jackman/wholeBrain
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Reorg

Browse Source
This commit is contained in:
ackman678
2014-06-04 10:01:30 -04:00
parent 954636a1c3
commit 5c6ecb08dd
3 changed files with 31 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

BIN
figure3.png
View File

Binary file not shown.
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 289 KiB

After

Width:  |  Height:  |  Size: 944 KiB

BIN
figure4.png
View File

Binary file not shown.
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 944 KiB

After

Width:  |  Height:  |  Size: 289 KiB

50
wholeBrain_main.md
View File

@@ -2,8 +2,7 @@ Author: James B. Ackman
Date: 2013-09-04 10:54:02
Tags: paper, draft, manuscript, literature, research, #results, retinal waves, spontaneous activity, development, calcium domains
# Structured population activity across developing neocortex
Coordination of cortical activity across the developing cerebral hemispheres
# Structured population activity across developing neocortex
# Abstract
@@ -14,7 +13,7 @@ The cerebral cortex exhibits spontaneous and sensory evoked patterns of activity
# Introduction
Brain development requires neural activity and calcium dynamics for establishing proper circuit structure and function. The importance of neural activity in the prenatal and neonatal period can be easily recognized in children exposed to chemical agents affecting neurotransmission during the fetal period that result in severe brain malformations, epilepsy, and mental retardation. Indeed, embryonic limb movements in species ranging from chick to human are thought to be initiated by spontaneous motor neuron activity in the spinal cord and thought to be crucial for activity-dependent development of motor synapses [Schoenberg:2003] [Marder,Lichtmann]. However it is only recently that we have begun to appreciate the underlying patterns of persistent neural activity that in fact exist in the developing brain in vivo. For example, sensori-motor feedback associated with spontaneous movement generated by spinal motor neurons triggers synchronized 'spindle-burst' potentials among cells in somatosensory cortex [Yang:2009][Khazipov:2004a] before the start of locomotion and tactile behavior. Correlated bursts of activity occur in the developing rat hippocampus in vivo [#Leinekugel:2002] [Mohns&Blumberg]. Spontaneous retinal waves drive patterned activation of circuits throughout immature visual system before the onset of vision [#Ackman:2012] [Hanganu,Colonnese?]. Furthermore, prenatal EEG recordings have demonstrated spindle burst oscillations and slow activity transients in the human infant somatosensory and occipital cortices before birth [#Vanhatalo:2005][#Tolonen:2007]. Nonetheless, a comprehensive account of the structural dynamics of persistent activity throughout the developing isocortex in vivo has not been undertaken.
Brain development requires neural activity and calcium dynamics for establishing proper circuit structure and function. The importance of neural activity in the prenatal and neonatal period can be easily recognized in children exposed to chemical agents affecting neurotransmission during the fetal period that result in severe brain malformations, epilepsy, and mental retardation. Indeed, embryonic limb movements in species ranging from chick to human are thought to be initiated by spontaneous motor neuron activity in the spinal cord and thought to be crucial for activity-dependent development of motor synapses [Schoenberg:2003] [Marder,Lichtmann]. However it is only recently that we have begun to appreciate the underlying patterns of persistent neural activity that in fact exist in the developing brain in vivo. For example, sensori-motor feedback associated with spontaneous movement generated by spinal motor neurons triggers synchronized 'spindle-burst' potentials among cells in somatosensory cortex [Yang:2009][Khazipov:2004a] before the start of locomotion and tactile behavior. Correlated bursts of activity occur in the developing rat hippocampus in vivo [#Leinekugel:2002] [Mohns&Blumberg]. Spontaneous retinal waves drive patterned activation of circuits throughout the immature visual system before the onset of vision [#Ackman:2012] [Hanganu,Colonnese?]. Furthermore, prenatal EEG recordings have demonstrated spindle burst oscillations and slow activity transients in the human infant somatosensory and occipital cortices before birth [#Vanhatalo:2005][#Tolonen:2007]. Nonetheless, a comprehensive account of the structural dynamics of persistent activity throughout the developing isocortex in vivo has not been undertaken.
@@ -27,6 +26,10 @@ Brain development requires neural activity and calcium dynamics for establishing
