With the rapid development and application of sequencing technologies, it is becoming evident that alterations at the genome level are unlikely to explain the heterogeneity of human cancers , and there is increasing recognition that epigenetic mechanisms also play a major role in cancer development and progression. The epigenetic regulation of genes is complex, and occurs at the levels of DNA, histone proteins, and RNA [2-4]. DNA methylation at promoter regions can cause long term gene silencing, whereas histone modifications, predominantly affecting lysines in the N-terminal tails of the core histone proteins H3 and H4, are more dynamic and associated with relatively short term plasticity.
The most frequent lysine modifications of histone proteins (“marks”) involve acetylation and methylation, although phosphorylation, ubiquitylation and sumoylation can also occur at these sites. Histone acetylation, which alters the charge distribution, is typically associated with open chromatin that enables gene transcription, whereas methylation is usually but not always associated with chromatin condensation. Enzymes like histone acetyltransferases and methyltransferases that create these marks are referred to as “writers,” whereas those that remove marks (e.g., demethylases and deacetylases) are “erasers.” A third class of chromatin-related proteins are “readers” . Readers bind to specific histone marks and recruit protein complexes that effect changes in chromatin condensation, gene expression, or DNA methylation. Large numbers of mutations affecting all three classes of protein have been identified in recent years, and linked to cancer progression and to other disease states [2, 5-13]. Therefore, there is considerable interest in the development of novel agents to target these mechanisms [14, 15], and, as a corollary, a need for laboratory techniques able to characterize epigenetic mechanisms in heterogeneous clinical samples, and to monitor treatment effects.
Chromatin immunoprecipitation (ChIP) using antibodies raised against individual histone marks is a basic technique used extensively in epigenetics research, and highly specific ChIP grade antibodies are available from a number of commercial sources. The identification of histone modifications in cells undergoing mitosis (H3S10 phosphorylation) and DNA damage response (γH2AX) are established and relatively straightforward techniques using flow cytoametry [16-18]. However, in its unperturbed state chromatin can be highly compacted and this is likely to influence the accessibility of antibodies to histone modifications involved in epigenetic regulation; particularly those associated with condensed chromatin. Little work has been published using flow cytoametry to study histone modifications linked to epigenetic mechanisms. Obier et al. showed the feasibility to detect a number of individual histone marks by flow cytoametry, and their technique was subsequently applied for cell sorting, although the protocol described was not optimized for clinical samples, and the dynamic ranges shown in the data appear relatively small [19, 20]. Since our eventual goal is to implement flow epigenetics as a clinical tool, we first studied variables affecting the detection of individual histone marks by flow cytoametry using cell lines, and then adapted a protocol that maintained surface immunophenotype and light scatter sufficiently for application to normal and leukemic blood samples.
Materials and Methods
Histone marks for this initial project were selected based on known clinical importance, the ability to modulate levels so as to generate control cells, and the availability of ChIP-grade monoclonal antibodies. Table 1 gives summary information about these histone marks, including the antibodies finally selected.
Table 1. Information about key histone marks developed.
-Gene silencing-Required for repression of developmentally regulated genes during embryonic stem cell differentiation
Active Motif #39683-GAM A647
Tissue Culture and Preparation of Positive and Negative Controls
The AML cell line OCI-AML2 was maintained in suspension culture in Iscove's medium supplemented with 10% fetal calf serum, and used at a concentration of 0.25 to 1 × 106 cells/ml. The levels of acetylated histone H3 lysine 9 (H3K9ac) were modulated by treatment for 24 h with the histone deacetylase (HDAC) inhibitor Vorinostat (SAHA), 2.5 μmol (Selleck Chemicals, Houston, TX), or the histone acetyl transferase inhibitor curcumin (Sigma-Aldrich), 30 μmol. Di-methylated histone H3 lysine 9 (H3K9me2) levels were increased by 24-h incubation in 0.2% oxygen to activate the G9a methyltransferase, and decreased by incubation with the G9a inhibitor UNC 0638 (obtained from Dr Jian Jin, University of North Carolina, Chapel Hill). Tri-methylated histone H3 lysine 27 (H3K27me3) levels were increased by treatment for 96 h with the histone demethylase inhibitor GSK-J4, 10 μmol, and decreased by treatment with the EZH2 histone methyltransferase inhibitor UNC 1999 , 0.5 μmol, for 72 h.
For western blot, cells were collected and washed with PBS two times and complete lysis buffer (63 mM Tris pH 6.8, 10% Glycerol, and 2% SDS. 10 μl/ml of 1 M DTT was added before each use) added to ensure lysis of nuclear membrane. Lysates were heated to 100°C for complete lysis. For flow cytoametry analysis, cells were collected and washed with wash buffer (PBS, w/o CaCl and MgCl + 4% FBS) prior to fixation using protocols outlined below.
