is that cephalexin treatment might alter the odors produced by bacteria, thereby changing their attractiveness. To control for this possibility, we performed a T-maze assay (Materials and methods) in which patches of normal and elongated bacteria were placed at the ends of each respective arm. The patches contained equal volumes of bacteria at an optical density (OD) of 1.0. Patches were spiked with sodium azide, a paralytic agent, to trap worms once they reached a patch. Worms were placed at the base of the T-maze and allowed to explore it for 60 min after which the number of worms in each patch was counted. A food preference index (-1≤I≤1\begin{document}$-1\leq I\leq 1$\end{document}) was calculated, where positive and negative values indicate preference for normal and elongated bacteria, respectively, and I≅0\begin{document}$I\cong 0$\end{document} denotes indifference between the two options (Materials and Methods). We found that worms under these conditions were indifferent (Figure 1F2; Table 1, row 6). This result indicates that cephalexin treatment does not alter the relative attractiveness of normal or elongated bacteria.

Having demonstrated that elongated bacteria require more effort to eat but are equivalent to normal bacteria in terms of nutrition and attractiveness, we next asked whether C. elegans exhibits a preference for the normal, easier-to-eat bacteria – evidence of effort discounting-like behavior. We tested this by measuring food preference in the T-maze assay using the same methodology as in Figure 1F but without the paralytic agent, allowing individual worms to sample both patches. One arm of the T-maze was baited with normal bacteria at an OD of 1.0, while the other arm contained elongated bacteria at ODs of 1.0, 1.5, or 2.0. Worms explored the maze for 60 min and the food preference index was computed based on worm counts at the end of this period. There was a significant overall effect of OD of elongated bacteria on preference (Figure 2B; Table 1, row 7). When normal and elongated bacteria were at the same concentration (OD 1), worms preferred normal bacteria (I = 0.10; Figure 2B; Table 1, row 8). This demonstrates that worms do, in fact, devalue elongated food. When normal and elongated bacteria were at OD 1.0 and 2.0, respectively, worms preferred elongated bacteria (I = –0.17; Figure 2B; Table 1, row 9). This suggests that, in the worm’s valuation system, higher food density compensates for greater effort. This outcome was not a foregone conclusion. For instance, it was possible that no amount of elongated food would be sufficient to make worms prefer it. Our results are consistent with findings in rodents, where high-effort options are chosen when the food reward density is sufficiently high (Floresco et al., 2008). When the optical densities of normal and elongated food were 1.0 and 1.5, respectively, worms were indifferent between the two options (I = 0.03; Figure 2B; Table 1, row 10). Indifference points are also observed in rodent studies of effort discounting (Salamone et al., 1994; van den Bos et al., 2006; Ostrander et al., 2011). Overall, we conclude that in feeding decisions, worms take into account both relative effort and relative food density.

Indifference points are significant because, by definition, the presented options have equal perceived value, allowing for the quantification of devaluation. This quantification is only possible at an indifference point. For example, consider a hypothetical human scenario in which a participant is asked to choose between receiving $20 immediately or $20 with the added cost of climbing 20 flights of stairs. Most participants would choose the immediate $20. Given a choice between $20 immediately and $100 at the cost of climbing 20 flights of stairs, some participants might choose the $100 option. However, suppose that when given a choice between $20 immediately and $80 at the cost of climbing 20 flights of stairs, a particular participant is indifferent between the two options. For this person, under these circumstances, we can conclude that the $20 option is perceived as equal in value to the $80 option. In other words, the $80 option has been discounted (devalued) to $20.

