Rapid nucleus-scale reorganization of chromatin in neurons enables transcriptional adaptation for memory consolidation

The interphase nucleus is functionally organized in active and repressed territories defining the transcriptional status of the cell. However, it remains poorly understood how the nuclear architecture of neurons adapts in response to behaviorally relevant stimuli that trigger fast alterations in gene expression patterns. Imaging of fluorescently tagged nucleosomes revealed that pharmacological manipulation of neuronal activity in vitro and auditory cued fear conditioning in vivo induce nucleus-scale restructuring of chromatin within minutes. Furthermore, the acquisition of auditory fear memory is impaired after infusion of a drug into auditory cortex which blocks chromatin reorganization in vitro. We propose that active chromatin movements at the nucleus scale act together with local gene-specific modifications to enable transcriptional adaptations at fast time scales. Introducing a transgenic mouse line for photolabeling of histones, we extend the realm of systems available for imaging of chromatin dynamics to living animals.


Introduction
Interphase chromatin is organized in a highly ordered spatial structure with individual chromosomes occupying discrete chromosome territories within the nucleus [1][2][3][4]. This threedimensional nuclear architecture is believed to reflect the transcriptional state of the cell as the positioning of individual genes relative to transcriptional hotspots or the nuclear lamina is associated with their level of expression [5][6][7][8][9][10][11]. Using microscopy, it has been demonstrated that the functional organization of interphase chromatin can undergo nucleus-scale plastic remodeling [12][13][14][15][16][17][18][19] linked to changes in gene expression [12,[20][21][22][23][24][25][26][27][28][29]. Most of this work relied on the experimental accessibility of cultured cells.   50 kHz and high-pass filtered at 200 Hz. Spikes were detected using a threshold-based detector set to a threshold of seven times the standard deviation of the noise level (MC_Rack, Multi Channel Systems). Electrophysiological recordings were performed in culture medium and temperature was maintained at 37˚C by a temperature controller (TC02, Multi Channel Systems). Spike datasets from all electrodes were imported into Matlab 7.7 (Mathworks, Natick, MA, USA) for analysis of single units using a custom written routine. Spike sorting was carried out as described previously [44]. Autocorrelation functions were applied to confirm spike sorting. If single units fired at least twice within the recording period, units were counted as active neurons. Average firing frequencies were calculated as arithmetic mean of individual firing frequencies of all identified units and number of active neurons was defined as number of individual single units with at least two action potentials per recording.

Immunocytochemistry of primary neurons (Figs 2, S1 and S2)
Dissociated cells grown on coverslips were fixed in 2% PFA in PBS for 5 minutes, then transferred to 4% PFA for an additional 15 minutes and then washed with PBS. Unspecific binding of antibodies binding was blocked with 7% normal donkey serum and 0.3% Triton diluted in PBS for two hours at RT. Overnight staining was performed at 4˚C with primary antibody diluted in 2% bovine serum albumin with 0.05% azide and 0.1% Triton. Thereafter, cells were washed three times with PBS and incubated with 1 μg/ml DAPI (Sigma-Aldrich), Alexa Fluor647-and DyLight488-coupled secondary antibody (Biomol, Hamburg, Germany) diluted in 2% bovine serum albumin with 0.05% azide in PBS for two hours at RT. Coverslips were washed in PBS and specimens were mounted with Fluoromount (Sigma-Aldrich). The following antibodies were used: Rabbit monoclonal anti c-Fos (Cell Signaling Technology, Inc., San-Diego, CA, USA) 1:800, Rabbit monoclonal anti γH2A.X (phosphor S139) antibody (Abcam, Cambridge, UK) 1:400, guinea pig polyclonal anti c-Fos (Synaptic Systems, Goettingen, Germany) 1:500. Imaging was done on an Olympus IX81 epifluorescence microscope (Olympus Life Sciences, Germany) or on a spinning disc confocal (Microscope (VisiScope 5-Elements, Visitron Systems, Germany) and images were subsequently analyzed with ImageJ or Matlab.

Live cell imaging of primary neurons (Figs 1 and 2)
Primary neurons transduced with recombinant AAV carrying either pAAV-hSyn-H2B:: mCherry alone or pAAV-hSyn-H2B::mCherry in combination pAAV-hSyn-H2B::PaGFP after one day in culture. Live cell confocal imaging of H2B::mCherry and H2B::PaGFP was performed on DIV 14-16. Images were acquired with a Visiscope 5-Elements spinning disk confocal system (Visitron Systems, Germany), and a 60x water immersion objective (CFI Plan Apo VC 60XWI, 1.2 NA). The system was equipped with a Yokogawa CSU-W1 scan head, a Prime BSI sCMOS camera (2048 x 2048 pixels, 6.5 μm pixel size, Photometrics), a FRAP module that was used for photo-activation of PaGFP, and with an incubator (Bold Line universal stage top incubator, okolab, Italy). Theoretical lateral and axial resolution was 163 nm and 490 nm, respectively, in a field of view of 228 x 228 μm. The laser lines of 488 nm and 561 nm were used for the fluorescence excitation and the fluorescence emission was filtered using filters 525/30 bandpass (Chroma) and 570 longpass (Chroma) for H2B::PaGFP and H2B::mCherry, respectively. The temperature in the incubator was kept at 37˚C and the CO 2 level at 5% during imaging. Penicillin and streptomycin (5 units/ml) was applied prior to the imaging session to avoid bacterial contamination. Cells were imaged every ten minutes for one hour before the application of pharmacological treatment, and up to two hours after the application. Cells were fixed immediately after the live cell acquisition for immunocytochemical analysis.
For experiments with primary cortical neurons co-expressing H2B::mCherry and H2B:: PaGFP, photolabeling of a narrow rectangle in the nucleus was performed with a 405 nm laser.

Calculation of the granularity index (Figs 1 and 2)
For each image stack of a nucleus, the optical section with the largest signal intensity was selfnormalized. A two-dimensional Fourier transform was calculated and a Gaussian function was fitted onto the two cardinal axes of the 2D Fourier transform. The full width half maximum of the average of the two Gaussian fits was termed Granularity Index (GI), which was normalized for each nucleus to its first time point.

Calculation of the circularity index (Figs 2-5 and S6-S8)
To quantify the remodeling of the fluorescence label, we modified a previously established method to measure the geometry of dendritic spines [45]. Briefly, maximum intensity projections of the image stacks of individual nuclei were background subtracted and normalized to maximum intensity values and only intensity values >0. 25 were considered for further analysis. A principal component analysis of the 2D pixel map was performed, where each pixel was weighted by its brightness. The circularity index (CI) was defined as the ratio of the smaller to the larger eigenvalue.

