Claims 39
From the paper Allen Institute - Locus coeruleus norepinephrine activity during dynamic foraging - Electrophysiology and fiber photometry · claims file · checked 2026-11-04: ready
- lc-definition-counts
LC, defined in the CCF as the region holding about two-thirds of pontine NE neurons, has a volume of 0.13 mm3 and holds 1507 +/- 55 NE neurons per mouse, with 367 +/- 95 more in sub-coeruleus (34,998 cells, 8 mice).
We defined LC in the CCF as a small dorsal region (0.13 mm 3 in volume) containing approximately two-thirds of all pontine NE neurons (67th density percentile, 1507 ± 55, mean ± SD). Additional NE neurons were found in sub-coeruleus (367 ± 95)
Results, LC-NE projections are spatially organized, paragraph 1; Figure 1a, S1 - fig1-dense-axon-distribution
Dense unilateral labeling shows LC-NE axons at high density in all CNS regions other than the striatum, with predominantly ipsilateral projections to the telencephalon and both ipsilateral and contralateral projections elsewhere.
We found a high density of LC-NE axons in all CNS regions other than the striatum, with varying densities across regions ( Figures 1c–1e and S2 ). Projections to the telencephalon were predominantly ipsilateral, from somata in the same hemisphere.
Results, LC-NE projections are spatially organized, paragraph 2; Figure 1c to 1e, S2How it will be judged
The paper gives no measure or test; a standard choice is used, axon density per CCF region from the expanded SPIM volumes. Reproduced when the striatum has the lowest axon density of the major regions, and in telencephalic regions the density ipsilateral to the labeled LC exceeds the contralateral density.
- axon-length-and-branching
Single LC-NE neurons have long axons (35.14 +/- 13.60 cm, up to 72.57 cm; 130 cells, 4 mice) and branch less per length than 1,227 MouseLight neurons (17.8 against 24.9 branches/cm, ANCOVA p < 0.001).
Many neurons had long axons, 35.14 ± 13.60 cm (mean ± SD; Figure S4a ) with the largest neurons reaching 72.57 cm of total length ... LC-NE neurons branched significantly less (ANCOVA, p < 0.001)
Results, LC-NE projections are spatially organized, paragraph 3; Figure S4a - fig2c-primary-targets
Individual LC-NE neurons have primary targets; neurons projecting to the forebrain rarely project to brainstem or spinal cord, and the reverse.
Consistent with largely nonoverlapping pathways, neurons with forebrain projections tended to project to other regions in the forebrain but not the brainstem or spinal cord ( Figure 2c )
Results, LC-NE projections are spatially organized, paragraph 3; Figure 2cHow it will be judged
From the fraction of each neuron's axon length in each CCF level-2 region (Figure 2c; Methods, Models of single-neuron axonal distributions), the rank correlation across neurons between region pairs (Figure 2c legend). Figure 2c has no spinal cord column (the legend lists OLF to MY). The paper gives no test or threshold; a standard choice is used. Reproduced when forebrain regions (OLF, CTX, HPF, CTXsp, CNU) correlate positively with each other and their correlations with P and MY are at or below zero.
- fig2ef-dorsal-ventral-projections
Dorsal LC neurons project mainly to the forebrain and ventral LC neurons to the hindbrain and spinal cord, in single-neuron reconstructions and in retrograde tracing.
somata of NE neurons with projections to cortex were preferentially in dorsal LC and NE neurons with projections to the spinal cord were in ventral LC
Results, LC-NE projections are spatially organized, paragraph 6; Figure 2e, 2f, S7How it will be judged
The paper gives no test; comparison of medians is a standard choice. Reproduced when the median dorsal-ventral CCF coordinate of somata retrogradely labeled from frontal cortex is dorsal to that of somata labeled from spinal cord (Figure 2f, S7), and the median of reconstructed neurons with primary forebrain projections (OLF, CTX, HPF) is dorsal to that of neurons with primary hindbrain projections (P, MY, spinal cord) (Figure 2e).
- fig3a-gene-expression-continuum
snRNAseq of 4,845 LC-NE neurons (40 mice) is best explained by a single cluster, with gene expression varying smoothly along a pseudospace axis.
