Isolation of neuronal chromatin from brain tissue
© Jiang et al. 2008
Received: 29 January 2008
Accepted: 28 April 2008
Published: 28 April 2008
Skip to main content
© Jiang et al. 2008
Received: 29 January 2008
Accepted: 28 April 2008
Published: 28 April 2008
DNA-protein interactions in mature brain are increasingly recognized as key regulators for behavioral plasticity and neuronal dysfunction in chronic neuropsychiatric disease. However, chromatin assays typically lack single cell resolution, and therefore little is known about chromatin regulation of differentiated neuronal nuclei that reside in brain parenchyma intermingled with various types of non-neuronal cells.
Here, we describe a protocol to selectively tag neuronal nuclei from adult brain – either by (anti-NeuN) immunolabeling or transgene-derived histone H2B-GFP fusion protein – for subsequent fluorescence-activated sorting and chromatin immunoprecipitation (ChIP). To illustrate an example, we compared histone H3 lysine 4 and 9 methylation marks at select gene promoters in neuronal, non-neuronal and unsorted chromatin from mouse forebrain and human cerebral cortex, and provide evidence for neuron-specific histone methylation signatures.
With the modifications detailed in this protocol, the method can be used to collect nuclei from specific subtypes of neurons from any brain region for subsequent ChIP with native/un-fixed or crosslinked chromatin preparations. Starting with the harvest of brain tissue, ChIP-ready neuronal nuclei can be obtained within one day.
An increasing number of neurodevelopmental and neuropsychiatric disorders are thought to result from defective DNA:protein interactions specifically in neurons; furthermore, sustained changes in neuronal gene expression and behavior after exposure to certain drugs or stimuli are likely to involve chromatin remodeling, including DNA methylation and histone modification changes [1–5]. However, even the most sensitive chromatin immunoprecipitation assays and most other approaches used to study the regulation of DNA and histone modifications, transcription factor binding etc., lack single cell resolution and instead require the preparation of homogenates from at least 103 – 107 nuclei. Consequently, detailed chromatin analysis was until now not feasible for nuclei of terminally differentiated neurons that typically reside in brain parenchyma intermingled with various types of glia and other cells mostly in a 2:1 – 1:2 ratio, dependent on species and brain regions [6, 7].
Recently, immunostaining in conjunction with fluorescence-activated cell sorting (FACS) was used successfully to selectively collect neuronal nuclei from human (postmortem) brain tissue for the purposes of retrospective birth dating  or assessment of age-related changes in DNA cytosine methylation . However, these studies utilized the nuclear harvest for highly sensitive radiation and PCR assays, and it remained unclear whether the protocol could be modified for the purposes of chromatin immunoprecipitation and other techniques that require comparatively larger amount of input (for example, 105 107 nuclei). We provide a detailed protocol for selective sorting of neuronal nuclei from mouse and human brain in quantities sufficient for immunoprecipitation with different chromatin preparations (enzyme-based digestion and crosslinking/sonication), followed by microarray or PCR studies. In addition, we introduce a transgenic mouse line for neuron-specific expression of GFP (enhanced green fluorescent protein)-tagged histone H2B. Evidence is presented that even under baseline conditions, promoter-bound histone methylation in neuronal samples is significantly different when compared to unsorted, or non-neuronal nuclei from the same brain region. Therefore, the methods presented here will be important for the study of molecular mechanisms governing epigenetic control of neuronal gene expression and chromatin remodeling specifically in mature brain.
Each of the major steps in this protocol – (1) extraction of nuclei from the tissue, (2) ultracentrifugation, (3) immunolabeling and (4) fluorescence-activated sorting – could result in loss of nuclei, together totaling perhaps up to 50% of the neuronal nuclei in the starting material. However, despite of these limitations, the alternative approach, such as nuclei harvesting from primary neuronal culture, is likely not to improve results in a higher yield and would have the additional disadvantage of any ex vivo preparation.
To find out whether or not neuronal chromatin, under baseline conditions, shows a histone methylation signature that is different compared to unsorted nuclei, we conducted ChIP for two histone lysine methylation marks, H3K4me2 and H3K9me3. All studies were conducted on the forebrain of male mice 8–12 weeks of age. Histone methylation levels were at 3 out of 4 – arbitrarily chosen – gene promoters significantly different in neuronal chromatin (Fig. 4C). Furthermore, in 2 out of 4 promoters (B2m, Mecp2), changes in the "open" mark, H3K4me2 and the "repressive" mark, H3K9me3, were in the opposite direction (Fig. 4C). Notably, H3K9me3 levels at the Mecp2 promoter – i.e., the Rett syndrome gene which is robustly and ubiquitously expressed in mature CNS neurons [11–15] – were higher in neuronal vs. unsorted chromatin. This particular finding was unexpected and requires further investigations.
The protocol presented here should be particularly useful to investigators who wish to study chromatin regulation, including epigenetic mechanisms of neuronal gene expression, in the mature and the aging brain. The alternative approach, i.e. the study of primary neuronal cultures or neuron-like cell lines, certainly is suited to model developmental mechanisms and neurodevelopmental disease [24–29]. However, neurons and other CNS cells are sustainable in culture, at best, for a few weeks [30, 31] and therefore are not ideal from the viewpoint of chronic neuropsychiatric diseases, which often involve a protracted course, late onset and predilection for a specific neuronal subtype. Furthermore, the complexities of brain connectivity and circuitry, among a myriad of other factors, is difficult or impossible to model ex vivo. For example, in vivo treatment with antipsychotic drugs induces chromatin remodeling in cerebral cortex and striatum, an effect that cannot be mimicked in cortical or striatal cell cultures lacking input from monoamine pathways [32–34]. Finally, chromatin assays in whole brain tissue are potentially confounded by cellular heterogeneity, including potential shifts in cellular composition during the course of aging or due to disease. Given that chromatin immunoprecipitation with genome-wide coverage from as little as 10,000 cells is now feasible , it should be possible to use transgenic mice – including the CAMIIK-H2B-GFP lines presented here – to selectively study the epigenome for highly defined subpopulations of neurons in the mouse. Furthermore, the relative resilience of nuclei in frozen-then-thawed postmortem brain tissue makes it possible to apply the techniques presented here to species other than small laboratory animals, including normal and diseased human brain. It remains to be determined whether or not the approach presented here is applicable to tissues other than brain. When compared to sorting protocols that are based on trypsinization of tissue and sorting of intact (neuronal or non-neuronal) somata , our approach has several advantages. First, it is applicable to frozen tissues. Second, the yield of recovered nuclei appears to be much higher, albeit direct comparisons are lacking. Third, it is comparatively simple as compared to related protocols.
To summarize, the perspective to develop chromatin assays for specific types of neuronal populations of the mature brain, as described here, will be relevant for a wide range of physiological and neurological conditions related to transcriptional regulation and chromatin remodeling.
A cassette, comprised of 8 kB of the CamIIK (α subunit) promoter and 1.4 kB of histone H2b-eGFP cDNA (CaMK II-H2B-EGFP), was used to generate 3 lines of transgenic mice by pronuclear injection. All the experiments were performed in strict accordance with Institutional Guidelines regarding use of experimental animals.
Chemicals used include the following: 1 × PBS (Cellgro, 21-040-CV), Triton X-100 (Fluka, 93426), Igepal CA-630 (NP-40) (Sigma, 1–3021), Protease inhibitor cocktail tablets (Roche, 11697498001), Phenylmethanesulfonyl Fluoride (PMSF) (Sigma, P7626), Mouse anti-NeuN (Upstate, MAB377), Goat anti-mouse IgG, Alexa Fluor 488 (Invitrogen, A11029), Bovine serum Albumin (BSA) (Sigma, B-4287), Normal goat serum (Vector, S-1000), Micrococcal nuclease (Sigma, N-3765), Protein G Agarose, (Upstate, 16–266), Rabbit Anti-H3 (Abcam, ab1791), Rabbit Anti-H3K4me2 (Upstate, 07–030), Rabbit Anti-H3K9me3 (Upstate, 07–442), Rabbit Anti-H3K4me3 (Upstate, 07–473), Normal rabbit IgG (Upstate, 120370)
Nuclei Extraction Buffer (NEB): 0.32 M Sucrose, 5 mM CaCl2, 3 mM Mg(Ac)2, 0.1 mM EDTA, 10 mM Tris-HCl (pH8), 1 X Protease inhibitor cocktail, 0.1 mM PMSF, w/o 0.1% Triton X-100, w/o 0.1% NP-40. Sucrose Cushion: 1.8 M Sucrose, 3 mM Mg(Ac)2, 10 mM Tris-HCl (pH8). Blocking Solution: 0.5%. Bovine serum albumin (BSA), 10% normal goat serum, in 1 × PBS.
Total nuclei were extracted via sucrose gradient ultracentrifugation. In the work presented here (mouse), each sample was derived from a single forebrain (adult males, 8–12 weeks of age); (human) 1000 mg of cerebral cortex. All the reagents used were pre-chilled and the entire procedure was performed on ice. Fresh or frozen samples were homogenized by douncing 50 times in 5 mL NEB with 0.1% Triton X-100, or alternatively, 0.1% NP-40. Triton X-100 was preferable if nuclei require immunotagging (with NeuN, for example), while NP-40 as a milder detergent left more nuclei intact and sufficient when working with nuclei expressing GFP. After douncing, brain homogenates were transferred into 14 mL ultracentrifuge tubes (Beckman, 14 × 95 mm, 344061), and 9 mL of Sucrose Cushion was carefully loaded directly to the bottom of the ultracentrifuge tube. Ultracentrifugation was performed at 24400 rpm for 2.5 hrs at 4°C (Beckman, L8-70M, SW28 rotor). After centrifugation, a nuclei pellet – thin and typically with a light yellow taint – was formed on the bottom of the tube. The sticky white tissue debris was restrained in the middle interface of the two sucrose layers. Supernatant, including debris, was carefully removed. 1200 μl of chilled 1 × PBS was then added into the tube and incubated without disturbing the pellet, for 20 min on ice. This incubation is recommended because the nuclei pellet is easier to dissociate and consequently, there will be less damage to the nuclei. After incubation, the nuclei were dissociated by pipetting up and down. FACS required single nuclei in solution but excess pipetting tended to break nuclei, at least in the unfixed preparations. Pipetting required optimization by checking the nuclei "quality" (intact single nuclei, non-adhering, without debris) under the light microscope. The total number of nuclei was counted by using a hemacytometer. One mouse forebrain typically yielded 20–30 × 106 unfixed nuclei after ultracentrifugation.
One alternative option was to fix the nuclei by directly douncing the tissue in NEB containing 1% formalin. Typically, we fixed with formalin at room temperature for 5–10 min, and then added glycine to a final concentration of 125 mM. The fixed nuclei were pelleted by centrifugation, re-suspended in 5 mL of NEB, and subjected to ultracentrifugation as described above. Nuclei clumping during isolation was – in our experience – usually to blame if the final yield was lower than expected. For unfixed samples, factors that appear to promote clumping are (i) contamination from brain debris, or (ii) DNA leaking from broken nuclei, which appears to be alleviated via careful handling and performing the whole procedure on ice. When a fixation step was included (see above), the only way to significantly alleviate the clumping was to remove Ca2+ from the NEB, albeit this increased the proportion of broken and damaged nuclei. Thus, lower post-FACS yield is to be expected when working with fixed samples.
Antibodies used for immunotagging of neuronal nuclei.
Primary/secondary antibody conjugate
anti-Ms IgG Alexa 488
1 × PBS
Next, the fluorescent nuclei (H2B-GFP transgene or NeuN labeled) were filtered through a 40 μm Nitex mesh to remove any remaining clumps, and then run through a FACS machine (Vantage SE/Diva, BD Biosciences) with proper gates settings based on the size and density of nuclei, to ensure that only single NeuN+ nuclei were sorted, which typically included 40% of the whole population (Fig. 2B, panel b). After FACS, the NeuN+ nuclei were reanalyzed to confirm the purity. Sample purity typically is > 95%. For selected samples, the purity of NeuN+ nuclei was checked under the microscope (Fig. 2A, panels c, d and g, h). FACS sorting should be performed on the same day as extraction and NeuN labeling of nuclei. For a subset of experiments, including those involving human postmortem brain, immunonegative (NeuN-) nuclei were collected and processed in parallel to the fraction of neuronal (NeuN+) nuclei.
We usually obtained – from a single mouse forebrain – 2.5 × 106 NeuN+ nuclei post-FACS if fixative was added to the NEB and up to 8 × 106 nuclei if processed unfixed. We usually obtained from 1000 mg of human (prefrontal) cerebral cortex 5 × 106 NeuN+ nuclei if processed unfixed.
After FACS, nuclei were collected in a large volume of 1 × PBS, roughly 4 × 106 nuclei in 10 mL of 1 × PBS. This was not ideal for ChIP because processing such a large volume is associated with very high expense and potential "wasting" of antibodies and other reagents. Therefore, pelleting was important, but the challenge thereby was that post-FACS nuclei became fragile and most of them were destroyed by centrifugation in 1 × PBS. Therefore, we tested different buffers and noticed that the presence of Ca2+ and Mg2+ is crucial for integrity of nuclei during pelleting. Hence, the concentration of sucrose, Ca2+, and Mg2+ in post-FACS nuclei solution was adjusted by adding the following to each 10 mL of post-FACS nuclei: 2 mL of 1.8 M Sucrose, 50 μl of 1 M CaCl2, and 30 μl of 1 M Mg(Ac)2. Samples were mixed gently, incubated on ice for 15 min, and centrifuged at 3000 rpm for 15 min at 4°C.
The FACS sorted nuclei were prepared either by micrococcal nuclease digestion for native chromatin immunoprecipitation (NChIP) as described , or in case of fixed nuclei, sheared by sonication (Branson Sonifier 250) for XChIP. (Typically, samples were sonicated at power level 6 (Branson, Danbury CT) in ice water, by applying 10 runs of 30 sec impulse with 1 min resting interval. Again, sonication conditions need to be optimized before each experiment.) In the work presented here, only NChIP was used. In brief, micrococcal nuclease (MNase) digestion was performed with a working concentration of 4U/mL at 37°C for 5 min. The resulting mono-nucleosomal preparation (around 146 bp) [see Additional file 1, panel A] was precleaned by incubating with protein G agarose (which should also remove the antibodies used for NeuN labeling of nuclei) and then subjected to immunoprecipitation using antihistone antibodies (see Reagents). Control experiments included samples processed with normal rabbit IgG, in parallel to samples with specific antibodies. Typically, we used 0.5 – 8 × 106 nuclei as input.
Quantification of DNA extracted from immunoprecipitates was done by real time PCR using custom-designed primers targeting the promoter regions of the following genes: (mouse) methyl CpG binding protein 2 (Mecp2), glutamate decarboxylase 1 (Gad 1), β2-microglobulin (B2m), and Glial fibrillary acidic protein (Gfap); (human) glutamate receptor, ionotropic, N-methyl D-aspartate 2B (GRIN2B), brain derived neurotrophic factor (BDNF), β2-microglobulin (B2M), and the locus control region of the globin genes (HBB). Mouse Mecp2 primer sequence (119 bp): forward GCCTCTTTTCCCTGCCTAAA, reverse CCCTTGCTCTTTGTCGAGAT; Gad 1 (98 bp): forward TGTCTCACCAAAGTCCCTGTC, reverse CACGTCTGGTTCGGTGTCT; B2m (112 bp): forward GGGAAAGTCCCTTTGTAACCT, reverse GCGCGCGCTCTTATATAGTT; Gfap (100 bp): forward TACCAGAAAGGGGGTTCCTT, reverse AACTCCTCTCACCCCACTGA. Human GRIN2B primer sequence (64): forward TCCTCTTCCATTCAGGTTGG, reverse GGCTATACCATTCCTGGGACA; BDNF (102): forward AGCCCAACAACTTTCCCTTT, reverse GAGAGCTCGGCTTACACAGG; B2M (99): forward GGGCACCATTAGCAAGTCAC, reverse GGCGCTCATTCTAGGACTTC; HBB (87): forward CCCCAGGTAGTTCCCTTT reverse TTCAAGGCCCTGTAGTTGCT. Quantification was done as described . For each of the antibodies used in this study (anti-H3K4me2, -H3K9me3, and -H3K4me3), specific signal was limited to input and chip fractions, and differed from controls by at least 3–4 cycle thresholds [see Additional file 1, panel B, C].
To study histone occupancy at gene promoters on a larger scale, a mouse promoter array, M8K (Aviva Systems Biology, San Diego, CA), containing 8000 40-mer oligonucleotide probes targeting gene promoters, was used. The DNA samples were amplified and labeled by Aviva's ChIP-GLAS system. Briefly, both input and immunoprecipated (ChIP) DNA were biotinylated and annealed to the M8K-Oligo-Mix. The DNA samples were then amplified by ligation mediated PCR with T3 and T7 based primers (25 cycles), labeled with Atto 550 or 647, and hybridized onto the chip. Amplification, labeling, hybridization, and washing was carried out according to the manufacturer's instruction (Aviva, AK-0524). Slides were scanned with DNA Microarray ScannerBA (Agilent Technologies, Santa Clara, CA, USA) and intensities extracted with Feature Extraction Software Version 9.1 (Agilent Technologies, Santa Clara, CA, USA). Raw data were read into the R statistical computation environment for preprocessing and data analysis (R: Development core team (2004). R: A language end environment for statistical computing. Vienna, Austria.URL found in Availability and requirements section). Local background intensities were subtracted from raw signals and negative values were replaced with small positive ones. Signals were then normalized using a variance stabilization method described in [38, 39] and implemented in the vsn Bioconductor package (Gentleman, RC. et. al., Bioconductor: open software development for computational biology and bioinformatics. Genome Biology 5, R80 (2004)).
We used the Genomeplex whole genomic amplification kit (WGA2, Sigma) to amplify DNA immunoprecipitates, which resulted in 2 to 5 ug of (amplified) DNA from starting material (10 ng of input DNA; ChIP DNA from 0.5 × 106 nuclei).
R:A language end environment for statistical computing: http://www.r-project.org
The authors thank Dr. Ralph Kuehn for the CAMIIK promoter and Dr. Konrad Hochedlinger for the H2B-GFP DNA; Dr. Ron Zielke and staff (Brain and Tissue Bank for Developmental Disorders) and Dr. Edward G. Jones and staff (University of California at Davis) for providing postmortem brain specimens. This study was supported by grants from National Institutes of Health (NIH) to S.A.
Core resources supported by the Diabetes Endocrinology Research Center Grant DK032520 were also used.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.