Skip to main content
PMC home page
eLife logoLink to eLife
. 2016 Jun 21;5:e15798. doi: 10.7554/eLife.15798

Target-specific membrane potential dynamics of neocortical projection neurons during goal-directed behavior

Takayuki Yamashita 1,2,*, Carl CH Petersen 1,*
Editor: Naoshige Uchida3
PMCID: PMC4915810  PMID: 27328320

Abstract

Goal-directed behavior involves distributed neuronal circuits in the mammalian brain, including diverse regions of neocortex. However, the cellular basis of long-range cortico-cortical signaling during goal-directed behavior is poorly understood. Here, we recorded membrane potential of excitatory layer 2/3 pyramidal neurons in primary somatosensory barrel cortex (S1) projecting to either primary motor cortex (M1) or secondary somatosensory cortex (S2) during a whisker detection task, in which thirsty mice learn to lick for water reward in response to a whisker deflection. Whisker stimulation in ‘Good performer’ mice, but not ‘Naive’ mice, evoked long-lasting biphasic depolarization correlated with task performance in S2-projecting (S2-p) neurons, but not M1-projecting (M1-p) neurons. Furthermore, S2-p neurons, but not M1-p neurons, became excited during spontaneous unrewarded licking in ‘Good performer’ mice, but not in ‘Naive’ mice. Thus, a learning-induced, projection-specific signal from S1 to S2 may contribute to goal-directed sensorimotor transformation of whisker sensation into licking motor output.

DOI: http://dx.doi.org/10.7554/eLife.15798.001

Research Organism: Mouse

eLife digest

Many animals can learn quickly to associate specific behaviors with rewards, such as food. Often, the animal’s senses of smell, taste, and touch trigger the behavior, and this allows an animal to respond favorably to changes in its environments. However, it is not clear exactly what happens in the animal’s brain to reinforce a behavior that ends in a reward, or how its senses help trigger the rewarding behavior.

Yamashita and Petersen have now studied what happens in the brains of mice that were taught to complete a task to get a reward. In the training, thirsty mice learned that they would receive a reward of water if they licked a water spout after they were briefly touched on one of their whiskers. Then, Yamashita and Petersen measured electrical changes in the brain cells of the trained mice and compared those with measurements from the brain cells of untrained mice. The measurements specifically focused on the brain cells that receive sensory information from the whiskers. These cells are in a region of the brain called the primary sensory cortex, which is known to help mice carry out the task. This brain area in turn sends signals to many downstream areas of the brain.

Yamashita and Petersen found that learning the task appeared to enhance the signaling of some cells in this area of the mouse brain. However, this was only the case for the cells that send signals to a region of the brain that further processes the sensory information (the so-called secondary sensory cortex). Other cells that are intermingled in this region but send signals to the part of the brain that controls movement (the motor cortex) were not affected in this way. Together the data suggest that routing signals from the primary sensory cortex to specific downstream areas might allow animals to learn tasks that depend on responding to sensory cues. More studies are now needed to understand exactly how these signals are generated and whether they contribute to triggering the licking behavior in the mice.

DOI: http://dx.doi.org/10.7554/eLife.15798.002

Introduction

Primary sensory cortex processes incoming sensory information flexibly in an experience, context and task-dependent manner (Gilbert and Li, 2013; Harris and Mrsic-Flogel, 2013). Functionally-tuned sensory information is signaled from primary sensory cortex to distinct cortical areas (Movshon and Newsome, 1996; Sato and Svoboda, 2000; Chen et al., 2013; Glickfeld et al., 2013; Yamashita et al., 2013), but the cellular mechanisms underlying specific cortico-cortical signals during goal-directed behavior are poorly understood.

Neuronal activity in primary somatosensory barrel cortex (S1) is known to participate in the execution of a simple whisker-dependent detection task, in which thirsty mice learn to lick a spout in order to obtain a water reward (Sachidhanandam et al., 2013). In well-trained mice, putative excitatory neurons in layer 2/3 of S1, on average, have a long-lasting biphasic depolarization after whisker deflection in hit trials, whereas in miss trials the late depolarization is smaller in amplitude (Sachidhanandam et al., 2013). However, there is considerable variability across different recordings (Sachidhanandam et al., 2013), which could in part relate to distinct types of excitatory projection neurons. Layer 2/3 of S1 barrel cortex has major anatomical ipsilateral cortico-cortical connections to primary whisker motor cortex (M1) and secondary somatosensory cortex (S2) (Aronoff et al., 2010). M1-projecting (M1-p) and S2-projecting (S2-p) neurons in layer 2/3 of S1 are likely to be distinct cell-types exhibiting differential patterns of gene expression (Sorensen et al., 2015), distinct intrinsic electrophysiological properties in vivo (Yamashita et al., 2013), and carrying functionally different signals (Sato and Svoboda, 2010; Chen et al., 2013; 2015; Yamashita et al., 2013). Retrograde labeling suggests that M1-p and S2-p neurons in S1 are largely non-overlapping types of excitatory neurons (Sato and Svoboda, 2010; Chen et al., 2013; Yamashita et al., 2013). Here, we investigate the cellular basis of selective signaling of sensorimotor information in distinct cortico-cortical pathways during the whisker detection task through membrane potential (Vm) recordings of M1-p and S2-p neurons, finding that task learning induces a licking-related depolarization specifically in S2-p neurons.

Results

Differential Vm responses in S2-p and M1-p neurons during task performance

Thirsty mice were trained to lick for water reward in response to a 1 ms deflection of the right C2 whisker (Sachidhanandam et al., 2013; Sippy et al., 2015), and whole-cell Vm recordings were targeted through two-photon microscopy to fluorescently-labelled M1-p and S2-p neurons in layer 2/3 of the C2 barrel column in S1 of the left hemisphere (Yamashita et al., 2013) (Figure 1A,B). We used two types of mice for recordings: (1) ‘Good performer’ mice that exhibited a high discriminability between test trials and catch trials (for details see Materials and Methods) during recordings (59 recordings in 27 mice; hit rate, 0.77 ± 0.03; false alarm rate, 0.17 ± 0.01; d’ = 2.12 ± 0.09; d’ > 1.1 for each recording; Figure 1C) learned through training sessions (typically 7–13 daily sessions prior to the recording day, but some mice learned more quickly), responding with a reaction time of 317 ± 17 ms (time from whisker deflection to tongue contact with the water spout) (Figure 1D); and (2) ‘Naive’ mice that were used for recordings on the first day of being exposed to the task and showed no apparent discrimination (36 recordings in 16 mice; hit rate, 0.31 ± 0.03; false alarm rate, 0.28 ± 0.03; d’ = 0.03 ± 0.09; d’ < 0.9, for each recording; Figure 1C), with a mean reaction time of 369 ± 23 ms which was significantly slower than ‘Good performer’ mice (p=0.0014; Figure 1E).

Figure 1. Target-specific Vm dynamics in S1 projection neurons during task performance.

(A) Top, the experimental setup. Bottom, a representative two-photon image of a CTB-labeled M1-p neuron (green) with a recording pipette (red). (B) Detection task trial structure (FA: false alarm, CR: correct rejection). (C) Behavioral performance during whole-cell recordings (HR: hit rates, FAR: false alarm rates, GP: ‘Good performer’, N: ‘Naive’). (D, E) Left, grand average changes in Vm (thick line: mean, thin lines: ± sem) and action potential (AP) firing rate for hit trials recorded from S2-p neurons (red) and M1-p neurons (blue) in ‘Good performer’ (D) and ‘Naive’ (E) mice (Arrow: 1 ms stimulation of the C2 whisker). A box plot indicates reaction time (1st lick). Right, box plots for postsynaptic potential (PSP) amplitude, secondary late Vm depolarization quantified at 0.05–0.25 s, Vm depolarization during the lick period (at 0.25–1.0 s), and evoked AP rates at early (0–0.05 s), late (0.05–0.25 s) and lick (0.25–1.0 s) periods.

DOI: http://dx.doi.org/10.7554/eLife.15798.003

Figure 1—source data 1. Data values and statistics underlying Figure 1.
DOI: http://dx.doi.org/10.7554/eLife.15798.004
elife-15798-fig1-data1.xlsx (27.3KB, xlsx)
DOI: 10.7554/eLife.15798.004
Figure 1—source data 2. Data values and statistics underlying Figure 1—figure supplement 3.
DOI: http://dx.doi.org/10.7554/eLife.15798.005
elife-15798-fig1-data2.xlsx (22.6KB, xlsx)
DOI: 10.7554/eLife.15798.005
Figure 1—source data 3. Data values and statistics underlying Figure 1—figure supplement 4.
DOI: http://dx.doi.org/10.7554/eLife.15798.006
elife-15798-fig1-data3.xlsx (18.7KB, xlsx)
DOI: 10.7554/eLife.15798.006

Figure 1.

Figure 1—figure supplement 1. Hit Vm traces from S1 projection neurons in ‘Good performer’ mice.

Figure 1—figure supplement 1.

(A) Left and middle, example Vm traces obtained from two individual S2-p neurons on two representative hit trials. Right, the averaged subthreshold Vm trace is superimposed with individual traces. APs are truncated. PSTHs for corresponding AP rates and the distribution of the first lick timings are also shown. (B) Same as A, but for M1-p neurons.
DOI: http://dx.doi.org/10.7554/eLife.15798.007
Figure 1—figure supplement 2. Hit Vm traces from S1 projection neurons in ‘Naive’ mice.

Figure 1—figure supplement 2.

(A) Left and middle, example Vm traces obtained from two individual S2-p neurons on two representative hit trials. Right, the averaged subthreshold Vm trace is superimposed with individual traces. APs are truncated. PSTHs for corresponding AP rates and the distribution of the first lick timings are also shown. (B) Same as A, but for M1-p neurons.
DOI: http://dx.doi.org/10.7554/eLife.15798.008
Figure 1—figure supplement 3. Average hit Vm traces and PSTHs.

Figure 1—figure supplement 3.

(A) Average hit Vm traces obtained from individual cells (thin lines) from S2-p (left) and M1-p (right) neurons of ‘Good performer’ mice. The grand average Vm traces (thick lines) are superimposed. The dotted line indicates the stimulus onset. (B) Left, grand average of subthreshold responses from S2-p (red) and M1-p (blue) neurons are shown at high temporal resolution (superimposed with the baseline Vm subtracted). An arrow indicates the onset of whisker stimulation. Middle, grand average PSTHs at high temporal resolution with the baseline AP rates subtracted. Time 0 is the onset of whisker stimulation. Right, quantification of evoked AP rate at initial 20 ms after whisker deflection. (C and D) Same as A and B, but for ‘Naive’ mice. (E) Quantification of △Vm and △AP rate at 0.05 – 0.35 s and 0.35 – 1.0 s on hit trials in ‘Naive’ mice. (F) PSP amplitude, Vm depolarization at the late (0.05–0.25 s) and lick (0.25 – 1.0 s) periods and evoked AP rates at early (0 – 0.05 s), late (0.05 – 0.25 s) and lick (0.25 – 1.0 s) periods were analyzed on a mouse-by-mouse basis for ‘Good performer’ mice (n = 9 mice for M1-p neurons; n = 17 mice for S2-p neurons). (G) Same as F, but for ‘Naive’ mice (n = 6 mice for M1-p neurons; n = 9 mice for S2-p neurons).
DOI: http://dx.doi.org/10.7554/eLife.15798.009
Figure 1—figure supplement 4. Target-specific changes of hit responses with task learning.

Figure 1—figure supplement 4.

(A and B) Left, grand average changes in Vm (thick line: mean, thin lines: ± sem) and AP rate for hit trials recorded from S2-p neurons (A) and M1-p neurons (B) in ‘Good performer’ (colored) and ‘Naive’ (black) mice (Arrow: 1 ms stimulation of the C2 whisker). Right, box plots for PSP amplitude, Vm depolarization at the late (0.05–0.25 s) and lick (0.25 – 1.0 s) periods and evoked AP rates at early (0 – 0.05 s), late (0.05 – 0.25 s) and lick (0.25 – 1.0 s) periods.
DOI: http://dx.doi.org/10.7554/eLife.15798.010
Open in a new tab

In trained ‘Good performer’ mice, whisker stimulation evoked a biphasic Vm depolarization in hit trials for both M1-p and S2-p neurons (Figure 1D, Figure 1—figure supplement 1,3). The early sensory response was not different comparing M1-p and S2-p neurons (p=0.40), but S2-p neurons had significantly larger Vm depolarization (△Vm) during the late phase (△Vm at 0.05 – 0.25 s after whisker deflection: S2-p=4.00 ± 0.59 mV, n = 31; M1-p=1.68 ± 0.44 mV, n = 22; p=0.0035) and during the subsequent lick period (△Vm at 0.25 – 1.0 s after whisker deflection: S2-p=3.10 ± 0.50 mV, n = 31; M1-p=0.73 ± 0.30 mV, n = 22; p=0.00038) (Figure 1D, Figure 1—figure supplement 1,3). The evoked action potential (AP) rate of S2-p neurons compared to M1-p neurons was also significantly higher during the lick period (p=0.015), but not during early (p=0.57) or late (p=0.98) response periods (Figure 1D, Figure 1—figure supplement 1,3).

In randomly licking ‘Naive’ mice (Figure 1C), M1-p neurons, compared to S2-p neurons, exhibited larger postsynaptic potentials (PSPs) in response to whisker stimulation in hit trials (PSP amplitude: S2-p=8.41 ± 0.69 mV, n = 14; M1-p=12.20 ± 1.25 mV, n = 12; p=0.043) and larger depolarizations during the licking phase (△Vm at 0.25 – 1.0 s after whisker deflection: S2-p=0.45 ± 0.62 mV, n = 14; M1-p=2.19 ± 0.57 mV, n = 12; p=0.043) (Figure 1E, Figure 1—figure supplement 2,3). The AP rates in ‘Naive’ mice during the lick period were also significantly larger in M1-p neurons compared to S2-p neurons (p=0.046) (Figure 1E, Figure 1—figure supplement 2,3).

Therefore, analyzed for hit trials, S2-p neurons were more strongly excited during licking compared to M1-p neurons in ‘Good performer’ mice, but, interestingly, the opposite was true for ‘Naive’ mice in which M1-p neurons were more excited during licking compared to S2-p neurons. Notably, the secondary long-lasting depolarization in S2-p neurons after whisker stimulation was seen only in ‘Good performer’ mice, not in ‘Naive’ mice (Figure 1D,E, Figure 1—figure supplement 4), while the small sustained depolarization of M1-p neurons in ‘Naive’ mice was attenuated in ‘Good performer’ mice (Figure 1D,E, Figure 1—figure supplement 4).

Differential hit vs miss responses in S2-p and M1-p neurons

We next examined whether the Vm dynamics of S1 projection neurons correlated with task execution on a trial-by-trial basis. In S2-p neurons of ‘Good performer’ mice, the amplitude of PSPs and the late △Vm were slightly larger in hit compared to miss trials (PSPs increased by 20%, p=0.026; late △Vm increased by 39%, p=0.029) (Figure 2A, Figure 2—figure supplement 1 and 2). Furthermore, the △Vmin S2-p neurons during the licking period was substantially larger (270% increase) in hit compared to miss trials (△Vm at 0.25 – 1.0 s: hit 2.63 ± 0.55 mV, miss 0.71 ± 0.47 mV, n = 19, p=0.0014) (Figure 2A, Figure 2—figure supplement 1 and 2). Thus, the Vm dynamics of S2-p neurons after whisker deflection were correlated with task execution in trained mice. However, hit and miss trials were not significantly different in M1-p neurons of ‘Good performer’ mice in early (p=0.23), late (p=0.43) or licking (p>0.99) phases (Figure 2B, Figure 2—figure supplement 1 and 2).

Figure 2. Target-specific Vm correlation with task execution.

(A) Left, grand average Vm traces (thick line: mean, thin lines: ± sem) of S2-p neurons during hit (red) and miss (black) trials for ‘Good performer’ mice. Right, data for each cell (thin lines) and box plots for PSP amplitude and Vm depolarization at the late (0.05–0.25 s) and lick periods (0.25–1.0 s) on hit (H) and miss (M) trials. (B) Same as A, but for M1-p neurons. (C) Same as A, but for S2-p neurons in ‘Naive’ mice. (D) Same as C, but for M1-p neurons.

DOI: http://dx.doi.org/10.7554/eLife.15798.011

Figure 2—source data 1. Data values and statistics underlying Figure 2.
DOI: http://dx.doi.org/10.7554/eLife.15798.012
elife-15798-fig2-data1.xlsx (19KB, xlsx)
DOI: 10.7554/eLife.15798.012
Figure 2—source data 2. Data values and statistics underlying Figure 2—figure supplement 2.
DOI: http://dx.doi.org/10.7554/eLife.15798.013
elife-15798-fig2-data2.xlsx (15.7KB, xlsx)
DOI: 10.7554/eLife.15798.013

Figure 2.

Figure 2—figure supplement 1. Representative hit/miss Vm traces.

Figure 2—figure supplement 1.

(A) Example average subthreshold Vm traces on hit (colored) and miss (black) trials from four individual S2-p neurons in trained ‘Good performer’ mice. Arrows: the onset of whisker stimulus. Insets: PSPs of corresponding traces at high temporal resolution. (B) Same as A, but for M1-p neurons. (C) Same as A, but for S2-p neurons in ‘Naive’ mice. (D) Same as C, but for M1-p neurons.
DOI: http://dx.doi.org/10.7554/eLife.15798.014
Figure 2—figure supplement 2. Mouse-by-mouse analysis of hit/miss responses.

Figure 2—figure supplement 2.

(A) Data from S2-p neurons in ‘Good performer’ mice averaged for each mouse (thin lines, n = 14 mice) and box plots for PSP amplitude and Vm depolarization at the late (0.05–0.25 s) and lick periods (0.25–1.0 s) on hit (H) and miss (M) trials. (B) Same as A, but for M1-p neurons (n = 7 mice). (C) Same as A, but for S2-p neurons in ‘Naive’ mice (n = 9 mice). (D) Same as C, but for M1-p neurons (n = 6 mice).
DOI: http://dx.doi.org/10.7554/eLife.15798.015
Open in a new tab

In contrast, for ‘Naive’ mice, S2-p neurons did not distinguish hit and miss trials in early (p=0.24), late (p=0.67) or licking (p=0.71) response phases (Figure 2C, Figure 2—figure supplement 1 and 2). M1-p neurons in ‘Naive’ mice also had similar hit and miss responses during early (p=0.42) and late (p=0.13) periods. However, M1-p neurons in ‘Naive’ mice had significantly larger △Vm during the lick period in hit trials compared to misses (△Vm at 0.25 – 1.0 s: hit 2.19 ± 0.57 mV, miss 0.30 ± 0.38 mV, n = 12, p=0.034) (Figure 2D, Figure 2—figure supplement 1).

Thus, S2-p neurons, but not M1-p neurons, in ‘Good Performer’ mice had larger depolarizing responses in hit trials compared to misses, whereas in ‘Naive’ mice M1-p neurons, but not S2-p neurons, had a larger depolarization during licking in hit trials compared to misses.

Differential depolarization of S2-p and M1-p neurons during spontaneous licking

Some S2-p neurons depolarized strikingly during spontaneous unrewarded licking (Figure 3—figure supplement 1). We therefore examined licking-related Vm dynamics and found that S2-p neurons of ‘Good performer’ mice depolarized during spontaneous unrewarded licking, peaking at around the time when the tongue first contacted the water spout, (△Vm at ± 0.1 s around tongue-spout contact: 3.48 ± 0.62 mV, n = 20). Licking-related depolarization was significantly (p=0.0045) smaller in M1-p neurons of ‘Good performer’ mice (△Vm: 0.83 ± 0.44 mV, n = 10) (Figure 3A, Figure 3—figure supplement 1,3). S2-p neurons of ‘Good performer’ mice also increased firing rate significantly during licking compared to M1-p neurons (p=0.027). Licking-related Vm and AP modulation was weak in ‘Naive’ mice, and it was not significantly different comparing S2-p and M1-p neurons (△Vm, p=0.060; △AP, p=0.30) (Figure 3B, Figure 3—figure supplement 2,3).

Figure 3. Target-specific Vm depolarization during spontaneous unrewarded licking.

(A, B) Left, grand average change in Vm (thick line: mean, thin lines: ± sem) and AP rate aligned at the onset of detected spontaneous licking (dotted line) in M1-p and S2-p neurons of ‘Good performer’ (A) and ‘Naive’ (B) mice. Right, quantifications at ± 0.1 s around the detected lick onset.

DOI: http://dx.doi.org/10.7554/eLife.15798.016

Figure 3—source data 1. Data values and statistics underlying Figure 3.
DOI: http://dx.doi.org/10.7554/eLife.15798.017
elife-15798-fig3-data1.xlsx (12.4KB, xlsx)
DOI: 10.7554/eLife.15798.017
Figure 3—source data 2. Data values and statistics underlying Figure 3—figure supplement 3.
DOI: http://dx.doi.org/10.7554/eLife.15798.018
elife-15798-fig3-data2.xlsx (15.3KB, xlsx)
DOI: 10.7554/eLife.15798.018

Figure 3.

Figure 3—figure supplement 1. Licking-related Vm dynamics of S2-p and M1-p neurons in ‘Good performer’ mice.

Figure 3—figure supplement 1.

(A) Left, an example Vm trace (below) from an S2-p neuron during spontaneous unrewarded licking and the corresponding lick-sensor signal (above). Right, individual lick-sensor signals (above) and Vm traces (below, thin lines; APs are truncated) aligned and superimposed at the onset of detected licking (the time when the tongue first contacted the water spout; dotted line), obtained from the cell shown on the left. The average subthreshold Vm trace (thick line) is superimposed. (B) Another example from an S2-p neuron different from the cell shown in A. (C) Same as A, but for an M1-p neuron. (D) Another example from an M1-p neuron different from the cell shown in C.
DOI: http://dx.doi.org/10.7554/eLife.15798.019
Figure 3—figure supplement 2. Licking-related Vm dynamics of S2-p and M1-p neurons in ‘Naive’ mice.

Figure 3—figure supplement 2.

(A) Left, an example Vm trace (below) from an S2-p neuron during spontaneous unrewarded licking and the corresponding lick-sensor signal (above). Right, individual lick-sensor signals (above) and Vm traces (below, thin lines; APs are truncated) superimposed and aligned at the onset of detected licking (the time when the tongue first contacted the water spout; dotted line), obtained from the cell shown on the left. The average subthreshold Vm trace (thick line) is superimposed. (B) Another example from an S2-p neuron different from the cell shown in A. (C) Same as A, but for an M1-p neuron. (D) Another example from an M1-p neuron different from the cell shown in C.
DOI: http://dx.doi.org/10.7554/eLife.15798.020
Figure 3—figure supplement 3. Further analysis of Vm dynamics during spontaneous unrewarded licking.

Figure 3—figure supplement 3.

(A, B) Left, grand average change in Vm (thick line: mean, thin lines: ± sem) and AP rate aligned at the onset of detected spontaneous licking (dotted line) in S2-p (A) and M1-p (B) neurons of ‘Good performer’ and ‘Naive’ mice. Right, quantifications at ± 0.1 s around the detected lick onset. (C) The lick-related change of Vm and AP rate was analyzed on a mouse-by-mouse basis for ‘Good performer’ mice (n = 7 mice for M1-p neurons; n = 15 mice for S2-p neurons). (D) Same as C, but for ‘Naive’ mice (n = 7 mice for M1-p neurons; n = 9 mice for S2-p neurons).
DOI: http://dx.doi.org/10.7554/eLife.15798.021
Open in a new tab

S2-p neurons, but not M1-p neurons, in ‘Good performer’ mice are therefore excited during spontaneous licking, whereas in ‘Naive’ mice there was little spontaneous licking-related activity in S2-p or M1-p neurons. The licking-related depolarization in S2-p neurons was significantly larger in ‘Good performer’ mice compared to that in ‘Naive’ mice (Figure 3—figure supplement 3), suggesting emergence of projection-specific excitation related to licking after task learning.

Discussion

Our projection-specific Vm measurements in mice with different levels of task proficiency suggest that cortico-cortical signals originating from S1 are bi-directionally modulated by task learning in a pathway-specific manner (Figure 4). In ‘Naive’ mice, whisker stimulation evoked the strongest signals in M1-p neurons during hit trials, whereas in ‘Good performer’ mice S2-p neurons showed the strongest excitation during hit trials. The largest differences in activity during task performance between S2-p and M1-p neurons were observed during the lick period, and task learning was accompanied by enhanced excitation during spontaneous licking specifically in S2-p neurons.

Figure 4. Schematic summary of target-specific output signals from S1 during execution of the whisker detection task, together with speculative hypothesis relating to possible cortico-cortical signalling pathways from S1 to tongue/jaw area of motor cortex.

Figure 4.

Open in a new tab

Left, In ‘Good performer’ mice, S2-p neurons, not M1-p neurons, in S1 develop depolarization correlated with task performance. The activities of S2-p neurons could be routed toward the tongue/jaw area of M1/M2 (tjM1/M2). Right, In ‘Naive’ mice, M1-p neurons, not S2-p neurons, exhibit depolarization correlated with task execution. wM1: whisker M1.

DOI: http://dx.doi.org/10.7554/eLife.15798.022

The lack of task-correlated activity in M1-p neurons in ‘Good performer’ mice is consistent with results from a previous study of a closely-related whisker detection task in which inactivation of whisker M1 did not reduce hit rates in trained animals, but rather increased false-alarm rates (Zagha et al., 2015). Thus signals from S1 to whisker M1 may not be essential for task execution.

Optogenetic inactivation of S1 during the late phase impairs task performance (Sachidhanandam et al., 2013), suggesting a causal role for late excitation. S2-p neurons exhibited a learning-induced depolarization at the late and lick phases of hit trials and during spontaneous unrewarded licking. The grand-averaged, late depolarization in S2-p neurons on hit trials peaked at 261 ms after whisker stimulation (Figure 1D), which was earlier than the mean reaction time (317 ms). The licking-related depolarization in S2-p neurons started shortly (260 ± 46 ms, n = 18) before tongue-spout contact during spontaneous licking (Figure 3A), which is consistent with the larger depolarization at early and late phases of their responses in hit compared to miss trials (Figure 2A). Interestingly, S2 has been suggested to be reciprocally connected to a tongue/jaw-related M1/M2 area (also termed anterior lateral motor cortex, ALM) a neocortical region known to be involved in goal-directed licking (Oh et al., 2014; Guo et al., 2014; Li et al., 2015). We therefore speculate that the licking-related signals in S2-p neurons in S1 might contribute to exciting neurons in tongue/jaw-related M1/M2 via S2 through reciprocally connected networks of excitatory long-range projection neurons, thus contributing to driving licking motor output (Figure 4). Consistent with such a hypothesis involving reciprocal excitation between S1 and S2, axons from S2 innervating S1 were found to exhibit strong task-related hit vs miss modulation in a closely-related whisker detection task (Yang et al., 2016).

It is important to note that there are many possible sensory/motor signals that might contribute to the lick-related depolarization of S2-p neurons in trained mice: movement of jaw and tongue must begin before tongue-spout contact, and we did not quantify orofacial movements during task performance. Furthermore, rodents are known to have reward-expecting orofacial movements such as sniffing and whisking (Deschênes et al., 2012). However, transection of the facial motor nerve that controls whisker movements has no impact on task performance or the late phase Vm (Sachidhanandam et al., 2013), suggesting that the late phase Vm might be generated by internal brain circuits rather than sensory reafference coming from associated whisker movements. In this study we primarily compare neurons projecting to different targets in mice with the same level of task proficiency (i.e. S2-p vs M1-p in ‘Good performer’ mice, or S2-p vs M1-p in ‘Naive’ mice), and the differences found comparing these projection neurons can therefore not reflect differences in sensorimotor behavior. In future experiments, it will be important to examine causal roles of S2-p neurons, as well as to investigate the synaptic mechanisms driving the target-specific Vm dynamics in M1-p and S2-p neurons associated with task learning.

Materials and methods

All experimental procedures were approved by the Swiss Federal Veterinary Office.

Animal preparation

Implantation of a metal head-restraint post on male C57BL6J mice (6-week-old or older), identification of the locations of the S1-C2 barrel column and whisker-S2 of the left hemisphere by intrinsic optical signal imaging, and the injection of CTB conjugated with Alexa-Fluor 488 or 594 (0.5% in PBS, weight/volume, Invitrogen) into left whisker M1 (1 mm anterior, 1 mm lateral from Bregma) and left S2 were performed as previously described (Yamashita et al., 2013). The injection volume of CTB was 50 – 100 nl for M1 and 25 – 50 nl for S2 at the depths of 300 and 800 µm, giving a total volume of 100 – 200 nl for M1 and 50 –100 nl for S2. Animals were kept with a light/dark cycle (12 hr/12 hr) in cages of four mice or less. Experiments were typically performed during the dark period.

Detection task training

At least one day after CTB injection, mice started to be water-restricted. The mice were adapted to head restraint on the recording setup through initial training to freely lick the water spout for receiving water reward (3 – 5 sessions, one session per day). Mice were then taught to associate whisker deflection with water availability through daily training sessions, essentially as described previously (Sachidhanandam et al., 2013; Sippy et al., 2015). For whisker stimulus, we used a brief (1 ms) magnetic pulse to elicit a vertical deflection of the right C2 whisker transmitted by a small metal particle glued on the whisker. The reward time window was 1 s after the whisker stimulus throughout training. Trials with whisker stimulation (test trials) or those without whisker stimulation (catch trials) were started without preceding cues at random inter-trial intervals ranging from 2 – 10 s. Catch trials were randomly interleaved with test trials, with 40 – 50% probability of all trials. If the mouse licked in the 2 s (or 3 s in some experiments) preceding the time when the trial was supposed to occur, then the trial was aborted. Catch trials were present from the first day of training. After each training session, 1.0 – 1.5 g of wet food pellet was given to the mouse in order to keep its body weight more than 80% of the initial value. Behavioral control and behavioral data collection were carried out with custom-written computer routines using a National Instruments board interfaced through LabView.

Whole-cell recordings in task-performing mice

Whole-cell patch-clamp recordings (95 recordings in total) were targeted to cell bodies of CTB-labeled neurons in the center of the C2 barrel column (as identified with intrinsic optical signal imaging) of adult C57BL6J mice (8-week-old or older) under visual control using a custom-built two-photon microscope, as previously described (Yamashita et al., 2013). Recordings were made at the subpial depth of 120 – 270 µm, and the recording depths for M1-p and S2-p neurons were similar. The recording pipettes had resistances of 5 – 7 MΩ and were filled with a solution containing (in mM): 135 potassium gluconate, 4 KCl, 10 HEPES, 10 sodium phosphocreatine, 4 MgATP, 0.3 Na3GTP (adjusted to pH 7.3 with KOH). For targeting CTB-labeled neurons, Alexa 488 or 594 (1 – 20 μM) was added to the pipette solution, depending on the color of the targeted cells. In most experiments, we targeted either M1-p or S2-p neurons. In one mouse, we injected CTB-Alexa 488 in M1 and CTB-Alexa 594 in S2 and targeted both M1-p and S2-p neurons. The Vm was measured using a MultiClamp 700B amplifier with a 10 kHz low pass Bessel filter, and digitized at 20 kHz by a National Instruments board. Vm was not corrected for liquid junction potential.

Short (1 min) sweeps of the Vm and the behavioral signals from the lick sensor together with TTL signals to control the water valve and the electromagnetic coil were recorded using Ephus in Matlab (Suter et al., 2010). We used two types of mice for recordings: (1) ‘Good performer’ mice that exhibited a high discriminability between test trials and catch trials during recordings (59 recordings in 27 mice; hit rate, 0.77 ± 0.03; false alarm rate, 0.17 ± 0.01; d’ = 2.12 ± 0.09) learned through training sessions (typically 7–13 daily sessions prior to the recording day, but some mice learned more quickly); and (2) ‘Naive’ mice that were used for recordings on the first day of being exposed to the task and showed no apparent discrimination (36 recordings in 16 mice; hit rate, 0.31 ± 0.03; false alarm rate, 0.28 ± 0.03; d’ = 0.03 ± 0.09; d’ < 0.9, for each recording). For calculating d’ when hit rate or false alarm rate was measured as 1.0 or 0.0, each value was corrected by subtracting or adding 1/(2N), where N is the trial number. Each recording typically lasted ~20 min or less, and we made multiple whole-cell recordings from one animal in most of the experiments. For each recording we routinely monitored the level of task performance by calculating d’ and discarded data with d’ < 1.1 in ‘Good performer’ mice or those with d’ > 0.9 in ‘Naive’ mice. The d’ values for recordings of M1-p and S2-p neurons was not significantly different (p=0.73; for ‘Good performer’ mice; p=0.50 for ‘Naive’ mice).

Data analysis

Subthreshold postsynaptic potentials (PSPs) were analyzed after removing APs by median-filtering (Crochet and Petersen, 2006). For analysis of Vm changes evoked by task-relevant whisker deflection, baseline Vm was defined as the mean Vm at 0 – 5 ms before the stimulus onset. The amplitude of PSPs was defined as the difference between the baseline Vm and the peak Vm of averaged traces. The △Vm at the late and lick periods was estimated as the difference between the baseline Vm and the mean Vm of the averaged traces at 0.05 – 0.25 s (late) or 0.25 – 1.0 s (lick) after whisker stimulus. APs evoked by whisker stimulation were estimated by subtracting spontaneous AP rate from the AP rate measured in the early (0 – 0.05 s), late (0.05 – 0.25 s) or lick (0.25 –1.0 s) periods after the stimulation for each cell. Baseline AP rates were computed as the mean of no-lick periods (2 s before test/catch trials) totaling over 16 s. Peristimulus time histograms (PSTHs) were computed by counting AP number in each 50 ms (or 10 ms) bin for each cell and averaging the number across cells recorded. Grand average PSTHs are shown in Hz after subtracting baseline AP rates. On average, 31 ± 2 hit trials (n = 53 cells) and 18 ± 2 miss trials (n = 29 cells) per recording were analyzed for ‘Good performer’ mice, and 14 ± 1 hit trials (n = 26 cells) and 31 ± 3 miss trials (n = 26 cells) per recording were analyzed for ‘Naive’ mice. In some recordings the well-trained mouse showed few misses and in such cases we only analyzed hit responses.

Lick bouts that occurred at least 3 s after whisker stimulation, and at least 1 s after the cessation of previous lick bouts were selected for analysis of Vm modulation induced by spontaneous unrewarded licking. On average, 64 ± 6 lick bouts (n = 30 cells) of ‘Good performer’ mice and 31 ± 3 lick bouts (n = 26 cells) of ‘Naive’ mice were analyzed for each recording. The individual Vm traces aligned at the onset of detected lick bouts (lick onset) were median-filtered to remove APs. Baseline Vm was defined as the mean Vm at 1.0 – 0.6 s before the lick onset, and the magnitude of Vm modulation was estimated by the difference between the baseline Vm and the mean Vm at ± 0.1 s around the lick onset. APs evoked during lick events were calculated by subtracting baseline AP rate (averaged at 0.6 – 1.0 s before the detected lick onset) from the AP rate measured within ± 0.1 s from the lick onset. PSTHs around lick events are shown in Hz after subtracting the baseline AP rate. The onset of the licking-related Vm depolarization was computed as the time point where Vm increased over 3 x SD of the baseline Vm for the 18 out of 20 S2-p cells with pre-lick depolarization.

All values (except for box plots) are presented as mean ± sem. Box plots indicate median and 1st/3rd quartile, with Tukey’s whiskers showing maximal and minimal data points within 1.5 times interquartile range away from 1st/3rd quartile. Statistical testing using two-tailed Wilcoxon rank-sum test for unpaired data (for example, M1-p vs S2-p for Figure 1D,E and Figure 3A,B; ‘Good performer’ vs ‘Naive’ for Figure 1C) and two-tailed Wilcoxon signed rank test for paired data (for example, hit vs miss trials for Figure 2; hit vs false-alarm rates for Figure 1C) was performed in IgorPro (WaveMetrics) without excluding any data points. Testing for the normality of data distribution was performed in IgorPro and we found that at least one of the samples in every two-sample comparison was not normally distributed. Non-parametric tests were therefore used for all figures. We analyzed data on a cell-by-cell basis unless otherwise noted. Neither randomization nor blinding was done for data collection or analysis.

Acknowledgements

We thank Taro Kiritani and Varun Sreenivasan for help with building the experimental setup, and Eloïse Charrière for help with mouse training. This work was supported by grants from the Swiss National Science Foundation and the European Research Council.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Funding Information

This paper was supported by the following grants:

  • Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung 310030_146252 to Carl CH Petersen.

  • European Research Council 293660 to Carl CH Petersen.

Additional information

Competing interests

The authors declare that no competing interests exist.

Author contributions

TY, Designed the project, Wrote the manuscript, Performed experiments, Analyzed data, Conceptionand design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

CCHP, Designed the project, Wrote the manuscript, Conception and design, Analysis and interpretation of data, Drafting or revising the article.

Ethics

Animal experimentation: All experimental procedures were approved by the Swiss Federal Veterinary Office, authorisation VD1628.

References

  1. Aronoff R, Matyas F, Mateo C, Ciron C, Schneider B, Petersen CCH. Long-range connectivity of mouse primary somatosensory barrel cortex. The European Journal of Neuroscience. 2010;31:2221–2233. doi: 10.1111/j.1460-9568.2010.07264.x. [DOI] [PubMed] [Google Scholar]
  2. Chen JL, Carta S, Soldado-Magraner J, Schneider BL, Helmchen F. Behaviour-dependent recruitment of long-range projection neurons in somatosensory cortex. Nature. 2013;499:336–340. doi: 10.1038/nature12236. [DOI] [PubMed] [Google Scholar]
  3. Chen JL, Margolis DJ, Stankov A, Sumanovski LT, Schneider BL, Helmchen F. Pathway-specific reorganization of projection neurons in somatosensory cortex during learning. Nature Neuroscience. 2015;18:1101–1108. doi: 10.1038/nn.4046. [DOI] [PubMed] [Google Scholar]
  4. Crochet S, Petersen CCH. Correlating whisker behavior with membrane potential in barrel cortex of awake mice. Nature Neuroscience. 2006;9:608–610. doi: 10.1038/nn1690. [DOI] [PubMed] [Google Scholar]
  5. Deschênes M, Moore J, Kleinfeld D. Sniffing and whisking in rodents. Current Opinion in Neurobiology. 2012;22:243–250. doi: 10.1016/j.conb.2011.11.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Gilbert CD, Li W. Top-down influences on visual processing. Nature Reviews Neuroscience. 2013;14:350–363. doi: 10.1038/nrn3476. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Glickfeld LL, Andermann ML, Bonin V, Reid RC. Cortico-cortical projections in mouse visual cortex are functionally target specific. Nature Neuroscience. 2013;16:219–226. doi: 10.1038/nn.3300. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Guo ZV, Li N, Huber D, Ophir E, Gutnisky D, Ting JT, Feng G, Svoboda K. Flow of cortical activity underlying a tactile decision in mice. Neuron. 2014;81:179–194. doi: 10.1016/j.neuron.2013.10.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Harris KD, Mrsic-Flogel TD. Cortical connectivity and sensory coding. Nature. 2013;503:51–58. doi: 10.1038/nature12654. [DOI] [PubMed] [Google Scholar]
  10. Li N, Chen TW, Guo ZV, Gerfen CR, Svoboda K. A motor cortex circuit for motor planning and movement. Nature. 2015;519:51–56. doi: 10.1038/nature14178. [DOI] [PubMed] [Google Scholar]
  11. Movshon JA, Newsome WT. Visual response properties of striate cortical neurons projecting to area MT in macaque monkeys. The Journal of Neuroscience. 1996;16:7733–7741. doi: 10.1523/JNEUROSCI.16-23-07733.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Oh SW, Harris JA, Ng L, Winslow B, Cain N, Mihalas S, Wang Q, Lau C, Kuan L, Henry AM, Mortrud MT, Ouellette B, Nguyen TN, Sorensen SA, Slaughterbeck CR, Wakeman W, Li Y, Feng D, Ho A, Nicholas E, Hirokawa KE, Bohn P, Joines KM, Peng H, Hawrylycz MJ, Phillips JW, Hohmann JG, Wohnoutka P, Gerfen CR, Koch C, Bernard A, Dang C, Jones AR, Zeng H. A mesoscale connectome of the mouse brain. Nature. 2014;508:207–214. doi: 10.1038/nature13186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Sachidhanandam S, Sreenivasan V, Kyriakatos A, Kremer Y, Petersen CCH. Membrane potential correlates of sensory perception in mouse barrel cortex. Nature Neuroscience. 2013;16:1671–1677. doi: 10.1038/nn.3532. [DOI] [PubMed] [Google Scholar]
  14. Sato TR, Svoboda K. The functional properties of barrel cortex neurons projecting to the primary motor cortex. The Journal of Neuroscience. 2010;30:4256–4260. doi: 10.1523/JNEUROSCI.3774-09.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Sippy T, Lapray D, Crochet S, Petersen CCH. Cell-type-specific sensorimotor processing in striatal projection neurons during goal-directed behavior. Neuron. 2015;88:298–305. doi: 10.1016/j.neuron.2015.08.039. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Sorensen SA, Bernard A, Menon V, Royall JJ, Glattfelder KJ, Desta T, Hirokawa K, Mortrud M, Miller JA, Zeng H, Hohmann JG, Jones AR, Lein ES. Correlated gene expression and target specificity demonstrate excitatory projection neuron diversity. Cerebral Cortex. 2015;25:433–449. doi: 10.1093/cercor/bht243. [DOI] [PubMed] [Google Scholar]
  17. Suter BA, O'Connor T, Iyer V, Petreanu LT, Hooks BM, Kiritani T, Svoboda K, Shepherd GM. Ephus: multipurpose data acquisition software for neuroscience experiments. Frontiers in Neural Circuits. 2010;4:e15798. doi: 10.3389/fncir.2010.00100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Yamashita T, Pala A, Pedrido L, Kremer Y, Welker E, Petersen CCH. Membrane potential dynamics of neocortical projection neurons driving target-specific signals. Neuron. 2013;80:1477–1490. doi: 10.1016/j.neuron.2013.10.059. [DOI] [PubMed] [Google Scholar]
  19. Yang H, Kwon SE, Severson KS, O'Connor DH. Origins of choice-related activity in mouse somatosensory cortex. Nature Neuroscience. 2016;19:127–134. doi: 10.1038/nn.4183. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Zagha E, Ge X, McCormick DA. Competing neural ensembles in motor cortex gate goal-directed motor output. Neuron. 2015;88:565–577. doi: 10.1016/j.neuron.2015.09.044. [DOI] [PMC free article] [PubMed] [Google Scholar]
eLife. 2016 Jun 21;5:e15798. doi: 10.7554/eLife.15798.023

Decision letter

Editor: Naoshige Uchida1

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Target-specific membrane potential dynamics of neocortical projection neurons during goal-directed behavior" for consideration by eLife. Your article has been favorably evaluated by Timothy Behrens (Senior editor) and three reviewers, one of whom, Naoshige Uchida, is a member of our Board of Reviewing Editors.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The authors performed whole-cell recording from neurons in the primary somatosensory cortex (S1) while mice performed a whisker deflection detection task. The main finding is that S2-projecting, but not M1-projecting, neurons show long-lasting depolarization at the late phase (0.05 – 0.25s) after whisker stimulation. Furthermore, this difference was observed only in well-trained mice but not in naïve mice. Overall, all the reviewers thought that the data are highly valuable, and that the findings are of great interest. However, it was pointed out that there are some interpretational difficulties. Also, some statistical analyses require more explicit justifications and explanations.