Photolabeling in fixed brain slices (S3B Fig)
To test the specificity of H2B::PaGFP expression and effectiveness of the stop-cassette, H2B:: PaGFP mice were crossed with EMX1-Cre mice. Cre-positive and Cre-negative littermates were transcardially perfused with PBS containing 10U/ml Heparin (Sigma-Aldrich) and 4% PFA, the brains removed and post-fixed with 4% PFA overnight at 4˚C. Coronal brain slices (70 μm thick) were cut with a vibratome (Leica Biosystems, Germany; VT-1000). Full frame photolabeling and subsequent imaging of PaGFP fluorescence was performed on a LSM780 confocal microscope. The full frame was illuminated using a 405nm laser and subsequent imaging of photolabeled neurons was performed using a 488nm laser.

Correlation of PaGFP expression and DAPI labeling (S3C Fig)
H2B::PaGFP mice were injected with a rAAV encoding for Cre recombinase under the control of a CMV promoter. Three weeks later cortical slices were prepared as described above. They were stained with 1 μg/ml DAPI, photolabeled and imaged using an LSM780 confocal microscope.

Craniotomy
Male and female H2B::PaGFP mice were used for in vivo imaging experiments and a small imaging window was implanted over the auditory cortex three weeks after virus injection. Briefly, mice were deeply anesthetized with a single dose of ketamine/medetomidine. After removal of an about 1 cm 2 patch of skin over the parietal bones on the right lateral part of the head, a small part of the musculus temporalis was removed to expose the temporal bone. Using a dentist's drill, the bones were smoothened, and part of the zygomatic process was removed and covered with a thin layer of Vetbond glue (3M). Next, a thin layer of dental cement (Lang Dental) was applied, except for the area over the temporal bone. The temporal bone was removed with a dental drill to expose a~2 mm × 3 mm part of the brain containing the auditory cortex. The craniotomy was subsequently sealed with a drop of liquid agarose, a small round cover glass, and additional dental cement. In addition, a small titanium head post was implanted for fixation of the head during imaging. activity during cocktail application. Scale: 40 mV, 2 min; scale inset: 30 mV, 100ms. (C) Analog control to the experiment shown in panel (b), however, normal ACSF was perfused during both baseline and application period. Note low activity levels in normal ACSF. (D) Quantification of burst events during baseline and application period in recordings from slices bathed with cocktail (red) and ACSF (black). (E) RNAseq following incubation (cocktail vs. control, six slices each) revealed a subcluster of significantly induced genes including many known IEGs. Additionally, corresponding relative expression levels of the genes are displayed that were observed in the auditory cortex following various behavioral treatments (data from [34]). Expression levels of some genes were measured with several probes on the microarray and are therefore mentioned multiple times. (F) Top panel: Timeline of the experiment to study chromatin dynamics in brain slices in the control group. Lower panels: Examples of time-lapse of two-photon image stacks (maximum intensity projections) of individual photolabeled neurons, sorted by their respective CIs at 120 min.

In vivo two-photon imaging (Figs 4, 5 and S3D)
In vivo imaging and photolabeling was performed using an Ultima in vivo multiphoton microscopy system (Prairie Technologies, WI, USA) with a 20x objective lens (Olympus, Tokyo, Japan; XLUMPlan FI, n.a. = 0.95) and a Chameleon Ultra Ti:Sapphire multiphoton laser (Coherent, CA, USA). Mice were anesthetized with isofluorane (Abbot Animal Health, IL, USA; IsoFlo) and H2B::PaGFP positive cells were identified by their weak basal For photolabeling of the whole nucleus, a ROI was placed over the soma. Photolabeling was performed at 750nm and imaging of photolabeled cells at 950nm. Photolabeling was done on a single focal plane and images were collected as z-stacks. To estimate the longterm stability of the photolabel in vivo, a ROI was placed over the soma of the neuron and the neuron was photolabeled. Individual labeled neurons were revisited and imaged at different time points at the indicated intervals. For analysis, images were background subtracted and the fluorescence was normalized to the fluorescence level measured directly after photolabeling.

Acute brain slice preparation and electrophysiology (Figs 3A-3D and S4)
4-week-old mice were deeply anesthetized using isoflurane (Abbot Animal Health, IL, USA; IsoFlo), decapitated and their brains quickly chilled in ice-cold oxygenated (95% O2/5% CO2) dissection buffer (110 mM choline chloride, 25 mM NaHCO3, 1.25 mM NaH2PO4, 2.5 mM KCl, 0.5 mM CaCl2, 7 mM MgCl2, 11.6 mM ascorbic acid, 3.1 mM pyruvic acid, and 25 mM D-glucose). Coronal brain slices (350μm) containing auditory cortex were cut in ice-cold dissection buffer using a Vibratome (Leica Biosystems, Germany; VT1000S) and immediately incubated for 15 min in dissection buffer at 32˚C. This was followed by a phase in which slices rested in oxygenated (95% O2/5% CO2) artificial cerebrospinal fluid (aCSF; 118 mM NaCl, 2.5 mM KCl, 26.5 mM NaHCO3, 1 mM NaH2PO4, 1 mM MgCl2, 2 mM CaCl2, and 20 mM Dglucose) for 45min at room temperature (RT). Individual brain slices were placed on the stage of an upright, infrared-differential interference contrast microscope (Olympus BX50WI) mounted on a X-Y experiments investigating incubation with pharmacological cocktail, coronal brain slices containing auditory cortex were prepared as described above. After the initial 15 minutes of recovery in Choline Chloride, the two hemispheres were separated by a cut along the midline under a dissection microscope and incubated in aCSF for 45 minutes. One hemisphere of each brain slice was then transferred to fresh aCSF for 1 hour; the other hemisphere was incubated in cocktail aCSF for 1h. Both, control aCSF and cocktail aCSF hemispheres, were finally exposed to another resting phase in bubbled aCSF for 1 hour. Solutions for each step described above were oxygenated with 95% O2/5% CO2 at 32˚C.

