Targeted analysis of recombinant NF kappa B (RelA/p65) by denaturing and native top down mass spectrometry

Abstract

Measuring post-translational modifications on transcription factors by targeted mass spectrometry is hampered by low protein abundance and inefficient isolation. Here, we utilized HaloTag technology to overcome these limitations and evaluate various top down mass spectrometry approaches for measuring NF-κB p65 proteoforms isolated from human cells. We show isotopic resolution of N-terminally acetylated p65 and determined it is the most abundant proteoform expressed following transfection in 293T cells. We also show MS¹ evidence for monophosphorylation of p65 under similar culture conditions and describe a high propensity for p65 proteoforms to fragment internally during beam-style MS² fragmentation; up to 71% of the fragment ions could be matched as internals in some fragmentation spectra. Finally, we used native spray mass spectrometry to measure proteins copurifying with p65 and present evidence for the native detection of p65, 71 kDa heat shock protein, and p65 homodimer. Collectively, our work demonstrates the efficient isolation and top down mass spectrometry analysis of p65 from human cells, and we discuss the perturbations of overexpressing tagged proteins to study their biochemistry.

Graphical Abstract

INTRODUCTION¹

Nuclear factor kappa B (NF-κB) is a transcription factor intimately involved in the inflammatory response, proliferation, and survival. Functionally, NF-κB exists as a dimer of proteins from the NF-κB family of genes: nfkb1, nfkb2, rel, rela, and relb. The most studied of these genes is rela, which encodes a 60 kDa protein, RelA/p65, referred to as p65 from here on. As with all NF-κB family members, p65 contains an N-terminal Rel Homology Domain (RHD) that functions in dimerization with NF-κB proteins. On its C-terminus, p65 contains a transcriptional activation domain similar to rel and relb. Under resting conditions, p65 is held in the cytoplasm in an inactive state through interaction with an inhibitor of kappa B protein (IκB). Upon stimulation, for example tumor necrosis factor (TNF) signaling, the IκB interaction is disrupted and p65-containing dimers translocate to the nucleus to induce the gene expression response[4–6].

The NF-κB transcriptional response is subject to multiple levels of regulation. In addition to IkB-mediated cytoplasmic sequestration, post-translational modification (PTM) of NF-κB proteins is also known to potentiate its function. p65 contains numerous sites for PTMs, including ten phosphorylation sites, seven acetylation sites, and five methylation sites[7]. Owing to a large body of work, the cellular pathways and enzymes responsible for many of these PTMs are known[7]; however, their functional significance is less defined. Notwithstanding, it has been shown that certain p65 modifications influence the addition of other p65 modifications[8–10]. Along these lines, and in analogy to the “histone code”[11], where specific combinations of PTMs on histone proteins provide the structural docking sites for other transcriptional regulatory proteins, it has been proposed that combinations of p65 modifications may similarly constitute an “NF-κB code”[12], a specific example of the more generalized “protein code”[13] hypothesis. If the NF-kB code hypothesis is true—where combinatorial modifications play a role in determining protein function—it becomes critical that we develop methods to readout these combinatorial modifications (so called “mod-codes” [14]).

Top down mass spectrometry is a powerful tool for measuring combinatorial PTMs. As opposed to the more routine bottom up mass spectrometry approach, where proteins are digested into peptides prior to measurement, top down omits enzymatic digestion and measures the mass of protein molecules in their intact, covalent state[15]. Top down therefore has a clear advantage over bottom up for measuring the overall stoichiometry of PTMs, as it is uniquely capable of providing the identity and relative abundance of individual “proteoforms”, a term used to describe the complete chemical formula and position of modifications to a protein product of a specific gene[3]. In theory, top down mass spectrometry is the best method to measure p65 proteoforms.

As of today, two main issues have impeded p65 proteoform measurement within human cells, as well as transcription factors in general: 1) top-down requires relatively large amounts of sample and p65 is expressed at low levels (1.2 × 10⁵ molecules p65 per cell[16] compared to 1.5 × 10⁸ molecules cytosolic actin per cell[17]); 2) mass spectrometry of intact proteins >30 kDa is associated with measurement challenges[18], including the difficulties of resolving isotopes on large proteins and generating fragmentation spectra in MSⁿ to aid in protein identification and characterization. For these reasons, it is unlikely that current high-throughput top down proteomics experiments will characterize transcription factor proteoforms, owing to their generally large size and low relative abundance[19]. Therefore, targeted approaches are required at present.

