Odor-dependent temporal dynamics in Caenorhabitis elegans adaptation and aversive learning behavior

Nematode culture and strains

Worms were grown and maintained at 20 °C on Nematode Growth Medium (NGM) plates seeded with E. coli OP50 as described previously (Brenner, 1974). Strains used for this study, N2, ceh-36, odr-7, pyIs500 ((p)odr-3::GFP::egl-4), were obtained from the Caenorhabditis Genetic Center (University of Minnesota, USA).

Behavior assays

A rough screen for behavioral attraction to a large set of odor chemicals was first conducted which identified dozens of attractive odors. From this smaller group, a subset of 17 highly attractive odors were tested for chemotaxis in wild-type and mutant strain worms, among which were many odors that have not been tested previously. Standard odor chemotaxis assays were carried out using previously established protocols with a few changes (Bargmann, Hartwieg & Horvitz, 1993). The following 17 odors in Table 1 were used: benzaldehyde, butanone, isoamyl alcohol (Sigma-Aldrich, St. Louis, MO, USA); isobutyric acid, 2-isobutyl thiazole, dimethylthiazole, 2,4,5-trimethylthiazole, 2-methylpyrazine, 2-heptanone, 4-chlorobenzyl mercaptan, butyric acid, 1-pentanol, benzyl mercaptan, 2-cyclohexylethanol, benzyl proprionate (Alfa Aesar, South Korea); 1-methylpyrrole, 2-ethoxythiazole (TCI, Tokyo, Japan); diacetyl (Acros Organics, Geel, Belgium). All odors were diluted from original stock to a 1:100 dilution with ethanol except for benzaldehyde (1:200) and 2,4,5-trimethylthiazole (1:1,000). A total of 4 µl of diluted odor or ethanol was placed on the assay plate as attractant and counterattractant, respectively. A total of 2 µl of NaN₃ was placed with the attractant and counterattractant. Adult worms were washed in S-basal buffer (5.85 g NaCl, 1 g K₂HPO₄, 6 g KH₂PO₄) three times, placed on the assay plate, and allowed to move freely for at least one hour before counting. We set a threshold for whether a neuron is required for detection as an AI in either ceh-36 or odr-7 mutants that was less than 50% of the wild-type AI.

Standard odor memory behavior assays were carried out similar to previous published protocols (Colbert & Bargmann, 1995; L’Etoile & Bargmann, 2000). Before assaying behavior, worms were washed in S-basal buffer three times, and placed in 1 ml of benzaldehyde or 2,4,5-TMT diluted at 1:10,000 in S-basal buffer for 0–120 min at 10 min or 20 min intervals. After odor exposure, worms were washed three times in S-basal and behavior was assayed.

Real-time odor behavior assay

Media for the assay plate is as follows: 1.6% Difco Granulated Agar (Beckton Dickinson, Franklin Lakes, NJ, USA) was dissolved in water by heating, and 1mM of CaCl₂, 1mM of MgSO₄ and 5 mM of KPO₄ buffer (108.3 g KH2PO4, 35.6 g K2HPO4, H2O to 1 liter) was mixed with the agar. Media was dispensed into 12.5 cm × 12.5 cm × 2 cm square plastic culture plates (SPL Lifescience, Pocheon, South Korea), and allowed to harden for at least 3 h. Odors were diluted with ethanol to the following concentrations: benzaldehyde (1:100), butanone (1:100), diacetyl (1:100), isoamyl alcohol (1:1,000), 2-heptanone (1:1,000), 2,4,5-TMT (1:500), 4-chlorobenzyl mercaptan (1:500), 2-ethoxythiazole (1:300), 2-cyclohexylethanol (1:5,000), 1-methylpyrrole (1:500), 1-pentanol (1:10,000), 2-isobutylthiazole (1:100), 2-methylpyrazine (1:10). Diluted attractant odors were placed in the middle of the “A” section of the assay plate (Fig. 1), and a dab of OP50 strain E. coli bacteria from a standard nematode growth media plate with a platinum worm picker was placed in the middle of the “F” section of the assay plate as a counterattractant, no larger than 3 mm diameter. The counterattractant itself is a very weak attractant (Fig. 1B, control), and overall aids this assay in resolving the attenuation of odor attraction after memory has formed. Washed worms are then placed in the middle of the plate, and dried by wicking the buffer using an unscented tissue. The plates were tapped somewhat vigorously which aids in worm movement away from the center of the plate. Worms were counted in each section A, B, E, and F every ten minutes for 120 min and AI was calculated (AI = [[2(#of worms in A) + (#of worms in B)] − [(#of worms in E) + 2(#of worms in F)]]∕2(#of worms in A + B + E + F)). 50% total behavior change was calculated as (30 min AI − 120 min AI)/2, and time to 50% behavior change for each odor and each trial was determined.

