Distinct Circuits for the Formation and Retrieval of an Imprinted Olfactory Memory

Abstract

Memories formed early in life are particularly stable and influential, representing privileged experiences that shape enduring behaviors. We show that exposing newly hatched C. elegans to pathogenic bacteria results in persistent aversion to those bacterial odors, whereas adult exposure generates only transient aversive memory. Long-lasting imprinted aversion has a critical period in the first larval stage and is specific to the experienced pathogen. Distinct groups of neurons are required during formation (AIB, RIM) and retrieval (AIY, RIA) of the imprinted memory. RIM synthesizes the neuromodulator tyramine, which is required in the L1 stage for learning. AIY memory retrieval neurons sense tyramine via the SER-2 receptor, which is essential for imprinted, but not for adult-learned, aversion. Odor responses in several neurons, most notably RIA, are altered in imprinted animals. These findings provide insight into neuronal substrates of different forms of memory, and lay a foundation for further understanding of early learning.

Figure 1. Pathogen training at the first larval stage (imprinting) induces long-term aversive memory

(A) Schematic illustration of bacterial choice assay, imprinting protocol, and adult pathogen training protocol. See also Figure S1. (B) Olfactory choice preference index of naïve, PA14-imprinted, and adult-trained animals. Each dot represents a single population assay calculated as shown; each line represents the mean value. (C) Learning index after imprinting on PA14, adult-training, imprinting on non-pathogenic PA*50E12, starving for 12 hours after hatching, and in F1 offspring of PA14-imprinted animals. Boxes represent median and first and third quartiles, and whiskers represent 10^th–90^th percentiles. (D) Learning index of mature (two-day old) adults after exposure to PA14 at different developmental stages. (E) Learning index of animals imprinted either on pathogenic PA14 or on an E. coli BL21 strain expressing the Pseudomonas translational inhibitor ToxA, then tested with choices between PA14/OP50 and ToxA/OP50. n, number of independent assays, 100–200 animals/assay. P values were generated by ANOVA with the Dunnett correction (B,C,D) or by the nonparametric Mann-Whitney test (E) (*** P <0.001, ** P <0.01, * P <0.05, ns not significant).

Figure 2. Imprinted memory formation and retrieval require distinct circuits

(A) Weighted wiring diagram of interneurons implicated in imprinted memory formation and retrieval. Synaptic strength is based on the number of chemical synapses from www.wormweb.org. See also Figure S1. (B) Schematic illustration of neuronal silencing either at the memory formation or memory retrieval stage using cell-specific expression of a histamine-gated chloride channel (HisCl1). (C–D) Neuronal silencing to identify neurons required either during memory formation (C) or during memory retrieval (D). n, number of independent assays, 100–200 animals/assay. P values were generated by ANOVA with the Dunnett correction (** P <0.01, * P <0.05, ns not significant). See also Figure S3.

Figure 3. Imprinting alters behavioural responses to the experienced pathogen

(A) A pirouette is a reversal coupled to a high-angle turn. The bearing angle θ is the animal’s direction of movement with respect to the odor source (here, PA14 lawn) before the pirouette. Each choice assay has two bacterial odor sources, which were examined separately (see Supplemental Experimental Procedures). (B) Normalized pirouette frequency of naïve and imprinted animals at different bearing angles with respect to a PA14 lawn (left) or OP50 lawn (right) in the choice assay. Naïve event frequency was compared to imprinted frequency at each bearing angle; P values were generated by ANOVA with the Sidak correction (* P <0.05). (C) Normalized pirouette frequency of naïve and PA14-imprinted animals navigating between a novel toxic bacterium, ToxA, and OP50. (D) Normalized pirouette frequency of naïve and PA14-imprinted animals navigating between PA14 and OP50 with AIY neurons silenced with HisCl1. Pirouette rates were calculated from 3–5 movies with 40–50 animals each and normalized to average pirouette rates across angles. See also Figure S2.