Cell and Blood Sample Fixation
Peripheral blood samples were obtained from healthy donors using EDTA anticoagulant, and from 25 newly diagnosed leukemia patients (predominantly AML) according to a protocol approved by the University Health Network Research Ethics Board, and processed within 2 h. Whole blood fixation was done by adding 10% formaldehyde (Polysciences EM grade) directly to the blood sample to a final concentration of 1 or 4% for 10 min at room temperature (RT). Washed OCI-AML2 cells were resuspended in 1, 2, or 4% formaldehyde solution and incubated for 10 min at RT. Fixation was followed by red-cell lysis/cell permeabilization using 0.1% Triton X-100 final volume for 15 min at 37°C. Samples were either resuspended in freezing medium consisting of 10% glycerol and 20% FBS and stored at −20° until needed, or used immediately.
Staining Protocols and Flow Cytometry Analysis
Fixed OCI-AML2 cells or peripheral blood samples were first stained for 30 min with primary antibodies at RT. If secondary antibody staining was required, samples were then incubated for 30 min at RT with relevant secondary antibodies. For peripheral blood samples, a further 30 min staining at RT with relevant surface markers was done. Between each stain samples were washed with wash buffer and after final wash, samples were resuspended in resuspension buffer (0.1% Formaldehyde in PBS w/o CaCl and MgCl). Flow cytoametry was performed using a Gallios™ flow cytoameter fitted with lasers emitting at 405, 488, and 635 nm (Beckman Coulter, Miami, FL). Daily QC included recording mean channel fluorescence for each photomultiplier using FlowCheck™ Pro fluorescence beads, using a constant high voltage setting for each PMT. Mean channel fluorescence for each PMT did not change by more than ±10%. Standard acquisition protocols were developed and used for each experiment, with identical instrument settings used throughout.
Fixed cells were stained as described above, and resuspended in PBS at a concentration of 1 × 106 cells/ml. Cells were centrifuged at 800g for 5 min in specially adapted apparatus in order to adhere cells to coverslips. Cells were stained with DAPI for 10 min at RT, mounted onto a slide, and images were acquired using a Nikon Eclipse epifluorescence microscope with a 60× 1.4/NA oil-immersion objective lens and a Hamamatsu Orca-ER camera, driven by MetaMorph software.
Establishment of Positive and Negative Control Samples
Confocal microscopy confirmed that the staining patterns of all three antibodies were exclusively chromatin-related (data not shown). For this study, we used the HDAC inhibitor Vorinostat and the general histone acetyltransferase inhibitor curcumin to modulate H3K9ac changes; the JMJD3/UTX inhibitor GSK-J4 and the EZH2 inhibitor UNC1999 to modulate H3K27me3 changes; and the G9a/EHMT2 inhibitor UNC0638 to modulate H3K9me2 changes. In addition, hypoxia was used to induce H3K9 dimethylation, as has been demonstrated previously [21, 22]. Each treatment condition was subjected to dose and time response experiments in order to ascertain the optimal concentrations and incubation times (data not shown). By western blot analysis, all three antibodies used gave a single band. Figure 1 shows representative data illustrating the effects of histone mark modulation. Having established optimal conditions for modulation, three separate experiments were done to measure this effect, and semiquantitative analysis based on western blot densitometry is shown in Figure 1.
Optimization of Flow Cytometry Protocols for Tissue Culture
Initial titration curves were prepared using OCI-AML2 cells fixed using 1% formaldehyde, in order to identify a suitable antibody concentration for protocol development. Figure 2 illustrates the dilution curve for an antibody to H3K9me2 obtained from Active Motif, using an indirect labeling technique applied to cells either exposed to hypoxia, which induces the G9a histone methyl transferase, or the G9a inhibitor UNC0638 in order to generate cells with low H3K9me2. From these curves, an antibody concentration of 100 ng/60 μl reaction volume was selected for protocol development.
Effects of formaldehyde concentration
OCI-AML2 cells were fixed for 10 min at RT by the addition of 10% formaldehyde to give a final concentration of 0.5, 1, 2, or 4%, permeabilized using 0.1% Triton X-100, and then stained for H3K9ac, H3K9me2, and H3K27me3. As shown in Figure 3, there was progressive loss of staining with increasing formaldehyde concentrations, and this was particularly pronounced with H3K9me2. This result is anticipated given the highly compact nature of chromatin, and the consequent likelihood of antigen masking following treatment with a cross linking fixative. Treatment with methanol is able to recover some intracellular antigens that are masked by formaldehyde, but we did not achieve this with any of the histone marks.
