Design, Construction, and Validation of Histone-Binding Effectors in vitro and in Cells

Abstract

Chromatin is a system of nuclear proteins and nucleic acids that plays a pivotal role in gene expression and cell behavior and is therefore the subject of intense study for cell development and cancer research. Biochemistry, crystallography, and reverse genetics have elucidated the macromolecular interactions that drive chromatin regulation. One of the central mechanisms is the recognition of post-translational modifications (PTMs) on histone proteins by a family of nuclear proteins known as “readers”. This knowledge has launched a wave of activity around the rational design of proteins that interact with histone PTMs. Useful molecular tools have emerged from this work, enabling researchers to probe and manipulate chromatin states in live cells. Chromatin-based proteins represent a vast design space that remains underexplored. Therefore, we have developed a rapid prototyping platform to identify engineered fusion proteins that bind histone PTMs in vitro and regulate genes near the same histone PTMs in living cells. We have used our system to build gene activators with strong avidity for the gene silencing-associated histone PTM H3K27me3. Here, we describe procedures and data for cell-free production of fluorescently tagged fusion proteins, enzyme-linked immunosorbent assay-based measurement of histone PTM binding, and a live cell assay to demonstrate that the fusion proteins modulate transcriptional activation at a site that carries the target histone PTM. This pipeline will be useful for synthetic biologists who are interested in designing novel histone PTM-binding actuators and probes.

Graphical Abstract:

Chromatin engineering is a burgeoning field with applications in disease treatment,¹ tissue engineering,² and development of cell lines for industrial production of biologicals.³ Recently, proteins that bind specific histone posttranslational modifications (PTMs) have been used as probes for histone PTM states in cells and to target gene regulators to sites based on histone PTM status.^4,5 Protein motifs that interact with specific acetylated and methylated residues on histone H3 have been used to design fusion fluorescent proteins to label genomic regions that contain high levels of H3K14ac,⁶ H3K9me3,⁷ and H3K4me3 and H3K27me3.⁸ The wealth of basic research that has supported the design and application of recombinant, chromatin-based proteins is discussed in detail in reviews from our group⁹ and others.¹⁰

Here, we describe a workflow to measure binding of recombinant histone-binding proteins in vitro and to validate their activity in live cells using techniques that do not require highly specialized equipment or high-yield protein preparations. Current approaches to validate the function of histone PTM-binding domains (HBDs) include fluorescence polarization (FP),¹¹ isothermal calorimetry (ITC),¹² surface plasmon resonance (SPR),¹³ and peptide arrays.¹⁴ The primary limitation of these approaches is the need for high yields of purified protein, which limits the number and variety of proteins that can be assayed efficiently. Recently, we demonstrated the rapid prototyping of novel multivalent histone-binding transcriptional activators built from a human orthologue called CBX8.¹⁴ The results showed that small volumes (<12 μL) of cell-free transcription−translation lysates (TXTL) can be used to produce sufficient RFP-tagged histone-binding fusion proteins to generate significant binding signal over background in an enzyme-linked immunosorbent assay (ELISA). Here as a proof of concept, we screened a small library of additional TXTL-expressed variants for H3K27me3 avidity. Open reading frames that encoded histone-binding candidates were then cloned into a mammalian expression vector and tested for gene-regulating activity in HEK293 cells. The cells used here include an ectopic heterochromatic site where the target histone PTM is located near the promoter of a reporter transgene (luciferase). Our report includes guidance for identifying other useful target PTMs and genomic loci.

MATERIALS AND METHODS

Constructs for Histone-Binding Protein Fusions.

BioBrick assembly¹⁵ was used to build open reading frames that encoded full-length fusion proteins in the general purpose vector V0120. Full-length, annotated sequences are available at the Haynes lab Benchling Web site (https://benchling.com/hayneslab/f_/rmSYkAAU-synthetic-chromatin-actuators-2-0/).

DNA Constructs for Cell-Free Transcription−Translation (TXTL) Expression.

Constructs were assembled via directional cloning of PCR-amplified and digested fusioncoding regions (from vector V0120) into pET28 as previously described¹⁴ with the following modifications: EcoRI and HindIII restriction sites were used instead of BamHI and XhoI sites. All histone PTM-binding domains (HBDs) were separated by a [GGGGS]x4 linker. Human (Pcα) = CBX8 (amino acids 1−62). Fish (Pcβ) = CBX2 or pc1 (amino acids 2−63). Fly (Pcγ) = Pc (amino acids 16−77).⁵ Plasmids were cloned in compatible strains: pET28 fusion-expressing constructs in Escherichia coli BL21 and σ70-T7 RNA polymerase plasmid in E. coli KL740.

Enzyme-Linked Immunosorbent Assays (ELISAs).

Capture antibodies included mouse anti-6-histidine (Abcam 18184) and chicken anti-mCherry (Novus biologicals NBP2-25158). Biotinylated peptides were H3K27me3 (21−44) (Anaspec AS-64367–025) and H3 (21−44) (Anaspec AS-64440–025). Primary antibodies were chicken anti-mCherry (Novus biologicals NBP2–25158) and mouse anti-6-histidine (Abcam 18184). Secondary antibodies were rabbit anti-chicken HRP (Gentel, 0.5 mg/mL, RCYHRP) and goat anti-mouse HRP (KPL 074–1806). Data shown in this report were generated using a Synergy H1 Hybrid Reader with the following key settings: standard 96-well plate option; detection, absorbance; read type, end point; read speed, normal. We have obtained very similar A₄₅₀ values [absorbance units (a.u.)] using two different plate reader models (Synergy H1 Hybrid Reader and PerkinElmer EnVision 2104 Multilabel Reader). For each Pc fusion, the A₄₅₀ value for each H3K27me3 well (four replicates) was divided by the mean A₄₅₀ value of four H3 (unmodified histone) control wells. “Histone peptide capture: relative binding” is the mean of the four normalized values. Nonspecific binding was assessed by using a fusion protein that contained no HBD (PcΔ¹⁴). We used two methods to determine histone peptide capture as percent fusion protein input, where either the mean six-His capture A₄₅₀ value or the TXTL RFP end point signal (533−610, a.u.) was used to represent input.