Formally, at an indifference point,V(E)=d(E)V0\begin{document}$$\displaystyle V\left (E\right)=d\left (E\right)V_{0}$$\end{document}

where E\begin{document}$E$\end{document} represents effort, V0\begin{document}$V_{0}$\end{document} represents the original value of the reward before devaluation, and V(E)\begin{document}$V\left (E\right)$\end{document} is the discounted value of V0\begin{document}$V_{0}$\end{document}. The quantity dE\begin{document}$d\left (E\right)$\end{document} is the discount factor, which, in economic models of discounting, is a function of E\begin{document}$E$\end{document}. From the above equation,d(E)=V(E)V0\begin{document}$$\displaystyle d\left (E\right)=\frac{V\left (E\right)}{V_{0}}$$\end{document}

with 0≤d(E)≤1\begin{document}$0\leq d\left (E\right)\leq 1$\end{document}; in the above example d(E)\begin{document}$d\left (E\right)$\end{document} = 20/80 or 1/4. In economic models, d(E)\begin{document}$d\left (E\right)$\end{document} is inversely related to E\begin{document}$E$\end{document} but the precise form of d(E)\begin{document}$d\left (E\right)$\end{document} varies between models (Figure 2—figure supplement 1).

In our experiments, E\begin{document}$E$\end{document} is determined by the degree of bacterial elongation. Although this quantity is unknown, it remained unchanged because we used the same cephalexin exposure protocol across experiments (see Materials and methods). It is analogous to the fixed effort of climbing 20 flights of stairs in the hypothetical example. Assuming a linear relationship between OD and perceived value, we can determine d(E)\begin{document}$d\left (E\right)$\end{document} from the indifference point shown in Figure 2B, where V0\begin{document}$V_{0}$\end{document} = OD 1.5 and V(E)\begin{document}$V\left (E\right)$\end{document} = OD 1.0, yielding d(E)\begin{document}$d\left (E\right)$\end{document} = 2/3. In the most basic models of discounting, d(E)\begin{document}$d\left (E\right)$\end{document} is independent of V0\begin{document}$V_{0}$\end{document}. Assuming independence, other values of V0\begin{document}$V_{0}$\end{document} can be inserted into equation 1 to predict novel indifference points. For instance, setting V0\begin{document}$V_{0}$\end{document} = OD 3 predicts indifference at V(E)\begin{document}$V\left (E\right)$\end{document} = OD 2, and setting V0\begin{document}$V_{0}$\end{document} = OD 0.75 predicts indifference at V(E)\begin{document}$V\left (E\right)$\end{document} = OD 0.5. That is, any pair of optical densities (normal versus elongated) in a 2:3 ratio should be an indifference point. We tested these predictions by measuring preference in T-mazes baited with normal and elongated food at OD ratios (normal:elongated) of 2:3 and 0.5:0.75, finding indifference as predicted (Figure 2C, Table 1, rows 11–14). The fact that standard discounting models correctly predict indifference points reinforces the conclusion that C. elegans exhibits discounting-like behavior.

We next investigated the behavioral mechanism through which effort influences food preference, examining how worms interact with patches of normal versus elongated food. We studied these interactions at the microscopic level of individual worms to explain behavior at the macroscopic level of T-maze assays. This was done by performing food-patch-leaving assays (Scheer and Bargmann, 2023) using either normal or elongated bacteria. Individual worms were transferred to a 1-cm diameter, nematode growth medium (NGM)-filled arena containing a small patch (3–4 mm diameter) of normal or elongated bacteria (OD 1.0) at its center. The patch size was comparable to those in the T-maze assay. Each worm was video recorded for 15 min while it entered and left the patch multiple times (Figure 3A).

For analysis, we took a kinetic approach to investigate the rates at which worms stochastically changed their position relative to the food patch. This approach is both simple, as it does not require assigning specific locomotory states at particular times to individual worms, and powerful, as it forms a bridge between the microscopic and macroscopic behavior of a stochastic system, thereby elucidating underlying dynamics. Accordingly, we defined three kinetic states based on occupancy of specific zones in the arena (Figure 3B):

This system contains four state transitions: F → B, B → O, O → B, and B → F. Behavior was analyzed in terms of state probabilities (PF,PB,PO\begin{document}$P_{\mathrm{F}}, P_{\mathrm{B}},P_{\mathrm{O}}$\end{document}) and rate constants (kFB,kBO,kOB,kBF\begin{document}$k_{\mathrm{F}\mathrm{B}}, k_{\mathrm{B}\mathrm{O}}, k_{\mathrm{O}\mathrm{B}},k_{\mathrm{B}\mathrm{F}}$\end{document}). State probability was defined as the fraction of total observation time the worm was in each state. Rate constants were defined as the number of transitions divided by the total time spent in the state of origin. Rate constants determine the probability per unit time of transitioning from one state to another and thus reflect the underlying physical causes of state transitions.

