Clinical significance and pre-clinical studies. To reveal the clinical significance of Hpa2 in gastric cancer, we subjected tissue array of gastric carcinoma biopsies and adjacent normal gastric tissue to immunostaining applying anti-Hpa2 antibody. We found that 99% (78/79) of the normal gastric tissues were stained positive for Hpa2 (Fig. 1A, left panels) compared with 56% (54/96) of the gastric carcinomas (Table 1), differences that are statistically highly significant (p=0.0001). This implies that about half of the gastric carcinoma cases lose, or exhibit reduced Hpa2 expression (Fig. 1A, middle and right panels), an expression pattern typical of a tumor suppressor. Moreover, Hpa2 staining intensity correlated inversely with tumor stage (p=0.04; Table 2) and tumor metastasis to lymph nodes (N). Thus, most (54%) of the tumors that were stained strongly for Hpa2 had only a few infected lymph nodes (N1-N2), whereas patients with low levels of Hpa2 were mostly (67%) diagnosed with multiple infected lymph nodes (N3-N4), differences that are statistically highly significant (p=0.01; Table 2). Such a correlation between Hpa2 and tumor metastasis to lymph nodes was noted earlier in head and neck carcinoma (24). Notably, patients that exhibited high levels of Hpa2 (n=56, median survival time of 72 months) survived longer than patients with low Hpa2 levels (n=79, median survival time of 23 months) (Fig. 1B; p=0.004). This suggests that in gastric cancer, Hpa2 functions to restrain tumor growth.
To further explore the role of Hpa2 in gastric cancer we next transfected gastric carcinoma cell lines with Hpa2 gene construct and examined their tumorigenic capacities in vitro. We found that proliferation of MKN-45 cells was attenuated markedly following Hpa2 overexpression (Fig. 2A; p<0.0001). Likewise, migration of MKN-45-Hpa2 cells was prominently reduced compared with control (Vo) cells (Fig. 2B; p<0.0001). Furthermore, the capacity of MKN-45 (Fig. 2C, upper panel), SGC-7901 (Fig. 2C, middle panels), and BGC-823 (Fig. 2C, lower panels) to form colonies in soft agar was attenuated following Hpa2 overexpression compared with control (Vo) cells. We next implanted control (Vo) and Hpa2 overexpressing MGC-803 cells intra-peritoneal (ip) and the survival time of the mice was followed. This model is most relevant because peritoneal metastases occur in 55-60% of patients with gastric cancer, associating with a low (2%) 5-year overall survival rate (3). Notably, mice implanted with Hpa2 cells survived significantly longer than control (Fig. 2D; p=0.0003). In a subsequent experiment, mice were similarly implanted with control (Vo) and Hpa2 cells, sacrificed 14 days later, and tumor lesions were collected. Remarkably, mice implanted with Hpa2 cells exhibited reduced tumor mass collected from the peritoneum (Fig. 2E), also evident by reduced amounts of ascites fluids collected (Suppl. Fig. 1A). Collectively, these results suggest that Hpa2 attenuates the tumorigenic properties of gastric carcinoma cells.
Hpa2 expression is induced by stress, involving HSF1 and AMPK. Reduced Hpa2 expression in some of the gastric carcinomas and its high expression in others (Fig. 1A) suggests that Hpa2 expression is tightly regulated. However, mechanisms that regulate Hpa2 gene expression have not been explored yet. We hypothesized that conditions of stress, which are often associated with the fast-growing tumor and the dynamic nature of the tumor microenvironment, are involved in Hpa2 gene regulation. To examine this possibility, we exposed gastric carcinoma cell lines to MG132, an inhibitor of the proteasome, which results in severe proteotoxic stress (32). Consistently, we found that Hpa2 expression was induced 10-30 folds by the proteotoxic stress elicited by MG132 in MKN-45, BGC-823, AGS, and MGC cells in a time-, and dose-dependent manner (Fig. 3A-D). Moreover, Hpa2 expression was induced to a comparable extent by bortezomib (Velcade) (Fig. 3E, F), a proteasome inhibitor that is most effective in the treatment of multiple myeloma patients and is also effective in gastric cancer (33). Hpa2 gene induction was similarly observed in non-transformed cells such as mouse embryonic fibroblasts (MEF, Suppl. Fig. 1B, upper panel), and in mouse Lewis lung carcinoma cells (Suppl. Fig. 1C, upper panel). In these cells, as well as in gastric carcinoma cells, the stress conditions also induced the expression of cytokines such as MIP2 (Suppl. Fig. 1B, C middle panels) and TNF-α (Suppl. Fig. 1B, C lower panels) that have a profound impact on the immune system and the tumor microenvironment.
