Pioneer factors and their in vitro identification methods

Abstract

Pioneer transcription factors are a special group of transcription factors that can interact with nucleosomal DNA and initiate regulatory events. Their binding to regulatory regions is the first event in gene activation and can occur in silent or heterochromatin regions. Several research groups have worked to define pioneer factors and studied their binding characteristics using various techniques. In this review, we will describe the in vitro methods used to define and characterize pioneer factors with particular attention to differences in methodology and how these differences can affect the results.

Eukaryotes such as humans are pluricellular organisms that are composed of multiple cell types, thus forming different tissues (Sapp, 2005). The human genome consists of more than 3 billion base pairs of DNAs packaged into 23 chromosomes, with approximately 30,000 protein-coding genes (Frazer, 2012; Pennisi, 2003). To establish competence for specific cell fates and guarantee the development of different cell lineages, cell lineage-specific spatial and temporal regulation of the expression of various genes is required (Pope & Medzhitov, 2018; Wittkopp, 2007). This cell type-specific regulation occurs largely at the transcriptional level, with transcription factor (TF) binding playing a dominant role. TFs regulate the transcription of genes via binding to specific regulatory DNA sequences, i.e., transcription factor binding sites (TFBS) (Spitz & Furlong, 2012). Typically, TFs recognize and bind to relatively short degenerative DNA sequences. Due to the degenerative nature of TFBS, they can appear millions of times in a genome. The majority of TFBS are never bound because most TFBS are inaccessible within chromatin.

Eukaryotic DNA is tightly packaged into a highly organized structure called chromatin. Chromatin is composed of nucleosomes that consist of nucleosome core particles of 147 bp double-stranded DNA that wraps 1.67 turns around a histone core containing 2 units of H2A, H2B, H3 and H4 (Jiang & Pugh, 2009; Kornberg & Lorch, 1999). Nucleosomes are then connected by a short stretch of DNA called a linker and to form a ‘beads-on-a-string’ structure (Olins & Olins, 2003). Chromatin can be further compacted by H1, which helps to stabilize the higher-order chromatin fiber through binding to the linker DNA (Ronald Berezney; R. Berezney, Mortillaro, Ma, Wei, & Samarabandu, 1995). Importantly, all these histone proteins are positively charged, allowing strong interaction with the negatively charged phosphate backbone of DNA. As a result, the DNA is wrapped tightly around the histone proteins, rendering the nucleosomal DNA inaccessible to binding by most TFs (Luger, Dechassa, & Tremethick, 2012; Vermaak, Ahmad, & Henikoff, 2003).

Reading inaccessible regulatory regions and creating accessible DNA for gene regulation is a prerequisite for most developmental processes. Pioneer factors are a special class of TFs that can bind to their target sequences even buried inside chromatin. Pioneer factors have been shown to be essential regulatory proteins during organismal development, are often misregulated or mutated in cancer, and are the key TFs in generating induced pluripotent cells. In this review, we focus mainly on the in vitro analysis methods being applied to study TF binding capabilities and identify pioneer factors. We believe that improvement and progress on these in vitro studies will finally contribute to understanding more complicated gene regulation mechanisms in vivo.

Different pioneer factors display various nucleosome binding patterns

The concept of pioneer factor first arose in 2002, when researchers found that FOXA and GATA4 can bind to nucleosome arrays in vitro, while other TFs such as NF1, C/EBP, and GAL4-AH cannot (Cirillo et al., 2002). They proposed that these nucleosome-binding-capable TFs might represent a new functional class of regulatory proteins and named them ‘pioneer’ factors, indicating that their binding occurs prior to other events. To date, the well-characterized pioneer factors include FOXA proteins (Cirillo et al., 2002; Cirillo et al., 1998), Yamanaka factors such as SOX2, KLF4 and OCT4 (Soufi, Donahue, & Zaret, 2012), and some other TFs such as ASCL1, PAX7, PU.1, GATA 4 and P53 (Barozzi et al., 2014; Budry et al., 2012; Cirillo et al., 2002; Iwafuchi-Doi & Zaret, 2016; Laptenko, Beckerman, Freulich, & Prives, 2011; Lidor Nili et al., 2010; Soufi et al., 2015). Although these pioneer factors can overcome the barrier of nucleosomes and bind to nucleosomal DNA, they display various binding patterns.

The canonical pioneer factor FOXA (Cirillo et al., 2002; Cirillo et al., 1998) can bind the nucleosome dyad and displace the linker histone to maintain nucleosome accessibility (Iwafuchi-Doi et al., 2016). SOX2 and OCT4 can bind to sites in heterochromatin regions (Soufi et al., 2012), but heterochromatin does impede their binding to some extent (Onder et al., 2012). p53 prefers binding sites located near the nucleosome edge rather than the center (Laptenko et al., 2011; Yu & Buck, 2019), and it binds target sequences on chromatin with higher affinity than sites located in naked DNA (Espinosa & Emerson, 2001). Another noteworthy point is that even for canonical pioneer factors such as SOX2 and OCT4, they can bind to only approximately 4% of their total available recognition sizes in the whole genome (Soufi et al., 2012). Overall, as additional pioneer factors are characterized, it is becoming clear that they are diverse in terms of nucleosome interaction and histone modification dependency.

Recent studies have started to uncover inconsistencies in pioneer factor binding specificity to nucleosomes. Early investigations of FOXA1 discovered that FOXA protein can interact with target sequences located near the nucleosome dyad (Cirillo et al., 1998; McPherson, Shim, Friedman, & Zaret, 1993). In contrast, a recent approach utilizing a new high-throughput approach, NCAP-SELEX, found that FOXA protein cannot bind directly to the nucleosome dyad (Zhu et al., 2018). Moreover, the results indicated that nucleosomal DNA length will also affect TF binding to nucleosomes, which might be a caveat regarding in vitro nucleosome binding assays (Zhu et al., 2018). Studies on p53 protein also show divergence. An in vitro study indicated that the rotational positioning of the binding site affects p53 binding, and p53 can bind to target sequences located near the nucleosome dyad when located properly (Sahu et al., 2010). Meanwhile, our previous recent study using both in vitro and in vivo approaches demonstrated that p53 displays much higher binding affinity to nucleosome edges than the dyad (Yu & Buck, 2019). The difference between the findings of these studies is likely due to differences in key aspects of the experimental approach.

Here, we want to compare the methods that have been applied to test TF-nucleosome binding in vitro to determine whether any differences in the experimental design or conditions utilized might in turn affect the result. The techniques or materials applied in each experiment might differ from laboratory to laboratory, which might be why the results of pioneer factor-nucleosome in vitro binding assays vary among researchers. The important elements that should be taken into consideration include 1) the DNA template utilized as the nucleosome positioning sequence; 2) histone variants and modifications; 3) the composition of the binding buffer; 3) the final nucleosome concentration in the reaction system; and 4) the method applied to detect the binding complex. Below, we will discuss each of these elements in detail.

1. Nucleosomal DNA

The nucleosome will form on a piece of double-stranded DNA in a characteristic manner depending on the sequence (Kaplan et al., 2009; Yuan et al., 2005). Some specific sequences will consistently position the nucleosome in the same manner and thus have strong nucleosome positioning capability (Lowary & Widom, 1998). Other sequences of DNA can have varied nucleosome positions, causing the population of nucleosomes on that DNA to be located in different places. Currently, several in vitro nucleosome positioning sequences are in common use, from natural sequences such as 5S rDNA (Simpson & Stafford, 1983), MMTV LTR (Richard-Foy & Hager, 1987), alpha satellite sequences (Yang, Hansen, Oishi, Ryder, & Hamkalo, 1982) and sequences derived from in vivo regions that have been found to be occupied by nucleosomes (Darvekar, Johnsen, Eriksen, Johansen, & Sjottem, 2012; Soufi et al., 2015) to artificial synthetic sequences such as Widom 601 DNA (Lowary & Widom, 1998), TG repeated units (Godde & Wolffe, 1996; Shrader & Crothers, 1989) and randomized DNA sequences (Zhu et al., 2018). Based on the goals of studying nucleosome behavior, different positioning sequences will be selected. Unlike other known nucleosome positioning templates that usually form labile nucleosomes that slide readily, Widom 601 DNA has the strongest nucleosome positioning ability and extremely high thermodynamic stability (as high as 55 degrees) (Senavirathne, Mahto, Hanne, O’Brian, & Fishel, 2017). Widom 601 DNA’s structural stability means it is not an ideal template to study nucleosome dynamics, but it can generate a homogenous population of nucleosomes and thus serves as an ideal model to test how TFBS location on a nucleosome can affect TF binding. Different DNA sequences being applied to assemble nucleosomes in vitro are summarized in Table 1. Alternatively, the micrococcus nuclease (MNase) enzyme can be used to obtain naturally existing nucleosomes from cells (Darvekar et al., 2012).

Table. 1.

DNA templates applied to reconstitute nucleosomes in vitro

DNA template	Sequence information	Sequence composition	Publication
In vivo fragments	Derived from genomic sequences	LIN28B	1. (Soufi et al., 2015)
		5S rRNA	2. (Angelov et al., 2004)
			3. (Howe & Ausio, 1998)
			4. (Steger & Workman, 1997)
		Albumin enhancer	5. (Cirillo et al., 1998)
		Chromatins from HeLa cells	6. (Darvekar et al., 2012)
	Derived from virus	HIV-1 5′ LTR	7. (Angelov et al., 2000)
			4. (Steger & Workman, 1997)
		MTV LTR	8. (Perlmann & Wrange, 1988)
Synthetic fragments	Containing repeated or phased units	TG motif	9. (H. Chen, Li, & Workman, 1994)
			10. (Li & Wrange, 1993)
			11. (Li & Wrange, 1995)
		TTAGGG repeat	12. (Galati et al., 2006)
		A-tracts	13. (Sahu et al., 2010)
	Artificial nucleosome positioning sequence	pBend vector	14. (Vettese-Dadey, Walter, Chen, Juan, & Workman, 1994)
		Widom 601 DNA	15. (Yu & Buck, 2019)
		Randomized DNA sequences	16. (Zhu et al., 2018)

2. Histone variants and modifications

The nucleosome template DNA is mixed with histone proteins at an optimized ratio, and nucleosomes are formed by salt gradient dialysis (Dyer et al., 2004; Lee & Narlikar, 2001), followed by sucrose gradient centrifuge or glycerol gradient centrifuge to pick purified nucleosomes from unbound DNA and histones (Lusser & Kadonaga, 2004; Noll & Noll, 1989; Schnitzler, 2001). In addition to commonly used histone proteins such as H2A, H2B, H3, and H4, other histone variants can also be applied to assemble nucleosomes in vitro. For example, H2A.Z and H3.3, which are variants that occupy approximately 10% of total H2A or H3 proteins in vivo (Pina & Suau, 1987), have been utilized to perform nucleosome reconstitution (Osakabe et al., 2018; Thakar et al., 2009). However, nucleosomes containing a combination of certain histone variants have been tested and found to be particularly unstable (Jin et al., 2009; Zlatanova & Thakar, 2008) and thus unsuitable for obtaining a homogeneous nucleosome population in vitro. For the purpose of examining histone posttranslational modifications, histone proteins with modified tails were also generated by applying chemically modified histone peptides (Chatterjee & Muir, 2010; Yun, Ruan, Huh, & Li, 2012).

3. Binding buffer

To mimic the biological environment where protein-DNA binding events occur in vivo, researchers utilized buffer systems such as Tris-HCl or HEPES to maintain a relatively stable physiological pH in the reaction mix (Durst & Staples, 1972; Vega & Bates, 1976). Salts are also included in the solution, but usually at low concentrations. As shown in Table 2, researchers usually included 50 to 100 mM sodium chloride or potassium chloride in the binding buffer to maintain sufficient ionic strength in solution to prevent nonspecific binding without breaking the specific interactions between proteins of interest and phosphate groups on the DNA backbone. In addition to monovalent cations, divalent salts were also added, such as Mg²⁺ and Zn²⁺, which have been shown to stabilize the DNA duplex structure (Chiu & Dickerson, 2000; Nakano, Fujimoto, Hara, & Sugimoto, 1999; Owczarzy, Moreira, You, Behlke, & Walder, 2008; Tan & Chen, 2006). On the other hand, salts in an appropriate concentration range can also enhance protein solubility and prevent protein aggregation (Bondos & Bicknell, 2003). Reducing agents, namely, DTT or 2-mercaptoethanol, are also added to mitigate oxidation damage (Lebendiker & Danieli, 2014; Wingfield, 2001). Moreover, additives such as BSA, glycerol, and nonionic detergents NP-40 can also be added to promote protein solubility and enhance protein-DNA stability (Lebendiker & Danieli, 2014). Protease inhibitors such as aprotinin, leupeptin, PMSF, or EDTA can be added when using cellular protein extracts. Some research groups also add poly(dI.dC), salmon sperm DNA or BSA to further reduce the nonspecific binding (Larouche, Bergeron, Leclerc, & Guerin, 1996). The table below summarizes the binding buffers utilized to test DNA-protein interactions (Table 2).

Table. 2.

Experimental conditions of various TF-nucleosome binding assays (Part I)

Reaction	Buffer (pH 7.0-8.0)		Monovalent salt (mM)				Divalent salt		Stabilization
Publication	Tris-HCl	HEPES	NaCl	KCl	Na2EDTA	K2HPO4	MgCl2 or MgSO4	ZnCl2 or ZnSO4	Glycerol	NP-40
1. (Soufi et al., 2015)	10 mM			10			1 mM	10 μM	5%
2. (H. Chen et al., 1994)	10 mM	10 mM	100	100				10 μM		0.10%
3. (Bowers, Calero-Nieto, Valeaux, Fernandez-Fuentes, & Cockerill, 2010)		20 mM	50	80			3 mM		10%
4. (Darvekar et al., 2012)		10 mM	100						5%	0.030%
5. (Angelov et al., 2000)		10 mM		50					5%
6. (Angelov et al., 2004)		20 mM	50						5%
7. (Cirillo et al., 1998)	10 mM			40			1 mM		0.50%
8. (Li & Wrange, 1993)	20 mM		50		1				10%
9. (Vettese-Dadey et al., 1994)	10 mM		100						20%	0.10%
10. (Perlmann & Wrange, 1988)	20 mM		50		1				10%
11. (Li & Wrange, 1995)	20 mM		50		1				10%
12. (Howe & Ausio, 1998)	20 mM		70					10 μM	6%	0.07%
13. (Sahu et al., 2010)	10 mM		50				2 mM		5%	0.025%
14. (Steger & Workman, 1997)		10 mM		50					5%
15. (Galati et al., 2006)		20 mM	100	50			1 mM		5%	0.10%
16. (Yu & Buck, 2019)	10 mM		50				2 mM		5%	0.0250%
17. (Zhu et al., 2018)		20 mM	5	50-140		1	2 mM	3 μM

4. Nucleosome concentration

The nucleosome concentration in the binding assay is another important element that affects binding and nucleosome stability. Previously, many studies utilized radiolabeled probes to detect supershifted TF-nucleosome complexes. Radioactive techniques have high sensitivity, allowing small quantities of nucleosomes to be used in the assays. For example, Angelov’s group applied as little as 0.5 nM nucleosomes to bind NF-κB (Angelov et al., 2004), and another group used nucleosomes at the picomolar level to interact with GR proteins (Li & Wrange, 1993, 1995). However, nucleosomes tend to partially disassemble at subnanomolar concentrations, and subnucleosomal structures such as hexasomes or tetrasomes appear (Y. Chen et al., 2017; Lai & Pugh, 2017; Rhee, Bataille, Zhang, & Pugh, 2014). These subtypes can also be obtained by adjusting the final salt concentration (Arimura, Tachiwana, Oda, Sato, & Kurumizaka, 2012; Krajewski, Li, & Dou, 2018).

A few studies have demonstrated the occurrence of subnucleosomal structures at low nucleosome concentrations (Claudet, Angelov, Bouvet, Dimitrov, & Bednar, 2005; Kelbauskas, Sun, Woodbury, & Lohr, 2008). The researchers observed the binding of NF-κB to its recognition site located near the nucleosome dyad at a nucleosome concentration of only ~7.5 nM, but they failed to detect binding at 40 nM nucleosome (Lone et al., 2013). They obtained partially disassembled subnucleosomal particles by diluting nucleosomes to approximately 10 nM. Under these conditions, H2A/H2B dimers were eliminated, and the researchers believed that this substructure decreased the hindrance to NF-κB access, thus allowing interaction to occur. Taking all these results into consideration suggests 10 nM as a threshold concentration for in vitro reconstituted nucleosome core particles to remain intact and relatively stable (Claudet et al., 2005; Kelbauskas et al., 2008; Lone et al., 2013).

5. Quantifying TF-nucleosome binding

The electrophoretic mobility shift assay (EMSA), a traditional technique that observes and quantifies protein-DNA interactions, can also be applied to detect TF-nucleosome binding in vitro (Dey et al., 2012). It is based on the principle that protein-DNA complexes are larger and thus run more slowly than unbound free DNA, allowing easy separation into at least two different-sized bands on the gel. Researchers have been using either agarose gel or nondenaturing polyacrylamide gel to visualize the TF-nucleosome complexes. Polyacrylamide gel is a better band-separating system than agarose gel because it has greater resolving power, presents sharper bands, and generates very pure recovered DNA (Guilliatt, 2002; Tietz, 1998). On the other hand, the nucleosome itself is already a DNA-protein complex, and when TFs bind, even more complicated protein-DNA-protein supercomplexes will be generated. The “cage effect” of the polyacrylamide gel contributes to stabilizing this large complex, thus efficiently preventing dissociation as the supercomplex moves through the gel (Nelson, Hendy, Reid, & Cavanagh, 2002; Vossen & Fried, 1997). This technique has been further enhanced to determine the affinity of different proteins to their corresponding target by applying radiolabeled or biotin-labeled probes. By measuring the signal emitted from each band, which corresponds to the concentration of probes being shifted away, researchers can quantify the templates bound by the tested protein, calculate the k_D values, and finally deduce the binding affinity.

One major limitation of EMSA is that it can test the TF binding to only one nucleosome type at a time. To test thousands of different nucleosomal DNA templates, researchers must perform thousands of binding assays and run thousands of gels, which is time consuming and not feasible. Recently, a few laboratories have used high-throughput sequencing to investigate TF binding to large nucleosome libraries simultaneously (Yu & Buck, 2019; Zhu et al., 2018). Our study tested p53 protein binding to multiple nucleosome types simultaneously through just one binding reaction and then utilized next generating sequencing (NGS) to obtain the binding frequency of each nucleosome type (Yu & Buck, 2019). Taipale’s group filtered randomized DNA sequences and retained the ones that were favorable for nucleosome formation and then tested the binding of 220 TFs to those random nucleosomes individually and performed computational analysis to discover motifs on those nucleosomal DNAs to define where the TFBS were located on each nucleosome (Zhu et al., 2018).

Future questions for studying pioneer factors

The overall goal of investigating pioneer TF-nucleosome binding in vitro is to address the following question: what are the rules that dictate pioneer factor binding? In general, several elements could affect pioneer factor binding to the nucleosome, including the following: 1) TFBS positioning. Every single nucleotide has a specific location within a nucleosome known as its translational position (Simpson, 1991), and the rotational position describes the orientation of each nucleotide relative to the histone core (Lowary & Widom, 1997; Simpson, 1991). 2) Histone modification on the histone core. Certain modifications that occur on histone tails, such as histone acetylation, can loosen histone-DNA contacts, destabilize nucleosomes, and finally facilitate TF binding (Fenley, Anandakrishnan, Kidane, & Onufriev, 2018; Henikoff, 2008). 3) Presence of cofactors or chromatin remodelers. The interactions between TFs and cofactors are important for transcription regulation (Goi, Little, & Xie, 2013; Siggers, Duyzend, Reddy, Khan, & Bulyk, 2011), while chromatin remodelers regulate chromatin accessibility and affect nucleosome positioning (Langst & Manelyte, 2015; Tyagi, Imam, Verma, & Patel, 2016).

During the past decades, the protein-DNA binding assay has been applied by different labs to test proteins of interest. This traditional technique has been further revised to test pioneer factor-nucleosome binding. However, the experimental conditions used in different labs are slightly different. As noted above, nucleosomes should be maintained above 10 nM in solution to prevent partial dissociation. It is also recommended to use a strong nucleosome positioning sequence to obtain a homogenous nucleosome population when studying the effect of TFBS location on the nucleosome. Taking all these points into consideration, a standard criterion for in vitro TF-nucleosome binding assays might be required in the future. Overall, we believe that these in vitro studies will enhance the understanding of pioneer factors and pave the road for future investigation.

Table. 2.

Experimental conditions of various TF-nucleosome binding assays (Part II)

Reaction	Reagents to reduce nonspecific binding					Protease inhibitors				Reducing agents
Publication	Ficoll	BSA (mg/ml)	Poly(dI.dC)	Pork insulin	Spermidine	EDTA or EGTA (pH 8.0)	PMSF (mM)	Aprotinin	Leupeptin	2-mercaptoethanol	DTT
1. (Soufi et al., 2015)		0.5									1 mM
2. (H. Chen et al., 1994)		1				0.4 mM	0.1			10 mM
3. (Bowers et al., 2010)		0.1	6.25 ng/μl				0.1	5 μg/ml	5 μg/ml		1%
4. (Darvekar et al., 2012)		0.15									1 mM
5. (Angelov et al., 2000)		0.2					0.5				5 mM
6. (Angelov et al., 2004)		100
7. (Cirillo et al., 1998)	1%	3									5 mM
8. (Li & Wrange, 1993)				0.1 mg/ml							5 mM
9. (Vettese-Dadey et al., 1994)		0.1				1 mM				1 mM
10. (Perlmann & Wrange, 1988)				0.1 mg/ml							5 mM
11. (Li & Wrange, 1995)				0.1 mg/ml							5 mM
12. (Howe & Ausio, 1998)		0.1	40 ng/μl								2.5 mM
13. (Sahu et al., 2010)					2 mM	1 mM					1 mM
14. (Steger & Workman, 1997)		0.25					0.5				5 mM
15. (Galati et al., 2006)		0.5				0.1 mM					1 mM
16. (Yu & Buck, 2019)		0.25									1 mM
17. (Zhu et al., 2018)						100 μM