DNA Constructs for Mammalian Cell Expression.

Constructs were assembled via nondirectional cloning of fusion-coding regions (from V0120) into vector MV10 as previously described.¹⁴ Colony PCR was used to screen for forward inserts: 1× GoTaq green master mix (Promega M7122), 0.4 μM forward primer 5′-caccatcgtggaacagtacg, and 0.4 μM reverse primer 5′-gcaactagaaggcacagtcg in a final volume of 25 μL. PCR was performed as follows: 95 °C for 2 min; 25 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 1 min; 72 °C for 5 min; and held at 4 °C. Candidate clones were grown in 3 mL of liquid LB and 100 μg/mL ampicillin. Restriction digestion followed by electrophoresis (1% agarose in TAE) was used to analyze purified plasmid DNA (Sigma PLN350–1KT): 500−1000 ng of plasmid DNA, 1 μL each of FastDigest enzymes XbaI and PstI, and 1× FastDigest buffer (Thermo Fisher FD0684, FD0614, and B72).

Cell Culture and Transfection.

HEK293 Gal4-EED/luc cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% tetracycline-free fetal bovine serum and 1% penicillin and streptomycin (pen/strep) at 37 °C in a humidified CO₂ incubator. Silencing of the reporter gene (Tk-luciferase) was induced by supplementing the medium with 1 μg/mL dox for 96 h. For wash-out of doxycycline (to allow depletion of Gal4-EED), growth medium was removed and replaced with dox-minus medium. Prior to transfection, doxtreated cells were plated in 12-well culture dishes at 40% confluency (~1.0 × 10⁵ cells/well) in 2 mL of pen/strep-free growth medium. Transient transfections were performed by adding 300 μL of DNA/Lipofectamine complexes to each well: 1 μg of plasmid DNA or ddH₂O for mock transfections (10 μL), 5 μL of Lipofectamine LTX (Invitrogen 15338100), or 285 μL of OptiMEM (Gibco 31985062). Plates were spun at 100g for 5 min to increase the transfection efficiency and then incubated at 37 °C in a humidified CO₂ incubator.

RFP, Hoechst, and Luciferase Plate Reader Assay Data Analysis.

For each channel, luciferase (chemiluminescence), RFP, and Hoechst 33342, the average signal from the 1× PBS wells was subtracted from the value for every sample well. Next, the average background-subtracted luciferase value for un-transfected (UT) cells was subtracted from each experimental luciferase value. UT-subtracted luciferase values were scaled by multiplying each value by [(mean UT Hoechst)/(sample Hoechst)] and then by [(mean Pcαα RFP)/(sample RFP)].

Statistical Analyses.

Reported p values were determined by a Student’s paired t test, with a two-tailed distribution.

RESULTS

Overview of Histone PTM-Binding and Gene Regulation Assays.

Binding of TXTL-expressed, fluorescently tagged fusion proteins with target histone peptides is assessed by a user-friendly ELISA. Previous work has shown that bacterially expressed, eukaryotic histone PTM-binding domains (HBDs) have specific avidity for their ligands (reviewed in ref 9). Therefore, TXTL in bacterial lysates is suitable for producing ligand-binding HBDs and allows researchers to circumvent cell transformation or transfection. Ligands (modified histone tail peptides conjugated with biotin) are immobilized in neutravidin-coated microwells, and TXTL-expressed fusions are incubated in the wells to allow binding. Next, candidate regulators are transferred into mammalian expression vectors to determine gene regulation activity in live cells, where the fusion proteins are expected to interact with whole nucleosomes. Excellent work reported by others has utilized single nucleosomes and nucleosome arrays as templates to study histone PTM-binding proteins.^16–19 Because our workflow aims to identify proteins that interact with histone PTMs and activate the transcription of target genes, we use chromatin in cellulo rather than reconstituted nucleosomes or nucleosome arrays. However, we encourage researchers to perform experiments using whole nucleosomes to, for instance, determine the impact of PTM spacing over the nucleosome on the kinetics of binding.

Design and Construction of Histone PTM-Binding Fusion-Expressing Plasmids.

In this report, we focus on PTMs within the unfolded N-terminal tail domain of histone H3. These PTMs represent some of the most well-characterized modifications in regard to the proteins that recognize these marks and their impact on gene regulation.⁹ Our group is interested in designing and optimizing histone-binding transcriptional activators that recognize elevated levels of H3K27me3 in cancer cell epigenomes and induce activation of repressed anticancer loci. The procedure described herein uses a library of fusion proteins built for this purpose as an example (Figure 1A). The fusions contain conserved, H3K27me3-binding polycomb chromodomains (PCD) from human (CBX8), fish (Danio rerio CBX2, pc1), and/or fly (Drosophila melanogaster CBX2, pc1) orthologues. These PCD orthologues contain 62 amino acids, including three conserved residues that make up an aromatic pocket that binds H3K27me3.^20,21 The 59 remaining residues show a lower level of conservation; reports suggest that some of these residues may contribute to stabilizing the interaction of PCD with the histone tail.^20,21 In our previous work, we directly compared the human, fly, and fish PCD orthologues by testing each as a histone-binding module in a fusion activator protein.⁵ The fusion protein that contained a single human CBX8 PCD activated target genes more effectively than the fish or fly homologues. Unknown differences in protein processing, such as folding and degradation, prevented a full understanding of the differences in fusion protein activity. In our current work, we investigated the three PCD orthologues to determine H3K27me3 binding under controlled conditions in vitro and to compare the results with gene regulation in live cells.