Here, we utilized previously developed HaloTag fusion protein technology[20, 21] to isolate and purify p65 prior to targeted proteoform analysis by top down mass spectrometry; HaloTag is a modified bacterial dehalogenase enzyme that enables covalent capture and fluorescent labeling of recombinant proteins. With the advantages of efficient protein yield in non-denaturing conditions afforded by HaloTag purification, we assessed the different MS¹ strategies for measuring the intact mass of the 62 kDa p65 target, including denaturing LC-Orbitrap Fourier-transform mass spectrometry (FTMS) and native direct infusion on an Orbitrap FTMS, and finally obtained isotopic resolution by a 12 Tesla Fourier-transform ion cyclotron resonance (FTICR) mass spectrometer. We observed a proclivity for p65 to fragment internally during beam-style MS² and demonstrate the effective implementation of our recently developed software for matching and displaying internal protein fragment ions [22]. We also observed a mass shift consistent with phosphorylation of p65. Finally, we used bottom up to complement our top down study and showed the copurification of proteins from four of the five NF-κB family genes as well as an IκB protein. Collectively, our data show the successful implementation of targeted top down to measure intact p65 species, proteoforms and even “complex-o-forms” from human cells and provide some perspectives on the analytical challenges and artifacts of present affinity-based approaches.

MATERIALS AND METHODS

Cell culture, protein expression, isolation, and imaging

Human embryonic kidney 293T cells were grown at 37°C and 5% CO₂ in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 u/mL penicillin, and 100 μg/mL streptomycin. Human RelA/p65 fused to a C-terminal linker peptide with the TEV protease recognition sequence followed by the HaloTag was obtained from Promega Corporation (Madison, WI) and was verified by DNA sequencing. For transfection, 50 – 70% confluent cells in either T175 flasks or 150 mm dishes were transfected with 24 μg DNA using PEI (poly(ethylene imine)) at a DNA:PEI ratio of 1:3 and a final transfection volume of 1/10^th of the total volume on cells, as described in the HaloTag® Mammalian Protein Purification System protocol (Promega, G6790). At one day post-transfection, cells were either harvested or treated with 20 ng/μL TNFα and 50 nM Leptomycin B for 1 h prior to harvest. HaloTag protein purification was performed following the manufacturer’s instructions with isolation buffer consisting of the following: 50 mM Hepes (pH 7.5) supplemented with 150 mM NaCl, 1mM DTT, and 0.005% Igepal CA-630. Protease inhibitor cocktail supplied with the HaloTag purification system (G6795) was included during lysis and binding. Imaging was performed using TMRDirect™ ligand following the manufacturer’s instructions (Promega, G6795) with additional fixation and staining with ProLong® Gold antifade reagent with DAPI (Invitrogen, P36931) where applicable. Confocal microscopy was performed using a Leica SP5 II laser scanning confocal microscope. Epifluorescence microscopy was performed with a Zeiss Axiovert 40 CFL microscope and affixed digital camera.