A single cluster best explained the data ( Figure S8 ), although gene expression varied smoothly across the LC-NE neuron population
Results, LC-NE neurons have graded gene expression in space, paragraph 1; Figure 3a, S8eHow it will be judged
MMIDAS and Gaussian mixture clustering as in the Methods select one cluster.
- fig3c-pseudospace-dorsal-ventral
In MERFISH data (2,221 LC-NE neurons), imputed pseudospace aligns with the dorsal-ventral axis of LC.
Gene expression pseudospace aligned with the dorsal-ventral axis of the LC
Results, LC-NE neurons have graded gene expression in space, paragraph 2; Figure 3c, S10How it will be judged
Reproduced when the CCA primary spatial axis of gene expression is closer to the dorsal-ventral axis than to either other axis.
- fig3d-retroseq-targets
LC-NE neurons projecting to cortex, cerebellum and spinal cord occupy different parts of gene expression space, cortex-projecting with dorsal and spinal cord-projecting with ventral neurons.
Projection target was imputed for each cell in snRNAseq based on the similarity mapping from retroseq data (Kruskal-Wallis, p < 0.001)
Results, LC-NE neurons have graded gene expression in space, paragraph 3; Figure 3d legendHow it will be judged
Pseudospace scores of snRNAseq cells grouped by imputed projection target differ (Kruskal-Wallis p < 0.001), ordered cortex, cerebellum, spinal cord.
- fig3e-axes-separated
The primary spatial axes of axonal projections (retrograde tracing) and gene expression (MERFISH) both run dorsal-ventral, separated by 22.20 degrees.
Both axes aligned with the dorsal-ventral direction, separated by 22.20◦ ( Figure 3e )
Results, LC-NE neurons have graded gene expression in space, paragraph 4; Figure 3e - figs14k-patchseq-by-target
In slice recordings (96 cells, 28 mice), spinal cord-projecting LC-NE neurons have action potential waveforms different from cortex- and cerebellum-projecting neurons, and cortex-projecting neurons have longer membrane time constants.
Neurons projecting to spinal cord had significantly different action potential waveforms than those projecting to isocortex or cerebellum. Neurons projecting to isocortex had significantly longer membrane time constants than those projecting to cerebellum or spinal cord
Results, Spatial variation of LC-NE activity, paragraph 5; Figure S14k - figs12-trials-per-session
Mice performed a few hundred trials per session (mean 342, range 94 to 558).
Each day mice performed a few hundred trials (mean, 342, range, 94–558 trials; Figure S12 )
Results, Spatial variation of LC-NE activity, paragraph 2; Figure S12aHow it will be judged
The paper does not say which sessions are counted. Readings (a) the sessions of the 241 recorded neurons after the code's behavior criteria (metrics/beh_only.json, at least 90 trials, 106 sessions); (b) all behavior sessions in the dataset, with and without recordings. A trial is a go cue (no-go cues are part of the inter-trial interval, Methods, Data analysis). Reproduced when, under one reading, the mean is within tolerance and the range lies within 94 to 558 widened by 10% of each bound.
- iti-mean
Inter-trial intervals averaged 4.83 s (range 1.1 to 20 s).
Inter-trial intervals were 4.83 s on average (range, 1.1–20 s, including a no-lick window
Methods, Behavioral task, paragraph 6How it will be judged
The inter-trial interval as the Methods define it, the time from the end of the 3.0 or 3.1 s trial to the next cue onset (a full cue-to-cue interval cannot be as short as the reported 1.1 s), over all task sessions. Open, run as readings, whether no-go cues count as cue onsets (Methods, Data analysis, treats them as part of the inter-trial interval). Reproduced when, under one reading, the mean is within tolerance and the range lies within 1.1 to 20 s widened by 10% of each bound.
- fig4c-choice-history
Mice used reward and choice history to guide choices; a logistic regression of choice on outcome history has R2 = 0.60 and predicts 0.88 of choices, with coefficients decaying over past trials (exponential fit coefficient 3.95).
