Article Figures & Data
Figures
- Figure 1.
Single-unit recordings of MB output neurons. A, After recording neuronal activity, one electrode wire was used to electrocoagulate the tissue at the electrode tip, resulting in a small spot of autofluorescence marked by the white circle (α-L: α-lobe; LH: lateral horn). B, The single electrode wires measured against the reference electrode show a considerable noise amplitude (channels 1 and 2, lower 2 traces). One source of the noise is illustrated by the simultaneously recorded myogram of the muscle M17 (second trace from the top), which is involved in proboscis extension. This noise can be partially diminished by using differential combinations of two electrode wires (upper channel). This differential channels were used for subsequent sorting of single units. C, Schematic wiring of the MB olfactory path. Olfactory input into the MB is provided by PNs of the AL. PNs converge onto MB intrinsic neurons, the KCs. KCs converge onto the MB ENs, providing one of the outputs of the MB. The α lobe (α-l) is one of the two MB lobes where the ENs leave the MB; its ventral side is the target of our recording electrode. Red arrows indicate the flow of olfactory information in the excitatory feedforward network. D, In a first experimental series, we presented 10 different odors (color code), 10 times each, to evaluate the general response characteristics of ENs. We separated 26 units out of 10 bees. A second experimental series was designed to investigate learning-induced changes of EN responses. Therefore, five different odors were presented pseudorandomized 10 times each in a pretest phase. Fifteen minutes later, two odors were randomly chosen to be presented rewarded (CS+) or unrewarded (CS−) during the conditioning (cf. Materials and Methods). In the posttest phase, 3 h after conditioning, all five odors were tested again similar to the pretest phase (cf. Materials and Methods). We separated 44 units out of 17 bees that were used to analyze initial response characteristics. We skipped six units for analysis of preconditioning versus postconditioning comparisons because of failed unit separation during the postconditioning phase (cf. supplemental material, available at www.jneurosci.org). Nineteen units were classified to be “switched” and 16 to be “modulated” after conditioning. Three units showed no response.
- Figure 2.
Unspecific odor responses in EN units. A, Relative numbers of odor-sensitive (pink) and odor-insensitive (gray) EN units. B, Percentage of responding units in relation to the number of odors that evoked responses in unconditioned bees (N = 70 units). Gray marked excitatory and light green inhibitory responses. Most odor-sensitive units were excited by five or more odors. C, Spike trains of an example unit in response to 10 different odors (2,9on = 2-nonanon; 8al = octanal; 6al = hexanal; cin = cineole; eug = eugenol; lim = limonene; lin = linalool; 1,7al = 1-heptanal; 2,8ol = 2-octanol; 1,9ol = 1-nonanol) and paraffin oil (oil), repeated 10 times each. The pink shading indicates the first second of odor presentation (3 s in total). D, Firing rate profiles averaged across all 10 stimulus repetitions per odor. Gray intensities from weak to strong correspond to the odor and oil stimuli in C, top to bottom.
- Figure 3.
Example units and M17 activity before, during and after differential conditioning. A, Dot displays of action potentials of one unit during all experimental phases (Pre = preconditioning, Acq = acquisition, Post = postconditioning) ordered by the different stimuli. Each line illustrates one trial, starting with the first (bottom of each plot). The five odors were presented as the rewarded (CS+), the unrewarded (CS−), or one of the three control odors (Ctr A–C). Odor stimulation time is marked in green, reward presentation in pink. PER behavior of the related bee is documented by recordings of the muscle M17 (Rehder, 1987). Insets show 10 trials of M17 activity during the 3 s of CS+ presentation before (i), during (ii), and after (iii) conditioning recorded simultaneously with the shown example unit. Each line corresponds to one trial starting with the first at bottom. Note that in this example, each of the CS+ trials evoked a PER in the posttest (100%). The M17 recording during the 10 test trials after conditioning (Post) were used to calculate the mean latency (460 ms, blue line) between stimulus onset and the first spike of M17 contraction (for mean latency of all bees, cf. supplemental material, available at www.jneurosci.org). The odor delay of 37 ms is always subtracted (cf. Materials and Methods). B, C, PSTHs (bin size = 50 ms) of four single units (rows) recorded in four different bees (bees 67, 87, 83, and 73) before (black) and after (red) conditioning. Stimuli (columns) in A–C are equivalent. B, Two examples of units that switched their odor responses. Unit 1 of bee 67 (same unit as in A) established sensitivity to the CS+ (“CS+ recruited”). Note that this unit rapidly decreased its spontaneous activity during the acquisition (cf. supplemental Fig. 1C, available at www.jneurosci.org as supplemental material). Unit 1 of bee 87 lost its initial sensitivity (i.e., “dropped” responses) most clearly to Ctr B and C. C, Two examples of units that modulated their odor response strength. Unit 1 of bee 83 “increased” its response strength to the CS+, and unit 1 of bee 73 “decreased” its response strength most clearly to the CS− representation. For spike-sorting quality and ISI distribution of the shown example units, compare supplemental Figures 3–6 (available at www.jneurosci.org as supplemental material). The percentage of the bees' PER to presentation of the CS+ in the posttest is indicated in the CS+ column. Bee 67 showed a PER during all of the 10 test trials after conditioning (cf. inset iii in A), bee 87 to 40%, bee 83 to 80%, and bee 73 to 60%.
