Novelty detection in early olfactory processing of the honey bee, Apis mellifera

Abstract

Animals are constantly bombarded with stimuli, which presents a fundamental problem of sorting among pervasive uninformative stimuli and novel, possibly meaningful stimuli. We evaluated novelty detection behaviorally in honey bees as they position their antennae differentially in an air stream carrying familiar or novel odors. We then characterized neuronal responses to familiar and novel odors in the first synaptic integration center in the brain-the antennal lobes. We found that the neurons that exhibited stronger initial responses to the odor that was to be familiarized are the same units that later distinguish familiar and novel odors, independently of chemical identities. These units, including both tentative projection neurons and local neurons, showed a decreased response to the familiar odor but an increased response to the novel odor. Our results suggest that the antennal lobe may represent familiarity or novelty to an odor stimulus in addition to its chemical identity code. Therefore, the mechanisms for novelty detection may be present in early sensory processing, either as a result of local synaptic interaction or via feedback from higher brain centers.

Fig 1. Left antenna responses to novel and familiar odors.

(A) The familiarization protocol using either Hexanol or Octanone as the repeatedly presented unreinforced odor stimulus. Tests were performed with each odor prior to and after familiarization. The headspace is divided into a clockwise circular range of 180 degrees for each antenna. (B) The average distribution densities of left antennae’s angular positions in response to odors before the familiarization process (left column plots) and 10 min after the familiarization process (right column plots). Plots are aligned by odor onset over 12 s, which includes 4 sec segments before, during and after odor delivery (360 frames; N = 24 animals, 13 Hex exposed, 11 Oct exposed). In both columns, the top plot shows responses to the familiar odor and the bottom plot shows responses to the novel odor. The solid line in each plot represents the mean angle of the left antenna across all 24 bees for each respective odor. The shaded region indicates the period during odor stimulus presentation. The colored scale to the right of the plots shows the standardized Kernel density scale across all plots. For statistical comparisons angular position was normalized by calculating the difference in the left antenna’s angular position at all time points from the mean antennal angle (when no odor present) across the three time-sections for each bee and odorant presentation. Statistical analyses: Two-way ANOVA, difference measure~ time period * odor, df = 2, n = 24, Tukey HSD post-hoc comparison, asterisks indicate statistical significance, **** p<0.00001, ns indicates not significant.

Fig 2. Separation of PNs and LNs using a statistical procedure outlined in [1].

(A) Separation of LNs and PNs using hierarchical clustering (see Methods) into LNs (blue) and PNs (orange). (B) Explanation of cell-based features used for clustering (see Methods) based on spike times (top) and response rate (bottom). (C) Spike rasters and, (D) firing rate time courses of separated cells. (E) Cell-based features visualized in reduced space using tSNE (t-distributed Stochastic Neighbor Embedding) [4].

Fig 3. Novelty and familiarity is evident in both PNs and LNs.

Familiarization potentiates the responses of PNs (A1, B1) and LNs (A2, B2) to novel odor (Hex), evident by the raster plots (top of each panel) and peristimulus time histograms (PSTHs) (bottom of each panel). The odor was delivered at time zero, lasting for 4 sec. The mean response of all PNs (orange bar graph, C1; log transformed) shows a decreasing trend to the familiar odor from pre- to post-familiarization (middle two bars, mean ± S.E., p = 0.06, n = 87, paired-t test), and a significant increase to the novel odor (right two bars, mean ± S.E., p = 0.01, n = 87, paired-t test); there is no significant change in response to the solvent control (left two bars, mean ± S.E., p = 0.98, n = 87, paired-t test). In LNs (C2), the familiarity effect is not significant (middle two bars, log transformed; mean ± S.E., p = 0.2, n = 115, paired-t test) but the novelty effect is significant (right two bars, mean ± S.E., p<0.01, n = 115, paired-t test). No significance was detected to solvent control (left two bars, mean ± S.E., p = 0.53, n = 115, paired-t test). The grey stars in (C1) and (C2) marks the median value, which was used to establish rough equivalence to the mean, as would be the case for a normal distribution needed for parametric tests. The pseudo-colored PSTHs of all Hex-familiarized PNs (D1, E1), ranked low (bottom) to high (top) according to firing rate pre-familiarization (D1), show clear variation across units in the novelty effect before and after familiarization. The novelty effect appears to be even more variable among the Hex-familiarized LNs (D2, E2).

Fig 4. Familiar-odor-biased PNs and LNs are novelty and familiarity detectors.

PNs and LNs display initial response biases to the two test odors Hex and Oct prior to familiarization (A1, A2). Some neurons respond more strongly to Oct than to Hex, i.e., data points above the diagonal line; some neurons respond more strongly to Hex than to Oct, i.e., data points below the diagonal line. Additionally, PNs are generally more responsive than LNs, evident from the wider distribution of data points along both X and Y axis. The Oct-biased units are above the diagonal line and the Hex-biased units are below the diagonal line. Then they are familiarized with Oct or Hex. Only the familiar-odor biased neurons showed consistent familiarity and novelty effects. In familiar-odor biased PNs (B1), whether it is Hex-familiarized or Oct-familiarized, most of the neurons decreased their responses to the familiar odor after the familiarization process, as revealed by the negative median values in the box plots (1st and 3rd boxes). In contrast, most of the neurons increased their responses to novel odor after the familiarization process–namely novelty effect, as shown by the positive median values in the box plots (2nd and 4th box). The familiarity and novelty effects are also evident in the familiar-odor biased LNs (B2). In Hex-familiarized familiar-odor biased PNs, 63% (n = 14) of the neurons exhibited the familiarity effect and 69% (n = 15) of the neurons exhibited the novelty effect; this distribution pattern is significantly different from random process (p = 0.02, McNemar test) (purple boxes, C1). In Oct-familiarized familiar-odor biased PNs, also seen were significantly higher percentage of neurons adapted by familiar odor (56%, n = 18) and potentiated by novel odor (69%, n = 22) (blue boxes, C1, p = 0.04, McNemar test). Similar phenomena were also observed among the familiar-odor biased LNs (C2). In novel-odor biased PNs and LNs, the familiar and novel odors did not produce contrasting changes after the familiarization process; none of the comparing pairs showed significant familiarity and novelty effects (D1, D2, McNemar test).

Fig 5. Familiarization enlarges the differences between familiar- and novel-odor representations in AL.

(A) Principal components of firing rate of PN responses to familiar- and novel-odor before and after the familiarization process (pooled across hex-familiarized and oct-familiarized PNs). Black stars mark the time points where the peak response was. After familiarization, the novel odor evoked a response that showed distinct prominence at peak response relative to all other presentation types. (B) Familiarization produced increases in the distance between the novel and familiar PN response (pooled across hex-familiarized and oct-familiarized PNs). Black star occurs at the same time as in A. Stimulus onset was at 2s.