Quantification of GFP::EGL-4 localization

Subcellular localization of GFP::EGL-4 in py Is500 ((p)odr-3::GFP::EGL-4; (p)odr-1::RFP) integrated strain worms during odor memory formation was performed as previously published (Lee et al., 2010). Briefly, py Is500 animals were exposed to odor dilutions of either benzaldehyde (1:10,000), 2-ethoxythiazole (1:10,000), 1-methylpyrrole (1:10,000), 4-chlorobenzyl mercaptan (1:5,000), or 2,4,5-TMT (1:10,000) in S-basal for 0–120 min at 20 min intervals. After odor exposure, worms were placed on slides containing a 2% agar dried pad with 1 µl of 1M NaN₃ added to anesthetize the animals and observed under a fluorescent microscope. 20 animals were counted per sample, and EGL-4 localization was determined as either cytoplasmic or nuclear in the AWC neuron for each animal. In the rare case when localization in an animal could not be categorized as either nuclear or cytoplasmic, we would count that worm as nuclear, and then counted an extra animal. The next time this occurred, we counted that worm as cytoplasmic, and counted an extra animal. After the data was plotted, a regression equation was determined (Microsoft Excel) for each odor as follows: benzaldehyde (y =  − 0.6424 × 3 + 7.3661 × 2 − 11.992 ×  + 13.304), 2-ethoxythiazole (y =  − 0.8924 × 3 + 9.7545 × 2 − 16.514 ×  + 19.071), 1-methylpyrrole (y =  − 0.9833 × 3 + 12.252 × 2 − 32.264 ×  + 31.657), 4-chlorobenzyl mercaptan (y =  − 0.4456 × 3 + 6.1379 × 2 − 14.315 ×  + 17.5), 2,4,5-TMT (y =  − 0.3003 × 3 + 3.4079 × 2 − 1.5239 ×  + 5.8048). Fit and statistical significance of the regressions were determined by calculating Pearson’s correlation coefficient and ANOVA analysis. Equations were used to calculate the approximate time at which 50% of animals display nuclear EGL-4.

Attraction to novel odors mediated by the AWC or the AWA primary sensory neurons

In order to test adaptation and aversive learning to a larger panel of odors than has been previously tested, we first identified 17 highly attractive odors (see Methods), among which were many odors that have not been tested previously (Table 1, Fig. S1). To identify the neuron responsible for sensing each odor, we subjected wild-type and olfactory sensory neuron mutants to these odors in a standard chemotaxis assay (Bargmann, Hartwieg & Horvitz, 1993; Lanjuin et al., 2003; Sengupta, Colbert & Bargmann, 1994). As predicted, diacetyl attraction is lost in the AWA- odr-7 mutants and maintained in AWC- ceh-36 mutants (Table 1; Fig. S1). On the contrary, benzaldehyde attraction disappears in the AWC- ceh-36 mutants but is normal in AWA- odr-7 mutants. Finally, 2,4,5-trimethylthiazole which is sensed by both AWA and AWC neurons (Bargmann, Hartwieg & Horvitz, 1993) remains attractive in both AWC- and AWA- mutants. When we tested the other odors, we found that attraction to most of the odors required the AWC sensory neurons, with only butyric acid and isobutyric acid sensed by the AWA neuron (Table 1; Fig. S1). Benzyl proprionate attraction was effectively lost in both AWC- and AWA- mutants. Hence, we have identified many novel attractive AWC-sensed odors.

Odor adaptation and aversive learning in the real-time odor behavior assay

Adaptation and aversive learning have been studied for the AWC-sensed odor benzaldehyde (Colbert & Bargmann, 1995; L’Etoile et al., 2002; Lee et al., 2010). However, the temporal dynamics of aversive learning and memory to other odors have not been tested thoroughly. Therefore, we sought to characterize the change in odor attraction over time to a multitude of AWC-sensed odors.

