Single-molecule analysis of combinatorial epigenomic states in normal and tumor cells (original) (raw)
Proc Natl Acad Sci U S A. 2013 May 7; 110(19): 7772–7777.
Patrick J. Murphy,a Benjamin R. Cipriany,b Christopher B. Wallin,c Chan Yang Ju,d Kylan Szeto,c James A. Hagarman,d Jaime J. Benitez,c Harold G. Craighead,b,c,1 and Paul D. Solowaya,d,1
Patrick J. Murphy
aGraduate Field of Genetics, Genomics, and Development,
Benjamin R. Cipriany
bDepartment of Electrical and Computer Engineering,
Christopher B. Wallin
cSchool of Applied and Engineering Physics, and
Chan Yang Ju
dDivision of Nutritional Sciences, Cornell University, Ithaca, NY, 14853
Kylan Szeto
cSchool of Applied and Engineering Physics, and
James A. Hagarman
dDivision of Nutritional Sciences, Cornell University, Ithaca, NY, 14853
Jaime J. Benitez
cSchool of Applied and Engineering Physics, and
Harold G. Craighead
bDepartment of Electrical and Computer Engineering,
cSchool of Applied and Engineering Physics, and
Paul D. Soloway
aGraduate Field of Genetics, Genomics, and Development,
dDivision of Nutritional Sciences, Cornell University, Ithaca, NY, 14853
aGraduate Field of Genetics, Genomics, and Development,
bDepartment of Electrical and Computer Engineering,
cSchool of Applied and Engineering Physics, and
dDivision of Nutritional Sciences, Cornell University, Ithaca, NY, 14853
Edited by Steven E. Jacobsen, University of California, Los Angeles, CA, and approved March 29, 2013 (received for review October 23, 2012)
Author contributions: P.J.M., H.G.C., and P.D.S. designed research; P.J.M., C.Y.J., and J.A.H. performed research; P.J.M., B.R.C., C.B.W., C.Y.J., K.S., and J.J.B. contributed new reagents/analytic tools; P.J.M., B.R.C., and C.B.W. analyzed data; and P.J.M. and P.D.S. wrote the paper.
Freely available online through the PNAS open access option.
Supporting Information
GUID: 08C53337-55DD-4DB9-89DE-227C7AD90333
GUID: 6CEBE674-49B9-442C-B132-F3153F4BA1E5
Abstract
Proper placement of epigenetic marks on DNA and histones is fundamental to normal development, and perturbations contribute to a variety of disease states. Combinations of marks act together to control gene expression; therefore, detecting their colocalization is important, but because of technical challenges, such measurements are rarely reported. Instead, measurements of epigenetic marks are typically performed one at a time in a population of cells, and their colocalization is inferred by association. Here, we describe a single-molecule analytical approach that can perform direct detection of multiple epigenetic marks simultaneously and use it to identify mechanisms coordinating placement of three gene silencing marks, trimethylated histone H3 lysine 9, lysine 27 (H3K9me3, H3K27me3), and cytosine methylation (mC), in the normal and cancer genome. We show that H3K9me3 and mC are present together on individual chromatin fragments in mouse embryonic stem cells and that half of the H3K9me3 marks require mC for their placement. In contrast, mC and H3K27me3 coincidence is rare, and in fact, mC antagonizes H3K27me3 in both embryonic stem cells and primary mouse fibroblasts, indicating this antagonism is shared among primary cells. However, upon immortalization or tumorigenic transformation of mouse fibroblasts, mC is required for complete H3K27me3 placement. Importantly, in human promyelocytic cells, H3K27me3 is also dependent on mC. Because aberrant placement of gene silencing marks at tumor suppressor genes contributes to tumor progression, the improper dependency of H3K27me3 by mC in immortalized cells is likely to be fundamental to cancer. Our platform can enable other studies involving coordination of epigenetic marks and leverage efforts to discover disease biomarkers and epigenome-modifying drugs.
Epigenetic marks are responsible for controlling the temporal and spatial pattern of gene expression throughout the genome. In a number of instances, these marks have been shown to act combinatorially (1, 2). Co-occurrence of epigenetic marks has been implicated in a variety of important processes including cell differentiation (2), gametogenesis (3), and DNA replication (4). Additionally, examples exist where epigenetic marks can directly promote or inhibit the presence of one another (5, 6). Consequently, reliably detecting epigenetic mark colocalization is an essential step for advancing a host of biological studies. Histone modifications and cytosine methylation (mC) are traditionally assayed by chromatin immunoprecipitation (ChIP) and bisulfite sequencing (BS), respectively. Typically, one assay is performed at a time and colocalization of marks is inferred by association. However, with this approach, it remains unknown if the inferred combinatorial states actually exist (7). Serial ChIP can detect combinations of histone modifications, but its low efficiency requires an abundant source of chromatin, and it is impractical for assaying more than two modifications; BS of ChIP DNA can report coincidence of histone modifications and mC (8); and mass spectrometry can quantify combinations of histone marks, if they reside nearby on the same histone (9). Each method is labor intensive and difficult to use when quantitative data are needed. Here, we describe a single-molecule analytical approach that can rapidly and quantitatively assay combinations of epigenomic marks.
