Association of patterns of class I histone deacetylase expression with patient prognosis in gastric cancer: a retrospective analysis

Background: Although histone deacetylases (HDACs) are known to have an important regulatory role in cancer cells, and HDAC inhibitors (HDIs) have entered late-phase clinical trials for the treatment of several cancers, little is known about the expression patterns of HDAC isoforms in tumours. We aimed to clarify these expression patterns and identify potential diagnostic and prognostic uses of selected class I HDAC isoforms in gastric cancer. Methods: Tissue samples from a training cohort and a validation cohort of patients with gastric cancer from two German institutions were used for analyses. Tissue microarrays were generated from tumour tissue collected from patients in the training group, whereas tissue slides were used in the validation group. The tissues were scored for expression of class I HDAC isoforms 1, 2, and 3. Overall expression patterns (gHDAC) were grouped as being negative (all three isoforms negative), partially positive (one or two isoforms positive), or completely positive (all isoforms positive), and correlated with clinicopathological parameters and patient survival. The main endpoints were amount of expression of each of the three HDAC isoforms, patterns of expression of gHDAC, effect of metastasis on expression of HDAC and gHDAC, and overall survival according to HDAC expression patterns. Findings: 2617 tissue microarray spots from 143 patients in the training cohort and 606 tissue slides from 150 patients in the validation cohort were studied. 52 of the 143 (36%) gastric tumours in the training cohort and 32 of the 150 (21%) gastric tumours in the validation cohort showed nuclear expression of all three HDAC isoforms. 60 (42%) of tumours in the training cohort and 65 (43%) in the validation cohort expressed one or two isoforms in the nuclei, whereas 31 (22%) of tumours in the training cohort and 53 (35%) in the validation cohort were scored negative for all three proteins. gHDAC expression in both cohorts was higher when lymph-node metastases were present (p=0·0175 for the training group and p=0·0242 for the validation group). Survival data were available for 49 patients in the training group and 123 patients in the validation group. In the validation cohort, 3-year survival was 44% (95% CI 34-57) in the HDAC1-negative group, 50% (39-64) in the HDAC2-negative group, and 48% (34-67) in the gHDAC-negative group. 3-year survival decreased to 21% (11-37) when HDAC1 was positive, 16% (9-31) when HDAC2 was positive, and 5% (1-31) when gHDAC (all isoforms) were positive. Those patients highly expressing one or two isoforms (the gHDAC-intermediate group) had an estimated 3-year survival of 40% (29-56). In multivariate analyses, high gHDAC and HDAC2 expression were associated with shorter survival in the training cohort (gHDAC: hazard ratio [HR] 4·15 [1·23-13·99], p=0·0250; HDAC2: HR 3·58 [1·36-9·44], p=0·0100) and in the validation cohort (gHDAC: HR 2·18 [1·19-4·01], p=0·0433; HDAC2: HR 1·72 [1·08-2·73], p=0·0225), independent of standard clinical predictors. Interpretation: High HDAC expression is significantly associated with nodal spread and is an independent prognostic marker for gastric cancer. Additionally, we postulate that immunohistochemical detection of HDAC as a companion diagnostic method might predict treatment response to HDIs, thereby enabling selection of patients for this specific targeted treatment in gastric cancer.