Tumorigenesis as well as drug-resistance phenotypes in humans are frequently associated with deregulated expression of Bcl-2 family proteins, best known for their ability to control mitochondrial apoptosis. The prosurvival function of Bcl-2 or related homologues, such as Bcl-xL or Mcl-1, is antagonized by distant cousins of the same family, referred to as “BH3-only” proteins, enabling activation of another class of Bcl-2 family members, Bax and/or Bak, that perturb mitochondrial integrity, triggering apoptotic cell death, e.g., in response to anticancer drug treatment.1 BH3-only proteins act as sentinels within the cell, integrating sheer endless numbers and types of stress signals to translate them into prodeath decisions at mitochondria. For example, the BH3-only protein Bim becomes activated in response to cytokine or growth factor deprivation, whereas Bid can get processed in a caspase-8–dependent manner after activation of certain tumor necrosis factor receptor (TNFR) family members by their cognate ligands, such as TNF itself, thereby linking the extrinsic (death receptor regulated) with the intrinsic (Bcl-2 regulated) apoptosis pathway. Others, such as Noxa or PUMA, are prime targets of the tumor suppressor p53 and critical for cell death in response to DNA damage–inducing agents.2 Many of these BH3-only proteins, including PUMA, show tumor suppressor potential in oncogene-driven mouse models of cancer, and loss of function or reduced expression of BH3-only proteins have been noted in human cancer, including Burkitt lymphoma, renal cell carcinoma, or melanoma. BH3-only protein expression levels can associate with overall and disease-free survival in some types of cancers, e.g., in colorectal carcinoma, and play rate-limiting roles for the efficacy of most anticancer drugs and regimens used today.2 All these observations are consistent with the well-accepted notion that induction of apoptosis acts as a barrier against malignant transformation and that resistance to apoptosis constitutes one of the hallmarks of cancer.3
In this issue of HEPATOLOGY, Qiu and colleagues4 demonstrate in an elegant study that there is also a “dark side” to this force, as lack of apoptosis upon DNA damage caused by the carcinogen diethylnitrosamine (DEN), due to loss of PUMA, can actually prevent the initiation of hepatocellular carcinoma (HCC) in mice (Fig. 1A). Loss of PUMA clearly delayed onset and disease burden in DEN-treated animals, and its absence directly affected the rate of cell death and subsequent compensatory proliferation observed after mutagenic challenge, which is considered critical for disease development.5 DEN treatment induced the induction of PUMA messenger RNA and protein that was, however, independent of p53, at least at the time points analyzed. It has been documented before that PUMA can be induced in a manner that does not require p53, but usually such stimuli impair the AKT/PI3K (phosphoinositide 3-kinase) signaling pathway, such as the lack of cytokines.6 As such, this study provides rare evidence that PUMA might be activated in response to DNA damage independent of p53. Qiu and colleagues suggest that the stress-activated c-Jun N-terminal kinase, JNK1, known to be involved in DEN-driven hepatocarcinogenesis,7, 8 is responsible for PUMA induction under these conditions. Consistently, pretreatment of mice with a pharmacological inhibitor of JNK, or loss of JNK1 impaired induction of PUMA, cell death, compensatory proliferation, and onset of disease. As a side note, it would have been interesting to see if inhibition of JNK signaling may have also impacted on increased expression of Bim, a related BH3-only protein, or possible Bid processing, because both molecules have been reported to play a prominent role in different settings of acute liver damage.9
Notably, older data has already provided some evidence that overexpression of Bcl-2 in hepatocytes,10 or loss of the BH3-only protein Bid, can delay DEN-driven liver carcinogenesis.11 Furthermore, hepatocyte-restricted loss of the prosurvival Bcl-2 homologue Mcl-1 was recently reported to facilitate spontaneous formation of liver cancer, even in the absence of overt inflammation.12 At first sight, these observations seem counterintuitive, because apoptosis resistance should favor tumor formation, whereas increased cell death rates should be protective. However, such phenomena are not unprecedented, even in human cancer where high levels of Bcl-2 expression were reported to correlate with better prognosis in some tumor types.2 Thus, it seems that apoptotic cell death of healthy tissue can contribute to the development of malignant disease, whereas resistance of tumor cells to apoptosis might be beneficial, at least in some settings. Of note, a recent study also suggests that tumor cell apoptosis caused by irradiation treatment (presumably mediated by PUMA) can foster the outgrowth of surviving cancer cells. This phenomenon appears to involve caspase-3–dependent activation-mediating cleavage of calcium-independent phospholipase A2, which leads to increased production of arachidonic acid and its derivatives, such as prostaglandin E2, known to promote tumor cell growth.13 Release of growth-promoting substances from dying cells and cellular competition for space and survival factors is a well-documented phenomenon in Drosophila,14 and the molecular basis of this response may reach as far as human cancer pathogenesis with important implications for treatment failure.13
How can these contradicting observations related to defective apoptosis and oncogenesis versus tumor suppression be reconciled? It is undisputed that repeated cycles of attrition and repletion of liver tissue, e.g., due to chronic viral infection or continuous alcohol abuse, drive pathogenesis and that associated inflammation accelerates the onset of liver cancer.5 A simple explanation for the delayed tumor onset in apoptosis-defective mice and an extrapolation to the human setting would be that mutations and other genetic alterations caused—e.g., by mycotoxins or integration of viral components into host DNA in hepatocytes, or their progenitors—are amplified and spread more rapidly when cell death occurs, leading to loss of genomic stability and to the selection of transformed clones. As such, lack of proliferative drive, due to lack of apoptotic cell death, might help to contain this damage, reducing the risk of accumulating additional hits and perhaps also allowing the immune system to clear transformers more readily (Fig. 1B). A similar phenomenon has recently been reported to account for delayed IR-driven lymphomagenesis in PUMA−/− mice. In this model system, however, lack of PUMA prevented the expansion of damaged stem/progenitor cells in the bone marrow, because T cell–restricted prevention of irradiation-driven cell death by overexpression of Bcl-xL failed to prevent onset of disease, whereas disease could in part be restored in PUMA−/− mice when stem cell proliferation was enforced by glucocorticoid-mediated depletion of more mature lymphocytes.15, 16 This suggests that irradiation-driven lymphomas in mice are diseases mainly derived from stem/progenitor cells. Because stem cells in the liver are less well-defined, it remains unclear in which cell type loss of PUMA may actually be critical to prevent onset of disease. Also, the role of JNK1 in triggering PUMA induction in hepatocytes, in relation to the observed effect, remains to be fully investigated. A recent study actually shows that combined loss of JNK1 and JNK2 restricted to the liver does not delay, but rather enhances, DEN-driven liver tumor burden.17 This phenomenon was associated with increased hepatocyte death and compensatory proliferation, assigning a prosurvival but also putative tumor suppressive role to JNK1 under these conditions. However, a neutralizing effect of JNK2 loss that can have opposing effects on death and proliferation, when compared with JNK1 deficiency,18 cannot entirely be excluded. Nonetheless, the anti-oncogenic effect of JNK may in part depend on the noted reduction of c-Myc and cyclin D1 expression in wild-type cells.
In contrast, loss of JNK in liver cells as well as in nonparenchymal cells by Mx-cre–mediated deletion ameliorated liver damage–associated inflammation and onset of disease, as noted before.17 These findings point to a key role for JNK1 in the inflammatory response of nonparenchymal cells and, together with the observations by Qiu etal., raise the question of where PUMA actually needs to become activated to exert its oncogenic effect. Keeping the data by Das and colleagues in mind,17 DNA damage–triggered, JNK1-dependent PUMA expression in liver cells, as suggested, cannot account for the observed effect, because loss of JNK in hepatocytes does not delay disease. However, PUMA-mediated hepatocyte death may still be key, but the effect of JNK1 on PUMA levels may be indirect by enabling cytokine production and/or signaling. Kupffer cell–derived inflammatory mediators such as interleukin-1 (IL-1), IL-6, or, more likely here, TNFα, might actually induce PUMA in hepatocytes, as noted before,19 and drive induction of this as well as activation of other BH3-only proteins, presumably Bim or Bid,9 in a JNK-dependent manner, thereby triggering hepatocyte death. This explanation appears plausible also in light of the findings that loss of p53 did not impair PUMA induction after DEN treatment.4 However, maybe the time points chosen for analysis were not optimal to reveal a p53 contribution to an early acute DNA damage response, and the induction of PUMA, as noted on day 3 after DEN treatment, may actually have been cytokine-driven.
In this context, it is worth mentioning that the related BH3-only protein Bid, in its full-length form, may also exert a nonapoptotic function and contributes to the production of inflammatory cytokines downstream of NOD (nucleotide-binding and oligomerization domain) receptor activation.20 Hence, it remains theoretically possible, albeit pure speculation, that the observed delay of HCC development in Bid-deficient mice might be due to a negative impact on cytokine production, rather than compensatory proliferation directly,11 which again is cytokine-dependent. Conditional ablation of BH3-only proteins in different cell types may help to dissect their true function in HCC pathogenesis.
Overall, it becomes increasingly evident that the induction of apoptotic cell death can no longer be considered as a tightly sealed barrier against malignant disease but that cell death, either early in disease pathogenesis, such as the one observed in response to chronic viral infections or other mechanisms triggering repeated tissue damage, can cause tumor formation.