A total of 231 cases of DLBCL treated with R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) or R-CHOP–like (with or without radiotherapy or surgery) chemotherapy were selected for the study. Cases were retrieved from the archival files from the Department of Pathology, Severance Hospital, from 2005 to 2011. All cases were independently reviewed by two pathologists (S.O.Y. and S.H.K.) based on current World Health Organization criteria [9], and discordant cases were consulted to other expert hematopathologists. In HOTAIR expression analysis, 164 cases were selected from the above 231 cases and investigated after quality assessment of extracted RNA. Clinical data were obtained from medical records. All study protocols were performed according to the ethical guidelines of the ‘‘World Medical Association Declaration of Helsinki–Ethical Principles for Medical Research Involving Human Subjects.’’ This study was approved by the Institutional Review Board of Severance Hospital.

Formalin fixed paraffin embedded (FFPE) tissue sections were prepared and stained with hematoxylin and eosin, and then the tumor areas were confirmed and marked under the microscope. The marked areas mainly contained packed tumor cells, and the stromal component was less than 10% of the marked area. The unstained slides of FFPE tissues were prepared after dissecting FFPE tissue blocks at 10-mm thickness using a microtome, and the marked area was scraped using a scalpel blade. Generally, three slices of tissue section per case were used for RNA extraction. Total RNA was isolated using an RNeasy FFPE Kit (Qiagen, Hilden, Germany) according to the supplier’s instructions. Extracts of RNA were verified by measuring the ratios ofA260/A280 and A260/A230 with a ND-1000 NanoDrop spectrophotometer (NanoDrop, Wilmington, DE, USA). Reverse transcription was performed using a QuantiTect Reverse Transcription kit (Qiagen). The expression patterns of HOTAIR were assayed by relative quantification using expression of the house-keeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Primers for HOTAIR and GAPDH were as follows: HOTAIR (forward, 5'-AGCCAGAGGAGGGAAGAGAG-3'; reverse, 5'-TCCCGTTCCCTAGATTTTCC-3') and GAPDH (forward, 5'-CAAATTCCATGGCACCGTCA-3'; reverse, 5'-ATCGCCCCACTTGATTTTGG-3'). Primers of HOTAIR were designed to detect all three transcript variants (transcript variant 1, 3, and 2). In brief, a 20 μL mixture containing 1.0 μL of cDNA, power SYBR Green PCR Master Mix (Applied Bio-systems, Carlsbad, CA, USA), 1.0 μL of 10 pmol/μL forward primer, 1.0 μL of 10 pmol/μL reverse primer, and 7.0 μL of tertiary distilled water was prepared. Amplification was performed using a Step One Plus Real-Time PCR instrument (Applied Biosystems) under the following conditions: denaturation at 94°C for 10 minutes, followed by 35 cycles of 94°C for 15 seconds, 55°C for 30 seconds, and 72°C for 30 seconds (fluorescence signal acquisition was performed at this phase). Immediately after amplification, melting curve analysis was performed for amplicon verification. All samples were analyzed in triplicate to confirm reproducibility. Using the Step One Plus Real-Time PCR System software ver. 2.1 (Applied Biosystems), the threshold cycle (Ct, beginning of the polymerase chain reaction exponential phase) value of amplified HOTAIR was normalized versus that of amplified GAPDH (2^-dCt). After quality assessment of extracted RNA and expression analysis of the housekeeping gene GAPDH, 164 cases were finally selected for the analysis of HOTAIR expression. Normal palatine tonsil tissue was obtained from 10 cancer-free individuals and used as age-matched cancer-free controls. The cutoff value for high expression of HOTAIR (HOTAIR^high) was determined at the uppermost value among those of normal control tonsil tissues.

The hematoxylin and eosin slides were reviewed and two representative core tissues of the tumor area were selected in each case. Core tissues (3 mm in diameter) were taken from the individual donor blocks and arranged in new recipient tissue microarray paraffin blocks using a trephine apparatus. Immunohistochemical staining and in situ hybridization were performed on 4-μm tissue microarray tissue sections. Immunohistochemistry of EZH2 (1:100, Invitrogen, Carlsbad, CA, USA), SUZ12 (1:50, Abcam, Cambridge, UK), EED (1:1,000, Abcam), H3K27me3 (1:100, clone C36B11, Cell Signaling Technology, Beverly, MA, USA) was performed using the Ventana BenchMark XT Autostainer (Ventana Medical Systems, Tucson, AZ, USA). Immunohistochemistry for c-MYC (1:50, clone Y69, Abcam), BCL2 (1:50, Novocastra, Newcastle, UK), CD10 (1:100, Novocastra), BCL6 (RTU, Novocastra), and MUM1 (1:200, Cell Marque, Rocklin, CA, USA) was performed using the LEICA BOND-III Autostainer (Leica Biosystems, Newcastle Upon Tyne, UK). For in situ hybridization for Epstein-Barr virus (EBV), the INFORM EBER probe (Ventana Medical Systems) was used with the Ventana BenchMark XT Autostainer (Ventana Medical Systems) and ISH iVIEW Blue Detection kit (Ventana Medical Systems).

For EZH2, SUZ12, EED, and H3K27me3, staining intensity (0, no staining to weak; 1, moderate to strong) and proportion of positive tumor cell nuclei (0, <10%; 1, ≥10% and <75%; 2, ≥75%) were semiquantitatively graded as in the previous study [8]. Based on the intensity multiplied by proportion, the protein expression was scored as low (0), intermediate (1), or high level (2). For CD10, BCL6, and MUM1, the positive cutoff value was determined according to the Hans classification criteria [10] and was considered positive if ≥30% of the tumor cells showed nuclear immunoreactivity for BCL6 and MUM1, and if ≥30% of cells showed membranous reactivity for CD10. Determination of the germinal center B-like (GCB) or non-GCB phenotype was based on the Hans algorithm 10.