Figure 1.

Constructs, plasmids, and a schematic for TXTL expression. (A) We constructed a library of nine fusion histone-binding protein open reading frames (ORFs) with tandem N-terminal HBDs separated by a 20-amino acid flexible linker (4×[GGGGS]). The 3D structure for CBX7 bound to H3K27me3 (Protein Data Bank entry 4X3K) is shown to represent CBX8/H3K27me3. (B) Plasmid maps, drawn to scale, used for expression in TXTL. Fusion protein ORFs were cloned into the EcoRI, HindIII site of pET28. A T7 RNA polymerase-expressing plasmid was included in the TXTL reactions to supply T7 polymerase, which is not present in the lysate mix. Legend: 6H, six-histidine; Term., transcription terminator sequence.

Generally, we recommend keeping the N- or C-terminal position of the histone PTM-binding domain (HBD) in the fusion protein consistent with its native context (e.g., the H3K27me3-binding PCD is an N-terminal motif) (Figure 1A). In our hands, a linker peptide is apparently unnecessary at the HBD−fluorophore junction in fusion proteins that show histone-binding activity. However, we do recommend the inclusion of linker peptides between tandem HBDs. Recent work has shown that combinatorial PTM-binding motifs increase the overall avidity of the fusion for its target^14,22,23 or allow for specific co-recognition of two distinct histone PTMs within a single nucleosome.⁸ Pairs of H3 as well as H4 nucleosome tails are expected to be oriented in cis so that they protrude away from the nucleosome in the same general direction (see Protein Data Bank entry 1AOI²⁴). These histone tails are flexible enough to come into the proximity of each other. Therefore, a flexible (GS repeat) or rigid (α helix) peptide of 20 amino acids is a good starting point for HBD− HBD linker design. A three-dimensional (3D) structure-based model of this format can be found in our previous work.¹⁴ It is critical to include in the fusion protein a fluorophore with efficient folding properties, such as the monomeric Cherry (mCherry) we have used in our work. The tag enables protein detection after expression in TXTL or in cultured mammalian cells. The transcription-regulating domain (activator or repressor) should be included at the in vitro testing stage so that any effects of whole protein folding on binding with the target ligand can be observed. A standard directional cloning approach (e.g., BioBrick assembly¹⁵) or scarless assembly (e.g., Golden Gate²⁵) is suitable for construction of the fusion protein-expressing plasmid DNA (Figure 1B). Two plasmids, one encoding a fusion protein and one expressing T7 RNA polymerase under the control of a constitutive σ70 promoter, are added to the TXTL reagent for the expression reaction.

Real Time Detection of Cell-Free Transcription− Translation (TXTL).

TXTL is a flexible system that enables expression of recombinant, ready-to-use proteins without the need for cell lysis and purification. In reaction mixtures containing ~100 ng of fusion-encoding plasmid, an increase in the intensity of the RFP signal began at just >2 h and continued to increase linearly up to 19 h (Figure 2). Our previous work with similar fusion proteins, including the Pcαα variant used here, showed a sigmoidal trend for RFP signal over time in TXTL reactions and began to plateau at 12−16 h.¹⁴ The slower, varying linear increases we observed in the current experiments could be due to incomplete translation, degradation, or misfolding of the fusion proteins. A Western blot with an antibody against mCherry (Figure S1) showed that the full-length proteins were the predominant product in a representative set of TXTL reactions. Another possible cause of variability is inefficient production of fusion proteins due to batch-to-batch differences in components of the TXTL reagent (e.g., ribosomes, ATP, nucleases, proteases, etc.) or variable plasmid quality (i.e., supercoiled or nicked DNA). In the Western blot, anti-mCherry signal intensities generally agreed with end point RFP signal values. We conclude that RFP signal variation is likely due to differences in production.

Figure 2.

Real time detection of RFP-tagged fusion protein expression in TXTL. Graphs show the RFP signal captured every 10 min over ~16 h at a temperature of 29 °C, detected by a Roche Light Cycler 480 real time thermal cycler. The negative control (T7) contains the polymerase vector without the fusion-expressing pET28 vector. The grid shows the mean end point RFP signal values (bold) and standard deviations.

The general procedure used for TXTL is as follows. (1) Set up a TXTL reaction by combining 9 μL of Sigma 70 Master Mix (myTXTL, Arbor Biosciences), 1 μL of 5 nM P70a-T7 RNA pol vector (Arbor Biosciences), and 2 μL of 50 ng/μL T7 promoter-driven template vector (pET28) in a total volume of 12 μL in one well of a white, 96-well plate (Roche 04729692001). (2) Seal the plate with a clear film (Roche 04729757001) and run a Roche 480 LightCycler II program (or machine of your choice) as follows: 29 °C for 10 min and 30 °C for 1 s, scan at 533−610 nm twice, and repeat 96 times (total of 16 h).