Bottom up LC-MS/MS

Two different bottom up LC-MS/MS experiments were performed: one in which a narrow gel band was excised and in-gel digested (Table 1), another in which the total p65-HaloTag purification was treated with protease in solution (Table 2). For in-gel digestion, p65-HaloTag purification was performed on transfected 293T cells. The resulting protein sample was resolved by SDS-PAGE and Coomassie stained. A small piece of the gel spanning the width of the lane and approximately 5 mm in height was excised and chopped into smaller pieces. In-gel trypsin digestion was then performed as follows: 1) the gel was destained by successive washes with water, acetonitrile (ACN), 100 mM ammonium bicarbonate (ABC), and 100 mM ABC/ACN (50:50 v:v), with the last two steps repeated until destained; 2) dehydrated using ACN and a speedvac; 3) reduced and alkylated using Tris (2-carboxyethl) phosphine and iodoacetamide, respectively; 4) washed and dehydrated again; 5) digested by overnight incubation with 250 ng trypsin at 37°C in 50 mM ABC. The resulting peptides were desalted by C₁₈ zip tip (Millipore, ZTC18S096) and resuspended in 20 μL Buffer A (95% water, 5% acetonitrile, 0.2% formic acid). Replicate LC-MS/MS injections were performed using nanoflow C18 reverse phase chromatography and a hybrid linear trap-7 Tesla FT-ICR mass spectrometer (Thermo). Peptides were eluted by ramping Buffer B (5% water, 95% acetonitrile, 0.2% formic acid). Data dependent acquisition with MS² of the 8 most abundant ions from each precursor scan was performed with the following settings: MS¹ = FTMS, 400 – 1600 m/z50,000 resolving power, 1 μscan, AGC 1e6; MS² = Ion Trap, CID at 35 NCE, 2 m/z isolation width, 2 μscans, AGC 1e4; dynamic exclusion = repeat count of 3, repeat duration of 45 s, exclusion duration of 120 s. For in-solution digestion, samples were acetone precipitated following HaloTag purification. The proteins were processed following standard in-solution digestion protocol, including cysteine alkylation with iodoacetamide and trypsin digestion. Peptides were analyzed using a VelosPro Oribitrap and 200 min. reverse phase gradient with the following instrument settings: MS¹ = FTMS, 400 – 2000 m/z, 60,000 resolving power, 1 μscan, AGC 1e6; MS² = Ion Trap, CID at 35 NCE, 1.5 m/z isolation width, 1 μscans, AGC 1e4; dynamic exclusion = repeat count of 2, repeat duration of 9 s, exclusion duration of 45 s. For data analysis and protein identification, Thermo .raw files were converted to Mascot generic format (.mgf) and Mascot was used to search against the Swiss-Prot human database. All peptide matches below an ion score cut-off of 30 were removed. Protein identifications required at least two peptides. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository[23] with the dataset identifier PXD001932 and 10.6019/PXD001932.

Table 1.

p65-HaloTag copurifying proteins identified by in-gel digestion of the predominant protein band^*

Accession #	Protein Description	Mass (Da)	Scans	Peptides
P08107	Heat shock 70 kDa protein 1A/1B	70009	60	11
P11142	Heat shock cognate 71 kDa protein	70854	58	12
Q04206	Transcription factor p65	60181	58	12
P38646	Stress-70 protein, mitochondrial	73635	46	10
P04264	Keratin, type II cytoskeletal 1	65999	28	7
P11021	78 kDa glucose-regulated protein	72288	26	9
P13645	Keratin, type I cytoskeletal 10	58792	12	3
P35908	Keratin, type II cytoskeletal 2 epidermal	65393	9	4
P02763	Alpha-1-acid glycoprotein 1	23497	8	3
Q12931	Heat shock protein 75 kDa, mitochondrial	80060	7	2
P35527	Keratin, type I cytoskeletal 9	62027	6	2
P52272	Heterogeneous nuclear ribonucleoprotein M	77464	6	2
O43390	Heterogeneous nuclear ribonucleoprotein R	70899	4	3
Q9NZI8	Insulin-like growth factor 2 mRNA-binding protein 1	63441	4	3
P13647	Keratin, type II cytoskeletal 5	62340	3	3
O60506	Heterogeneous nuclear ribonucleoprotein Q	69560	3	2

^*Gel piece analyzed spanned approximately 60 – 75 kDa

Table 2.

Proteins identified from in-solution digest of total protein isolated from a p65-HaloTag purification

Accession #	Protein Description	Mass (Da)	Scans	Peptides
Q04206	Transcription factor p65	60181	648	15
P08107	Heat shock 70 kDa protein 1A/1B	70009	44	11
P11142	Heat shock cognate 71 kDa protein	70854	37	12
P38646	Stress-70 protein, mitochondrial	73635	33	10
P11021	78 kDa glucose-regulated protein	72288	33	13
Q04864	Proto-oncogene c-Rel	68476	8	5
O00505	Importin subunit alpha-4	57775	6	4
O43390	Heterogeneous nuclear ribonucleoprotein R	70899	5	3
O60506	Heterogeneous nuclear ribonucleoprotein Q	69560	5	3
P19838	Nuclear factor NF-kappa-B p105 subunit	105290	5	3
O00629	Importin subunit alpha-3	57851	5	3
Q00653	Nuclear factor NF-kappa-B p100 subunit	96689	4	2
P52292	Importin subunit alpha-1	57826	4	2
Q15653	NF-kappa-B inhibitor beta	37748	4	2
Q14974	Importin subunit beta-1	97108	3	2
Q14739	Lamin-B receptor	70658	3	2
P52272	Heterogeneous nuclear ribonucleoprotein M	77464	2	2
…
117 total proteins identified ≥ 2 peptide matches (see Table S1)