- Figure 4.
Switching units encode odor–reward association. A, Single-unit responses to the control odors (A–C) and the conditioned odors (CS+ and CS−) in the preacquisition phase (PRE) and the postacquisition phase (POST). Each row illustrates one unit, and each box indicates its response (black: excitatory, light green: inhibitory, white: no response) to a single odor as specified at the top of each column. The CHANGE matrix illustrates the response difference between before and after stimulation. Red indicates a newly recruited excitatory response for a particular odor, and blue indicates the dropout of an excitatory response for a particular odor. B, Bar chart illustrates the percentage of units that show an excitatory response before (gray) and after (black) learning. The three control stimuli (A–C) were combined (ctr). The number of units responding to the CS+ was significantly increased (p < 0.001, binomial test) during conditioning from 7 to 16 of total 19, while the reduction of the number of units responding to CS− and control odors was not significant (p > 0.2). After conditioning, the number of units responding to the CS+ was significantly higher than the number of units responding to the CS− (p < 0.001) or the control odors (p < 0.001). C, Fraction of newly recruited (red) and dropped (blue) excitatory responses to the respective stimuli. The CS+ resulted in recruitments only.
- Figure 5.
Increased and decreased odor response strengths of modulated EN units. A, Binary responses (black = excitatory response; white = no response) of unit #20–35 (rows) to five odors (columns). Most units nonselectively responded to all five tested odors and the odor response spectra were the same before (PRE) and after (POST) conditioning. No inhibitory responses were detected. The CHANGE matrix illustrates the difference in response rate between preacquisition and postacquisition tests (blue: decreasing; red: increasing). B, Modulation effects on individual single-unit responses before (gray) and after (black) conditioning. The upper unit (#28) decreased its response to the nonconditioned stimuli (A–C) and the CS− but kept its response strength to the rewarded odor (CS+). Unit 29 kept its response strength to A–C but increased its response to both CS odors. In some units, modulation effects were negligible as illustrated for the lower unit (#23). C, Quantile distribution of rate changes illustrated by a box plot. After conditioning, overall responses to the CS+ were increased, whereas the overall response rates for the CS− and the control odors (ctr) showed almost no change. The median of all rate changes is significantly larger for the CS+ (p < 0.05; Wilcoxon signed rank test), but not for the CS− (p = 0.07) and the control odors (p = 0.35).
- Figure 6.