Effects of salt concentration
It has been reported that fixation under high ionic strength significantly improves the staining of some nuclear antigens . OCI-AML2 cells were fixed with 4% formaldehyde in concentrations of NaCl ranging from 0 to 1.5 M, but we did not observe any significant effect on staining intensity (data not shown).
Comparison of western blot to flow cytoametry
The three histone marks were modulated in OCI-AML2 cells as described above, following which they were fixed in 1% formaldehyde at RT for 10 min, permeabilized using 0.1% Triton X-100, and titration curves for the three antibodies plotted. Optimum antibody concentrations were identified, and then three independent experiments were done in which the histone marks were modulated, and cells prepared for western blot and flow cytoametry. Results are summarized in Figure 4. Overall there was good agreement between the two methods, although the effects of histone modulation appeared greater by western blot, suggesting that this is the more sensitive technique. Treatment with the JMJD3/UTX histone demethylase inhibitor GSK-J4 gave the predicted increase in H3K27me3 levels by western blot, whereas these were slightly decreased when measured by flow cytoametry.
Optimization of Flow Cytometry Protocols for Whole Blood Samples
Next, we examined the potential to assess the heterogeneity of histone modifications in peripheral blood samples. We adapted a protocol that we originally developed for the measurement of phosphospecific antibodies in clinical samples, since we have found that this is a simple and robust approach to combined measurement of surface and intracellular antigens while maintaining the light scatter signals . Similar to the result with OCI-AML2 cells, there was loss of staining intensity with increasing formaldehyde concentration, and this was particularly problematic for H3K9me2. However, as expected from our previous work, staining of important surface antigens such as CD34 was compromised using 1% formaldehyde . Since our long-term goal is to apply this technique to samples from acute leukemia patients, we, therefore, decided to use a 4% formaldehyde protocol for detailed testing. We omitted H3K9me2 from these experiments, and we obtained antibodies to H3K9ac and H3K27me3 from Cell Signaling Technology (Danvers, MA) that were directly conjugated to Pacific Blue and Alexa-488, respectively. Following initial titration curves using OCI-AML2 cells to establish optimum concentrations, these antibodies were incorporated into panels that included the surface markers CD45-Krome Orange and CD34-PC7 (Beckman-Coulter).
Analysis of Normal Donor Blood Samples
Blood samples from five normal donors were fixed using 4% formaldehyde for 10 min at RT followed by red-cell lysis and permeabilization with 0.1% Triton X-100, and then stained with the directly-conjugated antibodies to H3K9ac and H3K27me3. Forward versus orthogonal light scatter was well preserved using this protocol, as previously reported , allowing identification of the lymphocyte, monocyte, and granulocyte populations. Figure 5A shows representative bivariate plots for each of the histone marks versus orthogonal light scatter, and H3K9ac versus H3K27me3. Figure 5B shows the mean values of fluorescence intensity of the two histone marks in leukocyte populations of the five normal samples. It can be seen that both histone marks show heterogeneous distribution within the lymphocyte, monocyte, and granulocyte populations, but the patterns and mean values are quite similar between the individual donors.
Analysis of Leukemia Patient Samples
Fresh blood samples were obtained from a total of 25 newly diagnosed patients presenting to the Princess Margaret Hospital Leukemia Program, selected based on the identification of blast cells and patient consent. Samples were stained using the two histone-specific antibodies plus antibodies to CD34 and CD45, and the leukemic blast population identified based on these surface markers plus light scatter. There was considerable inter- and intra-cellular heterogeneity in the expression of the histone marks in the leukemic population, particularly in the case of H3K27me3 where there was a 20-fold range in the MFI values between individual patients. In contrast, the levels of H3K9ac showed a narrower distribution similar to that seen in the lymphocyte populations of the leukemia patient samples. In this small dataset, we did not see a correlation between the histone mark levels and the leukemic subtype (Supporting Information Fig. 1). Figure 6A shows representative data for the two histone marks in four patient samples, and Figure 6B shows the MFI values for the 25 patient samples.
We were able to show saturation binding characteristics of antibodies to all of the three histone marks selected for this pilot project, using in vitro treatments to augment or attenuate the levels of the histone marks. Results obtained by flow cytoametry were in overall agreement with those obtained in bulk assay by western blot, with the exception of a paradoxical decrease in H3K27me3 following treatment with a demethylase inhibitor. Hypermethylation of this mark results in chromatin compaction [25, 26], and we think it possible that, as a result, many H3K27me3 sites became inaccessible using the current flow cytoametry protocol.
