Article Figures & Data
Figures
- Figure 1.
Sniff change and concentration change have very similar effects on FR responses of MTCs. A, Stimulation paradigm during whole-cell recordings. PID traces show response of photoionization detector (magnitude proportional to odor concentration), while nasal flow traces show example sniffing behavior recorded using external flow sensor for the three types of trial. See Figs. 1-1 and 1-2 for details about sniff parameters. Black bar and gray box shows where odor is applied. B, Example odor responses recorded in each stimulus condition. Vm traces show example responses for cell a, while PSTHs below show averaged FR responses in 250 ms time bins for five trials in each case. Bottom-most PSTHs are calculated for a different example, cell b. Error bars show standard deviation (SD). All are aligned to first inhalation onset after odor onset. C, Scatter plot comparing mean FR response change for concentration change and sniff frequency change (normalized by the SD of baseline FR changes in the 2 s before odor stimulus for each cell-odor pair) across first second of odor stimulus. n = 20 cell-odor pairs. D, Heatmap of average FR responses for all cell odor pairs in the low concentration, slow sniff frequency condition, ordered by mean FR response. E, Heatmap of FR response differences (difference between PSTHs) normalized by the SD of baseline FR differences in the 2 s before odor stimulus for each cell-odor pair. Concentration increase = high concentration, slow sniffing, minus low concentration, slow sniffing. Faster sniffing = low concentration, fast sniffing, minus low concentration, slow sniffing. Asterisks indicate cell a and cell b examples. F, Top: R 2 values for correlations across all odor time bins as shown in E, between FR changes due to concentration change and those due to sniff frequency change. Histogram shows R 2 values for shuffle controls, “actual” shows R 2 value for real data. Red dotted line indicates value for correlation between FR changes due to concentration increase for two separate sets of high concentration trials. Bottom: as for above, but histogram showing p-values for the correlations (–log10). See Fig. 1-3 for analysis of membrane potential responses.
- Figure 2.
Faster inhalation causes temporal shifts similar to those caused by concentration increase. A, From top to bottom: example Vm traces, spike rasters, and mean spike counts for early excitatory responses for slow inhalation (black) and fast inhalation (pink), for two different cell-odor pairs. The left example is from a putative mitral cell (pMC) and the right example is from a putative tufted cell (pTC). Rasters are ordered (top to bottom) from slowest to fastest inhalation. Black bar and dotted line indicate odor onset aligned to the first inhalation onset. B, Comparison of response onset latencies for excitatory responses evoked by fast and slow sniffs for pMCs and pTCs. See also Fig. 2-1. C, Example Vm traces (above, for one cell) and mean spike counts (below, for two different cells) for early excitatory responses. Black shows response at low concentration evoked by slow inhalation, pink shows response at low concentration evoked by fast inhalation, and green shows response for high concentration evoked by slow inhalation. D, Left: heatmap to show spike counts of all 20 cell-odor pairs in response to low concentration odor stimulus and slow inhalation, for the first 250 ms of stimulation. Cell-odor pairs are sorted by the mean spike count during odor. Middle: heatmap to show difference in spike counts between high concentration and low concentration (evoked by slow inhalation). Left: heatmap to show difference in spike counts between fast inhalation and slow inhalation (low concentration stimulus). E, Top: R 2 values for correlations across all odor time bins as shown in D, between spike count changes due to concentration increase and due to faster inhalation. Histogram shows values for shuffle controls (see Methods), black bars show value for actual data. Red dotted line indicates value for correlation between spike count changes due to concentration increase for two separate sets of high-concentration trials. Bottom: as for above, but histogram showing p-values for the correlations (–log10). F, Histogram to show excitatory response onset latency changes due to concentration increase. Error bar in green shows mean and SD of this data, and in pink shows the distribution due to sniff changes (from dataset in panel B) for comparison. G, Euclidean distance between population spike count response vectors for high- versus low-concentration stimuli (where data for both came from slow inhalation trials; “slow sniff,” solid cyan), for high concentration and time-shifted low concentration (“slow sniff adv.,” where excitatory latency changes due to concentration change were undone via time shifting of the data; dotted cyan), and for high concentration and low concentration where low concentration data came from fast inhalation trials (“fast sniff,” solid purple).
- Figure 3.
Mice rapidly learn to discriminate concentrations on fast timescales. A, Diagram of head-fixed behavior setup. B, Average PID traces for concentration go/no-go stimuli. Shaded area shows SD. See Fig. 3-1A for odor outlet flow traces. C, Concentration go/no-go task sequence. See Fig. 3-1B for training protocol. D, Left: average learning curve for eight mice. Percentage correct is calculated as a moving average over 5 CS+ and 5 CS– trials. Shaded area indicates SD. Mice are initially trained on two concentrations of an odor mixture, and subsequently tested on the same two concentrations of vanillin. Right: distribution of learning times to criterion (four successive learning curve points at or above 80% correct), for the odor mixture and vanillin. E, Left: distribution of reaction times (RTs) calculated from licking behavior for the odor mixture and vanillin. Right: as for left, but for RTs calculated from sniffing behavior (see Methods). F, Left: example sniff traces for the 1st, 3rd, and 8th presentation of the CS+ and CS– concentrations for the initial concentration discrimination learning session. Note that in this session, the CS+ concentration is first presented 10 times to ensure retention of the lick pattern learned the day before, and then the CS– is interleaved in a pseudorandom order. Right: plot to show median sniff frequency across 8 mice (regardless of concentration-reward contingency) for presentations 1–10 of the CS+ and CS– concentration in the first 2 concentration discrimination sessions. Boxes show upper and lower quartiles.
- Figure 4.
Variance in inhalation has no overt impact on concentration discrimination performance. A, Diagram to show average PID traces for the five different concentrations and contingency schemes. Shaded area shows SD. To the left the contingencies are shown for “high-go” and “low-go” trained mice, with black crosses indicating CS– stimuli. B, Average go rate (percentage of trials with a go response) across mice for all five concentrations. C, Mean lick counts averaged across mice for the five different concentrations (darkest = strongest) for both “high go” and “low go” training contingencies. Black bar indicates odor stimulus, and blue bar indicates response period. D, Plot to show inhalation duration for first inhalation of the odor stimulus across trials, for the first session of one example mouse performing the five-concentration go/no-go task (“concentration GNG”), and for a passively exposed mouse (“passive”). Error bars show SD for each 10-trial block. Example representative nasal flow waveforms for single sniffs are shown to the left. E, Mean SD for the first inhalation duration (ms) during the odor stimulus, for seven mice performing five-concentration go/no-go in their first session, and for passively exposed mice (n = 23). SD is calculated for each 10-trial block of a session for each mouse. Error bars show standard error. F, Example histogram of inhalation durations of the first sniff during an odor stimulus across trials for one mouse. Data for each mouse is partitioned into fast inhalations (<30th percentile, red), slow inhalations (>70th percentile, cyan), and other (gray). Example representative nasal flow waveforms for a single sniff of each subset are shown. G, Go rate as a function of concentration when splitting trials according to duration of first inhalation as in F. Dotted line shows mean go rate for sniffs with inhalation between 30th and 70th percentile. H, Top: average difference in lick-histograms between CS+ and CS– (highest versus lowest concentration) averaged across all seven mice for slow sniff trials (cyan data) and fast sniff trials (red data) partitioned as in F. Shaded area indicates SD. Dashed line indicates odor stimulus onset aligned to the first inhalation. Bottom plot shows difference in reaction times as measured by licking for fast and slow sniff trials for all seven mice. See also Fig. 3-1D. J, Example sniff traces for one animal for a puff trial (a trial in which an air puff to the flank was used to evoke fast sniffing) and an adjacent control trial of the same concentration. Blue ticks indicate licks. K, Mean go rate as a function of concentration across mice for puff trials (orange) versus control trials (black). L, As for H, but now comparing lick distributions and reaction times between puff trials and control trials. See also Fig. 3-1E.