Due to the limitations of the standard odor assay, we designed a new odor behavior assay we call the “real-time odor behavior assay” described in the Methods (Fig. 1A), in which worms are tracked every 10 min as they move towards or away from the odor in a two-hour time frame. When tested with benzaldehyde, worms spent the first 30–40 min moving towards the odor, which was demonstrated by the increase in attraction index (AI). AI remained high for another 30 min, until it started to decrease at 60 min and continued to fall during the remainder of the assay (Fig. 1B). In contrast, worms given only the diluent ethanol showed no such patterns of movement and their attraction index remained steady throughout the assay (Fig. 1B). The pattern of initial attraction followed by decrease in attraction was similar to what is observed in the standard assay (Fig. 2A). To make a side-by-side comparison with the standard assay, we took into consideration that both the tracking of worm movement and odor exposure occur concurrently with the assay. Therefore, we made the 30-minute time point to be the “beginning” of the real-time assay. When compared this way, similar patterns of overall decrease in attraction could be observed in both assays.

To test whether this assay could indeed be used to assess adaptation and aversive learning to an odor, we took advantage of mutants that are defective in each of these odor behaviors. In the traditional assay, adaptation-defective cng-3 mutants showed high attraction for the first 60 min followed by a precipitous drop, demonstrating selective deficit for adaptation but not aversive learning (Fig. 2A). cng-3 worms showed a similar pattern in the real-time odor assay, where decline in attraction index occurred at a later time point compared to wild type, although the difference did not seem as drastic at first glance. However, when we compared attraction at the 0 and 40 min time points in the standard assay to the 30 and 70 min time points in the real-time assay, the decrease in attraction seen in wild-type N2 animals was absent in cng-3 mutants for both the standard and real-time assays (Fig. 2C).

Next, we looked at whether aversive learning could be assessed in the real time assay by using the aversive learning-defective adp-1 mutant. Consistent with previous studies, adp-1 mutants displayed no change in attraction even after a 120 min odor exposure in the standard assay (Colbert & Bargmann, 1995; Fig. 2A). Likewise, in the real time assay, adp-1 mutants maintained high attraction throughout the course of the assay (Fig. 2B). Thus, both adaptation and aversive learning can be resolved in the real-time odor behavior assay.

Timing of adaptation and aversive learning varies by odor

Since we demonstrated that our assay can accurately assess the timing of benzaldehyde adaptation and aversive learning, we conducted a comprehensive temporal analysis of behavior change to a palette of 12 other odors, nine of which have never been tested for odor adaptation or learning. These include alcohols, ketones, thiols, thiazoles, aromatics, and pyrazines. 30 min into the assay when we start the analysis, we observe differences in attraction to each odor ranging from 0.91 AI (2,4,5-trimethylthiazole) to 0.50 AI (2-heptanone). However, over time, the total decrease in attraction to all of the odors is similar over the entire assay (average AI decrease = −0.475 ±0.025) with one exception, the odor cyclohexylethanol, which displays only a modest attenuation (AI decrease = −0.19).

In all the odors tested, one can observe a distinct pattern of change in worm behavior towards these twelve odors over time. For AWC-sensed odors, behavior change generally occurs relatively slowly at the beginning of the assay then proceeds more quickly towards the end (Fig. 3). Interestingly, for the odor diacetyl, which is the only odor here not sensed by the AWC neuron, behavior change appears to arise early and fairly consistently during the entire assay (Fig. 3).

Although overall changes in odor behavior patterns over the 120 min assay is similar for most of the odors tested, the timing of adaptation and aversive learning varies depending on the odor. For instance, benzaldehyde attraction decreases overall from 0.79 AI to 0.27 from 30 to 120 min, and 50% of this behavior change occurs at about 80 min (Fig. 3; Fig. S2). We then calculated the time at which 50% behavior change occurs for each of the odors (Fig. 3; Fig. S2A), and observed early trends in behavior change in diacetyl attraction (76.0 min) and 1-methylpyrrole attraction (76.3 min), as well as other odors (Fig. 3, Fig. S2A). On the other hand, late trends in behavior change were observed in 2,4,5-trimethylthiazole (2,4,5-TMT) attraction (100.0 min) and 4-chlorobenzyl mercaptan (4-CB) attraction (105.5 min) (Fig. 3, Fig. S2A). To confirm whether the late change in behavior was actually a result of aversive learning, we tested the attraction of adp-1 mutants to 4-CB and 2,4,5-TMT attraction in the real-time assay (Fig. S3). Wild-type animals showed large decreases in attraction over the whole assay for both 4-CB (change in AI =  − 0.478) and 2,4,5-TMT (change in AI =  − 0.349). However, the change in behavior over time was minimal in adp-1 mutants for both 4-CB attraction (change in AI =  − 0.226) and 2,4,5-TMT attraction (change in AI =  − 0.009). Thus, the late trend in decreased odor attraction for the two odors is likely due to late odor learning. We also confirmed that the early trend in benzaldehyde associative learning and the late trend in 2,4,5-TMT learning can be observed using either the real-time assay or the standard assay (Fig. S4). Thus, temporal dynamics of adaptation and aversive learning varies depending on the specific odor.

