Behavioral Forgetting of Olfactory Learning Is Mediated by Interneuron-Regulated Network Plasticity in Caenorhabditis elegans

Abstract Forgetting is important for animals to manage acquired memories to enable adaptation to changing environments; however, the neural network in mechanisms of forgetting is not fully understood. To understand the mechanisms underlying forgetting, we examined olfactory adaptation, a form of associative learning, in Caenorhabditis elegans. The forgetting of diacetyl olfactory adaptation in C. elegans is regulated by secreted signals from AWC sensory neurons via the TIR-1/JNK-1 pathway. These signals cause a decline of the sensory memory trace in AWA neurons, where diacetyl is mainly sensed. To further understand the neural network that regulates this forgetting, we investigated the function of interneurons downstream of AWA and AWC neurons. We found that a pair of interneurons, AIA, is indispensable for the proper regulation of behavioral forgetting of diacetyl olfactory adaptation. Loss or inactivation of AIA caused the impairment of the chemotaxis recovery after adaptation without causing severe chemotaxis defects in the naive animal. AWA Ca2+ imaging analyses suggested that loss or inactivation of AIA interneurons did not affect the decline of the sensory memory trace after the recovery. Furthermore, AIA responses to diacetyl were observed in naive animals and after the recovery, but not just after the conditioning, suggesting that AIA responses after the recovery are required for the chemotaxis to diacetyl. We propose that the functional neuronal circuit for attractive chemotaxis to diacetyl is changed temporally at the recovery phase so that AIA interneurons are required for chemotaxis, although AIAs are dispensable for attractive chemotaxis to diacetyl in naive animals.


Introduction
Animals are able to learn and form memories, depending on the experience they gain from their surroundings; however, to adapt to changing environments, it is essential that dispensable information is discarded to manage accumulating memories. Recent studies reveal that memories can be actively forgotten by interference with other memories or by activating forgetting in neurons that are important for maintaining memories (Hardt et al., 2013;Davis and Zhong, 2017). However, the molecular mechanisms and neural networks engaged in forgetting are not well understood.
In invertebrates, despite their simple neural networks, several studies showed that forgetting is actively regulated. In Drosophila, dopamine neurons regulate both learning and active forgetting through distinctive dopamine receptors in mushroom body neurons (Berry et al., 2018(Berry et al., , 2012. One of the dopamine receptors in mushroom body neurons, dDA1, leads to memory formation (Berry et al., 2012), while for forgetting, another receptor, DAMB, activates Scribble scaffold to initiate forgetting by actin cytoskeleton remodeling (Shuai et al., 2010;Cervantes-Sandoval et al., 2016;Berry et al., 2018Berry et al., , 2015Berry et al., , 2012. C. elegans is also used to study active forgetting through olfactory adaptation, a form of associative learning (Colbert and Bargmann, 1995;Bargmann, 2006).The forgetting of diacetyl olfactory adaptation, which is sensed by AWA olfactory neurons, is regulated by the TIR-1/JNK-1 pathway in another type of olfactory neuron, AWC. In wild-type and mutant animals, which are defective in the TIR-1/JNK-1 pathway, the sensory Ca 21 response of AWA neurons to diacetyl is positively correlated with behavioral change through memory formation and forgetting, suggesting that the sensory response can be considered as the sensory memory trace and that the memory trace in AWAs is actively and non-cell-autonomously regulated by AWCs (Inoue et al., 2013;Kitazono et al., 2017). In addition, a membrane protein, MACO-1, and a tyrosine kinase pathway, SCD-2/HEN-1, regulate the forgetting (Kitazono et al., 2017). Another study showed that, similar to Rac1 in Drosophila, the Arp2/3 complex, which regulates the actin cytoskeleton in AVA interneurons, is important for forgetting downstream of the RNA binding protein Musashi (Hadziselimovic et al., 2014). Although these studies indicate that active forgetting is important, even in simple learning paradigms of model organisms, the corresponding neural network has not been fully revealed.
Here, we demonstrate that a pair of neurons in C. elegans, AIA interneurons, which are the first-layer interneurons in olfactory circuits, is required to regulate forgetting processes of olfactory adaptation. Although absence or inactivation of functional AIA interneurons (AIA -) slightly affect attractive chemotaxis to diacetyl, it caused prolonged retention of the olfactory adaptation to diacetyl, suggesting that AIAs accelerate forgetting. Calcium imaging analyses showed that, although the behavioral response in AIAanimals did not recover after cultivation for 4 h, the calcium responses to diacetyl in AWA animals were recovered. These results suggest that AIAs are indispensable for the behavioral response of the olfactory adaptation forgetting mechanism, probably because the functional neuronal circuit is changed temporally so that AIAs are required for the chemotaxis.