We previously described SCAN (Single Chromatin molecule Analysis in Nanochannels), a nanofluidic approach that enabled high-throughput fluorescent measurements of single DNA and chromatin molecules (10). When used to analyze native chromatin from GFP tagged histone H2B (H2B-GFP) expressing HeLa cells, we showed that molecules bound with a fluorescent DNA intercalator also carried GFP, demonstrating that the chromatin remained intact during the analysis. When we analyzed mixtures of methylated and unmethylated DNAs that were combined with a fluorescently tagged methyl binding domain protein-1 (MBD1)1 protein, we observed specific detection of methylated DNA. These results suggested SCAN could be used for rapid, quantitative epigenomic measurements, and that it could be used to detect the presence of combinations of epigenetic features on individual chromatin molecules. Here, we reduce this objective to practice and apply SCAN to demonstrate the interdependence of histone modifications on DNA methylation status. We show that mC is needed for proper H3K9me3 placement and that it antagonizes H3K27me3 in primary cells; however, the effects of mC are reversed in immortalized and transformed cells where mC is required for full H3K27me3 placement. This improper cooperation between mC and H3K27me3 deposition can potentially lead to aberrant placement of gene silencing marks on tumor suppressors and disease progression.
Results
We first established conditions for binding fluorescent MBD1 and antibodies recognizing histone features to chromatin with high specificity (Fig. 1_A_). As an initial test, we labeled an antibody recognizing the unmodified N-terminal tail of histone-H3 (α-H3) with AlexaFluor647, bound it to native chromatin isolated from HeLa cells (Fig. S1) expressing an H2B-GFP fusion protein, and then analyzed the mixture by SCAN. We ensured the total concentration of fluorophores in our analyte remained at or below 1 nM so that the probability of detecting only a single molecule or complex was greater than 99.5% (Fig. 1_B_). To promote binding of the antibody to chromatin, typical reactions were performed with 3–10 nM chromatin and a minimum of 15-fold molar excess of antibody followed by a cross-linking step. Under these conditions, even antibodies with a relatively unfavorable _K_d of 10 nM will bind to 95% of their chromatin target molecules, even when the epitope is present on only 0.01% of the molecules. Unbound labeled antibodies will remain in the analyte, and bound complexes between chromatin and unlabeled antibody molecules escape detection. If SCAN were capable of detecting antibody-chromatin complexes, we expected to observe individual molecular complexes that emitted both green (GFP) and red (AlexaFluor647) photons. This is precisely what we observed demonstrating that labeled antibodies could be used to detect chromatin features on single molecules using SCAN (Fig. 2 A and B). To confirm the presence of complexes and measure their abundance, coincidence analysis was performed (Fig. 2_B_ and Fig. S2).
SCAN workflow. (A) Native chromatin bearing epigenetic marks is mixed with fluorophore (e.g., AlexaFluor488) labeled antibody specific to a given mark. After binding, the chromatin is labeled with an intercalator (e.g., TOTO-3). Finally, the chromatin is driven by voltage through a nanoscale channel fabricated in fused silica and fluorescent measurements of individual molecules are taken in a 150-aL inspection volume. A more detailed schematic of laser setup can be found in our previous publication (10). (B) Probability of erroneously interrogating more than a single molecule increases with analyte concentration according to a Poisson distribution less than 0.5% at the concentrations used here (≤1 nM). _P_c(m), probability of m molecules residing in the 150-aL interrogation volume at any one time, given concentration c of fluorescent molecules in analyte, expressed as molecule count x in 150 aL, where NA is Avogadro’s number. The curve shows the probability that more than one molecule is in the inspection volume.
SCAN detects chromatin features with high specificity. AlexaFluor dye-tagged antibodies or MBD1 protein were bound to chromatin before analysis by SCAN. (A and B) Chromatin was from H2B-GFP expressing HeLa cells, and antibody (α-H3) was H3-specific. (A) Time-resolved photon counts (0.5 s) reporting GFP (Upper) and α-H3 (Lower) fluorescent emissions, depicted as peaks. Shading identifies antibody-bound chromatin complexes emitting both fluorophores nearly simultaneously. (B) Time-offset histogram for 7,065 GFP chromatin molecules analyzed in A. GFP fluorescent events are placed at time 0 and the time offset identifies how soon before or after the GFP emission AlexaFluor dye was detected. The peak centered at time 0 with width less than the transit time for molecules passing through the inspection volume (∼1 ms) identifies antibody–chromatin complexes. See Fig. S2 for description of coincidence calculations. (C and D) Chromatin was from WT-ESC bound to α-H3 (C), or preimmune mouse serum (D). (E and F) Antibody probe was specific for H3K27me3 and chromatin was from WT-ESC (E) or _Eed_−/− ESC, which are deficient for H3K27me3 (F). (G and H) MBD1 probe was specific for mC in duplex DNA, and chromatin was from WT-ESC (G) or DNMT TKO ESC, which are deficient for mC (H). Chromatin in C_–_F was detected by labeling with the intercalator YOYO-1, whose emission defined the 0 time in the offset plots.
