Cortical Membrane Potential Signature of Optimal States for Sensory Signal Detection

Abstract

The neural correlates of optimal states for signal detection task performance are largely unknown. One hypothesis holds that optimal states exhibit tonically depolarized cortical neurons with enhanced spiking activity, such as occur during movement. We recorded membrane potentials of auditory cortical neurons in mice trained on a challenging tone-in-noise detection task while assessing arousal with simultaneous pupillometry and hippocampal recordings. Arousal measures accurately predicted multiple modes of membrane potential activity, including rhythmic slow oscillations at low arousal, stable hyperpolarization at intermediate arousal, and depolarization during phasic or tonic periods of hyper-arousal. Walking always occurred during hyper-arousal. Optimal signal detection behavior and sound-evoked responses, at both sub-threshold and spiking levels, occurred at intermediate arousal when pre-decision membrane potentials were stably hyperpolarized. These results reveal a cortical physiological signature of the classically observed inverted-U relationship between task performance and arousal and that optimal detection exhibits enhanced sensory-evoked responses and reduced background synaptic activity.

Figure 1. Peripheral and central state fluctuate rapidly and continuously in a highly correlated manner in the waking mouse

a, Head-fixed mice spontaneously stood or walked on a cylinder during whole-cell recordings from ACtx neurons. b, Pupil diameter was monitored via infrared camera and found to vary over a wide range (e.g. 0.5 to 1.8 mm in this animal) during the waking state. c, Pupil diameter is highly correlated with the rate of ripples in the CA1 area of the hippocampus. Light pink trace is an overlay of the inverted pupil diameter (upwards = smaller pupil diameter) with rate of hippocampal ripples. Lines highlight example peaks in both variables. d, Increased pupil diameter is associated with a large decrease in hippocampal ripple rate (R=−0.94 +/− 0.03; N=6). e, Coherence between pupil dilation and CA1 ripple rate exhibits an average peak of 0.78 (N=6) at low frequencies, demonstrating that pupil dilation is strongly associated with a decreased rate of occurrence of hippocampal ripples. f, The coherence between pupil diameter and ripple rate is consistently anti-phase (~180°) across the full range of relevant frequencies. Gray shaded areas indicate 68% CI.

Figure 2. Global state fluctuates together with cortical membrane potentials and their oscillations

a, An example ACtx membrane potential (Vm; top), the standard deviation and mean of the Vm in a 1 second sliding window (middle) and pupil dilation and walking patterns (bottom) on the same time base. Vm mean and S.D. are blanked during action potentials, indicated by asterisks at top. Brief periods of increased pupil diameter are highly correlated with depolarizations in the ACtx neuronal membrane potential. Just prior to, and during, walking the membrane potential depolarizes and pupil diameter increase together. When pupil diameter becomes small, slow oscillatory synaptic activity becomes prominent. b, Expansion in time of the relationship between neuronal membrane potential, pupil diameter, and walking. Note the brief depolarizations time-locked to pupil dilations. c, Coherence between pupil diameter and membrane potential exhibits an average peak of 0.68 (n=9) at low frequencies (0.03 Hz) and a second peak at 0.3 Hz that corresponds to the brief periods of pupil dilation (microdilations). Depolarization of Vm preceded pupil dilation by 1.4 seconds, as indicated by the lag in the peak of the cross-correlation between pupil and Vm (inset), for the example trace shown in panels a and b. Gray shaded area indicates 68% CI.

Figure 3. Neuronal membrane potentials and their variability exhibit an inverted-U relationship with arousal, as indexed by the pupil diameter

a, As pupil diameter decreases below a critical zone (grey box), microdilations tend to stop and large fluctuations in synaptic activity at low (2–10 Hz) frequencies become prominent (at bottom). These low frequency synaptic barrages are largely suppressed during pupil dilation and the correlation between microdilations and brief depolarizations of the membrane potential are observed. At intermediate arousal when microdilations are absent, neurons are stably hyperpolarized (green horizontal dashed line). Asterisks indicate action potentials. b, Dependence of low-frequency and high-frequency Vm fluctuations on pupil diameter. Small diameters are associated with increased relative levels of low frequency fluctuations in membrane potential, while large diameters are associated with increased relative levels of power at higher frequencies (see also Fig. S7). c, Two-dimensional heatmap (each vertical slice normalized to an area of 100%) showing group data (n=9 recordings) for the relationship between change in membrane potential from the minimum (lower bound of a 98% confidence interval) potential reached during the recording and normalized pupil diameter. Note that the membrane potential exhibits large variations for either small or large pupil dilations, with the smallest variability occurring at intermediate pupil dilation. Also note that membrane potentials are most hyperpolarized at intermediate pupil diameters. Zero ‘Δ Membrane Potential’ corresponds to an average Vm of −74 ± 1.2 mV. Inset shows the color scale of the heat map in % units. d, Standard deviation of membrane potential exhibits a non-monotonic relationship with pupil diameter. e, Spontaneous multiple unit (MU) firing rate in auditory cortex exhibits a U-shaped relationship with pupil diameter. Some of the increased unit activity with large pupils is associated with walking. Shaded areas indicate 68% CI.

Figure 4. Subthreshold and suprathreshold evoked responses in auditory cortex are of maximal amplitude and reliability at intermediate levels of arousal

a, Example whole-cell recording of responses to repeated presentation of a complex sound (TORC), sorted by pupil diameter. Sound presentation evokes the most spikes and largest, most reliable membrane potential depolarizations at intermediate pupil diameters. Tick marks on left are at the same membrane potential for reference. b, Group data (n=6 recordings, N=5 animals) illustrating that the maximal evoked action potential response in the Vm occurs at intermediate pupil dilation. c, Average TORC-evoked PSP amplitude (over baseline; see methods) peaks at intermediate pupil diameters. d, Reliability of synaptic responses (trial-to-trial cross-correlation of membrane potential) is highest at intermediate pupil dilations. e, Example PSTH responses of ACtx multiunit activity to a complex (TORC) sound are largest and most reliable at intermediate levels of pupil dilation. f, Group (n=18 recordings, N=6 animals) data illustrate that the average evoked firing rate is largest for intermediate pupil diameters. Both small and large pupil diameters are associated with decreased evoked (over spontaneous) responses. Walking is associated with decreased auditory responses, but to an extent not significantly different from responses during periods of non-walking with the same pupil diameters. g, Response reliability (average trial-to-trial cross correlation) is highest at mid-pupil diameters. Walking increases reliability over periods of non-walking with correspondingly large pupil diameters (p<0.05). h, The increase in variance explained (over the PSTH alone) by incorporating walking status (left), pupil diameter (middle) or both (right). Incorporating both walking and pupil explains more variance than walking along, but not more than pupil diameter alone (n=18; p<0.05). Shaded areas indicate 68% CI.

Figure 5. Cortical and thalamic gain peaks at intermediate levels of arousal

a, Increases in pupil diameter are associated with modest increases in MU spontaneous activity in the medial geniculate (MG). Walking was not associated with an additional increase in spontaneous activity compared to pupil dilation alone. b, TORC-evoked average firing rate over baseline peaks at intermediate pupil diameters and decreases with arousal. Walking is associated with a strong additional suppression of evoked MG responses. c, Reliability (trial-to-trial cross correlation) of MG responses peaks at intermediate pupil diameters. d, Example histogram of a MU MG response to a TORC stimulus. e,h, Average action potential counts in 20 msec bins were sorted and normalized from 0 (no spikes) to 1 (largest response) for the x-axis. The data was then resorted according to pupil diameter (see Methods) for the y-axis. If arousal (pupil diameter) had no effect on response, then all the data should fit along the diagonal with a slope of 1. Both MG and ACtx responses were strongly affected by pupil diameter, with peak responses occurring at intermediate pupil diameters. f,g, Gain, as measured by our gain model (see Methods), is modulated as changes in slope in MG (top) and cortex (bottom), while offset was relatively unchanged. This model is derived from un-binned data points, and therefore differs slightly from the plots of e,h, which are based on binned and averaged data for illustrative purposes (see Methods). Shaded areas indicate 68% CI.

Figure 6. Sound detection performance fluctuates moment-to-moment with state and is optimal at intermediate arousal levels

a, The ability of the animal to perform the auditory detection task and the pupil diameter vary widely throughout the training session. Note that false alarms appear prevalent because many more reference than target sound stimuli were presented (see Methods). b, Example trial structure for the detection task. Mice were trained to lick for a reward upon detection of a pure tone embedded in a complex (TORC) sound. The level of the tone varied randomly from trial to trial. The frequency of the tone varied from day to day. c, Plotting pupil diameter over the course of the session revealed a broad range that centered around 50% diameter (N=5 animals; n=30 sessions). Walking was always associated with large pupil diameters. d, Plotting response latency on hit trials versus pupil diameter, for an example session, reveals that intermediate pupil diameters are associated with short latency licks and a large decrease in variability of lick latency. e, Group data demonstrating that lick latency on hit trials is shortest for intermediate pupil diameters (n=30 sessions; N=5 animals). On false-alarm trials, lick latency from the start of the most recent TORC are significantly longer. Variability of lick latency (measured as average standard deviation) is lowest at mid pupil diameters for both hit and false alarm responses (dashed lines and open symbols). f, Performance rate (hit rates or false alarm rate; FAR) is maximal for intermediate pupil diameter for all tone levels, with the greatest modulation occurring between small and medium pupil diameters for quiet sounds. Increases in pupil diameter are associated with a large increase in false alarm rate (FAR). g, Decision bias (c) is minimal at intermediate pupil diameters for all tone levels. h-, Group data, averaged across all tone levels, demonstrating that intermediate pupil diameters are associated with maximal hit rate and sensitivity and minimal bias. False alarm rate is reproduced from panel f. Walking does not have additional effects to non-walking with large pupils, except for an additional increase in bias. j, Accounting for pupil diameter substantially increases the accuracy of estimates of d′, bias, and latency over knowing tone intensity alone. Shaded areas indicate +/− 1 SEM.

Figure 7. Auditory cortical membrane potentials are most hyperpolarized and least variable before correct target identifications during the detection task

a, An example membrane potential recorded during behavior. Lick patterns and reward administration are indicated at top. The sound spectrogram is recreated at bottom. Note horizontal green bars indicating the presence of target sounds and trial outcomes noted above. Asterisks indicate action potentials. b, The average membrane potential time course (+/− 1 S.D.) just before and after the start of sound, for all hits (green), misses (blue) and false alarms (FA; red) from an example recording during behavior. c, The average membrane potential trajectory sorted by trial type for the population of membrane potential recordings during behavior. Membrane potentials are hyperpolarized before hit trials, depolarized before false alarms, and intermediate before miss trials. Evoked responses are largest (from baseline) for hits and smallest for false alarms. d, The variability (S.D.) across trials of the membrane potentials, before and during the early component of sounds, is smallest for hit trials, largest for false alarms, and intermediate for miss trials. Data for Hit and Miss trials in b–d are resampled (1000 times with replacement) to have equivalent target tone level distributions. Shaded areas indicate 68% CI unless noted.

Figure 8. Influence of internal state on neuronal and behavioral accuracy and responsiveness

a, Brainstem/hypothalamic mechanisms of state control simultaneously modulate: pupil diameter, through the superior cervical ganglion (SCG) and the Edinger-Westphal (EW) nucleus; activity within the central auditory pathway, including the medial geniculate nucleus (MG) and primary auditory cortex (A1); and the animal’s behavior. b, Relationship between state, behavioral performance, and auditory neural responsiveness. When pupil size is small, cortical slow frequency oscillations are prominent (slow waves), resulting in low gain, low reliability of sound-evoked neuronal responses in the auditory cortex, and thus poor performance on the auditory detection task (low percent correct choices, frequent misses, variable latency behavioral responses). Moderate arousal (quiet, alert) is associated with a suppression of slow wave activity and an increase in the gain and reliability of auditory cortical evoked neuronal responses. Behavioral performance is optimal, with short latency, correct choices, and less variability. The hyper-aroused state (with or without walking), is associated with further ACtx neuronal depolarization, the appearance of fast as well as arrhythmic synaptic activity, and a decrease in MG evoked responses. Together, these effects result in decreased amplitude and reliability of ACtx sound evoked responses, and this is also reflected by diminished behavioral accuracy, with increased false alarms, misses, and more variable and longer response latencies.