Logistic regression coefficients for choice as a function of outcome history for all mice (error bars are smaller than points). R 2 = 0.60, successful choice prediction, 0.88 ... exponential fit (coefficient of 3.95)
Results, Spatial variation of LC-NE activity, paragraph 3; Figure 4c legendHow it will be judged
Methods, Analyses and models of behavior; the model equation is lost from the text. Following the code (F_behavior_w_FP.ipynb), choice is regressed on rewarded and unrewarded choice history and choice history over past lags (5 in prepare_glm_session; the README says up to 15). Open details, run as readings, are the number of lags (5, 10, 15), the R2 measure (McFadden pseudo-R2, or squared correlation of predicted and actual choices) and whether the fit pools all mice or averages per-mouse fits. Reproduced when the rewarded-history coefficients are positive and decay over lags, and, under one reading, R2 and the fraction predicted are within tolerance.
- fig5a-engagement-reward-history
Whether mice respond to a go cue depends on reward history.
A logistic regression predicting whether mice responded to a go cue with a lick revealed a dependence on reward history ( Figure 5a )
Results, Spatial variation of LC-NE activity, paragraph 7; Figure 5aHow it will be judged
Methods, Analyses and models of behavior, engagement model (response or ignore regressed on past rewards regardless of direction); the code (F_hit_miss.ipynb) fits it per session and per mouse over several lags. Reproduced when the coefficient for the previous trial's reward has a 95% CI excluding zero across mice, as plotted in Figure 5a.
- fig5e-switch-history
Mice were likely to repeat choices when they had recently been rewarded for that lick direction, as shown by a logistic regression of switch against stay on no-reward and choice history.
Mice were likely to repeat choices when they received recent rewards from those lick directions ( Figure 5e )
Results, Spatial variation of LC-NE activity, paragraph 8; Figure 5eHow it will be judged
Methods, Analyses and models of behavior, switch model, whose equation is lost from the text. Per mouse, logistic regression of switch s(t) on no-reward history N_ipsi (1 for an unrewarded choice ipsilateral to the choice at t - 1, -1 for an unrewarded contralateral choice, 0 for a rewarded choice) and choice history (Figure 5e legend). The number of lags is open; readings 5, 10 and 15, as in claim fig4c-choice-history. The paper gives no test (error bars are smaller than the points); the 95% CI across mice is a standard choice. Reproduced when, under one reading, the lag-1 no-reward coefficient is positive (an unrewarded previous choice raises the probability of switching) with a CI excluding zero.
- fig6c-choice-change-rpe
Changes of choice on consecutive trials correlate with the model's RPE (rank correlation rho = 0.84).
Relationship between how choices changed on consecutive trials and RPE (rank correlation, ρ = 0.84, p = 0.0)
Figure 6c legend; Results, Activity of LC-NE projections, paragraph 1How it will be judged
RPE from the hierarchical Q-learning model (aind_stan_fit_sim, stan_qLearning_5params) as the dataset holds it, or refitted with that model. Change of choice likelihood as the Methods define it (Analyses and models of behavior; equation lost from the text), centred per session. The code (F_behavior_w_FP.ipynb, Figure6c_version2) has no rank correlation; it bins RPE per session into 8 or 9 equal-width bins and averages over sessions. Readings (a) Spearman over the session-averaged bin means; (b) Spearman over all trials pooled. Reproduced when rho is positive and within tolerance under one reading. Reading (b) is unlikely to give 0.84, and a match only under (a) is recorded as resting on the binning.
- fig4g-go-cue-excitation
Nearly all LC-NE neurons (240 of 241, 19 mice) increased their spike rate in the 300 ms after the go cue.
Nearly all LC-NE neurons showed an increase in spike rate in a 300 ms window following the go cue (paired t -tests, 240/241 cells, 19 mice
Results, Spatial variation of LC-NE activity, paragraph 6; Figure 4gHow it will be judged
Population as in the code's metrics/beh_all.json and beh_only.json (241 units, 19 mice). Per neuron, a paired t-test over trials of the rate 0.01 to 0.3 s after the go cue against the rate 2 s to 0.01 s before it (Figure 4g legend; basic_ephys_generation.py). Readings (a) two-sided p < 0.05, as the code counts; (b) p < 0.05 with response above baseline, as "increase" states. Reproduced when the count is within tolerance under one reading.