We observed a variety of changes in kinetic parameters on elongated versus normal food. PB\begin{document}$P_{\mathrm{B}}$\end{document} and PO\begin{document}$P_{\mathrm{O}}$\end{document} increased at the expense of PF\begin{document}$P_{\mathrm{F}}$\end{document}, consistent with decreased preference for elongated food in T-maze assays (Figure 3C; Table 1, rows 15 and 16). The rate constants exhibited two key features. First, in control animals, between adjacent states, the mean rate constants leading to patch entry (O → B → F transitions) were greater than the mean rate constants leading to patch leaving (F → B → O transitions). In particular, kBF>kFB\begin{document}$k_{\mathrm{B}\mathrm{F}}\gt k_{\mathrm{F}\mathrm{B}}$\end{document} and kOB>kBO\begin{document}$k_{\mathrm{O}\mathrm{B}}\gt k_{\mathrm{B}\mathrm{O}}$\end{document} (Figure 3, D1 vs D3 and D2 vs D4; Table 1, rows 17–23). This indicates a greater propensity to enter food patches than to leave them, which is consistent with efficient foraging. Second, changes in rate constants brought about by elongated food exhibited a distinctive pattern: the rates for patch leaving (kFB,kBO\begin{document}$k_{\mathrm{F}\mathrm{B}}, k_{\mathrm{B}\mathrm{O}}$\end{document}) were increased whereas the rates for patch entry (kOB\begin{document}$k_{\mathrm{O}\mathrm{B}}$\end{document}, kBF\begin{document}$k_{\mathrm{B}\mathrm{F}}$\end{document}) were unchanged (Figure 3, D1 and D2 vs D3 and D4; Table 1, rows 24–28). Thus, worms exhibited a heightened propensity to leave elongated food, consistent with previous observations (Scheer and Bargmann, 2023). The fact that rates for patch entry were unaffected by elongated bacteria rules out a mechanism in which worms, exiting patches of elongated food (which is relatively undesirable), become reluctant to re-enter them. Finally, we found that speed on elongated food increased, whereas speed off food remained unchanged (Figure 3E; Table 1, rows 29–33). Applying the kinetic data of individual worms to the population level in T-maze assays suggests that effort discounting-like behavior could result from feeding decisions based on the local food environment rather than a direct comparison of patch contents across the arms of the T-maze.

A salient aspect of effort discounting in rodents is its sensitivity to changes in midbrain dopamine signaling (Salamone et al., 2018). Increased signaling promotes preference for high-value, high-effort rewards, whereas decreased signaling promotes preference for low-value, low-effort rewards (Bailey et al., 2016; Salamone et al., 2018; Hart and Izquierdo, 2019); in other words, increased dopamine signaling results in greater effort tolerance. Dopamine signaling is a candidate mechanism for the regulation of effort discounting-like behavior in C. elegans because it affects how worms respond to the presence and absence of food as they enter or leave food patches. Worms slow down when they encounter a food patch, a response that requires intact dopamine signaling (Sawin et al., 2000). When worms exhaust their food supply or are removed from a food patch, they engage in a bout of frequent, high-angled turns called area-restricted search (Hills et al., 2004). This response also requires dopamine signaling.

