Ultrasound and Transcriptomics Identify a Differential Impact of Cisplatin and Histone Deacetylation on Tumor Structure and Microenvironment in a Patient-Derived In Vivo Model of Gastric Cancer

Abstract

The reasons behind the poor efficacy of transition metal-based chemotherapies (e.g., cisplatin) or targeted therapies (e.g., histone deacetylase inhibitors, HDACi) on gastric cancer (GC) remain elusive and recent studies suggested that the tumor microenvironment could contribute to the resistance. Hence, our objective was to gain information on the impact of cisplatin and the pan-HDACi SAHA (suberanilohydroxamic acid) on the tumor substructure and microenvironment of GC, by establishing patient-derived xenografts of GC and a combination of ultrasound, immunohistochemistry, and transcriptomics to analyze. The tumors responded partially to SAHA and cisplatin. An ultrasound gave more accurate tumor measures than a caliper. Importantly, an ultrasound allowed a noninvasive real-time access to the tumor substructure, showing differences between cisplatin and SAHA. These differences were confirmed by immunohistochemistry and transcriptomic analyses of the tumor microenvironment, identifying specific cell type signatures and transcription factor activation. For instance, cisplatin induced an "epithelial cell like" signature while SAHA favored a "mesenchymal cell like" one. Altogether, an ultrasound allowed a precise follow-up of the tumor progression while enabling a noninvasive real-time access to the tumor substructure. Combined with transcriptomics, our results underline the different intra-tumoral structural changes caused by both drugs that impact differently on the tumor microenvironment.

Keywords: HDAC Inhibitors; SAHA; cancer immunity; cisplatin; epigenetic; gastric cancer; microenvironment; p53; transcriptomic; ultrasound.

Figure 1

Three measurement techniques on primary gastric adenocarcinoma tumors grown in nude mice (PDX): caliper and ultrasound on living mice, and caliper on ex vivo tumors (gold standard). (a) Histological hematoxylin eosin staining (magnification 40×) of the primary human tumor compared to tumors issued from it grown in a PDX mouse (1st and 8th passage). The PDX tumors retained the major architectural and histological traits of the primary tumor and maintained across passages from mouse-to-mouse. (b) The caliper on the live mouse only measured the length and width, and the depth was estimated from the width. (c) An ultrasound enabled the measurement of all three dimensions of the tumor (length, width, and depth). (d) The caliper on the ex vivo tumor was the gold standard for length, width, and depth measures. Formulas used for tumor volumes are indicated below.

Figure 2

Evolution of the gastric tumor growth between treated and non-treated tumors. (a) Average tumor volume growth over time, measured by a manual caliper on living mice from day 0 to day 23, between suberanilohydroxamic acid (SAHA, HDAC inhibitor), standard chemotherapy using cisplatin, which targets DNA, and a control group (n = 12, 12, 8). p < 0.05 between control and treated groups as established by a t-test. (b) Illustration of representative ultrasound images and calculations showing the evolution of a GCX-004 tumor from the control group over time (day 0, day 6, day 16, and day 23).

Figure 3

Correlation and comparison of the tumor volume among the gold standard, caliper, and ultrasound measurements, and for each treatment group (control, cisplatin and SAHA). At day 23, final tumor volumes were measured and compared for each measurement method. (a,b) Gold standard measurements were correlated with caliper or ultrasound measurements using GraphPad Prism, with r values 0.95 and 0.94, respectively, and with a p value < 0.0001. Then measurements were compared among each treatment group for the gold standard (c), caliper (d), and ultrasound (e), and for each treatment group among the different measurement methods, control (f), cisplatin (g), and SAHA (h). Control n = 12, cisplatin n = 12, SAHA n = 6. Data were subjected to statistical analyses using GraphPad software version 6.0 (ns = not significant, * = p < 0.05, *** = p < 0.001).

Figure 4

Histology confirms treatment-related sub-structural changes observed with ultrasound. Ultrasound images of a representative GXC-004 tumor from each treatment group (upper panel, day 23) compared to their corresponding histological slices (hematoxylin eosin staining, magnification ×2 (upper), ×20 (middle) and ×40 (lower)). Control tumors had sharper margins with a white peripheral rim (empty arrow), while treated tumors, especially SAHA-treated ones, were more lobulated. Some tumors (especially in the control group) presented scarce intraglandular necrosis, which could potentially explain the comet-tail artifacts observed on ultrasound (i.e., small central hyperechoic punctiform foci, plain arrow). Histology revealed a stromal predominance (circled) in cisplatin-treated tumors compared to the primary tumor and controls, with a stroma/gland ratio > 50%. Cysts (star) were frequently observed with an ultrasound (visible when >2 mm) in the periphery of SAHA-treated tumors, less frequently in cisplatin-treated tumors. This finding was confirmed by histology (star).

Figure 5

Immunohistochemical (IHC) staining and intensity measures for Ki67, p53, H3K9, and α-actin. (a) Representative photomicrographs of the expression of Ki67, p53, H3K9, and α-actin for treatment group (control, cisplatin, and SAHA (magnification ×20 with close up ×40). (b) Quantifications by Fiji software of each condition. Graphs represent number of pixels for each light intensity from 0 to 255. Red lines indicate the relevant values as considered, based on negative control (e.g., no primary antibody).

Figure 6

Transcriptomics analyses confirm differences in the mouse stroma/microenvironment of the human gastric tumor caused by cisplatin and SAHA. (a) Total RNAs were extracted from the tumors. Sequences obtained were specifically aligned on the mouse genome. Deregulated genes with fold change >1.5 and p-value < 0.05 were analyzed. False discovery error <0.05 and Z score above 2 were chosen as the cut-off in bioinformatics analyses and AltAnalyze was used to find gene signatures corresponding to given processes representing biomarkers selective of each condition (b), typical signatures for specific cell types (c), regulation of specific transcription factors (d).