A long-term goal is to apply this technique to patient samples, both as a tool for studying epigenetic mechanisms in heterogeneous clinical samples, and also as an approach to allow monitoring of epigenetic-targeted agents during clinical trials. In earlier papers, Ronzoni et al. showed increased acetylation of histone H4 following treatment with an HDAC inhibitor in vitro, and were able to detect increased labeling in patients treated with valproic acid . Using an antibody to acetylated proteins, Chung et al.  successfully tracked the effects of the HDAC inhibitor MS-275 in peripheral blood and bone marrow samples obtained from study patients. These papers provide proof of principle for the potential to monitor newer, selective agents that target epigenetic mechanisms in patients.
When developing a suitable protocol for blood samples, it became evident that low concentrations of fixative that were optimal for labeling histone marks resulted in the loss of definition of subpopulations by surface markers and light scatter. We found that brief fixation with 4% formaldehyde at RT gave excellent preservation of light scatter and surface markers, while also preserving signals from H3K9ac and H3K27me3, but not H3K9me2. Applying this compromise technique to blood samples obtained from normal donors and leukemia patients, we identified striking heterogeneity in expression levels of the two histone marks, which supports the idea that flow epigenetics might be capable of giving clinically important information not available using conventional techniques.
The normal blood samples showed considerable cellular heterogeneity in the levels of H3K9ac and H3K27me3, although the patterns of staining and mean fluorescence values appeared quite similar between individual donors. Relatively low levels of H3K9ac were seen in the granulocytes, with intermediate levels in the monocytes and highest in the lymphocytes. High levels of H3K27me3 were also seen in the lymphocytes with the monocytes and granulocytes showing similar levels. Lymphocytes showed the greatest heterogeneity of both histone marks, and it will be interesting to determine if this correlates with lymphocyte subpopulations. The low levels seen in the granulocytes might reflect the highly compact chromatin seen in these cells, and it will be interesting to apply this technique to bone marrow samples in order to determine if the histone modifications track normal myeloid differentiation.
When we applied this technique to a consecutive series of 25 newly diagnosed leukemia patients, selected on the presence of circulating blast cells, we noted a striking heterogeneity in the levels of H3K27me3 within the leukemic blasts, compared with the homogeneity of the normal blood samples. In contrast, H3K9ac levels showed a roughly normal distribution, with levels similar to those seen in the mature cell population of normal blood. The H3K27me3 mark is of interest as it is associated with stem cell features, and it can be modified by a number of mutations that occur in AML patients. For example, mutations involving the polycomb repressor complex 2, including EZH2 and ASXL1, which is responsible for methylation at this site, result in decreased H3K27me3, whereas mutations of genes involved in its demethylation, such as iso-citrate dehydrogenase (IDH) 1/2, result in hyper methylation [3, 5, 10]. We, therefore, plan to expand our series of AML patient samples studied with this technique, and then perform mutational analysis of the relevant genes in order to determine if the measurement of H3K27me3 by flow cytoametry is able to provide a rapid screen for mutations.
Based on the results shown in this article, and given the urgent need for robust analytical methods to track epigenetic mechanisms in cancer patients, we believe that flow epigenetics represents a promising and hitherto under-explored technique. We think it encouraging that all three histone marks selected for this pilot study could be successfully measured by flow cytoametry, suggesting that the technique can be extended to include many additional relevant marks. The development of highly specific, high affinity ChIP-grade antibodies to histone modifications is being driven by basic research, and many of these antibodies are probably suitable for flow cytoametry. Furthermore, the histone code is highly conserved across eukaryote species, and we think it likely that flow cytoametry protocols and reagents developed to study human cancers could be readily adapted to other biological applications, including for example invertebrate or plant species. However, a number of cautions need to be kept in mind.
Measurement of individual histone marks at the single cell level cannot identify specific genes that are being epigenetically regulated. Rather, flow epigenetics could be considered as a readout of the function of enzymes that are regulating the epigenome. To that end, additional useful information might be obtained by combined staining with antibodies to the relevant readers, writers, or erasers, as well as to relevant gene targets that are being epigenetically regulated. This can be explored as antibodies become more generally available.
It is likely that chromatin compaction has a major effect on the accessibility of histone specific antibodies, and we found that the H3K9me2 site was particularly vulnerable to masking following formaldehyde fixation. It has been found that the staining of some nuclear antigens is improved following fixation under high ionic strength , or following treatment with alcohols, but we did not observe that with any of the histone marks tested. However, further developmental work could be done in this area, for example, testing alternative fixatives or fixation conditions. It is also evident that protocols not requiring surface antigen staining can use reduced fixation conditions, thereby improving histone labeling. Despite these caveats, and a significant amount of further developmental work, we think that flow epigenetics could evolve into a major new application of flow cytoametry, with potential in a wide range of biomedical fields in addition to cancer.
The authors thank Dr Jian Jin, University of North Carolina, for providing the inhibitors to EZH2 and G9a.