c-Fos immunohistochemistry in coronal brain slices (S6 Fig)
c-Fos/DAPI staining was performed on coronal brain slices incubated in normal aCSF or cocktail aCSF as described above. After fixation in 4% PFA at 4˚C for 2 hours, free floating slices were permeabilized with 0.2% Triton X-100 and blocked with 2% normal goat serum and 2% BSA for 2 hours at room temperature. Slides were then incubated with a primary antibody against c-Fos (1:1000, rabbit; Abcam, Cambridge, UK; ab7963) for 24 hours at 4˚C, followed by washing and incubation with an Alexa-488 conjugated secondary antibody (1:1000, goat, Invitrogen, CA, USA; A11008) against rabbit at room temperature. Labeled probes were cover slipped with DAPI-Fluoromount-GTM (Biomedica, Vienna, Austria). Images were acquired using the Zeiss LSM 700 point scanning confocal microscope. For detection and quantification of c-Fos signal, a semi-automated, machine learning based approach was implemented with the Pannoramic Viewer image processing software (3D Histech, Budapest, Hungary; Developer XD, Definiens, Munich, Germany). The DAPI signal was used as nuclear reference for the immunodetection of (nuclear) c-Fos. Nuclei were segmented based on the DAPI channel using a LoG-Filter and watershed algorithms. On a set of training images, nuclei objects were manually classified as positive and negative and used as samples for training a machine-learning algorithm. Mean intensity, standard deviation and local contrast parameters were extracted from these as input for a decision tree-based classifier. This classifier was then applied to all nuclei objects on a larger set of images to identify positive cells. A similar approach was used to segment and classify cytoplasmic stain around the nuclei. Training was done separately for each channel. The output was given as number of positive cells in each channel, either c-Fos or DAPI, co-staining of cells, total cell number and mean intensity per cell for further analysis. To quantify differences in cumulative fluorescence probability of c-Fos between control and cocktail samples, average c-Fos fluorescence values from each DAPIstained nucleus were normalized to the individual background of each image using R (R foundation, Vienna, Austria). Fluorescence data from all mice in each group were pooled to generate an empirical cumulative distribution function (ECDF) for each group.

Imaging chromatin dynamics in coronal brain slices (Figs 3F-3K and S8)
An individual slice was transferred to a slice recording chamber in an Ultima multiphoton microscopy system (Prairie Technologies, WI, USA), gently immobilized by a silver grid with attached nylon mesh, and fully submerged and continuously perfused with oxygenated aCSF at room temperature. Imaging was performed using a 20x objective lens (XLUMPlan FI, 0.95 N.A., Olympus) and a Ti:Sapphire multiphoton laser (Coherent, CA, USA). Neurons of layer II/III were identified by baseline PaGFP fluorescence at 950 nm, and were labeled by applying 750 nm laser light on a rectangular ROI over the nucleus at a single focal plane. After labeling of a population of neurons, z stacks were acquired at 0.5 μm steps. Imaging parameters were: 1.2 us pixel dwell time, 2048x2048 pixel resolution at 0.1344 microns per pixel. Photolabeled populations of neurons were revisited and imaged every hour for six hours. After one hour of aCSF perfusion, experimental slices were perfused with aCSF including the pharmacological cocktail described above for one hour and returned to normal aCSF for the rest of the experiment. Slices in the Cocktail + BDM group were treated with 20 mM 2,3-butanedione monoxime (Sigma) in addition to the described pharmacological cocktail. Control slices were perfused with normal aCSF during the whole course of the experiment including a 'mock switch' at the same timing as the experimental groups. Individual labeled neurons were manually identified in z-stacks of different time points and isolated by cropping of ROIs of 100x100 pixels. Maximum intensity z-projections were used for calculation of the CI using the same custom written Matlab (Mathworks, Natick, MA, USA) script as for the in vivo data.

RNA preparation, sequencing, and expression analysis (Figs 3E and S7)
Brain slice hemispheres that underwent incubation protocol in cocktail or control aCSF, as described above, were placed under a dissection microscope and auditory cortices dissected while submerged in bubbled aCSF. Total RNA from dissected auditory cortices was isolated by Trizol (800μl/auditory cortex; Life technologies, Grand Island, NY, USA). Total RNA was quantified and quality assessed with a Bioanalyzer (Agilent Technologies, CA, USA) RNA 6000 Nano kit. One microgram of total RNA was used for poly-A selection with a Dynabeads mRNA purification kit (Invitrogen, CA, USA) followed by cDNA generation (NEB 1st strand and NEB 2nd strand ultra directional Kits; New England BioLabs, Frankfurt am Main, Germany). Libraries were prepared using a NEBNext Ultra DNA kit (New England BioLabs, Frankfurt am Main, Germany) and quantified with a Library Quantification kit for Illumina platforms (Kapa Biosystems, Inc., Wilmington, MA, USA). Sequencing was performed on a HiSeq 2500 system (Illumina, San Diego, CA, USA) following the manufacturer's guidelines in single read modus and at a read length of 50bp. All samples were multiplexed in one lane returning 15 to 27 million reads with an average Q score above 30 per sample. The strand specific reads were screened for ribosomal RNA by aligning with BWA (v0.6.1) against known rRNA sequences (RefSeq). The rRNA subtracted reads were aligned with TopHat (v1.4.1) against the Mus musculus genome (mm10) and a maxiumum of 6 missmatches. Maximum multihits was set to 1 and InDels as well as Microexon-search was enabled. Additionally, a gene model was provided as GTF (UCSC RefSeq mm10). rRNA loci are masked on the genome for downstream analysis. Aligned reads are subjected to FPKM estimation with Cufflinks (v1.3.0). In this step bias detection and correction was performed. Furthermore, only those fragments compatible with UCSC RefSeq annotation (mm10) of genes with at least one protein coding transcript were allowed and counted towards the number of mapped hits used in the FPKM denominator. Furthermore, the aligned reads were counted with HTSeq (0.6.1p1) and the genes were subjected to differential expression analysis with DESeq2 (v1.6.3; Bioconductor). The effect of samples originating from the same mouse was blocked.

Auditory cued fear conditioning and memory testing (Figs 6 and S5)
Behavioral experiments were performed in an isolation cubicle (

Intrinsic imaging (Fig 5B)
Mice were lightly anesthetized with isofluorane (Abbot Animal Health, IL, USA; IsoFlo) and the area under the cranial window was imaged with a 780nm light emitting diode and a CCD camera (Vosskühler, Osnabrück, Germany). Sound stimuli consisted of white noise, pure tones and the complex sound used for auditory cued fear conditioning. Sounds were presented 30 times and the change in light reflectance was computed and averaged using custom Matlab (Mathworks, Natick, MA, USA) software.