Because most chromatin samples of interest will not come from cells with GFP-tagged histones, we extended this α-H3 binding test using native chromatin from wild-type mouse embryonic stem cells (WT-ESC) labeled with the fluorescent intercalator YOYO-1, and then used SCAN to detect complexes carrying YOYO-1 and α-H3AlexaFluor647. We could easily detect these complexes, demonstrating the utility of SCAN for a variety of chromatin sources (Fig. 2_C_). As a negative control, we repeated this test using AlexaFluor647-labeled preimmune mouse serum in place of α-H3 and observed no antibody–chromatin complexes, demonstrating specificity of antibody binding (Fig. 2_D_).
α-H3 should bind most if not all chromatin molecules in a mixture of complex chromatin. To determine if we could use SCAN to detect less-common epigenomic features of particular interest, we bound AlexaFluor647-labeled α-H3K27me3 specific antibodies to native chromatin from WT-ESC, labeled it with YOYO-1, and then performed SCAN. We could readily detect YOYO-1-labeled chromatin bound to α-H3K27me3 antibodies (Fig. 2_E_). As a negative control for α-H3K27me3 binding specificity, we repeated this test using chromatin from ESC homozygous for a mutation in Eed, which encodes a polycomb repressive complex 2 (PRC2) component needed for efficient H3K27me3 (11). In _Eed_-deficient cells, the number of YOYO-1 labeled molecules bound to antibody was greatly diminished (Fig. 2_F_). Therefore, detection of H3K27me3 using SCAN and α-H3K27me3 antibodies is specific. We performed a similar test using fluorescently tagged MBD1 protein to detect DNAs harboring mC in WT-ESC, and included as a negative control chromatin from DNA methyltransferase triple knockout (DNMT TKO) ESC that are deficient for the three mammalian DNA methyltransferases, which have ∼2% of the mC content of WT-ESC (12). The MBD1 bound chromatin from WT-ESC, but binding to DNMT TKO ESC chromatin was greatly diminished (Fig. 2 G and H). Like H3K27me3, detection of mC by SCAN and MBD1 is specific.
We next used SCAN to detect combinations of epigenetic features on chromatin. For these experiments, we bound to chromatin two different probes recognizing distinct epigenetic features, both labeled with spectrally distinct fluorophores. In our first tests we bound antibodies recognizing H3 and H2B to WT-ESC, which are expected to bind virtually all chromatin fragments. As a negative control for antibody aggregation, we performed an identical binding reaction without chromatin. Only in the binding reactions that included chromatin could we detect complexes with both antibodies, demonstrating that we can use SCAN for simultaneous detection of multiple chromatin features (Fig. 3 A and B). We extended this test incrementally by substituting α-H3K27me3 for the α-H3 antibody and were able to detect chromatin molecules binding both antibodies (Fig. 3_C_). We then performed binding reactions using α-H3K9me3 and MBD1 and detected chromatin molecules bearing two combined epigenetic features (Fig. 3_D_). As a negative control, we tested chromatin for coincidence of H3K4me3 and mC, which are thought to be mutually exclusive. We could detect no statistically significant coincident events (Fig. 3_E_). As combinations of epigenetic features are fundamental to genomic regulation (13), their multiplexed detection with tools like SCAN can provide new insights into the epigenome.
Detection of two epigenetic marks simultaneously. (A and B) Antibodies against H3 and H2B were labeled with spectrally distinct fluorophores and bound to chromatin (A), or included in a binding reaction without chromatin as a negative control (B). (C_–_E) Binding reactions were performed as in A using antibodies against H2B and H3K27me3 (C), using MBD1 protein and an antibody against H3K9me3 (D), or using MBD1 protein and an antibody against H3K4me3 (E). Chromatin was from WT-ESC; the central peak identifies chromatin molecules bound to two fluorescent probes as described in Fig. 2.
Having shown that mC and H3K9me3 were commonly detected on the same individual chromatin molecules, we wondered if H3K9me3 was dependent on mC for its placement. There is precedent for cross-regulation of the two marks: H3K9me3 is needed for normal mC deposition in Neurospora crassa and mice (5, 14); in a reciprocal way, mC positively affects H3K9me3 placement at normally silenced loci in Arabidopsis thaliana (15). However, in some human cell cultures, mC antagonizes H3K9me3 placement (16). In none of these studies was the magnitude of these effects quantified. We used quantitative SCAN to measure the relative abundance of H3K9me3 on chromatin in WT-ESC vs. DNMT TKO cells, using H2B antibody to normalize the signals from the two cell types. Our results showed that 60% of the H3K9me3 levels in ESC depend on mC placement (Fig. 4_A_). H3K27me3, like H3K9me3 and mC, is commonly associated with gene silencing; therefore, we also measured changes in H3K27me3 levels when mC was diminished. In contrast with H3K9me3, H3K27me3 levels rose ∼220% when mC was lost (Fig. 4_A_). This is consistent with previous reports that mC antagonizes H3K27me3 placement in primary mouse fibroblasts at Rasgrf1 (17) and is consistent with global suppression of H3K27me3 in mouse ESC by mC (8, 18).
Loss of DNA methylation affects histone modification states. (A) ESC, (B) MF primary mouse fibroblasts, (C) MF3T3 immortal mouse fibroblasts, (D) MF60.1 transformed tumorigenic mouse fibroblasts, or (E) HL-60 human myeloleukemia cells had normal mC levels, or mC impaired by mutation (A), or 5-Aza-2′-deoxycytidine (5AzadC) treatment (B_–_E). Chromatin from cells was analyzed by SCAN after labeling with an intercalator (TOTO-3 or YOYO-1) and fluorescent MBD1 protein or α-H2B, α-H3K9me3 (K9) or α-H3K27me3 (K27). Coincidence values between α-H2B and intercalator were used to normalize relative abundance of H3K9me3, H3K27me3, and mC in cells with normal and impaired mC. In box plots, whiskers represent 5th and 95th percentile.
Our observation that mC antagonizes H3K27me3 in primary cells is in contrast with results from transformed cells. For example, mC requires H3K27me3 at several loci in HeLa cells, a cervical carcinoma line (19); additionally, mC is not found at many PRC2 binding sites in primary cells, but mC is aberrantly acquired at PRC2 binding sites in cancerous cells (20–26). Generally, coordination between mC and H3K27me3 in transformed vs. normal cells varies by genomic location (27). It is possible to reconcile these results if the antagonism between mC and H3K27me3 seen in primary cells breaks down in transformed cells. To test this possibility, we analyzed the effects of DNA demethylation on levels of H3K27me3 in additional mouse cell types including primary fibroblasts (MF), immortal fibroblasts generated by 3T3 selection (MF3T3), and tumorigenic fibroblasts (MF60.1), fully transformed with Ha-ras and _v-_myc (28). We used the DNMT inhibitor 5-Aza-2′-deoxycytidine (5AzadC) to partially demethylate the cells, and then used SCAN to compare H3K27me3 levels in 5AzadC-treated cells relative to untreated controls (Fig. S3). Primary MF behaved similarly to primary ESC––depletion of mC led to a 570% increase in H3K27me3 abundance (Fig. 4_B_). However, in immortal MF3T3 and oncogene-transformed MF60.1, depletion of mC led, respectively, to a 98% and 81% depletion of H3K27me3 (Fig. 4 C and D). These changes were confirmed by quantitative Western blotting (Fig S4). This demonstrates that as mouse cells transition from normal to cancer-associated phenotypes, the widespread antagonism between mC and H3K27me3 that is normally present is lost. To determine if human cancerous cells similarly lose this antagonism, we repeated these measurements using the human acute promyelocytic leukemia cell line HL-60 treated with 5AzadC (Fig. S3). Similar to MF3T3 and MF60.1, depletion of mC in HL-60 cell caused a 24% decrease in chromatin associated H3K27me3, indicating that antagonism between mC and H3K27me3 that exists in primary cells breaks down in human cancer (Fig. 4_E_).
There are two general mechanisms by which mC might antagonize H3K27me3. In a direct mechanism, mC might exclude the recruitment of PRC2, preventing deposition of H3K27me3; when mC is lost, PRC2 is no longer excluded and H3K27me3 can be deposited. In support of this, mC has been shown to impair binding of PRC2 to DNA in biochemical assays (29). It is likely that this direct mechanism operates in ESC because more than 99% of sites where H3K27me3 was acquired in mC-deficient TKO ESC were sites where mC was normally found in WT-ESC (8, 18). Alternatively, mC might antagonize H3K27me3 deposition indirectly by affecting activities of PRC2, PRC2 regulatory factors, effectors of H3K27 demethylation, or higher complexity mechanisms controlling global chromatin organization. In support of this is the fact that cells treated with a DNMT inhibitor acquire H3K27me3 even at sites where there was no preexisting mC (16). It is clear that no single mechanism exists by which mC affects H3K27me3 as the influence of mC can vary by direction, locus, and cell type (8, 16, 19).
To determine if the effect of decreased mC on H3K27me3 abundance in immortalized and transformed MF was due to reversal of direct antagonism seen in primary cells, we performed relative quantification of chromatin possessing both H3K27me3 and mC simultaneously (Fig. 5). For these experiments we compared the MF samples (where 5AzadC caused increased H3K27me3 abundance) to the MF3T3 samples (where H3K27me3 levels decreased following 5AzadC treatment). Rather than comparing the number of bound complexes relative to total YOYO-1-labeled molecules, we compared the number of bound complexes relative to either total α-H3K27me3 molecules (Fig. 5 A and D), or to total MBD1 molecules (Fig. 5 B and E), both of which were present at a 100-fold molar excess over chromatin and which served as internal spike in controls. Analyzing the data in this manner allowed us to compare across cell types without relying on the DNA intercalator, which would require reengineering the SCAN platform to accommodate analyses of a third fluorophore. Applying this approach to data used to generate Fig. 4 B and C demonstrated that MF3T3 cells had significantly more H3K27me3 and significantly less mC relative to MF (Fig. 5 A_–_C). The increase in H3K27me3 levels following immortalization was confirmed using quantitative Western blotting (Fig. S5). If the observed changes in H3K27me3 abundance were due to reversal of direct antagonism, then coincidence of α-H3K27me3 and MBD1 would be higher in MF3T3 relative to MF chromatin samples. Our data show there were very few chromatin molecules bound by both MBD1 and α-H3K27me3, and that there were no significant differences between the two cell types (Fig. 5 D_–_F). This is consistent with the notion that mutual exclusion of the H3K27me3 and mC was maintained following immortalization, and that the observed increases in H3K27me3 level in immortal and transformed MF after 5AzadC treatment were not due to reversal of direct antagonism. In this case, the effects of 5AzadC on H3K27me3 are more likely due to general reorganization of histone modifications that occurs as a consequence of hypomethylation. However, it is possible that analysis of many more molecules would identify subtle differences not seen in these analyses. Regardless of mechanisms by which mC affects H3K27me3 levels differently in primary vs. immortal and transformed cells, the changes after immortalization are likely to be fundamental to the epigenetic chaos associated with cancer and contribute to tumor suppressor silencing.
Immortalization increases H3K27me3 abundance and reduces mC abundance without changing mC and H3K27me3 coincidence. Chromatin was bound to either a single epigenetic probe and YOYO-1 (A_–_C and F), or to two different probes simultaneously (D and E). (A) Relative abundance of H3K27me3 in MF and MF3T3 chromatin. Chromatin from the two cells was labeled with the intercalator YOYO-1 and AlexaFluor647 labeled α-H3K27me3. Epigenetic mark abundance was measured as molecules labeled with both fluorophores relative to total α-H3K27me3, which was present at a 100-fold molar excess over input chromatin and served as a spike in control. H3K27me3 abundance in MF3T3 was normalized to signal from MF chromatin. (B) Relative abundance of mC in MF and MF3T3 chromatin. Analysis done as in A with AlexaFluor488 labeled MBD1, present at a 100-fold molar excess relative to chromatin, and TOTO-3-labeled chromatin. (C) Relative abundance of MF and MF3T3 chromatin used in A and B. Analysis done in parallel and as in A and B with AlexaFluor647 labeled α-H2B and YOYO-1-labeled chromatin. Lack of statistical difference indicates similarity in chromatin loading for A and B. (D and E) Relative abundance of chromatin marked by both α-H3K27me3 and MBD1. Chromatin was bound to AlexaFluor488 labeled MBD1 and AlexaFluor647 labeled α-H3K27me3, with epigenetic probes present at a 50-fold molar excess relative to chromatin. Relative abundance of chromatin bearing H3K27me3 and mC marks is shown relative to total α−H3K27me3 (D), or to total MBD1 (E). (F) Relative abundance of MF and MF3T3 chromatin used in D and E; analysis done in parallel and as in D and E with AlexaFluor647 labeled α-H2B and YOYO-1-labeled chromatin. Lack of statistical difference indicates similarity in chromatin loading for D and E.
Discussion
Here, we describe a single-molecule analytical approach to characterize epigenetic states. This allowed us to detect combinations of epigenetic features, measure epigenetic mark abundance rapidly, identify coordinated regulation among mC, H3K9me3, and H3K27me3, and monitor effects of epigenome-modifying drugs on cancer cells.
Our platform directly detected the colocalization of H3K9me3 and mC on individual molecules in ESC. The colocalization is relevant mechanistically: H3K9me3 requires mC for its normal placement in the genome. The requirement is not absolute, as nearly half of H3K9me3 is placed in the absence of mC. It is possible that without mC, the nucleosome remodeling deacetylase (NuRD) complex is not recruited to mC by the MBD protein in the complex, and H3K9Ac persists, preventing H3K9me3 placement.
In contrast with H3K9me3, mC antagonizes H3K27me3 in primary cells. One possible mechanism underlying this is that mC impairs chromatin recruitment of PRC2, which is needed for H3K27me3 deposition (29). In immortalized mouse and human cells effects on H3K27me3 from 5AzadC treatment are reversed, whereby H3K27me3 becomes partially dependent on mC. How this shift occurs is unknown. It is possible that the composition of PRC2, or posttranslational modifications to its components, changes upon immortalization, allowing EZH2, the H3K27 histone methyltransferase, to be stimulated rather than inhibited by mC. It is also possible that aberrant H3K27me3 regulation in immortal and tumorigenic cells is a consequence of altered energy metabolism in those cells. Many cancers have increased glycolysis and reduced tricarboxylic acid (TCA) cycle activity (30–32), which can lead to accumulation of succinate that in turn inhibits the H3K27 demethylase JMJD3 (33). Additionally, impaired TCA cycle use might result in reduced conversion of isocitrate to α-ketoglutarate, potentially limiting activities of H3K27 demethylases requiring this cofactor (34). If DNA hypomethylation caused by 5AzadC treatment partly reverses these cancer-associated metabolic defects, H3K27 demethylase activity could increase, leading to simultaneous reductions in both mC and H3K27me3 levels, which we observed. Regardless of the mechanisms by which normal antagonism of mC and H3K27me3 goes awry in cancer, synergy in their placement can potently and inappropriately silence tumor suppressors, leading to disease progression.
Our studies of coordination among mC, H3K9me3, and H3K27me3 represent just one application of our single-molecule analytical platform. Other applications for the SCAN platform include enabling discovery of additional epigenome-based cancer biomarkers, facilitating quantitative assays for the effects of epigenome modifying agents in a drug discovery pipeline, and serving in clinical assays to measure patient responses to such drugs. There are several additional embodiments possible for our system that can expand its capabilities, including performing parallel analyses in multiple channels to increase throughput; adding a third fluorescent color, which would allow for absolute quantitation of epigenetic marks; and implementing a recently developed method to sort single DNA molecules based on epigenetic state (35) which would allow for downstream sequencing. Because individual molecules are queried, only small amounts of cellular material are required, raising the possibility of epigenomic analyses of cells that are rare, impossible to culture, or even of single cells (36).
Materials and Methods
Chromatin Preparation.
Native chromatin was prepared as described (10). Briefly, 3 × 107 cells were homogenized using a Dounce homogenizer and Triton X-100 containing PBS buffer. Isolated nuclei were treated with micrococcal nuclease, and chromatin was solubilized in a high-salt, EGTA-containing buffer. Presence of nucleosome ladders was verified and sizes estimated using agarose gels loaded with DNA purified from chromatin (Fig. S1). Samples used for analysis had size distributions centered on 2 kbp. Concentration was assessed spectrophotometrically.
Labeling Epigenetic Probes.
Antibodies were purchased from Active Motif (61013, 39155, 61037, 39763) and the MBD1 probe was prepared as previously described (10). All reagents were then labeled using Invitrogen’s Microscale Protein Labeling kits (A10238, A30009, A30006) according to manufacturer’s instructions and purified from free dye by seven cycles of 10-fold dilution into PBS and reconcentration in Amicon Ultra Centrifuge concentration columns (UFC501008, UFC51008). Labeling was assayed using fluorometry and absorbance, and label density was approximately two dye molecules per antibody. Protein concentrations were confirmed by bicinchoninic acid (BCA) assay.
Binding Reactions.
Native chromatin (3–10 nM) and fluorescent antibodies (125–400 nM) were mixed such that there was a 15–100-fold molar excess of antibody or MBD1 and then left to incubate overnight at 4 °C on a rotator. The following morning 0.75% formaldehyde was added and samples incubated at room temperature (RT) for 15 min, after which 160 mM glycine was added to quench the cross-linking reaction. Once cross-linked, fluorescent DNA intercalator dye (TOTO-3 or YOYO-1, Invitrogen) was added such that there was one molecule of dye for every five DNA base pairs. Samples were left to incubate for 1 h at RT followed by 1 h at 4 °C. Before SCAN, binding reactions were diluted into flow buffer such that the total concentration of fluorescent molecules was below 1 nM.
SCAN and Data Analysis.
SCAN data were collected as described (10). To measure the abundance of epigenetic marks, we determined the fraction of intercalator-stained material bound to the epigenetic probe, and normalized this value to the fraction bound to α-H2B for both test and control cells, and then took the ratio of normalized values for the two cell types according to the formula
where NAE = normalized abundance of epigenetic marks; B e = count of molecules bound to both intercalator and epigenetic probes; I e = count of intercalator bound molecules in epigenetic mark probed sample; B 2 = count of molecules bound to both intercalator and α-H2B; I 2 = count of intercalator bound molecules in α-H2B probed sample; T = test cells; C = control cells. For experiments where combinations of epigenetic marks were quantified the NACE equation was used, where NACE = normalized abundance of combined epigenetic marks; B de = count of chromatin molecules bound to two different epigenetic probes; E = count of total (free and bound) molecules for a given epigenetic probe. All data analysis was performed using MATLAB 2010b (Mathworks). Raw intensity vs. time traces were autocorrelated and fit to the 1D FCS equation with the addition of intersystem crossing and directed flow terms to determine the characteristic transit time of molecules through the focal volume. A noise level of random background photons was established and used to establish a baseline for the intensity traces. Single-molecule events were identified above the baseline by applying a signal threshold and locating local maxima among contiguous points above the threshold. Coincident molecules were identified by temporally translating one channel trace relative to a second, over a time window whose size was determined from the characteristic transit time of the single-molecule events. The number of overlapping local maxima from both channels was determined at each time offset, generating a time distribution of coincident events. By analyzing the time distribution of coincident events from multiple datasets a signal-to-noise ratio was determined and used to calculate background levels and correct the values for abundance of epigenetic mark detection in experiments requiring quantitative analyses (Figs. 3–5). Due to an increased noise level from falsely detected coincident events for experiments where two different colored probes were used, a similar background correction was applied by uniformly subtracting the bulk of background events from total events, effectively shifting the y axis on coincidence plots.
Drug Treatment.
Cells (3 × 107) were cultured in media containing 1.1 ng/mL (HL-60) or 10 ng/mL (MF lines) 5-Aza-2′-deoxycytidine (Sigma A3656) for 72 h with the addition of fresh drug containing media every 24 h. HL-60 cells were cultured in RPMI media; fibroblasts were cultured in DMEM containing 5% (vol/vol) FBS. Chromatin was prepared as described above. Concentrations were chosen that permitted cell accumulation at a level equal to 40–60% that of untreated cells (Fig. S3).
Western Blotting.
Histones were acid-extracted from cells and 3 μg was run on a 15% SDS/PAGE gel. Proteins were transferred to nitrocellulose membranes and blocked with tris buffered saline with 0.1% triton-X-100 (TBST) + 5% (wt/vol) dried milk, then incubated with primary antibody and secondary antibodies for 2 h at RT in TBST +1% milk. Primary antibodies recognized H3K27me3 (Active Motif 39155), H3K9me3 (Active Motif 39162), or H2B (Active Motif 61037). Secondary HRP-labeled antibodies were against either rabbit or mouse Ig, depending on the primary antibody (Millipore 12–348 and Abcam 102448). Chemiluminescent detection of HRP was done using the SuperSignal West Dura Extended Duration Substrate (Thermo 34075). Quantification was performed using ImageJ software.
Supplementary Material
Acknowledgments
We thank Jonathan D. Flax for helpful discussions; Adrian Bird for the MBD1 expression vector; John Schimenti for the 3T3 cells; Geoffrey Wahl for the H2B-GFP-expressing HeLa cells; and Millipore, Active Motif, and Cell Signaling Technology for gifts of antibodies. We are also grateful for funding from National Institutes of Health (NIH) Grants DA025722 and HG006850 (to P.D.S. and H.G.C.) and the Cornell Center for Vertebrate Genomics. P.J.M. was supported by NIH Training Grant GM007617; C.Y.J. was supported by the Empire State Stem Cell Fund through New York State Department of Health Contract C026075.
Footnotes
References
1. Murr R. Interplay between different epigenetic modifications and mechanisms. Adv Genet. 2010;70:101–141. [PubMed] [Google Scholar]
2. Bernstein BE, et al. A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell. 2006;125(2):315–326. [PubMed] [Google Scholar]
3. Ooi SKT, et al. DNMT3L connects unmethylated lysine 4 of histone H3 to de novo methylation of DNA. Nature. 2007;448(7154):714–717. [PMC free article] [PubMed] [Google Scholar]
4. Eaton ML, et al. Chromatin signatures of the Drosophila replication program. Genome Res. 2011;21(2):164–174. [PMC free article] [PubMed] [Google Scholar]
5. Tamaru H, et al. Trimethylated lysine 9 of histone H3 is a mark for DNA methylation in Neurospora crassa. Nat Genet. 2003;34(1):75–79. [PubMed] [Google Scholar]
6. Zilberman D, Coleman-Derr D, Ballinger T, Henikoff S. Histone H2A.Z and DNA methylation are mutually antagonistic chromatin marks. Nature. 2008;456(7218):125–129. [PMC free article] [PubMed] [Google Scholar]
7. Ernst J, et al. Mapping and analysis of chromatin state dynamics in nine human cell types. Nature. 2011;473(7345):43–49. [PMC free article] [PubMed] [Google Scholar]
8. Brinkman AB, et al. Sequential ChIP-bisulfite sequencing enables direct genome-scale investigation of chromatin and DNA methylation cross-talk. Genome Res. 2012;22(6):1128–1138. [PMC free article] [PubMed] [Google Scholar]
9. Johnson L, et al. Mass spectrometry analysis of Arabidopsis histone H3 reveals distinct combinations of post-translational modifications. Nucleic Acids Res. 2004;32(22):6511–6518. [PMC free article] [PubMed] [Google Scholar]
10. Cipriany BR, et al. Single molecule epigenetic analysis in a nanofluidic channel. Anal Chem. 2010;82(6):2480–2487. [PMC free article] [PubMed] [Google Scholar]
11. Montgomery ND, et al. The murine polycomb group protein Eed is required for global histone H3 lysine-27 methylation. Curr Biol. 2005;15(10):942–947. [PubMed] [Google Scholar]
12. Meissner A, et al. Reduced representation bisulfite sequencing for comparative high-resolution DNA methylation analysis. Nucleic Acids Res. 2005;33(18):5868–5877. [PMC free article] [PubMed] [Google Scholar]
13. Ruthenburg AJ, Li H, Patel DJ, Allis CD. Multivalent engagement of chromatin modifications by linked binding modules. Nat Rev Mol Cell Biol. 2007;8(12):983–994. [PMC free article] [PubMed] [Google Scholar]
