A transient postnatal quiescent period precedes emergence of mature cortical dynamics
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
The authors use dense electrode recordings in young mice and EEG recordings in human infants to quantitatively describe the transition from immature patterns of brain activity in sleep to more mature patterns. Interestingly, they find an intervening period when overall activity declines in both species. This study is interesting because it enriches our relatively impoverished view of how mature activity patterns emerge during development. However, reviewers expressed concerns that further work was need to rule out potential artifacts of the surgical and recording techniques used in the animal experiments.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #3 agreed to share their names with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Neuroscience (eLife)
Abstract
Mature neural networks synchronize and integrate spatiotemporal activity patterns to support cognition. Emergence of these activity patterns and functions is believed to be developmentally regulated, but the postnatal time course for neural networks to perform complex computations remains unknown. We investigate the progression of large-scale synaptic and cellular activity patterns across development using high spatiotemporal resolution in vivo electrophysiology in immature mice. We reveal that mature cortical processes emerge rapidly and simultaneously after a discrete but volatile transition period at the beginning of the second postnatal week of rodent development. The transition is characterized by relative neural quiescence, after which spatially distributed, temporally precise, and internally organized activity occurs. We demonstrate a similar developmental trajectory in humans, suggesting an evolutionarily conserved mechanism that could facilitate a transition in network operation. We hypothesize that this transient quiescent period is a requisite for the subsequent emergence of coordinated cortical networks.
Article activity feed
-
-
Author Response:
Reviewer #1:
The authors use dense electrode recordings in young mice and EEG recordings in human infants to quantitatively describe the transition from immature patterns of brain activity in sleep to more mature patterns. Interestingly, they find an intervening period when overall activity declines in both species. Although primarily concerned with describing the phenomenology of this transition, this study is interesting because it enriches our relatively impoverished view of how mature activity patterns emerge during development.
Reviewer #2:
The authors employ sophisticated electrophysiological techniques and analyses to investigate ontogenetic patterns of brain activity in sleep. This is a major strength of the study.
Although this topic has been explored many times over the last 50-60 years, the authors …
Author Response:
Reviewer #1:
The authors use dense electrode recordings in young mice and EEG recordings in human infants to quantitatively describe the transition from immature patterns of brain activity in sleep to more mature patterns. Interestingly, they find an intervening period when overall activity declines in both species. Although primarily concerned with describing the phenomenology of this transition, this study is interesting because it enriches our relatively impoverished view of how mature activity patterns emerge during development.
Reviewer #2:
The authors employ sophisticated electrophysiological techniques and analyses to investigate ontogenetic patterns of brain activity in sleep. This is a major strength of the study.
Although this topic has been explored many times over the last 50-60 years, the authors make some interesting observations. The first is that there is a window of time when immature cortical activity changes from immature forms to more mature forms. The 2nd major finding is a transient condition of diminished brain activity that appears between these stages.
Major weaknesses:
The first finding seems incremental in nature. Especially as no mechanistic insights are provided. It is well known that the 2nd postnatal week in rodents is when many cortical and sub cortical events coincide with a change in sleep organization--including cortical manifestations. Therefore, the first finding is more detailed than earlier studies, but not especially surprising when put in proper context.
Our goal in this work was to investigate emergence of network processes associated with cognitive functions in adults over the course of development. Such features are not necessarily strictly linked to sleep organization. In our opinion, identifying the timelines for such processes merits specific investigation due to the likely implications for derangement of cortical function in neurodevelopmental disorders. Furthermore, we show that these processes change abruptly over a short period of time, rather than progressively or in a staggered fashion during the second postnatal week.
The 2nd finding is interesting, but its significance is unknown.The significance of this 'state' or 'condition' is a bit overstated. For example, the authors state in their discussion that this state 'enables' the emergence of mature brain organization, but they provide no evidence for this. Their study, as interesting as it is in places, is descriptive and provides no direct evidence of mechanism or function.
