Decision letter: Activity in developing prefrontal cortex is shaped by sleep and sensory experience

Abstract

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract In developing rats, behavioral state exerts a profound modulatory influence on neural activity throughout the sensorimotor system, including primary motor cortex (M1). We hypothesized that similar state-dependent modulation occurs in prefrontal cortical areas with which M1 forms functional connections. Here, using 8- and 12-day-old rats cycling freely between sleep and wake, we record neural activity in M1, secondary motor cortex (M2), and medial prefrontal cortex (mPFC). At both ages in all three areas, neural activity increased during active sleep (AS) compared with wake. Also, regardless of behavioral state, neural activity in all three areas increased during periods when limbs were moving. The movement-related activity in M2 and mPFC, like that in M1, is driven by sensory feedback. Our results, which diverge from those of previous studies using anesthetized pups, demonstrate that AS-dependent modulation and sensory responsivity extend to prefrontal cortex. These findings expand the range of possible factors shaping the activity-dependent development of higher-order cortical areas. Editor's evaluation This manuscript examines neural activity in several cortical areas (such as the primary and secondary motor cortex and the medial prefrontal cortex) across sleep-wake states and under anesthesia. The quality of the recordings in infant rats is excellent, evidence is solid, and results are important in the field of research into the role of active sleep in the neuronal and circuit mechanisms of early cortical development. Some of the findings presented and the hypothesis developed are novel, and some should hopefully prompt future developmental studies to look at sleep as an essential component that cannot be replaced by using anesthetics. https://doi.org/10.7554/eLife.82103.sa0 Decision letter Reviews on Sciety eLife's review process Introduction The functional development of cerebral cortex is a sinuous and often surprising process, even for those structures with the most transparent of adult functions. Consider primary motor cortex (M1), whose name reflects its well-established role in adult motor control: In early development, M1 does not contribute to motor control at all, but instead functions exclusively as a sensory structure (Bruce et al., 1980; Chakrabarty and Martin, 2005; Young et al., 2012; Tiriac et al., 2014; Dooley and Blumberg, 2018; Singleton et al., 2021). The early-emerging somatosensory map in M1 provides the foundation upon which its later-emerging motor map is built (Dooley and Blumberg, 2018). Another surprising aspect of M1 in early development is that its activity is modulated by behavioral state, in particular active sleep (AS, or REM sleep). In infant rats, this modulation reflects AS-dependent increases in neural activity that are enhanced by limb movements during AS, called twitches, that discretely and preferentially trigger sensory feedback to M1 (Dooley and Blumberg, 2018; Glanz et al., 2021). Importantly, AS-dependent modulation of activity is not unique to M1 but is seen in many developing sensorimotor structures (Blumberg, 2015; Blumberg et al., 2020). Given that AS predominates in early life (Jouvet-Mounier et al., 1969; Gramsbergen et al., 1970), it has been posited that this sleep state plays an outsized role in typical and atypical development (Blumberg et al., 2022). That sleep so profoundly modulates neural activity in developing sensorimotor structures raises the possibility that it also modulates activity in cortical areas that are directly or indirectly influenced by sensorimotor input, including higher-order areas like prefrontal cortex. Of particular interest here are two areas with which M1 forms connections: secondary motor cortex (M2) and medial prefrontal cortex (mPFC) (Van Eden et al., 1992; Bedwell et al., 2014; Bedwell et al., 2017). As its name implies, M2 has a particularly close functional and anatomical connection with M1: It integrates multimodal sensory cues for motor planning and modulates M1 activity during goal-directed action (Yin, 2009; Omlor et al., 2019; Barthas and Kwan, 2017; Morandell and Huber, 2017; Wang et al., 2020). Like M1, M2 develops a somatotopic map, further highlighting its dependence on sensory input (Yin, 2009; Kunori and Takashima, 2016; Omlor et al., 2019; Barthas and Kwan, 2017; Chen et al., 2017; Singleton et al., 2021). Thus, we hypothesize that, in infant rats, M2 is similar to M1 with respect to sensory responsiveness and modulation by behavioral state. In contrast with M1 and M2, mPFC in adults is not closely associated with sensorimotor functions, but rather with cognitive processes such as decision-making and attention (Tanji and Hoshi, 2001; Tanji and Hoshi, 2008; Miller et al., 2002; Barbas and Zikopoulos, 2007; Euston et al., 2012). In infant rats, it is not known whether behavioral state modulates activity in mPFC, nor is it known whether mPFC processes sensory input. In fact, it has been theorized that prefrontal cortex, including mPFC, develops its unique higher-order functions precisely because it develops independently of sensory input (Johnson et al., 2015). What is currently known about functional development in mPFC derives primarily from neural recordings from rat pups under urethane anesthesia (Brockmann et al., 2011; Bitzenhofer et al., 2015). Although urethane precludes natural sleep–wake cycles, it does not prevent expression of spindle bursts in mPFC (Brockmann et al., 2011). Spindle bursts are brief thalamocortical oscillations that, in primary sensory areas, are closely associated with the processing of sensory stimuli (Khazipov et al., 2004; Hanganu et al., 2007; Dooley et al., 2020). In the mPFC of urethanized pups, however, spindle bursts appear to occur spontaneously. Here, we determine if this is also the case in unanesthetized pups—as well as test the hypothesis that the infant mPFC, like M1, is modulated by behavioral state. Using unanesthetized rats at postnatal days (P) 8 and P12, we find that M2 and mPFC exhibit state-dependent modulation such that neural firing rates are highest during AS, especially during periods of twitching. We also find that neurons in M2 and mPFC respond to sensory input arising from limb movements that are self-generated (i.e., reafference) or other-generated (i.e., exafference). Finally, to explain discrepancies between the present findings and those reported earlier, we show that urethane administration at P8 prevents expression of behavioral state and brain–behavior relations. Altogether, these findings demonstrate that previously documented effects of behavioral state and sensory experience on somatosensory activity in M1 extend to M2 and mPFC, thus pointing toward new directions for conceptualizing activity-dependent development of higher-order cortical areas. Results We recorded extracellular unit activity in M1, M2, and mPFC in head-fixed rats at P8 and P12 (Figure 1A). For each pup, dual recordings were performed first in the forelimb regions of M1 and M2 (also referred to as the caudal and rostral forelimb areas, respectively) for 40 min, followed by 50 manual stimulations of the forelimb contralateral to the recording sites. Next, the M2 electrode was repositioned in mPFC and the recording and stimulation procedure was repeated but now with dual recordings in M1 and mPFC. Electrode locations in M1, M2, and mPFC were confirmed histologically (Figure 1B). At P8, we collected eight M1–M2 recordings (107 M1 units, 118 M2 units) and eight M1–mPFC recordings (117 M1 units; 103 mPFC units); at P12 we collected nine M1–M2 recordings (217 M1 units; 204 M2 units) and eight M1–mPFC recordings (222 M1 units; 179 mPFC units). Neural activity, electromyographic (EMG) activity in the nuchal and biceps muscles, and high-speed video (100 frames/s) were recorded as pups cycled between sleep and wake (Figure 1C, D). As expected (Dooley et al., 2021; Glanz et al., 2021; Gómez et al., 2021), pups spent more time in AS than wake at P8 (AS: 57.7 ± 2.5%; wake: 30.9 ± 1.9%) and P12 (AS: 44.0 ± 3.6%; wake: 39.3 ± 3.5%). Also, the transition from discontinuous cortical activity at P8 to continuous activity at P12 was evident in all three areas, as described previously in primary somatosensory, motor, and visual cortex (Golshani et al., 2009; Rochefort et al., 2009; van der Bourg et al., 2017; Glanz et al., 2021; Riyahi et al., 2021). Figure 1 Download asset Open asset Representative neural activity in M1, M2, and mPFC in P8 and P12 rats. (A) Illustration showing the surface locations of M1 (blue), M2 (magenta), and mPFC (gold). These color codes are used in all other figures. (B) From left to right, illustrations of coronal sections of M1, M2, and mPFC beneath corresponding brightfield coronal sections that show a fluorescent electrode track in each area. (C) Representative 20-s segments of data from paired recordings in M1 and M2 at P8 (left) and P12 (right) across behavioral states. For each record from the top, data are presented as follows: M2 local field potential (LFP) (magenta trace), M2 unit activity (magenta ticks), M1 LFP (blue trace), M1 unit activity (blue ticks), forelimb movement, and nuchal electromyography (EMG). Bottom row: Behavioral states marked as active sleep (dark gray) or wake (light gray). (D) Same as in C, but for paired recordings in M1 and mPFC (mPFC LFP, gold trace; mPFC unit activity, gold ticks). Neural activity in M2 and mPFC is modulated by behavioral state At P8, representative recordings in M1, M2, and mPFC illustrate substantial and often abrupt increases in neural activity during AS (Figure 2A). In each area, the mean firing rate was significantly higher during AS than wake (t(7)s ≥ 4.38, ps ≤ 0.003, Cohen’s Ds ≥ 1.55; Figure 2B; see Figure 2—figure supplement 1 for data from representative recordings). State-dependent modulation of cortical activity continued through P12 (Figure 2C); once again, the mean firing rate in each area was significantly higher during AS than wake (t(7–8)s ≥ 3.17, ps ≤ 0.016, Cohen’s Ds ≥ 1.12, Figure 2D). Similarly, at P8, the rate of spindle bursts was higher during AS than wake for all three areas (t(7)s ≥ 3.805, ps ≤ 0.007, Cohen’s Ds ≥ 1.35; Figure 3) and were associated with increases in unit activity (Figure 3—figure supplement 1). (Spindle bursts were not analyzed at P12 as they are not clearly discernable at this age.) Thus, like M1, neural activity in M2 and mPFC is modulated at both ages in a state-dependent manner. Figure 2 with 1 supplement see all Download asset Open asset State-dependent unit activity in M1, M2, and mPFC in P8 and P12 rats. (A) Representative 10-min segments of data from a P8 rat showing mean firing rate (2-s bins) in relation to active sleep (dark gray) and wake (light gray). Top: Units in M1 and M2. Bottom: Units in M1 and mPFC. (B) Top: Mean firing rates for M1 and M2 units during active sleep (AS) and wake (W). Bottom: Mean firing rates for M1 and mPFC units during AS and wake. Mean firing rates for individual pups are shown as gray lines. Means ± standard error of the mean (SEM). Asterisks denote significant difference between states, p ≤ 0.025. (C) Same as in A, but for a P12 rat. (D) Same as in B, but for P12 rats. (For M2, the values for one data pair exceed 8 spikes/s and are not shown.) Figure 3 with 1 supplement see all Download asset Open asset State-dependent spindle-burst activity in M1, M2, and mPFC in P8 rats. (A) Left column: Representative 50-s segment of local field potential (LFP) data showing spindle bursts in the spectrogram for a paired M1 (top) and M2 (bottom) recording across active sleep (dark gray) and wake (light gray). Right column: Same as for left column, but for a paired M1 (top) and mPFC (bottom) recording. (B) Illustration to show method for detecting spindle bursts and calculating their amplitude and duration. Spindle bursts were defined when the median LFP amplitude exceeded, for at least 100 ms, an established threshold (horizontal dashed lines). (C) Bar graphs showing mean spindle-burst rate in M1, M2, and mPFC during active sleep (AS) and wake (W). Mean firing rates for individual pups are shown as gray lines. Means ± standard error of the mean (SEM). Asterisks denote significant difference between states, p ≤ 0.025. Neural activity in M2 and mPFC increases during periods of self-generated movement In infant rats, AS-dependent increases in M1 activity correspond with periods of limb movement (e.g., Glanz et al., 2021). Thus, we next determined whether the same is true for M2 and mPFC (Figure 4A). At both ages, the mean firing rate in each area increased significantly during periods of movement (Figure 4B). For all cases, repeated-measures analyses of variance (ANOVAs) revealed significant main effects of behavioral state (F(1,7–8)s ≥ 16.90, ps ≤ 0.005, ηp2s ≥ 0.71) and movement (F(1,7–8)s ≥ 6.83, ps ≤ 0.035, ηp2s ≥ 0.49). None of the state × movement interactions was significant (F(1,7–8)s ≤ 5.35), except for one of the M1 tests at P8 (F(1,7) = 6.11, p = 0.043, ηp2 = 0.47). Figure 4 Download asset Open asset Movement-dependent unit activity in M1, M2, and mPFC in P8 and P12 rats. (A) Representative 20-s segment of data showing unit activity (blue ticks), movement data (black trace), movement periods (green blocks), movement-detection threshold (horizontal dotted line), and behavioral state. (B) Bar graphs showing mean firing rates for neurons in M1, M2, and mPFC during periods of movement (Mvt) or no movement (No Mvt) across active sleep (AS) and wake (W). Mean firing rates for individual pups are shown as gray circles. Means ± standard error of the mean (SEM). Brackets denote significant difference between groups, p ≤ 0.0125. (C) Same as in B, but for P12. For each of the eight repeated-measures ANOVAs across P8 and P12, four planned comparisons were conducted to compare firing rates within behavioral state across movement conditions and within movement conditions across behavioral state (Figure 4B, C). Of the 32 planned comparisons, 31 were significant (t(7–8)s ≥ 3.51, ps ≤ 0.01, Cohen’s Ds ≥ 1.24). The general pattern was for firing rates to be highest during AS-related periods of movement (i.e., twitching), intermediate during periods of AS-related periods of no movement and wake-related periods of movement, and lowest during wake-related periods of no movement. In summary, at P8 and P12, neural activity in M1, M2, and mPFC reflects the interactive effects of behavioral state and movement. Given that all three areas exhibited similar movement-related increases in activity and that M1 is known to respond to movement-related sensory feedback (Dooley and Blumberg, 2018; Gómez et al., 2021), we determined next whether M2 and mPFC are also responsive to sensory input. Neurons in M2 and mPFC respond to sensory input We quantified neural responses in M1, M2, and mPFC to forelimb twitches, wake movements, and stimulations (Figure 5A). As expected (and in M1, consistent with previous results; Tiriac et al., 2014; Dooley and Blumberg, 2018; Glanz et al., 2021; Gómez et al., 2021), units in M1 and M2 at both ages responded to sensory input arising from twitches, wake movements, and stimulations; surprisingly, so did units in mPFC. The percentage of responsive units in all three areas varied by age and event type. At P8, M1 generally exhibited the highest percentage of responsive units, followed by M2 and then mPFC (t(7)s ≥ 4.44, ps ≤ 0.003, Cohen’s Ds ≥ 1.57; Figure 5B). Mean M2 responsiveness was 60.9 ± 10.2% for twitches, 46.5 ± 11.9% for wake movements, and 36.7 ± 9.1% for stimulations; for mPFC, these values were 37.8 ± 8.0%, 20.6 ± 7.5%, and 9.4 ± 5.5%, respectively. At P12, responsiveness declined to low levels in all three areas, but M1 was still more responsive than M2 or mPFC (t(7–8)s ≥ 3.13, ps ≤ 0.017, Cohen’s Ds ≥ 1.11; Figure 5C). Figure 5 Download asset Open asset M1, M2, and mPFC neural responses to sensory input in P8 and P12 rats. (A) Methodology for determining sensory responsiveness of individual units. Left: Perievent time histogram (PETH) of unit firing rate (blue line) relative to a sensory event, showing the baseline window (BLW) and response window (RW). Event onset denoted by dotted line at Te. Middle: Bar graph showing mean unit activity during the BLW and RW, response threshold (dotted line), and the threshold calculation. Right: Stacked plot showing the percentage of units that exceeded the response threshold (blue) and percentage of units that did not (gray). (B) Stacked plots showing mean percentage of responsive (colored) and unresponsive (gray) units across pups at P8. Top: Data from M1 (blue) and M2 (magenta) recordings. Bottom: Data from M1 (blue) and mPFC (gold) recordings. Means + standard error of the mean (SEM). Asterisks denote significant difference between cortical areas, p ≤ 0.017. PETHs to the right of each stacked plot show normalized unit firing rates for responsive units only in each cortical area. (C) Same as in B, but at P12. Regardless of the mean responsiveness of a cortical area at a given age, when units were responsive they exhibited response profiles (i.e., perievent time histograms, PETHs) that were strikingly similar to each other. These profiles indicate sensory responding because the peaks in activity occurred after the movement or stimulation (Figure 5B, C). Such response profiles were observed even when responsive units were rare (e.g., mPFC at P12). To assess whether M1, M2, and mPFC were similarly activated by sensory events, we next measured each area’s activation rate to compare the reliability with which each area responded to sensory events. The activation rate was defined as the percentage of twitches, wake movements, or stimulations for which at least 30% of responsive units in the area showed an increase in activity (Figure 6A); this threshold was chosen based on prior study of population responses in M1 to self-generated movements at P8 and P12 (Glanz et al., 2021). At P8, M2 and mPFC showed mean activation rates that were similar to those in M1 (t(7)s ≤ 2.28) (Figure 6B), with one exception: mPFC had a significantly lower mean activation rate than M1 for stimulations (t(7) = 5.91, p < 0.001, Cohen’s D = 2.09). At P12, mean activation rates in M2 and M1 were also similar (t(8)s ≤ 2.53) (Figure 6C); for mPFC, the mean activation rate for twitches was similar to that for M1 (t(7) = 1.73), but mean activation rates were significantly lower in mPFC for wake movements and stimulations (t(7)s ≥ 4.06, ps ≤ 0.005, Cohen’s Ds ≥ 1.43). Thus, whereas M1 and M2 had comparable activation rates to all three kinds of sensory events at both ages, mPFC was generally less responsive than M1. Figure 6 with 1 supplement see all Download asset Open asset Response rates of M1, M2, and mPFC to sensory events in P8 and P12 rats. (A) Methodology for determining the activation rate of cortical areas to sensory events. Left: Illustration of responsive (orange) and unresponsive (gray) units within an area. Right: Illustration of activity of responsive neurons (opaque) to individual sensory events. For each of the three sensory events indicated, the percentage of responsive units is determined. Based on the percentage of events that exceeds threshold (>30%; check marks), the activation rate is calculated. (B) Activation rates in M1, M2, and mPFC at P8 to twitches (Twitch), wake movements (Wake), and stimulations (Stim). Left: Bar graphs showing percentage of sensory events that evoked a response in M1 and M2. Right: Same as at left, but for M1 and mPFC. Mean activation rates for individual pups are shown as gray lines. Means ± standard error of the mean (SEM). Asterisks denote significant difference between areas, p ≤ 0.017. (C) Same as in B, but at P12. Spindle bursts in all three areas were associated with sensory events. In M1 and M2, spindle bursts were significantly more likely to occur following twitches, wake movements, and stimulations when compared to shuffled data (t(7)s ≥ 5.79; ps ≤ 0.001; Cohen’s Ds ≥ 0.93) (Figure 6—figure supplement 1). Likewise, spindle bursts in mPFC reliably followed wake movements and stimulations (t(7)s ≥ 4.27; ps ≤ 0.004; Cohen’s Ds ≥ 1.00), though not twitches (t(7) = 2.16). In most cases, sensory events were equally likely to trigger spindle bursts across areas, with the exception of twitches for M2 and stimulations for mPFC (t(7)s ≥ 3.51, ps ≤ 0.01, Cohen’s Ds ≥ 1.24). Finally, the fact that M1, M2, and mPFC all exhibited sensory responses at these ages led us to consider the sources of this sensory input, as done previously for M1 and primary somatosensory cortex (S1; Gómez et al., 2021). Specifically, we performed analyses to determine if sensory input is conveyed in parallel to these structures or serially between them (e.g., from M1 to M2). However, unlike with M1 and S1, this analysis yielded a null result: Individual sensory events did not reliably trigger contemporaneous unit activity in M1 