Strains and culture
All strains were cultured on nematode growth medium (NGM) agar plates seeded with Escherichia coli strain OP50 (Brenner, 1974) and were grown at 20°C before experiments. In all experiments, we used young adult hermaphrodites prepared as described in each section (Table 1).

Behavioral assay
Chemotaxis toward attractive odorants was performed on assay plates (2% Bacto agar, 50 mM NaCl, 10 mM K 2 HPO 4 , pH 6, 1 mM MgSO 4 , 1 mM CaCl 2 ) with 1:100 dilutions of odorants (diacetyl and isoamyl alcohol; Bargmann et al., 1993). During behavioral assays, animals were placed in the middle of the assay plate while the odorant and control solution (ethanol, the odorant diluent) were spotted on opposite sides of the plate. The chemotaxis index was calculated as (A -B)/N, where A refers to the number of animals within 1.5 cm of the odorant spot, B refers to the number of animals within 1.5 cm of the control spot, and N is the total number of animals. In the forgetting assay (Inoue et al., 2013), adult animals were first washed three times with S-basal buffer (100 mM NaCl, 50 mM K 2 HPO 4 , pH 6, 0.02% gelatin; naive) and pre-exposed to 1:5000 diluted diacetyl or isoamyl alcohol in S-basal buffer with slow rotation for 90 min at room temperature (adaptation). Next, the worms were washed once and allowed to recover on OP50-seeded NGM plates for 4 h (recovery). In the extended forgetting assay, animals were recovered on OP50seeded NGM plates for 24 h, and behavioral assays after recovery were conducted after 4, 8, and 24 h of recovery.

Calcium imaging
Calcium imaging of AWA neuron responses toward diacetyl was performed using AWA-cameleon YC3.60-expressing animals (Inoue et al., 2013). The day before imaging, 30-45 animals (L4 -young adults) were picked and cultured at 20°C. A 1:10 À7 dilution of diacetyl was used for odor stimulation, and a 1:10 À3 dilution of diacetyl was used for adaptation. Adaptation and recovery were conducted as described for the behavioral assay. During Ca 21 imaging, odor stimulation was applied to the animal for 60 s (30th to 90th second of recording). Fluorescence images of AWA sensory neurons were acquired using a microscope (model BX53-FL, Olympus) equipped with a 60Â objective lens (UPLSAPO 60XW, Olympus) and a dual CCD camera (model ORDA-D2, Hamamatsu). Cameleon YC3.60 was excited using X-Cite 120 fluorescence lamp illuminators (EXFO). The fluorescence ratio of yellow fluorescent protein (YFP) to cyan fluorescent protein (CFP) in Cameleon YC3.60 was analyzed using an AQUACOSMOS system (version 2.60; Hamamatsu). The  calculation (R max -R 0 )/R 0 was performed as the peak amplitude of the YFP/CFP ratio during the first 10 s interval after stimulation (R max ) relative to the mean basal ratio (R 0 ) during the 10 s interval before stimulation. The relative Ca 21 response was evaluated by normalized (R max -R 0 )/R 0 with respect to the average naive value. For calcium imaging on AIA neurons, we used animals expressing GCaMP6f in AIA neurons by gcy-28.d promoter. Animals were cultivated as for AWA imaging. A 1:10 À7 dilution of diacetyl was used for odor stimulation, and a 1:10 À3 diacetyl was used for adaptation. Fluorescent images were acquired using a microscope (model BX53-FL, Olympus) equipped with a 60Â objective lens (UPLSAPO60XS2, Olympus) and an ORCA-Flash camera with extended focus device of Gemini-2c (Hamamatsu). GCaMP6f was excited using a 470 nm laser (LDI, 89 North) with a dichroic mirror (488/543/ 635, Semrock) and fluorescence images were captured at 50 ms through an emission filter (512/25, Semrock). The neurite of AIA neurons was analyzed. The fluorescent intensities were normalized by the average response (R 0 ) of a 5 s time period prior the stimulation.
plates 16-20 h after transfer (Ishihara et al., 2002). Injection markers, such as myo-3p::gfp, lin-44p::gfp, unc-122p:: mCherry, and unc-122p::gfp, were used to distinguish transgenic animals with extrachromosomal transgenes, and the values of animals with extrachromosomal transgenes were compared with those of animals without transgenes on the same plates as internal controls.
All values are presented as either the mean 6 SEM in a line graph or box plot. Data analyses were performed using Bell Curve for Excel (version 3.22; Social Survey Research Information Co., Ltd.). Statistical significance between means was determined by Student's t test or two-way ANOVA followed by a post hoc t test with Bonferroni's correction. Sample sizes and statistical values are noted in the figure legends (Table 2).

Results
AIA interneurons are required to regulate forgetting in AWA olfactory adaptation C. elegans shows strong attractive chemotaxis to diacetyl, which is mainly sensed by AWA sensory neurons. After animals are exposed to diacetyl without food for 90 min, they show significantly weaker responses to diacetyl (olfactory adaptation; Colbert and Bargmann, 1995). The conditioned animals are able to recover the attractive chemotaxis toward diacetyl to a level similar to that of naive animals after cultivation with food for 4 h (recovery), and we consider this recovery as forgetting (Inoue et al., 2013). Consistent with the behavioral change, the Ca 21 responses to diacetyl in AWA neurons are decreased after conditioning and recover with cultivation. This correlation between behavior and sensory responses is also observed in mutants defective in the TIR-1/JNK-1 pathway, which function in AWC sensory neurons. In the tir-1 (tm3036)-null mutant, naive animals show the sensory response to diacetyl and, after conditioning, the Ca 21 responses decrease to levels similar to those of wild-type animals. However, in tir-1-null animals, similar to the behavioral changes, the Ca 21 responses in AWAs to diacetyl do not recover with cultivation. Therefore, the forgetting of diacetyl olfactory adaptation in AWA neurons is regulated by AWC sensory neurons via the TIR-1/JNK-1 pathway (Inoue et al., 2013). AWC neurons do not make direct connections to AWAs; therefore, other neurons may be involved in this regulation. Figure 1A shows the olfactory circuit including olfactory sensory neurons, AWAs, and AWCs, and their downstream interneurons (White et al., 1986;Chalasani et al., 2010Chalasani et al., , 2007Cook et al., 2019;Dobosiewicz et al., 2019). As shown in Figure 1A, AWCs and AWAs mainly relay signals to the first-layer interneurons AIA, AIB, and AIY. Among these, we first examined whether AIB and AIY interneurons, the main synaptic target of AWCs, are involved in forgetting by using animals with genetically ablated AIB and AIY, in which cell-specific cell death is promoted by expressing mouse Caspase 1 (Casp1; Satoh et al., 2014). However, in animals without AIBs and/or AIYs, we detected no significant differences in changes of chemotaxis to diacetyl among naive, conditioned, and recovered animals (Fig. 1B,C). This indicated that AIB and AIY interneurons are dispensable for the regulation of forgetting in this olfactory adaptation.

AIA interneurons accelerate forgetting of olfactory adaptation
Next, we examined whether AIAanimals completely lost the ability to forget, or decelerated the forgetting progress, as in tir-1(tm3036)-null animal ( Fig. 2; Inoue et al., 2013). To test this, we analyzed the time course of memory retention for up to 24 h of recovery (4, 8, and 24 h after conditioning; Fig. 2). In the first 4 and 8 h of recovery, although wild-type animals showed full recovery of chemotaxis, AIAand tir-1 (tm3036)-null animals showed very weak recovery. After 24 h, AIAanimals, similar to tir-1(tm3036)-null mutants, showed almost full recovery to diacetyl, suggesting that, even without AIAs, animals can AWA neurons recovered their sensory response to diacetyl after adaptation even in the absence of AIAs The diacetyl-evoked Ca 21 response in AWAs is correlated with behavioral change in naive, conditioned, and recovered animals. Therefore, the weakened Ca 21 response in AWAs after conditioning can be considered a sensory memory trace. Consistent with this, in tir-1 (tm3036)-null mutants, similar to its behavioral response, a weakened Ca 21 response in AWAs after conditioning did not recover after 4 h of recovery ( Fig. 3; Inoue et al., 2013). To examine whether the ablation of AIAs causes prolonged weakened Ca 21 responses in AWAs after  -1(tm3036), and AIAanimals in naive, adaptation, and recovery phases (n ! 21). The black line represents the application of odor stimulation (1:10 À7 dilution of diacetyl). B, Relative Ca 21 responses of AWAs in wild-type, tir-1(tm3036), and AIAanimals. Values are normalized to the average naive value in respective animals. Boxes, First to third quartiles (Q1 25th to Q3 75th percentile) of each dataset; black line in the boxes, medians; black dots, mean; whiskers, minimum and maximum, excluding outliers (beyond 1.5-fold interquartile range from Q3 and Q1); x, outliers (n ! 21, two-way ANOVA, F strain(2,247) = 3.6626, p = 0.0271, h 2 = 0.0288; mean 6 SEM) e . Post hoc t test with Bonferroni's correction. A, Error bars represent the SEM.
conditioning, we analyzed Ca 21 responses of AWAs to diacetyl in AIAanimals (naive, adapted, and recovered). In contrast to the behavioral response, the Ca 21 response in AIAanimals was recovered after recovery for 4 h (Figs. 1D, 3), suggesting that the loss of AIAs decelerates forgetting but not through the inhibition of sensory recovery in AWAs.