Essential revisions:

1) It was discussed that "the licking-related signals in S2-p neurons in S1 might contribute to exciting neurons in tongue/jaw-related M1 via S2 through reciprocally connected networks of excitatory long-range projection neurons, thus contributing to driving licking motor output (Figure 4)". While this is an interesting possibility, alternative possibilities are not considered. What if the learning-related changes in membrane potential dynamics are secondary to changes in behavior? For instance, in addition to expected changes in licking behavior, other orofacial movements and whisking might change during learning. More importantly, sensation associated with these behaviors may change during learning and cause alterations in membrane potential dynamics. If the altered membrane potential dynamics are caused by these secondary changes, the interrelation will be very different. Ideally, these issues should be addressed by additional experiments or measurements. For instance, the authors might record from S2-projecting neurons while paralyzing the whisker pad and removing whisker movement (or whisker pad movement). We understand that this experiment may take more than two months, and you might decide not to do these experiments. If you choose not to perform these experiments, these alternative interpretations must be discussed in the revised manuscript very carefully and in a very visible and balanced way.

2) Animal-to-animal differences in behavior/learning may cause biases in neuronal responses. The authors need to make a convincing argument that neurons from the same mice can be treated independently. More details need to be provided on the statistical analysis and how these confounding factors may have been dealt with.

Reviewer #1:

In the manuscript entitled "Target-specific membrane potential dynamics of neocortical projection neurons during goal-directed behavior", Yamashita and Petersen performed whole-cell recording from a large number of neurons (n = 95 neurons from at least 27 mice) in the primary somatosensory cortex (S1) while mice performed a whisker deflection detection task. Notably, whole cell recording was targeted to neurons retrogradely labeled from their projection target (secondary somatosensory cortex [S2] and primary motor cortex). Furthermore, the authors also compared the data from well-trained (after 7-13 days of training) versus naïve (the first day of training) mice.

The authors found that S2-projecting, but not M1-projecting, neurons show long-lasting depolarization at the late phase (0.05 – 0.25s) after whisker stimulation. Furthermore, this difference was observed only in well-trained mice but not in naïve mice. The authors also found that S2-projecting neurons in well-trained mice show depolarization immediately before spontaneous, unrewarded licking. In naïve mice, M1-projecting neurons in naïve mice showed stronger response than S2-projecting neurons. The authors discuss that S2-projecting S1 neurons may be responsible in generating licking response in well-trained mice, and the activation of S2-projecting neurons before licking may be due to the reciprocal connection between S2 and the tongue/jaw-related area in M1.

The authors collected a large data set and the conclusions are convincing. The manuscript is written clearly and the conclusions are very interesting. I have a few relatively minor comments.

1) The definition of "good performer" and "naïve" mice is done only in the Materials and methods section. It would be very helpful if the authors explain this in the main text when they appear.

2) The histogram in Figure 1D indicates that S2-projecting neurons show on average a higher rate of firing during the late response phase than M1-projecting neurons. However, the box plot does not show any difference. First, is there any difference in the data between these plots? Second, this result may suggest that action potentials sent to the postsynaptic neurons are not different between the two populations. Is it possible that whole cell recording artificially reduced the firing rate responses in recorded neurons?

3) Are there S1 neurons that project to both S2 and M1? How much would this affect the results?

Reviewer #2:

This paper is part of a substantial body of work from the Petersen lab, investigating L2/3 networks in the barrel cortex in the context of tactile passive detection tasks. Yamashita and Petersen (2013) previously reported that S1 neurons projecting to primarymotor cortex (M1) and those projecting to secondary somatosensory cortex (S2) have distinct intrinsic membrane properties and exhibit different membrane potential dynamics during spontaneous behavior. Here, by making whole-cell recordings in primary somatosensory, they compare these two projection classes in the context of a simple tactile detection behavior. They also compare behavior-related activity in the projection classes in naive and performing mice.

They make interesting observations. In expert mice, S2-p neurons show larger slow depolarizations after whisker deflection, and also depolarize with licking. In contrast, in naive mice, M1-p neurons show larger deflections. So there is a change in the response properties with learning, suggesting changing routing of signals with learning. This is hard-won data. Whole cell recordings are rarely applied to behaving mice, much less in a projection-specific manner. The experimental approach is pioneering. However, there are some weaknesses that make the data difficult to interpret. It's not clear to me is there is a change in routing with learning; the mouse might instead employ distinct sensorimotor strategies, which recruit the different projection classes in a different manner. It's nice to see differences in coding across projection classes, but the implications beyond this interesting observation are unclear.

1) What might the learning-related changes mean? The paper is silent on any interpretation. Is the idea that some stimulus-response association is coded in the network (whisker – S2p – tongue M1)? Does Figure 4 suggest that functional connectivity changes to recruit different loops? If so how? How about the alternative explanation that S2-p neurons report orofacial sensation related to movement (which differs for different training conditions). As is, it seems that the behavioral design does not allow separating these and other interpretations.

2) In the 2013 paper differences between M1p and S2p neurons are dramatic. Here the differences are subtle. In particular, the data in Figure 2 (trial-trial analysis) appears marginal. This is somewhat worrisome given some uncertainties with stats.

3) Related: It's not clear to me if the stats were done properly. They need to convince me that neurons from the same mice can be treated independently. This is critical here because behavior was not monitored (for example, whisker movement) and there likely are strong correlations across neurons of one animal, imposed by idiosyncratic behavior. In addition, it's not clear that S2p and M1p projecting neurons were recorded in the same proportions in each animal etc. A lot more details needs to be provided on the statistical analysis and how these confounding factors may have been dealt with.

4) Finally, it's a bit surprising that whisker movements, details of tongue movements, and learning-related changes thereof, where not monitored. How do we know that behavior (learning; hit vs. miss) are not associated with different movements and sensory input related to movement? Clearly the licking will be very different in expert mice compared to naive mice. Similar for other orofacial movements.

5) What is the implication of the data of Figure 3. Is this a sensory or motor signal? Does activity precede the earliest signs of movement?

Other points:

Why does the 'lick period' extend long after the all-important first lick?

Reviewer #3:

This is a brief paper from Yamashita and Petersen on the change in intracellular response on neurons in vS1 that project to vS2 versus vM1. The authors find differences – a greater change in vS2 versus vM1 projecting cells – which is of importance in deciphering the rules of neuronal plasticity in cortex and in explaining the variance observed in prior experiments with functionally unlabeled cells. This submission extends earlier work from the Petersen laboratory (Sachidhanandam et al. Nature Neuroscience 2013 and Yamashita et al. Neuron 2013) as well as the Helmchen laboratory (Chen et al. Nature 2013, Chen et al. Nature Neuroscience 2015). The conclusions of the present work are similar to those in Chen 2015. With no slight toward the Chen 2015 paper, I find the present work more compelling as it is based on intracellular measurements, which reveals significant structure about the nature of the response, as opposed to calcium-dynamics in Chen 2015, which is largely a means to report bursts of spikes.

The authors present a wealth of intracellular data on the response to deflection of a vibrissa in trained versus naive animals. There are a number of important differences in the records in the data that are unreported, so I would ask the authors to revise the Results section to address the following.

1) With respect to Figure 1 – the effect of "good performer" versus "naive" appears to be mainly a substantial increase in the depolarization of S2-projecting cells as opposed to a change in M1-projecting cells, whose response is largely unchanged (compare parts D and E). The text describes only the difference in the relative depolarization; I think this statement can be strengthened.

2) With respect to the depolarization for the "good performers", there is a significant delayed peak at about 250 ms into the trial. This is never discussed, but should be; it seems like a long time for a recurrent signal.

In a real sense, this paper and the Chen 2015 complement each other. The large calcium imaging set in Chen 2015 allow those authors to delineate an increase in discrimination by S2 projecting neurons (Chen 2015 Figure 5D and 6), while the present work shows the nature of the change in subthreshold response.

eLife. 2016 Jun 21;5:e15798. doi: 10.7554/eLife.15798.024

Author response

Essential revisions:

1) It was discussed that "the licking-related signals in S2-p neurons in S1 might contribute to exciting neurons in tongue/jaw-related M1 via S2 through reciprocally connected networks of excitatory long-range projection neurons, thus contributing to driving licking motor output (Figure 4)". While this is an interesting possibility, alternative possibilities are not considered. What if the learning-related changes in membrane potential dynamics are secondary to changes in behavior? For instance, in addition to expected changes in licking behavior, other orofacial movements and whisking might change during learning. More importantly, sensation associated with these behaviors may change during learning and cause alterations in membrane potential dynamics. If the altered membrane potential dynamics are caused by these secondary changes, the interrelation will be very different. Ideally, these issues should be addressed by additional experiments or measurements. For instance, the authors might record from S2-projecting neurons while paralyzing the whisker pad and removing whisker movement (or whisker pad movement). We understand that this experiment may take more than two months, and you might decide not to do these experiments. If you choose not to perform these experiments, these alternative interpretations must be discussed in the revised manuscript very carefully and in a very visible and balanced way.

This is an important point. The “licking-related” signal in S2-p neurons of Good Performer mice could have various origins including both sensory and motor signals. In a previous study of the same detection task, we carried out facial motor nerve section, and this had no impact on licking behavior or the late phase Vm (Sachidhanandam et al., 2013). Furthermore, in the same study, we also ascribed a causal role of the late activity in S1 for our detection task, through temporally-precise optogenetic inactivation during the late phase (Sachidhanandam et al., 2013). We therefore think that it is not inappropriate to speculate that the licking-related signal in S2-p neurons might contribute to task performance by enhancing the lick probability.