To test whether dopamine signaling is involved in effort discounting-like behavior in C. elegans, we examined food preference in a variety of mutants with altered dopamine signaling. The gene cat-2 encodes a tyrosine hydroxylase essential to the main synthesis pathway for dopamine; cat-2 mutants are commonly used to reduce dopamine signaling in C. elegans. The gene dat-1 encodes a dopamine transporter that mediates the reuptake of dopamine into presynaptic neurons; dat-1 mutants are used to increase dopamine signaling. The effects of reduced dopamine signaling can also be investigated in dopamine receptor mutants. C. elegans has four genes encoding dopamine receptors with homology to dopamine receptors in mammals. Of these, dop-1, dop-2, and dop-3 are the most thoroughly characterized to date. dop-1 is a DRD1 homolog, whereas dop-2 and dop-3 are DRD2 homologs. The genes dop-1 and dop-3 are required for normal responses to the presence of food (Chase et al., 2004); dop-2 is required for increased levels of activity in the absence of food (Bastien et al., 2024). We found a significant overall effect of strain on food preference measured in the T-maze assay when both foods were at OD 1.0 (Figure 4; Table 1, row 34). Post hoc testing showed that cat-2, dat-1, dop-1, and dop-2 were indistinguishable from N2, whereas dop-3 mutants were significantly different from N2 (Table 1, rows 36–40). In particular, they were indifferent between the two food options, indicative of decreased effort discounting-like behavior (Table 1, row 41). We conclude that fully intact dopamine signaling is required for normal effort discounting-like behavior in C. elegans, as in rodents. However, the fact that impaired dopamine signaling rendered worms indifferent between the food options suggests that decreased dopamine signaling increases tolerance for high-effort food, in contrast to its effect in rodents. The absence of an effect of the cat-2 and dat-1 mutations is puzzling in view of the requirement for dop-3. One possibility is that residual dopamine in the cat-2 mutant (Sanyal et al., 2004) is sufficient for normal effort discounting-like behavior and that normal levels of dopamine signaling are saturating, such that dat-1 mutations have no effect.

Finally, we considered whether effort discounting-like behavior may reflect adaptation to the laboratory environment. The reference strain used in this study, N2, has been cultured for over 50 years in the laboratory setting, where it is typically fed small, easily consumed bacteria, such as Escherichia coli strain OP50. Consequently, it may have lost the ability to consume larger bacteria efficiently. To address this question, we tested whether any natural isolate strains would prefer normal over elongated bacteria. Accordingly, we performed T-maze assays on five such strains with normal and elongated food at OD 1.0. Natural isolate strains proved to be challenging to work with as they frequently escaped the maze or burrowed in the substrate. These traits reduced the number of countable worms per assay, leading to considerable variance in the data. Nevertheless, there was a significant overall effect of strain on mean preference index (Figure 5; Table 1, row 42). Post hoc testing revealed that only the strain JU775 had a preference index that was significantly different from N2 (Table 1, row 48) and that JU775’s preference was not distinguishable from zero (Table 1, row 49). Thus, JU775 was indifferent between normal and elongated food. On the other hand, mean preference of DL238 was both different from zero (Table 1, row 50) and indistinguishable from N2 (Table 1, row 44). That is, DL238 behaved like N2. We conclude that N2’s preference is probably not an adaptation to the laboratory environment. Between DL238 and JU775, there was a range of different responses, suggesting the possibility of genetic variation in preference for normal versus elongated food. This could reflect differences in effort tolerance or ease of consumption of elongated bacteria.

An increase in patch-leaving frequency in response to elongated bacteria has been reported previously (Scheer and Bargmann, 2023). There are some significant differences between that study and ours: (1) E. coli was used instead of Comamonas; (2) bacteria were elongated with the antibiotic aztreonam which renders food inedible to C. elegans (Gruninger et al., 2008; Ben Arous et al., 2009), whereas the cephalexin treatment we used resulted in elongated but edible food (Millet et al., 2022); (3) worms were video recorded for 40 min instead of 15 min; (4) food-patch leaving was defined using different behavioral criteria; (5) patch-entry rates were not considered. Nevertheless, our results confirm and extend the previous findings by showing that the effect of elongated food on patch-leaving rate is robust across bacterial species, varied degrees of edibility of elongated food, and differences in details of behavioral analysis.

