Apoptotic DNase network: Mutual induction and cooperation among apoptotic endonucleases

Abstract DNA fragmentation produced by apoptotic DNases (endonucleases) leads to irreversible cell death. Although apoptotic DNases are simultaneously induced following toxic/oxidative cell injury and/or failed DNA repair, the study of DNases in apoptosis has generally been reductionist in approach, focusing on individual DNases rather than their possible cooperativity. Coordinated induction of DNases would require a mechanism of communication; however, mutual DNase induction or activation of DNases by enzymatic or non‐enzymatic mechanisms is not currently recognized. The evidence presented in this review suggests apoptotic DNases operate in a network in which members induce each other through the DNA breaks they produce. With DNA breaks being a common communicator among DNases, it would be logical to propose that DNA breaks from other sources such as oxidative DNA damage or actions of DNA repair endonucleases and DNA topoisomerases may also serve as triggers for a cooperative DNase feedback loop leading to elevated DNA fragmentation and subsequent cell death. Therefore, mutual induction of apoptotic DNases has serious implications for studies focused on activation or inhibition of specific DNases as a strategy for therapeutic intervention aimed at modulation of cell death.


| D ISCOVERY OF AP OP TOTIC DNA S E S
Apoptotic DNases are endonucleases responsible for fragmentation of DNA in apoptosis and other types of cell death. 1 Initial studies observed delayed changes in DNA following irradiation cell injury, but preceding cell death, might be the result of 'enzyme activation'. 2 Non-necrotic cell death associated with normal physiological involution or triggered by noxious agents was recognized as programmed cell death and defined as apoptosis in the seminal work on apoptosis by Kerr, Wyllie and Currie in 1972. 3  internucleosomal DNA fragmentation pattern similar to isolated nuclei exposed to endonuclease digestion. Additionally, the observed morphological chromatin condensation of apoptosis correlated closely with the products of DNA digestion. The DNase involved in early studies of chromatin structure that produced internucleosomal DNA fragmentation was initially called 'Ca/Mg-dependent endonuclease' due to its cation requirements. 6 This activity could be identified by the production of a 200-bp-ladder pattern in agarose gel electrophoresis of DNA from isolated cell nuclei incubated in the presence of Ca 2+ and Mg 2+ ions. 7 Wyllie suggested apoptosis may be induced by activation of this Ca/Mg-dependent endonuclease. 5 However, the studies of the next two decades did not confirm the existence of a single apoptotic endonuclease, but instead showed that nine DNases acting together perform apoptotic DNA fragmentation. This group of the endonucleases included the venerable deoxyribonuclease 1 (DNase I), well known since the 1940s, 8 three DNase I homologs, DNase 2 with its two homologs, endonuclease G (EndoG) and caspase-activated DNase (CAD). These DNases are covered in greater detail in the review by Keyel. 9

| E VIDEN CE OF AN AP OP TOTI C DNA S E NE T WO RK
When studied individually by different research groups, each of the DNases behaved as the central regulator of all apoptotic DNA fragmentation because its overexpression induced apoptosis and its genetic inactivation inhibited DNA fragmentation-associated cell death in different models. [10][11][12][13] However, this 'necessary and sufficient' character seems to be shared by different DNases even within the same experimental model. For example, graphene cytotoxicity in vitro and cisplatin nephrotoxicity in vivo were shown to be alleviated by inhibition of either DNase I or EndoG. 14,15 Three enzymes, DNase I, DNase γ (DNase1L3) and CAD, were shown to be important for γ irradiation-induced cell death in vivo. 16,17 Both EndoG and CAD were shown to be necessary for cardiac hypertrophy, 18,19 while both DNase I and DNase γ were shown to be crucial for acetaminopheninduced acute liver injury. 17,20 In the latter model, 20 DNase I knockout mice were protected against drug-induced liver injury; however, the protection of DNase I knockouts was an apparent off-target effect because the most active and abundant endonuclease expressed in liver is DNase γ. 21 While individual DNases can appear to be necessary and sufficient for DNA fragmentation leading to cell death, even when tested within the same model, temporal relationships in DNase induction suggested a cooperative relationship. Follow-up studies demonstrated that in some models, the induction of different apoptotic DNases occurred at different time-points after a cell death stimulus. Apoptotic DNA fragmentation induced by the protein kinase inhibitor staurosporine in proliferating N1E-115 neuroblastoma cells was associated with induction of CAD followed by its disappearance that coincided with the induction of DNase γ. 22 In kidney ischaemia-reperfusion in rats and mice, DNase I was induced with a peak at 16 hours during the reperfusion stage, while EndoG was induced only when DNase I went down several hours later. 23 In both of these models, DNases appear to have a way of signalling and regulating the expression of each other.
In acute kidney injury induced by cisplatin, EndoG could be induced only in DNase I-expressing wild-type mice, while DNase I knockouts did not have EndoG induction in the kidney. 15 Again, this clearly suggested a signalling link between the two enzymes. DNase I has preferential specificity to AT-rich sequences, 26 like those found in promoter TATA boxes. 27 On the other hand, the entire EndoG gene, including its promoter, is highly GC-rich and presents as one large CpG island, 28 which would be relatively resistant to DNase I cleavage. The question remains, why is DNase I able to induce EndoG? It may be DNase I scissions are located outside of EndoG gene. These reports were the first evidence that endonucleases may be induced by the DNA-degrading activity of DNase I.