One possibility that may explain the variance in odor-dependent behavior changes over time is the differences in attractive strength for each odor. Odors elicit differential attraction in the standard assay (Table 1) and in our real-time behavior assay (Fig. 3) when worms first encounter the odors. This can be due to both inherent qualities of the odor and odor concentration (Bargmann, Hartwieg & Horvitz, 1993). To investigate the relationship between the initial attraction and subsequent behavior changes to the odor, we first grouped odors into either “late-trending” odors in which behavior change occurs late and “early-trending” odors in which behavior change occurs earlier (Fig. S3). The three late-trending odors, 4-CB, 2,4,5-TMT, isobutylthiazole, had an average initial AI of 0.857 ±0.045, which was higher than the initial AI of the early-trending odors (average AI = 0.721 ± .098; Fig. S5). However, the difference between initial AI of late and early-trending odors was not significant (Fig. S5).

We next wondered whether animals that initially displayed a low attraction to late-trending odors could change their behavior toward the odor faster than animals that initially displayed a high attraction to the same concentration of odor. However, animals that had an initially low AI to 4-CB (average AI = 0.71 ± .088) and 2,4,5-TMT (average AI = 0.81 ± .043) did not display faster behavior changes than animals that had an initially high AI (Fig. S6). Therefore, we did not find a significant relationship between temporal patterns of behavior change and initial attraction to the odors.

Timing of aversive learning formation correlates with the timing of EGL-4 nuclear localization

Odor adaptation and aversive learning to AWC-sensed odors requires EGL-4 activity: cytoplasmic EGL-4 activity directs short-term adaptation whereas aversive learning requires EGL-4 to translocate to the nucleus and phosphorylate nuclear targets (Juang et al., 2013; L’Etoile et al., 2002). Because we showed that behavior changes over time in response to both benzaldehyde and 2,4,5-TMT require adp-1 (Fig. 2B, Fig. S3), we wondered whether these changes in odor attraction were correlated to the sub-cellular localization of a functional GFP-tagged EGL-4 in the AWC neuron (Fig. 4A). We narrowed down our pool of odors to 5 AWC-sensed odors that induced consistent changes in behavior and grouped them into the early-trending odors benzaldehyde, 2-ethoxythiazole, 1-methylpyrrole and the late-trending odors 2,4,5-TMT and 4-CB. We find that these three early-trending odors induce a 50% change in behavior after an average of 83.3 min whereas the two late-trending odors induced a 50% change at 101.8 min, significantly later than early-trending odors (Fig. 5). We then exposed worms to early and late-trending odors and observed EGL-4 localization in the AWC neuron every 20 min for 2 h (Figs. 4B–4F). Whereas the odor butanone is sensed by one of the AWC neuron pair and EGL-4 nuclear translocation is induced in only one of the AWC neurons, all the odors tested here could induce EGL-4 translocation in both AWC neurons indicating that they are likely sensed by both AWC neurons. The maximal percent of animals with nuclear EGL-4 ranged from 60% to almost 80% for each odor (Figs. 4B–4F). However, neither longer odor incubation nor increasing odor concentration could substantially increase maximal nuclear EGL-4 for 4-chlorobenzyl mercaptan or 2,4,5-trimethylthiazole (Figs. S7 and S8).

The temporal dynamic of EGL-4 nuclear localization in response to the early- and late-trending odors seems to vary depending on the odor. To analyze this in detail, we calculated a regression for the time curves for each odor (Figs. 4B–4F), and then estimated the time at which 50% EGL-4 nuclear localization was reached for each odor (Fig. 5). We found that EGL-4 nuclear localization occurred much faster in the early trending odors (58 min, 70 min, 75 min for 2-ethoxythiazole, benzaldhehyde and 2-methylpyrrole, respectively), but significantly later in the late-trending odors (90 min and 89 min for 4-CB and 2,4,5-TMT, respectively). Thus, we saw a correlation between the timing of behavior change and EGL-4 nuclear translocation (Fig. 5). Taken together, the timing of odor behavior changes for AWC-sensed odors may be regulated by EGL-4 sub-cellular localization and aversive learning mechanisms.