Imaging chromatin dynamics in vivo (Figs 4 and 5F)
For in vivo chromatin imaging, large nuclei in layer II/III were identified for labeling using PaGFP baseline fluorescence. Photolabeling was performed sound responsive areas, as determined via intrinsic imaging. For photolabeling, a 5x zoom factor was used (x,y-pixel size 0.215 x 0.215 microns) and a small ROI in the pattern of a thin stripe was placed over the nucleus. Then a single focal plane was illuminated with 750nm leading to a thin fluorescent rod in the nucleus. Photolabeled cells were imaged as z-stacks (1um step size) in duplicates.To analyze the three-dimensional spread of fluorescence and to assess the three-dimensional pattern in labeled cells, lines were photolabeled and brains fixated after 1h or 18h. When performing photolabeling using two-photon excitation, the minimal extent of the label is determined by the point-spread function of the focus. The point spread function is affected by the specific imaging conditions in vivo and is typically in the submicron range in the xy-dimension, whereas it extends to about 2 μm along the z-dimension. Thus, the extent of the photolabeled lines along the z-axis is likely larger than their lateral extent. As in many cells a clear rotational component became apparent, we wondered if the transformation of the line pattern to a more homogeneous labeling of the nucleus can be accounted for by a rigid 3D rotation of the whole chromatin along the x or y axis. Labeled nuclei were re-identified in fixed brain slices and were imaged at high magnification using an LSM780 confocal microscope. Images were deconvolved using Huygens software (Scientific Volume Imaging (SVI), Hilversum, Netherlands). After deconvolution, image stacks of individual nuclei were re-sampled to obtain an isotropic resolution of 0.368 μm for all three dimensions. Again, pixel intensities were background subtracted and normalized to maximum intensity values. A 2D principal component analysis was performed on a maximum intensity projection as described above. To assess the three-dimensional pattern a principal component analysis of the 3D pixel map was performed considering intensity values >0.1 and the eigenvalues calculated describing the spread of fluorescence along the three principal axes.

Tracking of chromatin dynamics in vivo using bolus-loading of cells with Hoechst (S12 Fig)
A craniotomy was performed on 8-week-old mice and 200nl of a 10 μg/ml solution of Hoechst 33342 (ThermoFisher) was injected in multiple spots in 200-300 μm depth. Image stacks of stained cells were acquired in 15 minutes intervals for three hours in 200-300 μm depth at 750 nm wavelength.

Chromatin photolabeling and behavior (Fig 5D-5H)
Mice were habituated for 3 days prior to the experiment. This included handling by the experimenter, placing into the behavior chambers, anaesthesia with isoflurane (Abbot Animal Health, IL, USA; IsoFlo) and head fixation in the two-photon microscope. The behavior and imaging paradigm lasted for five days (anesthesia, context, unpaired, paired, and memory test). Anesthesia (day one): Mice were anesthetized and placed into the two-photon microscope. Nuclei were labeled with a line pattern and z-stacks of the population were acquired (Pre). The mice remained under the microscope and the same neurons were imaged again two hours later (Post). Context (day two): Mice were anesthetized and placed into the two-photon microscope. A different group of neurons from the first day were labeled with a stripe and zstacks of the population were acquired (Pre). Afterwards, the mice were removed from the microscope, put back into their cage and allowed to wake up. Two hours after the first imaging session, mice were anesthetized, placed back under the microscope and stacks of the same ensemble of neurons were recorded (Post). Unpaired (day three): Mice were anesthetized and placed into the two-photon microscope. Another ensemble of nuclei was labeled with a stripe and imaged (Pre). Afterwards the mice were put back into their cage and allowed to wake up. Then the mice were subjected to one unpaired conditioning session and put back in their cage. Two hours after the first imaging session, mice were anesthetized and put back into the microscope and stacks of the same ensemble were recorded (Post). Paired (day four): Mice were anesthetized and placed under the two-photon microscope. Another ensemble of nuclei was labeled with a stripe and z-stacks of the population were recorded (Pre). Afterwards, the mice were put back into their cage and allowed to wake up. Then the mice were subjected to one paired conditioning session and put back in their cage. Two hours after the first imaging session, mice were anesthetized and put back under the two-photon microscope and stacks of the same neurons were recorded (Post). 24 hours after the paired session, the acquisition of an associative memory was assessed in a memory test session (day five).

Guide cannula implantation and infusion combined with auditory cued fear conditioning and memory testing (Fig 6)
Male 2-4 month old C57BL/6J mice (Charles River) were stereotaxically implanted with 26GA guide cannulas bilaterally (PlasticsOne) 1mm above infusion target area under isoflurane anesthesia. Cannulas were mounted on the skull with dental cement (Super-Bond C&B, Sun Medical) and animals let recover for 1 week. Animals received analgesics (Carprofen, Rimadyl) via drinking water post-surgery. Bilateral infusion of 250nl aCSF (40nl/min each) or 250nl 2,3-butanedione monoxime (BDM, 20 mM) via a 33GA infusion cannula (PlasticsOne) by an infusion pump (Harvard apparatus) was done 20 minutes prior behavioral experiments. All behavioral experiments were performed in an isolation cubicle which was equipped with white LEDs as house light and a CCD camera (Basler acA1300-30gc) with infrared LEDs. The conditioning chamber (17 × 17 × 42 cm) was combined with either a grid floor (context A) or a cylindrical inset and a stainless-steel shock grid (context B). Mice were habituated in a context A and context B on day 1. On day 2 mice underwent auditory cued fear conditioning 1 in context B after infusion of aCSF (Group 1) or BDM (Group 2), where each of 5 sound presentations (2s, 70dB) was immediately followed by a foot-shock (0.75 mA, 1 second; Precision Animal shocker, Coulbourn Instruments) with randomized inter-stimulus time intervals ranging from 50 to 75 seconds. On day 3 memory was tested in context A (recall 1) by 2 blocks of the 20s sound (10x 2s) after infusion of aCSF to both groups. On day 4 mice were re-conditioned (conditioning 2) in context B after aCSF infusion to both groups, followed by a memory test (recall 2) after infusion with inverted group treatments, compared to conditioning 1, on day 5. Freezing was tracked in Observer v11 (Noldus) by an experimenter blind to the treatment. Freezing values are given as % time of the duration of all freezing bouts during the respective experimental periods (baseline and CS periods). Baseline freezing was assessed during silence between 30 and 60 seconds of each protocol run.