AIA neurons can respond to diacetyl in naive animals and after the recovery
The loss of the functional AIA neurons caused a defect in chemotaxis to diacetyl after the recovery, but not before the conditioning. Recently, Larsch et al. (2015) reported that AIA neurons respond to diacetyl stimulation. Therefore, by using animals expressing GCaMP6f specifically in AIA, we analyzed the responses of AIA neurons to diacetyl in naive animals, immediately after conditioning, and after the recovery. As shown in Figure 4, we found that in naive animals and after the recovery, the fluorescent intensities of AIA neurons responding to the diacetyl stimulation were significantly increased, but such changes were not seen in those immediately after conditioning. This result is consistent with the importance of AIA in the chemotaxis to diacetyl after the recovery.
AIA interneurons regulate forgetting downstream of the TIR-1 pathway AIA interneurons are not required to regulate the Ca 21 responses in AWAs after recovery; therefore, we suspected that AIA interneurons might regulate forgetting independently of the TIR-1/JNK-1 pathway. We examined Figure 4. The Ca 21 responses to diacetyl of AIA interneurons. A-C, Left, Ca 21 responses of AIAs in wild-type animals in naive, adaptation, and recovery phases (n = 36; mean 6 SEM). Black line represents the application of odor stimulation (1:10 À7 dilution of diacetyl). Right, Ca 21 responses of AIAs before (color; 2.5-5 s) and after (pattern; 7.5-10 s) stimulation in wild-type animals in naive, adaptation, and recovery phases. Boxes, First to third quartiles (Q1 25th to Q3 75th percentile) of each dataset; black line in the boxes, medians; black dots, mean; whiskers, minimum and maximum, excluding outliers (beyond 1.5-fold interquartile range from Q3 and Q1); x, outliers (naive: n = 36, t (35) = 2.2336, p = 0.032; adaptation: t (35) = 1.5505, p = 0.13; recovery: t (35) = 2.5616, p = 0.0149) f . Student's t test. A-C, Left, Error bars represent the SEM. the genetic relationship between AIA interneurons and the TIR-1/JNK-1 pathway in the forgetting mechanism, by analyzing genetic epistasis using tir-1(ok1052 gf) animals, which show weak adaptation after conditioning probably because of hyperforgetting (Chuang and Bargmann, 2005;Inoue et al., 2013). Consistent with previous studies, tir-1 (ok1052 gf) animals showed weak adaptation after conditioning ( Fig. 5; Inoue et al., 2013). We made tir-1 (ok1052 gf) animals without AIA interneurons and found that the animals showed normal adaptation and also prolonged retention of the adaptation (Fig. 5). These phenotypes cannot be distinguished from those of AIAanimals, suggesting that AIA interneurons regulate forgetting downstream of the TIR-1/ JNK-1 pathway.

Discussion
Forgetting is important for animals to manage information to properly respond to changing environments. Yet, the neuronal mechanisms for forgetting are not fully understood. In this study, we discovered that a pair of interneurons, AIA interneurons, is required to regulate behavioral forgetting of olfactory adaptation.

Interneurons accelerate forgetting of olfactory adaptation
We found that AIA interneurons are required to accelerate forgetting of olfactory adaptation (Figs. 2, 6). Without functional AIAs, conditioned animals were unable to regain chemoattraction toward diacetyl after cultivation with food for 4 h. However, after cultivation with food for 24 h, chemoattraction was recovered in AIAanimals, suggesting that, even in the absence of the functional AIA interneurons, animals can slowly forget. Therefore, AIAs are important to accelerate forgetting of olfactory adaptation.
AIA interneurons are indispensable for chemotactic behavior to diacetyl only after recovery and, thereby, for behavioral forgetting AIA interneurons are part of the olfactory sensory circuit (Larsch et al., 2015;Dobosiewicz et al., 2019). We observed a minor defect in chemoattraction of naive AIAanimals to diacetyl (Fig. 1D,E), indicating that the neuronal circuit for chemotaxis can function in naive animals in the absence of AIA interneurons. However, after recovery for 4 h, AIAanimals still showed a defect in chemotaxis to diacetyl (Fig. 1D), observed as a defect in behavioral forgetting, although the sensory memory trace declined normally in AWA sensory neurons (Fig. 3). These observations suggest that, in AIAanimals, the sensory response of AWAs cannot induce attractive chemotaxis to diacetyl after conditioning. These results raise two possibilities. One is that although redundant neuronal circuits can regulate chemotaxis to diacetyl in naive animals, after conditioning, the circuit that does not include AIAs becomes nonfunctional so that the AIAs become indispensable for the chemotaxis (Fig. 7A). Another one is that, only after conditioning does the neuronal circuit for chemotaxis to diacetyl require the activity of AIAs, which is distinct from the naive circuit (Fig. 7B). In these hypotheses, the functional neuronal circuit that does not include AIAs may recover slowly so that chemotaxis to diacetyl recovers after conditioning for 24 h. Our Ca 21 imaging analyses of AIA might support the model for the redundant neuronal circuits in naive animals (Fig. 7A) because AIA responses are similar to those after the recovery (Fig. 4). These kinds of circuit plasticity, which are based on internal states, are important for behavioral plasticity in higher organisms (Herry et al., 2008;Ramaswami, 2014;Joshua Kim et al., 2017;Kuchibhotla et al., 2017). Furthermore, we suspect that such circuit plasticity involving AIA interneurons Figure 5. Genetic epistasis between TIR-1 and AIA interneurons. Chemotaxis to diacetyl was analyzed in tir-1(gf), AIA -, and tir-1(gf) animals with no functional AIA in naive, adaptation, and recovery phases. Boxes, first to third quartiles (Q1 25th to Q3 75th percentile) of each dataset; black line in the boxes, medians; black dots, mean; whiskers, minimum and maximum, excluding outliers (beyond 1.5-fold interquartile range from Q3 and Q1); x, outliers (n = 8, two-way ANOVA, F strain(5,126) = 38.4268, p = 8.41e À24 , h 2 = 0.6039) g . Post hoc t test with Bonferroni's correction.
Research Article: New Research might also be used by other olfactory adaptation mechanisms because we also observed that AIAanimals displayed a defective forgetting phenotype toward AWC-sensed isoamyl alcohol without causing a severe chemotactic defect (Fig. 6). To clarify the precise role of AIA interneurons in both circuit and behavior plasticity, additional experiments including optogenetic inactivation or activation of the olfactory circuits in naive and conditioned animals are required to examine these hypotheses.
Our genetic epistasis experiment indicates that AIAs function downstream of the TIR-1 pathway in the regulation of forgetting (Fig. 5). TIR-1 is required to accelerate the forgetting of olfactory adaptation of diacetyl; therefore, the adaptation defect to diacetyl in the tir-1(gf) mutant might be caused by forced chemotactic recovery from adaptation during conditioning (Inoue et al., 2013). If this is the case, the suppression of the adaptation defect by AIAis consistent with the role of AIAs in chemotaxis during the recovery phase.
Our study shows that AIA interneurons in C. elegans are required to regulate behavioral forgetting of olfactory adaptations. This indicates that intact neural circuits are important for simple forgetting regardless of the simplicity of the neural system. Studies that reveal learning, memory formation, and forgetting pathways in invertebrates might be conserved across species (Stein and Murphy, 2014;Vorster and Born, 2015;Lipina et al., 2016;Costa et al., 2020;Rahmani and Chew, 2021); therefore, we believe that studies in invertebrates are important to elucidate the mechanisms of forgetting in higher organisms. Figure 6. Chemotaxis to AWC-sensed isoamyl alcohol in AIAanimals. Chemotaxis to isoamyl alcohol was analyzed in two AIAtransgenic animals, AIAp::mec-4(d) and AIAp::unc-103 (gf), in naive, adapted, and 4 h recovery phases. Boxes, First to third quartiles (Q1 25th to Q3 75th percentile) of each dataset; black line in the boxes, medians; black dots, mean; whiskers, minimum and maximum, excluding outliers (beyond 1.5-fold interquartile range from Q3 and Q1); x, outliers (n = 8, two-way ANOVA, F strain(3,82) = 11.4853, p = 2.31e À6 , h 2 = 0.2959) h . Post hoc t test with Bonferroni's correction. Figure 7. AIAs may regulate forgetting of diacetyl olfactory adaptation via circuit plasticity. A, B, Two hypothetical neural circuits of AIA-dependent behavioral plasticity in the forgetting of olfactory adaptation to diacetyl. In the two models, naive chemotactic behavior might be regulated along with (A) or independent from (B) an AIA-dependent functional neural circuit. In both models, after conditioning, the AIA-dependent functional neural circuit is required to regulate the corresponding behavioral output after the animal recovers from adaptation.