ELISA Capture for Quantification of Protein Generated by TXTL.

Prior to histone PTM-binding experiments, total input (TXTL-generated fusion protein) should be determined with a reliable assay. In this report, we consider two parameters to represent input: end point fluorescent signal from the TXTL reaction (Figure 2) and signal values from control ELISA wells where antibodies are used to capture the TXTL products. The TXTL RFP signal value is readily available and does not require additional experiments. However, this value may not correspond precisely with total full-length, histone-binding protein if incomplete folding of mCherry has occurred. Control ELISA wells using high-affinity antibodies against the fusion proteins circumvent the problem of misleading RFP signal values.

To determine the best epitopes for capture and detection, we compared different capture antibodies (Figure 3A): one against the fluorophore within the fusion (mCherry RFP) and the other against the six-histidine motif that appears at both the N-and C-termini of the fusion proteins. The ELISA HRP signals generated from the six-histidine capture were stronger than those from the anti-RFP capture. This is perhaps due to the higher ratio of six-histidine epitopes: two per fusion protein molecule, compared to the single mCherry tag (Figure 1B). The six-histidine capture data and the end point RFP signal values from real time detection of RFP in TXTL reactions show a modest correlation [R² = 0.61 with the Pcγα outlier removed (Figure S2B)]. Therefore, the RFP signal may be used to roughly approximate the amount of fusion protein. However, epitope capture (e.g., via anti-six-histidine) should be used as a control to more accurately determine fusion protein input.

Figure 3.

Relative protein levels and H3K27me3 binding of TXTL-expressed fusions determined by an ELISA. (A) To determine relative protein levels, three replicate TXTL reaction mixtures per fusion protein (from Figure 2) were combined and 1.4 μL of the total TXTL reaction mixture was brought to 50 μL with 5% skim milk in 0.2% PBST and applied to each well containing immobilized anti-six-histidine (anti-his) or anti-mCherry (anti-RFP). Primary and secondary antibodies used in the detection step are shown in the table above each graphic. The grid shows mean values (bold) and standard deviations (±). (B) Interaction of each Pc fusion with the target ligand was assessed by histone peptide capture using H3K27me3 (green/blue) or unmodified H3K27 (negative control, not shown) anchored to immobilized neutravidin (orange). “Histone peptide capture: relative binding” is the mean A₄₅₀ value from four replicate H3K27me3 wells, each normalized to the mean A₄₅₀ for H3K27. (C) The top chart shows relative H3K27me3 capture scaled to six-His capture (A, bottom left). The bottom chart shows relative H3K27me3 capture scaled to mean RFP (from Figure 2). Legend: error bars, standard deviation; HRP, horseradish peroxidase; FP, fluorescent protein; ED, effector domain for gene regulation.

The general procedure for the ELISA control experiment is as follows. On the day the TXTL reaction is being set up and run (1 day before the ELISA experiment), prepare ELISA plates. (1) Add 50 μL of 2 μg/mL capture antibody in 200 mM carbonate buffer (pH 9.6) to each well. (2) Seal well with sealing foil (Cryostuff FS100) and incubate at 4 °C overnight. All subsequent incubations take place at 800 rpm at room temperature unless otherwise noted. (3) On the day of the ELISA, remove foil, shake out the coating solution, and tap the inverted plate on paper towels to remove excess liquid. (4) Rinse the plate with 200 μL of 0.2% PBST (10 mM Na₂HPO₄, 2 mM KH₂PO₄, 2.7 mM KCl, 137 mM NaCl, and 0.2% Tween 20) three times for 3 min each. (5) Block the plate with 200 μL of 5% BSA (Cell Signaling Technologies 9998S) in 0.2% PBST for 30 min. (6) Rinse the plate with 200 μL of 0.2% PBST three times for 3 min each. (7) Dilute the TXTL reaction mixture (~11 μL after evaporation) into 400 μL of 5% skim milk in 0.2% PBST. (8) Add 50 μL of the TXTL/skim milk solution to each well and incubate for 1 h. (9) Rinse the plate with 200 μL of 5% skim milk in 0.2% PBST three times for 3 min each. (10) Add 100 μL of the 1:3000 primary antibody diluted in 5% skim milk in 0.2% PBST to each well and incubate for 1 h. (11) Rinse the plate with 200 μL of 5% skim milk in 0.2% PBST three times for 3 min each. (12) Add 100 μL of the 1:3000 secondary, HRP-conjugated antibody diluted in 5% skim milk in 0.2% PBST to each well and incubate for 30 min. (13) Wash the plates five times with 200 μL of 0.2% PBST for 3 min each. (14) Remove buffer and add 100 μL of the one-step Ultra TMB-ELISA solution (Thermo 34029) to each well, protect from light, and incubate for 15 min. (15) Stop the reaction with 100 μL of 2.0 M sulfuric acid and incubate for 2 min. Measure A₄₅₀ using a plate reader of choice (see details in Materials and Methods). See Figure 3A for an illustration of immunocapture and detection.

Initial First-Pass Screening for Histone PTM Binding.

An enzyme-linked immunosorbent assay (ELISA) is used to assess histone PTM binding by measuring the amount of fusion protein that remains bound to an immobilized target ligand (histone peptide) after washing away unbound and nonspecifically bound proteins.