Post hoc immunohistochemistry of in vivo photolabeled neurons (S10 Fig)
After the photolabeling session, mice were transcardially perfused as described with a twohour postfix protocol in 4% PFA. 70 μm thick slices were cut on a vibratome (VT-1000, Leica) parallel to the imaging plane. Free floating sections were incubated for 2 hours at room temperature in PBS containing 10% normal goat serum (NGS) and 1% Triton-X 100. They were washed with PBS and incubated overnight at 4˚C in PBS with 5% NGS, 0.1% Triton-X 100 and an anti NeuN antibody (raised in guinea pig). On the next day, sections were washed three times with PBS for ten minutes at room temperature and incubated for two hours at room temperature in PBS containing 5% NGS, 0.1% Triton-X 100 and an anti-guinea pig secondary antibody coupled to Alexa Fluor 647. Next, sections were washed three times with PBS for 10 minutes at room temperature and coversliped. Photolabeled cells could be readily identified in the slice by eye based on their strong fluorescence signal. Confocal stacks were obtained on a LSM 780 confocal laser scanning microscope (Zeiss) and co-localisation was analyzed using ImageJ.

Statistical analysis
The statistical analyses were performed using Graphpad Prism 6 (Graphpad Software, La Jolla, CA, USA) and Matlab (Mathworks, Natick, MA, USA). We used an appropriate statistical test for the given statistical analysis. For the data to analyze three dimensional distribution of the chromatin label, we used an unpaired t-test with multiple comparisons. For the comparison of the immunohistochemical analysis of c-Fos expression in brain slices, we used a Mann-Whitney test. For all comparisons of the circularity index between groups and time points, we used a two-way ANOVA with a post-hoc Sidak's multiple comparison test. All values are given as mean±sem unless otherwise noted. Statistical parameters including the exact value of n, precision measures and statistical significance are reported in the text, the figures, and the figure legends. The significance threshold was placed at α = 0.05 (n.s., p>0.05; � , p<0.05).

Histological verification of cannula position (S14 Fig)
Mice were transcardially perfused with 4% PFA in PBS. The brains were extracted, and fixed with 4% PFA over night at 4˚C. The next day, brains were coronally sliced and stained with DAPI, as described above. Epifluorescence images of the section with the largest extent of cannula damage were acquired with an AF7000 Widefield Fluorescence Microscope (Leica, Germany) and an N PLAN 5x/0.12 DRY objective to estimate the position of the cannula posthoc. The final position of the cannula in the tissue was estimated from the surface of the brain and the length of the cannula and animals with a mis-hit were excluded (group 1: 12/16 included, group 2: 10/16 included).

Nucleus-scale chromatin dynamics in primary cortical neurons
To investigate and visualize nucleus-scale chromatin dynamics in neurons upon induction of neuronal activity, we followed a widely used strategy that preserves endogenous nucleosome organization [46]: We transduced cultured primary cortical neurons [44] with a recombinant adeno-associated virus (rAAV) encoding for the core nucleosome component histone H2B fused to the fluorescent marker mCherry (H2B::mCherry) (Fig 1A), and imaged labeled chromatin with a confocal microscope at a ten minutes interval. When cultures were kept in standard media, H2B::mCherry remained uniformly distributed in the nucleus except for regions within the nucleus that contain no or only little chromatin, e.g., the nucleolus [47] (Fig 1B  top). When adding a pharmacological cocktail (medium containing 50μM (-)-Bicuculline methiodide, 100μM DL-Norepinephrin hydrochloride, 50μM Carbamoylcholine chloride, 100μM Dopamine hydrochloride, 40μM Ascorbic Acid and 100μM Serotonin hydrochloride) activating major neuromodulatory receptors and increasing synchronized network activity through a bicuculline-mediated block of inhibitory GABA A -receptors [48] (S1 Fig), we observed a fast increase in granularity of the fluorescent signal (Fig 1B bottom, S1 Movie). Interestingly, this effect appeared to be transient as the chromatin label became more homogeneous again within the two-hour observation period.
In order to quantify these changes in the nuclear distribution of H2B::mCherry, we calculated a Granularity Index (GI) based on the 2D Fourier transform of the fluorescent signal (for details see Methods) with larger GI values describing an increasingly uneven, granular distribution. GI values were normalized to time-point t 0, corresponding to the addition of the cocktail. When quantifying GI values in groups of neurons under cocktail conditions, we observed a systematic and significant increase in the GI, which peaked for most cells 40 minutes after application (Fig 1C and 1D, S2 Movie). This increase was completely absent in cells under control conditions. We furthermore tested if the observed changes in chromatin condensation are dependent on actomyosin activity and included a third condition in which we applied the pharmacological cocktail together with 20mM 2,3-butanedione monoxime (BDM). This cell-permeable drug impairs myosin ATPase activity and has been shown to interfere with nuclear actomyosin dynamics underlying the movement of nuclear bodies [49][50][51][52][53]. The simultaneous application of the pharmacological cocktail with BDM significantly inhibited the activity-induced restructuring of chromatin (Fig 1C and 1D)

Chromatin reorganization correlates with induction of c-Fos and increased γH2AX labeling
Next, we tested if activity-induced restructuring of chromatin is correlated with transcriptional activation. We repeated the previous experiments with the same conditions as described above (medium only, medium with the pharmacological cocktail, and medium with the pharmacological cocktail and BDM) and found a strong increase in the expression of the immediateearly gene product c-Fos after incubation with the pharmacological cocktail. The induction of c-Fos was blocked by the addition of BDM in a dose-dependent manner (Figs 2A-2D, S1E and S1F) (control, n = 90: 0.909± 0.051; cocktail, n = 372: 6.556± 0.251; cocktail+BDM, n = 106: 1. Recent studies have shown increased occurrence of DNA double-strand breaks following neuronal stimulation and behavioral experience [54,55], a phenomenon classically associated to pathological conditions. Indeed, we observed an increase of γH2AX, a marker for DNA double-strand breaks, after induction of neuronal activity with the pharmacological cocktail. Consistent with our previous observations on c-Fos induction, the addition of BDM also inhibited the increase of γH2AX (S2 Fig).

H2B::PaGFP facilitates tracking of chromatin reorganization
We wondered if nucleus-scale rearrangements of chromatin would be also triggered in the living brain by behaviorally relevant stimuli that are associated with changes in gene expression. The detection of chromatin reorganization using H2B::mCherry as marker requires imaging techniques with high spatial resolution restricting its applicability for in vivo approaches, where imaging conditions are typically less favorable compared to cell culture. To overcome this limitation, we followed a strategy to visualize chromatin dynamics by fusing H2B with a photoactivatable version of GFP (PaGFP) [56]. This approach leads to the integration of PaGFP into nucleosomes and allows for conditional photolabeling of fiducial makers to facilitate tracking of chromatin dynamics [57][58][59].
We produced a rAAV encoding for the H2B::PaGFP fusion protein under the human Synapsin1 promoter, transduced primary cortical cultures with both the H2B::PaGFP and the H2B::mCherry viruses and performed similar time-lapse experiments as previously described. After the first imaging time point, we photolabeled thin lines across the nuclei of individual cells. We reasoned that the nucleus-scale reorganization of chromatin structure would be robustly detected as transformations of the shape of the photolabeled line. Similar to our previous observations, we only detected minimal changes in GI values in the H2B::mCherry channel during one hour of baseline imaging in standard medium and the labeled line in the H2B:: PaGFP channel remained largely stable (Fig 2E, S3 Movie). After addition of the pharmacological cocktail, we again observed an increase in granularity in the H2B::mCherry signal, which was paralleled by a drastic transformation of the photolabeled lines of H2B::PaGFP. In order to quantify changes of the photolabel, we derived a second metric, termed circularity index (CI). This metric is calculated as the ratio of the second to the first eigenvalue of a principal component analysis of the intensity distribution of the photolabel. This index is 0 for an infinitely thin, straight line and 1 for a circle. CI values began to rise sharply immediately after the application of the pharmacological cocktail, similar to the GI values (Fig 2F; n = 466: before: 0.170±0.007, after: 0. 403±0. 012; ANOVA: d.f.: 1, F = 288.23, p<0.001, differences between groups: before vs. after: p<0.001; post-hoc multiple comparison test using Dunn and Sidák's approach). This demonstrates that the transformation of the photolabeled fiducial marker captures a related phenotype of the same underlying process we had previously observed with increases in GI in the pan-nucleosome labeling approach using H2B::mCherry. In summary, using H2B::PaGFP for photolabeling of a fiducial marker and subsequent tracking of the transformation of this label allows robust detection of chromatin dynamics in neurons that is less dependent on optical resolution.

Activity-induced reorganization of chromatin in acute brain slices from transgenic mice expressing H2B::PaGFP
To compare the effects of neuronal activity on chromatin organization between dissociated neurons and the transgenic mouse model, we took advantage of an in vitro slice preparation, which allows measuring chromatin dynamics in isolated brain tissue. First, using paired patchclamp whole-cell recordings in current-clamp mode, we observed that application of the pharmacological cocktail lead to reliable induction of synchronous bursting activity, similar to the previously described effects on dissociated cells (Fig 3A-3D; p<0.001, Fisher's exact test). Furthermore, we observed that this treatment led to a robust induction of c-Fos immunoreactivity (S6 Fig). Additional bulk RNAseq in cocktail-treated and control slices showed differential expression of 205 genes (S7 Fig). We compared these results with a previous study in which the effects of various behavioral treatments on gene expression patterns in the auditory cortex in vivo were analyzed [34] and found that the application of the pharmacological cocktail leads to robust induction of many known immediate-early genes in brain slices similar to behavioral experiences in vivo (Fig 3E).
Next, we investigated the effect of increased neuronal activity on changes in chromatin structure. We labeled thin lines across nuclei of layers II/III neurons in the auditory cortex in acute brain slices obtained from H2B::PaGFP x CaMKIIa-Cre mice and acquired images at a one-hour interval for a total of four hours. Consistent with our findings in primary cortical neurons, we observed that the magnitude of chromatin reorganization as measured by changes in the CI was markedly different between control slices and slices treated with the pharmacological cocktail. Co-application of the cocktail together with BDM inhibited this effect ( p<0.001, 120min: p<0.001; control vs. BDM: -60min: p = 1, 0min: p = 1, 60min: p = 1, 120min: p = 1; cocktail vs. BDM: -60min: p = 1, 0min: p = 1, 60min: p<0.002, 120min: p<0.001; post-hoc multiple comparison test using Dunn and Sidák's approach). In contrast to dissociated cell cultures, in brain slices the treatment effects were more heterogeneous across neurons (S9 Fig): Whereas we also observed few cells acquiring high CI values even under control conditions, also many cells showed a stable photolabel following cocktail treatment. Together, these results demonstrate that patterns of neuronal activity that induce changes in gene expression induce at the same time nucleus-scale reorganization of chromatin in brain tissue.

Dynamic chromatin organization in vivo
In order to investigate chromatin dynamics in a living animal, we injected rAAV-Cre into the auditory cortex of H2B::PaGFP mice, implanted a small cranial window above the auditory cortex [45,60], and photolabeled the chromatin of cortical neurons with thin lines across the nucleus. We found that this strategy lead, in the vast majority of cases, to the labeling of neurons expressing the neuronal marker NeuN. (S10 Fig). When revisiting the same cells 24 hours after photolabeling, we observed that the line-pattern was essentially stable in many cells, while in others its shape was preserved as well, but rotated. In yet other cells, the photolabel was distributed across the nucleus, reminiscent of the observations in dissociated neurons and acute brain slices (Fig 4A). A post-hoc analysis of in vivo photolabeled nuclei confirmed that an increase in the CI value was associated with a three-dimensional redistribution of the photolabel throughout the nucleus (S11 Fig). To obtain a better measure of the relevant time scales of chromatin reorganization in vivo, we monitored the transformation of the photolabel in an anesthetized mouse for three hours at a 15-minute interval (Fig 4B). Similar to the findings in brain slices, chromatin dynamics were highly heterogeneous in neighboring neurons in vivo. In many cells, the label remained largely stable, while in others, the labeled line transformed and redistributed within the nucleus. When plotting the CI values of individual cells over time, we noted that phases of relative stability were interrupted by rapid and substantial transformations of the photolabel within few minutes (e.g. cells #5, #6 in Fig 4B) [12,14].
Does the transformation of the shape of a photolabeled line in vivo reflect direct movements of chromatin or rather unbinding and re-binding of nucleosomes? In previous studies introducing photolabeling of tagged histones, only little diffusion of unbound nucleosomes has been reported and therefore, photolabeled fiducial markers have been used to track chromatin movements within the nucleus [57,58]. Consistently, we could track individual speckles of photolabeled chromatin in our time-lapse data suggesting that changes in the CI reflect movements of chromatin rather than histone diffusion or turnover, which occurs on much slower time scales [61]. Furthermore, the transformations of the patterns typically occurred in a directed manner, i.e. changes from image to image followed clear trajectories. This is in contrast to stochastic changes from image to image as they would be expected from a random walk-like diffusion process and rather suggests an actively driven process [15,62]. We observed similar dynamics in wild-type mice using Hoechst to directly stain DNA in vivo, as an alternative method to track nucleus-scale movements of chromatin independent of nucleosome components tagged with a fluorophore (S12 Fig). Passive diffusion of the photolabeled chromatin is expected to occur continuously in all cells. However, the observation that in some cells the photolabel can maintain a remarkable degree of stability over many hours indicates that, at the nucleus scale, only little chromatin motility results from passive processes alone. This suggests that under basal conditions chromatin appears to be highly viscous in vivo and effectively restrains diffusion [63].

Behavioral experiences induce nucleus-scale chromatin reorganization
To identify the physiological relevance of nucleus-scale chromatin dynamics, we combined in vivo imaging of chromatin with specific behavioral paradigms previously shown to induce plasticity in the mouse auditory cortex: Auditory cued fear conditioning and variants thereof affect the dynamics of synapses [45] and induce specific global patterns of gene expression (S13 Fig) [34]. We stereotactically injected rAAV-Cre into the auditory cortex of H2B::PaGFP mice and implanted a cranial window. We used intrinsic imaging of sound-evoked activity patterns to map areas within the auditory cortex activated by the sound (CS+) used for conditioning (Fig 5A-5C) [45]. We followed a longitudinal imaging strategy and exposed individual mice to a series of behavioral paradigms with increasing behavioral relevance (Fig 5D). On each experimental day, we photolabeled thin lines on a set of cells in a small patch of layer II/ III in the auditory cortex and acquired image stacks of those labeled cells directly after photolabeling and two hours later.
On the first day, animals were kept under the microscope under isoflurane anesthesia during the two hours as baseline control. On the second day, mice were woken up after the first imaging time point and returned to their home cage, a familiar environment. On the third day, experimental subjects underwent unpaired auditory cued fear conditioning between the two imaging time points, in which five sound presentations and foot shocks were applied temporally separated by more than a minute. Unpaired auditory cued fear conditioning does not lead to the specific formation of a tone-shock association but represents a stressful experience. It has previously been shown that unpaired auditory cued fear conditioning induces an extensive change of the transcriptional program [34], and increased rewiring of neuronal connections [45] in the auditory cortex. On the fourth day, mice underwent paired auditory cued fear conditioning between the two imaging time points, in which the conditioned sound was directly followed by a mild foot shock. Successful memory formation was assessed in a memory test session in a neutral environment on the next day. Here, we observed an increase in freezing behavior during the presentation of the conditioned sound, demonstrating the successful formation of an associative memory (Fig 5E; Freezing (Fig 5F-5I).
Increases in the CIs were observed during the two hours imaging interval in all groups but did not reach statistical significance for cells of mice which were anesthetized or were put into the familiar context. (Fig 5I; anesthetized, n = 62, pre: 0.105±0.005, post: 0.164±0.012, p = 0.122; context, n = 40, pre: 0.113±0.008, post: 0.163±0.019, p = 0.793; unpaired, n = 32, pre: 0.098±0.008, post: 0.234±0.036, p<0.001; paired, n = 42, pre: 0.113±0.007, post: 0.285±0.034, p<0.001; post-hoc multiple comparison test using Dunn and Sidák's approach). The fraction of neurons showing high CI values was significantly larger with increased behavioral relevance of the paradigm. In the two-hour time interval, significantly more cells showed stronger increases when the mice underwent unpaired and paired conditioning, the two paradigms that lead to large changes in transcriptional activity, in comparison to the anesthetized and context groups (Fig 5I; anesthetized  Sidák's approach). We did not detect a significant difference between the unpaired and the paired group, the two paradigms with the highest behavioral relevance. These results show that not only increased neuronal activity in primary neurons and brain slices, but also behaviorally relevant experiences that induce plasticity in synapses and gene expression profiles control the rate of nucleus-scale chromatin dynamics in vivo.

Inhibition of chromatin dynamics in auditory cortex impairs memory formation but not memory retrieval
Changes in the transcriptional program are necessary for the acquisition and consolidation of long-term memories [30]. Based on our findings that the inhibitor of nuclear actomyosin dynamics BDM efficiently blocks activity-induced chromatin reorganization and c-Fos expression, we hypothesized that interference with chromatin dynamics could impair memory acquisition. To test this, we bilaterally implanted cannulas on top of the auditory cortex of wild-type mice (S14 Fig) and subjected them to an auditory cued fear conditioning task. Immediately before the first conditioning session, we either infused BDM to block chromatin reorganization or with ACSF as control and probed the memory on the next day (Fig 6A). Consistent with our hypothesis, freezing levels were significantly lower in the BDM group during CS presentation but not baseline, indicating an impairment of fear memory (Fig 6B left). Freezing levels in the BDM group during CS presentation were nevertheless higher than during baseline, suggesting that memory formation was not completely blocked. To test if this impairment reflects a specific deficit in memory formation and not an impairment of memory maintenance or sound processing necessary for cued fear expression [64], we subjected the same cohort of mice to a second conditioning session without pharmacological treatment. Prior to the memory test session on the next day, we infused the auditory cortex with either BDM or ACSF and subjected the two groups to the conditioned stimulus to assess memory retrieval. While Group 2 recovered from the memory deficit, we did not detect significant differences in freezing levels within between the BDM and ACSF treated groups. These data indicate that BDM does not interfere with memory retrieval or a behavioral fear response ( Fig 6B right; freezing levels: memory test 1, baseline: group 1, n = 12, 10. p>0.999; memory test 1 CS, group 1 vs group 2: p = 0.019, memory test 2 CS, group 1 vs. group 2: p = 0.611; differences within groups: memory test 1, group 1, CS vs. memory test 2, group 1, CS: p = 0.564; memory test 1, group 2, CS vs. memory test 2, group 2, CS: p = 0.032; post-hoc multiple comparison test using Sidák's approach). Taken together, inhibition of chromatin dynamics in vivo significantly impairs memory formation but not memory retrieval in mice.

Discussion
Here, we show that changes in patterns of neuronal activity, as well as specific behavioral experiences, are correlated with a rapid, nucleus-scale reorganization of chromatin as well as changes in the transcriptional program of the cells. Furthermore, BDM, an inhibitor of nuclear forms of myosin, blocks chromatin dynamics, prevents the induction of the IEG c-Fos, and impairs the formation of long-term memories, whose consolidation is known to depend on transcriptional regulation. Together, this study demonstrates the dynamic nature of chromatin at the microscopic level in vivo and provides evidence consistent with an essential role of these dynamics to allow transcriptional changes underlying experience-dependent behavioral adaptation.
To extend the systems available for the study of chromatin dynamics from cultured cells to tissue preparations and living animals, we generated a novel transgenic mouse line that allows conditional expression of the photoactivatable fluorescent protein, PaGFP, fused to the histone H2B. Apart from the analysis of chromatin per se, this mouse line offers substantially enhanced labeling of individual cells for various photolabeling applications where reliable tracking and re-identification of cells is crucial [65].
How could nucleus-scale reorganization of chromatin contribute to alterations in the transcriptional program necessary for behavioral adaptation? There are three scales to consider. First, the molecular scale of individual genes: There is good evidence that the transcriptional activity in neurons, similar to other cell types, is controlled by epigenetic modifications and changes in affinities of promoters and enhancers to transcription factors [35,[66][67][68][69][70][71]. Second, the nucleus scale: The localization of individual genes into functionally active domains, such as euchromatin near nuclear speckles, or functionally repressed domains, such as heterochromatin close to the lamina, occurs at the micrometer scale [1-4, 62,72]. Third, the temporal scale: The activation of IEGs [31][32][33][34] and corresponding chromosome conformational changes [39] are reported within minutes after induction. Our observation that the chromatin architecture at the nucleus scale appears to be little affected by passive diffusion-like processes (Fig 4A, orange arrowhead) but can be rapidly adapted in response to behavioral experience, connects these three scales in the living animal.
Our findings are consistent with a model in which neuronal activity triggers processes dependent on nuclear myosin that put the nucleus in a permissive state allowing local, gene specific modifications of promoters and other regulatory elements to fall into a new organizational structure of the nucleus, resulting in an adaptation of the transcriptional program. It should be noted, however, that BDM has been shown not only to block nuclear forms of myosin, but can also interfere with other non-nuclear processes in cells [73] and therefore the interpretation of the effects should be done with care. We believe that impairment of nuclear motility being the primary effect of BDM in our experiments provides nevertheless the most parsimonious explanation for our observations. Interestingly, nuclear forms of myosin have been recently implicated as autism spectrum disorder risk genes [74]. In addition to myosin-dependent processes, recent studies implicate a phase separation mechanism to create such a permissive state [75][76][77][78][79][80][81][82][83][84]. Furthermore, our findings relate to the report of increased occurrence of double-strand breaks in neurons of mice exposed to a novel environment [55]: Global nucleus-scale movements likely lead to increased chromatin tension causing topological constraints for transcription of activated genes which would require resolution by double-strand breaks [54].
In recent years, HiC studies have allowed detailed insight into molecular features of chromatin reorganization upon transcriptional activation [41]. However, as these techniques typically rely on pooling of cells and fixating them at a given time point, our in vivo imaging approach provides a valuable complement enabling the longitudinal study of chromatin dynamics in individual cells and thereby to explore their heterogeneity within a tissue.
In summary, we report rapid nucleus-scale chromatin dynamics in a living animal which, are associated with fast changes in transcriptional states. Although, we focused in our study on memory formation, we believe that nucleus-scale reorganization of chromatin could be a general process associated to fast changes in the transcriptional program also in other contexts. Heatmap display showing significantly differentially expressed genes analyzed by RNAseq in six acute brain slices treated with the pharmacological cocktail and six control slices (column 'cocktail vs. control'). In a previous study, we analyzed the effect of various behavioral treatments (paired auditory cued fear conditioning, unpaired auditory cued fear conditioning, shock presentation only, sound presentation only, temporary housing in neutral context, housing in home cage) on gene expression using a microarray-based analysis. Here, we reported, considering the relative expression levels of individual genes, that many known immediateearly genes were induced in those mice undergoing a treatment involving a shock [34]; EMBL-EBI ArrayExpress, Accession number: E-MTAB-661). For comparison, corresponding relative expression levels from this in vivo study are displayed as additional columns. Expression levels of some genes were measured with several probes on the microarray and are therefore mentioned multiple times. Fig 3E of  Correlation of the CI calculated using the maximum intensity projection along the z-axis and the third eigenvalue estimated in the analysis considering all three dimensions in neurons fixed 1 hour (red) and 18 hours (blue) after photolabeling. In cultured cells, brain slices as well as in vivo, we observed rotational movements of the chromatin in many cells. This became particularly obvious in cells in which the photolabeled line was preserved but rotated. This phenotype suggests that nuclear chromatin can move entirely as a rigid body without changing its internal architecture. To test if increases in CI values describing a loss of the line structure are associated with a three-dimensional redistribution of the photolabeled chromatin throughout the nucleus, we performed in vivo photolabeling of neurons 18 and 1 hour before fixation, leading to two groups of cells that will, on average, have undergone more or less spontaneous reorganization of chromatin. The photolabeled cells were re-identified in the fixed tissue and imaged using confocal microscopy and subsequent image deconvolution to improve the z resolution of the images. We analyzed the three-dimensional spread of the fluorescence in the image stack using a principal component analysis. Rigid rotational movement would leave the pattern of the photolabel unaltered and, therefore would not affect the eigenvalues describing the spread along the three orthogonal principal components. We observed, however, that the variance along the third and smallest component significantly increased in the group of nuclei photolabeled 18 hours before fixation (1st, 18hrs, n = 6, 34.0±2.3, 1hr, n = 7, 31.2±1.4, n.s.; 2nd, 18hrs, 25.5±3.3, 1hr, 15.8 ±2.5, n.s.; 3rd, 18hrs, 10.7±0.8, 1hr, 3.9±0.6, p<0.001, unpaired t-test with multiple comparisons). Furthermore, there was a clear correlation between the CI, calculated on the two-dimensional maximum intensity projection along the z-axis, and the third eigenvalue, taking into account the three-dimensional distribution of the photolabel. These results show that, despite clear rotational components in the observed dynamics of chromatin in neuronal nuclei, increases in the CI are correlated with a three-dimensional distribution of the photolabel. A re-analysis of the data from the previous microarray study [34] taking into account the expression pattern of all genes on the microarray revealed that the biological replicates of mice undergoing either paired or unpaired conditioned could be identified based on the analysis of gene expression patterns. (A) Spearman correlation plot of the complete expression profiles, three independent biological replicates per treatment. Paired and unpaired conditioning and to some extend shock presentation only result in a good correlation of their overall expression profile within replicates but is particular distinct from each other as well as from sound presentation and housing in a neutral context.