Figure 4. Tyramine in RIM neurons and the tyramine receptor SER-2 in AIY neurons are required for imprinted aversion

(A) Imprinted aversion and adult learned aversion in mutants for the serotonin biosynthetic enzyme TPH-1, the serotonin receptor MOD-1, the vesicular glutamate transporter EAT-4, the glutamate receptors GLR-1, GLR-3 and NMR-1, the CREB homolog CRH-1 (two alleles), and the orphan G-protein coupled receptor SRA-11. Red bars mark assays with a significant learning deficit. (B) Biosynthetic pathways for tyramine (produced in RIM and RIC neurons) and octopamine (produced in RIC neurons). Cells of the somatic gonad also make tyramine and octopamine. (C) Learning index of tyramine/octopamine mutants and rescued strains. See also Figure S3. (D) Learning index after exogenous tyramine or histamine administration to tdc-1 mutants and RIM::HisCl1 strains. (E) Learning index of tyramine receptor mutants and rescued strains. (F) Cell-specific rescue of ser-2 using intersectional promoters. Cre expression and inversion allows ser-2 expression in subsets of ser-2p2-expressing cells. n, number of independent assays, 100–200 animals/assay. P values were generated by ANOVA with the Dunnett correction. (*** P <0.001, ** P <0.01, * P <0.05, ns not significant).

Figure 5. Calcium responses of AIB, RIM, AIY and RIA neurons after aversive imprinting

(A–C) Average (A) AIB (B) RIM and (C) AIY calcium responses to 60 s or 10 s alternations between OP50- and PA14-conditioned medium in naïve (black) and imprinted (red) animals. See also Figure S5. (D) Altered RIA calcium response after imprinting. Top: illustration of RIA showing nrD and nrV axonal compartments. Calcium dynamics in nrD and nrV are correlated with local input from dorsal and ventral head motor neurons, respectively (Hendricks et al, 2012); odors or bacterial conditioned media acutely synchronize nrD and nrV. Bottom: average calcium responses in RIA nrV compartment of naïve (black) and imprinted (red) animals to alternating 10 s pulses of OP50- and PA14-conditioned medium. (E) Calcium dynamics of individual RIA nrV responses to odor transitions between PA14- to OP50-conditioned medium. Traces were ordered according to the time derivatives of response at odor transitions (t=0). Arrowhead indicates threshold for calcium activation (dF/dt>0.01 %s⁻¹), suppression (dF/dt <−0.01 %s⁻¹), or no response. (F) Average synchronous calcium flux rate of nrD and nrV compartments of RIA neurons. (G–I) AIY and RIA responses in naïve and imprinted animals whose RIM neurons were silenced during the L1 stage. (G) Illustration of the experiment. (H) Average AIY responses. (I) Average synchronous calcium flux rate of RIA. Blue background: PA14-conditioned medium; yellow: OP50-conditioned medium. Calcium traces were normalized on a 0–1 scale, see Experimental Procedures. Average differences for 10 s (AIB, RIM) or 1 s (AIY, RIA) before and after odor transitions were compared in naïve and imprinted animals. Shaded regions around traces are ±SEM. P values were generated by two-way ANOVA with the Bonferroni correction (*** P <0.001, ** P <0.01, ns not significant). See Figure S6 for further analysis of RIA compartmentalized dynamics and synchrony.

Figure 6. Weighted wiring diagram of imprinting neurons AIB, RIM, AIY, RIA, and their synaptic partners

Synaptic strengths are based on the number of chemical synapses from www.wormweb.org. The four imprinting interneurons receive input from many sensory neurons (in grey) that represent different sensory modalities, and send output to motorneurons (in brown) to produce behaviors. Many additional neurons are synaptically connected to this network (Figure S1A) (White et al., 1986). Adult learning requires either AIB or AIY neurons, whereas aversive imprinting requires both AIB and AIY. Both adult learning and aversive imprinting appear to require AWC, AWB, ADF, RIM, and RIA neurons. Among the neurons shown, AWC, AIB, RIM, and RIA are glutamatergic; AIY, SMD, and RMD are cholinergic; ADF is serotonergic, RIM is tyraminergic, and all neurons express one or more neuropeptides. The SRA-11 receptor required for positive imprinting is required in AIY.