The procedure for measuring histone peptide capture includes the steps from “ELISA Capture for Quantification of Protein Generated by TXTL” with the following modifications. (1) Coat a 96-well plate (Greiner Bio-one 655101) with 20 ng/ μL neutravidin (Invitrogen 31000) in PBS [10 mM Na₂HPO₄, 2 mM KH₂PO₄, 2.7 mM KCl, and 137 mM NaCl (pH 8.0)] and add 50 μL to each well. (2) Follow steps 2−6 and then dilute biotinylated peptides to 1 μM in 0.2% PBST. Add 50 μL of the appropriate peptide solution to each well and incubate the plate at 800 rpm for 1 h. Following the peptide incubation, rinse the plate with 200 μL of 0.2% PBST three times for 3 min each. Block the plate with 200 μL of 5% skim milk (Apex 20241) in 0.2% PBST for 30 min. After this step, continue the protocol starting at step 7. See Figure 3B for an illustration of capture via immobilized histone peptides and detection.

We used the ELISA-based histone peptide capture procedure to generate data for the library of fusion proteins described in Figure 1. Normalized HRP signal values from the H3K27me3 capture showed that tandem human orthologues of the PCD motif (variant Pcαα) conferred the strongest binding (Figure 3B). ELISA trials using additional replicate TXTL products agreed with this general observation (Figure S3). For the other variants, we observed relative binding values that were significantly higher (p < 0.029) than PcΔ, but at no more than half the level of Pcαα. Overall, these data suggest that Pcαα has the strongest preference for H3K27me3 versus unmodified H3 and that this activity is reduced when the second PCD is changed to the fly for fish orthologue. These results also provide the first in vitro evidence to suggest that low level of relative binding of the fly and fish HDBs with H3K27me3 may contribute to the weaker activity of the fusion gene regulators in human cells, as previously reported.⁵

Next, we investigated histone peptide-binding activity by determining the proportion of input protein that remained bound in the H3K27me3 and H3 histone peptide capture wells (Figure 3C). When six-histidine capture was used to represent input, the values for all variants with the fish (β) or fly (γ) orthologue in position 1 were at least half of or slightly higher than Pcαα (Figure 3C, top chart). These variants also showed H3 peptide capture values significantly higher than Pcαα (p < 0.017), suggesting higher levels of nonspecific binding. When the data were calculated using the TXTL end point RFP signal to represent input, relatively high or low RFP values changed the relationships between values; e.g., the values for Pcβγ and Pcγα become higher and lower, respectively, than Pcαα (Figure 3C, bottom chart). Discrepancies between the six-histidine epitope capture values and RFP signal lead to different interpretations of the data and underscore the importance of using a reliable internal control to determine input protein levels in the ELISA-based experiments.

Determination of Target Specificity and Apparent K_d.

After first-pass screening for fusions that preferentially bind the histone PTM of interest, ELISA and microspot arrays can be used to determine apparent dissociation constants (K_d^app)¹⁴ as well as target histone PTM selectivity. Both histone peptide capture techniques detect the amount of protein that remains bound after washing away nonspecifically bound proteins. Therefore, the calculated K_d^app will be biased toward the off-kinetics. Previously, we used E. coli-expressed, purified fusion proteins and an HRP ELISA to assess binding as a function of histone PTM concentration in mixtures of immobilized on-target PTMs (H3K27me3) with unmodified peptides (H3). We used direct visualization of mCherry (RFP) on microspot arrays to assess binding as a function of fusion protein concentrations.

TXTL potentially produces a suitable amount of protein for quantitative analysis. For our previously reported HRP ELISA binding curves,¹⁴ 100 nM (50 μL) E. coli-expressed, purified Pcαα showed a ratio of ~5:1 for signal (binding to 100% H3K27me3) to background (100% H3K27), as well as significant signal for 20−90% H3K27me3. In the current study, we observed that 1.4 μL of TXTL-expressed Pcαα produced a HRP ELISA signal:background ratio of ~6:1 (Figure 3). Therefore, we would expect to detect a range of binding values (above background) if TXTL-produced Pcαα protein were exposed to a H3K27me3 dilution series (as previously described¹⁴). This should hold true for other Pc fusion variants with similar HRP ELISA values.

An important feature of histone PTM-binding domains (HBDs) is their ability to discriminate between distinct histone tags within a cellular chromatin landscape that is decorated with a wide variety of modifications. Therefore, it is important to perform histone peptide capture using a variety of potential “off-target” PTMs, e.g., H3K27me3 versus K4me3, K9me3, and K36me3, or different degrees of methylation (me1−me3). In our previous work, ELISA experiments showed that Pcα fusions had no interaction with K4me3 or K9me3 that was significantly stronger than that of unmodified H3.¹⁴ Microspot binding assays have a substantial advantage over an ELISA in that the former supports miniaturization, which allows a variety of histone PTMs at several concentrations to be tested in a single application of soluble protein. One challenge for this approach is that it does not allow signal amplification (i.e., via HRP) prior to detection and thus requires a sufficient amount of protein to visualize linked tags (e.g., mCherry used here) or immunostaining of the histone-peptide-bound fusion proteins. In our previous work, purified Pcαα¹⁴ produced a detectable signal (~2000 fluorescence units) on H3K27me3 microspots (1 mm each) with no background signal from unmodified H3 peptides when 0.5 mL of 100 nM fusion protein was applied over an area of 187 cm². Theoretically, if 1.4 μL of TXTL is equivalent to 50 μL of 100 nM purified protein (as suggested by the HRP ELISAs that used purified proteins), we could expect to require at least ~35 μL of TXTL to produce a detectable signal on a microspot array. Because the reaction volume used here is 11 μL, a few reactions may need to be combined for a sufficient yield. We encourage researchers to review the work of Tekel et al.¹⁴ and Filippakopoulos et al.²⁶ for examples of glass slide- and cellulose membrane-based histone peptide capture experiments, respectively.

Construction of Plasmids for Mammalian Cell Expression.

We have constructed a vector called MV10¹⁴ for the overexpression of fusion proteins in frame with a nuclear localization sequence and six-histidine tag. A single XbaI cloning site is included to accept BioBrick RFC23 standard inserts flanked by XbaI and SpeI overhangs (Figure 4). Nondirectional cloning theoretically yields equal proportions of forward (desired) and reverse insertions. We recommend the following colony PCR protocol to quickly screen for forward insertions. (1) Assemble constructs via nondirectional cloning of fusion-coding regions into vector MV10 using restriction digests and ligations as previously described.¹⁴ (2) Prepare colony PCR mixes as follows: 1× GoTaq green master mix (Promega M7122), 0.4 μM forward primer 5′-caccatcgtggaacagtacg, and 0.4 μM reverse primer 5′-gcaactagaaggcacagtcg in a final volume of 25 μL in each 0.5 mL tube. (3) Pick colonies grown from ligation-transformed E. coli with a sterile, disposable micropipette tip, streak onto a small area (1 cm²) on a gridded and labeled LB agar, 100 μg/mL ampicillin plate, swirl the same tip in each PCR mix, and discard the tip. Incubate the streak plate overnight at 37 °C. (4) Perform PCR as follows: 95 °C for 2 min; 25 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 1 min; 72 °C for 5 min; and held at 4 °C. (5) Analyze 20−50% of each reaction mixture via gel electrophoresis. (6) Inoculate 5 mL of liquid LB, 100 μg/mL ampicillin cultures with candidate clones from the streak plate, grow for 18 h at 37 °C with shaking, and miniprep the plasmid DNA (Sigma PLN350–1KT). (7) Perform restriction digests as follows: 500−1000 ng of plasmid DNA, 1 μL of each of FastDigest enzymes XbaI and PstI, and 1× FastDigest buffer (Thermo Fisher FD0684, FD0614, and B72). Analyze the products using standard agarose gel electrophoresis with TAE buffer.

Figure 4.

Vector for mammalian cell expression and scheme for verifying constructs. The MV10 vector is designed to accept XbaI − SpeI donor fragments at a XbaI site so that the open reading frame (red) is cloned in frame with the translation start site within the yeastderived Kozak (K) sequence and the downstream nuclear localization sequence, six-histidine tag, and stop codon. The downstream SpeI/ XbaI junction is a mixed site (5′-actaga) that cannot be recut. Concurrence of both results shown for the validation assays is unique for a single forward insert for any construct that includes the Cterminal mCherry-VP64 sequence used here.

Determination of Transcriptional Regulation in Mammalian Cells: Microwell Assay for Luciferase.

Standard transfection techniques (polyplex or electroporation) can be used to deliver fusion-expressing plasmids into cells. In our hands, transiently transfected cells (nonhomogeneous) for which the transfection efficiency is ≥50% (determined by fluorescence microscopy and flow cytometry) is sufficient for quick validation of large sets of plasmids. For more rigorous quantitative analyses of histone PTM-binding candidate fusion regulators, we discuss a strategy for generating a stable transgenic line in Limitations.

We have used a H3K27me3-enriched locus as a target to test the function of PCD fusion regulators (Figure 5A). In this system, accumulation of H3K27me3 near the promoter of a luciferase target gene in HEK293 cells is controlled by adding doxycycline to the growth medium (as described in Cell Culture and Transfection). Other mammalian ectopic chromatin systems have been developed for H3K27me3,²⁷ H3K9me3,^27,28 and acetylation of H3 and H4.²⁹ These systems provide a powerful platform for comparing the activity of libraries of histone PTM-binding regulators but represent a very limited set of PTMs; we discuss endogenous loci as an alternative option in the next section.

Figure 5.

Microwell plate reader assay for assessing fusion protein expression and regulation of a target reporter gene. (A) Gal4-EED/luc cells were treated with dox to induce ectopic recruitment of polycomb repressive complex 1 (PRC1) and accumulation of H3K27me3 and PRC2 at a luciferase reporter. (B) After luciferase silencing, cells were transfected with each fusion-expressing plasmid, harvested, and aliquoted into a 96-well plate. In this procedure, RFP is used to assess fusion protein expression, the signal from the Hoechst 33342 DNA stain is a proxy for cell loading, and luciferase activity indicates gene expression induced by each Pc activator fusion. (C) Bar charts show mean background-subtracted values from triplicate wells, normalized and scaled as described in Materials and Methods. Legend: error bars, standard deviation; UT, untransfected cells.

Regulation of a reporter gene at ectopic chromatin is performed as follows. (1) Transfect cells and grow for 3 days to allow fusion protein expression and regulation of the target. (2) Collect semiadherent cells with 1× PBS washes, pellet (200g, room temperature, 5 min), and resuspended in 1.5 mL of 1× PBS. (3) Transfer cells to Black Costar Clear Bottom 96-Well Plates (Corning 3631). To test each sample in triplicate, load three wells with a 200 μL cell suspension for RFP detection, three wells with a 100 μL cell suspension and 100 μL of 2× Hoechst 33342 stain (2 μg/mL, Invitrogen H3570), and three wells with 100 μL of cells and 100 μL of prepared Luciferase Assay Buffer with D-luciferin substrate (Biotium 30085). Add 200 μL of 1× PBS to one or more columns of wells to determine background noise. Allow the plate to incubate at room temperature protected from light for 10 min, allowing cells to settle for optimal fluorescence reads. (4) Scan the plate in a Biotek Synergy H1 microplate reader with the following parameters: read, fluorescence at 580−610 nm (RFP), gain set to 100; read fluorescence, 360−460 nm (Hoechst 33342), gain set to 75; orbital shake, 5 min; read chemiluminescence (luciferase activity). (5) Determine fluorescence signal detection limits of the plate reader with a set of standard samples as described in Figure S4. (6) Analyze the data from each channel separately. Subtract the average background (1× PBS wells) from each value. See Figure 5B for an illustration of a microwell plate configuration.

Results from the in vitro histone peptide capture assays suggest that Pcα confers H3K27me3-selective binding to HDB fusion proteins (Figure 3B). The other orthologues might enhance general, nonspecific interaction with unmodified histone H3, as indicated by relatively high input-normalized values (Figure 3C). To investigate the significance of the observed histone peptide-binding activities for gene regulation activity, we tested five fusion protein variants in the mammalian cell assay: two that showed high and moderate levels of relative binding with H3K27me3 and low levels of H3 binding (Pcαα and Pcββ), one that showed moderate levels of relative binding and high levels of H3 binding (Pcγα), other variants that showed low levels of relative binding and contained at least one human PCD orthologue (Pcαβ, Pcαγ, or Pcβα), and the truncated PcΔ protein (negative control). Cells expressing Pcαα showed the highest levels of luciferase reactivation (luciferase enzyme activity), and the only mean value that was higher (p = 0.058) than PcΔ after scaling by the RFP signal (Figure 5C). This result suggests that a high relative binding level of a fusion protein with H3K27me3 from the ELISA-based assay may predict effective transcriptional activation at a silenced chromatin site in cells. Overall, our results demonstrate that an ELISA using TXTL-expressed proteins is an effective method for identifying histone PTM-binding regulators that can activate target genes in live cells.

Determination of Gene Regulation in Mammalian Cells: RT-qPCR Assay for Endogenous Genes.

If the Gal4-EED/luc system or other similar reporters are not available or relevant for one’s own project, we recommend an expression assay based on an endogenous target locus (or several loci) with a well-characterized epigenetic state. These genes can be identified as those reported to show changes in expression when histone-modifying enzymes are disrupted by inhibitors or RNAi-mediated knockdown. For instance, our group identified a set of 14 genes for which polycomb-mediated silencing was supported by genetic and pharmacologic disruption studies in human cancers and stem cells.⁴ Epigenetic regulation can be confirmed by investigating genomic chromatin feature maps. Table 1 presents publicly available data from whole-genome chromatin profiling experiments (chromatin immunoprecipitation followed by deep sequencing, ChIP-seq) to assist in the identification of target loci in widely used model cell lines.

Table 1.

Recommended Human Cell Lines for Identifying Endogenous Target Sites^a

		ChIP-seq data sources
histone H3 PTM	corresponding HBDb	HEK293 (kidney)	MCF7 (breast)	K562 (blood)	H1-hESC (stem)
H3K4me3	PHD8	ENCSR000DTU	ENCSR000DWJ	ENCSR668LDD	ENCSR000AMG
H3K9me3	CD31	ENCSR000FCJ	ENCSR000EWQ	ENCSR000APE	ENCSR000APZ
H3K14ac	BRD6	c	c	c	ENCSR057BTG
H3K27me3	PCD5,8,14	c	ENCSR000EWP	ENCSR000EWB	ENCSR000ALU
H3K27acd	YEATS32	ENCSR000FCH	ENCSR000EWR	ENCSR000AKP	ENCSR880SUY

^aDetailed information for HBDs tested in vitro and in cells is available in the cited references and in our recent review.⁹ Data ID numbers from the ENCODE project online database at https://www.encodeproject.org/experiments are listed. These can be used to access and download BAM files. To view ChIP enrichments, we recommend loading the BAM file into the Integrated Genome Viewer from the Broad Institute (free downloadhttp://software.broadinstitute.org/software/igv/).

^bAbbreviations: PHD, plant homeodomain; PCD, polycomb chromodomain; CD, chromodomain; BRD, bromodomain; YEATS, “Yaf9, ENL, AF9, Taf14, Sas5” domain.

^cNot available in ENCODE.

^dAlso expected to interact with H3K9ac and H3K18ac.

Reverse transcription followed by quantitative PCR (RT-qPCR)³⁰ is a standard, flexible method for determining changes in gene transcription. For each experiment, a minimum of three transcripts must be measured: a housekeeping gene (e.g., GAPDH), the fusion regulator, and the regulated target gene. We have successfully used RT-qPCR to measure changes in luciferase expression in the Gal4-EED/luc reporter cell line to validate luciferase assay results¹⁴ and to measure the expression of several endogenous genes in the presence of a H3K27me3-binding fusion activator called PcTF.^4,5 In our hands, it is not necessary to sort for fusion protein-positive cells prior to RT-qPCR because the gene expression values can be scaled by RFP signal (flow cytometry) or fusion mRNA levels as needed. (1) Transfect ~1.0 × 10⁵ cells per well as described in Materials and Methods. Include a mock-transfection control: empty vector or water instead of fusion protein-expressing DNA. Grow for 2−3 days in an appropriate cell culture medium. (2) Prepare total RNA from ~1.0 × 10⁶ cells using a method similar to that previously described by our group.^2,3 We recommend direct lysis of pelleted cells with 500 μL of TRIzol (Thermo Fisher 15596026) followed by spin column purification of RNA from the aqueous phase of chloroform-extracted samples (Qiagen RNeasy Mini kit 74104). (3) Generate cDNA libraries by using 2 μg of total RNA and the SuperScript III first-strand synthesis system (Invitrogen 18080051). (4) Perform qPCR (15 μL each) using 2 μL of a 1:10 cDNA dilution for the endogenous gene(s) of interest or a 1:1000 dilution for the highly expressed genes (i.e., reference gene GAPDH, ACTB, etc., or the overexpressed fusion transgene). Follow the manufacturer’s recommendations for other reagents (primers, probes, polymerase, dNTPs, and buffer). (5) Calculate the mean quantification cycle (C_q, also described as C_t and C_p) for three replicate wells per unique reaction. Calculate the expression level with the equation ΔC_q = 2^{mean Cq reference−mean Cq target.} Calculate the log 2-fold change in gene expression as log 2(ΔC_{q transfected cells}/ΔC_{q mock}).

Chromatin Immunoprecipitation To Determine Fusion Protein Binding at Target Loci.

A critical step in validating the behavior of fusion proteins that interact with histone PTMs is to detect the interaction of the fusion protein with target loci in cells. For fusion transcriptional regulators, it is important to identify instances in which the histone PTM-binding fusion interacts with chromatin sites but does not influence gene expression. Furthermore, several HDB fusion proteins have been used to label histone PTM-enriched regions and are not intended to artificially regulate genes.^6,8,31 Crosslinked chromatin immunoprecipitation (X-ChIP), first reported in 1988,³³ has become a gold standard method for measuring the accumulation of proteins at specific genomic loci. In brief, cells are treated with formaldehyde to cross-link chromatin proteins with target DNA, the chromatin is sheared (sonicated or enzymatically digested) into soluble particles, particles containing the protein of interest (e.g., the fusion protein) are immunoprecipitated, and the co-immunoprecipitated DNA is purified and analyzed by quantitative PCR of specific sites (ChIP−qPCR) or by next-generation sequencing and computational alignment of all DNA fragments in the sample (ChIPseq). Fusion proteins expressed from transgenes are highly amenable to ChIP because they can be designed to include a reliable epitope tag, and uniform expression across all cells can be achieved in a stable transgenic cell line. In our previous work, we used ChIP-seq to profile the genomewide distribution of a Pcα fusion (called PcTF) and its target histone PTM H3K27me3 in a transgenic U2OS cell line.⁴ As an alternative to the resource-intensive ChIP-seq assay, we have used ChIP−qPCR to measure changes in Pc fusion and histone PTM levels over time near a specific PcTF-responsive gene.⁴ In cases in which the HBD fusion shows strong, specific localization within nuclei, immunofluorescence cytology with antibodies against specific histone PTMs can be used to assess the co-localization of HBD fusion proteins and their targets.³⁴

Limitations.

TXTL May Not Be Sufficient for Single-HBD Fusion Proteins.

In previous trials, we observed very little HRP signal over background for a TXTL-expressed Pc fusion called PcTF, which carried a single copy of Pcα (unpublished). However, using overexpressed and purified protein from transformed E. coli, we demonstrated specific binding of PcTF to H3K27me3 (vs H3Kme3 and H3K9me3) in an ELISA. We also observed H3K27me3 binding of PcTF at levels higher than that of the negative control (mCherry-VP64) using a histone peptide microspot array. These assays showed a weaker K_d^app value (~0.5-fold) for PcTF than for the double-HBD variant Pc₂TF [identical to Pcαα (Figure 1)]. Thus, the lower-affinity fusion PcTF functions in vitro, but the TXTL reaction as described here may not produce a sufficiently high yield for a detectable HRP signal after binding and washing during the ELISA.

Fusion Protein Dose and Cell Heterogeneity Are Difficult To Measure and Control in Transiently Transfected Cells.

The transiently transfected HEK293 cells that we generated for this study produced RFP signal levels that varied between expression plasmids and were near the lower end of the detection limits for the microwell plate reader that we used (Figure S4). Therefore, we recommend transient transfections only for preliminary assessment of expression levels and fusion regulator activity in a model system. For follow-up studies that require precise measurements of gene expression in response to the dose (expression level) of the fusion regulator or require a homogeneous population of transgenic cells for optimal results (e.g., ChIP-seq), we recommend generating stable cell lines. Inducible transgene expression systems prevent premature expression of the fusion, which may broadly regulate genes in the host cells prior to treatment and analysis. There are many available options that do not require viral transduction. For instance, we have used the popular T-REx system (Invitrogen) to express PcTF in U-2 OS cells.^4,5 We have also successfully used an all-in-one doxycycline-inducible expression cassette based on the sleeping beauty transposon-based system pSBtetGP³⁵ to express PcTF in MCF7 cells.

ABBREVIATIONS

PTM

post-translational modification

HBD

histone PTM-binding domain

PCD

polycomb domain

PHD

plant homeodomain

HRP

horseradish peroxidase

EED

embryonic ectoderm development

VP64

viral protein 64

BSA

bovine serum albumin

RFP

red fluorescent protein

ELISA

enzyme-linked immunosorbent assay

Pc

polycomb

TXTL

E. coli-based in vitro transcription and translation system

mCh

mCherry

BRD

bromodomain

CD

chromodomain

PCR

polymerase chain reaction