Proteomic dissection of cell type-specific H2AX-interacting protein complex associated with hepatocellular carcinoma

Abstract

The replacement histone variant H2AX senses DNA double-strand breaks (DSBs) and recruits characteristic sets of proteins at its phosphorylated (γ-H2AX) foci for concurrent DNA repair. We reasoned that the H2AX interaction network, or interactome formed in the tumor-associated DNA DSB environment such as in hepatocellular carcinoma (HCC) cells, where pre-neoplastic lesions frequently occur, is indicative of HCC pathogenic status. By using an in vivo dual-tagging quantitative proteomic method, we identified 102 H2AX-specific interacting partners in HCC cells that stably expressed FLAG-tagged H2AX at close to the endogenous level. Using bioinformatics tools for data-dependent network analysis, we further found binary relationships among these interactors in defined pathway modules, implicating H2AX in a multi-functional role of coordinating a variety of biological pathways involved in DNA damage recognition and DNA repair, apoptosis, nucleic acid metabolism, Ca²⁺-binding signaling, cell cycle, etc. Furthermore our observations suggest that these pathways interconnect through key pathway components or H2AX interactors. The physiological accuracy of our quantitative proteomic approach in determining H2AX-specific interactors was evaluated by both co-immunoprecipitation/ immunoblotting and confocal co-localization experiments performed on HCC cells. Due to their involvement in diverse functions, the H2AX interactors involved in different pathway modules, such as Poly(ADP-ribose) polymerase1, 14-3-3ζ, coflin1, and peflin1, were examined for their relative H2AX binding affinities in paired hepatocytes and HCC cells. Treatment with the DSB-inducing agent bleomycin enhanced binding of these proteins to H2AX, suggesting an active role of H2AX in coordinating the functional pathways of each protein in DNA damage recognition and repair.

Introduction

DNA double-strand breaks (DSBs) induced by various stresses^1–5 represent a common genomic damage/lesion that may lead to genomic instability and ultimately, to cancer development^{2, 6}. The replacement histone variant H2AX^{7 8} plays a central role in both cellular responses to DNA damage and in repairing damaged DNA^{2, 3}. In the early cellular response to DNA DSBs, H2AX is phosphorylated at serine 139, probably by ataxia telangiectasia mutated (ATM), triggering various signal transduction cascades for DNA damage recognition and DNA repair^{9, 10}. The phosphorylated form of H2AX or γ-H2AX can recruit particular sets of proteins such as MRE11, RAD50, and NBS1 (MRN) in a complex with Brca1, 53BP1, and NFBD1/DNA damage checkpoint 1(MDC1) to form γ-H2AX foci at DNA DSB sites^11–13. For example, MDC1 recognizes γ-H2AX through its BRCT domain, and facilitates recruitment of additional MRN and ATM proteins that subsequently leads to the phosphorylation of additional H2AX and MDC1 molecules^{14, 15}. The MRN complex can interact directly with MDC1, then MDC1 binds ATM through its FHA domain and NBS1 through its Ser-Asp-Thr repeat^{14, 15}. Other evidence suggests that the ubiquitin-interacting-motif-containing protein RAP80 binds K63-linked ubiquitin chains of ubiquitinated proteins and, therefore, is able to recognize ubiquitinated H2A/H2AX, resulting in formation of a larger BRCA1/BARD1/CCDC98/RAP80 protein complex targeted to DNA damage foci^{14, 16, 17}. In addition to these H2AX interacting proteins, many pivotal DNA repair-related proteins such as 53BP1¹⁸ have also been found to interact with H2AX through different recruiting mechanisms.

H2AX and its interacting proteins play synergistic roles in tumourigenesis^{19, 20}. H2AX knockout mice were found with increased genomic instability and a higher risk of developing cancers^{11, 21}. Mutations or deletions in the H2AX gene are frequently found associated with various human cancers including acute myeloid leukemia, acute lymphoid leukemia, head and neck squamous carcinoma, etc^{8, 22–26}. In tumor cells from clinical specimens and cell culture, emerging evidences also showed increased levels of DNA DSBs and mutations in H2AX-interacting partner genes^{27, 28}. Defects in DNA damage response pathways, such as the MRN complex or the kinases ATM and RAD3-related (ATR), are associated with cancer predisposition syndromes in humans, and disruption of DNA repair systems in mouse models leads to an increased risk of cancer^{29, 30}. H2AX and its interacting proteins such as ATM, Chk2, p53 were found activated in bladder cancer³¹. Immunohistochemistry studies have shown DNA damage signal activation in precancerous bladder lesions, which is lost on progression, suggesting that the damage signaling acts as a brake to further tumorigenesis³². In addition, expression of RAD51, one of the critical H2AX interactors, is higher in breast cancer than in normal breast tissue³³. We therefore reasoned that different compositions of H2AX complexes may be specifically correlated to different human cancers, wherein the H2AX-associated pathways for DNA damage recognition and repair could be dys-regulated leading to un-repairable DNA DSBs.

Emerging mass spectrometry(MS)-based proteomic approaches with improved sensitivity, signal specificity, and throughput have been extended to efficiently and systematically analyze protein complexes and decipher protein interaction networks^34–36. Metabolic labeling strategies such as SILAC/AACT²⁶ have improved the signal specificity in distinguishing genuine protein-protein interactions in protein complexes^{37, 38}. In our previous studies of ionizing radiation-induced dynamic changes in the H2AX interactome using a dual-tagging quantitative proteomic approach³⁸, we found several novel binding partners of H2AX in human embryonic kidney (HEK) 293T cells under defined conditions, and demonstrated that the H2AX protein complex underwent dynamic changes upon induction of DNA damage and during DNA repair. Further analysis revealed a critical role for Ca²⁺/calmodulin in ionizing radiation-induced cell cycle arrest in HEK cells. Despite growing evidence of correlations between an abnormal H2AX interactome, and carcinogenesis, little is known how a characteristic H2AX interactome is structured in cancer cells and how it is dysregulated in respond to DNA DSBs and repair. Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related death worldwide³⁹. This high mortality is mainly due to unknown mechanisms of HCC pathogenesis and lack of precise HCC-characteristic markers for early diagnosis and therapeutic intervention^{39, 40}. One of the conspicuous features of HCC development is the frequent occurrence of chromosomal abnormalities³⁹. Microarray analyses revealed that expression of many DNA repair-related proteins was up-regulated in HCC cells⁴⁰. Also HCC tissues accumulate higher endogenous levels of γ-H2AX foci⁴¹. Some evidence suggests that the regulatory level of γ-H2AX may be tissue-specific^{42, 43}. Also, variable expression patterns of an H2AX partner, NBS1, in testis, thymus, spleen telencephalon, diencephalon, liver, lung, kidney, and gut are observed in mice⁴⁴. Organ-specific factors may have an impact on PARP-1, another H2AX interactor, to protect against genotoxic damage in mice⁴⁵. We therefore hypothesize that H2AX may recruit particular sets of protein interactors in HCC cells whereby H2AX coordinates complex cellular signals indicative of the pathological environment in HCC.

We describe here the use of DNA nuclease digestion of nuclei and a dual tagging (both epitope and SILAC tag) quantitative proteomics strategy^{37, 38}, to profile proteins interacting with H2AX in HCC cells. Using bioinformatics tools, we performed data-dependent network analysis to first validate protein interactions with H2AX and second to sort out their binary interactions. As an immediate outcome, multiple pathway modules were mapped and their interconnected links were established. The proteomic dataset and the selected components of particular pathway modules were further validated by co-immunoprecipitation/immunoblotting. The endogenous interactions between H2AX and selected interacting partners representing the pathways involved in DNA damage recognition and repair were localized by confocal scanning. Using bleomycin (BLM), the most used agent to induce DNA DSBs^{46, 47} we then examined the possible role of certain H2AX interactors in DNA damage recognition and repair in HCC cells. We also compared the H2AX binding affinities of the proteins in paired hepatocytes and HCC cells⁴⁸. All of these results provide insight into an HCC-characteristic H2AX interactome indicative of the pathological status of hepatocellular carcinoma.

Experimental Procedures

Materials

Deuterium-labeled leucine (5, 5, 5-d₃) was purchased from Cambridge Isotope (Andover, MA). Trypsin was purchased from Promega (Madison, WI). Hoechst 33342 was purchased from Sigma (St. Louis, MO). All components of cell culture medium were purchased from Invitrogen Corporation (Carlsbad, CA) except for puromycin, which was obtained from Calbiochem (San Diego, CA) and fetal bovine serum (FBS), which was obtained from PAA Laboratories GmbH (Linz, Austria). Chemicals for IP, SDS-PAGE gel electrophoresis, visualization, peptide extraction, and sample preparation of LTQ-Orbitrap were purchased from Sigma (St. Louis, MO). All the chemicals were sequence- or HPLC -grade unless specifically mentioned.

The antibodies used in immunoblotting and immunofluorescence studies were purchased from different companies, such as H2AX(ab11175, ab22551), γ-H2AX(ab2893), β-actin (ab16039) and GAPDH(ab8245) from Abcam; PARP-1 (sc-1561), 14-3-3ζ(sc-1019) from Santa Cruz Biotechnology; CALR(10292-A-AP), NONO(11058-1-AP), CFL1(10960-1-AP), PEF1(10151-1-AP) and NCL(10556-1-AP) from Protein Tech Group; anti-FLAG(F1860) from Sigma; FITC-conjugated goat anti-rabbit secondary antibody (75230) and TRITC-conjugated goat anti-mouse secondary antibody (61551) from Jackson ImmunoResearch Laboratories, Inc.

The cell line QSG-7701 and QGY-7703 were purchased from Shanghai Institute of Biochemistry and Cell Biology (SIBCB, Shanghai, China). QSG-7701 and QGY-7703 were originally established in QiDong, the highest morbidity area of HCC in China. The hepatoma cell line QGY-7703 was derived from primary carcinoma of a 35 years old female, and its normal liver counterpart QSG-7701 was from peripheral non-tumor tissue of the same surgery specimen.

Plasmids, Stable Cell Lines, and Dual-tagging (FLAG tag and isotope tag) Quantitative Proteomics Approach

Expression plasmids for H2AX were described previously^{37, 38}. The construct containing the FLAG tag alone or FLAG-H2AX was transfected into human HCC cells (QGY-7703) using the Lipofectemin™ 2000 reagent according to the manufacturer, and the cells were selected in RPMI-1640 supplemented with 10% dialyzed FBS, 100U/ml penicillin and streptomycin, and 1.0 μg/ml puromycin. Whereas the cells containing the tag alone (control cells) were cultured in regular unlabeled RPMI-1640 medium, the cells containing FLAG-tagged H2AX (FLAG-H2AX cells) were maintained in RPMI-1640 containing Leu-d₃ to isotopically label the proteome. A detailed procedure has been described previously^{37, 38}.

Histone Isolation

Histone extraction was performed as described previously¹⁵. Briefly, cells were lysed in 0.5% NP40, 100 mM NaCl, 50 mM Tris, 1 mM EDTA. After centrifugation, histones were extracted by resuspending pellets with 0.2 N HCl. The HCl extracts were centrifuged, and the supernatants (histone extracts) were then neutralized with 1 N NaOH and blotted with anti-H2AX or anti-γ-H2AX antibodies.

Protein Extraction and Purification of H2AX Protein Interaction Partners

Protein extraction and purification were performed as described previously with minor modifications³⁸. Briefly, approximately 8×10⁸ FLAG-H2AX cells and control cells were harvested, then equal numbers of each cell line were mixed. The mixed cells were lysed. After nuclear isolation and nuclease digestion, the nuclear extract was incubated with 200 μl of M2 anti-FLAG beads at 4°C for 6 h with 20% glycerol and 150 mM NaCl. The beads were then washed four times with TBS. The bound protein complexes were then solubilized in sample buffer, and separated by 12% SDS-PAGE. After staining with Coomassie Brilliant Blue, visible bands were cut for in-gel digestion. The extracted peptides were lyophilized for LC-MS/MS.

LC-MS/MS Analysis

LC-MS/MS experiments were performed on a LTQ-Orbitrap hybrid mass spectrometer (Thermo Finnigan, Bremen, Germany) equipped with a Finnigan Dynamic nano-spray source. Sample was injected via an SIL-20 AC auto-sampler (Shimadzu Corporation, Tokyo, Japan) and loaded onto a CAPTRAP column (0.5×2 mm, Michrom Bioresources Inc., Auburn, CA) for 5 min at a flow rate of 60μL/min. The peptide mixtures was subsequently separated by a PICOFRIT C18 reverse-phase column (0.075×100 mm, New Objective Inc., Woburn, MA) at a flow rate of 300nL/min with a 110 min-gradient. Buffer A was 5% ACN, 0.1% formic acid; buffer B was 95% ACN, 0.1% formic acid. Samples were loaded in solvent A, and peptides were eluted by 5% solvent B for 5 min followed by a linear gradient to 45% solvent B in the next 90 min, ramping to 95% solvent B in 4 min and dropping to 90% solvent B for 4 min before re-equilibrating the system to 10% solvent B for 7 min. The LTQ-Orbitrap mass spectrometry was operated in the data-dependent mode using the TOP3 strategy. In brief, a scan cycle was initiated with a full scan of mass range (m/z 400–2000) in the Orbitrap under the target mass resolution of 60,000, which was followed by MS/MS scans in the linear ion trap on the 3 most abundant precursor ions. Single charged ions were excluded from MS/MS analysis.

Database Search and Protein Identification

Applying the TurboSequest V.27 (rev.12) search engine, we searched all MS/MS spectra against human reference database (IPI 3.27). Searches were performed with the following parameters: Trypsin enzyme specificity with allowing one missed cleavage; a precursor tolerance of 10 ppm and fragment tolerance of 1 Da; dynamic modifications of d₃-labeled Leu (+3.018). Positive identifications were made according to a rigorous statistical model as Peptide Prophet and reverse database search^{49, 50}. In detail, for SEQUEST engine, all peptides must have a ΔCn of at least 0.1 and cross correlation values (XCorr) were assigned with at least 1.90 (+1 charge), 3.23 (+2 charge) and 3.66(+3 charge) by reverse database search evaluation (p=0.05). Additionally, the computed probability through Peptide Prophet of trans-proteomic pipeline (TPP) tools must be over 0.99 for each peptide⁵¹. Protein assignments were made only if the protein had at least two different peptides passing the above criteria. In order to eliminate redundant protein identifications that matched the same set of peptides (protein group), a parsimonious interpretation method was applied to result in a minimized dataset. Reproducibility was assessed primarily according to the method of Blagoev et al⁵².

Data-dependent bioinformatics for network analysis

Functional classification of identified proteins was accomplished by using Entrez Gene (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db_gene) and DAVID (http:/www.david.abcc.ncifcrf.gov). Proteins with multiple functions were assigned to those that are best known. Pathway module analysis of the functional clusters was performed by STRING (http://www.string.embl.de), a database of known and predicted protein-protein interactions.

Immunoblotting Analysis

Cells (7×10⁷ cells/assay) were harvested for nuclei isolation/digestion, immunoprecipitation, and immunoblotting. The procedures for nuclei isolation/digestion and immunoprecipitation were essentially the same as described as above, except that buffer volumes in each step were reduced ~10-fold. The immunoprecipitated proteins were eluted by boiling the beads and separated on SDS-PAGE. The proteins were then transferred to a PVDF membrane (Bio-Rad) and incubated with the specified primary antibody followed by incubation with a secondary antibody conjugated to horseradish peroxidase. The ECL substrate was then added and the blot was developed.

Immunofluorescence

Cells were fixed in 4% paraformaldehyde for 15 min, washed in PBS, and blocked in 10% normal goat serum with 0.1% Triton X-100 for 1 h at room temperature. The fixed cells were incubated with anti-γ-H2AX (phosphorylated form) antibody overnight at 4°C, followed by incubation with TRITC-conjugated goat anti-mouse secondary antibody for 1 h at room temperature. Cells were washed in PBS. Slides were incubated sequentially with other primary antibodies overnight at 4°C and FITC-conjugated goat anti-rabbit secondary antibody for 1 h at room temperature. After washing, cells were counterstained with Hoechst 33342. Then slides were mounted in glycerol/PBS solution. All primary antibodies were used at dilutions of 1:200 to 1:800. The secondary antibodies included anti-rabbit FITC and anti- mouse TRITC and were used at dilutions of 1:200. Immunofluorescence images were captured using a laser scanning confocal microscope (Zeiss, LSM 510 system or Leica TCS-SP15, Germany).

Bleomycin Stimulation

HCC cells were seeded in 10 cm dishes. When they were 90–95% confluent, the cells were treated with BLM. γ- H2AX was detected with different concentrations of BLM and different times of treatment. The cells treated with optimized conditions (40mU/ml BLM and harvested after 3 hours) were selected for immunoblotting analysis.

Binding Comparative Assay

A stable cell line of hepatocytes (QSG-7701) was produced as described above. Hepatocytes and HCC cells were harvested for nuclei isolation digestion, immunoprecipitation, and immunoblotting analysis.

Results and Discussion

In vivo expression and subcellular location of epitope-tagged H2AX in hepatocellular carcinoma (HCC) cells

To identify H2AX-interacting proteins from HCC cells, we generated a HCC cell line stably expressing human FLAG tagged-H2AX at close to the natural level, using the retroviral gene transfer method which we reported previously^{37, 38}. In the stable cell line, we first examined the subcellular location of the tagged-H2AX by using a sequential histone extraction method¹⁵. Each fraction of the extraction was examined by immunoblotting against H2AX antibody. As shown in Fig.1A, FLAG-tagged H2AX was detected only in the nuclear/histone components, suggesting that the FLAG-tagged H2AX was incorporated into chromatin similar to our previous observations³⁸. Also expression of FLAG-tagged H2AX was found at a level similar to that of endogenous H2AX (Fig.1B, left). The phosphorylated form of FLAG-tagged-H2AX, γ-H2AX, was also detected at an abundance slightly higher than that of its endogenous counterpart (Fig. 1B, middle). Similarly, anti-FLAG was used to detect ectopically expressed ɣ-H2AX and H2AX (Fig. 1B, right).

Figure 1. Expression of FLAG-tagged H2AX in stable QGY-7703 cells.

(A). FLAG-tagged H2AX was detected only in the nuclear fraction, not in the cytoplasmic fraction. Equal amounts of protein (40μg) extracted from cytoplasm and nuclear fraction of cells stably expressing FLAG-tagged H2AX (FLAG-H2AX cells) were loaded onto a 12% SDS-polyacrylamide gel for Western blotting (WB) analysis. GAPDH and endogenous H2AX were detected and used as the control markers of the cytoplasmic and nuclear fractions, respectively. (B). FLAG-tagged H2AX and its phosphorylated form were expressed at a level comparable to that of its endogenous counterpart. Equal amounts of protein (15μg) extracted from the parental 7703 cells and the FLAG-H2AX cells were loaded onto a 15% SDS-polyacrylamide gel for Western blotting (WB) analysis using either anti-H2AX or anti-phospho-H2AX or anti-FLAG antibody. The bands appearing close to 15 or 25 kDa corresponded to the endogenous or FLAG-tagged H2AX, respectively.

Given the fact that H2AX senses DNA DSBs through site-specific phosphorylation⁵³, we expected to observe a portion of γ-H2AX in HCC cells where many un-repairable DSBs may exist. In fact, after several passages the expression level of the epitope-tagged H2AX became even lower than the untagged endogenous form (Table S1). Furthermore, the stable HCC cells expressing the FLAG-tagged H2AX showed a growth rate and morphology similar to the parental cells, indicating that expression of the FLAG-tagged H2AX had no effect on the phenotype of the stable cells. Collectively the above results suggested that the FLAG-tagged H2AX, like its endogenous counterpart, was correctly packaged into nucleosomes and functioned in a physiologically relevant condition in chromatin.

Identification of H2AX-interacting components formed in HCC cells using a dual-tagging (FLAG tag and isotope tag) quantitative proteomics approach

The strategy for affinity purification and identification of H2AX-interacting proteins is, with a few modifications, similar to that described previously³⁸. Briefly, the HCC cell line stably expressing FLAG-H2AX was maintained in a “heavy” medium containing leucine-d₃, whereas the control cells were cultured in the regular (or “light”) medium. Equal amounts of cells from each culture were combined. After nuclei isolation/digestion, extraction and IP, the immunoprecipitated proteins were eluted and separated on SDS-PAGE (Fig 2B). All gel bands were digested with trypsin followed by LC-MS/MS for peptide separation and protein identification. In our MS analysis, all of the leucine-containing peptides appeared as pairs in the MS spectrum with one set of peaks from the stable isotope-labeled cells and the other set of peaks from the unlabeled cells. The intensity ratios of the paired peaks reflect the binding profiles of the protein to the bait protein H2AX. For those H2AX-interacting proteins, the heavy isotope signals should be more intense than their lighter counterparts, suggesting enrichment of these proteins around the H2AX bait. Because background binding occurs randomly in both IP purifications, the ratio of the “heavy” (H) versus “light” (L) peptide intensities (H/L) for bait-nonspecific binding proteins should be 1:1.

Figure 2. AACT/SILAC-based MS analysis of H2AX complexes.

(A) The distribution of H/L ratio of the identified proteins in the H2AX complex formed in HCC cells. The threshold for distinguishing specific interactors from non-specific contaminants was an H/L of 1.34. The sign of diamond, circle and square indicate those proteins with H/L over 1.34, suggesting specificity in associating with H2AX, while a triangle represents a protein with H/L less than 1.34 or a non-specific protein in the complex. The green square represents the bait protein. The red circles indicate proteins known previously to interact with H2AX in other types of cells. The blue diamonds suggest newly identified H2AX-interacting partners. (B) SDS-PAGE analysis of FLAG-immunoprecipitates as visualized by Coomassie Brilliant staining. The band of FLAG-tagged H2AX is indicated by arrow.

The procedure for protein identification and quantitative measurements was carried out similarly to that described previously^{37, 54}. As shown in Figure 2A, for the distribution of H/L for all proteins identified, the proteins previously known to interact with H2AX such as Calreticulin (CALR) were found with H/L ratios > 1.34. Empirically, we selected this value as the threshold for distinguishing specific interactors from non-specific contaminants as all proteins with H/L less than 1.34 were considered as non-specific background. We then searched the available information about the known functional category and the possible H2AX-interacting nature of our identified interactors by using public interaction databases such as oPHID, IntAct, BioGRID, and HPRD. As summarized in Table 1, among 102 proteins distinguished as the H2AX interactors 15 of them are known H2AX interacting partners.

Table 1.

Identified H2AX-interacting Proteins in HCC QGY7703 Cells.

IPI no.	Gene symbol	Protein namea	No.of peptide-matched	Enrichment-foldb	S.D.
IPI00219037.5	H2AX	Histone H2AX (bait)	4	9.11
Apoptosis
IPI00012011.6	CFL1	Cofilin-1	2	1.90	0.014
IPI00291262.3	CLU	Clusterin precursor	2	2.26	0.56
IPI00027230.3	HSP90B1	Endoplasmin precursor	8	1.60	0.07
IPI00304925.4	HSPA1A	Heat shock 70 kDa protein 1	3	1.50	0.16
IPI00555692.2	ANXA4	ANXA4 protein	2	1.49	0.72
IPI00329801.12	ANXA5	Annexin A5	6	1.36	0.11
IPI00025277.5	PDCD6	Programmed cell death protein 6	3	1.745	0.19
IPI00171438.2	TXNDC	Thioredoxin domain-containing protein 5 precursor	2	1.62	0.08
IPI00291467.7	SLC25A6	ADP/ATP translocase 3	2	1.72	0.057
Response to protein stimulus, oxidative stress and hypoxia
IPI00023137.1	TOR1B	Torsin B precursor	2	2.71	0.70
IPI00382470.3	HSP90AA1	Heat shock protein HSP 90-alpha 2	3	2.57	0.19
IPI00301277.1	HSPA1L	Heat shock 70 kDa protein 1L	3	1.45	0.13
IPI00030275.5	TRAP1	Heat shock protein 75 kDa mitochondrial precursor	9	2.45	0.24
IPI00220301.5	PRDX6	Peroxiredoxin-6	3	1.73	0.38
IPI00027192.5	PLOD1	Procollagen-lysine,2-oxoglutarate 5-dioxygenase precursor	2	2.46	0.03
IPI00022314.1	SOD2	Superoxide dismutase90,	2	2.56	0.22
DNA repair (14)
IPI00000690.1	AIFM1	Isoform 1 of Apoptosis-inducing factor 1, mitochondrial precursor	6	2.93	0.37
IPI00304596.3	NONO	Non-POU domain-containing octamer-binding protein	2	1.93	0.02
IPI00449049.5	PARP1	Poly [ADP-ribose] polymerase 1	2	1.81	0.07
IPI00020599.1	CALR	Calreticulin precursor	7	1.34	0.20
IPI00021263.3	YWHAZ	14-3-3 protein zeta/delta	3	1.37	0.11
IPI00000816.1	YWHAE	14-3-3 protein epsilon	3	1.71	0.10
IPI00413293.5	TOR1A	Torsin A precursor	2	1.81	0.56
IPI00000877.1	HYOU1	150 kDa oxygen-regulated protein precursor	4	2.68	0.28
IPI00029744.1	SSBP1	Single-stranded DNA-binding protein, mitochondrial precursor	2	1.74	1.12
IPI00169383.3	PGK1	Phosphoglycerate kinase 1	4	2.58	0.61
IPI00219365.3	MSN	Moesin	2	1.96	0.76
IPI00025491.1	EIF4A1	Eukaryotic initiation factor 4A-I	2	1.41	0.04
IPI00008438.1	RPS10	40S ribosomal protein S10	2	2.14	0.03
IPI00011062.1	CPS1	Isoform 1 of Carbamoyl-phosphate synthase [ammonia], mitochondrial precursor	4	1.44	0.05
DNA metabolism
IPI00005198.2	ILF2	Interleukin enhancer-binding factor 2	3	1.71	0.12
IPI00453473.6	HIST2H4A	Histone H4	5	2.27	0.44
Ca2+ binding signaling (6)
IPI00418169.3	ANXA2	annexin A2 isoform 1	3	1.67	0.27
IPI00334627.3	ANXA2P2	Similar to annexin A2 isoform 1	6	1.84	0.20
IPI00075248.11	CALM2	Calmodulin	6	5.35	0.97
IPI00032313.1	S100A4	Protein S100-A4	2	1.87	0.01
IPI00018235.3	PEF1	Peflin	2	3.08	0.54
IPI00002459.4	ANXA6	annexin VI isoform 2	8	2.19	0.22
Cell cycle
IPI00216318.5	YWHAB	tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, beta polypeptide	2	1.48	0.04
IPI00017297.1	MATR3	Matrin-3	2	2.44	0.02
Alcohol metabolic process
IPI00465439.5	ALDOA	Fructose-bisphosphate aldolase A	2	2.015	0.049
IPI00220644.8	PKM2	pyruvate kinase 3 isoform 2	6	2.45	0.27
IPI00479186.5	PKM2	pyruvate kinase 3 isoform 1	2	2.34	0.17
IPI00465248.5	ENO1	enolase 1	2	1.945	0.16
IPI00058192.1	POFUT1	Isoform 1 of GDP-fucose protein O-fucosyltransferase 1 precursor	3	1.86	0.05
IPI00219217.3	LDHB	L-lactate dehydrogenase B chain	2	2.15	0.19
RNA metabolism
IPI00301936.3	ELAVL1	ELAV-like protein 1	3	1.47	0.22
IPI00028888.1	HNRPD	Isoform 1 of Heterogeneous nuclear ribonucleoprotein D0	3	1.39	0.49
IPI00011274.2	HNRPDL	heterogeneous nuclear ribonucleo protein D-like	2	1.84	0.10
IPI00013877.2	HNRPH3	Isoform 1 of Heterogeneous nuclear ribonucleoprotein H3	2	3.12	0.36
IPI00012074.3	HNRPR	Heterogeneous nuclear ribonucleoprotein R	5	3.26	0.31
IPI00106509.2	HNRPAB	Isoform 4 of Heterogeneous nuclear ribonucleoprotein A/B	2	4.57	0.13
IPI00012066.2	PCBP2	poly(rC)-binding protein 2 isoform b	2	1.86	0.77
IPI00008524.1	PABPC1	Isoform 1 of Polyadenylate-binding protein 1	2	4.18	0.54
IPI00221035.3	BTF3	Isoform 1 of Transcription factor BTF3	2	2.63	0.05
IPI00073713.3	MSI2	Isoform 1 of RNA-binding protein Musashi homolog 2	2	2.81	0.65
IPI00328840.9	THOC4	THO complex subunit 4	3	4.76	0.70
IPI00018140.3	SYNCRIP	Isoform 1 of Heterogeneous nuclear ribonucleoprotein Q	9	2.62	0.35
IPI00215965.2	HNRPA1	heterogeneous nuclear ribonucleoprotein A1 isoform b	2	1.42	0.20
IPI00386854.5	HNRPA2B1	HNRPA2B1 protein	4	2.59	0.17
IPI00216592.2	HNRPC	Isoform C1 of Heterogeneous nuclear ribonucleoproteinsC1/C2	3	1.84	0.04
Establishment of cellular localization
IPI00018931.6	VPS35	Vacuolar protein sorting-associated protein35	3	1.845	0.15
IPI00024911.1	ERP29	Endoplasmic reticulum protein ERp29 precursor	5	1.49	0.16
IPI00156689.3	VAT1	Synaptic vesicle membrane protein VAT-1 homolog	3	2.27	0.45
IPI00021405.3	LMNA	Isoform A of Lamin-A/C	3	1.53	0.06
IPI00045839.3	LEPRE1	Isoform 3 of Prolyl 3-hydroxylase 1 precursor	3	1.35	0.15
IPI00009904.1	PDIA4	Protein disulfide-isomerase A4 precursor	11	1.58	0.08
IPI00020436.4	RAB11B	Ras-related protein Rab-11B	3	1.47	0.60
IPI00023748.3	NACA	Nascent polypeptide-associated complex subunit alpha	2	1.50	0.63
Cellular macromolecule metabolism
IPI00021728.3	EIF2S2	Eukaryotic translation initiation factor 2 subunit 2	2	1.36	0.13
IPI00178440.3	EEF1B2	Elongation factor 1-beta	2	1.52	0.58
IPI00299571.5	PDIA6	Isoform 2 of Protein disulfide-isomerase A6 precursor	9	1.60	0.05
IPI00010720.1	CCT5	T-complex protein 1 subunit epsilon	2	1.57	012
IPI00011229.1	CTSD	Cathepsin D precursor	3	3.61	0.39
IPI00386755.2	ERO1L	ERO1-like protein alpha precursor	2	1.56	0.11
IPI00010796.1	P4HB	Protein disulfide-isomerase precursor	6	1.47	0.42
IPI00011937.1	PRDX4	Peroxiredoxin-4	3	2.19	0.27
IPI00419585.9	PPIA	Peptidyl-prolyl cistrans isomerase A	4	1.44	0.06
IPI00028004.2	PSMB3	Proteasome subunit beta type 3	2	2.26	0.74
IPI00008529.1	RPLP2	60S acidic ribosomal protein P2	2	1.40	0.01
IPI00555956.2	PSMB4	Proteasome subunit beta type 4 precursor	2	1.76	0.43
IPI00303300.3	FKBP10	FK506-binding protein 10 precursor	3	2.39	0.96
IPI00171412.1	SUMF2	Isoform 1 of Sulfatase-modifying factor 2 precursor	2	2.38	0.02
IPI00003128.1	P4HA2	Isoform IIb of Prolyl 4-hydroxylase subunit alpha-2 precursor	3	2.3	0.14
IPI00028635.4	RPN2	Dolichyl-diphosphooligosaccharide—protein glycosyltransferase 63 kDa subunit precursor	7	2.20	0.02
Cell redox homeostasis
IPI00026328.3	TXNDC12	Thioredoxin domain-containing protein 12 precursor	3	1.35	0.04
IPI00303568.3	PTGES2	Prostaglandin E synthase 2	6	1.71	0.34
Carboxylic acid metabolism
IPI00017726.1	HSD17B10	hydroxysteroid (17-beta) dehydrogenase 10isoform 1	5	2.15	0.35
IPI00643920.2	TKT	Transketolase	3	3.52	0.31
Carbohydrate catabolism
IPI00011454.1	GANAB	Isoform 2 of Neutral alpha-glucosidase AB precursor	3	1.44	0.42
IPI00441344.1	GLB1	Beta-galactosidase precursor	2	4.08	0.38
IPI00383046.3	CMBL	carboxymethylenebutenolidase-like	7	4.15	0.29
IPI00026154.2	PRKCSH	Glucosidase 2 subunit beta precursor	4	2.21	1.00
IPI00023673.1	LGALS3BP	Galectin-3-binding protein Precursor	3	3.22	0.29
IPI00027851.1	HEXA	Beta-hexosaminidase alpha chain Precursor	2	3.01	0.16
Others
IPI00020075.4	ABHD10	CDNA FLJ11342 fis, clone PLACE1010800	3	2.10	0.05
IPI00168479.2	APOA1BP	apolipoprotein A-I binding protein precursor	3	3.36	0.22
IPI00550363.3	CCDC19	Transgelin-2	3	1.44	0.67
IPI00329696.1	FAM82B	Protein FAM82B	3	3.08	0.08
IPI00007765.5	HSPA9	Stress-70 protein, mitochondrial precursor	9	1.63	0.74
IPI00024919.3	PRDX3	Thioredoxin-dependent peroxide reductase, mitochondrial precursor	3	2.81	0.34
IPI00377161.7	HIBCH	3-hydroxyisobutyryl-Coenzyme A hydrolase isoform 2	2	4.32	0.63

^aOnly the proteins with two or more peptides matched are listed.

^bFold of enrichment was calculated as the ratio of Leu-d₃-labeled peptide to unlabeled peptide.

Functional categories of the identified H2AX-interacting proteins in HCC cells

By searching Entrez Gene (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db_gene) and DAVID (http://www.david.abcc.ncifcrf.gov), we found that most of the H2AX interacting proteins identified by our proteomic method were clustered in diverse functional categories. As shown in Table 1 and Fig. 3, these functional categories include DNA damage recognition/repair, cell cycle, apoptosis, cellular localization, response to stimulus/oxidative stress/hypoxia, alcohol metabolic processes, DNA and RNA metabolism, cellular macromolecule metabolism, redox homeostasis, Ca²⁺-binding signaling, etc.

Figure 3. The distribution of HCC cell H2AX interacting proteins in different functional categories.

The identified proteins were analyzed by bioinformatics tools using Entrez Gene and KEGG databases. A total of 102 proteins identified as H2AX-interacting partners were found to be associated with fourteen functional categories or biological processes. Proteins with multiple functions were assigned to the best known function.

The distribution of the identified H2AX interactors in these functional categories is given in Figure 3. Using a similar dual-tagging strategy, those proteins including Calmodulin, CALR, PARP-1, histone H4, etc, were also found to interact with H2AX in the human embryonic kidney (HEK) 293T cells³⁸. The identification of these H2AX-interacting partners in both HEK and hepacellular carcinoma cell types suggested there are core components in the H2AX complex regardless of the difference in cell types. Because of their functional relevance to Ca²⁺ signaling-related cell cycle, apoptosis, and DNA damage recognition and repair, etc, the core role of H2AX-interacting complexes in coordinating these biological processes was clearly suggested. In addition we have identified many more H2AX-intercating partners in HCC cells in which some may be HCC-specific and others may be due to the higher sensitivity of LTQ-orbitrap compared to the ABI Qstar/Qtof used previously. The largest portion of the H2AX interactome in HCC cells was represented by proteins involved in macromolecule metabolism including both nucleic acid and protein metabolism, and carbohydrate catabolism. This observation suggested that H2AX may recruit these proteins for readily activating multiple functional pathways such as energy generation, protein synthesis, RNA splicing and transport, translational control, gene regulation, etc^{24, 55–57}. Interestingly, in our dataset (Table 1) many H2AX interactors such as CFL1, HSP90B1 and ANXA5 possess anti-apoptotic functions^{58 59}. A possible explanation is that H2AX may recruit these proteins in a complex to protect the damaged or un-repairable cells from apoptosis in hepacellular carcinoma. In addition, one group of interactors has known functions in alcohol metabolic processes and oxidative stress. Alcohol has long been recognized as a major risk factor for HCC development^{60, 61 62}, and acetaldehyde, the main metabolite of alcohol, was reported to cause hepatocellular injury and is an important factor in causing increased oxidative stress, which damages DNA⁶². Examples of our identified H2AX interactors in this category are enolase 1 (ENO1), Fructose-bisphosphate aldolase A (ALDOA), and Isoform 1 of GDP-fucose protein O-fucosyltransferase 1 precursor (POFUT1), which may take part in alcohol metabolism in liver physiology and pathology.

Data-dependent network analysis of the H2AX-interacting proteins identified in HCC cells reveals that H2AX coordinates cross-talk among diverse pathway modules

To elucidate how H2AX coordinates multiple functional pathways in the tumor environment and the possible functional role of the H2AX interactome in HCC, we explored the inter-connected relationship among the H2AX interactors and further identified each of different pathway modules. By using the STRING mapping tool, the sub-networks operating each pathway module could be mapped and specifically defined with the associated function. As shown in Figure 4, H2AX was found to be the critical network node in a variety of the identified pathway modules associated with apoptosis and cell cycle control (Fig. 4A), DNA repair (Fig. 4B), alcohol metabolism (Fig. 4C), carbohydrate catabolism and cellular macromolecule metabolism (Fig. 4D), stress response (Fig. 4E), and RNA processing (Fig. 4F). Further, given their previously known multi-functional nature, certain H2AX interactors, for instance, YWHAB and YWHAE (also named as 14-3-3 isomers), were found in multiple pathway modules related to apoptosis, cell cycle, and DNA repair (Fig. 4A&4B). EIF4A1 and RPS10 were shown in all pathway modules associating with carbohydrate catabolism, macromolecule metabolism, and DNA repair (Fig. 4B–4D). TOR1A, LMNA, and HYOU1 were shared by both pathway modules related to stress response and DNA repair (Fig.4B&4E). Notably, we have detected some vital proteins such as PARP-1, HSP90AA1, etc, previously known as components of multiple signaling pathways related to DNA damage recognition and repair. Although our network analysis suggested that p53, a key tumor suppressor protein,^{63, 64} could be a critical node in a number of pathway modules (Fig.4), it was not identified in our mass spectrometry experiments. In response to DNA damage, p53 accumulates in differentiated cells, and activates particular target genes that initiate cell cycle arrest or apoptosis⁶³. The p53-regulated tumor suppressor pathway is frequently inactivated in cancer⁶⁵. A possible explanation of our results was that most of the pathway modules identified in our networks analysis could be more p53-independent than p53-dependent in the HCC cells, or both mechanisms may co-exist. PARP-1 is multi-functional in many pathways including those related to DNA damage/repair^{3, 66, 67}. HSP90AA1 is a molecular chaperone involved in both formation and maintenance of proper conformation of proteins and promoting cell survival in response to stress^{26, 68}.

Figure 4. The H2AX-associating pathway modules found in the HCC-specific H2AX interactome.

The regulatory/signaling pathway modules of identified proteins were analyzed by DAVID and STRING 8.1 (http://www.string.embl.de/). (A) apoptosis and cell cycle, (B) DNA repair, (C) alcohol metabolism, (D) carbohydrate catabolism and cellular macromolecule metabolism, (E) stress response, (F) RNA processing and protein synthesis.

This data-dependent network analysis not only provided further support for the accuracy of our determination of the H2AX-interacting partners but also the binary relationships among these interactors, as most of them could be identified in the corresponding pathway module(s). More importantly, H2AX was shown to promote cross-talk in these interconnected pathway modules. Clearly, these pathway modules coordinated by H2AX may sense intrinsic stress such as un-repaired DNA DSBs in HCC cells, transduce the signal in recognizing DNA damage, and selectively activate biological processes such as apoptosis, cell cycle arrest, or DNA repair.

Validation of the HCC-specific H2AX interactome dataset using immunoblotting

We used co-immunoprecipitation (co-IP) and concurrent immunoblotting to evaluate the physiologically relevant accuracy of our proteomics dataset of the HCC cell H2AX-interacting complex. In our selected pool of H2AX interactors, proteins such as PARP-1, CALR, non-pou domain-containing octamer-binding protein (NONO), 14-3-3ζ, CFL1, and PEF1 were known to be involved in individual or multiple pathway modules associated with DNA damage recognition/repair, apoptosis, and cell cycle control respectively and could represent each of individual pathway modules. The known H2AX interactors such as PARP-1, CALR and NONO were found to interact with H2AX in HCC cells while other proteins previously unknown to interact with H2AX such as 14-3-3ζ, CFL1 and PEF1 were also identified as H2AX-inetracting partners in HCC cells.

Poly(ADP-ribose) polymerases, or PARPs, which showed 1.8 fold enrichment in the H2AX complex (Table 1) are a class of cell signaling enzymes functioning in poly(ADP-ribosylation) of DNA-binding proteins⁶⁹. PARP-1 is best known to be involved in the cellular response to DNA damage^{69, 70} in addition to its associations with various pathways including DNA replication, DNA repair, recombination, gene transcription, cell proliferation and death⁷¹. PARP-1 participates in DNA DSB repair through either nonhomologous end-joining (NHEJ) or homologous recombination (HR)³, and is believed to be a `sensor' in detecting DNA damage and initiating DNA repair⁷². Earlier reports suggested the increased level of PARP-1 in cancer cells might promote their survival^{72, 73}. PARP-1 inhibitors have been used in cancer chemotherapy in combination with anti-tumor drugs having DNA-binding properties^{74, 75}. A previous study also indicated that PARP-1 expression was significantly increased in human HCC compared to its surrounding liver tissue^{76, 77}. Also PARP-1 was found to contribute to HBV-related hepatocarcinogenesis⁷⁸. Our study here has provided direct evidence suggesting that the role of PARP-1 in HCC might be due to its interaction with H2AX in the HCC cells.

14-3-3ζ is one of the isoforms in the 14-3-3 family of regulators of diverse cellular responses in both physiological and pathological conditions⁷⁹. 14-3-3ζ is an important regulatory protein in intracellular signaling pathways and is known to interact with more than 100 cellular proteins, including many oncogene and proto-oncogene products⁸⁰. 14-3-3ζ blocks apoptosis by inhibiting activation of p38 mitogen-activated protein kinase (MAPK) and could be an anti-apoptotic factor in cells⁸¹. 14-3-3ζ was also reported to interact with β-catenin, and to enhance or to inhibit β-catenin-dependent transcription. It also facilitates activation of β-catenin through Akt, and is possibly involved in stem cell development⁸². Furthermore, 14-3-3ζ is up-regulated in a number of cancer cell types^{83, 84}. Evidence indicates that 14-3-3ζ plays a role in keratin filament organization in hepatocytes, and regulates mitotic progression^{66, 85}. In our analysis (Table 1), 14-3-3ζ was identified as an intrinsic interactor of H2AX in HCC cells.

Cofilin-1 (CFL1) is a small ubiquitous protein that binds both monomeric and filamentous actin, and, through its ability to sever actin filaments, is an essential regulator of actin dynamics at the plasma membrane during cell migration⁸⁶. Aberrant regulation of cell migration drives the progression of many diseases, including cancer invasion and metastasis⁸⁷. CFL1 may be involved in cancer cell migration, invasion, and metastasis^88–90. In the highly metastatic human HCC cells such as MHCC97-H⁹¹, a decreased level of CFL1 was found. Furthermore, some studies suggest CFL1 may relate to radio-sensitivity by altering DNA repair capacity⁸⁷ since key components for repair of DNA DSBs, including RAD51, RAD52, and Ku70/Ku80, were down-regulated in CFL1 over-expressing cells following ionizing radiation⁸⁷. Interestingly, CFL1 was identified with 1.9 fold increase in its binding to H2AX in HCC cells.

One of the penta-EF-hand (PEF) proteins, PEF1, is a Ca²⁺-binding protein⁹² functioning in Ca²⁺-mediated signaling irrespective of cell type⁹³. PEF1 dimerizes with apoptosis-linked gene 2 (ALG-2) and modulates the function of ALG-2 in Ca²⁺ signaling⁹². Our previous study in HEK 293 cells showed that the H2AX interactome plays a important role in ionizing radiation-induced cell cycle arrest through Ca²⁺ signaling pathways³⁸.

Similar to the design of validation experiments³⁸, the immunoblotting assay using corresponding protein antibodies was performed on the same amount of the immunoprecipitates derived from FLAG-alone HCC cells (parental) and FLAG-H2AX stable HCC cells, respectively. H2AX-specific interactions for the selected proteins found in HCC cells are shown in Fig. 5, and are consistent with our quantitative proteomic dataset.

Figure 5. Immunochemical analyses of selected proteins identified as H2AX-interacting partners in HCC cells.

The immunoprecipitated proteins from FLAG-QGY-7703 cells and FLAG-H2AX-QGY-7703 cells were analyzed by immunoblotting (IB). Anti-FLAG antibody was used to detect H2AX with nuleolin (NCL) as a loading control. Lane 1: the QGY-7703 cells expressing FLAG alone; lane 2: the QGY-7703 cells stably expressing FLAG-H2AX.

In situ localization of endogenous H2AX-specific interactions using confocal laser scanning

With the positive control of the known H2AX-PARP1 interaction, by using confocal scanning, we further examined the endogenous interactions of H2AX with its newly identified partners such as 14-3-3ζ, CFL1, and PEF1 in `resting' HCC cells. As shown in Figure 6, endogenous phosphorylated (γ)-H2AX stained with red fluorescence was observed in the nucleus. Stained with green fluorescence, either 14-3-3ζ, CFL1, PEF1, or PARP-1 were also found in the nucleus (in blue). Although in a small population, γ-H2AX was found to co-localize with either PARP-1, or 14-3-3ζ, or CFL1 or PEF1 in the nucleus of HCC cells as seen by the bright merged spots (arrowhead in Fig 6). As a result, for the first time, the interactions of H2AX with 14-3-3ζ, CFL1, PEF1 and PARP-1 were observed at the endogenous level in HCC cells based on both proteomic and immunoassay/confocal results.

Figure 6. Confocal laser scanning analysis of co-localized endogenous interactions between γ-H2AX and either 14-3-3ζ, or PEF1, or PARP-1 or CFL1 respectively using.

The parental QGY7703 cells were stained with TRITC for γ-H2AX (red), and with FITC for either PARP-1, or 14-3-3ζ, or PEF1 or CFL1 respectively (green). The nucleus was stained with Hoechst 33342 (blue). With a scale bar of 10 μm the merged images are indicated by white foci. All foci are indicated by white arrows. The confocal images of co-localized (A). γ-H2AX and PARP-1, (B). γ-H2AX and 14-3-3 ζ, (C). γ-H2AX and PEF1, and (D). γ-H2AX and CFL1.

The correspondence between methods indicates the accuracy and reliability of our dual-tagging quantitative method for profiling bait-specific interacting partners. The results also suggest that these H2AX interactors function in DNA damage recognition/repair, apoptosis, and cell cycle control through their interactions with H2AX or mediated by H2AX.

H2AX recruits particular interacting proteins in response to bleomycin (BLM)-induced DNA damage in HCC cells

To understand the functional role of the HCC-specific H2AX-interacting complexes in diverse functions of DNA damage recognition and repair, apoptosis, and cell cycle control in response to DNA DSBs, we next studied the changes in H2AX binding by the interactors representing each functional category. Bleomycin (BLM) was used to trigger the response to instant DNA DSBs^{47, 94}. Because H2AX can sense the extent of BLM-induced DNA DSBs, we first examined both dose- and time course-dependent BLM-induced phosphorylation at H2AX to determine the condition of maximum effect of BLM. As shown in Fig.7A, the abundance of phosphorylated H2AX or γ-H2AX increased along with BLM treatment at a defined dose of 40 mg/ml for an extended period of time until a maximum level was reached after approximately 3 hours. Under this condition, the H2AX-interacting complex was first pulled down through immunoprecipitation using anti-FLAG beads. Using the same approach, the immunoprecipitate was obtained from non-stimulated HCC cells stably expressing FLAG-tagged H2AX. Using the antibodies against each of the H2AX interactors such as PARP-1, 14-3-3ζ, CFL1 and PEF1, immunoblotting experiments were performed on equal amounts of the immunoprecipitates derived from BLM-stimulated versus non-stimulated HCC cells. In these co-immunoprecipitation and immunoblotting experiments, the amount of FLAG tag detected was used as the control. As shown in Figure 7B, the abundance of PARP-1, CFL1, PEF1, or 14-3-3 ζ in the H2AX-interacting complex increased proportionally along with the BLM-induced increase of γ-H2AX (Table S2), suggesting that these proteins were recruited through interaction with H2AX to the complex that participates in the processes of recognizing BLM-induced DNA DSBs or DNA damage.

Figure 7. Immunoblotting analysis of bleomycin (BLM)-induced changes in the H2AX binding to its interacting partners.

(A) Time course-dependent BLM-induced phosphorylation of H2AX in HCC cells, and (B) Immunoprecipitates isolated from either untreated or BLM-stimulated cells (40 U/ml for 3 hours), respectively, were analyzed by immunoblotting (IB). The FLAG tag was detected and used as the loading control. Lane 1: from the untreated FLAG-H2AX-QGY-7703 cells; lane 2: from the BLM-stimulated FLAG-H2AX-QGY-7703 cells.

In addition to its known role in DNA repair, with still largely unknown mechanism γ-H2AX may participate in regulation of a variety of biological processes such as cell cycle, chromatin remodeling, etc^{9, 95}. For example, two distinct γ-H2AX foci were found in non-irradiated or `resting' cells⁹⁶. One predominant population of small foci did not co-localize with DSB repairing proteins, while another small population of large foci co-localized with many repair proteins and resembled IR-induced foci⁹⁶. It has also been shown that H2AX can be phosphorylated independently of DNA DSBs, and that this phosphorylation could be involved in regulating the cell cycle independently of sensing DNA damage or promoting repair^{96, 97}. Also H2AX may function to promote chromatin remodeling⁹⁸.

Our results for BLM-inducible γ-H2AX interactions in HCC cells revealed that upon BLM-induced DSBs, γ-H2AX could interact with those proteins representing diverse functional categories of DNA DSB recognition such as PARP-1, the cytoskeleton such as CFL1, Ca²⁺ signaling such as PEF1, etc. The BLM-induced composition of γ-H2AX foci provides an insight into the molecular mechanisms underlying how γ-H2AX turnover keeps cell cycle checkpoints active until any DNA damage is repaired.

Comparative analysis of the H2AX interacting partners in paired hepatocytesversus HCC cells

On the basis of the profiling dataset of the H2AX interacting partners, we selected certain core components identified in the complex to compare their relative binding affinities to H2AX in the paired hepatocytes and HCC cells. These comparative studies were performed on two cell lines derived from healthy liver and carcinoma tissue of the same donor with the same genetic background, i.e., QSG-7701 for normal and QGY-7703 for HCC respectively⁴⁸. We also generated the QSG-7701 or QGY-7703 cell lines expressing only FLAG (parental) and FLAG-tagged H2AX respectively. Co-immunoprecipitation and immunoblotting experiments were performed on the pair of QSG-7701 and QGY-7703 both stably expressing FLAG-tagged H2AX in contrast to their corresponding parental cells. As shown in Figure 8, with respect to the abundance of FLAG tag in the immunoprecipitates as the loading control, the expression of γ-H2AX did not show a significant difference between these cell lines. Measured by Fujifilm Las-3000 Luminescent image analyzer, the H2AX interacting partners such as PARP-1, 14-3-3ζ, and CFL1 displayed enhanced binding to H2AX in QGY-7703 HCC cells compared to that in normal QSG-7701 hepatocytes (Table S3). PARP-1 was among those interactors showing a relatively stronger enhancement for its binding to H2AX in HCC cells.

Figure 8. Comparative analysis of the H2AX binding with selected protein partners in paired hepatocytes (QSG-7701)versusHCC (QGY-7703) cells.

Using the antibodies as indicated in the Figure, immunoprecipitates were obtained from QSG-7701 cells expressing FLAG alone, QSG-7701 cells expressing FLAG-H2AX, the QGY-7703 cells expressing FLAG alone and QGY-7703 cells expressing FLAG-H2AX respectively. In immunoblotting (IB) analysis, the FLAG tag was detected and used as the loading control. Lane1: from the FLAG-QSG-7701 cells; lane2: from the FLAG-H2AX-QSG-7701 cells; lane3: from the FLAG-QGY-7703 cells; lane4: from the FLAG-H2AX-QGY-7703 cells.

Defects in the DNA damage recognition and repair pathways contribute to genome instability and promote tumorigenesis³. It has been documented that the processes of DNA damage response and repair remain active during tumourigenesis^{31, 99}, probably due to stress-inducing DNA damage in the rapidly dividing pre-neoplastic lesions^{31, 99 100, 101}. Improper DNA damage recognition and DNA repair could contribute to the survival and apoptotic resistance of cancer cells. Clearly, our observations support these notions at the molecular level with a systems view, suggesting that not only the phosphorylated H2AX but its interactors with characteristic binding strength are responsive to intrinsic DSBs/DNA damage in tumor cells.

Conclusions

Using a dual-tagging quantitative proteomic approach we have dissected a cell type-specific H2AX interactome. The HCC profile of the H2AX-interacting partners identified by our approach first suggested a multi-functional role of H2AX in mediating many biological pathways/processes such as DNA damage recognition, DNA repair, apoptosis, cell cycle, protein metabolism, cellular localization, etc. Network analysis further indicated the cross-talk among functional clusters. The physiologically relevant accuracy of these interactions was validated by immunoassys and the cellular localization of these interactions at the endogenous level was determined by confocal scanning microscopy. The stimulation of HCC by a DNA DSB-inducing agent, BLM, allowed us to determine the role of those interactions in the cellular response to DNA damage. The differential strengths of H2AX in recruiting its interactors were studied in paired hepatocyte and HCC cells, suggesting that H2AX mediates the differential cellular response in carcinogenesis through interacting with different strengths with its partners. Our results of dissecting HCC-characteristic H2AX interactome have provided insight into possible markers indicative of the pathological status of HCC.

Abbreviations

DSBs

double-strand breaks

AACT

amino acid-coded tagging

SILAC

stable isotope labeling with amino acid in cell culture

14-3-3ζ

14-3-3zeta

CFL1

coflin1

PEF1

peflin1

PARP-1

Poly(ADP-ribose) polymerase1

HCC

hepatocellular carcinoma

γ-H2AX

phosphorylation H2AX

MS

mass spectrometry

BLM

bleomycin

CALR

Calreticulin

NONO

non-pou domain-containing octamer-binding protein

NCL

nucleolin

MRN

MRE11-NBS1-RAD50

UIM

ubiquitin-interaction motif

ATM

ataxia telangiectasia mutated

MDC1

DNA damage checkpoint 1

SDT

Ser-Asp-Thr

Co-IP

co-immunoprecipitation

IB

immunoblotting

WB

Western blotting

FITC

fluorescein isothiocyannate

ECL

electrochemiluminescence

HPLC

high performance liquid chromatography

MS/MS

tandem mass spectrometry

LTQ-Orbitrap

linear trap quadrupole-Orbitrap