Neocortical organization consists of cortical modules tiled across the cortical surface in a topographic fashion such that vertical arrays of cells concerned with specific sensory features are grouped together as columns [#Mountcastle:1997]. Most evidence to date suggests that columns/hyper/macro columns are 300-500µm diameter across species.
The patterns of activity that exist in the developing brain are unknown. We performed recordings from mice expressing the genetic calcium indicator GCaMP (gcamp3 or gcamp6) throughout cortical neurons. We performed our recordings in three age groups: P2-P5, P8-P9, and P12-13. We found that cortical activity was characterized by discrete domains of activation during the first two postnatal weeks. These activity domains ranged from 200-800 µm in diameter (*Ns*, *fig*), with larger sized domains of activation in the visual cortex and motor cortex (*Ns*, *fig*). In the second postnatal week the size of activations in the hindlimb/trunk representation was larger, The average duration of these domain activations was about 600 ms – 1 s (*Ns*, *fig*) with longer activations on the order of seconds to tens of seconds in visual cortex driven by retinal waves [#Ackman:2012].
We parcellated the brain into distinct anatomical boundaries by using reference coordinates from a mouse line that expressed the tdtomato reporter in thalamocortical afferents. The expression can be used to parcellate out areal boundearies of primary sensory cortical areas (wong riley 1979). We matched these parcellations to a Allen brain atlas adult mouse reference image and than linearly scaled the remaining parcellations in our FOV on to the images of our recordings that contain fucntional boundaries (like in the domain centroid activation plot and in the normalized domain frequency plots).
* Cortical column (mini/hyper columns) history (20th century anatomists-- Lorente de No, Mountcastle, Hubel & Wiesel, rakic, etc).
* Column physiology-- Hubel and Wiesel. Rodent V1?
@@ -38,7 +41,7 @@ Neocortical organization consists of cortical modules tiled across the cortical
* Calcium imaging-- Konnerth 'waves' in Ent cortex [#Adelsberger:2005]. For visual cortex, domains activity in extrastriate cortex (Ackman Nature 2012). But S1-- [#Golshani:2009] work in later postnatal-- but activity not obeying domains in barrel cortex-- problem with spatial sampling in the xy and the z for this study?
![**Figure 1.** Calcium domains throughout neonatal mouse isocortex. **a** Experimental schematic. **b** Single image frame showing calcium domains in both hemispheres at P3 and automatically detected domain masks. **c** Domain centroid positions for 10 min recording. Points are overlaid on a normalized pixel activation frequency map and primary sensory areas determined by thalamocortical inputs at P7 are outlined in red. Notice rows of whisker barrels are evident in the structure of domain centroid positions. **d** Functional activity map at P3. Based on pixel activation frequency from detected domains in a single 10 min recording. Map is overlaid on cortical areal parcellations. Notice localized maxima and minima of functional activity between areas that approximate known anatomical cortical area boundaries and the mirroring of map structure bilaterally. **e** Histograms showing the distribution of spatial diameters and durations for calcium domains. **f** Domain duration map.](figure1.png)
![ **Figure 1.** Calcium domains throughout neonatal mouse isocortex. **a** Experimental schematic. **b** Single image frame showing calcium domains in both hemispheres at P3 and automatically detected domain masks. **c** Domain centroid positions for 10 min recording. Points are overlaid on a normalized pixel activation frequency map and primary sensory areas determined by thalamocortical inputs at P7 are outlined in red. Notice rows of whisker barrels are evident in the structure of domain centroid positions. **d** Functional activity map at P3. Based on pixel activation frequency from detected domains in a single 10 min recording. Map is overlaid on cortical areal parcellations. Notice localized maxima and minima of functional activity between areas that approximate known anatomical cortical area boundaries and the mirroring of map structure bilaterally. **e** Histograms showing the distribution of spatial diameters and durations for calcium domains. **f** Domain duration map.](figure1.png)
metric | mean | min | max | unit
------------- | ----- | ---- | ------ | --------------------
@@ -59,9 +62,11 @@ frequency | 2.9 | | | domains/sec/hemisphere
* Regional effects, more corr anticorr in certain regions?
* State dependent corr?
To understand the patterns and how they interact we first looked at correlation between the hemispheres. Cortical activity exhibited high temporal cortelation between the hemispheres () . In additon this activity was highly correlated in the spatial dimension. For example epochs of time would exhibit high correlation in the medial-lateral dimension or in the rostral-caudal dimension. This strength of correlation temporally and spatially increased between the hemsipheres with a function of age.
<!--- * Each hemisphere 'training' the other one in preparation for behaviorally relevant sensory-motor imitations '[[mirror_neurons]]' hypothesis? --->
![**Figure 2.** Cortical domain activity exhibits bilateral symmetry. **a** Examples of domains exhibiting spatially symmetric activations. Notice most timepoints contain a mixture of symmetric and asymmetric domain activations. **b** Cortical active fraction timecourses for both hemispheres. **c** Hemispheric domain centers of mass for coactive frames in a recording along medial-lateral (ML) and anterior-posterior (AP) extents. Bottom left panels show the periods indicated by black bars at expanded view. Pearson's correlation: ML, p = 1.1591e-28; AP, p = 7.0982e-07. **d** Hemispheric autocorrelation and cross-correlation functions for cortical activity at all and short time lags. Notice the peaks above gaussian distributed noise (blue traces).](figure2.png)
![ **Figure 2.** Cortical domain activity exhibits bilateral symmetry. **a** Examples of domains exhibiting spatially symmetric activations. Notice most timepoints contain a mixture of symmetric and asymmetric domain activations. **b** Cortical active fraction timecourses for both hemispheres. **c** Hemispheric domain centers of mass for coactive frames in a recording along medial-lateral (ML) and anterior-posterior (AP) extents. Bottom left panels show the periods indicated by black bars at expanded view. Pearson's correlation: ML, p = 1.1591e-28; AP, p = 7.0982e-07. **d** Hemispheric autocorrelation and cross-correlation functions for cortical activity at all and short time lags. Notice the peaks above gaussian distributed noise (blue traces).](figure2.png)
@@ -69,7 +74,7 @@ frequency | 2.9 | | | domains/sec/hemisphere
The big secondary peaks around ±30 sec is present in both autocorrs and xcorrs and is far above the random normal xcorr baseline (blue trace). In fact there is a periodicity seen in the autocorrs and the xcorrs where there is a dampening oscillation about on this interval! (See ideal dampening frequency in random sine wave example above). This corresponds to a 1/30sec == 0.033 Hz ultra-slow oscillation.
Looking at the above plot showing lags from [–1000, 1000] frames which is ± 200 s, we can see about 5.5 cycles of this underlying dampening oscillation in both autocorr plots. This corresponds to (1000fr*0.2sec/fr)/5.5 => 36.36 sec/cycle => 0.0275 cycles/sec or ~0.03 Hz
{==Looking at the above plot showing lags from [–1000, 1000] frames which is ± 200 s, we can see about 5.5 cycles of this underlying dampening oscillation in both autocorr plots. This corresponds to (1000fr*0.2sec/fr)/5.5 => 36.36 sec/cycle => 0.0275 cycles/sec or ~0.03 Hz==}{>>Crazy!<<}
**Conclusions:** So the activity in both hemispheres at postnatal day 3 (P3) clearly exhibits significant spatial correlations in both in the medial-lateral and anterior-extent. This is consistent with and complementary to the fact that the active pixel fraction in each hemisphere exhibits a strong temporal correlation as I found earlier in this report [Temporal correlation of activity][]. The medial-lateral positional correlation is stronger than the anterior-posterior (higher *R* and lower *p* value). The total number of coactive frames is `numel(y1(~isnan(y1)&~isnan(y2)))` == **1114 frames**. This is accounts to **37.13%** of the movie or **222.8 s**. Cortex.L had 1635 actvFrames and cortex.R had 1677 actvFrames which means that each hemisphere was coactive with the other hemisphere 1114/1635 == **68.13%** and 1114/1677 == **66.43%** of the active time respectively.
@@ -78,8 +83,6 @@ Looking at the above plot showing lags from [–1000, 1000] frames which is ± 2
### Percentage of cortical activity which exhibits corr with motor signal?
lenActvFraction>0 | fracCorr | timeCorr_s | fracCorrPos | timeCorrPos_s | fracCorrNeg | timeCorrNeg_s
@@ -89,29 +92,38 @@ lenActvFraction>0 | fracCorr | timeCorr_s | fracCorrPos | timeCorrPos_s | fracCo
## Developing cortical activity consists of distinct subnetworks
![**Figure 3.** Subnetworks in developing isocortex. **a** Areal trace examples. Synchronized activations between hemispheres indicated by red links. **b** Correlation matrix of domain activity among cortical areas. Colormap indicates Pearson's r correlation coefficient values. **c** Map of cortical area associations for r > 0.2. **d** Graph of functional connections for r > 0.1. Community structure detected with hierarchical clustering are indicated in 4 colors. Notice that S1-limb/body regions form a distinct subgraph (green).](figure3.png)
## Cortical domain activity is state dependent
* EEG slow oscillations not detectable until P10 in rodent.
* Previously demonstrated that general anesthesia abolishes spontaneous activity in visual system [#Ackman:2012].
* What about ongoing activity in other cortical areas during early brain development? Surgical procedure relevance.
* No population calcium activity found during gen'l anesthesia, only slow traveling waves.
* During anesthesia induction, rapid (<30 s) knock down of discrete domain activity (P3 mouse <120518_09.tif>). Cingulate, retrosplenial activations the last to go-- default mode/resting state network areas last.
* During anesthesia induction, there is rapid (<30 s) knock down of discrete domain activity (P3 mouse <120518_09.tif>). Cingulate, retrosplenial activations the last to go-- default mode/resting state network areas last.
<120518_09_mjpeg.mov>
![**Figure 4.** Cortical domains are state dependent. **a** Experimental schematic. Red light illumination measured with a photodiode (PD) was used to monitor motor activity. **b** Cortical activity (active fraction) in each hemisphere after onset of gas anesthetic. **c** Cortical activity and coincident motor activity signals. Gray shading indicates active and quiet motor periods determined by the half-rise and decay times from peak in the low pass filtered motor signal. Active pixel fraction traces for motor (M1,M2), somatosensory (HL,FL,T; barrel), and visual (V1,V2) cortex shown at bottom of panel. Red links show synchronized motor movements and brain activity with different cortical regions. **d** Single frame domain masks for times indicated in **c**. **e** Autocorrelation and cross-correlation functions for motor and all cortical activity or just S1-limb/body and motor cortex regions. Notice the general negative correlation between motor activity and all cortex activity signals (r = -0.1368, p = 5.1919e-14) and the high positive correlation between motor and S1-limb/body signals (r = 0.3019, p = 2.7861e-64). **f** Pixel activation frequency maps during quiet and active motor periods.](figure4.png)
Variation in the strength of correlation between cortical areas and the motor movement signal depended on brain region (p < 2.2e-16, anova) and age (p = 1.627e-05, anova) The first age group in which motor cortex exhibited signficant positive correlation with motor movements was at P12-13 (r=0.06±0.02, p-value = 0.001449, t-test).
![ **Figure 3**. Cortical domains are state dependent. **a** Experimental schematic. Red light illumination measured with a photodiode (PD) was used to monitor motor activity. **b** Cortical activity (active fraction) in each hemisphere after onset of gas anesthetic. **c** Cortical activity and coincident motor activity signals. Gray shading indicates active and quiet motor periods determined by the half-rise and decay times from peak in the low pass filtered motor signal. Active pixel fraction traces for motor (M1,M2), somatosensory (HL,FL,T; barrel), and visual (V1,V2) cortex shown at bottom of panel. Red links show synchronized motor movements and brain activity with different cortical regions. **d** Single frame domain masks for times indicated in **c**. **e** Autocorrelation and cross-correlation functions for motor and all cortical activity or just S1-limb/body and motor cortex regions. Notice the general negative correlation between motor activity and all cortex activity signals (r = -0.1368, p = 5.1919e-14) and the high positive correlation between motor and S1-limb/body signals (r = 0.3019, p = 2.7861e-64). **f** Pixel activation frequency maps during quiet and active motor periods.](figure3.png)
Low pass filtered Moving averages of cortical and motor activity at 10 s and >70 s windows.
## Developing cortical activity consists of distinct subnetworks
We then calculated a matrix of pearsons correlation coefficients based on the pixel active fraction timecourses for each pair of parcellations. The resulting assocaition matrix was run through a hierarchal clustering alogtithm to reveal functional modules of of activation. These functional modules typically consisted of 3 distinct subnetworks-- a frontal motor network, a posterior parietal network, a S1-body/limb network, and an auditory A1 network at P12.
We found many similarities but some striking differences as a function of age.
![ **Figure 4.** Subnetworks in developing isocortex. **a** Areal trace examples. Synchronized activations between hemispheres indicated by red links. **b** Correlation matrix of domain activity among cortical areas. Colormap indicates Pearson's r correlation coefficient values. **c** Map of cortical area associations for r > 0.2. **d** Graph of functional connections for r > 0.1. Community structure detected with hierarchical clustering are indicated in 4 colors. Notice that S1-limb/body regions form a distinct subgraph (green).](figure4.png)
# Conclusions
* Neural population activity constitutes discrete spatial and temporal activations among developing cortical areas