Mutual induction by directly targeting promoter regions of other
DNases is not the only mean by which DNases can cooperate. In terms of sequence specificity, DNases often overlap with each other.
However, due to some specificity of the initial cleavage, endonucleases may cooperate with each other in a manner where the product of one may act as the substrate of another. In particular, DNase I was shown to stimulate the ability of human recombinant EndoG to produce double-stranded DNA breaks in naked DNA and chromatin in vitro. 29 Cooperative activity between CAD and DNase γ resulting in internucleosomal DNA fragmentation was reported in tumour necrosis factor-induced apoptosis in HT-29 cells. 30

| OTHER CONTRIBUTOR S TO THE DNA S E NE T WORK
DNases are just one source of DNA breaks, and it is logical to propose that DNA breaking mechanisms in general, including radiation, mitochondria-and drug-induced reactive oxygen species (ROS), apurinic/apyrimidinic (AP) DNA repair endonucleases and DNA topoisomerases (stopped mid-reaction by topoisomerase inhibitors), may contribute to apoptotic DNase induction through a feedback loop or act cooperatively with apoptotic DNases (Figure 1). There is overwhelming collective evidence that ROS induce direct DNA breaks as well as DNA modifications, for example 8-hydroxyguanosine requiring DNA repair with participation of AP-endonucleases. This entire process is associated with DNA breaks and activation of DNA damage pathways and apoptosis, indirectly leading to enzymatic DNA fragmentation. Similarly, topoisomerase inhibitors induce proteinbound DNA breaks and activate DNA damage pathways and apoptosis with its own DNase-mediated DNA fragmentation. 31,32 Human AP endonuclease Ape 1 and its N-terminally truncated form were shown to participate in apoptotic DNA degradation and potentially cooperate with CAD. 33 CRN-1, a C elegans homologue of human AP endonuclease FEN-1 that is normally involved in DNA replication and repair, was reported to cooperate with C elegans Endo G (CPS- 6) in DNA fragmentation, utilizing the endonuclease activity of CPS-6 and both the 5'-3' exonuclease activity and gap-dependent endonuclease activity of CRN-1. 34 Whether initial oxidative DNA breaks, DNA repair endonucleasemediated DNA breaks or topoisomerase-bound DNA breaks directly contribute to the pool of breaks and directly activate or cooperate with apoptotic endonucleases has not been studied. The absence of research on such relationships could be because the interactions are difficult to distinguish and study separately due to lack of appropriate research tools or an apparent lack of interest. However, even in a simple in vitro model, the anticancer drug bleomycin able to directly induce DNA strand breaks, induced transcription of DNase I and EndoG genes inserted in an expression vector as effectively as recombinant endonucleases. 24,25 Placing DNA breaks at the centre of communication between all DNA-damaging agents makes it easy to imagine that a single endonuclease activated by an external stimulus, extracellular or intracellular ROS, or DNA proliferation and repair agent could trigger a vicious cycle of apoptotic DNases to destroy host cell DNA and induce cell death ( Figure 1).

| CON CLUS ION
The precise mechanisms that enable apoptotic DNases to act simul-

ACK N OWLED G EM ENTS
This study was supported by National Institutes of Health grant 2P20 GM109005-06 and VA Merit Review grant 2I01 BX002425.

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.