The associative effect to the CS+ occurs only after consolidation. A, Time-resolved single-trial firing rates for 10 initial trials (PRE), 10 acquisition trials (ACQ), and 10 retention trials (POST) during stimulation with the CS+. Color code indicates whether the firing rate is below (blue) or above (red) the threshold (baseline rate + 2.5 × SD). Gray lines indicate stimulus onset and offset. Plastic changes become evident only during the POST test 3 h after conditioning. This is illustrated by the differences D of trial-averaged firing rates (bottom panels) calculated between the acquisition phase and the pretest (black) or the posttest and the acquisition phase (red). During acquisition the units show a response to the unconditioned stimulus presented ∼2 s after stimulus onset. Units 10 and 13 (left panels) became recruited to the CS+ odor (switched units), and unit 35 increased its response rate for the CS+ (modulated unit). Unit 37 is nonresponsive to all odors before and after learning but shows a clear US response. B, Response rate difference between the preconditioning phase and the acquisition phase (black line, SD in gray) and the preconditioning phase and the postconditioning phase (red line, SD in pink), averaged across all switched (left) or modulated (right) units due to the CS+ stimulations. The odor started at time 0 and lasted 3000 ms. Both unit types showed their CS+ sensitivity only during the posttest 3 h after conditioning (red line). Both unit types showed a high average response rate for the sucrose reward presented ∼2 s after odor onset (black line). C, Peak response rates for all 30 trials averaged across the population of switched and modulated neurons shows that the CS+ response maintained an increased level during all 10 retention tests. Green shading indicates acquisition phase.
- Figure 7.
Different tuning characteristics of “switched” and “modulated” units. A, SL for switched (SWITCH) and modulated (MOD) units before (black) and after (gray) conditioning. Switched units increased their SL significantly from 0.23 ± 0.13 before to 0.31 ± 0.16 after conditioning (p < 0.05; Wilcoxon rank sum). Modulated units show a small and nonsignificant increase (p = 0.13) with an average of 0.08 ± 0.11 before and 0.12 ± 0.13 after conditioning. Note that the SL values of modulated units are comparably small. B, SNR of both unit types. Switched units showed a significantly increased SNR after conditioning from average 0.03 ± 0.12 to 0.47 ± 1.0 (p < 0.05; Wilcoxon rank sum). Modulated units increased the SNR from 0.56 ± 1.21 to 1.24 ± 1.71 (p = 0.12).
- Figure 8.
Rapid encoding of reward association in the EN population. A, Time-resolved firing rate responses in the EN population (N = 35) during the first 700 ms of odor stimulation during pretests (top rows) and posttests (bottom row). The top rows display the average and color-coded population responses to the five tested odors (as indicated) during the initial test phase before conditioning (PRE). Switching units 1–19 (average: black curve) generally exhibited smaller rate amplitudes than modulated units 20–35 (average: gray curve). The bottom row compares the dynamic population responses of the three control stimuli (ctrl), and the CS+ and CS− after conditioning (POST). Generally, switched units developed a response preference to CS+ and diminished responses to the control odors, and modulated units established increased rate responses to the CS+. After conditioning, response rates to CS+ were generally higher as indicated by the increased color rate scale. Firing rates were estimated with a triangular kernel (σ = 50 ms). Note that in the POST phase (lower row), stimuli were grouped by their quality (CS+, CS−, ctrl) and not by the odor identity as in the PRE phase (upper rows). B, Time evolution of the Euclidean distance between different population rate vectors averaged across pairs of different stimuli after conditioning (POST). The red curve indicates the average distance (±SD) between the rewarded odor (CS+) and the nonrewarded control odors. It peaks during the initial epoch of the phasic population response and stays at a high level throughout most of the 3 s stimulus period. The distance between the CS+ and the CS− (green) exhibits a similar temporal profile. The black and blue curves represent the pairwise distances among control odors and between a control odor and the CS− after conditioning, respectively. Both curves show the same temporal profile, which stays well below the distance to the CS+ (red). The dashed gray line shows the Euclidean distance averaged across all pairs of odors before conditioning (PRE). C, Estimate of the latencies for reward encoding. Euclidean distances during the first 500 ms of the odor response as in B but computed from causal rate estimates (exponential kernel, τ = 25 ms) for CS+ versus control (red) and control versus control odors (black). The conditioned odor rapidly reaches a significantly distinct ensemble representation within ∼60 ms (solid vertical line). This distance becomes significantly higher than the distance among control odors 141 ms after odor onset (vertical dashed line), and it reaches 90% of its maximum 177 ms after odor onset (vertical dotted line). The horizontal line indicates the average baseline distance as estimated before odor onset; the horizontal dotted line indicates the threshold of 5 SDs above baseline.
Additional Files
Supplemental Material
Files in this Data Supplement:
- supplemental material - Supplemental Material