Denaturing Top down LC-MS and LC-MS/MS

In all cases except the 12 Tesla FT-ICR (Figure 2C), chromatography was performed using a monolithic RP-4H column from Thermo; for the 12T FT-ICR, a PLRPs column was packed in house and nanoflow was performed at 300 nL/min. For the data in Figure 2A and 2B, LC elution was performed at nanoflow rates (200 – 300 nL/min); the data in Figure 3A was obtained using 1 μL/min. Mass spectrometry instrument settings varied by analyzer and were as follows: Ion Trap (Orbitrap Elite, Figure 2A) = 1.8 kV source voltage, 320°C capillary temperature, 15 V SID, 500 – 2000 m/z, AGC 3e4, 25 μscans; FTMS (Orbitrap Elite, Figure 2B) = 1.8 kV source voltage, 320°C capillary temperature, 15 V SID, 501 – 2000 m/z, 15,000 resolving power, AGC 1e6, 4 μscans; FT-ICR (12T VelosPro/FT-ICR custom hybrid, Figure 2C) = 1.8 kV source voltage, 320°C capillary temperature, 15 V SID, 650 – 1500 m/z, 171,000 resolving power (100,000 resolving power setting × (12/7) to account for custom upgrade to 12 T magnet from the factory settings of 7 T), AGC 1e6, 6 μscans. The data in figure 3 were obtained on a Q-Exactive modified for large protein analysis[24]. The instrument settings were as follows: 15 V SID, nanospray 2.1 kV, 320°C capillary temperature, MS¹ = 500-2000 m/z, AGC target 1e6, 25 μscans, resolving power 7,500 (at m/z 200), MS² = isolation window 100 m/z around 1013 m/z, HCD 10 and 22 volts, resolving power 60,000 (at m/z 200), AGC target 1e6, 6 microscans.

Figure 2.

MS¹ spectra of intact p65HC following transfection and purification from 293T cells. p65HC refers to the p65 plus 13 amino acid C-terminal linker resulting from TEV protease treatment. Reverse phase LC-MS was performed and the spectra were generated by averaging scans across the chromatographic peak. Red is the theoretical isotopic distribution of the 64+ charge state of N-acetyl p65HC; black is the observed data. Res = resolution as calculated from the data. (A) Ion trap data collected on an Orbitrap Elite. (B) FTMS data collected on an Orbitrap Elite set at 15,000 resolving power at 400 m/z. (C) FTMS data collected on a VelosPro/12 T FT-ICR set for 170,000 resolving power at 400 m/z. 61,746.1 –0 refers to the theoretical monoisotopic mass of N-acetyl p65HC. * FT noise.

Figure 3.

(A) MS¹ spectra of intact p65HC from transfected 293T cells treated with TNFα and Leptomycin B for 1 hour prior to harvest. LC-MS was performed using reverse phase chromatography and a custom modified Q Exactive quadrupole/Orbitrap hybrid set to a transient acquisition length of 7,500 resolving power at 200 m/z. The spectrum was generated by averaging scans across the chromatographic peak. Red is the theoretical isotopic distribution of the 64+ charge state of N-acetyl p65HC; black is the observed data. Res = resolution as calculated from the data. (B) Fluorescence microscopy of p65-HaloTag in transfected 293T cells. Live cells were stained with TMR direct ligand, imaged (untreated), co-treated with TNFα and Leptomycin B for 45 min., and imaged again (+ TNF/LepB) using an epifluorescence microscope. (C & D) Spectra produced by higher energy collisional dissociation (HCD) of a 100 m/z window around the 1013 m/z peak from 3A above. Collisional energies of 10 and 22 volts were used. Transient acquisition length was set for a resolving power of 60,000 at 200 m/z. Pie charts describing the ions were produced by manual inspection. (E & F) Fragment heat maps combined with graphical fragment maps for terminally matching b- and y-fragment ions. Each amino acid is color-coded based on the number of times it is included in an internal fragment ion match, ranging from no color (not included) to dark red (included the most).

Manual spectral analysis and modeling of theoretical isotopic distributions was performed using mMass open source software[25–27]. Fragment ion matching was performed with the freeware ProSight Lite (http://prosightlite.northwestern.edu/) and our recently developed in-house software for searching and displaying internal fragment ions [22]. In all cases, fragment mass tolerance was 10 ppm at most; internal fragments matched within 4 ppm following internal calibration. Fragment ion spectra were manually validated. RAW files supporting this paper will be accessible at http://www.topdownproteomics.org/data.

Native Top down MS

For native spray MS, HaloTag-purified sample was subjected to HiPPR detergent removal resin column kit (Thermo, Product # 88305) followed by buffer exchange with a 30 kDa molecular weight cut-off filter into 100 mM ammonium acetate pH 7. Samples were electrosprayed into a modified Q-Exactive HF (Thermo Fischer Scientific) as described previously[24] using a simplified sheath flow capillary electrophoresis setup[28]. RAW files supporting this paper will be accessible at http://www.topdownproteomics.org/data.

RESULTS AND DISCUSSION

To increase protein yield, we exogenously expressed p65-HaloTag, tagged on the C-terminus. HaloTag is a protein labeling and purification technology consisting of a 34 kDa modified bacterial dehalogenase enzyme fused to a protein of interest by way of a linker containing the TEV protease recognition sequence. Labeling and purification of HaloTag fusion proteins are based on the same concept, where the dehalogenase activity of the tag produces a covalent bond between the fusion protein and the ‘bait’, either a fluorescent molecule or a resin. In the case of purification, after the fusion protein is isolated, tag-free protein is removed from the resin in its native state by TEV protease[29, 30]. Others have shown that adding the HaloTag to p65 does not significantly disrupt p65 function; p65-HaloTag is able to translocate to the nucleus and associate with DNA following cytokine stimulation[29]. Another study showed an increased yield and purity of p65 using the HaloTag strategy compared to His-tag and Flag-tag[30]. Granted the addition of HaloTag to p65 adds an unnatural protein species to the cellular milieu, and could interfere with post-translational modification of p65, the advantages of ease of expression, high yield and purity lead us to choose exogenous expression of p65-HaloTag for our developmental studies.

To evaluate p65-HaloTag expression and isolation efficiency, we transfected human 293T cells and performed fluorescent labeling and purification. Transfection efficiency was 50% and localization was predominantly cytoplasmic (Figure 1A). Consistent with previous studies, co-treatment with tumor necrosis factor (TNF)α and Leptomycin B (LepB), a nuclear export inhibitor, resulted in p65-HaloTag nuclear accumulation in transfected cells (Figure 3B)[29]. Purification followed by SDS-PAGE and silver staining yielded two predominant bands at approximately 65 kDa (Figure 1B). Subsequent in-gel trypsin digestion and LC-MS/MS identified p65 (60 kDa) with 24% sequence coverage (Figure 1C). Each p65 peptide was manually validated (two examples shown in Figure 1D). In addition to p65, 15 other proteins were identified in the excised gel band, including heat shock proteins (70 kDa), mitochondrial proteins (70 – 80 kDa), ribonucleoproteins (70 – 80 kDa), and keratins (60 – 70 kDa) (Table 1). Heat shock proteins and p65 were the most abundant based on the number of spectral matches (Table 1). These data demonstrate successful HaloTag purification of p65 and copurifying proteins from transfected 293T cells at sufficient purity and quantities for MS analysis.

Figure 1.

(A) Fluorescent imaging of p65-HaloTag in 293T cells. Cells were transfected with p65-HaloTag, stained with TMR direct ligand (red) and DAPI (blue), and imaged by confocal microscopy. Scale bar is 22 μm. (B) SDS-PAGE and silver staining of p65-HaloTag purification from transfected 293T cells. (C) Schematic representation of p65 peptides identified by in-gel digestion and LC-MS/MS. Transfection, purification, and SDS-PAGE were performed as in Figure 1B above. The gel was stained with Coomassie and a single gel piece ranging from 60 – 75 kDa was excised and trypsin digested. The sample was analyzed with a hybrid linear trap/7T FT-ICR and reverse phase chromatography with MS² of the 8 most abundant ions from each precursor scan. The full list of protein identifications is presented in Table 1. (D) MS² spectra matching to p65, with matching b- and y-ions highlighted in red.

To assess the ability of different mass analyzers to measure intact p65 by top down, we subjected purified p65-HaloTag (referred to as p65HC) to reverse phase LC-MS using an ion trap (Figure 2A), an Orbitrap using a short transient (Figure 2B), and a 12 Tesla FT-ICR (Figure 2C). A charge state envelope of a large protein was detected using each of the analyzers (Figure 2A – C). The resolution at 967 m/z was 1,400 for the ion trap (Figure 2A), 3,800 for the Orbitrap (Figure 2B), and 64,000 for the FT-ICR (Figure 2C). Isotopes were partially resolved in the FT-ICR (Figure 2C). Application of Xtract (a high-resolution deconvolution algorithm) to the FT-ICR data resulted in a monoisotopic mass of 61,746.1 Da, within 2 ppm of the mass predicted for N-terminally acetylated p65HC (Figure 2C). Subsequent MS² confirmed the identity as N-acetyl p65HC (vide infra, Accession no. Q04206, plus 13 C-terminal residues from linker). The Orbitrap and FT-ICR outperformed the ion trap in mass accuracy and were similarly accurate (Figure 2, compare red to black). Potential mass shifts were observed at 25 – 50% relative abundance in the ion trap and FT-ICR (Figure 2A & C), but not in the Orbitrap (Figure 2B). These data show isotopic resolution of a 62 kDa protein by FT-ICR and suggest N-acetyl p65HC with no other modifications is the most abundant p65 proteoform following transfection in 293T cells.

To further evaluate top down analysis of p65, we co-treated transfected 293T cells with TNFα and Leptomycin B and subjected the HaloTag-purified sample to LC-MS using a resolving quadrupole/Orbitrap hybrid mass spectrometer set to the lowest possible transient acquisition length, to enable maximum sensitivity yet retain the mass accuracy of Orbitrap FTMS. As stated earlier, TNFα/LepB treatment resulted in p65-HaloTag nuclear accumulation compared to untreated cells (Figure 3B). Similar to LC-MS of untreated cells (see Figure 2), a charge state envelope of a 62 kDa protein was detected (Figure 3A, left) and the mass of the most abundant species was that of N-acetyl p65HC (Figure 3A right, compare red to black). The resolution at 966.5 m/z was 2,800 and there was a potential mass shift of +79 Da at ~25% relative abundance (Figure 3A), which is within tolerance of a phosphorylation (+80 Da) at this resolution.

Higher energy collisional dissociation (HCD) of a 100 m/z window around the 1013 m/z peak (Figure 3A, left) produced fragment ions using collisional energies of 10 and 22 volts (Figure 3C & D). By visually examining the MS² spectra in Figure 3, at least 8 fragment ions were present at both HCD energies, but the majority are unique between the two (Figure 3C & D). In general, the fragment ions generated at higher HCD energy were of lower m/z compared to those generated with lower HCD energy (compare Figure 3D to C).Fragments ions from both energies were matched to termini of p65HC (Figure 3E & 3F), however, greater than 80% of both spectra were left unexplained by searching solely for terminal fragments (Figure 3C & D, pie charts).

Employing our in-house internal fragment search software [22] to MS² spectra in Figure 3C & D indicated the majority of ions from each of these spectra were internal fragment ions; after manual validation, 55% of the HCD 10 spectra and 71% of the HCD 22 spectra could be explained as internals (Figure 3C & D, pie charts). The images in Figure 3E & F are a combination of graphical fragment maps — showing matching b- and y-ions—and fragmentation heat maps, where each amino acid is color coded based on the number of matched internal fragment ions that contain it, with darker red indicating more matches and white indicating no matches. Due to chemical formula redundancy, only one internal fragment ion from the HCD 10 spectra could be unambiguously assigned (Figure 3E); the remaining internal fragment ions matched to multiple p65HC peptide sequences of the same chemical formula and were omitted from the display.

At an HCD energy of 22 volts, p65HC sequence coverage approached 100% when internal fragment ions were included in the search, a dramatic increase compared to terminal fragments alone (Figure 3F). Internal fragmentation clustered around two regions, one on each end of the protein (Figure 3F). In general, regions nearby acidic amino acids (D and E) and prolines were covered more frequently, consistent with D/E and P being favorable sites of cleavage during collisional fragmentation[31]. Combined with the observation that the HCD 22 spectra (Figure 3D) produced an increased number of lower m/z ions than the HCD 10 (Figure 3C), these data indicate that increasing fragmentation energy increases the number of p65HC internal fragment ions, drastically increasing the characterization of the protein.

Because the HaloTag strategy offers the unique ability to obtain target proteins in their native state, we performed native mass spectrometry on the p65-HaloTag purified sample (Figure 4A). The spectrum in Figure 4A was generated using the same quadrupole/Orbitrap hybrid used in Figure 3, which we have customized for analysis of native protein complexes[24]. Four species were detected by charge-state deconvolution of MS¹ spectra: 61,783 Da, 70,865 Da, 123,637 Da, and 165,660 Da (Figure 4A). MS² spectra were acquired, but, due to low signal intensity, did not allow for protein identification or subunit characterization. However, the average mass of N-acetyl p65HC is 61,784 Da suggesting that the observed 61,783 Da species (Figure 4A) within a Dalton is the same N-acetyl p65HC proteoform. While MSⁿ fragmentation will be required for unequivocal identification of the 61,783 Da protein as p65, our MS¹ detection of four protein species >60 kDa is sufficient for us to conclude native MS is possible using the HaloTag strategy; performing multiple stages of tandem MS to enable direct, top down identification of complex subunits is exceedingly difficult at present, especially for complexes formed in vivo and not reconstituted in vitro.

Figure 4.

(A) Native spray MS of p65-HaloTag purification from transfected 293T cells. A 30 kDa molecular weight cut-off filter was used to buffer exchange into 100 mM ammonium acetate, pH 7. The spectrum was generated using a custom modified Q Exactive quadrupole/Orbitrap with a resolving power setting of 15,000 at 200 m/z. Over 400 scans were averaged. (B) Total ion chromatogram with MS² insets of in-solution digestion of p65-HaloTag purification from transfected 293T cells. Data were acquired using reverse phase LC-MS/MS on a VelosPro Orbitrap with MS² of the top 10 most abundant ions from each precursor scan. For the MS² insets, red denotes ions that were matched to the respective proteins.

To further study the species in a p65-HaloTag purified sample detected by native MS, we repeated the purification and performed bottom up LC-MS/MS on the entire sample using in-solution digestion (Figure 4B). The resulting tryptic peptides were analyzed with a VelosPro Oribitrap over the course of a 200 min. reverse-phase LC gradient. Figure 4B displays the MS¹ total ion chromatogram with MS² spectra as insets. Over 100 proteins were identified by 2 or more peptides, 17 of which are presented in Table 2; the full list with a corresponding protein interaction network created by STRING (string-db.org) can be found in Table S1 and Figure S1, respectively.

p65 was the most frequently identified protein with 648 total spectral matches from 15 different peptide sequences (Table 2). Other proteins included those observed prior by in-gel digestion of the 60 – 75 kDa region (Table 1), such as heat shock 70 kDa protein 1A/1B, heat shock cognate 71 kDa protein, mitochondrial stress-70 protein, 78 kDa glucose-regulated protein, and the heterogeneous nuclear ribonucleoproteins M, Q, and R (Table 2). In addition, known interactors including c-Rel, NF-κB p105, NF-κB p100, NF-κB inhibitor beta (IκBβ), importin subunits alpha-4 and beta-1, and lamin-B receptor were identified (Table 2), all of which are consistent with p65 being a dimeric transcription factor regulated through an interaction with IκBβ[6].

According to the native MS data shown in Figure 4A, a mass within 1 Da of p65HC (61,783 Da) was detected. In addition to the 61,783 Da mass, which is likely p65HC, a mass of 70,865 Da was detected (Figure 4A). 70,865 Da is 11 Da heavier than heat shock cognate 71 kDa protein identified as a copurifying protein (Table 2). Therefore, it is possible given the low resolution of the native data that the 70,865 Da mass is heat shock cognate 71 kDa protein plus 11 Da error. Regarding the next heaviest mass, p65 can form homodimers[32]. The 123,637 Da mass detected by native MS is within experimental error (delta mass 71 Da) of a homodimer of the putative p65HC mass (Figure 4A). No single copurifying protein could explain the 165,660 Da mass detected by native MS (Figure 4A). Therefore, the 165,660 Da mass is likely a yet-to-be-characterized multimer resulting from a combination of p65HC copurifying proteins. Collectively, these data, both native MS and bottom up LCMS/ MS, indicate the protein sample resulting from HaloTag purification contains multiple known p65 interacting proteins and that a fraction of these can be detected by native spray MS. Although one of the four detected masses is likely p65HC (61,783 Da, Figure 4A), proving this, and identifying the other three masses unequivocally, will require future MS² fragmentation of ionic species generated by native MS.

Our work demonstrates the possibility of targeted purification and top down mass spectrometry of the 60 kDa NF-κB p65. Combinatorial PTMs to p65 play a crucial role in regulating NF-κB activity[12], therefore, having the ability to measure these combinatorial PTMs, and the relative abundance of the proteoforms we find them on, is key to advancing our understanding of molecular mechanisms governing gene regulation. However, our results did not yield the myriad p65 proteoforms one might expect given the numerous sites for post-translational modification to p65 (i.e. 10 phosphorylation, 7 acetylation, and 5 methylation sites). Why?

The analytical choices available for measuring proteoforms within a cell are many; we have chosen transfection of tagged recombinant protein (Figure 5). While overexpressing a protein facilitates advanced measurement techniques, including denaturing and native top down mass spectrometry (Figure 5), it is likely that overexpressing a protein introduces artifacts into the system, as the stoichiometry between enzymes and target are now altered. Accordingly, the phosphorylated p65 proteoform we detected (Figure 3A) should be considered in this light. Given the protein characterization power of top down, the major challenge moving forward is to achieve similar protein characterization on endogenous protein amounts.

Figure 5.

Three dimensional research space of targeted transcription factor analysis by mass spectrometry. Aspects to consider include expression type (transfection, transduction, or endogenous), biological state (inactive or active), and mass spectrometry (MS) measurement type (bottom up, denaturing top down, or native top down).

Others have used low levels of transfection DNA to achieve a compromise between overexpression and endogenous protein amounts when identifying protein phosphorylation sites[33]. We propose that future studies will employ targeted genome editing (e.g. CRISPR/Cas) to insert the HaloTag at the endogenous RelA/p65 locus. Having optimized the top down mass spectrometry characterization of p65 here, leveraging HaloTag capture from endogenously expressed p65 will enable biologically relevant proteoform description. It will then be informative to empirically determine the influence overexpression has on protein modification state.

CONCLUSION

Our data demonstrate the effective use of HaloTag for a targeted proteoform analysis by top down mass spectrometry of the 60 kDa NF-κB protein, p65. This initial foray into targeted transcription factor proteoform measurement lays an important analytical foundation for similar studies to follow; the foundation includes the HaloTag purification, internal fragmentation, and proteoform detection by denaturing and native top down mass spectrometry.

SIGNIFICANCE.

Characterizing transcription factor proteoforms in human cells is of high value to the field of molecular biology; many agree that post-translational modifications and combinations thereof play a critical role in modulating transcription factor activity. Thus, measuring these modifications promises increased understanding of molecular mechanisms governing the regulation of complex gene expression outcomes. To date, comprehensive characterization of transcription factor proteoforms within human cells has eluded measurement, owing primarily—with regard to top down mass spectrometry—to large protein size and low relative abundance. Here, we utilized HaloTag technology and recombinant protein expression to overcome these limitations and show top down mass spectrometry characterization of proteoforms of the 60 kDa NF-kB protein, p65. By optimizing the analytical procedure (i.e. purification, MS¹, and MS²), our results make important progress toward the ultimate goal of targeted transcription factor characterization from endogenous loci. Graphical abstract

HIGHLIGHTS.

• Measuring transcription factors by top down mass spectrometry is challenging.

• Using HaloTag, we purified sufficient amounts of NF-κB p65 for top down MS.

• We resolved isotopes and detected a phosphorylated p65 protein species by FTMS.

• We show p65 fragments internally during HCD, increasing protein coverage.

• HaloTag purification is amenable to native MS; p65 and p65 dimer were detected.

ABBREVIATIONS

ACN

acetonitrile

AGC

automatic gain control

CID

collisional induced dissociation

FT-ICR

Fourier transform ion cyclotron resonance

HCD

higher energy collisional dissociation

MS

mass spectrometry

MS1

intact/precursor scan

MS2 or MSn or MS/MS

tandem mass spectrometry scan, fragmentation

LC

liquid chromatography

SDS

sodium dodecyl sulfate

SID

source induced dissociation