- fig4h-tonic-phasic-anticorrelation
Neurons whose go-cue rate is more anticorrelated with their baseline rate have larger go-cue responses relative to baseline (r = -0.38).
Change in go-cue spike rate relative to baseline, normalized by baseline vs. the correlation between go-cue and baseline spike rates ( r = –0.38, p = 8.3 × 10 −11 )
Figure 4h legendHow it will be judged
Following the code (F_basic_ephys.ipynb), x is the trial-by-trial Pearson correlation of the rate 0.5 s before the go cue with the rate 0.01 to 0.3 s after it (Methods, Comparison of phasic and tonic responses), y is (response rate minus 2 s baseline rate) over baseline rate, and r is Pearson across neurons. The population is open; readings (a) the 241 neurons of Figure 4g; (b) the code's basic_ephys_all.json population (274 units, 21 mice), which includes neurons outside the task criteria. Reproduced when r is negative with p < 0.05 and within tolerance under one reading.
- figs15f-tonic-modulation-vs-engagement
Across neurons, the correlation between baseline and go-cue spike rates (tonic modulation strength) correlates with the increase of spike rate on lick-response trials over trials without a response, and with the respond-versus-ignore t-statistics of the baseline rate (r = -0.59 for each).
Top, increase of spike rate in trials with lick responses vs. those without against tonic modulation strength from e ( r = –0.59, p = 4.63 × 10 −21 )
Figure S15f legendHow it will be judged
Pearson across the 241 neurons (A&O notebook cells 62 to 63) of the 0.5 s baseline against 0.3 s response correlation (claim fig4h-tonic-phasic-anticorrelation) with, as the legend gives the same r for both, (top) the response-window rate difference between respond and ignore trials and (bottom) the respond-versus-ignore t-statistic of claim fig5b-baseline-predicts-ignore. Reproduced when either is negative and within tolerance; the other is reported.
- fig5b-baseline-predicts-ignore
LC-NE neurons had higher spike rates before ignored go cues than before responded ones; baseline rates in the 2 s before the go cue predicted the response in 46.5% of neurons.
LC-NE neurons had higher spike rates preceding ignored trials relative to those with a lick response ... (46.5% of neurons had p -values < 0.05)
Results, Spatial variation of LC-NE activity, paragraph 7; Figure 5bHow it will be judged
Methods, Linear regression models of spike rates; per neuron, OLS of the rate 2 s to 0.01 s before each go cue on response (any lick in the response window) against ignore, with the drift regressor (per-trial waveform amplitude). The code (response_tstats_generation.py) adds switch-versus-stay as a regressor and drops sessions with fewer than 2 ignored trials. Readings, crossed (a) regressors response and drift (Methods) or response, drift and switch (code); (b) denominator all 241 neurons or the neurons with a defined coefficient (216 in the code). Reproduced when the fraction with p < 0.05 is within tolerance under one reading and the mean baseline rate before ignored trials exceeds that before responded trials across neurons (paired t-test across neurons, p < 0.05, a standard choice; the paper gives no test for the direction).
- fig5c-ignore-ventral
Neurons with higher activity before ignored trials lie more ventral in LC (103 cells, 15 mice; linear model R2 = 0.152).
Neurons with higher activity preceding ignore vs. respond trials (blue) were located more in ventral than dorsal LC (103 cells, 15 mice; t -statistics shown in d ; kNN cross-validated R 2 = 5.57 × 10 −2 , p = 8.4 × 10 −3 ; linear model R 2 = 0.152, p = 1.4 × 10 −3 )
Results, Spatial variation of LC-NE activity, paragraph 7; Figure 5c legendHow it will be judged
The respond-versus-ignore t-statistics of claim fig5b-baseline-predicts-ignore for neurons with CCF locations, medial-lateral mirrored by absolute value (Methods, Analyzing relationships with space). Linear trend, OLS of t on x, y, z with a one-sided permutation p over 5,000 permutations (Methods; the code uses 10,000); kNN, distance-weighted, k = 20, 5-fold cross-validated R2 with a permutation p. The code also applies FDR across six features; the paper names no correction, so uncorrected p is the main reading. Reproduced when the linear p < 0.05 and the fitted gradient puts higher pre-ignore activity (negative t, in the code's sign) ventrally, with R2 within tolerance; the kNN result is reported.
- fig5d-ignore-on-anatomy-axes
Respond-versus-ignore t-statistics vary along the primary spatial axes of axonal projections and gene expression (r = 0.23 for each).
Respond vs. ignore t -statistics projected onto the primary spatial axes from axonal projections and gene expression ( Figure 3e ; left, r = 0.23, p = 6.1 × 10 −3 ; right, r = 0.23, p = 5.8 × 10 −3 )
Figure 5d legend; Results, Spatial variation of LC-NE activity, paragraph 7How it will be judged
Each neuron's CCF coordinate projected on the unit axis by dot product and correlated (Pearson) with its t-statistic (Methods, Relating behavioral features to spatial axes). The axes come from the retrograde tracing (LDA) and MERFISH (CCA) data; the code prints them (F_spatial-axis-comparison.ipynb cell 25) as [-0.248, 0.499, 0.830] and [0.024, 0.332, 0.943] and centres coordinates before projection. Reproduced when both r are positive with p < 0.05 and within tolerance.
- fig5g-switch-dorsal
Neurons with higher activity in the 500 ms after the go cue on switch than on stay choices lie more dorsal in LC (114 cells, 15 mice; linear model R2 = 0.168).
Neurons with higher activity during switch vs. stay choices (magenta) were located more in dorsal than ventral LC (114 cells, 15 mice; t -statistics shown in h ; kNN cross-validated R 2 = –4.96 × 10 −2 , p = 0.15; linear model R 2 = 0.168, p = 3.0 × 10 −4 )
Results, Spatial variation of LC-NE activity, paragraph 8; Figure 5f, 5g legendHow it will be judged
Per neuron, OLS of the rate 0.01 to 0.5 s after the go cue on switch against stay over responded trials, with the drift regressor (Methods, Linear regression models of spike rates). Open, run as crossed readings (a) switch defined against the previous trial (code, response_tstats_generation.py) or against the previous responded trial (as in the code's Figure 5f and photometry); (b) regressors switch and drift (Methods) or switch, drift, choice side and their interaction (code). Spatial tests as in claim fig5c-ignore-ventral. Reproduced when, under one reading, the linear p < 0.05 with switch-preferring neurons dorsal and R2 within tolerance; the kNN result is reported.
- fig5h-switch-on-anatomy-axes
Switch-versus-stay t-statistics vary along the primary spatial axes of axonal projections (r = 0.32) and gene expression (r = 0.30).
Switch vs. stay t -statistics projected onto the primary spatial axes from axonal projections and gene expression (left, r = 0.32, p = 6.1 × 10 −5 ; right, r = 0.30, p = 1.7 × 10 −4 )
Figure 5h legend; Results, Spatial variation of LC-NE activity, paragraph 8How it will be judged
As claim fig5d-ignore-on-anatomy-axes, on the t-statistics of claim fig5g-switch-dorsal under each of its readings. Reproduced when both r are positive with p < 0.05 and within tolerance under one reading.
- fig5k-cortical-input-switch
LC-NE axonal fluorescence in prelimbic cortex and spiking of LC-NE neurons projecting to cortex are higher on switch than on stay choices.
LC-NE axonal fluorescence changes in PL and single-neuron spiking projecting to frontal cortex were significantly higher during switch versus stay choices ( Figures 5j and 5k ; Fisher's exact test, p = 7.8 × 10 −5 )
Results, Spatial variation of LC-NE activity, paragraph 9; Figure 5j, 5kHow it will be judged
Photometry per session with axonal signal passing the Methods' quality threshold (go-cue increase above 0.5 z-scored dF/F, photometry_combine.py); OLS of dF/F in 0.75 s after the go cue on switch against stay (Methods, Linear regression models of axonal GCaMP). Open, run as readings (a) with the isosbestic signal as regressor (Methods) or with outcome and without the isosbestic signal (code, photometry_generation.py); (b) switch against the previous responded trial (code) or the previous trial. Fisher's exact test on the counts of positive and negative switch t-statistics; the groups compared are open, readings (i) photometry sessions against all 241 neurons of claim fig5g-switch-dorsal (code, A&O cell 61); (ii) photometry sessions with cortex-projecting neurons pooled, against the other neurons. Reproduced when the photometry sessions and cortex-projecting neurons have mostly positive switch t-statistics and the Fisher test gives p < 0.05 under one reading.
- figs17i-baseline-axons-vs-spiking
Before go cues, axonal fluorescence differs from single-neuron spiking in how it relates to responding (Fisher p = 4.2e-4), while cortex-projecting neurons do not differ significantly from the population of single neurons.
Axonal fluorescence vs. single-neuron spiking, Fisher's exact test with permutation, p = 4.2 × 10 −4 . Single neurons projecting to cortex compared to the population of single neurons, Welch test with permutation, p = 6.7 × 10 −2 ; Fisher's exact test with permutation, p = 8.8 × 10 −2
Figure S17i legendHow it will be judged
Respond-versus-ignore t-statistics on the 2 s before the go cue for neurons (claim fig5b-baseline-predicts-ignore) and photometry sessions (Methods, Linear regression models of axonal GCaMP). As in the code (A&O cells 57 and 61, 5,000 permutations), the sign comparison is Fisher's exact test on positive against non-positive counts with a label permutation (legend); the Methods (Comparison between single neuron activity and axonal GCaMP) describe a two-proportion z-test and a label-permutation test on the difference in positive rates, run as a variant. The Welch test is two-sided with a label permutation (Methods, Comparison between single neuron activity and axonal GCaMP). Reproduced when axons against neurons gives p < 0.05, and, as a claim of absence, cortex-projecting against other neurons gives p >= 0.05 in both tests; effect sizes reported.
- fig6-outcome-responsive-fraction
76.8% of LC-NE neurons had significant regressions of outcome-period activity.
Across the population, LC-NE neurons showed varied responses to the reward or its absence (76.8% cells with significant regressions
Results, Activity of LC-NE projections to frontal cortex, paragraph 1; Figure 6a, S15cHow it will be judged
Per neuron, OLS of z-scored spike counts on outcome, choice side, their interaction and chosen value Q, with the drift regressor, in a 1.5 s bin centred at the outcome time plus the auROC lag for reward and for no reward, keeping the window with the larger |t| of outcome (Methods, Linear regression models of spike rates; beh_combined_outcome_generation.py and auc_windows.json). Q from the Q-learning model (aind_stan_fit_sim). The code does not compute 76.8%; readings (a) outcome coefficient p < 0.05; (b) whole-model F-test p < 0.05; over the 241 neurons. Reproduced when the fraction is within tolerance under one reading.
- fig6a-outcome-space
Neurons excited by reward lie more dorsal and anterior in LC than neurons excited by no reward; their distribution in space is nonuniform (p = 3.0e-4).
Neurons excited by reward (cyan) or no reward (pink) were distributed nonuniformly in space ( p = 3.0 × 10 −4 )
Results, Activity of LC-NE projections to frontal cortex, paragraph 2; Figure 6a legendHow it will be judged
The outcome t-statistics of claim fig6-outcome-responsive-fraction for neurons with CCF locations, tested as in claim fig5c-ignore-ventral. The code (A&O cell 48) runs both the linear and the kNN test and does not say which gives p = 3.0e-4; both are readings. Reproduced when one gives p < 0.05 and the linear gradient puts positive (reward-excited) t-statistics dorsal and anterior.
- figs18d-outcome-not-on-anatomy-axes
Reward-versus-no-reward t-statistics are not correlated with neuron position along the primary axes from other modalities (r = -0.01 and 0.01).
How it will be judged
A claim of absence, judged on the paper's test, Pearson correlation of the outcome t-statistics with position along each axis as in claim fig5d-ignore-on-anatomy-axes, at p < 0.05. The axes are open; readings are the retrograde and MERFISH axes (as in Figure 5d) and the waveform axis. Reproduced when no correlation reaches p < 0.05 on the two axes of a reading.
- fig6f-outcome-angles-nonuniform
The angles of outcome and chosen-value coefficients cluster nonuniformly across LC-NE neurons (Rayleigh z = 19.53), separating RPE-like neurons from neurons excited by no reward.
Polar histogram of coefficients from regressions in ( e ) shows a nonuniform distribution of activity (Rayleigh test, z = 19.53, p = 2.3 × 10 −9 )
Results, Activity of LC-NE projections to frontal cortex, paragraph 2; Figure 6e, 6fHow it will be judged
Angle arctan2(chosen-value coefficient, outcome coefficient) per neuron from the regressions of claim fig6-outcome-responsive-fraction, over the 241 neurons, Rayleigh test (A&O cells 55 to 56). Reproduced when p < 0.05 and z is within tolerance, with modes near RPE (positive outcome, negative chosen value) and near negative outcome.
- fig6f-cortex-projecting-rpe-biased
LC-NE neurons identified as projecting to isocortex are biased toward RPE coding compared with other LC-NE neurons (Watson-Wheeler W = 0.04, permutation p = 0.01).
Purple: neurons projecting to isocortex are RPE-biased (Watson-Wheeler test, W = 0.04, permutation p = 0.01)
Figure 6e, 6f legendHow it will be judged
Angles of claim fig6f-outcome-angles-nonuniform for cortex-projecting neurons (antidromic criteria of the Methods) against the other neurons; rank-based Mardia-Watson-Wheeler statistic with 5,000 label permutations, p = (k + 1)/(N + 1) (Methods, Circular distribution comparison). Reproduced when p < 0.05 and the cortex-projecting neurons' mean angle is closer to the RPE direction than that of the others.
- fig6g-rpe-vs-no-reward-populations
Neurons excited by reward scale with RPE, whereas neurons excited by no reward respond to lack of reward with little modulation by expected reward.
Average responses of neurons of each population showed modulation by RPE or lack of reward ( Figure 6g )
Results, Activity of LC-NE projections to frontal cortex, paragraph 2; Figure 6gHow it will be judged
Neurons split by outcome response sign (auROC above or below 0.5, Methods, Linear regression models of spike rates); mean outcome-window rate against RPE quartiles (the binning of the Figure 6d examples; the Figure 6g binning is not stated). Reproduced when the reward-excited group's rate increases monotonically with RPE across quartiles and the no-reward-excited group's rate on unrewarded trials does not differ across chosen-value quartiles (one-way ANOVA p >= 0.05, a standard choice; the paper gives no test).
- fig6h-opposing-behavior
Neurons correlated with RPE and neurons excited by lack of reward have opposing correlations with choice behavior on subsequent trials.
Neurons that correlated with RPE or lack of reward had opposing correlations with choice behavior on subsequent trials ( Figure 6h )
Results, Activity of LC-NE projections to frontal cortex, paragraph 2; Figure 6hHow it will be judged
For each group of claim fig6g-rpe-vs-no-reward-populations, the change of choice likelihood (claim fig6c-choice-change-rpe) against outcome-window rate binned by quantile (the binning is not stated; quartiles, as in Figure 6d, are a standard choice). Reproduced when the slopes of the two groups have opposite signs; the paper gives no test, so a sign difference of the group mean slopes with a two-sample t-test p < 0.05 is the standard choice used.
- fig6i-axons-rpe
LC-NE axonal activity in prelimbic cortex after the choice outcome correlates with RPE, with regression angles that differ from the single-neuron population (Watson-Wheeler W = 0.1).
LC-NE axonal activity in frontal cortex 1.5 s after choice outcome correlated with RPE ( Figures 6i–6l )
Results, Activity of LC-NE projections to frontal cortex, paragraph 3; Figure 6i to 6lHow it will be judged
Per photometry session passing the quality threshold, OLS of dF/F on outcome, choice side, their interaction and chosen value, with the isosbestic signal as regressor, in a 2 s window after the outcome (Methods, Linear regression models of axonal GCaMP). The code aligns to the choice time with a bin centred near 0.7 s; the Results say 1.5 s after outcome; these are readings. Unit of sample open, per session (code) or per mouse (legend, 9 mice; the code has 7). Reproduced when the mean outcome t-statistic is positive and the mean chosen-value t-statistic negative (RPE sign), and the Watson-Wheeler comparison of axonal and single-neuron angles (5,000 permutations) gives p < 0.05, under one reading.
- fig6j-cortical-input-more-positive
Outcome regression t-statistics of cortex-projecting neurons and of axonal fluorescence are more positive than those of the single-neuron population (Fisher's exact test, p < 0.01).
Regression t -statistics ( Figure 6j ) and coefficients ( Figure 6k ) were more positive than the population of single neurons (Fisher's exact test, p < 0.01)
Results, Activity of LC-NE projections to frontal cortex, paragraph 3; Figure 6j, 6kHow it will be judged
Fisher's exact test on counts of positive and non-positive outcome t-statistics (claims fig6-outcome-responsive-fraction and fig6i-axons-rpe), with the code's binary_shift_X_vs_Y. Groups open, readings (a) photometry sessions against all neurons; (b) cortex-projecting neurons against other neurons. Reproduced when the group named has the higher positive fraction with p < 0.01 under one reading; under both readings is the full statement.
- figs14-waveform-space
Dorsal LC-NE neurons have more symmetric spike waveforms and longer peak-to-trough times than ventral neurons; slope symmetry is spatially organized (kNN and linear p = 5.0e-4).
dorsal LC-NE neurons tended to have more symmetric spike waveforms and longer peak-to-trough times than ventral LC-NE neurons
Results, Spatial variation of LC-NE activity, paragraph 4; Figure S14b, S14cHow it will be judged
Waveforms as in the Methods (Spike waveform extraction), from raw traces filtered 50 to 8,000 Hz, tetrode and Neuropixels 2.0 units only, ISI violation ratio below 0.5 and peak below -50 uV, aligned to peak and normalized. If the dataset holds only mean waveforms, those are used and the difference is a deviation. Spatial tests as in claim fig5c-ignore-ventral. Reproduced when slope symmetry has linear and kNN p < 0.05 with more symmetric waveforms dorsally, and peak-to-trough time increases dorsally.
- figs15q-spontaneous-vs-task-licks
LC-NE peak spike rates around spontaneous licks correlate with those around in-trial licks across neurons (r = 0.49).
Peak spike rate around spontaneous vs. in-trial licks ( r = 0.49, p < 0.001)
Figure S15q legendHow it will be judged
Lick bouts separated by inter-lick intervals above 0.5 s (Methods, Lick bout detection), split into in-trial and spontaneous bouts; per neuron the peak of the bout-aligned rate for each; Pearson across neurons (F_spont_choice_lick_neuron.ipynb). The peak window and bin are open; the code's are used. Reproduced when r is positive with p < 0.001 and within tolerance.
- figs15t-crosscorrelogram-outcome-axis
The first PC of cross-correlograms between simultaneously recorded LC-NE neurons varies along the spatial axis of outcome responses, but not of waveform, respond-versus-ignore or switch-versus-stay.
Spatial analysis shows PC1 of cross-correlogram changes along axes of outcome, but not other task responses (kNN cross-validation bootstrapped p -values
Figure S15t legendHow it will be judged
Spike counts in 50 ms bins, cross-correlation for each simultaneously recorded pair, PCA over pairs (Methods, Spike count cross-correlations). How a pair is placed on each axis is not stated; readings (a) the midpoint of the two neurons' CCF positions projected on the axis; (b) the difference of their projections. Reproduced when, under one reading, the outcome axis gives p < 0.05 and, as absence on the paper's test, the other three p >= 0.05.
- figs16-pupil-spiking
LC-NE spiking correlates with pupil size; pupil autocorrelation decays with a time constant of 3.37 +/- 1.03 s across sessions.
τ had a mean of 3.37 and SD of 1.03
Figure S16 title and panels i, j; Methods, Pupil auto-correlogramHow it will be judged
Pupil diameter from left and right edges with confidence at least 0.9, 5 Hz low-pass (Methods, Pupil tracking); exponential fit, A exp(-t/tau) + C, to the whole-session autocorrelogram per session, computed with a 0.5 s sliding window and with periods of poor time alignment or edge detection removed (Methods, Pupil auto-correlogram). For spiking, per neuron, Pearson correlation over trials of the rate and pupil diameter 1 s before the go cue (Methods, Pupil correlation with neuron activity). The paper gives no population test; reproduced when the mean tau is within tolerance and more neurons have significant positive than negative baseline correlations (binomial test p < 0.05, a standard choice).
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none