- Figure 5.
Inhalation duration transforms mean baseline MTC activity in a large proportion of cells. A1–C1 refer to one example cell, while A2–C2 refer to a different example cell. A1–2, Example nasal flow traces and Vm traces in absence of odor. Sniffs are color coded according to inhalation duration (blue = slow, and red = fast). Black ticks indicate inhalation onset. B1–2, Left: average spike count histograms triggered by inhalations of different durations (denoted on each histogram). Right: mean spike count per sniff as a function of inhalation duration. Error bars = standard error (SE). C1–2, Left: inhalation-triggered mean Vm waveforms for sniff cycles of different inhalation duration. Right: mean Vm and timing of Vm peak for membrane potential waveforms averaged across all sniffs as a function of inhalation duration. Error bars = SE. D, Top: heatmap of R values for correlations between inhalation duration and 3 different activity parameters (spike count, mean membrane potential, and timing of peak membrane potential, rows 1–3, respectively), for 45 MTCs. Cells are sorted left to right from largest number of significant correlations to lowest number. Black squares show where the correlation was insignificant (p > 0.01, linear regression). Two lowest heatmaps show the same data but for two example shuffle controls, where inhalation durations were shuffled with respect to the physiology, and the data reanalyzed. This gives an indication of false-positive rates in this analysis. Bottom: histogram to show proportion of cells with 0–3 significant correlations between the different activity parameters and inhalation duration. Gray shows proportion for shuffle controls. E, Scatter plot between inhalation duration predicted by a simple linear model using peak spike rates of 25 cells (see Methods) and the actual (true) inhalation duration for all 7 sniff cycles tested in each category. See Fig. 5-1 for the impact of cell type on responses to inhalation change, Fig. 5-2 for further analysis regarding detecting inhalation change, and Fig. 5-3 for a hypothetical relative timing code using this information to infer environmental concentration change.
Tables
Location Data structure Statistical test 95% confidence intervals a Paired response onset latencies (fast vs slow sniffs), n = 13 cells Paired t test –25 to –7 ms b Paired response onset latencies (fast vs slow sniffs), n = 5 pMCs Paired t test –39 to –22 ms c Paired response onset latencies (fast vs slow sniffs), n = 8 pTCs Paired t test –16 to 1 ms d Normal distributions of equal variance Unpaired Student’s t test, 2-tailed 11 < 23 < 34 ms e Paired response onset latencies (high vs low concentration, n = 4) Paired t test 2.3 to 33 ms f SD in inhalation duration for passive (n = 23) and concentration go/no go mice (n = 7), calculated for each block (1 block = 10 trials) Two-way ANOVA on SD in inhalation duration [factors: block #, behavior (passive vs concentration go/no go)] Multiple comparison test: –10 < 5 < –2 ms g Go rate for fast and slow sniff trials for each concentration (5), for n = 3 mice trained on low Go contingency Three-way ANOVA on go rates [factors: mouse, concentration, sniffing (fast vs slow)] Multiple comparison test: –20 < –10 < –1% h Go rate for fast and slow sniff trials for each concentration (5), for n = 4 mice trained on high Go contingency Three-way ANOVA on go rates (factors: mouse, concentration, sniffing (fast vs slow)) Multiple comparison test: –22 < –15 < –7% i Difference in go rate between fast and slow sniff trials for each concentration (5), for mice trained on two different contingencies: “low go (n = 3 mice)” and “high go (n = 2 mice)” Two-way ANOVA on differences in go rate (factors: contingency, concentration) Multiple comparison test: –16 < –4 < 7% j Normal distributions of equal variance Paired t-test –15 to 4% k Go rate for probe trials and control trials for each concentration (5), for n = 3 mice trained on low Go contingency Three-way ANOVA on go rates [factors: mouse, concentration, trial type (probe vs control)] Multiple comparison test:
–16 < –7 < 3%l Go rate for probe trials and control trials for each concentration (5), for n = 4 mice trained on high Go contingency Three-way ANOVA on go rates [factors: mouse, concentration, trial type (probe vs control)] multiple comparison test: –19 < –8 < 3% m Difference in go rate between probe and control trials for each concentration (5), for mice trained on two different contingencies: “low go (n = 3 mice)” and “high Go (n = 2 mice)” Two-way ANOVA on differences in go rate (factors: contingency, concentration) Multiple comparison test: 13 < 1 < 16 % n Paired reaction time data (fast vs slow sniffing, n = 5 mice) Paired t test 0.0 to 70 ms o Paired reaction time data (puff vs control, n = 5 mice) Paired t test –61 to 50 ms p Contingency table (significant vs non-significant R 2, actual data vs shuffle controls) Fisher’s exact test 3.4 to 18.3 q Contingency table (significant vs non-significant R 2, actual data vs shuffle controls) Fisher’s exact test 3.5 to 23.3 r Contingency table (significant vs non-significant R 2, actual data vs shuffle controls) Fisher’s exact test 5.9 to 33.3
Extended Figure 1-1
Effect of puff stimulus on sniff behaviour. (A) Example nasal flow traces from one animal during a control trial (no puff stimulus accompanying 2 s odor stimulus) and a trial with a puff stimulus. Shaded area shows odor stimulus. Black ticks indicate inhalation onsets. (B) Plots to show average change in sniff frequency, mean inhalation duration, and first inhalation duration between five control (black) and five puff (orange) trials for all 20 cell-odor pairs. Download Extended Figure 1-1, TIF file.
Extended Figure 1-2
Relationships between different sniffing parameters. (A) Example nasal flow trace during an inter-trial interval (no odor), with sniffs colored according to their inhalation duration (blue to red = long to short duration). Black ticks show time of inhalation onset, orange plot shows peak inhalation slope for each inhalation aligned to the inhalation onset, and green plot shows inhalation duration for each inhalation. (B) Example correlation between inhalation duration and peak inhalation slope for 1988 sniffs in 1 animal (top), and histogram of correlation R values between inhalation duration and peak inhalation slope across 50 animals. Black bars indicate significant correlations. (C) As for B, but for correlation between sniff duration and inhalation duration. As expected from constraints on duty cycle during respiration, we found a biphasic relationship between sniff duration and inhalation duration, with a linear correlation for sniffs < 250 ms duration (magenta) and a plateau for sniffs > 250 ms duration (blue). This resulted in high R values for sniffs < 250 ms duration and low R values for sniffs > 250 ms duration. (D) As for B, but for the correlation between the previous sniff duration and the current inhalation duration. Download Extended Figure 1-2, TIF file.
Extended Figure 1-3
Changes in subthreshold response are more inhibitory for concentration increase than for fast sniffing. (A) Example average subthreshold response traces for low concentration, slow sniffing (black), high concentration, slow sniffing (green) and low concentration, fast sniffing (magenta), for two different cells, cell c (top) and cell d (bottom). Each trace is the average of 5 spike-subtracted trials. (B) Scatter plot to show average change in membrane potential response for the first 1 s of the odor stimulus for concentration increase (high conc.-low conc.) and sniff frequency change (fast sniffing-slow sniffing). (C) Cumulative histograms of membrane potential response change for concentration increase (green) and sniff frequency increase (magenta). P = 0.03, paired t-test. Download Extended Figure 1-3, TIF file.