Although C. elegans patch-leaving behavior has been described in detail before, placing it in an explicit economic context is new. This context provided a quantitative framework that enriched our study in two respects. First, it gave us a metric – the discount factor – for quantifying the extent to which effortful food was devalued by the worm. We observed that worms were indifferent when normal and elongated bacteria were at ODs 1.0 and 1.5, respectively, yielding a discount factor of 2/3. Second, this framework made testable predictions. At the level of effort determined by our cephalexin exposure protocol, any combination of optical densities of normal and elongated food in a 2:3 ratio should be an indifference point. This prediction was verified by measuring preference under conditions in which the OD of elongated food was both higher and lower than the original value of 1.5. This result shows that standard models of effort discounting may apply to C. elegans. It will now be interesting to test whether the discount factor is sensitive to effort, as assumed in many discounting models (Figure 1—figure supplement 1). This could be done by identifying indifference points when effort is manipulated by changing cephalexin exposure conditions to increase or decrease the extent of elongation.

We propose a model of effort discounting that builds upon previous studies of foraging in C. elegans. Once worms left the starting point of the T-maze, they were free to move back and forth between food patches at the end of each arm, encountering one patch then the other in serial fashion. Thus, our procedure was a foraging assay. In a simple model to explain the food preferences we observed, worms make an on-line assessment of food value in the current patch, which in turn alters patch-leaving dynamics, increasing the exit rates from cephalexin-treated patches as shown in Figure 3. The decision of whether to stay or leave a patch of food while foraging has been well studied in C. elegans. (Shtonda and Avery, 2006; Bendesky et al., 2011; Milward et al., 2011; Olofsson, 2014; Scheer and Bargmann, 2023). These studies point to a similar mechanism. The key advance here is the demonstration that the worm’s foraging decisions are consistent with formal effort-discounting models under our conditions.

Stay-or-leave decision-making based on assessment of current conditions has been observed in the context of effort discounting in other organisms. One procedure widely used in rodent studies is the progressive ratio task (Salamone et al., 2018). In this task, animals can either lever press for a preferred food or consume a less preferred food that is freely available nearby. However, the number of lever presses required to obtain preferred food increases as a function of the cumulative number of lever presses until the effort cost of obtaining preferred food becomes too high and the animal switches to a freely available food. This is a foraging assay in the sense that the lever and the freely available food are patches, and the animal decides whether or not to leave the ‘lever’ patch. Our assay also has parallels in the human effort-discounting literature. In one example (Bonnelle et al., 2015), participants are presented with a series of virtual apple trees. They can see how many apples are in the current tree and how much effort (squeezing a handgrip) is required to gather them. Their task is to decide whether or not to gather apples based on the perceived cost and benefit. Here, the trees are equivalent to patches, and the decision to gather apples or not is analogous to a stay–leave decision in patch foraging. Thus, stay-or-leave decision-making may be a conserved solution to effortful foraging tasks.

Fully intact dopamine signaling is required for normal effort discounting-like behavior in C. elegans. This finding is broadly consistent with the known requirements of dopamine signaling for foraging behavior in C. elegans, as our T-maze assay is essentially a foraging assay. C. elegans dopamine receptors DOP-1, DOP-2, and DOP-3 are required for normal responses to the presence or absence of food (Chase and Koelle, 2007; Suo et al., 2009; Bastien et al., 2024). Dopamine itself is required for area-restricted search in which worms, suddenly finding themselves outside a food patch, engage in a period of frequent locomotory reorientations in an attempt to relocate the food (Hills et al., 2004).

However, the relationship between dopamine signaling and effort discounting-like behavior in C. elegans remains an area for future research. For example, although mutations in the dopamine pathway cause a range of locomotor phenotypes (Chase et al., 2004; McDonald et al., 2007; Ezak and Ferkey, 2010) and may affect metabolism (de Almeida Barros et al., 2014), only dop-3 displays a phenotype in our effort-discounting assay, suggesting a specific role of dop-3 in effort discounting in C. elegans. Understanding the precise role of dop-3 and other components of the dopamine signaling pathway in effort discounting will require more sensitive behavioral assays as well as other technical approaches, including cell-specific rescues. The fact that decreased dopamine signaling in the dop-3 mutant rendered worms indifferent between the food options suggests that decreased dopamine signaling increases tolerance for high-effort food, in contrast to its effects in rodents. This discrepancy may reflect a true difference in the function of dopamine in effort discounting in C. elegans and rodents, or methodological differences. In particular, manipulations of dopamine signaling in rodent studies mainly involve acute experiments using pharmacological agents, whereas we studied worms with chronic, lifelong alterations in dopamine signaling.

In summary, our study provides strong evidence that C. elegans exhibits effort discounting-like behavior. By demonstrating that worms devalue effortful food like higher organisms, we extend the phylogenetic boundary of this behavior. The application of an economic framework allowed us to quantify this effect and make testable predictions, confirming its robustness. Future research can now exploit the utility of C. elegans as a genetically tractable system to explore how environmental factors and genetic variations influence effort-based decision-making. Such an approach could provide deeper insights into the fundamental principles and molecular basis of cost–benefit decisions.

Two bacterial species were used: E. coli (OP50) is used for worm culture, and Comamonas spp. (DA1877) for assays. Bacterial cultures were grown in liquid lysogeny broth (LB) under sterile conditions at 37°C with agitation. For all assays, DA1877 was grown for 1 day, then ‘Control’ cultures were solubilized with M9 buffer (3 g KH₂PO₄, 6 g Na₂HPO₄, 5 g NaCl, 1 ml 1 mol/l MgSO₄, H₂O to 1 l); ‘cephalexin’ cultures were supplemented with a solution of M9 and cephalexin hydrate 97% to a concentration of 450 µmol/l and incubated for 2 hr. Cultures then underwent three rounds of concentration by centrifugation, followed by rinsing with 10 ml of M9 buffer before being resuspended to their final concentration. Concentration was defined as OD at 600 nm, as measured with a cell density meter (Laxco, DSM, Bothell, WA, USA). All measurements were performed on samples diluted into the linear range of the instrument (OD 0.1–1.0).

All strains were maintained at 20°C on 5 cm plates containing NGM seeded with E. coli OP50 bacteria as a food source (Brenner, 1974). For synchronization, 20 young adult worms were transferred to fresh 5 cm NGM plates. Adults were removed from the plates after laying eggs for 3 hr. The plates were maintained at 20°C until worms reached the young adult stage.

Pharyngeal pumping was measured electrophysiologically using the ScreenChip microfluidic system (InVivo Biosystems, Eugene, OR, USA). Synchronized young adults were preincubated with control or cephalexin-treated bacteria at OD 1 for 20 min before being transferred into the reservoir of a microfluidic device fitted with electrodes. Individual worms were moved one at a time from the reservoir into the recording channel and given 1–2 min to acclimate. Voltage transients associated with pharyngeal pumping were recorded for 5 min; such recordings are called electropharyngeograms. All recordings were performed within 90 min of the worms being loaded into the microfluidic reservoir. Mean pumping frequency was extracted using custom code written in Igor Pro (WaveMetrics, Lake Oswego, OR, USA). This code is included as a Supporting File titled Source code 1. The code operates on a set of voltage time courses obtained from a cohort of worms. It identifies the individual pumping events in each time series and computes the mean pumping frequency across the cohort.

T-mazes (Levichev et al., 2023) were constructed by cutting masks from sheets of 2 mm thick ethylene-vinyl acetate foam (Darice Craft Foam, Strongsville, OH, USA) using a fabric-cutting machine (Cricut, South Jordan, UT, USA). Masks were placed on blank 5 cm NGM plates. Maze arms were baited with 4.5 µl of various bacterial suspensions as described in the text. Synchronized young adults were washed by five rounds of centrifugation (30 s, 300 × g) followed by aspiration of the supernatant, then deposited by liquid transfer at the starting point of the maze. A transparent plastic disc was placed over the maze to eliminate air currents. Twelve plates were placed on a flatbed scanner (EPSON, model J221B, Los Alamitos, CA, USA) and imaged after an elapsed time of 60 min. Imaged worms were counted manually. The food preference index was computed as I=(nN−nL)/(nN+nL)\begin{document}$I=\left (n_{\mathrm{N}}-n_{\mathrm{L}}\right)/\left (n_{\mathrm{N}}+ n_{\mathrm{L}}\right)$\end{document}, where nN\begin{document}$n_{\mathrm{N}}$\end{document} and nL\begin{document}$n_{\mathrm{L}}$\end{document} are the number of worms in the normal and elongated patches, respectively. In some experiments, a paralytic agent (sodium azide, NaN₃, 3 µl at 20 mM) was added to each food patch to prevent animals from leaving the patch of food after reaching it. Sodium azide diffuses through the agar over time, and its action is not instantaneous. To account for these effects, all worms within 5 mm of the end of the maze arm, rather than only those on food, were counted when calculating the preference index.

Custom 8-well plates (25 mm × 45 mm) were fabricated from acrylic plastic. Wells (diameter: 10 mm; depth: 6 mm) were filled with NGM agar. Bacterial suspension (2 μl, OD 1) was placed at the center of each well, forming a patch 3–4 mm in diameter. Synchronized 1-day-old adult worms were cleaned by transferring them to an unseeded NGM plate for several minutes. One worm was transferred to each well using a pick and placed at the center of the patch. The well plate was immediately placed on a stage 8 cm above a diffuse white-light source and imaged (2 or 5 frames per second, 23.4 μm/pixel) for 15 min using a video camera (AmScope, 1080P, Irvine, CA, USA) fitted with a telephoto lens (Nikon USA, AF Micro Nikkor 60 mm, Melville, NY, USA). The distance between the lens and the stage was 50 cm. Worm tracks and locomotion parameters were extracted using WormLab software (MBF Bioscience, Williston, VT, USA) and analyzed using custom routines in Igor Pro (WaveMetrics, Lake Oswego, OR, USA). To define the worm’s position with respect to the patch, its margin was traced manually.

Bacteria were grown under sterile conditions at 37°C with agitation in 500 ml of LB overnight, pelleted (5000 rpm, 3 min), washed with water, and re-pelleted. Pellets were stored at –80°C prior to analysis. Bacterial stocks were weighed and then lysed with CelLytic B Plus kit (Sigma-Aldrich, CB0050, Saint Louis, MO, USA) on the day of the assay.

Control and cephalexin-treated bacteria were prepared as described above. BacLight Green dye dissolved in DMSO was added to the bacteria at a final concentration of 0.1 µM. Following a 15-min incubation, bacteria were mounted between a microscope slide and a coverslip, then imaged on a spinning disk confocal microscope (Nikon Eclipse Ti-2E with Yokogawa CSU-W1; objective CFI60 Plan Fluor 40x Oil Immersion, N.A. 1.3; Melville, NY, USA).

Fat storage was assessed using the procedures of Soukas et al., 2009 and Stuhr and Curran, 2020. A 0.5% isopropanol solution of Oil-Red-O (Sigma-Aldrich, Oil-Red-O powder 00625, Saint Louis, MO, USA) was prepared several days before the assay. Bacterial cultures (Control and cephalexin) were prepared as described above and adjusted to OD 3 before being loaded onto fresh NGM plates. Seventeen young adult worms were placed on these plates and allowed to lay eggs for 5 hr before being removed. The plates were incubated at 20°C for 2 days. L4 larvae were harvested and washed in PBS + 0.01% Triton X-100. Worms were fixed with a 60% isopropanol solution and incubated with the Oil-Red-O solution for 2 hr on a rotating plate. Worms were washed twice with PBS + Triton 0.01% solution, first for 30 min, then for 1 hr, on a rotating plate. Stained worms were imaged on a microscope slide at 32×, and staining intensity was quantified using custom routines in Igor Pro (WaveMetrics, Lake Oswego, OR, USA).