We agree that a key opportunity for future investigation provided by this work is for perturbation of the transition period to identify functional ramifications. Most likely, multiple molecular, genetic, and cellular mechanisms underlie such a profound transition in electrophysiological features. We have now highlighted both of these points in our Discussion. We would suggest, however, that the fact that network properties change swiftly and simultaneously during a quiescent state provides key clues about the ways by which neural networks can shift their properties. On a practical level, an identifiable marker of developmental maturation, such as a quiescent state, allows matching of cortical development timelines across species, and in instances of putative cortical pathology. Therefore, we posit that identification and characterization of this state are functionally useful, regardless of whether a specific function is ascribed to the state.
There are also methodological issues that make the interpretation of the mouse data extremely difficult.
Performing in vivo electrophysiologic recordings in immature organisms remains challenging due to various experimental and technical considerations. We employed evidence-based practices and verified the health of the neural networks being monitored to minimize confounders related to any specific methodology. We have included more detail on each of these practices in the Methods section of the manuscript to facilitate robust experimentation in developing organisms.
Overall, the analyses are meticulous and suggest an important phase of brain organization occurs at about the 2nd postnatal week in rodents--and possibly humans. This study could be very informative, provided that additional control experiments are performed, and direct mechanistic or functional questions are addressed.
Reviewer #3:
This paper is, to my knowledge, the first to suggest that there may be 'regressive' or at least non-progressive steps in the general thrust of early activity and functional development, at least before the later stages of net synaptic elimination. The authors show that in mouse somatosensory cortex that the period after spindle-burst elimination (an early activity pattern associated with sensory stimulation either self-generated or evoked) is characterized by a 2-day 'nadir' in total activity before firing rates and synchronization as well as surface EEG power and spread begin again to increase toward adult levels. This pattern was echoed in EEG recordings from human infants, which showed a similar decrease in activity around 45 weeks of gestation (on parietal electrodes). This careful analysis of activity done similarly in the two species is a real strength and overall my confidence is high that this is a real phenomenon in the regions examined. The number of animals and analysis methods are impressive and largely appropriate. Overall the data presented make a solid and important contribution to our understanding of the developmental dynamics of neural activity development.
To my mind, there are a couple of critical analyses that need to be included to fully support the authors' conclusions.
- The mouse experiments call for some control of developmental changes in arousal state especially as regards twitching and other movement. With the current presentation, the quiescent period could as easily be a result of reduced twitching at P8 before extensive volitional (and whisking) emerges starting on P10 as it could be explained by circuit changes in the ascending pathways. Likewise, shifts in the proportion of quiet and active sleep (which are related to twitch amount) could largely account for the differences.
Thank you for identifying potential confounders for our observations. To address these, we first quantified twitching rate in each animal and examined whether there were any systematic changes across age groups. There was no significant difference in twitching rates across age groups (ANOVA p = 0.0861), though a weak trend toward decrease in twitching over time (P5 to P14) was found, in agreement with other studies of twitches in neonatal rodents (1-3). The lack of statistically significant change in twitch rate across groups, and the lack of nadir in twitching during our identified transition period argues against our results being a function of less twitching. This data is presented in Supplementary Figure 5D with relevant statistical testing.
We furthermore analyzed the proportion of time spent in active/quiet sleep across this developmental period. As known from the literature, the most mature animals had less active sleep than the most immature animals (3-5). Although the exact amount of quiet sleep in early development remains unclear, our results fits the increasing trend of quiet sleep reported and described by other groups (4). This data is presented in Supplementary Figure 5C with relevant statistical testing. ANOVA with post-hoc testing did not reveal a significant difference in active sleep proportion between P5-7 and P8-9 animals, or between P8-9 and P10-12 animals, indicating lack of an abrupt change in sleep proportions during the transition period that could explain our results. Furthermore, we specifically analyzed data from periods of immobility lasting 10 seconds or more to facilitate analysis of comparable states given the difficulty in precise scoring of active and quiet sleep in neonatal rodents (5-6). Therefore, any potential effects related to sleep state are minimized.
There was no sharp transition (or statistically significant group difference) in either feature that could account for the unique electrophysiologic features exhibited by the animals at the beginning of the second postnatal week. It would also be difficult to explain the differences in oscillation spatial extent, interspike interval, phase locking, and cross-frequency coupling that we observe during this time as a function of twitching or sleep state. Taken together, these data do not support the notion that the pausing phenomenon is an artifact of twitching or sleep state distributions across ages.
- The location of the analyzed contacts is incompletely described and justified. In the mouse they are described as 'somatosensory cortex' but the pictures suggest that barrel cortex is the most likely location. Better descriptions of how the locations for analysis were chosen and controlled over the wide age range are necessary. Were the contacts analyzed verified as barrel cortex by whisker deflection? Is there any possibility the quiescent period is a result of shifting the location of the grid or analyzed channels. The infant data surprisingly are taken primarily from parietal electrodes, which are not the location of sensory-evoked twitches (Milh et al 2007). Why was the analysis limited to parietal? Are the results dependent on this localization?
We used vGLUT2 immunohistochemistry to identify primary somatosensory and primary visual cortex. Barrel cortex has the most striking histological appearance using this method, and we centered our NeuroGrids over this particular region of primary somatosensory cortex. However, we did not perform functional testing by whisker deflection, which is why we prefer to use the more generic term “somatosensory cortex” than “barrel cortex” because we cannot exclude that some channels were in forelimb, hindlimb or other regions of somatosensory cortex. We note in Supplementary Figure 3 that channels identified histologically as recording from somatosensory cortex displayed spindle bursts in the immature mice, concordant with literature on this region (for instance, 13). NeuroGrids were large enough to extend past somatosensory cortex in all ages, allowing us to consistently identify channels recording from this region and making it essentially impossible to “miss” somatosensory cortex during surgical placement.
For the human data, we used electrodes that relatively correspond to somatosensory cortex in rodents. In Milh et al 2007, a double distance neonatal montage is used because the recordings are from very premature infants (29-31 wks), where head size precludes placement of a full 10-20 electrode montage. In this case, the “C” or central electrodes are located over the somatomotor area. In a conventional 10-20 montage, the somatosensory area is expected to lie between the central electrodes and the parietal electrodes. We chose to use parietal electrodes because they had the most consistent high-quality data across our patient group, but similar results are obtained if central electrodes are used. We replicated power analysis based on central electrodes in Supplementary Figure 11D, and there is no change to the result.
We have included this additional information regarding location of the analyzed contacts in the Methods section.
- The authors do a number of analyses of cross-frequency co-modulation and spike-frequency modulation that are limited to 'spindle frequencies'. These results are often extrapolated to make general statements about the precision of spiking or spread of activity etc but are really just smaller snapshots of the larger activity. This would be justified if there was good reason to believe that early spindle-bursts and later sleep spindles are the same network activity. However this proposition has only weak support (and is not argued for explicitly here). In essence, the authors end up analyzing three different patterns: spindle-bursts in P5-7, unknown activity in spindle band (P8-10), and sleep spindles (P11+). That these are in the same broad range of frequencies doesn't mean they are making similar measurements across ages. It would strengthen the case that P8-10 is a unique quiescent period to show differences in power spectra and spiking not limited to spindle frequencies. Some of these are presented, but difficult to extract from the spindle analyses. In addition spiking data from layers, 4-6 are used, but these layers are both very diverse in their behavior, and the least likely to be strongly correlated with spindle-bursts (maximal in layer 2-4). A more consistent and limited analysis of spiking is important to confirm the general vs specific nature of this quiescence.
We do make several analyses that are independent of spindle activity:
- Continuity (Figure 2C, Figure 6D)
- Wide band power (Figure 2D, Figure 6C)
- Spiking rate (Figure 3A)
- Interspike interval (Figure 3B)
- Spike autocorrelation (Figure 3C-D)
Therefore, the transient quiescent period is not limited to spindle band oscillations. To clarify this point, we have included power spectra as suggested by the reviewer, which demonstrate a paucity of oscillatory power across the physiologic frequency spectrum between P8-9 (Supplementary Figure 6), as well as in humans during the transient period (Supplementary Figure 13). We have also clarified in the Results and Methods that these analyses are derived from any activity above the noise floor, not just those in the spindle band. We have also rearranged the results text to improve the clarity of these analyses.
The rationale for subsequently focusing on the spindle band frequency is that well identified oscillations exist in this band in immature and mature animals. Certainly, this does not presuppose that these oscillations are serving a similar purpose or are generated by similar underlying mechanisms, and as the reviewer notes, we do not espouse this notion here. However, it does allow us to reliably detect discrete oscillations across development for the purpose of investigating the spike/LFP relationship in a relatively controlled fashion.
We quantified spiking activity from superficial cortical layers to address the last point mentioned here. We used the grouping of layers 2-3 and 4-6 for this purpose to maximize integration with the data obtained using surface arrays, which capture activity primarily from the superficial cortical layers (Khodagholy et al., Nature Neuroscience 2015). We were also mindful of the precision of the histological methods used, and thus did not separate into more than 2 groups. We found that the superficial cortical layers followed a similar pattern to the deeper layers in regard to the spiking measures analyzed (firing rate, interspike interval, recruitment into spindle-band oscillations). The results of these analyses are presented, with complete quantification, in Supplementary Figure 10 and referenced in the Results text. A nadir at the beginning of the second postnatal week was demonstrated in each analysis.
- How generalizable these results are, and how they comport with previous studies is unclear. The paper is written as if this quiescent state is universal, and its identification in two species in likely different regions adds to the argument that this is the case. However, it has not been observed in similarly detailed developmental studies in other rodent regions (multiple papers by the Hanganu-Opatz lab, Minlebeav et al Science 2011, Shen and Colonnese J Neuro 2016) nor in the clinical literature. Some more careful and nuanced discussion of the relationship between these findings or expansion of the regions surveyed to show they were wrong would help situate the current findings and better comport the claims and evidence.
We have carefully reviewed the literature to address this point.
Hanganu-Opatz and colleagues have performed detailed work on the development of prefrontal cortex (some examples in 18-20). Whether this association cortex, which does not receive sensory information directly from thalamus, would be expected to follow a similar developmental trajectory to a primary sensory cortex is unclear. They group ages across multiple developmental delays (18), or sample with wider intervals (19). With such a method, it is possible that short transitions in neural activity could be obscured. From such studies, it can be discerned that a discontinuous pattern is present in the prefrontal cortex of rats as late at P9, with a continuous appearance by P12. However, the transition between these states is not delineated. Interestingly, their recent study showed that optogenetically increasing mPFC activity around the beginning of the second postnatal week disrupts developmental trajectory and results in functional deficits in adulthood (20).
Minlebeav and colleagues characterized gamma band activity (14). Importantly, they actually reported an abrupt decrease in gamma band activity at P8 and the disappearance of early gamma oscillations (EGOs) around P8 before “adult” gamma patterns emerge later during development. This is concordant with our timeline of spindle band activities, potentially suggesting potentially a shift in cortical dynamics around the timepoint.
Shen and Colonnese investigated elicited response from primary visual cortex during the first weeks of neonatal development (21). Although they did not identify a transient nadir in continuity, they reported that LFP continuity in superficial layers did rise sharply after P8 with a highly non-linear trajectory. The depth profile of spontaneous activity changed in cortex between P8-10, accompanied by a change in anticorrelated activity and spectral features at this timepoint. Also concordant with our results, an earlier work by Colonnese and colleagues investigated correlation of neural firing in mouse visual cortex, and reported that activity is the least synchronous at the beginning of the second postnatal week (22), similar to our reported nadir in temporal precision of spiking.
This literature supports an abrupt change in cortical network function around a similar timepoint to what we identify, with some indicators of a nadir in synchrony, spectral features, and oscillatory patterns. It is possible that our use of surface arrays, which sample summated local field potential activity from the undisturbed superficial cortical layers (I-III) highlights the transient quiescent state compared to penetrating probes that disrupt the cortical surface upon implantation. Furthermore, our conducting polymer-based electrodes have lower impedance than the silicon probes used in most neonatal rodent studies, potentially increasing the sensitivity to changes in oscillatory power and continuity.