14. Lehnertz B, et al. Suv39h-mediated histone H3 lysine 9 methylation directs DNA methylation to major satellite repeats at pericentric heterochromatin. Curr Biol. 2003;13(14):1192–1200. [PubMed] [Google Scholar]
15. Johnson L, Cao X, Jacobsen S. Interplay between two epigenetic marks. DNA methylation and histone H3 lysine 9 methylation. Curr Biol. 2002;12(16):1360–1367. [PubMed] [Google Scholar]
16. Komashko VM, Farnham PJ. 5-azacytidine treatment reorganizes genomic histone modification patterns. Epigenetics. 2010;5(3):229–240. [PubMed] [Google Scholar]
17. Lindroth AM, et al. Antagonism between DNA and H3K27 methylation at the imprinted Rasgrf1 locus. PLoS Genet. 2008;4(8):e1000145. [PMC free article] [PubMed] [Google Scholar]
18. Hagarman JA, Motley MP, Kristjansdottir K, Soloway PD. Coordinate regulation of DNA methylation and H3K27me3 in mouse embryonic stem cells. PLoS ONE. 2013;8(1):e53880. [PMC free article] [PubMed] [Google Scholar]
19. Viré E, et al. The Polycomb group protein EZH2 directly controls DNA methylation. Nature. 2006;439(7078):871–874. [PubMed] [Google Scholar]
20. Keshet I, et al. Evidence for an instructive mechanism of de novo methylation in cancer cells. Nat Genet. 2006;38(2):149–153. [PubMed] [Google Scholar]
21. Schlesinger Y, et al. Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de novo methylation in cancer. Nat Genet. 2007;39(2):232–236. [PubMed] [Google Scholar]
22. Ohm JE, et al. A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing. Nat Genet. 2007;39(2):237–242. [PMC free article] [PubMed] [Google Scholar]
23. McCabe MT, Lee EK, Vertino PM. A multifactorial signature of DNA sequence and polycomb binding predicts aberrant CpG island methylation. Cancer Res. 2009;69(1):282–291. [PMC free article] [PubMed] [Google Scholar]
24. Widschwendter M, et al. Epigenetic stem cell signature in cancer. Nat Genet. 2007;39(2):157–158. [PubMed] [Google Scholar]
25. Ohm JE, et al. Cancer-related epigenome changes associated with reprogramming to induced pluripotent stem cells. Cancer Res. 2010;70(19):7662–7673. [PMC free article] [PubMed] [Google Scholar]
26. Doi A, et al. Differential methylation of tissue- and cancer-specific CpG island shores distinguishes human induced pluripotent stem cells, embryonic stem cells and fibroblasts. Nat Genet. 2009;41(12):1350–1353. [PMC free article] [PubMed] [Google Scholar]
27. Hon GC, et al. Global DNA hypomethylation coupled to repressive chromatin domain formation and gene silencing in breast cancer. Genome Res. 2012;22(2):246–258. [PMC free article] [PubMed] [Google Scholar]
28. Soloway PD, Alexander CM, Werb Z, Jaenisch R. Targeted mutagenesis of Timp-1 reveals that lung tumor invasion is influenced by Timp-1 genotype of the tumor but not by that of the host. Oncogene. 1996;13(11):2307–2314. [PubMed] [Google Scholar]
29. Wu H, et al. Dnmt3a-dependent nonpromoter DNA methylation facilitates transcription of neurogenic genes. Science. 2010;329(5990):444–448. [PMC free article] [PubMed] [Google Scholar]
30. Mullen AR, DeBerardinis RJ. Genetically-defined metabolic reprogramming in cancer. Trends Endocrinol Metab. 2012;23(11):552–559. [PMC free article] [PubMed] [Google Scholar]
31. Cardaci S, Ciriolo MR. TCA cycle defects and cancer: When metabolism tunes redox state. Int J Cell Biol. 2012;2012:161837. [PMC free article] [PubMed] [Google Scholar]
32. Cao Z, et al. Analysis of succinate dehydrogenase subunit B gene alterations in gastric cancers. Pathol Int. 2010;60(8):559–565. [PubMed] [Google Scholar]
33. Cervera AM, Bayley J-P, Devilee P, McCreath KJ. Inhibition of succinate dehydrogenase dysregulates histone modification in mammalian cells. Mol Cancer. 2009;8:89. [PMC free article] [PubMed] [Google Scholar]
34. Kruidenier L, et al. A selective jumonji H3K27 demethylase inhibitor modulates the proinflammatory macrophage response. Nature. 2012;488(7411):404–408. [PMC free article] [PubMed] [Google Scholar]
35. Cipriany BR, et al. Real-time analysis and selection of methylated DNA by fluorescence-activated single molecule sorting in a nanofluidic channel. Proc Natl Acad Sci USA. 2012;109(22):8477–8482. [PMC free article] [PubMed] [Google Scholar]
36. Benítez JJ, et al. Microfluidic extraction, stretching and analysis of human chromosomal DNA from single cells. Lab Chip. 2012;12(22):4848–4854. [PMC free article] [PubMed] [Google Scholar]
Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences