p38-MAPK/MSK1-mediated overexpression of histone H3 serine 10 phosphorylation defines distance-dependent prognostic value of negative resection margin in gastric cancer

Tissue samples and histopathological analysis

Frozen tissue sections (n = 36) and formalin-fixed paraffin-embedded (FFPE) tissue blocks (n = 115) of the tumor (T), PRM, and DRM from each gastric adenocarcinoma patients were obtained. All the patients were operated between 2009 and 2012 at Tata Memorial Hospital, Mumbai, India. Histopathological analysis including determination of tumor content (% of tumor cells) was done by a blinded gastrointestinal pathologist. Based on the histopathological analysis, frozen tumor tissues with ≥70 % tumor content, FFPE tumor tissues with ≥10 % tumor content, and negative resection margins (without any tumor cell) were included in the study. Finally, the study was conducted on paired tumor, PRM, and DRM frozen tissues (n = 10), and FFPE tissue blocks (n = 101). The protocol of the study was reviewed and approved by the Institutional Review Board and Ethics Committee (project number-466). All patients provided a written informed consent.

Immunohistochemistry

Immunohistochemical (IHC) staining was performed using VECTASTAIN® ABC kit (Vector Lab-P6200). FFPE tissue blocks were sectioned at a thickness of 4 μm and mounted on poly-l-lysine-coated glass slides. The sections were deparaffinized through a graded series of xylene and rehydrated through a graded series of absolute ethanol to distilled water. Endogenous peroxidase was quenched with 3 % hydrogen peroxide in methanol at room temperature for 30 min in the dark. Microwave antigen retrieval was carried out with 0.01 M Sodium citrate buffer (pH 6). Anti H3S10ph (Abcam-1776), H3K16ac (Millipore-07-329), H4K20me3 (Abcam-9053), and ph-MSK1 (Abcam-32190) antibodies were applied for 16 h at 4 °C at the dilution of 1:100. Immunoreactive proteins were chromogenically detected with diaminobenzidine (DAB; Sigma-D5537). The sections were counterstained with Harris’s hematoxylin, dehydrated, and mounted. In parallel, control staining was performed without adding a primary antibody.

Evaluation of immunohistochemistry

The nuclear IHC staining for all the antibodies were scored using H-score which is based on intensity of staining (ranges from zero to three) and percentage of stained cells using the following formula: H-score = [(0 × % of cells with staining intensity of zero) + (1 × % of cells with staining intensity of one) + (2 × % of cells with staining intensity one) + (3 × % of cells with staining intensity two)]. H-score was further divided into three groups: (i) 0–100: low level, (ii) 100–200: intermediate level, and (iii) 200–300: high level. The IHC staining was examined by two independent researchers (SAK and MR), one of whom is a senior consultant pathologist (MR). Both the researchers were blinded to all clinicopathological and outcome variables.

Cell culture, transfection, and treatment

GC cell lines, AGS (CRL-1739) and KATOIII (HTB-103), were procured from ATCC. AGS and KATOIII were cultured in RPMI1640 (Invitrogen) and F12K (Himedia) media, respectively, at 37 °C with 5 % CO₂ supplemented with 10 % FBS, 100 U/ml penicillin, and 100 mg/ml streptomycin (Sigma). AGS cells were transfected using CaCl₂ method with 10 μg of the pCMV-flag-MSK1 construct. AGS and KATOIII cells were treated for 1 h with chemical inhibitors against ph-p38 (SB203580, Calbiochem), ph-ERK (PD98059, Calbiochem), and ph-MSK1 (H89, Millipore) at 10 and 20 μM concentrations for 1 h, respectively.

Total RNA isolation and RT-PCR

Total RNA was extracted (Thermo scientific-0731) from 25 mg of the frozen tumor and resection margin (PRM or DRM) tissues with the maximum distance from the site of the tumor. Total RNA (1 μg) was used for cDNA synthesis (Fermentas-K1632) using random hexamers. RT-PCR amplification was done using specific primers for c-Jun (F:CCCCAAGATCCTGAAACAGA, R:TCCTGCTCATCTGTCACGTT), c-Fos (F:CGGGGATAGCCTCTCTTAC, R:CCCTTCGGATTCTCCTTTTC), cyclin-E1 (F:AGCGGTAAGAAGCAGAGCAG, R:TTTGATGCCATCCACA GAAA), cyclin-B1 (F:CGGGAAGTCACTGGAAACAT, R:CCGACCCAGACCAAAGTTTA), cyclin-D1 (F:GATCAAGTGTGACCCGGACT, R:AGAGATGGAAGGGGGAAAGA), and 18S rRNA (F:AAACGGCTACCACATCCAAG, R:CCTCCAATGGATCCTCGTTA) with an initial denaturation step at 95 °C for 2 min, followed by 15 cycles of denaturation at 95 °C for 45 s, primer annealing at 55 °C for 30 s, primer extension at 72 °C for 30 s, and a final extension at 72 °C for 10 min. Amplified products were resolved on 1.5 % agarose gels and visualized by EtBr staining.

Preparation of total cell lysate, nucleo-cytosolic fraction, and histones

Cells were harvested from 90-mm plates and lysed in 200 μl of Laemmli buffer to prepare the total cell lysate (TCL). Nucleo-cytosolic fraction (NCF) was prepared by homogenizing 100 mg of frozen tissue in 2 ml of ice-cold lysis buffer (20 mM Tris-Cl pH 8, 2 mM EDTA pH 8, 10 mM EGTA, 5 mM MgCl₂, 0.1 % TritonX-100, 1 mM sodium orthovanadate, 1 mM sodium fluoride, 20 mM β-glycerophosphate, 1 mM DTT, 1 mM PMSF, 10 μg/ml leupeptin, 10 μg/ml aprotinin), and for cell lines, 2 ml of ice-cold MKK lysis buffer (10 mM Tris-Cl, 1 mM EDTA, 1 mM EGTA, 1 % Triton X-100, 10 μg/ml leupeptin, 10 μg/ml aprotinin, 1 mM PMSF, 1 mM sodium orthovanadate, 10 mM sodium fluoride, 10 mM β-glycerophosphate) [10] was used for cells harvested from a 90-mm plate. Both the homogenates were then kept at 4 °C for 30 min with intermittent mixing and then was clarified by centrifugation at 16,000 rpm for 30 min. The supernatant was collected as NCF and stored at −20 °C until it is required, and the pellet was used for histone isolation by acid extraction method [11]. For the preparation of NCF and histones, along with tumor, resection margin (PRM or DRM) tissues with the maximum distance from the site of the tumor were used.

Electrophoresis and immunoblotting

TCL and NCF, and histones were resolved on 10 and 18 % polyacrylamide SDS-PAGE, respectively, and transferred to PVDF membrane. Proteins on PVDF membrane were hybridized with anti-H3 (Upstate-06-755; 1:2000 dilution), H4 (Millipore-07-108; 1:4000 dilution), H3S10ph (Millipore-06-570; 1:7000 dilution), H4K16ac (Millipore-07-329; 1:8000 dilution), H4K20me3 (Abcam-9053; 1:4000 dilution), β-actin (Sigma-A5316; 1:10,000 dilution), MSK1 (Santacruz-9392; 1:2000 dilution), ph-MSK1 (Abcam-31190; 1:3000 dilution), ERK1/2 (Santacruz-292838; 1:2000 dilution), ph-ERK (Cell signaling-9910; 1:2000 dilution), p38 (Santacruz-728; 1:2000 dilution), ph-p38 (Cell signaling-9910; 1:2000 dilution), and anti-flag (Sigma-F3165; 1:5000 dilution). Signal was visualized using horseradish peroxidase-conjugated anti-rabbit/mouse secondary antibody and ECL plus chemiluminescence kit (Amersham). Wherever required, the densitometry analysis was done on immunoblot and membrane to determine their mean intensities using ImageJ software. For native proteins, mean intensity of immunoblot was normalized with the stained PVDF membrane; for phosphorylated forms, mean intensity of immunoblot was normalized with immunoblot of native proteins. The resulted value was used to express their mean relative levels in resection margin and tumor.

Immunofluorescence microscopy

Cells grown on glass coverslips were fixed with 4 % paraformaldehyde for 20 min. Cells were then permeabilized in PBS containing 0.5 % Triton X-100 for 20 min at RT and then blocked with PBS containing 3 % BSA and 0.1 % NP-40 for 1 h. Next, cells were incubated with a primary antibody against H3S10ph (Millipore-06-570) and ph-MSK1 (Abcam-31190) and appropriate secondary antibodies for 2 h each. Dilution of primary (1:100) and secondary antibodies (Alexa-568 or Alexa-488; Cell signaling; 1:300 dilution) was made in blocking buffer. After the addition of secondary antibody, all the steps were performed in the dark and at room temperature. Finally, coverslips were mounted in VECTASHIELD (Vector lab). Fluorescence intensity was analyzed using a fluorescence microscope (IX81; Olympus, Japan).

Cell cycle analysis

Frozen tissue (50 mg) was powdered using mortar and pestle in liquid nitrogen. The powder was homogenized in 2 ml of nuclear buffer A (15 mM Tris-Cl pH 7.5, 60 mM KCl, 15 mM NaCl, 2 mM EDTA, 0.5 mM EGTA, 0.34 M sucrose, 0.15 mM β-ME, 0.15 mM spermine, and 0.5 mM spermidine) using a Dounce homogenizer. In the case of cell lines, cells were trypsinized and harvested in PBS. Tissue homogenate and cells were then centrifuged (5000 rpm for 15 min at 4 °C) to pellet nuclei and cells, respectively; the supernatant was discarded. The pellet was then fixed with 70 % chilled ethanol and stored at −20 °C. When required, tissue nuclei/cells were washed with PBS and suspended in 500 μl of PBS containing 0.1 % TritonX-100 and 100 μg/ml of RNaseA and incubated at 37 °C for 30 min. After incubation, propidium iodide (25 μg/ml) was added and nuclei were incubated at 37 °C for 30 min. DNA content analysis was carried out in a FACSCalibur flow cytometer (BD Biosciences). Cell cycle analysis was performed using the Mod-fit software.

Mitotic index analysis

Based on the morphology of nuclei, mitotic cells were counted in 10 consecutive high power fields (HPF; ×40) of H&E-stained tissue (tumor and resection margin) section slides. Average of mitotic cells from 10 HPF was represented as mitotic index. H&E staining was done as per the standard protocol on 4-μm-thick FFPE tissue sections.

Statistical analysis

To test the statistical significance of paired and unpaired resection margin and tumor tissues, Wilcoxon matched pair and Mann-Whitney test were used, respectively. To test whether variables differed across groups, we used the chi-square test. Chi-square analysis was used to find the correlation between H3S10ph levels of the tumor, PRM, and DRM tissues and clinical parameters. Survival curves were plotted using the Kaplan-Meier method, and the significance of the differences between the survival curves was determined using a univariate log-rank test. To test the statistical independence and significance of predictors, a multivariate survival analysis was performed using the Cox proportional hazard regression model. All p values were two-sided, and p < 0.05 was considered significant. All statistical analyses were performed with GraphPad and/or SPSS software.

Level of H3S10ph in tumor and resection margin tissues of GC

Histones were prepared from freshly resected paired (n = 10) tumor and R0 resection margin (RM) tissues of GC patients, for a pilot study. Histones and their respective paraffin blocks were subjected to immunoblotting and IHC analysis with site-specific acetylation, methylation, and phosphorylation marks of histone H3 and H4 (data not shown). H3S10ph showed a most consistent (9/10 patients) increase in tumor compared to resection margin tissues in immunoblot analysis (Fig. 1a). Further, the loss of H4K16ac and H4K20me3 is a hallmark of the tumor [7]; however, it was not reported in paired GC tissue samples. Our immunoblot and IHC analysis in the paired tissues confirmed the decrease of H4K16ac (8/10 patients) and H4K20me3 (8/10 patients) in GC as well (Fig. 1a, b and ​Additional file 1: Figure S1).

The status of H3S10ph was further studied in a validation set (n = 101) among tumor and histopathologically negative PRM and DRM tissues using IHC. IHC analysis showed a high level of H3S10ph in tumor compared to both the resection margins (Fig. 1c). H-score-based analysis of frequency distribution of tumor and PRM and DRM tissue samples of GC patients showed 76, 57, and 42; 19, 40, and 44; and 6, 4, and 15 samples with a low, intermediate and high level of H3S10ph, respectively (Fig. 1d). Further, comparison of H-score showed a significant high level of H3S10ph in tumor compared to PRM (p < 0.001) and DRM (p < 0.001) tissues (Fig. 1e).

Correlation of H3S10ph levels of tumor, PRM, and DRM with clinicopathological variable of GC patients

Correlation of H3S10ph levels of tumor and resection margins with survival of GC patients

Overall survival (OS) and disease-free survival (DFS) rate among groups with the low, intermediate, and high level of H3S10ph was compared by log-rank test/Kaplan-Meier survival analysis (Fig. 2). Analysis showed a significant negative correlation of H3S10ph levels of tumor (p = 0.004 and 0.011), PRM (p = 0.014 and 0.004), and DRM (p = 0.026 and 0.006) with OS and DFS, respectively (Fig. 1a–c). Moreover, H3S10ph levels of the tumor, but not the PRM and DRM, were found to be independent predictors of overall survival (Additional file 2: Table S1). Therefore, together, data of this and previous sections confirm the association of a high level of H3S10ph of resection margins along with tumor tissues with poor prognosis of GC.

Relation of H3S10ph levels of resection margins and their distance from the site of the tumor in GC

Our observation of the low level of H3S10ph in resection margin compared to tumor tissues led us to examine whether the decrease had any relation with the distance of resection margin from the site of the tumor. To answer this, we first grouped the resection margin samples as per their distance from the tumor site and compared the mean H-score of H3S10ph immunostaining of each group with the mean H-score of tumor samples (Fig. 3). For both PRM and DRM, a significant reduction in HS10ph (p < 0.05) was observed for patient’s group with resection margin distance >4 cm (Fig. 3a, b, left panel). Further, patients were divided into two groups based on the distance of resection margin ≤4 or >4 cm and their mean H-scores were compared with the tumor tissues. Interestingly, further analysis showed H3S10ph levels of resection margins with the distance ≤4 cm were almost equal to those of the tumor tissues; however, resection margins with >4 cm showed a significant (p < 0.001) reduction in both PRM and DRM (Fig. 3a, b, middle panel). Additionally, immunoblot analysis also confirmed the reduction of H3S10ph levels of resection margins if the distance is >4 cm from the site of the tumor (Fig. 3a, b, right panel).

Effect of resection margin distance on prognostic value of H3S10ph in GC

In the case of OS, patients with PRM ≤4 cm showed a significant (p = 0.003) difference among the group of high, intermediate, and low levels of H3S10ph (Additional file 3: Figure S2A) and no difference was observed in the case of DRM (Additional file 2: Figure S1C). However, in the case of DFS, distance seems to have no effect as patients with both ≤4 or >4 cm resection showed significant difference in survival among the group of high, intermediate, and low levels of H3S10ph for both PRM (p = 0.028 vs 0.006) and DRM (p = 0.041 vs 0.005). Moreover, when the patients were grouped just based on the distance of the resection margin from the site of the tumor (≤4 or >4 cm), no significant difference was observed in the case of both OS and DFS (Additional file 4: Figure S3).

Taken together, these data indicate that the distance of resection margin is an important factor in GC prognosis and H3S10ph could be a potential biomarker in predicting the association between distance of resection margin and clinical parameters. However, H3S10ph cannot be used to predict the survival difference based on the distance of the resection margin for both PRM and DRM.

Association of an increase of H3S10ph with the phase of cell cycle in GC

H3S10ph is a very dynamic histone mark and its level changes throughout the cell cycle with the highest level in mitotic phase and the lowest level in the interphase of the cell cycle [12, 13]. Therefore, to determine whether the increase of H3S10ph in GC is dependent on the cell cycle profile of the tissues samples, cyclin levels, mitotic index, and cell cycle profile of the tumor and resection margin tissues were studied (Fig. 4). Cyclin B1, D1, and E1 levels are known to peak at the time of G2/M phase transition, mid-S phase, and G1/S phase transition, respectively. RT-PCR analysis showed the increase in the messenger RNA (mRNA) levels of all the cyclins in tumor than the resection margin tissues; however, no changes were observed in their ratios between the tumor and resection margin tissues (Fig. 4a). The mitotic index also did not show any significant increase in mitotic cells in tumor compared to resection margin tissues (Fig. 4b). Flow cytometry-based cell cycle analysis of tissue samples showed an equal percentage of G1, S, and G2/M cells in the tumor and resection margin tissues with a maximum population of cells in the G1 phase (Fig. 4c). These results indicate that the observed increase of H3S10ph in GC is not because of the enrichment of cells in any cell cycle phase in tumor compared to resection margin tissues.

In the mitotic phase, H3S10ph is associated with chromatin condensation and transcription silencing while in the interphase of cell cycle an increase of H3S10ph is associated with chromatin relaxation and transcription upregulation of mainly immediate early (IE) genes [12, 13]. Cell cycle analysis revealed about 80 % cells of the tumor and resection margin tissues were in G1 phase (Fig. 4c). Therefore, to determine whether the increase in H3S10ph in GC is an interphase-associated phenomenon or not, we checked the levels of IE genes (c-jun and c-fos) using RT-PCR and immunoblotting. The data showed an increase in the levels of c-jun and c-fos in tumor compared to resection margin tissues (Fig. 4d). Therefore, taken together, these data confirm that increase in H3S10ph levels in GC is not due to the alteration in the cell cycle phase, but an interphase-associated phenomenon.

MSK1 phosphorylates H3S10 through p38-MAPK pathway in GC

Several kinases are known to phosphorylate H3S10 [12]; however, only mitogen- and stress-activated protein kinase-1 (MSK1)-mediated phosphorylation of H3S10 is known to be involved in cellular transformation [11] which is activated through p38 and/or ERK1/2 MAP kinase pathway [14]. In addition, overexpression of c-jun and c-fos as observed in our experiments (Fig. 4d) has also been linked to MSK1-mediated phosphorylation of H3S10 at their promoters [15]. Therefore, ph-MSK1, ph-p38, and ph-ERK1/2 levels in tumor and resection margin tissues of GC patients were analyzed. Immunoblot (Fig. 5a, upper panel) as well as its densitometry analysis (Fig. 5a, lower panel) showed the significant increase of ph-MSK1 (p < 0.001), p38 (p < 0.01), and ph-p38 (p < 0.001), while ph-ERK1/2 (p < 0.001) levels significantly decrease in tumor compared to resection margin tissues, thus, indicating p38-mediated activation of MSK1 in GC. The increase of ph-MSK1 levels in GC was further confirmed by IHC analysis of the same tissues (Fig. 5b). The observed increase of H3S10ph on the overexpression of MSK1 in AGS cells by immunoblot (Fig. 5c and Additional file 5: Figure S4A) and, moreover, decrease of H3S10ph- on H89-mediated biochemical inhibition of MSK1 by immunoblot studies in AGS and KATOII cell lines (Fig. 5d and Additional file 5: Figure S4B) and immunofluorescence studies in AGS cells (Fig. 5e) confirmed MSK1-mediated phosphorylation of H3S10 in GC. Further, immunoblot analysis with specific antibodies showed a decrease of ph-MSK1 and H3S10ph only on the treatment of p38 inhibitor (SB203580) in AGS and KATOIII cells but not on the treatment of ERK1/2 inhibitor (PD89059) (Fig. 5f). And immunofluorescence studies on inhibitor-treated AGS cells validated that p38 is responsible for phosphorylation of MSK1 in GC (Fig. 5g), thus, confirming p38-MAPK/MSK1-mediated increase of H3S10ph in GC.