However, we cannot rule out the possibility that orofacial sensorimotor activity contributes to the licking-related signal in S2-p neurons, and indeed it is possible that some orofacial sensory-reafference signals are important for task performance, and for generating the licking-related signal in S2-p neurons. In the Discussion section we now write:

“Optogenetic inactivation of S1 during the late phase impairs task performance (Sachidhanandam et al., 2013), suggesting a causal role for late excitation. […] In future experiments, it will be important to examine causal roles of S2-p neurons, as well as to investigate the synaptic mechanisms driving the target-specific Vm dynamics in M1-p and S2-p neurons associated with task learning.”

2) Animal-to-animal differences in behavior/learning may cause biases in neuronal responses. The authors need to make a convincing argument that neurons from the same mice can be treated independently. More details need to be provided on the statistical analysis and how these confounding factors may have been dealt with.

This is an interesting point, and the reviewer is correct to point out that animal-to-animal differences could contribute to some aspects of our dataset.

We therefore pooled data of multiple cells for each mouse and analysed the Vm and APs during task performance for S2-p and M1-p neurons. Our main conclusions that S2-projecting, but not M1-projecting, neurons in trained mice show long-lasting depolarization at the late phase after whisker stimulation (Figure 1—figure supplement 3), that hit trials are more depolarized than miss trials during late and licking periods for S2-p neurons (Figure 2—figure supplement 2), and licking-related depolarization (Figure 3—figure supplement 3) remain prominent and statistically significant even when analysed on a mouse-by-mouse basis. Our results are therefore robust.

For main figures, we would like to keep analysis on a cell-by-cell basis. The reasons for this are as follows: (1) In our data set, there is no clear evidence that there is stronger correlation across neurons of one animal over correlation across animals; (2) The number of cells we recorded in one animal is small (2 cells on average), so animal-to-animal differences, if any, could not effectively affect results; (3) The cell-to-cell variability of membrane potential properties such as spike rates is high in these projection neurons (Yamashita et al., Neuron, 2013), presumably overwhelming animal-to-animal variability in our experimental condition; (4) As far as we know all published papers using similar techniques (for example, Cohen et al., Nature 2012; Haider et al., Nature 2013; Polack et al., Nature Neuroscience, 2013; Zagha et al., Neuron 2015; Yang et al., Nature Neuroscience, 2016) also treat cells independently.

We also added more detailed information on how we collected and analyzed data in the Materials and methods section of the revised manuscript.

Reviewer #1:

[…] The authors collected a large data set and the conclusions are convincing. The manuscript is written clearly and the conclusions are very interesting. I have a few relatively minor comments.

1) The definition of "good performer" and "naïve" mice is done only in the Materials and methods section. It would be very helpful if the authors explain this in the main text when they appear.

We have now added statements at the beginning of the Results section to explain the definition of "good performer" and "naive" mice. We write: “We used two types of mice for recordings: (1) ‘Good performer’ mice that exhibited a high discriminability between test trials and catch trials (for details see Materials and methods) during recordings (59 recordings in 27 mice; hit rate, 0.77 ± 0.03; false alarm rate, 0.17 ± 0.01; d’ = 2.12 ± 0.09; d’ > 1.1 for each recording; Figure 1C) learned through training sessions (typically 7–13 daily sessions prior to the recording day, but some mice learned more quickly), responding with a reaction time of 317 ± 17 ms (time from whisker deflection to tongue contact with the water spout) (Figure 1D); and (2) ‘Naive’ mice that were used for recordings on the first day of being exposed to the task and showed no apparent discrimination (36 recordings in 16 mice; hit rate, 0.31 ± 0.03; false alarm rate, 0.28 ± 0.03; d’ = 0.03 ± 0.09; d’ < 0.9, for each recording; Figure 1C), with a mean reaction time of 369 ± 23 ms which was significantly slower than ‘Good performer’ mice (p = 0.0014; Figure 1E).”

2) The histogram in Figure 1D indicates that S2-projecting neurons show on average a higher rate of firing during the late response phase than M1-projecting neurons. However, the box plot does not show any difference. First, is there any difference in the data between these plots? Second, this result may suggest that action potentials sent to the postsynaptic neurons are not different between the two populations. Is it possible that whole cell recording artificially reduced the firing rate responses in recorded neurons?

The grand average Vm and PSTH (Figure 1D left) and the box plots (Figure 1D right) show different analyses of the same data. For the box plots we divide the post-stimulus timeline into three epochs: early (0 – 50 ms), late (50 – 250 ms) and licking (250 – 1000 ms). We found larger Vm depolarization during the late and licking periods, and enhanced AP firing in the licking period for S2-p neurons compared to M1-p neurons. In our view, the box plots are consistent with the PSTH.

So far across our different studies, whole-cell recordings have generated similar data to cell-attached recordings (Gentet et al., 2010; Sachidhanandam et al., 2013).

3) Are there S1 neurons that project to both S2 and M1? How much would this affect the results?

There are very few layer 2/3 S1 neurons double-labeled with retrograde tracers injected into S2 and M1 (0.7% , Yamashita et al., 2013). We are in the process of reconstructing the full axonal arborisations of M1-p and S2-p neurons, and, so far across our small sample of 15 neurons, we have not found any that project to both targets. On P. 3 we now write: “Retrograde labeling suggests that M1-p and S2-p neurons in S1 are largely non-overlapping types of excitatory neurons (Sato and Svoboda, 2010; Chen et al., 2013; Yamashita et al., 2013).”

Reviewer #2:

This paper is part of a substantial body of work from the Petersen lab, investigating L2/3 networks in the barrel cortex in the context of tactile passive detection tasks. Yamashita and Petersen (2013) previously reported that S1 neurons projecting to primarymotor cortex (M1) and those projecting to secondary somatosensory cortex (S2) have distinct intrinsic membrane properties and exhibit different membrane potential dynamics during spontaneous behavior. Here, by making whole-cell recordings in primary somatosensory, they compare these two projection classes in the context of a simple tactile detection behavior. They also compare behavior-related activity in the projection classes in naive and performing mice.

They make interesting observations. In expert mice, S2-p neurons show larger slow depolarizations after whisker deflection, and also depolarize with licking. In contrast, in naive mice, M1-p neurons show larger deflections. So there is a change in the response properties with learning, suggesting changing routing of signals with learning. This is hard-won data. Whole cell recordings are rarely applied to behaving mice, much less in a projection-specific manner. The experimental approach is pioneering. However, there are some weaknesses that make the data difficult to interpret. It's not clear to me is there is a change in routing with learning; the mouse might instead employ distinct sensorimotor strategies, which recruit the different projection classes in a different manner. It's nice to see differences in coding across projection classes, but the implications beyond this interesting observation are unclear.

1) What might the learning-related changes mean? The paper is silent on any interpretation. Is the idea that some stimulus-response association is coded in the network (whisker – S2p – tongue M1)? Does Figure 4 suggest that functional connectivity changes to recruit different loops? If so how? How about the alternative explanation that S2-p neurons report orofacial sensation related to movement (which differs for different training conditions). As is, it seems that the behavioral design does not allow separating these and other interpretations.

We agree with the reviewer that the functional meaning of the learning-induced changes in Vm is unclear. Our results are consistent with the possibility that the licking-related depolarization in S2-p neurons might be routed to tongue/jaw M1 via S2. However, the reviewer is correct to point out that alternative explanations could be possible. We made this point clearer by adding a question mark in Figure 4, and adding discussion: “It is important to note that there are many possible sensory/motor signals that might contribute to the lick-related depolarization of S2-p in trained mice: movement of jaw and tongue must begin before tongue-spout contact, and we did not quantify orofacial movements during task performance. […] In this study we primarily compare neurons projecting to different targets in mice with the same level of task proficiency (i.e. S2-p vs M1-p in ‘Good performer’ mice, or S2-p vs M1-p in ‘Naive’ mice), and the differences found comparing these projection neurons can therefore not reflect differences in sensorimotor behavior.”

2) In the 2013 paper differences between M1p and S2p neurons are dramatic. Here the differences are subtle. In particular, the data in Figure 2 (trial-trial analysis) appears marginal. This is somewhat worrisome given some uncertainties with stats.

In this study we found many robust and prominent differences between M1-p and S2-p neurons. The differences in late and lick Vm (Figure 1D) and spontaneous lick-related Vm changes (Figure 3A) are obvious and have very high statistical significance. The hit/miss differences in Figure 2 are also robust, and are also found when analysed on a mouse-by-mouse manner (see below). Indeed, each of the three main figures shows data with statistical significance P < 0.01.

3) Related: It's not clear to me if the stats were done properly. They need to convince me that neurons from the same mice can be treated independently. This is critical here because behavior was not monitored (for example, whisker movement) and there likely are strong correlations across neurons of one animal, imposed by idiosyncratic behavior. In addition, it's not clear that S2p and M1p projecting neurons were recorded in the same proportions in each animal etc. A lot more details needs to be provided on the statistical analysis and how these confounding factors may have been dealt with.

This is an interesting point, and the reviewer is correct to point out that animal-to-animal differences could contribute to some aspects of our dataset.

We therefore pooled data of multiple cells for each mouse and analysed the Vm and APs during task performance for S2-p and M1-p neurons. Our main conclusions that S2-projecting, but not M1-projecting, neurons in trained mice show long-lasting depolarization at the late phase after whisker stimulation (Figure 1—figure supplement 3), that hit trials are more depolarized than miss trials during late and licking periods for S2-p neurons (Figure 2—figure supplement 2), and licking-related depolarization (Figure 3—figure supplement 3) remain prominent and statistically significant even when analysed on a mouse-by-mouse basis. Our results are therefore robust.

For main figures, we would like to keep analysis on a cell-by-cell basis. The reasons for this are as follows: (1) In our data set, there is no clear evidence that there is stronger correlation across neurons of one animal over correlation across animals; (2) The number of cells we recorded in one animal is small (2 cells on average), so animal-to-animal differences, if any, could not effectively affect results; (3) The cell-to-cell variability of membrane potential properties such as spike rates is high in these projection neurons (Yamashita et al., Neuron, 2013), presumably overwhelming animal-to-animal variability in our experimental condition; (4) As far as we know all published papers using similar techniques (for example, Cohen et al., Nature 2012; Haider et al., Nature 2013; Polack et al., Nature Neuroscience, 2013; Zagha et al., Neuron 2015; Yang et al., Nature Neuroscience, 2016) also treat cells independently.

We also added more detailed information on how we collected and analyzed data in the Materials and methods section of the revised manuscript.

4) Finally, it's a bit surprising that whisker movements, details of tongue movements, and learning-related changes thereof, where not monitored. How do we know that behavior (learning; hit vs. miss) are not associated with different movements and sensory input related to movement? Clearly the licking will be very different in expert mice compared to naive mice. Similar for other orofacial movements.

The “licking-related” signal could have many origins including both sensory and motor signals. We have already carried out facial motor nerve section, and this had no impact on behavior or the late phase Vm (Sachidhanandam et al., 2013). The projection-specific nature of the signal is seen for the same type of mice (Good performer or Naive), and a large part of our story is thus independent of what correlated changes in behavior take place over learning. We added discussion on this point in the Discussion section (last paragraph), see our reply to point 1.

5) What is the implication of the data of Figure 3. Is this a sensory or motor signal? Does activity precede the earliest signs of movement?

The licking-related depolarization in S2p neurons starts slightly before tongue- spout contact. In a previous study we found that optogenetic inactivation of S1 during the late period caused deficits in the same detection task (Sachidhanandam et al., 2013). The licking-related activity of S2-p neurons could therefore contribute as a motor-related signal. However, there could be movements of tongue and other orofacial parts that occur earlier than tongue- spout contact and give rise to sensory-reafference signals. It is therefore not possible to make a definitive statement based on our current data. We added discussion on this point in the revised manuscript (last paragraph), see our reply to point 1.

Other points:

Why does the 'lick period' extend long after the all-important first lick?

The “lick period” covers the licking period to measure licking-related signals. The “late” period on the other hand precedes the first lick.

Reviewer #3:

This is a brief paper from Yamashita and Petersen on the change in intracellular response on neurons in vS1 that project to vS2 versus vM1. The authors find differences – a greater change in vS2 versus vM1 projecting cells – which is of importance in deciphering the rules of neuronal plasticity in cortex and in explaining the variance observed in prior experiments with functionally unlabeled cells. This submission extends earlier work from the Petersen laboratory (Sachidhanandam et al. Nature Neuroscience 2013 and Yamashita et al. Neuron 2013) as well as the Helmchen laboratory (Chen et al. Nature 2013, Chen et al. Nature Neuroscience 2015). The conclusions of the present work are similar to those in Chen 2015. With no slight toward the Chen 2015 paper, I find the present work more compelling as it is based on intracellular measurements, which reveals significant structure about the nature of the response, as opposed to calcium-dynamics in Chen 2015, which is largely a means to report bursts of spikes.

The authors present a wealth of intracellular data on the response to deflection of a vibrissa in trained versus naive animals. There are a number of important differences in the records in the data that are unreported, so I would ask the authors to revise the Results section to address the following.

1) With respect to Figure 1 – the effect of "good performer" versus "naive" appears to be mainly a substantial increase in the depolarization of S2-projecting cells as opposed to a change in M1-projecting cells, whose response is largely unchanged (compare parts D and E). The text describes only the difference in the relative depolarization; I think this statement can be strengthened.

To show the learning-induced changes in membrane potential dynamics of projection neurons, we added new supplementary figures (Figure 1—figure supplement 4 and Figure 3—figure supplement 3) comparing between naive and good-performer mice (i.e., M1p naive vs. M1p experts and S2p naive vs. S2p expert).

We accordingly added statements on this, as suggested (Results):

“Notably, the secondary long-lasting depolarization in S2-p neurons after whisker stimulation was seen only in ‘Good performer’ mice, not in ‘Naive’ mice (Figure 1D, E, Figure 1—figure supplement 4), while the small sustained depolarization of M1-p neurons in ‘Naive’ mice was attenuated in ‘Good performer’ mice (Figure 1D, E, Figure 1—figure supplement 4).”

“The licking-related depolarization in S2-p neurons was significantly larger in ‘Good performer’ mice compared to that in ‘Naive’ mice (Figure 3— figure supplement 3), suggesting emergence of projection-specific excitation related to licking after task learning.”

2) With respect to the depolarization for the "good performers", there is a significant delayed peak at about 250 ms into the trial. This is never discussed, but should be; it seems like a long time for a recurrent signal.

The long-lasting depolarization for “good performers” is quantified across late (50 – 250 ms period after whisker stimulation) and licking (250 – 1000 ms) periods. The peak of the grand average Vm trace is at 261 ms, which precedes licking (mean reaction time was 317 ms). As suggested, we now added discussion on the peak time of the late depolarization:

“S2-p neurons exhibited a learning-induced depolarization at the late and lick phases of hit trials and during spontaneous unrewarded licking. […] We therefore speculate that the licking-related signals in S2-p neurons in S1 might contribute to exciting neurons in tongue/jaw-related M1/M2 via S2 through reciprocally connected networks of excitatory long-range projection neurons, thus contributing to driving licking motor output (Figure 4).”

The timescale is indeed long, but even simple reciprocal long-range feedback interactions can take place on long time-scales. For example, in the study of Manita et al. (2015) they find ~100 ms in an S1->M2->S1 loop. We think it is likely that there might be more complex interactions across different areas in our behavior, perhaps involving both S1, S2 and tjM1, which could give rise to even longer timescales. The mechanisms driving slow, behaviorally relevant membrane potential depolarization are indeed of enormous interest to investigate in the future.

In a real sense, this paper and the Chen 2015 complement each other. The large calcium imaging set in Chen 2015 allow those authors to delineate an increase in discrimination by S2 projecting neurons (Chen 2015 Figure 5D and 6), while the present work shows the nature of the change in subthreshold response.

We agree with the reviewer.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Figure 1—source data 1. Data values and statistics underlying Figure 1.

    DOI: http://dx.doi.org/10.7554/eLife.15798.004

    elife-15798-fig1-data1.xlsx (27.3KB, xlsx)
    DOI: 10.7554/eLife.15798.004
    Figure 1—source data 2. Data values and statistics underlying Figure 1—figure supplement 3.

    DOI: http://dx.doi.org/10.7554/eLife.15798.005

    elife-15798-fig1-data2.xlsx (22.6KB, xlsx)
    DOI: 10.7554/eLife.15798.005
    Figure 1—source data 3. Data values and statistics underlying Figure 1—figure supplement 4.

    DOI: http://dx.doi.org/10.7554/eLife.15798.006

    elife-15798-fig1-data3.xlsx (18.7KB, xlsx)
    DOI: 10.7554/eLife.15798.006
    Figure 2—source data 1. Data values and statistics underlying Figure 2.

    DOI: http://dx.doi.org/10.7554/eLife.15798.012

    elife-15798-fig2-data1.xlsx (19KB, xlsx)
    DOI: 10.7554/eLife.15798.012
    Figure 2—source data 2. Data values and statistics underlying Figure 2—figure supplement 2.

    DOI: http://dx.doi.org/10.7554/eLife.15798.013

    elife-15798-fig2-data2.xlsx (15.7KB, xlsx)
    DOI: 10.7554/eLife.15798.013
    Figure 3—source data 1. Data values and statistics underlying Figure 3.

    DOI: http://dx.doi.org/10.7554/eLife.15798.017

    elife-15798-fig3-data1.xlsx (12.4KB, xlsx)
    DOI: 10.7554/eLife.15798.017
    Figure 3—source data 2. Data values and statistics underlying Figure 3—figure supplement 3.

    DOI: http://dx.doi.org/10.7554/eLife.15798.018

    elife-15798-fig3-data2.xlsx (15.3KB, xlsx)
    DOI: 10.7554/eLife.15798.018

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES