Purification & Characterization of Transcription Factors

Abstract

Transcription factors (TFs) are essential for the expression of all proteins, including those involved in human health and disease. However, TFs are resistant to proteomic characterization because they are frequently masked by more abundant proteins due to the limited dynamic range of capillary liquid chromatography-tandem mass spectrometry and protein database searching. Purification methods, particularly strategies that exploit the high affinity of TFs for DNA response elements on gene promoters, can enrich TFs prior to proteomic analysis to improve dynamic range and penetrance of the TF proteome. For example, trapping of TF complexes specific for particular response elements has been achieved by recovering the element DNA-protein complex on solid supports. Additional methods for improving dynamic range include two- and three-dimensional gel electrophoresis incorporating electrophoretic mobility shift assays and Southwestern blotting for detection. Here we review methods for TF purification and characterization. We fully expect that future investigations will apply these and other methods to illuminate this important but challenging proteome.

Nature of the TF Proteome

The transcription factor (TF) proteome is an attractive target for future studies to understand human health and disease. The Central Dogma of Biology is that genes, protein-coding stretches of DNA, are transcribed from DNA to RNA and translated from RNA to protein. Transcription is tightly regulated by a variety of mechanisms at the DNA, RNA, and protein levels to determine TF abundance and post-translational modifications (PTMs). TFs are sequence-specific DNA-binding proteins and represent 3–6% of the human genome (Venter et al., 2001), which act as transcriptional activators and repressors. To regulate transcription, TFs bind to specific DNA sequences, typically 6–12 base pairs (bp), called the response element (RE). Typically, parts of the TF protein insert into the major groove of the DNA double helix to recognize the RE sequence, though minor groove interactions are also known. Less than 100 of the 1500 genes encoding TFs in humans have been purified and characterized (Gadgil et al., 2001). The limited dynamic range of proteomic analysis with capillary LC/MS/MS and protein database searching, combined with the masking of low abundance TFs by more abundant proteins, is a major technological barrier to our quest for knowledge of the human proteome (Rusconi et al., 2002). Consequently, much of our current knowledge of TFs has been gleaned from genetic studies in genetically engineered model organisms, such as yeast, which allow n-hybrid interaction assays and facile gene knockout experiments. Many of these studies are not possible or ethical in mammals. Moreover, it has become clear that a detailed understanding of TFs must include PTMs, such as phosphorylation, due to the combinatorial nature of these interactions with DNA, RNA and other proteins. For example, multiple phosphorylation sites have been discovered in the c-Jun subunit of the activator protein 1 (AP1) TF (Behrens et al., 2000).

Here and elsewhere (Jarrett, 2012, Jiang et al., 2009) we review recent advances in the purification and characterization of TFs, with particular emphasis on purification strategies that exploit the high affinity of TFs for DNA response elements on gene promoters to enrich for low abundance TFs prior to proteomic analysis with capillary LC/MS/MS and protein database searching. We also present an introduction to TF families.

TF Families

TFs are grouped into families based on homologous protein domains that mediate binding to DNA or other proteins. In many cases, homo- or hetero-dimerization is required to control transcriptional activation or repression. Some of the major families of TFs are described below although the discussion is not comprehensive.

The Basic Helix-Loop-Helix and Basic Leucine Zipper Families

The basic helix-loop-helix (bHLH) TF family is highly conserved from yeast to humans, and it plays a pivotal role in developmental events such as sex determination, lineage commitment, and cell differentiation. This family of proteins contains an evolutionary conserved basic region that promotes binding to the deoxyribose phosphate backbone of DNA in addition to the ability to directly interact with other bHLH TFs (Murre et al., 1989). One important group in the bHLH family contains a basic alpha helix that binds the E-Box RE sequence (CANNTG) and a hydrophobic HLH that allows these proteins to form homo- or heterodimers. Members of this family include Myoblast determination protein 1 (Davis et al., 1987), c-Myc (Bishop, 1982), and Hypoxia inducible factor 1 alpha (Wang & Semenza, 1995). The bHLH proteins can also contain a leucine zipper domain of amphipathic alpha helices that homo- or hetero-dimerize to form coiled-coils to enable the TF to interact with the major groove of DNA in a sequence-specific manner. The phrase ‘leucine zipper’ derives from the occurrence of a leucine at every seventh amino acid residue position to place leucine residues on roughly the same side of the α-helix where they interact with another leucine zipper during dimerization by hydrophobic interactions (Vinson et al., 1989). Activator protein 1 is a member of the bHLH family, and exists as a heterodimeric complex with members of c-Jun, c-Fos, ATF, and other protein families. This family of TFs regulates gene transcription in response to cytokines, growth factors, stress signals, and infection (Hess et al., 2004).

The related basic leucine zipper (bZIP) family of TFs are the largest family in eukaryotic cells and also contain the simplest known protein-DNA recognition motif (Fujii et al., 2000). These proteins contain a basic region and a leucine zipper region but not a bHLH motif. The bZIP basic region is a highly conserved region spanning fourteen to sixteen amino acids and mediates sequence-specific DNA binding. The leucine zipper is characterized by the several leucine residues, which are necessary for dimerization by forming a chopstick like structure around the major groove of palindromic sequences (Fujii & Shimizu & Toda & Yanagida & Hakoshima, 2000). For example, C/EBP is a bZIP protein which is made of a single, long alpha-helix. One region of that helix inserts into the major groove of DNA, and nearby is a basic region which binds the deoxyribose-phosphate DNA backbone. Elsewhere this long helix has several heptad leucine repeats that allow it to dimerize. In C/EBP, these regions are in the order discussed from N- to C-terminus. This family of TFs regulates cell development, differentiation, as well as environmental stresses (Hu et al., 2002).

The Winged-Helix Family

The winged-helix family is a subfamily of the helix-turn-helix (HTH) family of TFs. This group is classified by three β-strands, three α-helices and two “wings” (Gajiwala & Burley, 2000). The wings are found at the C-terminus along with three strands forming a twisted antiparallel β-sheet whereas the N-terminus is mostly helical. The forkhead-box (FOX) members of this family are characterized by their winged-helix DNA-binding domain of 80–100 amino acid residues (Myatt & Lam, 2007). However, they differ in other domains involved in transactivation and transrepression. Although the FOX members do have some differences from the original winged-helix family, this nomenclature is still accepted in the scientific community (Kaestner et al., 2000). Members of this family of TFs modulate a variety of cellular functions such as cell cycle progression, differentiation, DNA damage repair, and programmed cell death (Accili & Arden, 2004).

The Homeobox Family

The homeobox family is another subfamily of the helix-turn-helix (HTH) family of TFs. The homeobox is a 180 bp DNA sequence that encodes a 60 amino acid residue protein domain referred to as the homeodomain. The most famous examples in this family are the well-studied Hox TFs that govern development, including development of larval segments into specific body parts in Drosophila (Lewis, 1978). In mammals, Hox genes have redundant functions that regulate anterior-posterior axis formation and are grouped into four clusters, based on their chromosomal location. A recent proteomic investigation identified the Paired related homeobox protein as a regulator of another TF, known as Osterix, during bone differentiation (Lu et al., 2011).

The Zinc Finger Family

Zinc finger TFs facilitate the binding of zinc by coordination to cysteine and histidine residues. These residues form the base of the ‘finger’. that will recognize a specific DNA sequence, the RE. The first zinc finger TF, protein TFIIIA, was identified in Xenopus oocytes (Miller et al., 1985). The zinc finger domain itself encompasses an independently folding region of the protein that utilizes one or more zinc ions to stabilize packing of the antiparallel β-sheet with the α-helix finger (Berg, 1988, Lee et al., 1989, Miller & McLachlan & Klug, 1985).

Transcriptional Complexes and Transcription Activation

For protein-encoding genes, transcription is carried out by RNA polymerase II (Pol II) and we restrict our discussion to only Pol II promoters. The promoter region of any gene contains multiple REs and bind TFs of at least two recognizable types, here called the specific and general TFs. The specific TFs bind to REs with very high affinity (typically ≤ nM) that include Specificity protein 1, AP1, CAAT enhancer binding protein (C/EBP) and many others, which have already been discussed. In current models of transcription, this binding occurs early and recruits the general transcription machinery. The general transcription complex involves many individual proteins and subunits. Also contained within these promoters are other DNA sequences that bind components within the general transcription complex. Promoters typically contain core promoter sequences, including: the ATATAA consensus sequence bound by the TATA-binding protein subunit within the general transcription complex, the TFIIB recognition element, the downstream promoter element, and the initiation start site where mRNA transcription begins. The pre-initiation complex assembly starts with binding at the TATA box. In some species, assembly also involves a novel conserved DNA element such as the GA element for Saccharomyces cerevisiae (Seizl et al., 2011). Current models involve binding of specific TFs to TFIID to assemble over a DNA region 40–60 bp upstream and downstream of the RNA transcriptional initiator element (Chen & Hahn, 2004, Rani et al., 2004). This assembly ultimately includes Pol II and the mediator complex properly positioned at the transcriptional start site, and it is referred to as the pre-initiation complex (PIC). The TFII complex depends upon relatively weak interactions (lower than nM affinity) among core promoter elements, the TFII protein components, the specific TFs, and Pol II. Altogether, there are at least 60 proteins and subunits involved in the formation of the PIC. The convergence of these proteins leads to hyperphosphorylation within the C-terminus of the largest subunit of pol II by protein kinases incorporated in the PIC. Subsequent partial dissociation of mediators in PIC generates what is known as an ‘open-complex’. The open complex allows rNTPs to bind RNApolII and catalyze the formation of phosphodiester bonds to link rNTPs together as linear chain, driving transcription at the transcriptional initiation (or start) site. At this point, most of the TFII complex remains at the promoter, is referred to as the ‘scaffold complex’, and recruits more Pol II, TFIIB and TFIIF to re-initiate transcription (Spangler et al., 2001). A recent review on Pol II machinery highlights the importance of this complex (Chen & Hahn, 2004).

TF Purification

There are several methods available to enrich for low abundance TFs, including TFs present at less than 10⁵ copies per cell. Because all purifications rely on a specific assay for the protein of interest, the assays used must be first discussed. The most popular method to determine whether a TF binds to particular DNA sequence is the electrophoretic mobility assay (EMSA), also called a gel shift assay. This procedure uses either radio- or fluorescence-labeled DNA, incubated with pure or crude (nuclear) extract prior to resolution with native (non-denaturing), polyacrylamide gel electrophoresis (PAGE), to detect TF-DNA complexes. If a TF is present and binds the labeled DNA, then the corresponding TF-DNA complex will be detected from its slower migration on the gel (i.e., a supershift) (Fried & Crothers, 1981, Lane et al., 1992).

The labeled DNA can also be used for detection with Southwestern blotting (SWB), where an electrophoresis gel is blotted and TFs partially renature and bind to labeled DNA (Jiang et al., 2009). TFs are the only nuclear proteins that have ever been reported to bind DNA with nM or less affinity. Therefore, the concentration and sequence of DNA used in the incubation and washing steps of SWB can be tailored to the affinity of target TFs. For example, a low concentration of a DNA oligonucleotide with the sequence CAAT specifically targets C/EBP (Jiang & Jia & Zhou & Jarrett, 2009). SWB is demonstrated in Figure 2C where an electro-blotted PVDF membrane was blocked with 5% milk to reduce non-specific binding and then probed with radio-labeled C/EBP promoter DNA. A single spot for purified C/EBP demonstrated the specificity of the method for this example. However, SWB has limitations due to the potential inability of a protein to fully renature, and other limitations related to oligomerization of TFs such as the inability to detect TFs that bind as hetero-/homo-dimers. Figure 1C was treated in the same manner as Figure 2C however it was probed with radio-labeled human telomerase reverse transcriptase (hTERT) promoter DNA and later visualized with autoradiography. Each of the spots represents different hTERT-binding TFs.

Figure 2. TF purification with electrophoretic mobility shift assay-three dimensional electrophoresis (EMSA-3-DE).

A) In this example, the CAAT enhancer binding protein (C/EBP) promoter was ³²P-labeled at both the 5’ and 3’ ends. Once the DNA-TFs complexes were formed, TF purification was performed on a 3% EMSA gel by separating the Complex from the free DNA. Unbound proteins are not visualized since the radiolabel is only on the oligonucleotide. Characterization of TFs in DNA-TF complexes revealed by their higher molecular mass (M_r) on the EMSA gel (top) was performed with a variety of methods (B-D). B) TF characterization with two-dimensional gel electrophoresis (2-DE). 2DE-SDS-PAGE was performed as previously described in promoter trapping. Silver stained spots indicate TF components of DNA-TF complexes and other proteins that that were purified with EMSA. The single spot that was visualized was in excellent agreement with the predicted pI of 8.97 and M_r of 42,569 for C/EBP however, further characterization was necessary to confirm (shown in figures 2C and 2D). C) TF characterization with two-dimensional Southwestern blotting (2-D SWB). The 2-DE gel shown in B) was electroblotted onto a PVDF membrane, blocked with 5% milk, and hybridized with the radiolabeled C/EBP promoter. The single spot indicates C/EBP that was purified with EMSA. D) TF characterization (identification) with capillary liquid chromatography-tandem mass spectrometry (LC/MS/MS) and protein database searching in a manner similar that described in Figure 1. C/EBP was identified, and the amino acid sequence assignment of product ions for the top-ranked tryptic peptide from C/EBP, GSHMASMTGGQQMGR, spanning amino acid residues 18–32, and a representative MS/MS spectrum are shown.

Figure 1. TF purification with trapping combined with a variety of TF characterization methods.

A). In this example, the full hTERT promoter containing single stranded (GT)₅ tails at both the 5’ and 3’ ends was used to purify TFs. DNA-TF complexes are purified by annealing (GT)₅ tails to a (CA)₅-Sepharose column. The column was washed with 4 mL of low salt buffer (1X binding buffer), and DNA-TF complexes were eluted with 1 mL of high salt buffer (1X TE 0.5M NaCl pH 7.5). Characterization of TFs in these DNA-TF complexes was performed with a variety of methods (1B-1E). B) TF characterization with two-dimensional gel electrophoresis (2-DE). Isoelectric focusing was performed for the first dimension followed by separating the proteins by their relative molecular mass (M_r) with the use of a 12% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) Silver stained spots indicate TF components of DNA-TF complexes and other proteins that that were purified with promoter trapping. C) TF characterization with two-dimensional Southwestern blotting (2-D SWB). The 2-DE gel shown in B) was electroblotted onto a PVDF membrane and hybridized with the radiolabeled hTERT promoter. Spots indicate TF components of the DNA-TF complexes that were purified with promoter trapping. D) TF characterization (identification) with capillary liquid chromatography-tandem mass spectrometry (LC/MS/MS) and protein database searching¹. Two separate spots were excised from the gel and in-gel trypsin digestion was performed. 1 µg of extracted peptide was injected onto a C18 column and characterized with a hybrid linear ion trap-Fourier-transform tandem mass spectrometer. MS/MS spectra were generated via high-energy C-trap dissociation (HCD). HMGB1 (upper spectra) was identified with a Mascot protein score of 99 and sequence coverage of 37%. The amino acid sequence assignment of product ions for the top-ranked tryptic peptide from HMGB1, IKGEHPGLSIGDVAK, spanning amino acid residues 113–127, yielded an expectation value of 0.00118. The observed monoisotopic mass of the unmodified, doubly-charged precursor ion for IKGEHPGLSIGDVAK at 760.9258 m/z was 1519.8371 Da. Transcription factor CP2 (TFCP2) (IPI00037599) was also identified (lower spectra) with an expectation value of the top-ranked tryptic peptide (LHDETLTYLNQGQSYEIR) was 0.0000034 (Cunningham & Vanin & Tran & Valentine & Jane, 1995, Lim & Swendeman & Sheffery, 1992, Ramamurthy & Barbour & Tuckfield & Clouston & Topham & Cunningham & Jane, 2001). E) TF characterization with one-dimensional Western blotting (1-D WB). A 1D-SDS-PAGE gel was electroblotted onto a PVDF membrane followed by probing with specific antibodies (anti-TBP, anti-PolII, anti-USF2, anti-Sp1 and anti-beta(b)-actin). The different lanes of the SDS-PAGE gel were loaded with whole cell lysate (WCL); nuclear extract (NE); hTERT promoter trapping eluent (hTERT PT Eluate) and flow-through (hTERT PT FT) from 1A); c-Jun promoter trapping eluent (c-Jun PT Eluate) and flow-through (c-Jun PT FT) were from a previous experiment. Bands indicate components of DNA-TF complexes.

¹ Unpublished data were collected in the RCMI Proteomics & Protein Biomarkers Cores at UTSA.

The presence of particular TFs can be determined with one- or two-dimensional Western blotting (1D-WB or 2D-WB) with many commercially available antibodies specific for TFs present in the PIC (specific TFs, general transcription machinery and Pol II subunits). Major limitations arise from the limited availability of antibodies specific for each TF of interest. The recovery of TFs with various purification strategies can be assessed using protein stains (Coomassie, silver, and others). An example of 1D-WB is shown in Figure 1E. First, sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed to resolve proteins on a 12% gel by M_r (gel not shown). Second, the SDS-PAGE gel was electro-blotted onto a PVDF membrane, blocked with 5% milk to reduce non-specific binding, and probed with specific antibodies. The success of the purification experiment was verified by the use of positive and negative controls loaded into different lanes of the SDS-PAGE gel. Bands indicated TF components of DNA-TF complexes that were purified with promoter trapping. Absent or dim bands observed with anti-TBP-antibodies for WCL and NE, compared to the dark bands for hTERT PT Elute and c-jun PT-Elute, provided strong evidence for TF purification with promoter trapping (described below).

DNA Affinity Chromatography

In one version of this method, duplex DNA that contains several identical REs (a concatemer) is prepared and covalently immobilized to cyanogen-bromide activated Sepharose beads (Sepharose 4B or CL-2B). TF purification is carried out in the presence of a competitor DNA in the mobile phase, such as poly dI:dC (Kadonaga, 1991), to increase the specificity of binding to the concatemer on the stationary phase. Partially purified or crude nuclear extract passes through the column while TFs bind to the stationary phase. Gradient elution of TFs is performed by disrupting DNA-TF interactions with increasing concentrations of salt (e.g., NaCl). DNA binding activity of the protein fractions is monitored with EMSA. A major obstacle of DNA-affinity chromatography is the co-purification of non-specific proteins that bind to either the TF of interest or its respective RE. One study showed that a concatemer is no more effective than a single copy of a RE (Gadgil et al., 2001). To address non-specific protein binding, the sample can be pre-cleared with a non-specific DNA column made by attaching fragmented salmon or herring sperm genomic DNA or mutated RE DNA. Pre-cleared samples have been successfully used for the purification of C/EBP, the lac repressor, and various nuclear factors (Gadgil et al., 2001, Gadgil & Taylor & Jarrett, 2001). Sequential purification steps have yielded a 1000-fold improvement in TF purity (Yang, 1998). For example, SP1 can be enriched 500–1000 fold with 90% homogeneity and 30% yield with just two successive purifications. Lastly, tandem DNA affinity chromatography, with at least two different REs, has successfully purified SP1 and CAAT-binding TFs from the same sample (Kadonaga & Tjian, 1986, Kerrigan & Kadonaga, 2001).

DNA Affinity Chromatography with Biotin-Avidin

Avidin- or streptavidin- solid supports can be employed to capture biotinylated DNA as a stationary phase to capture DNA-bound TFs. The affinity of biotin for avidin is extremely high and allows for stringent washing conditions (i.e., high salt and/or urea). A nuclear extract that contains biotinylated DNA is bound to the (strept)avidin, the sample is washed, and TFs are eluted with a high ionic strength buffer (Gadgil & Taylor & Jarrett, 2001). A major limitation of this technique is that avidin is highly charged and might non-specifically bind other proteins, decreasing TF purity (Franza et al., 1987, Gadgil & Taylor & Jarrett, 2001, Jarrett & Foster, 1995). Alternatively, Neutravidin, a de-glycosylated form of avidin with a near-neutral isoelectric point, can be selected to reduce non-specific binding problems (Chen et al., 2008).

Trapping

Trapping methods have been developed by our laboratory to allow more efficient purification of TFs at more physiologically relevant concentrations (Figure 1). The specific TFs typically bind to their respective REs with nM-pM affinity in vivo, and a more specific approach to purification is to mimic these concentrations in vitro. In DNA affinity chromatography, immobilized REs present at µM or higher concentrations, encourage undesirable non-specific binding. In the trapping method, duplex DNA, such as an oligonucleotide RE or an entire promoter that contains a single stranded poly(GT)₅ tail is incubated at nM concentrations with nuclear extract or a crude cell lysate. The specific TF-RE complex is allowed to form in solution. Next, the TF-RE complex is ‘trapped’ with a poly(AC)₅ Sepharose column which anneals with the poly(GT)₅ present on the duplex DNA (Gadgil & Jarrett, 2002, Jiang et al., 2008, Moxley & Jarrett, 2005). In order to reduce non-specific binding on trapped material, modifiers that diminish non-specific binding (heparin, poly dI:dC, T18, and non-ionic detergents) are added to the mixture. Trapped TF-DNA complexes are washed with low salt, and proteins are eluted with the addition of high salt buffer that disrupts TF-DNA interactions. Conditions need to be optimized for each oligonucleotide or promoter used in the trapping method. Compared with DNA affinity chromatography, trapping yields TFs of higher purity (Gadgil & Jarrett, 2002, Jiang & Zhou & Moxley & Jarrett, 2008, Moxley & Jarrett, 2005), and in some cases only requires a one-step purification from crude nuclear extract.

Oligonucleotide Trapping

Trapping with an oligonucleotide RE will result in less non-specific binding than trapping with an entire promoter based on size limitations of the duplex region. Typically, trapping experiments are performed with oligonucleotides between 18–40 bp in length. This method has been used to successfully purify rat liver C/EBP, Xenopus B3 TF (Gadgil & Jarrett, 2002, Moxley & Jarrett, 2005), MafA (Matsuoka et al., 2003) and a complex that binds to a novel RE present in the c-Jun promoter (Jiang & Zhou & Moxley & Jarrett, 2008). Figure 2A shows oligonucleotide trapping where the reaction mixture consist of bacterial crude extract, radio-labeled C/EBP oligonuclueotide DNA, binding buffer (1X TE, 50 mM NaCl, 10 mM HEPES, 10 mM MgCl₂, 1 mM EDTA, 50 µM ZnSo₄, 1 mM DTT, 0.1% Tween, pH 7.5) and poly dI:dC. The use of poly dI:dC reduces the degree of non-specific binding. The reaction mixture incubated at room temperature for 30 minutes to allow assembly of DNA-TF complexes. DNA-TF complexes were revealed by their higher molecular mass (M_r) on the EMSA gel (top band). Characterization on this band was performed with a variety of methods such as liquid chromatography-tandem mass spectrometry, Southwestern Blots, and Western Blots. (data shown in Figures 2B–2D).

Promoter Trapping

Oligonucleotide trapping uses short DNA sequences (typically 18–40 bp) whereas promoter trapping uses longer DNA sequences (100–1000 bp). Because of the increased length and relative number of REs present on a promoter sequence, there exists a greater chance for non-specific binding; thus, modifying conditions will need to be more carefully optimized. This method has been used to successfully purify the PIC and specific TFs for the c-Jun promoter (Jiang et al., 2006). The PIC and specific TFs that bind to the c-Jun promoter were successfully purified and characterized with the c-Jun minimal promoter sequence (−200 to +81). The modifying conditions involved optimization of the concentration of both promoter and nuclear extract. In our experience, the c-Jun promoter trapping technique is good starting point for promoter trapping of other TFs. An example of promoter trapping is shown in Figure 1A. A reaction mixture containing nuclear extract (cell line HEK293), human telomerase reverse transcriptase (hTERT) promoter DNA containing poly(GT)₅ tail, binding buffer (1X TE, 50 mM NaCl, 10 mM HEPES, 10 mM MgCl₂, 1 mM EDTA, 50 µM ZnSo₄, 1 mM DTT, 0.1% Tween, pH 7.5), and poly dI:dC. The mixture is incubated at room temperature for 30 minutes to allow for assembly of DNA-TF complexes. DNA-TF complexes are captured by annealing (GT)₅ tails to a (CA)₅-Sepharose column. Unbound proteins are then washed with 20 column volumes of low salt buffer (1X binding buffer) followed by elution of DNA-TF complexes with 5 column volumes of high salt buffer (1X TE 0.5M NaCl pH 7.5).

1-DE

Sodium-dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), isoelectric focusing (IEF), or EMSA gels can be used for further purification of TFs by size, charge, and DNA-binding, respectively. Furthermore, these one-dimensional methods can be combined to provide multiple dimensions of separation.

2-DE

Two-dimensional electrophoresis (2-DE) combines isoelectric focusing (1^st dimension) with SDS-PAGE (2^nd dimension) to separate proteins by charge and size (O'Farrell, 1975). This technique has been utilized for the MS identification of proteins and also has been adapted for EMSA. Two-dimensional electrophoretic mobility shift assay (2D-EMSA), is a variant of this method useful for TFs or other nucleic acid-binding proteins from crude extract (Stead & McDowall, 2007). In this method, partially purified or crude extract was separated with 2-DE. The isoelectric point (pI) and the relative molecular mass (M_r) for a protein can be estimated. Examples of 2DE-gels are shown in Figures 1B and 2B. Both samples were first resolved by their isoelectric point (pI). Isoelectric focusing was performed with a 7cm, pH 3–10 IPG strip. The second dimension resolved the TFs on a 12% SDS-PAGE by separating the proteins by their relative molecular mass (M_r). Corresponding silver-stained spots on the 2-DE can be excisedelectro-eluted, re-natured, and assayed with EMSA. An alternative to 2D-EMSA is to replace the IEF dimension of 2-DE with EMSA. The TF-RE complex can be resolved from free RE due to its slower electrophoretic mobility. SDS-PAGE is used as the second dimension and TFs that were bound to the RE are observed as spots on the gel. These spots are excised, digested and subjected to MS analysis for protein identification. For example, the Streptomyces coelicolor TF AtrA was successfully purified and characterized from crude bacterial extract with this method (Stead et al., 2006, Stead & McDowall, 2007). This technique, unlike SWB, has the advantage that renaturation after spot excision is unnecessary.

EMSA-3-DE

The successful development of 2D-EMSA has allowed our laboratory to develop another technique, three-dimensional EMSA (EMSA-3-DE) (Figure 2). EMSA-3-DE combines non-denaturing EMSA with traditional 2-DE to purify TFs. The EMSA gel is electroblotted to PVDF, and the specific TF-RE complex band is eluted and applied to isoelectric focusing and SDS-PAGE for additional 2-DE separation. The advantage of EMSA-3-DE is that it combines the specificity of EMSA with the resolution of 2-DE and allows for purification of TFs from a crude extract. This method was used to successfully isolate GFP-C/EBP fusion protein constructed to bind the C/EBP canonical sequence CAAT from a crude bacterial extract and C/EBP from a nuclear extract of HEK293 cultured cells (Jiang et al., 2011).

Characterization

Methods for TF identification and quantification with capillary LC/MS/MS and protein database searching are described below.

Identification

TF identification is typically performed, following TF purification, with femtomole (10⁻¹⁵ mol = fmol) amounts of low-abundance TF peptides injected on-column in microliter (µL) volumes at nanomolar (10⁻⁹M = fmol/µL) concentrations. However, low-abundance protein identification has been achieved with attomole (10⁻¹⁸ mol = amol) amounts of peptide injected on-column at picomolar (10⁻¹²M = amol/µL) concentrations (Haskins et al., 2001, Thakur et al., 2011). Careful attention to each stage of identification, from sample preparation to validation, is essential to achieve the low mass- and concentration-limits of detection needed to identify TFs. The major steps involved in TF identification are described below and illustrated by the examples shown in Figure 1D.

Sample Preparation

In vitro proteolysis is performed in gel-, in-solution or on-blot to generate protease-specific peptides amenable to capillary LC/MS/MS and protein database searching, such as tryptic peptides with C-terminal lysine or arginine residues. For TFs, high recovery of their corresponding low-abundance TF peptides, particularly large and/or hydrophobic peptides that adsorb to surfaces such as microcentrifuge tubes and autosampler vials, is critical for success. Recovery can be maximized by being mindful of sample-handling surfaces, sample-storage buffers, temperature, and additives to minimize adsorptive losses. Just as critical is the removal of more-abundant peptides that originate from more-abundant nuclear proteins such as heterogeneous nuclear ribonuclear proteins, a problem encountered even when TFs are purified with affinity tagging (Grosveld et al., 2005, Sebastiaan Winkler et al., 2002) or antibody-based methods such as immunoprecipitation, where masking of TFs by tags or antibodies can also be problematic. Thus, it is important to maximize both the absolute and relative amount of TF peptides recovered during sample preparation. Pooling peptides from multiple different sample preparation strategies, including various proteases and/or chemical techniques such as cyanogen bromide cleavage, is an important strategy for low abundance protein identifications (Labugger et al., 2003).

Capillary LC

The capacity of capillary LC columns is typically on the order of picomole (10⁻¹² mol = pmol) amounts of peptides injected on-column. Because the majority of these peptides are high-abundance peptides from masking proteins, a dynamic range of 3–6 orders of magnitude is required to identify fmol-amol (10⁻¹⁵⁻¹⁸) amounts of low-abundance peptides from TFs. Miniaturization of columns, stationary phase particles, and integrated nanoelectrospray emitters for capillary LC/MS/MS at low flow rates (~10–400 nL/min) improves dynamic range through improved separation of low-abundance peptides from high-abundance peptides and other molecules (e.g., salts, detergents, metabolites, etc.) that suppress ionization. Improved separation and ionization efficiency also provide more tandem mass spectrometer time for gas-phase fragmentation of closely eluting peptides.

MS/MS

Comprehensive gas-phase fragmentation of multiply-charged peptide precursor ions into complementary N- and C-terminal sequence-specific product ions, cleavage of bonds between every amino acid residue along the peptide backbone, enables unambiguous assignment of amino acid sequences from MS/MS spectra (Novak et al., 2005). Various fragmentation mechanisms, once considered uncommon, are now standard practice in the modern mass spectrometry laboratory, including: collision-induced dissociation (CID), electron transfer dissociation (ETD) and high-energy C-trap dissociation (HCD). The advantages and disadvantages of these mechanisms are beyond the scope of this review. However, ETD and other non-ergodic mechanisms should be seriously considered if labile PTMs, such as phosphorylation, are of interest. Pooling sequence-specific product ions from multiple different fragmentation mechanisms is a valuable means to bolster one’s confidence in low-abundance protein identifications (Wu et al., 2007). Figure 1D depicts a spot excised from the 2-DE gel shown in Figure 1B. The gel plug was further processed by in-gel digestion with trypsin. Capillary LC/MS/MS was performed by injecting 1 µg of the extracted peptides onto a 50 µm-i.d. column packed with 7 cm of 3 µm-o.d. C18 particles and an integrated nanoelectrospray emitter with a flow rate of 350nL/min. Gas-phase fragmentation of the most abundant peptides were performed with a hybrid linear ion trap-Fourier-transform tandem mass spectrometer via high-energy C-trap dissociation (HCD) to generate MS/MS spectra. Examples of the spectra that can be acquired through this technique are shown in Figures 1D and 2D.

Protein Database Searching

Several transcription factor protein databases, relatively small compared to whole proteome databases, enable rapid, automated assignment of TF sequences from MS/MS spectra by protein database searching algorithms (Fogel et al., 2005, Matys et al., 2006, Messina et al., 2004, Murali et al., 2011, Pfreundt et al., 2010, Schaefer et al., 2011, Wilson et al., 2008, Zhang et al., 2011, Zhang et al., 2011). Pooling results from multiple different databases, including our own in-house human TF database at http://proteomics.utsa.edu, and/or multiple database search algorithms is a particularly valuable means to maximize true-positive and minimize false-positive protein identifications (Kwon et al., 2011, Searle, 2010). Protein database searching was performed to identify putative TFs that bind the hTERT promoter in Figure 1D, albeit much about the hTERT promoter remains unknown. For example, the low abundance protein high-mobility group protein B1 (HMGB1), an important transcriptional activator and repressor (Agresti et al., 2003, Grasser et al., 2007, Mouri et al., 2008, Naghavi et al., 2003, Riuzzi et al., 2012, Ueda et al., 2004) was identified (IPI00419258). HMGB1 was the 24^th –ranked protein family with a Mascot protein score of 99 and sequence coverage of 37%, while the most abundant, 1^st-ranked protein, heterogeneous nuclear ribonucleoprotein A2/B1 (HNRNPA2B1), was identified with a Mascot protein score of 2030 and sequence coverage of 70%. The amino acid sequence assignment of product ions for the top-ranked tryptic peptide from HMGB1, IKGEHPGLSIGDVAK, spanning amino acid residues 113–127, yielded an expectation value of 0.00118 (i.e., probability of a random match in the IPI human protein database). A representative MS/MS spectrum for this peptide is shown in Figure 1D, where b- and y-type ion annotations designate charge retention on N- or C-terminal sequence-specific product ions, respectively. Validation was performed by accurate mass measurement. The observed monoisotopic mass of the unmodified, doubly-charged precursor ion for IKGEHPGLSIGDVAK at 760.9258 m/z was 1519.8371 Da, and the theoretical value of 1519.8358 Da was in excellent agreement, with less than 1 ppm mass error. Likewise, product ions were assigned with less than a 1 ppm (rms) mass error from predicted values. Confidence in ion assignments was also reflected by 36/253 sequence-specific product ions and assignment of the same amino acid sequence to other MS/MS spectra from the triply- and quadruply-charged precursor ion. As another example, protein database searching also identified transcription factor CP2 (TFCP2) (IPI00037599) (Cunningham et al., 1995, Lim et al., 1992, Ramamurthy et al., 2001). The amino acid sequence assignment of product ions for the top-ranked tryptic peptide from TFCP2, LHDETLTYLNQGQSYEIR, spanning amino acid residues 81–98, yielded an expectation value of 0.0000034. A representative MS/MS spectrum for this peptide is also shown.

Validation

Validation of low-abundance TF identifications by accurate mass measurements is an important means to improve dynamic range because it reduces the probability of false-positive assignment of amino acid sequences from MS/MS spectra with protein database searching (Lu et al., 2008). De novo sequencing, synthetic peptides, and MS/MS/MS are also strong tools for validation. For example, if the 1^st-ranked de novo-derived amino acid sequence matches the 1^st-ranked database-derived amino acid sequence, then it is highly unlikely that the 1^st-ranked protein identification is a false-positive (Bern et al., 2007, Johnson & Taylor, 2000, Lu & Chen, 2003, Ma et al., 2003, Sheng et al., 2000, Standing, 2003, Zhang & McElvain, 2000). Alternatively, the peptide can be synthesized to compare MS/MS and MS/MS/MS spectra, known as spectral library searching (Zhang et al., 2011). A comparison of several of these methods to validate low-abundance protein identifications has been shown previously (Haskins et al., 2001). Finally, combining different sequence-specific proteases for in vitro proteolysis and/or combining different gas-phase fragmentation methods, as described above, are also important means to validate TF identifications.

Quantification

Determination of the relative and absolute abundance of proteins has received unprecedented attention in recent years with the advent of a plethora of quantitative gel-free proteomic techniques, including label-free, isotopic labeling, and isobaric labeling. For an example of isotopic labeling, the absolute quantification of c-Fos was determined to be only 6000 copies per HeLa cell with stable isotopic labeling of amino acids in culture (SILAC) (Zeiler et al., 2011). For an example of isobaric labeling, the relative phosphorylation of STAT3 was determined in a cancer stem cell model exposed to a kinase inhibitor by a workflow that incorporates phospho-protein/-peptide enrichment and tandem mass tagging (TMT) (Nilsson et al., 2010). In another example of isobaric labeling, quantification of SP1, c-MYC, and p53 in tissue specimens from cancer patients was achieved with isobaric tagging for relative and absolute quantification (ITRAQ) (Cai et al., 2012). For an example of label-free quantification method, FOXO3A interactions with 14-3-3 proteins were quantified from peptide intensity profiles in sequentially diluted samples (Rinner et al., 2007). It is beyond the scope of this review to speculate as to which technique has the largest dynamic range for TF characterization at this time. Based on our experience with several of these methods, TF identification remains more sensitive than TF quantification unless specific TFs are targeted for analysis by selected reaction monitoring. Despite the improved sensitivity of targeted analyses (Stergachis et al., 2011) and with the exception of subunit stoichiometry within a well-characterized TF complex, it is usually of interest to discover previously unknown binding partners in TF complexes. Therefore, targeted quantification is expected to be uncommon in future studies, whereas unbiased quantification is expected to become of more interest.

Promising New Strategies

Sample preparation, instrumentation and computer algorithms for capillary LC/MS/MS and protein database searching continue to evolve at a phenomenal rate. Microwave-assisted digestion strategies for sample preparation are particularly appealing for TF analysis. In vitro proteolysis in the microwave requires seconds instead of hours and limits time-sensitive adsorptive losses of low-abundance TF peptides (Lill et al., 2007, Sandoval et al., 2007). Improvements in ion transmission to, and the dynamic range of, hybrid tandem mass spectrometers (Hossain et al., 2011, Saba et al., 2009), now ubiquitous, are also very promising. Likewise, improvements in computer algorithms for data analysis, particularly algorithms for validation of protein database searching results, as described above, are expected to enable more unexpected isoform measurements, including alternative splicing variants and PTMs, from characteristic mass shifts at specific amino acid residue(s). The dramatic evolution of top-down proteomics for capillary LC/MS/MS and protein database searching of low-abundance PTMs of histones (Kelleher, 2004, Parks et al., 2007) and other small-to-medium-sized proteins suggests that top-down proteomics might also be viable for TF characterization. In top-down proteomics, gas-phase fragmentation of intact proteins is performed without in vitro proteolysis, eliminating the ambiguity inherent when inferring protein information from peptide information. Finally, stable isotope standards and capture with anti-peptide antibodies (SISCAPA) is an attractive alternative to enrich for low-abundance TF peptides after in vitro proteolysis (Anderson et al., 2009).

Competing Technologies

Several platforms that incorporate antibodies conjugated to micrometer- or nanometer-sized beads have been developed for high-throughput, multiplexed analysis of low-abundance proteins, including TFs. For example, kits are available to simultaneously quantify ~40 TFs in a 96-well plate format, where the identity of each TF is deduced from the emission wavelength of dye(s) encapsulated within a luminescent bead conjugated to a corresponding anti-TF antibody. In nanoparticle-based immunoPCR, and variations thereof, the identity of each TF is deduced from the oligonucleotide sequence bound to a corresponding anti-TF antibody-nanoparticle conjugate (Malou & Raoult, 2011, Nam et al., 2003). The major advantage of these competing technologies and others (e.g., reverse-phase protein arrays) is improved dynamic range and sample throughput relative to capillary LC/MS/MS and protein database searching. However, just as the case for immunoblotting, the major disadvantage is that non-cross-reacting antibodies must be developed for each TF of interest, including isoforms that arise from alternative splicing and PTMs. ‘Catch and release’ technologies that can leverage the incredible improvements in massively parallel DNA microarray and high-throughput DNA sequencing, seek to quantify TFs by measuring the abundance of TF-bound DNA probes. However, akin to antibody technologies, these suffer from unknown specificity and cross-reactivity, and PTMs cannot be investigated.

Conclusions

Notable Achievements

Tremendous progress towards the illumination of the TF proteome has been made in recent years. For example, with 2DE-SWB and on-blot digestion, C/EBP was identified directly from an unfractionated nuclear extract (Jiang, 2009), whereas AP1 subunits and C/EBP were both identified directly from a nuclear extract when combined with EMSA-3-DE (Jiang, 2011). AP1 is a complex TF composed of homo- and heterodimers of c-Jun, c-Fos, ATF family members, and others. Previously, c-Jun had not been identified with proteomic analysis with the exception of one recent paper where biotin-avidin chromatography proved adequate for identification with MALDI-TOF-TOF-MS/MS and protein database searching (Yaneva & Tempst, 2003). The TF AP1 components are some of the least-abundant TFs and successful characterization has proved especially challenging. Our recent EMSA-3-DE identification of c-Jun required not only sophisticated electrophoretic methods for purification, but also state-of-the-art tools for protein identification with capillary LC/MS/MS and protein database searching.

Future Progress

Synergy among the TF purification and characterization methods described herein has partially overcome the dynamic range limitations of proteomic analysis. However, further improvements in dynamic range are needed to enable widespread investigation of the TF proteome, particularly for scarce and heterogeneous clinical specimens. Innovative methods must be harnessed in order to reveal the subtle but important changes in TFs involved in human health and disease. Given the importance of transcriptional regulation to the overall central dogma of Biology and the progress on TF proteomics thus far, TFs are fully expected to serve as prognostic/predictive biomarkers and novel therapeutic targets in future studies.

List of Abbreviations

2-DE

two-dimensional electrophoresis

AP1

activator protein 1

bp

base pair

bHLH

basic helix-loop-helix

bZIP

basic leucine zipper

C/EBP

CAAT enhancer binding protein

CID

collision-induced dissociation

EMSA

electrophoretic mobility shift assay

ETD

electron-transfer dissociation

FOX

forkhead-box

GFP

green fluorescent protein

HCD

high-energy C-trap dissociation

HEK293

human embryonic kidney cell line 293

HMGB1

high-mobility group protein B1

HNRNPA2B1

heterogeneous nuclear ribonucleoprotein A2/B1

hTERT

human telomerase reverse-transcriptase

HTH

helix-turn-helix

IEF

isoelectric focusing

ITRAQ

isobaric tags for relative and absolute quantification

LC/MS/MS

liquid chromatography-tandem mass spectrometry

Mr

relative molecular mass

MS

mass spectrometry

PAGE

polyacrylamide gel electrophoresis

PCR

polymerase chain reaction

PolII

RNA polymerase II

PIC

pre-initiation complex

PTM

post-translational modification

RE

response element

rNTP

ribonucleotidetriphosphate

SDS-PAGE

sodium-dodecyl sulfate-polyacrylamide gel electrophoresis

SILAC

stable isotope labeling of amino acids in culture

SWB

southwestern blotting

TF

transcription factor

TFCP2

transcription factor CP2

TMT

tandem mass tagging

WB

western blot