and M2, or M1 and mPFC above chance (data not shown). This result suggests that the pathways through which sensory input reaches M2 and mPFC are distinct from those that reach M1 and S1, and presently remain unknown. In summary, as found previously in M1, both M2 and mPFC respond to sensory input in early development with increases in spiking activity and spindle bursts. Urethane anesthesia suppresses behavior and neural activity in M1 and mPFC The prefrontal activity described thus far does not resemble that reported previously in urethanized pups (Brockmann et al., 2011; Bitzenhofer et al., 2015; Chini et al., 2019). To determine whether the use of urethane accounts for this disparity, we recorded M1 and mPFC activity in an additional set of P8 rats (n = 6/group) before and after administration of urethane (1.0 mg/g b.w. SC) or sterile saline. Urethane administration produced rapid and dramatic effects on behavior and neural activity (Figure 7A). Before urethane injection, pups cycled between sleep and wake, as evidenced by alternating periods of high and low muscle tone accompanied by bouts of wake movements and twitches, respectively. In contrast, urethane (but not saline) injection produced muscle atonia (with occasional spasmodic increases in muscle tone) and suppressed limb movements, thus precluding identification of sleep–wake states. Compared with saline, urethane injection produced significant percentage decreases in limb movements (urethane: −87.14 ± 6.75%; t(5) = 11.90, p < 0.001, Cohen’s D = 4.86; saline: +10.84 ± 5.84%, t(5) = 1.00). Of the limb movements that remained after urethane, most occurred during brief whole-body spasms; twitch-like movements were rarely observed. Figure 7 with 1 supplement see all Download asset Open asset Urethane anesthesia suppresses unit activity in M1 and mPFC in P8 rats. (A) Representative 20-s segments of data from recordings in M1 and mPFC before (left) and after (right) injection of urethane (1.0 mg/g b.w.). For each record from the top, data are presented as follows: mPFC local field potential (LFP) (gold trace), mPFC unit activity (gold ticks), M1 LFP (blue trace), M1 unit activity (blue ticks), forelimb movement, and nuchal electromyography (EMG). (B) Representative 75-min segment of data showing mean unit firing rate (in 2-s bins) in M1 (blue) and mPFC (gold) before and after injection of urethane (vertical dashed line). (C) Bar graphs showing mean firing rates of neurons across pups in M1 (left) and mPFC (right) during the pre-injection (Pre) and post-injection (Post) periods for the urethane (UR) and saline (SAL) groups. Mean firing rates for individual pups are shown as gray circles. Means ± standard error of the mean (SEM). (D) Left: Survivor plots of pooled interspike intervals (ISIs) for M1 units during the pre- and post-injection periods for pups in the urethane (solid blue line) and saline (dashed blue line). Right: Same as at left but for mPFC during the pre-injection (dark gold) and post-injection (light gold) periods. Asterisks denote significant difference (p ≤ 0.025) between urethane and saline groups for ISI values at the bottom fifth percentile (dashed horizontal lines). Urethane administration also disrupted neural activity in M1 and mPFC (Figure 7A, B), causing reductions in firing rate of over 85% (Figure 7C). Mean reductions in firing rate were significant for both M1 (t(10) = 3.83, p = 0.003, Cohen’s D = 2.21) and mPFC (t(10) = 5.01, p < 0.001, Cohen’s D = 2.94). Urethane also dramatically and significantly reduced the mean rate of spindle bursts in the two areas (t(10)s ≥ 3.18, ps ≤ 0.01, Cohen’s Ds ≥ 1.83) (Figure 7—figure supplement 1). Urethane also changed the temporal patterning of neural activity (Figure 7A). In the absence of urethane, neural activity in both areas exhibited the discontinuous pattern characteristic of cortical activity at P8 (Golshani et al., 2009; van der Bourg et al., 2017; Glanz et al., 2021). In contrast, urethane injection produced a burst-suppression pattern that is characteristic of general anesthesia as well as coma, hypothermia, and neonatal trauma (Grigg-Damberger et al., 1989; Steriade et al., 1994; Hellström-Westas et al., 2006; Shanker et al., 2021). This