Extended Figure 2-1
Additional data for sniff-induced temporal shifts in odor response. (A) Heatmaps of mean spike count for 13 cell-odor pairs showing early excitation in response to the odor presented, for both slow inhalation (top) and fast inhalation (middle). White dashed line indicates odor onset aligned to the first inhalation onset. Cell-odor pairs are sorted from short to long response onset latency (during slow inhalation). Bottom heatmap shows the difference between the two above (fast-slow). White solid and dotted line indicates onset latency of each cell-odor pair for slow inhalation. Blue line indicates onset latency for fast inhalation. (B) Histogram of onset latency changes (fast-slow) for all 13 cell-odor pairs. Errorbar shows mean and SD. (C) Scatter plot to show relationship between onset latency for slow inhalation, and the onset change between fast and slow inhalation (ΔOnset). (D) Correlation between response onset latency and peak spike count (analysed within 10 ms time bins) for early excitatory odor responses evoked by a slow sniff. Blue data comes from pTCs and red data comes from pMCs. Boxplots compare the two parameters for pTCs and pMCs. (E) Comparison of response onset latency change (fast-slow sniff) for pMCs and pTCs. (F) Above and below plots are for two different example cells. Left: plot to show first inhalation duration during odor stimulation sorted from shortest to longest for all trials for one cell. Right: heatmap of spike count for the cell during odor stimulation for trials sorted by first inhalation duration as in left plot. White dotted line indicates where odor is on (aligned to first inhalation onset). Download Extended Figure 2-1, TIF file.
Extended Figure 3-1
Additional behavioral data. (A) Mean flow change recorded 1 mm from olfactometer output for high concentration stimulus (red) and low concentration stimulus (blue). Average of 10 trials; shaded area shows standard deviation. Y scale bar is compared to that of nasal flow traces recorded in the same manner to demonstrate the negligible nature of flow change from the olfactometer. (B) Diagram to show training sequence for mice (described in methods). (C) Comparison of changes in the first inhalation duration between physiological and behavioural experiments. Black ('phys.') shows distribution of mean change used for analysis of odor responses for 20 cell-odor pairs recorded in passive mice (as in Figure 2). Purple ('F vs S') shows mean difference between red and cyan sections of the inhalation distribution as in Figure 4F for all mice and concentrations (n = 7 mice x 5 concentrations). Orange 'puff' shows average changes in mean first inhalation duration for puff vs control trials during behaviour (n = 7 mice x 5 concentrations, as in Figure 4J). (D) Average difference in lick-histograms between CS+ and CS- (concentration 4, 2.6%, vs concentration 2, 1.4%) averaged across all 7 mice for slow sniff trials (cyan data) and fast sniff trials (red data) partitioned as in Figure 4F. Dotted line indicates onset of odor stimulus (aligned to the first sniff onset). Right plot shows difference in reaction times as measured by licking for fast and slow sniff trials for all 7 mice. Mean difference in RT (fast-slow) = -92 ± 235 ms, p = 0.34 paired t-test. (E) As for panel D, but now comparing lick distributions and reaction times between puff trials (orange) and control trials (black), as in Figure 4L. Mean difference in RT (puff-control) = -36 ± 125 ms, p = 0.61, paired t-test. Download Extended Figure 3-1, TIF file.
Extended Figure 5-1
Cell type specificity of effect of inhalation as defined by sniff-phase preference. In absence of applied odor, putative mitral cells (pMCs) respond to faster sniffs with increases in inhibition, and putative tufted cells (pTCs) with increases in excitation. (A1) Reconstructed morphology of a tufted cell recorded in awake mouse. 'Bb' refers to brain border, 'EPL' refers to external plexiform layer and 'MCL' refers to mitral cell layer (these morphologies have been previously published in Jordan et al. 2018 for different purposes). (B1) Example nasal flow and Vm trace during a rapid sniff bout (blue to purple represents longer to shorter inhalation duration on flow trace. Spikes have been cropped for display. (C1) Mean membrane potential waveform for different bands of inhalation duration: blue = long inhalation duration, purple = short. (A2)-(C2) as for A1-C1, but for a filled mitral cell recorded in an awake mouse. (D) R values for correlations between inhalation duration and mean Vm as a function of phase preference. Only strong correlations have been included (p<0.05 and R2>0.6). Grey line shows mean R value for all cells within a 2 radian moving window (centred), to give an idea of the phase modulation strength of the data. There was a significant organisation according to phase (p<0.01, bootstrapping, see methods). Boxplots to the right compare all values within the putative MC (red) and putative TC (blue) phase boundaries (mean Vm: pMC: median = 0.93, IQR = 0.84 to 0.95, n = 6; pTC: median = -0.83, IQR = -0.96 to -0.82, n = 10; p = 0.002, Ranksum). These phase boundaries are based upon those used in previous studies (Jordan et al., 2018; Fukunaga et al., 2012). (E) As for panel D, but for mean spike count per sniff. Again, there was a significant organisation according to phase preference (p<0.001; bootstrapping, see methods), and R values were significantly different between pMC and pTC boundaries (spike count: pMC: median = 0.84, IQR = -0.88 to 0.96, n = 22; pTC: median = -0.92, IQR = -0.94 to -0.89, n = 12; p = 0.008, Ranksum). Download Extended Figure 5-1, TIF file.
Extended Figure 5-2
Detection of inhalation duration change and cell 'weights' in linear model (A) Top: diagram to show construction of sniff sequences of different inhalation duration: either three of 95 ms inhalation duration (blue), or two of 95 ms with the final sniff of 55 ms duration (purple). Below shows two example PSTH sequences (from two different cells) averaged from random subsets of 25 sniffs that show the particular inhalation duration. Blue plot shows the PSTH sequence for 4 sniffs of 95 ms, and purple plot shows sequence in which the last inhalation is of 55 ms. Bottom trace shows mean Euclidean distance calculated between population vectors containing all cells constructed from the two sniff sequences as in panel A. Plot shows the average of 5 different subsets of data (made by averaging different sniff subsets for each cell), and shaded area shows standard deviation. Dashed red line indicates time of significant detection of change. (B) As for panel A, this time comparing PSTHs for a smaller inhalation duration change (95 ms, blue to 75 ms, purple). (C) Regression coefficients (weights) for individual MTCs in the linear model used to predict inhalation duration based on peak spike rates (related to Figure 5E; see methods). Download Extended Figure 5-2, TIF file.
Extended Figure 5-3
Diagram of a potential relative timing code for concentration. (A) Highly simplified diagram of the olfactory bulb, depicting only glomeruli and MTCs. Green glomeruli/MTCs show those receiving odor input (odor responsive), while grey glomeruli/MTCs show those with absence of such input (unresponsive population). While odor inputs are sparse, mechanical input (response to the pressure change associated with each sniff) is widespread. (B) Diagram to show how a relative time code may work. Each instance shows one sniff prior to and during odor stimulation. 'Nasal flow' shows these two sniffs, with grey showing slow sniffs (>95 ms inhalation duration) and red showing a fast sniff (e.g. 55 ms inhalation duration). 'Odor concentration' trace shows a step increase in concentration from zero just prior to the second sniff. Grey traces show low concentration and red shows high concentration (2 x low concentration). 'Unresponsive population' shows a hypothetical population average FR for all cells without odor inputs during the odor stimulus. Dotted traces in second and third column for the second sniff cycle show the trace for the first column, for sake of comparison. 'Odor responsive' shows the average population FR for cells receiving direct odor input. Dotted traces in second and third column for the second sniff cycle show the trace for the first column, for sake of comparison. Δt shows the difference in peak population activity for unresponsive and odor-responsive cells during the stimulus. Note that Δt remains stable unless the concentration changes, and not when the sniff alone changes. A change in Δt allows perception of a different concentration. Download Extended Figure 503, TIF file.
In this issue