To examine the molecular mechanism underlying Hpa2 induction, we subjected gastric carcinoma cells to MG132 for increasing periods and protein extracts were subjected to immunoblotting. Consistently, we found that MG132 treatment resulted in a profound increase in the phosphorylation levels of AMPK (Fig. 4A, upper panels), JNK (Fig. 4A, fifth panels), and p70S6K (pS6K; Fig. 4A, seventh panels), the latter is indicative of mTOR activation. We also found that MG132 enhances the phosphorylation of heat shock factor-1 (HSF1; Fig. 4A, third panels), while HSF1 expression was not affected (Fig. 4A, fourth panels). Notably, the increased phosphorylation of HSF1 was associated with a profound increase in the expression levels of heat shock protein (HSP) 40, 105, 27, and 70 (Suppl. Fig. 2A), a consequence of HSF1 activation because this increase in HSP expression was abrogated by HSF1 inhibitors KRIBB 11 and KNK-437 (Suppl. Fig. 2B). To tie between Hpa2 induction and the signaling pathways elicited by the stress, we treated MKN-45 cells with MG132 in the absence or presence of inhibitors specific for each signaling pathway. We found that Hpa2 induction by MG132 was attenuated markedly by KRIBB 11, an inhibitor of HSF1, and by dorsomorphin (Dor), an inhibitor of AMPK (Fig. 4B, red arrows; Fig. 4C, upper and middle panels). In contrast, Hpa2 gene induction by MG132 was not affected by rapamycin (inhibitor of mTOR) (Fig. 4B) or sp600125 (a JNK inhibitor), thus pointing to AMPK and HSF1 as modulators of Hpa2 expression. Moreover, silencing of HSF and AMPK-beta resulted in reduced Hpa2 expression (Fig. 4C, lower panel), further supporting the significance of these pathways in Hpa2 gene regulation. Notably, subjecting control (Vo) and Hpa2 overexpressing MKN-45 cells to MG132 revealed increased levels of cleaved caspase 3 and cleaved PARP in Hpa2 cells vs control (Vo) cells (Fig. 4D), suggesting that Hpa2 cells are more sensitive to proteotoxic stress conditions, resulting in increased apoptosis.
Hpa2 enhances AMPK phosphorylation. We next attempted to reveal signaling pathways that are modulated by Hpa2 and may be responsible for the observed reduced tumorigenic properties (Figs. 1, 2) and higher sensitivity to conditions of stress (Fig. 4D). Immunoblot analyses of cell extracts derived from control and Hpa2 overexpressing MKN-45 (Fig. 5A, left panels), BGC823 (Fig. 5A, middle panels) and SGC7901 (Fig. 5A, right panels) cells revealed a consistent increase in the phosphorylation of AMPK (Fig. 5A, upper panels) and its substrate, acetyl CoA carboxylase (pACC; Fig. 5A, third panels). Increased ACC phosphorylation by Hpa2 was also evident by immunofluorescent staining (Fig. 5B). Moreover, we found that the increase in ACC phosphorylation in Hpa2 cells was reversed by adding heparin to the cell culture medium (Fig. 5C), suggesting that this effect involves the interaction of Hpa2 with cell membrane HS (20). Enhanced phosphorylation of AMPK is relevant to the anti-tumorigenic properties of Hpa2 because AMPK activation is found in correlation with good prognosis of cancer patients, including gastric cancer patients (1, 34). Also, we found reduced phosphorylation of HSF1 in Hpa2 cells (pHSF1; Fig. 5A, fifth panels, BGC, SGS). This finding is important because HSF1 exerts a pro-tumorigenic effect in many types of cancer, including gastric cancer (35, 36). In order to study the significance of AMPK in gastric carcinoma, we applied metformin, a drug that induces the phosphorylation of AMPK (Suppl. Fig. 3A) and is used in the clinic to treat diabetic patients (37, 38). Notably, the proliferation of MKN-45 cells was attenuated markedly by metformin in a dose-dependent manner (Suppl. Fig. 3B) while inducing the expression of Hpa2 (Suppl. Fig. 3C). Furthermore, we found that metformin treatment reduced the invasion (Fig. 6A), migration (Fig. 6B, C), and colony formation by gastric carcinoma cells to an extent comparable to Hpa2 (Fig. 6A-D, Vo+Met vs Hpa2). These results suggest that the anti-tumorigenic properties of Hpa2 in gastric carcinoma are mediated, at least in part, by enhancing the phosphorylation of AMPK.
Given that Hpa2 is highly expressed in normal gastric tissue (Fig. 1A, left), we next examined the correlation between Hpa2 levels and AMPK phosphorylation in this tissue. We found that AMPK is phosphorylated in about half of the gastric tissues examined (43/93; Table 3) and, like Hpa2, is decreased prominently in gastric cancer (Fig. 5D; N=normal, T=Tumor). Importantly, gastric tissues that exhibit high levels of AMPK phosphorylation (Fig. 5D, E) are also stained positive for Hpa2 (Table 3; p=0.02). These results imply that Hpa2 functions to modulate AMPK phosphorylation and metabolic aspects in normal gastric epithelial cells.