From a clinical perspective, longitudinal studies of neonatal EEG activity are understandably few. The disappearance of trace alternant and emergence of sleep spindles have been used as boundaries for “perinatal” and “infantile” pattern classifications, and this transition is described to occur over a few weeks between 44-49 wks (23), coinciding with our observed transition period. Trace alternant is identified by a pronounced change in amplitude of the EEG signal with a characteristic burst and interburst interval. We propose that the high amplitude activity is lost during the transition period, leading to a low voltage state with lack of organized oscillatory activity, which is replaced by the typical organized patterns of more mature NREM sleep. When quantified, such a trajectory would resemble that of our data in Figure 6. We support this notion by including power spectra derived from the human data in Supplementary Figure 13, which shows a paucity of periodic or aperiodic activity during the transition period. Therefore, we posit that our results are not at odds with clinical literature, but more clearly define the transition period.
-
Evaluation Summary:
The authors use dense electrode recordings in young mice and EEG recordings in human infants to quantitatively describe the transition from immature patterns of brain activity in sleep to more mature patterns. Interestingly, they find an intervening period when overall activity declines in both species. This study is interesting because it enriches our relatively impoverished view of how mature activity patterns emerge during development. However, reviewers expressed concerns that further work was need to rule out potential artifacts of the surgical and recording techniques used in the animal experiments.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #3 …
Evaluation Summary:
The authors use dense electrode recordings in young mice and EEG recordings in human infants to quantitatively describe the transition from immature patterns of brain activity in sleep to more mature patterns. Interestingly, they find an intervening period when overall activity declines in both species. This study is interesting because it enriches our relatively impoverished view of how mature activity patterns emerge during development. However, reviewers expressed concerns that further work was need to rule out potential artifacts of the surgical and recording techniques used in the animal experiments.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #3 agreed to share their names with the authors.)
-
Reviewer #1 (Public Review):
The authors use dense electrode recordings in young mice and EEG recordings in human infants to quantitatively describe the transition from immature patterns of brain activity in sleep to more mature patterns. Interestingly, they find an intervening period when overall activity declines in both species. Although primarily concerned with describing the phenomenology of this transition, this study is interesting because it enriches our relatively impoverished view of how mature activity patterns emerge during development.
-
Reviewer #2 (Public Review):
The authors employ sophisticated electrophysiological techniques and analyses to investigate ontogenetic patterns of brain activity in sleep. This is a major strength of the study.
Although this topic has been explored many times over the last 50-60 years, the authors make some interesting observations. The first is that there is a window of time when immature cortical activity changes from immature forms to more mature forms. The 2nd major finding is a transient condition of diminished brain activity that appears between these stages.
Major weaknesses:
The first finding seems incremental in nature. Especially as no mechanistic insights are provided. It is well known that the 2nd postnatal week in rodents is when many cortical and sub cortical events coincide with a change in sleep organization--including …
Reviewer #2 (Public Review):
The authors employ sophisticated electrophysiological techniques and analyses to investigate ontogenetic patterns of brain activity in sleep. This is a major strength of the study.
Although this topic has been explored many times over the last 50-60 years, the authors make some interesting observations. The first is that there is a window of time when immature cortical activity changes from immature forms to more mature forms. The 2nd major finding is a transient condition of diminished brain activity that appears between these stages.
Major weaknesses:
The first finding seems incremental in nature. Especially as no mechanistic insights are provided. It is well known that the 2nd postnatal week in rodents is when many cortical and sub cortical events coincide with a change in sleep organization--including cortical manifestations. Therefore, the first finding is more detailed than earlier studies, but not especially surprising when put in proper context.
The 2nd finding is interesting, but its significance is unknown.The significance of this 'state' or 'condition' is a bit overstated. For example, the authors state in their discussion that this state 'enables' the emergence of mature brain organization, but they provide no evidence for this. Their study, as interesting as it is in places, is descriptive and provides no direct evidence of mechanism or function.
There are also methodological issues that make the interpretation of the mouse data extremely difficult.
Overall, the analyses are meticulous and suggest an important phase of brain organization occurs at about the 2nd postnatal week in rodents--and possibly humans. This study could be very informative, provided that additional control experiments are performed, and direct mechanistic or functional questions are addressed.