For c-MYC and BCL2, staining intensity (0, no staining to weak; 1, moderate to strong) and areas of positive tumor cell nuclei by 10% increments (1 of <10% to 10 of 90%–100%) were semiquantitatively graded. The cutoff value for high expression was determined by log-rank tests (Mantel-Cox) for overall survival as in the previous report, and the threshold value was determined for c-MYC at a score ≥4 (≥40% of tumor cells with moderate to strong expression), and BCL2 at a score ≥7 (≥70% of tumor cells with moderate to strong expression) [11,12]. For EBV in situ hybridization, the threshold value was determined when ≥ 10% of the tumor cells showed moderate to strong nuclear expression.

The features of expression of polycomb repressive complex proteins (EZH2, SUZ12, and EED), H3K27me3, c-MYC, and BLC2 in HOTAIR^high or HOTAIR^low cases are presented in Fig. 1.

The t test and chi-square test were used to analyze the differences between the variables examined. Overall survival times were measured from the date of lymphoma diagnosis to the date of death or last follow-up visit. Patient survival rates were determined using the Kaplan-Meier method, and the differences in survival rates were compared using the log-rank test. Multivariate analysis was performed using the Cox proportional hazards model. A two-sided p-value<.05 was considered to be statistically significant. When a two-sided p-value was ≤.05 and <.10, a trend toward statistical significance was considered. All statistical analyses were carried out using SPSS software ver. 20.0 for Windows (IBM Corp., Armonk, NY, USA).

The DLBCL cases frequently showed high expression of PRC2 proteins and H3K27me3 (28.9%, 47.1%, 74.3%, and 35.7% for EZH2^high, SUZ12^high, EED^high, and H3K27me3^high, respectively). High expression of all PRC2 proteins (EZH2^high/SUZ12^high/EED^high) was noted in 16.1% of cases (Fig. 2A). The expression of H3K27me3 was positively correlated to PRC2 markers. When compared to cases of H3K27me3^low, those of high H3K27me3 expression more frequently showed high expression of EZH2 (41.2% vs 22.2%, p=.003), SUZ12 (62.5% vs 43.0%, p=.053), EED (88.6% vs 69.7%, p=.026), and all PRC2 (EZH2^high/SUZ12^high/EED^high) (32.3% vs 11.3%, p=.005). These results are summarized in Fig. 2B.

The expression of c-MYC was positively correlated to PRC2 markers and H3K27me3. When compared to cases of c-MYC^low, those of high c-MYC expression more frequently showed high expression of EZH2 (39.1% vs 24.6%, p=.036), SUZ12 (72.2% vs 32.1%, p<.001), EED (85.7% vs 66.7%, p=.011), all PRC2 (EZH2^high/SUZ12^high/EED^high) (26.4% vs 9.5%, p=.009), and H3K27me3 (40.6% vs 21.2%; p=.004). BCL2 was positively correlated with c-MYC (49.3% vs 26.8%; p=.002). These results are summarized in Fig. 2C. BCL2 expression showed no significant correlation with the tested PRC2-related markers.

The expression level of HOTAIR was significantly higher in cases of EZH2^high than those of EZH2^low (mean ratio value, 207 vs 66; p=.027) (Fig. 2D). For other markers, no statistical significance was observed according to HOTAIR expression level.

In the present study, 23.8% of cases (39/164) showed high HOTAIR expression levels (HOTAIR^high). In the correlation analysis between HOTAIR^high and clinicopathological characteristics of DLBCL (Table 1), high expression of c-MYC protein was less frequent in HOTAIR^high than in HOTAIR^low (p=.037). The rate of co-expression of c-MYC and BCL2 was significantly lower in HOTAIR^high than in HOTAIR^low (p=.015). Other clinicopathological factors including EBV (p=.660) and Hans classification (p=.746) showed no significant correlation with HOTAIR expression (Table 1).

In univariate analysis for overall survival of DLBCL patients (Table 2), the known prognostic factors including age >60 (p<.001), Eastern Cooperative Oncology Group performance status ≥2 (p<.001), elevated lactate dehydrogenase (p<.001), extranodal sites ≥2 (p=.001), Ann-Arbor stage III–IV (p=.003), and International Prognostic Index (IPI) risk ≥3 (p<.001) showed significant association with reduced overall survival rate. Expression of c-MYC and co-expression of c-MYC/BCL2 were also related to inferior overall survival rate (p=.032 and p=.006, respectively). Among the PRC pathway markers (EZH2, SUZ12, EED, and H3K27me3), SUZ12^high and H3K27me3^high were related to inferior overall survival rate (p=.028 and p=.010, respectively) (Fig. 3A, B). HOTAIR^high, however, was significantly related to superior overall survival rate than HOTAIR^low (p=.004) (Fig. 3C).

In multivariate analysis (Table 2), H3K27me3^high revealed an independent effect on poor overall survival (hazard ratio, 2.0; p=.023). IPI risk ≥3 was still determined as an independent prognostic factor (hazard ratio, 3.5; p<.001), while HOTAIR^high showed a tendency to be related to improved survival with marginal significance (hazard ratio, 0.5; p=.086).