pattern, comprising population bursts separated by periods of relative silence lasting 10 s or longer, is illustrated by survivor plots of interspike intervals (ISIs) (Figure 7D): In both areas, whereas the pre-injection ISI distributions for urethane and saline are indistinguishable, the post-injection distributions for the urethane group deviates substantially from the saline group, especially for longer ISIs where the pronounced shoulders in the plots, indicative of inter-burst silence, are evident. For the bottom fifth percentile of ISIs, we found a significant difference between urethane and saline groups during the post-injection period in both M1 and mPFC (t(5)s ≥ 3.36; ps ≤ 0.007; Cohen’s Ds ≥ 1.94), but not during the pre-injection period (t(5)s ≤ 1.64). In summary, urethane anesthesia at P8 eradicates sleep–wake cycling, suppresses behavior, and produces atypical neural activity. Discussion We demonstrate here in P8 and P12 rats that neurons in two prefrontal areas—M2 and mPFC—exhibit state-dependent neural activity and responsivity to somatosensory stimuli. First, at both ages, neural activity in M2 and mPFC increases specifically during AS, similar to previous findings at these ages in M1 and S1 (Tiriac et al., 2014; Dooley et al., 2020; Glanz et al., 2021). Second, we find that neurons in M2 and mPFC respond to reafference arising from twitches and wake movements, and exafference arising from manual stimulation, with the proportion of responsive neurons generally being highest in M1 and decreasing across M2 and mPFC. Finally, we show that urethane thwarts accurate assessments of brain–behavior relations in developing cortex by suppressing neural activity and abolishing sleep–wake states, thus explaining discrepancies between the present and previous findings. Altogether, these results highlight the potential importance of sleep and sensory experience for the functional development of prefrontal cortex. Prefrontal cortex is most active during sleep In developing rats, AS modulates spiking and oscillatory activity in M1 and S1 (Blumberg et al., 2020; Dooley et al., 2020; Glanz et al., 2021), findings that we now extend to M2 and mPFC. Neural activity in these two areas was highest during movement-related periods of AS, but it was also higher during AS than wake even in the absence of movement. These findings suggest that state-dependent neuromodulation is a general feature of infant cortical activity. Although neuromodulators like acetylcholine and serotonin, respectively, influence early cortical activity (Hanganu et al., 2007; Janiesch et al., 2011) and development (Kolk and Rakic, 2022), it is not yet known whether these and other neuromodulators are released in a state-dependent fashion, as is known to occur in adults (Lee and Dan, 2012; Jones, 2020). Prefrontal cortex responds to sensory input Sensory experience in early life scaffolds developing sensory and sensorimotor systems, providing information about the growing body and the world it inhabits (Blumberg, 2015). Notably, recent evidence from the visual system of both rodents and primates suggests that how sensory input reaches cortex is fundamentally different in infants and adults. In adults, sensory information flows through a hierarchical network, from primary cortical areas to higher-order cortical areas. But, in the developing visual system, both primary and higher-order visual areas receive parallel sensory input directly from thalamus (Warner et al., 2012; Murakami et al., 2022). Likewise, in the developing sensorimotor system, both M1 and S1 receive parallel sensory input (Dooley and Blumberg, 2018; Gómez et al., 2021). This ascending sensory input to M1 and S1 may refine somatotopy and connectivity within and between these cortical areas, thus laying a foundation for their further development, including the later emergence of M1’s motor functionality (Dooley and Blumberg, 2018; Gómez et al., 2021). M2 is a higher-order sensorimotor area with a somatotopic representation, though its organization is coarser than M1’s (Mohammed and Jain, 2014; Mohammed and Jain, 2016). Thus, it is perhaps not surprising that we found that M2 units, like those in M1, exhibit short-latency sensory res