-
Reviewer #3 (Public Review):
This paper is, to my knowledge, the first to suggest that there may be 'regressive' or at least non-progressive steps in the general thrust of early activity and functional development, at least before the later stages of net synaptic elimination. The authors show that in mouse somatosensory cortex that the period after spindle-burst elimination (an early activity pattern associated with sensory stimulation either self-generated or evoked) is characterized by a 2-day 'nadir' in total activity before firing rates and synchronization as well as surface EEG power and spread begin again to increase toward adult levels. This pattern was echoed in EEG recordings from human infants, which showed a similar decrease in activity around 45 weeks of gestation (on parietal electrodes). This careful analysis of activity …
Reviewer #3 (Public Review):
This paper is, to my knowledge, the first to suggest that there may be 'regressive' or at least non-progressive steps in the general thrust of early activity and functional development, at least before the later stages of net synaptic elimination. The authors show that in mouse somatosensory cortex that the period after spindle-burst elimination (an early activity pattern associated with sensory stimulation either self-generated or evoked) is characterized by a 2-day 'nadir' in total activity before firing rates and synchronization as well as surface EEG power and spread begin again to increase toward adult levels. This pattern was echoed in EEG recordings from human infants, which showed a similar decrease in activity around 45 weeks of gestation (on parietal electrodes). This careful analysis of activity done similarly in the two species is a real strength and overall my confidence is high that this is a real phenomenon in the regions examined. The number of animals and analysis methods are impressive and largely appropriate. Overall the data presented make a solid and important contribution to our understanding of the developmental dynamics of neural activity development.
To my mind, there are a couple of critical analyses that need to be included to fully support the authors' conclusions.
The mouse experiments call for some control of developmental changes in arousal state especially as regards twitching and other movement. With the current presentation, the quiescent period could as easily be a result of reduced twitching at P8 before extensive volitional (and whisking) emerges starting on P10 as it could be explained by circuit changes in the ascending pathways. Likewise, shifts in the proportion of quiet and active sleep (which are related to twitch amount) could largely account for the differences.
The location of the analyzed contacts is incompletely described and justified. In the mouse they are described as 'somatosensory cortex' but the pictures suggest that barrel cortex is the most likely location. Better descriptions of how the locations for analysis were chosen and controlled over the wide age range are necessary. Were the contacts analyzed verified as barrel cortex by whisker deflection? Is there any possibility the quiescent period is a result of shifting the location of the grid or analyzed channels. The infant data surprisingly are taken primarily from parietal electrodes, which are not the location of sensory-evoked twitches (Milh et al 2007). Why was the analysis limited to parietal? Are the results dependent on this localization?
The authors do a number of analyses of cross-frequency co-modulation and spike-frequency modulation that are limited to 'spindle frequencies'. These results are often extrapolated to make general statements about the precision of spiking or spread of activity etc but are really just smaller snapshots of the larger activity. This would be justified if there was good reason to believe that early spindle-bursts and later sleep spindles are the same network activity. However this proposition has only weak support (and is not argued for explicitly here). In essence, the authors end up analyzing three different patterns: spindle-bursts in P5-7, unknown activity in spindle band (P8-10), and sleep spindles (P11+). That these are in the same broad range of frequencies doesn't mean they are making similar measurements across ages. It would strengthen the case that P8-10 is a unique quiescent period to show differences in power spectra and spiking not limited to spindle frequencies. Some of these are presented, but difficult to extract from the spindle analyses. In addition spiking data from layers, 4-6 are used, but these layers are both very diverse in their behavior, and the least likely to be strongly correlated with spindle-bursts (maximal in layer 2-4). A more consistent and limited analysis of spiking is important to confirm the general vs specific nature of this quiescence.
How generalizable these results are, and how they comport with previous studies is unclear. The paper is written as if this quiescent state is universal, and its identification in two species in likely different regions adds to the argument that this is the case. However, it has not been observed in similarly detailed developmental studies in other rodent regions (multiple papers by the Hanganu-Opatz lab, Minlebeav et al Science 2011, Shen and Colonnese J Neuro 2016) nor in the clinical literature. Some more careful and nuanced discussion of the relationship between these findings or expansion of the regions surveyed to show they were wrong would help situate the current findings and better comport the claims and evidence.
-
