Avenues to molecular imaging of dying cells: Focus on cancer

Abstract Successful treatment of cancer patients requires balancing of the dose, timing, and type of therapeutic regimen. Detection of increased cell death may serve as a predictor of the eventual therapeutic success. Imaging of cell death may thus lead to early identification of treatment responders and nonresponders, and to “patient‐tailored therapy.” Cell death in organs and tissues of the human body can be visualized, using positron emission tomography or single‐photon emission computed tomography, although unsolved problems remain concerning target selection, tracer pharmacokinetics, target‐to‐nontarget ratio, and spatial and temporal resolution of the scans. Phosphatidylserine exposure by dying cells has been the most extensively studied imaging target. However, visualization of this process with radiolabeled Annexin A5 has not become routine in the clinical setting. Classification of death modes is no longer based only on cell morphology but also on biochemistry, and apoptosis is no longer found to be the preponderant mechanism of cell death after antitumor therapy, as was earlier believed. These conceptual changes have affected radiochemical efforts. Novel probes targeting changes in membrane permeability, cytoplasmic pH, mitochondrial membrane potential, or caspase activation have recently been explored. In this review, we discuss molecular changes in tumors which can be targeted to visualize cell death and we propose promising biomarkers for future exploration.

explored. In this review, we discuss molecular changes in tumors which can be targeted to visualize cell death and we propose promising biomarkers for future exploration.

INTRODUCTION
A living organism can be considered as a complicated machine, which requires constant maintenance, modernization, and restructuring or reconstruction. Subunits of the organism, such as cells, are continuously produced, exploited, altered, utilized and exchanged. Billions of cells die daily as a part of natural processes in the adult human body, and even more cells die during embryonic development. Under physiological conditions, superfluous, dangerous, or damaged cells are killed and dismantled in a discrete and highly orchestrated manner. For instance, squamous epithelial cells are removed via cornification, 1 Müllerian duct in males or Wolffian duct in females via apoptosis, and pronephric kidney tubes also via apoptosis. 2,3 A mainstay of the body's homeostasis is a proper decision on cellular fate: death or survival.
It is thus not surprising that perturbations of cell death processes are an underlying factor of many pathologic conditions. Cell death is enhanced in ischemia, 4 sepsis, 5 type-1 diabetes, 6 transplant rejection, 7 neurodegenerative disorders, 8 and autoimmunity (e.g., AIDS). 9 In contrast, reduced cell death is observed in persistent inflammation (as occurs in chronic obstructive pulmonary disease and asthma), 10,11 autoimmunity (e.g., rheumatoid arthritis), 12 and cancer. 13 With nondestructive and minimally invasive medical imaging techniques like PET (positron emission tomography) and SPECT (single photon emission computed tomography), cell death in organs and tissues of the human body can be visualized and quantified. Such quantification may be important in cancer treatment, since monitoring of the increase in cell death early after the onset of antitumor therapy can serve as a predictor of the eventual therapeutic outcome.
In the following review, we describe molecular changes in tumors related to cell death and we provide an overview of the wide range of PET and SPECT tracers which have been developed to monitor these changes. We discuss the potential and the limitations of the existing tracers and we propose some promising biomarkers of dying cells which deserve to be explored in future imaging research.

Canonical classification of cell death modes
There are many ways for a cell to die. In recent years our concepts of cell death have changed. In this chapter, we first describe the canonical classification of cell death modes and we subsequently summarize new observations which have led to a revised classification.
The classical concept of cell death (proposed in 1973) is based on morphologic features of dying cells and makes a distinction between three death types: apoptosis (type I), autophagic cell death (type II), and necrosis (type III) (see Table 1). 14 Even nowadays, cell death is still frequently classified in these three subroutines. Apoptosis and autophagy are considered as "regulated" and necrosis as "accidental" cell death. 15

Apoptosis
Apoptosis was considered to be a noninflammatory, highly orchestrated, and inherently controlled process. Since its identification in 1972, 16 apoptosis has been the most investigated type of cell death. Apoptosis can be activated by intra-or extracellular stimuli and is then coined as "intrinsic" or "extrinsic" apoptosis. Both these apoptotic scenarios mutations in the TP53 gene and/or defects in the p53 signaling pathway (e.g., upregulation of the p53 inhibitor mouse double minute 2, mouse double minute 2 homolog [E3 ubiquitin-protein ligase]) result in uncontrolled proliferation and a brake on apoptosis. This may have a subsequent impact on both initiation of oncogenesis and development of treatment resistance. Although apoptosis is the best-characterized cell death mechanism, in many cancers it is not the main cause of cell loss induced by DNA damaging agents. 28

Autophagic cell death
Autophagy is a natural, regulated process for disassembly of dysfunctional or damaged cellular organelles and proteins.
Such damaged components are contained inside a double-membrane vesicle called an autophagosome. After fusion of an autophagosome and a lysosome to an autolysosome, the contents of the organelle are digested by acidic lysosomal hydrolases. 29 Even today, there is much controversy on the question whether in vivo autophagy is a type of cell death or fulfills a pro-survival function, for example, by limiting cell constituents during nutrient starvation. This question is raised because most inhibitors of autophagy accelerate (and not retard) cell death. [30][31][32][33][34] For this reason, autophagic cell death has now been defined as cell death inhibited by inactivation of autophagy genes or by autophagy inhibitors, such as 3MA, rather than cell death judged by simple morphological classification. 35 This definition is based on studies which have elucidated molecular mechanisms of autophagic cell death. 36,37 Tissue-specific knockout models of genes controlling autophagy in mice have provided much information about the role of autophagy in the development and differentiation of mammalian tissues and organs. 38 In some tissues (e.g., mouse liver) autophagy seems to suppress tumorigenesis, 39 but in most cases, autophagy facilitates the formation of tumors and increases tumor growth and aggressiveness. 40 Autophagy seems to be particularly induced when cancers progress to metastasis. 41 Inhibitors of autophagy may thus be useful as adjuvants in cancer therapy.

Necrosis
Necrosis is the consequence of irreversible damage to cells caused by factors such as mechanical trauma, infections, toxins, and shortage of oxygen and nutrients. Necrosis is traditionally thought to be an uncontrollable and accidental type of cell death, which is highly immunogenic and elicits an inflammatory response due to leakage of cytosolic contents. It was considered the death mode of cells which displayed no characteristics of apoptosis. In most cases necrosis affects not a single cell but spreads over a group of cells, as in gangrene or ischemia. Morphologic features of necrosis are listed in Table 1. At the biochemical level, necrosis is accompanied by a massive production of reactive oxygen species and reactive nitrogen species, besides a marked drop of cellular ATP. 35 About 10 years ago, studies on genes that could control necrosis led to the conclusion that a regulated form of necrosis must exist. Regulated necrosis ("necroptosis") can occur as the result of activation of death receptors, for example, by TNF, first apoptosis signal ligand, or TRAIL, 42 and is controlled by two key regulators:TNF receptor-associated factor 2 and receptor-interacting protein kinases 1 and 3. 35,43 Besides the activation of death receptors, necroptosis requires inhibition of the apoptotic signaling. 44 This type of necrosis occurs not only in disease (e.g., in systemic inflammatory response syndrome), but also in normal physiology (e.g., in immunologically silent maintenance of T-cell homeostasis). 45,46 In cancer, necrosis occurs when rapid tumor growth is accompanied by insufficient vascularization or the cancer cell population becomes very dense. 47 It can also be a consequence of successful immunotherapy, for example, with oncolytic viruses. 48 The triggering of nonapoptotic cell death modes, such as regulated necrosis, is currently explored for treatment of apoptosis-resistant cancer cells. 49 However, clinical application of regulated necrosis in cancer treatment has not yet been achieved.

Revised classification of cell death modes
Canonical (morphologic) features of a particular cell death mode can be inhibited while death is only deferred. 15 Under certain circumstances, a dying cell can even switch between different cell death programs, for example, the response to F I G U R E 1 Physiologic, molecular, and morphologic events during the time-course of cell death DNA damage changes from apoptosis to mitotic catastrophe in p53-expressing ovarian cancer treated with cisplatin versus cisplatin and checkpoint kinase 2 (required for checkpoint-mediated cell cycle arrest) inhibitor [50][51][52] or from apoptosis to (secondary) necrosis in conditions of insufficient phagocytosis. This suggests that an interplay and/or a fluidic switch may exist between various types of cell death. 53 Apparently, cell death may differ not only in its main morphologic features but also in biochemical features, cell types involved, and activating mechanisms. Moreover, morphologic features are hardly quantifiable and do not take functional, biochemical, and immunological variables into account. Therefore, scientists have shifted from a morphological to a biochemical classification of cell death. 35 As a consequence, the canonical distinction of three different cell death modes has been revised and expanded to comprise 14 subroutines (see Table 2), of which ten play a proven role in treatment-induced cancer cell death. 15,35,54 These include: apoptosis (divided into: intrinsic caspase-dependent, intrinsic caspase-independent, extrinsic by death receptors, extrinsic by dependence receptors), unregulated necrosis, regulated necrosis (necroptosis), pyroptosis, autophagic cell death, mitotic catastrophe, and anoikis. It is still hotly debated whether some of these processes (e.g., autophagic cell death and mitotic catastrophe) are true subroutines or associated phenomena preceding cell death (for more information, see). 35,55 Furthermore, it is still not clear which of these subroutines predominates in cell death induced by antitumor treatment and which route should be activated for the most effective treatment of a particular type of cancer. 28 Nevertheless, this new classification of cell death allows a better separation of molecular pathways and the linking of pathways to functional consequences.
In order to properly classify cell death, several parameters should be determined since many biochemical processes that were initially considered to be hallmarks of apoptosis appear also in other death modes (Table 2). Despite this complexity, five main biochemical parameters appear to define dying cells: (1) changes of membrane asymmetry (exposure of phosphatidylethanolamine [PE] and phosphatidylserine [PS]), (2) loss of transmembrane potential, (3) permeabilization of the mitochondrial membrane with associated potential changes, (4) increased proteolysis, and (5) DNA fragmentation. We will discuss these in the following chapter.

HALLMARKS OF CELL DEATH
As listed in Table 2, each of the five characteristics of apoptosis occurs in more than one cell death mode. However, the order of their appearance on the scenario of cell death is generally well preserved (see Figure 1).

Changes in membrane asymmetry
The cell membrane is a highly specialized bilayer of asymmetrically distributed phospholipids. In the resting state, cationic phospholipids prevail in the outer, and anionic phospholipids in the inner membrane leaflet. The cell membrane functions as: a barrier (allowing passage of only a selected set of molecules), an organizer (assembling, co-localizing, and controlling activity of signaling components), and a sensor and communicator (processing and conducting signals between the cell and its environment). 56 Multiple cellular activities are accompanied by changes in morphology or An asterisk (*) indicates cell death modes known to apply to therapy-induced cancer cell death, ++ = process strongly increased, + = process increased, − = process not increased, n.d.
composition of the cell membrane. These activities include the regulation of immunity, coagulation and bone formation, for example, by changing the conformation, interactions, localization, and destination of proteins. [57][58][59][60][61] A hallmark of apoptosis is the disturbance of membrane asymmetry, and specifically, the translocation of phospholipids, such as PE and PS, from the inner to the outer leaflet of the membrane. Under basal conditions, PE is predominantly and PS is almost exclusively confined to the inner leaflet of the cell membrane (in erythrocytes, 80-85% and >96%, respectively). 62 Once on the cell surface, exposed PE may regulate actin-dependent blebbing and the formation of apoptotic bodies, 63-65 whereas exposed PS serves as a recognition and docking site, for example, for phagocytes, and facilitates the removal of apoptotic cells. [66][67][68] Although disturbance of membrane asymmetry is a feature of apoptosis, disturbed asymmetry also appears early after activation of other cell death modes, such as anoikis, autophagic cell death, pyroptosis and mitotic catastrophe ( Table 2). [69][70][71][72] In death modes such as necrosis, PE and PS may become accessible only at later time points, when cell membrane integrity has been lost. 73

Phosphatidylethanolamine exposure
PE is a neutral (zwitterionic) molecule which accounts for 40-50% of total membrane phospholipids. 74 Most PE molecules are cone-shaped and do not organize themselves into membrane bilayers in an artificial setting, but rather form monolayers, 75 although PE is kept in bilayer configuration in biological membranes by interaction with other phospholipids. This feature enables PE to "coat" lipophilic regions of membrane proteins and to participate in membrane fusion and fission. In hepatocytes, the presence of PE in the bilayer was shown to result in a less tight packing of the membrane lipids and increased membrane permeability. 76 The dynamics of PE play a role in membrane reorganization during cytokinesis, 77,78 stress and apoptosis, 63,79 and possibly also in hemostasis 80 and the physiology of the mitochondrial inner membrane. 81,82 The appearance of PE on the surface may be a more sensitive biomarker of cell stress than PS, since PE is more abundant than PS and could deliver a stronger signal. 64,82 Moreover, PE is present on the luminal surface of tumor blood vessels. Exposed PE in the vessel wall may represent a biomarker for imaging response to antivascular cancer therapy. 64

Phosphatidylserine exposure
PS is an anionic molecule accounting for 2-10% of the total membrane phospholipids. 83,84 It has a cylindrical shape, which promotes formation of membrane bilayers. However, at elevated pH or [Ca 2+ ], PS can adopt a conical shape to form hexagonal membrane structures. [85][86][87] PS is inhomogenously distributed in the plasma membrane, forming 11 nm clusters. 88 As mentioned above, PS exposure is a hallmark of apoptosis and an "eat me" signal for phagocytosis of dying cells.
Many biochemical assays (e.g., in vitro staining of cells with Annexin A5) use PS exposure as a marker of apoptosis. Since annexin is not able to selectively identify apoptosis, Annexin A5 is then used in combination with propidium iodide to identify necrotic cells from apoptotic cells. Early in apoptosis, 10 6 -10 9 PS molecules become accessible to Annexin A5 after translocation to the outer leaflet of the cell membrane. 89,90 However, PS exposure also occurs in normal physiology. For example, binding of proteins to intracellular PS can localize their signaling pathways to the proximity of the cell membrane (e.g., PS-PKC [protein kinase C] interaction) 91,92 and/or can promote membrane fusion and fission (e.g., PS-synaptotagmin-I interaction). 93 PS exposure plays a role in physiological processes such as cell activation (platelets in clotting cascade, lymphocytes in immune response), membrane fusion in phagocytosis, 94 release of membrane-encapsulated nuclei during maturation of erythroblasts, 95 and cellular stress responses. 96,97 Up to 50% of blood vessels in untreated tumors are positive for exposed PS, likely due to oxidative stress in their environment. 98,99 This fraction generally increases after anticancer treatment. 100 In recent years it has become apparent that different forms of PS play unique and important signaling roles in the cell. Oxidized PS was shown to promote recognition of apoptotic cells by macrophages via interaction with CD36 (cluster of differentiation 36 [fatty acid translocase]) 101 or the bridging protein lactadherin (aka milk fat globule-epidermal growth factor 8 protein, MFGE8). 102 Up to 20% of the PS in neutrophils is endogenously converted to PS with only a single acyl chain lyso-phosphatidylserine (lysoPS), in a -nicotinamide adenine dinucleotide (reduced) oxidase-dependent manner. LysoPS plays a role in the clearance of PS-expressing, nonapoptotic neutrophilic cells. 103 1-Lyso-2-acyl-PS and 1-acyl-2-lyso-PS (PS with deletions of the first or second acyl chain) perform different cellular functions. 104-106 1-Lyso-2-acyl-PS can signal platelet degranulation, mast cell activation, and T-cell growth suppression; and 1-acyl-2-lyso-PS may accompany histamine release from peritoneal mast cells and neuronal differentiation.
However, our understanding of the role of different forms of lysoPS in cancer cell death is still rudimentary.

Mechanism of PE and PS exposure
Currently, there are two models describing PS exposure during apoptosis: a recently proposed model of increased phospholipid vesicle trafficking (involving lysosomes, 107 or bidirectional endosomes 108 ) and a widely accepted model of disturbed phospholipid transport. [109][110][111][112] According to the first model, PS externalization reflects phospholipid vesicle trafficking between plasma membrane and cytoplasm rather than an activity of phospholipid transporters. 108 This model is supported by the finding that PS externalization during apoptosis is derived from a newly synthesized pool, and the rate of PS synthesis is then ∼twofold increased. 113,114 Furthermore, altered lipid packing in shrinking cells can prompt PS exposure. 115 According to the second model, localization of PE and PS is regulated by a common set of transporters, such as scramblases, 116,117 ATP-binding cassette (ABC) transporters, 118 and aminophospholipid translocases. 119 Scramblases carry out Ca 2+ -dependent bidirectional and nonspecific transport of phospholipids, whereas ATP-dependent ABC transporters (floppases) and aminophospholipid translocases (flippases) transport PS and PE appropriately between the two leaflets of the cell membrane, that is, in outward or inward direction. The more specific localization of PS than PE to the intracellular leaflet under baseline conditions may be attributed to the fact that aminophospholipid translocases have a somewhat lower affinity for PE than for PS. It is generally accepted that apoptosis leads to deactivation of aminophospholipid translocases and activation of scramblases and ABC transporters. [109][110][111][112] Scramblases are activated by elevation of cytosolic Ca 2+ , an upstream event in, for example, apoptosis and blood coagulation. However, the identity of the transporters that are activated during cancer cell apoptosis has been the subject of a long debate.
The speed, strength, persistence, and reversibility of the signal are the best-characterized features of PS exposure.
Exposure of PS to the outer leaflet has been shown to occur within a few hours after induction of apoptosis. 120 In human promyelocytic leukemia cells and Jurkat cells (immortalized line of human T lymphocytes) treated with various apoptosis inducers (e.g., anti-Fas antibody or camptothecin), the content of PS in the outer leaflet increased 25-280fold (from <0.9 to >240 pmole/million cells). 67,120 At least an eightfold increase in externalized PS had to be reached to initiate phagocytosis of these cells, which is in line with the threshold model. 120 In myocardial ischemia in mice, PS exposure on apoptotic cardiomyocytes was shown to persist for about 6 hours (hr) after reperfusion. 121 The upstream signaling cascade leading to PE and PS externalization in apoptosis has also been examined. PS exposure is usually accompanied by other molecular events, such as caspase activation, 121-123 cathepsin D activation, 124 perturbed Ca 2+ homeostasis, [125][126][127][128] and PKC activation. 129,130 Whether these processes may occur in parallel or are required in combination to initiate PE and PS exposure is not yet clear. 108 A direct role of caspases in PS exposure during apoptosis has been suggested by the discovery of Kell blood group precursor-related protein 8, which requires a caspase-3 cleavage site to support presentation of PS on the surface of a dying cell followed by phagocytosis. 131 In the human myeloid leukemia cell line KBM7, the P4-ATPases ATPase phospholipid transporting, type 11C and cell division cycle protein 50A were shown to act as flippases and to transport aminophospholipids from the outer to the inner leaflet of the plasma membrane. 132 ATPase phospholipid transporting, type 11C is a caspase substrate. Caspasemediated apoptotic exposure of PS is irreversible and leads to cellular engulfment by macrophages.
PS exposure is not under all circumstances closely related to cell death and phagocytic removal. PS can be exposed by viable cells, but is then likely an insufficient trigger for phagocytosis. 133 However, blocking PS on dying cells can abrogate their clearance by phagocytosis. Therefore, phagocytes recognize cell surface PS on dying cells most likely only within strongly curved membrane areas (i.e., in blebs). However, little is known about membrane morphology surrounding exposed PE and PS and how these phospholipids are engaged by specific receptors, for example, lactadherin. 66,134,135 Furthermore, several tumor cell lines have been identified that lack PS exposure during apoptosis 108 and PS exposure can be reversible. 97

Loss of cellular transmembrane potential
Scrambling processes in early apoptosis reduce the pH of the external membrane leaflet and cytoplasm (acidification), and reduce the energy barrier of the cell membrane (depolarization). 138,139 The mechanism of cytoplasm acidification is not yet completely understood. A change in PS localization during apoptosis may affect the function of H + -ATPases, increase proton (H + ) transport across the cell membrane, and reduce cytoplasmic pH. 140,141 Under basal conditions the cytoplasm has a pH of about 7.2 which decreases by about 0.3 to 0.4 pH units in early apoptosis. This drop promotes the activity of important enzymes involved in cell death, such as proteases and DNase II. 142 A loss of plasma membrane potential can be due to a change in cationic and anionic phospholipid distribution, an altered balance between extracellular Na + and intracellular K + (e.g., impaired function of Na + /K + -ATPase) and export of intracellular Cl − . The impairment of Na + /K + ATPase function in apoptotic cells was shown to be caspase-dependent and coincided with mitochondrial depolarization. 143

Change in mitochondrial transmembrane potential ( m )
Ca 2+ is a very powerful regulator of many biochemical processes. Therefore, its cellular concentration must be tightly controlled. Increases in cytoplasmic Ca 2+ (e.g., caused by calcium release from the endoplasmic reticulum [ER]) can be resolved by mitochondria. 144 Mitochondria are one of the largest stores of intracellular Ca 2+ (after the ER), and centers of cellular energy production by oxidative phosphorylation. The functioning electron transport chain facilitates the creation of an electrochemical gradient ( pH) across the inner mitochondrial membrane and the creation of an MMP (Δ m ). The highly negative charge generated at the inner mitochondrial membrane by oxidative phosphorylation is strongly reduced when cells are energetically compromised and on their way to death. Certain apoptotic stimuli (e.g., ER stressors, death receptors, DNA damage) may cause a mitochondrial Ca 2+ overload and spillage of Ca 2+ into the cytoplasm. Ca 2+ efflux is regulated by the Na + /Ca 2+ exchanger and the permeability transition pore complex formed by proapoptotic Bcl-2 family members. A disturbance in Ca 2+ homeostasis and transition pore formation was shown to result in inhibition of oxidative phosphorylation and electron transport, dissipation of Δ m and/or generation of mitochondrial outer membrane permeability, a decrease in cellular ATP, release of proteins from the mitochondrial intermembrane space, and activation of cytoplasmic Ca 2+ -dependent endonucleases. 145,146 Factors which are then released from mitochondria include ATP, reactive oxygen species, and facilitators of caspase-9 activity, such as CytC, apoptosis-inducing factor, and second mitochondria-derived activator of caspase (see Section 1.1). The release of such factors is thought to be "a point-of-no-return" in the apoptotic cascade. 147,148 Changes in mitochondrial transmembrane potential can be both the cause and a consequence of apoptosis. They are the cause if certain agents induce mitochondrial damage and downstream activation of caspase-9, and a consequence if mitochondria amplify the apoptotic cascade downstream death receptors and caspase-8 has already become activated.
Depolarization (or, in rare cases, hyperpolarization) of the mitochondrial membrane occurs in response to a cellular insult. 149,150 Changes in Δ m are frequently monitored as an indicator of cell viability. Almost each form of cell death results in declined m, either at an earlier or a later stage, but an interesting study has shown that release of certain proapoptotic molecules (such as CytC) may occur in the absence of changes in mitochondrial outer membrane potential. 151

Increased caspase proteolysis
Cell death is frequently mediated by a proteolytic cascade, in which caspases play a pivotal role. Caspases have been demonstrated to cleave as much as 5% of the cellular proteome during apoptosis. 152,153 The caspases are a family of enzymes with the ability to sever a myriad of peptides and proteins at residues C-terminally to aspartate (Asp, D). They contain a catalytic Cys-His pair with Cys285 acting as the nucleophile and His237 acting as the general base to abstract the proton from the catalytic Cys and promote the nucleophile. Caspases recognize and cleave proteins after the tetrapeptide motif Asp-x-x-Asp. The enzymes occur as dimers and are mostly present in the cytoplasmic compartment of the cell.
In the absence of a demand for proteolytic activity, caspases are present in an inactive zymogen form (procaspases).
Upon specific cellular insults, two procaspases are cleaved in a highly controlled manner into two small and two large subunits, assembled into a heterotetramer and activated. By cleaving a specific range of assigned protein substrates, caspases render a controlled loss, gain, functional change, or altered localization of client proteins. This in turn leads to the appearance of typical apoptotic characteristics, such as disturbance of cell membrane lipid asymmetry, cell shrinkage, nuclear chromatin condensation, and DNA fragmentation.
Synthetic caspase-3/7 substrates should consist of at least five amino acid residues. Caspase substrates are selected based on protein primary, secondary, tertiary, and quaternary structure. 152 The design of synthetic caspase substrates is based on the preference of caspases for individual peptide sequences (subsite preference). 157

DNA fragmentation
DNA fragmentation is a major step of cellular disassembly. The process may be induced by cell death-inducing factors (e.g., cytolytic T-cells) or by irreparable errors or damage to DNA (e.g., radiation damage). Genomic DNA can be hydrolyzed either inside or outside a dying cell. 158 DNA hydrolysis occurs at different time points and has a different pattern in different cell death modes.
Cleavage of DNA is executed by certain enzymes, DNA endonucleases, which are also known as DNases. These DNases are divided into three groups: (a) Ca 2+ /Mg 2+ endonucleases (e.g., DNase I and DNAS1L3), (b) Mg 2+ endonucleases (e.g., endonuclease G and DFF40/caspase-activated DNase), and (c) cation-independent/acid endonucleases (e.g., DNase II). The activity of these DNases is controlled by various means, such as protease activation (caspases or serine proteases), poly(ADP ribosylation), phosphorylation, or ubiquitination, and by physicochemical conditions, such as a change of cytoplasmic pH. 142,159 Activation of various DNases results in different DNA fragmentation patterns.
(Inter)nucleosomal DNA fragmentation yielding low molecular weight (MW) DNA fragments ("laddering pattern") almost always accompanies apoptosis. Caspase-activated DNase (present in extrinsic and intrinsic apoptosis) and endonuclease G (present in intrinsic apoptosis) produce various laddering patterns. [160][161][162][163][164] The selection of a certain DNase seems to be stimulus-and cell type-dependent. DNA is processed in two steps during apoptosis. In the early stage, DNA is cleaved into high MW fragments (50-300 kb). Here DNA condensation takes place. Subsequently, these molecules are further broken up into oligonucleosome-sized fragments (repeats of 180-200 bp). 165 Free DNA termini present as a consequence of apoptosis can be detected by a TdT-mediated-dUTP nick end labeling assay. 166 However, DNA breaks detected by this assay need not to be a consequence of apoptosis. The TdT-mediated-dUTP nick end labeling assay cannot discriminate among apoptosis, necrosis, and autolytic cell death.
A more random form of DNA fragmentation, yielding a "smear pattern," is observed in nonapoptotic cell death modes, such as necrosis, or cellular disassembly after phagocytosis. This pattern results from the activity of lysosomal DNases, for example, DNase II.

CELL DEATH IMAGING
Since the mechanisms underlying cell death are complex, the question arises how treatment-induced cell death, for example, in cancer, should be quantified with medical imaging. The majority of tracers monitoring cell death are designed to probe: (1) disturbances in membrane asymmetry, (2) reductions in the membrane energetic barrier, (3) changes in MMP, and (4) activation of apoptotic caspases. Although these phenomena were initially considered hallmarks of apoptosis, similar processes occur in other forms of cell death. Thus, most imaging probes are not selective for one particular form of cell death. Increased uptake of such probes may be the net result of cells dying by various mechanisms.

Exposure of PE
Several imaging probes have been developed to monitor the translocation of PE to the outer leaflet of the cell membrane during apoptosis. A few lantibiotics have been radiolabeled and tested for imaging of exposed PE; these include cinnamycin and duramycin.

Duramycin
Duramycin (PA48009, a peptide of 2,013 kDa and 19 amino acids) differs from cinnamycin by only one amino acid residue: Lys2 → Arg2. 168,169 Duramycin takes its name from being resistant to high temperatures and proteolysis.
Soon after its discovery, duramycin was shown to interact with biological membranes and to have a high affinity (K d , 4-11 nM) to PE. 170 The PE binding is specific and occurs in an equimolar and Ca 2+ -independent manner. 171 Duramycin binding to PE depends on membrane curvature and may alter both the curvature and permeability of the membrane.
The mechanism by which duramycin induces these changes is unknown. 172  The results presented in Table 3 suggest that radiolabeled duramycin but not cinnamycin is suitable for SPECT imaging of exposed PE. However, the tracer has not yet been tested in patients or in healthy human volunteers.

Exposure of PS
Since PS exposure accompanies apoptosis, PS has been extensively studied as a target for the imaging of dying cells.
Thus far, five families of protein or peptide-based PS imaging probes have been employed: Annexin A5, the C 2A domain of synaptotagmin I, lactadherin, PS-binding peptide 6, and bavituximab. Annexin A5 is the only probe that has proceeded to the clinical stage of testing. Imaging data for probes targeting exposed PS are presented in Tables 3 and 4.

Annexin A5
Annexin A5 (earlier called Annexin V or "placenta protein 4") is an endogenous 36 kDa protein which was originally isolated from human placenta. 174 Other tissues, such as endothelial cells, kidneys, myocardium, skeletal muscle, skin, red blood cells, platelets, and monocytes contain lower quantities of the protein. 175 Annexin A5 was identified as a potent anticoagulant which could displace and inhibit coagulation factors from biological membranes. 176  None Seems to bind also to (nonapoptotic) immune cells.
was attributed to a Ca 2+ -dependent interaction with negatively-charged PS molecules on the cell surface. Annexin A5 has no absolute specificity for PS, but binds with lower affinities to other targets, such as PE, 177 membrane products of lipid peroxidation, 178 vascular endothelial growth factor receptor 2, 179 and integrin 5. 180 For this reason, some Annexin A5 binding may be observed even in viable cells.
Although annexin A5 has been extensively tested in experimental animals and in cancer patients (see Table 3), for various reasons the original probe failed to meet clinical expectations 181-183 : 1. The radiolabeling procedures for Annexin A5 are rather elaborate and complex, which has limited application of the radiolabeled probe in a clinical setting.

2.
Since Annexin A5 binds to exposed PS, an annexin scan cannot discriminate between apoptosis and necrosis. This caveat is true for all PS-and PE-binding radiotracers. In a treatment response setting, the lack of specificity is not necessarily a problem, and may rather be an advantage, since PS-and PE-probes can provide a stronger signal than pure apoptosis tracers and both apoptosis and necrosis can be desirable consequences of antitumor therapy.

3.
Since the binding of Annexin A5 to exposed PS is calcium-dependent, fluctuations (or regional differences) of intracellular Ca 2+ concentrations may affect the binding of the tracer. This impact of calcium may result in high intraindividual variability of probe binding and an impaired test-retest reproducibility of annexin scans.

4.
The magnitude of Annexin A5 uptake in target lesions and the target-to-background (or signal-to-noise) ratios of Annexin A5 scans are usually rather low. Low uptake of the tracer may be partially due to poor penetration of Annexin A5 into tumor tissue. Poor image contrast may be caused by slow clearance of radiolabeled Annexin A5 from nontarget regions and blood, and by an increased uptake of the probe in normal tissues after antitumor therapy. In order to address this problem, Annexin V-128 was developed, which shows a significantly lower kidney retention than Annexin A5 and is currently being evaluated in clinical trials.

5.
High nonspecific accumulation of Annexin A5 in the liver and the kidneys makes it hard to detect tumors in the abdomen.

6.
Tracer accumulation in areas far from known tumor sites may indicate the presence of unknown tumors or metastases, but may also be false positives, since Annexin A5 can accumulate in various benign lesions, such as infections and inflammations, capillary haemangioma, platelet-rich thrombi, and unstable atherosclerotic plaques. Uptake of the tracer in such sites could be misinterpreted as indicating the presence of malignant lymph nodes.

7.
The optimal timing of a post-therapy Annexin A5 scan is frequently unknown or uncertain (which is true for all existing cell death-targeting tracers), and a complex protocol with multiple scans may be necessary for correct evaluation of the response of a tumor to therapy. A protocol involving three separate injections of radiolabeled annexin and six whole-body SPECT scans has been proposed for studies in cancer patients, in order not to miss an early response of the tumors to chemotherapy. 184

Annexin B1
Annexin B1 is a PS-binding protein isolated from the pork tapeworm (Cysticercus cellulosae, the larval stage of Taenia solium). The protein has a distinct N-terminus and only 32 to 44% homology to other annexins, including Annexin A5. 185 Radiolabeled Annexin B1 has been tested for SPECT and PET imaging of apoptosis (

Zinc coordination complexes
Zinc-dipicolylamine (Zn-DPA) coordination complexes contain two meta-oriented bivalent zinc cations and were created as mimetics to the domain of Annexin A5 which binds to PS via two bridging bivalent calcium cations. 186 These small-molecule complexes associate with negatively-charged phosphorylated molecules, based on electrostatic interaction. 187,188 PSS-380 has a binding site with high and a binding site with low affinity for Zn 2+ ; coordination of the second Zn 2+ molecule occurs only after association of the probe with the anionic membrane surface. 189

PSS-380 has only been used in an in vitro setting. In vitro and in vivo studies with a similar NIR probe (PSS-794) demonstrated that
Zn-DPA complexes can detect human cells dying by apoptosis or necrosis, and bacterial infections. [190][191][192][193] The small molecular size of zinc coordination complexes could be an advantage and is one of the reasons why PET and SPECT analogues of these compounds were tested for apoptosis imaging (it could, e.g., lead to improved probe entry into tumor tissue). However, a high nonspecific binding of the labeled molecules in healthy tissue 194 and/or a high uptake and retention of radioactivity in liver and intestines 195,196 was found to limit the usefulness of Zn-DPA probes for visualization of cell death. Moreover, since Zn-DPA complexes can bind to all kinds of anionic surfaces, positive SPECT or PET signals may not always reflect exposed PS.

Synaptotagmin I
Synaptotagmin I is a 65 kDa transmembrane protein primarily present in synaptic vesicles where it binds to negativelycharged phospholipids in a Ca 2+ -dependent manner to facilitate vesicle fusion and recycling during neurotransmitter release. [197][198][199] The two cytoplasmic C 2 domains (C 2A and C 2B ) of this protein have homology to PKC. 198,199 These domains interacting with Ca 2+ , phospholipids, and soluble N-ethylmaleimide-sensitive factor attachment protein receptor are involved in membrane fusion during synaptic vesicle cycling. 87  Initial experiments with a fluorescent probe showed that C 2A derivatives had much lower background binding in viable cells than Annexin A5 and were fourfold more specific in imaging cell death. 201 However, since the affinity of C 2A for PS-containing membranes (K d = 20 to 71 nM) is much lower than that of Annexin A5 (K d = 1 to 7 nM), a >50 times higher protein concentration may be necessary for good images. 201 The preclinical imaging results described in Table 4 have indicated that C 2A -based probes are potentially useful for evaluation of antitumor treatment, but have also some drawbacks: 1. High levels of radioactivity in liver, kidney, and abdomen may complicate the evaluation of tracer uptake in these areas, particularly at short intervals after injection. The C 2A domain labeled with 18 F 202 has shown a better clearance profile than the 99m Tc-labeled analogue. 203

2.
Because of the large size of the C 2A molecule, tracer uptake is limited by the rate of diffusion into tissue. Radiochemists could try to produce probes with a reduced size and charge which may show a more rapid tissue entry.

Although in vitro experiments indicated a low background binding of C 2A derivatives in viable cells, target-to-
background ratios of the radiolabeled compounds in the mammalian body were rather unfavorable. These low ratios could be related to a low affinity of the probes to PS-containing membranes. C 2A domain probes with higher specificity and lower nonspecific retention have recently been developed, and as expected, these probes showed improved tumor-to-background ratios. 204

Lactadherin (MFG-E8, milk fat globule epidermal growth factor 8 protein)
MFGE8, a 46 kDa extracellular glycoprotein, is secreted by a subset of macrophages and dendritic cells. As a soluble molecule, it participates in the opsonization of apoptotic cells and their phagocytosis, adhesion between sperm and the egg coat, repair of intestinal mucosa, mammary gland branching, morphogenesis, and angiogenesis. 205 The protein acts as a potent anticoagulant in blood 206 and was linked to Alzheimer's disease and autoimmunity. It is a bridging molecule between apoptotic and phagocytic cells, has the ability to bind to integrins ( v 3 and v 5 ) on immune cells via its arginylglycylaspartic acid motif of the glutamic acid-leucine-arginine domain, 207 and also binds to membrane PS on apoptotic cells (with preference to PS in membrane areas of spiky morphology). The binding to membrane PS occurs via its F5/8-type C1 and C2 domains (Kd C1C2 = 4.9 nM, Kd C2 = 2.0 nM) and does not require Ca 2+ . The C2 domain has about 100-fold lower affinity toward soluble than membrane PS (Kd = 2.8 M) and has a much higher affinity toward phosphatidyl-L-serine than phosphatidyl-D-serine. 208 Despite functional similarity of the C-domains present in synaptotagmin-1 and lactadherin, they do not share any sequence homology, 207 but there is homology between the C2 domains shared by lactadherin and blood coagulation factor VIII and V. 209,210 An in vitro study showed that lactadherin can specifically detect PS and has a higher affinity for PS than Annexin A5. 211 Imaging data for a SPECT tracer based on bovine lactadherin are presented in Table 4.
Lactadherin binding to apoptotic HL60 cells was reported to be related to PS exposure and not to an interaction of the probe with integrins. 212

PS-binding peptides
Using phage display technology, peptides were identified which can bind with considerable affinity to exposed PS. Clusters of the basic amino acids Arg (R) and Lys (K) appeared to be critical for (ionic?) interaction with this phospholipid.
A peptide called PSBP-6 has been radiolabeled for SPECT and PET imaging. The amino acid sequence of this peptide is based on the 14-amino-acid sequence from the C2 domain shared by PKC, PS decarboxylase, and synaptotagmin I. 215 PS-binding peptides are in theory an attractive alternative to PS-binding proteins such as Annexin A5. The procedures for radiolabeling of peptides can be simpler, and the radioactive probes may show a more rapid entry into tumor tissue because of their smaller size. This reduced size can also result in a more rapid clearance of unbound probe from tissue and from blood. Moreover, peptides can be structurally modified, in order to improve their pharmacokinetic properties and metabolic stability. However, the currently available PS-binding peptides seem to have insufficient affinity [216][217][218][219][220] and/or specificity 221 for their target phospholipid (see Table 4).

Bavituximab family of antibodies
An indirect option for imaging of externalized PS is provided by the generation of antibodies for 2 -glycoprotein 1.
This protein is abundant in plasma and was shown to bind to negatively charged compounds, such as heparin, anionic phospholipids, and dextran sulfate. Two molecules of 2 -glycoprotein 1 are required for the interaction with PS (Kd ∼ 1 nM). 222 Several murine monoclonal antibodies (e.g., 3G4 and 2aG4), 99,223-225 a chimeric monoclonal antibody (mAb) (bavituximab), 226 and a human mAb (PGN635) 227 were generated to detect PS exposure on tumor vessels. All of these antibodies have been explored preclinically and in clinical trials for treatment of different types of malignancy. Radiolabeled bavituximab, PGN635, and PGN650 have been used for noninvasive in vivo imaging of PS exposure (Table 4). PGN635 is a first-in-class PS-targeting fully human mAb. The F(ab') 2 fragment of PGN635 was used to produce PGN650, which has similar affinity for PS-2 -glycoprotein 1 complexes as 3G4 and bavituximab. 227 In an animal model of human prostate tumors, 74 As-bavituximab displayed very high tumor-to-muscle ratios and specific binding in the tumor (

'Betabodies'
'Betabodies' are fusion products based on the PS-binding domain(s) of 2 -glycoprotein 1 and the constant region of an antibody. 228 The recombinant 'betabody' KL15 is expressed in a dimeric form and consists of the domain I and V from 2 -glycoprotein 1 fused with the CH2 and CH3 constant (F ) domains of a mouse IgG2a antibody. Only a few preclinical data concerning this probe have been published (see Table 4).

ApoSense family
The ApoSense family ( Figure 2) is a group of small-molecule compounds (size 300 to 700 D) that can be used to detect altered membrane permeability in apoptotic cells. The family comprises two different generations of molecules. ApoSense molecules were initially thought to detect both apoptotic and necrotic cell damage, but later studies have suggested that they specifically accumulate in apoptotic cells. 230 Since ApoSense family members can cross the intact blood-brain barrier, they can be used to image the response of brain tumors to treatment, and loss of neurons after stroke or neurodegeneration in diseases like Alzheimer's disease. ApoSense compounds accumulate in the cytoplasm. 231 Correlation between the in vitro uptake of DDC and Annexin A5 has suggested that scrambling processes in early apoptosis reduce the energetic barrier of the cell membrane and allow DDC to enter the cell. DDC uptake is thought to be the result of the following sequence of events: Scrambling → membrane acidification → (mono)protonation of ApoSense molecules → flip-flop of the molecule through the membrane by active scramblases and cell membrane depolarization → binding of the molecule to cytoplasmic proteins.
However, this proposed mechanism is not yet fully supported by experimental data. Imaging results acquired with ApoSense probes are summarized in Table 5.
Advantages of the Aposense family of compounds are: their small molecular size, the minimal number of functional groups, and the absence of chemically reactive, undesired labeling sites. 232 Disadvantages are: the rather poorly defined mechanism of uptake and the requirement of a high administered dose. This last aspect raises concern about potential toxicity, since the dose is in the therapeutic rather than the tracer range. Some findings in animal models have suggested that the uptake of ML-10 is pH-sensitive. 233 If ML-10 uptake is indeed dependent on protonation, a decreased pH of the blood (e.g., due to failure of multiple organs after anti-Fas antibody treatment) may result in a high nonspecific uptake of ML-10 in viable tissues, whereas an increased extracellular pH (e.g., due to cyclophosphamideinduced necrosis in treated tumors) could be associated with a decrease of ML-10 uptake. Such factors may complicate the interpretation of PET images acquired with [ 18 F]ML-10.

Changes of mitochondrial transmembrane potential
Several lipophilic phosphonium cation-based tracers (arylphosphonium salts) have been developed for in vivo imaging of treatment-induced changes of MMP (Δ m ). 234 Loss of negative charge at the inner mitochondrial membrane leads to reduced uptake of these lipophilic cationic tracers. Thus, radiolabeled arylphosphonium salts will generate a negative contrast.

[ 18 F]fluorobenzyl triphenyl phosphonium
[ 18 F]fluorobenzyl triphenyl phosphonium (FBnTP) accumulates in cells with normal mitochondrial potential and washes out when this potential is impaired by apoptosis. When the baseline uptake of the tracer in tumor tissue is low, another imaging modality must be used for tumor localization. 235 The signal of the tracer has been reported to be stable up to 45 min after injection. 236

[ 99m Tc]sesta-methoxyisobutylisonitrile
The SPECT perfusion tracer [ 99m Tc]sesta-methoxyisobutylisonitrile (mibi) has been tested as a probe of reduced membrane potential in dying cells. An early study reported that the uptake of this tracer in human breast cancer cells (MCF7) was reduced when cells were treated with a cytostatic agent (sodium phenylacetate), and the decline of tracer uptake was correlated to the fraction of apoptotic cells. 243 Another study reported that tumor uptake of [ 99m Tc]sestamibi was dose-dependently reduced in mice bearing Ehrlich carcinomas that were subjected to radiotherapy. At 24 hr after irradiation, tumor-to-background ratios were inversely correlated with apoptosis index and the percentage of necrotic area, but at longer intervals (72 hr and 144 hr post irradiation) these ratios were inversely correlated only with the percentage of necrotic area. 244 Although this study confirmed that [ 99m Tc]sestamibi is a "negative contrast tracer of dying cells," another investigation performed in the same year showed that the absolute uptake values of [ 99m Tc]sestamibi in carcinomas are six-to eightfold smaller than those of a phosphonium cation like TPP. 238 Thus, [ 99m Tc]sestamibi scans will show a considerably lower signal-to-noise ratio than TPP scans. tumors. 387 Mouse model of chemotherapy-induced enteropathy. 388 Uptake is specific for cells involved in apoptosis. 388 None (fluorescent probes are only suitable for studies in cells and experimental animals).
Detects response of tumors (and rapidly dividing cells) to chemotherapy (uptake up to sevenfold increased).

F Membrane permeability
Jurkat cells. 389 LY-S (mouse lymphoma cell line) tumors. 389 Only results for fluorescent probe reported, not yet for the PET probe. 390 None.
Detects response of tumors and tumor cells to radio-and immunotherapy (uptake up to 12-fold increased).

NST-729 (fluorescent)
Membrane permeability Mouse models of Alzheimer's disease and ALS 391 None (fluorescent probes are only suitable for studies in cells and experimental animals).
Co-localizes with amyloid plaques in Alzheimer's disease and regions with axonal apoptosis in ALS. None (ML-9 cannot be labeled with a positron emitter. Its alkyl chain was modified in order to allow such labeling, resulting in the derivative ML-10).

ML
Detects response of tumors and tumor cells to chemo-and immunotherapy (uptake up to 10.6-fold increased). ALS, amyotrophic lateral sclerosis; Jurkat, immortalized line of human T lymphocytes.

ML
In summary: PET and SPECT probes of mitochondrial transmembrane potential have shown limited success. The uptake of such tracers is affected by the activity of transporters involved in multidrug resistance and by changes of the physical properties of target tissue. Changes in the uptake of such probes after antitumor therapy may not always reflect changes in mitochondrial transmembrane potential of tumor cells.

Increased proteolysis
Extrinsic and intrinsic apoptotic pathways converge at the level of caspase-3 and caspase-7 activation. The detection of activated caspases could be a valuable and specific tool for identifying dying cells before morphological features of cell death occur. Quantitative imaging of activated caspase-3 and -7 may be more useful for monitoring tumor responses to therapy than for diagnosis and localization of unknown tumors. In vivo imaging of activated caspases is possible via two different approaches: 1. use of caspase inhibitors (Z-valine-alanine-DL-aspartate or isatin-derivatives, for example, [ 18 F]WC-II-89) 245 ; and 2. use of caspase substrates (Z-aspartate-glutamate-valine-aspartate-derivatives, for example, [ 18 F]CP18). [246][247][248] The main benefits of radiolabeled substrates over radiolabeled inhibitors are (in theory): (a) no problem of saturation of the binding sites, and: (b) signal amplification. Since a single enzyme molecule can convert several substrate molecules within the time frame of a PET or SPECT scan, the use of a substrate may result in a higher sensitivity for the detection of an active enzyme. However, in a comparative study between a caspase substrate and activity-based probes (inhibitor-based), signal amplification at the site of proteolysis did not have a dramatic enhancing effect on imaging. The authors believe that this was due to slow diffusion of the substrates into tissues and cells. 249 In another study with inhibitor-based probes, the abundance of active proteases in tumor tissues was found to be sufficient for the generation of images with acceptable contrast, therefore no saturation of binding sites occurred. 250

Caspase inhibitors
Radiolabeled inhibitors bind to a finite number of sites resulting in saturability of the probe binding. [251][252][253] The amount of accumulation is dependent on the ratio of the concentration of active caspases and the affinity of the inhibitor for these caspases (B max /K d ). The addition of a sulfonamide group confers isatins (i.e., derivatives of 1H-indole-2,3-dione) a high affinity for caspase-3 and -7. 254 The chemical structures of some isatin-based caspase inhibitors are shown in Figure 3, whereas imaging results acquired with these tracers are summarized in Table 6.
Radiolabeled isatins have been shown to bind specifically to activated caspases, but their sensitivity as PET probes was limited. Injected isatins can be trapped in blood (either due to apoptosis in lymphocytes, or to released, circulating caspases).
Further optimization of the pharmacological properties of isatin-based caspase inhibitors seems therefore necessary, but unfortunately, literature indicates that the list of chemical alternatives for existing caspase-3/-7 tracers is almost exhausted.

F I G U R E 3
Chemical structures of radiolabeled isatins which have been tested as PET probes for caspase-3

Caspase substrates
The cellular trapping of radiolabeled caspase substrates is less sensitive to competition by physiological substances than the binding of radiolabeled caspase inhibitors, but intracellular retention of the cleaved substrate is necessary for successful imaging. In the first attempts at probe development, a Tat sequence (e.g., Tat49-57, RKKRRQRRR) was added to ensure cellular uptake. It was demonstrated that insertion of yDEVDG at the C-terminus of Tat was preferable, but the mechanism of uptake which is triggered by addition of that sequence is caspase-independent. 255 An elegant solution to the problem of intracellular retention of the cleaved substrate has recently been provided by the so-called "smart probes" which display intramolecular macrocyclization and in situ nanoaggregation upon activation by caspase-3. [256][257][258] Due to sequence homology among the caspases, most caspase probes are not specific for caspase-3 or caspase-7. However, recent research on activity-based probes has shown that the selectivity of such probes for a single caspase can be greatly improved by introducing several unnatural amino acids in the peptide recognition sequence. [259][260][261][262] Imaging results acquired with radiolabeled caspase substrates are presented in Figure 4 and Table 6. Although the preclinical data presented in Table 6 (particularly those of [ 18 F]CP18) have indicated that it is possible to image apoptosis and therapy-induced increases of apoptosis with a radiolabeled substrate for caspase-3, concentrations of radioactivity in target tissues were usually very low. Thus, the currently available caspase substrates seem to have not fulfilled their promise of significant signal amplification with respect to radiolabeled caspase inhibitors.

DNA damage and repair
As explained in Section 2.5 of this review, fragmentation of DNA is a process which accompanies both apoptosis and   Table 7 and DNA damage and repair will not always lead to cell death. Thus, PARP-1 inhibitors will have a limited specificity for dying cells.

Phosphorylated X isoform of the histone H2A ( H2AX)
When double-strand breaks in DNA occur, the X-form of histone H2A (H2AX) is phosphorylated ( H2AX) and several hundreds of phosphorylated protein molecules accumulate around each break site. The formation and accumulation of H2AX is necessary for recruitment and activation of the subsequent processes of DNA repair. The expression levels of H2AX are very low under normal physiological conditions, but show a strong and rapid rise after the induction of DNA damage. For this reason, H2AX is an attractive target for SPECT and PET imaging. Imaging of this target may be used to visualize the impact of antitumor therapy.
Anti-H2AX antibodies can be used to quantify phosphorylated H2AX in permeabilized or lysed cells, but are not useful in living cells since such antibodies do not cross intact cell membranes. However, when the antibodies are linked to a cell penetrating peptide ("TAT sequence"), they are internalized in living cells and targeted to the nucleus (see Table 7).
A recent review on imaging of the DNA damage response 263 concluded that several important issues still need to be addressed before anti-H2AX-TAT antibodies can be applied in clinical studies:

1.
A humanized version of the antibodies should be prepared, since the preclinically tested antibodies were raised in rabbits and will cause an immune response when they are injected in humans;

2.
Since the currently used H2AX-TAT antibodies have a rather high nonspecific in vivo binding, it may be necessary to improve the target-to-nontarget ratio of these probes, for example, by using smaller antibody fragments rather than full antibodies, or by the application of a pretargeting strategy;

3.
Quantification of the exact number of DNA double strand breaks may be difficult, since the local increase of H2AX is not directly or linearly related to the number of strand breaks. More information about the biology of H2AX is required to properly interpret PET or SPECT images acquired with anti-H2AX-TAT. 263

Ataxia telangiectasia and Rad3-related threonine serine kinase
Another important enzyme involved in the initiation and orchestration of the repair of DNA damage is ataxia telangiectasia and Rad3-related threonine serine kinase (ATR kinase). A radiolabeled analog of the ATR kinase inhibitor Ve-821 has been prepared but the results were disappointing (Table 7). Apparently, the pharmacokinetic properties of radiolabeled ATR kinase inhibitors need to be improved before they can be applied as PET tracers.

Other processes involved in cell death
Several imaging probes have been developed which may visualize necrosis. Imaging findings concerning these probes are summarized in Table 8 and the chemical structures of some probes are shown in Figure 6. The probes in question targeted the following processes:

Exposure of histone H1
Apoptosis-targeting peptide-1 (ApoPep-1), a hexapeptide identified by phage display, binds in a Ca 2+ -independent manner to histone H1, which is exposed by apoptotic cells or becomes accessible in the nucleus of necrotic cells. 264 The translocation of histone H1 during apoptosis proceeds in a caspase-dependent manner and occurs at the early stage of apoptosis (before DNA fragmentation). The R3 residue was shown to determine binding and the ApoPep-1 sequence was homologous to the G-protein-coupled receptor 83.

Redistribution of La autoantigen
La autoantigen is a nuclear protein with an MW of 47 (or 48) kDa which is overexpressed in cancer cells with respect to cells of the tissue of origin. The La protein is cleaved by caspase-3 during apoptosis, resulting in translocation of the NH 2 terminus part of the molecule (MW 43 kDa) to the cytoplasm 265 and accessibility of this part to anti-La antibodies. 266 Since the expression of the La autoantigen is selectively induced in dead or dying cancer cells after DNA-damaging chemotherapy, imaging of this target is an interesting strategy for the detection of tumors and the evaluation of anti- Duodenal drainage catheter is required to reduce the intestinal radiation dose. [458][459][460] The dose to thyroid and lungs is then still considerable. 461 Intratumoral tracer administration has been proposed to circumvent these problems. 462 Formation of aggregates should be avoided by addition of PEG400 or sodium cholate. [463][464][465][466] Various hypericin analogs with better solubility have been proposed. 454,[467][468][469] Uptake mechanism is poorly defined.
Uptake mechanism is poorly defined.

F I G U R E 6
Chemical structures of some compounds which have been used to target tissue necrosis

Accessibility of myosin
Radiolabeled Fab fragments of monoclonal antibodies against myosin ([ 111 In]antimyosin) have been widely used for the detection of myocardial cell injury and necrosis. Membrane disruption of myocytes makes it possible for such fragments to enter the dying cell and to interact with myosin heavy fragments.

Exposed histones
[ 99m Tc]Glucarate ([ 99m Tc]-D-glucaric acid) is a six-carbon dicarboxylic acid with a structural similarity to fructose. This SPECT tracer has been reported to accumulate in areas of acute ischemic injury where necrosis occurs, both within the brain 268 and heart. [269][270][271][272] For this reason, [ 99m Tc]glucarate has also been tested as a cell death tracer in animal models of human cancer and in cancer patients (see Table 8).

Extracellular DNA
Hypericin is a red pigment with antraquinone-like structure (MW 504 Da), which has been isolated from St. John's wort (Hypericum perforatum). Hypericin has been tested in many studies as a photosensitizer for photodynamic therapy.
Since the compound accumulates in necrotic cells and tissues, hypericin has also been radioiodinated or labeled with 64 Cu for the imaging of tumors and infarctions in experimental animals and humans ( Table 8). Because of its polyphenolic polycyclic structure, hypericin has fluorescent properties and the compound can be detected in cell or preclinical experiments by optical imaging. 273-275

Unknown target (Pamoic acid derivatives)
The bis-DTPA derivative of pamoic acid (4,4 ′ -methylenebis[3-hydroxy-2-naphtoic acid]) is a necrosis avid contrast agent. The mechanism underlying accumulation of this compound in necrotic tissue is unknown. 276 Various derivatives of pamoic acid have been radiolabeled and evaluated for visualization of necrosis with SPECT or PET (Table 8).
Unfortunately, most necrosis-targeting probes seem to lack adequate specificity (see Table 8). They may accumulate in tissues by mechanisms unrelated to cell death (e.g., inflammation, ischemia, hypoxia, or hypoglycemia), and the uptake mechanism of these probes is poorly defined. Only the peptide ApoPep-1 seems to deserve further evaluation.

CONCLUSIONS AND PERSPECTIVES
Although a large number of PET and SPECT probes for imaging of cell death have been developed, only a few radiopharmaceuticals have proceeded to the clinical stage of testing, viz. radiolabeled Annexin A5, PGN650, ML-10, CP18, antimyosin antibodies, glucarate, and hypericin. Of these seven, the first four are the most likely candidates for translation to the clinic, and results of ongoing clinical trials with Annexin V-124 and PGN650 are eagerly awaited.
An important issue concerning cell death imaging is the question whether radiopharmaceuticals should be specific for a particular death mode and biochemical process (e.g., activated caspase-3 or caspase-7), or can have limited specificity (e.g., detect exposed PS or anionic phospholipids). The required specificity will probably depend on the intended use of the tracer. In a basic science setting (visualizing of dying cells in animal and in vitro models of human disease), specificity of the used probe is very important in order to acquire specific information about the mechanisms underlying cell death (apoptotic vs. nonapoptotic, noninflammatory, or pro-inflammatory, etc.). However, in a clinical setting (assessment of a patient's response to antitumor treatment), specificity of the probe may be of less importance. In this case, a probe with limited specificity that provides a stronger signal than a specific probe may be preferred. Here the main question to answer is whether cells have died. The question via which mechanism cell death was induced is then only a secondary issue.
In the extensive work performed with radiolabeled Annexin A5, two important difficulties were noted which will be of general concern in treatment response evaluation with any cell death tracer: (i) since the optimal timing of a post-therapy scan is frequently unknown or uncertain, a complex (multi-scan) protocol may be required for correct evaluation of tumor responses, and (ii) increases in cell death occur rapidly after the onset of therapy and correlate with early tumor shrinkage, but the magnitude of this early response to treatment is not always predictive for the long- information about this subject, since valid predictive tools will allow clinicians to change therapy in nonresponding patients at an early stage, avoiding unnecessary toxicity and increasing treatment efficacy.
Since a limited probe entry into tumor tissue was frequently encountered in previous research (probably due to a large molecular size of the probes), radiolabeled protein domains or antibody fragments may be more promising as tracers than full-length proteins or antibodies. Some novel potential tracer candidates have been identified in recent years, but have not yet been widely explored for PET and SPECT imaging. These include the Tim family of proteins which bind to PS via their IgV domain (but show a higher affinity to oxidized PS) 277 ; Bai-1, which binds to PS via thrombospondin domains 278 ; and sRAGE, which binds PS via a V-type domain. 279 Other possible candidates are: antibodies against CXCL1, which is released during the unfolded protein response, 280 the high mobility group box 1 (HMGB1) protein,which interacts with PS in an integrin-dependent manner, 52 and imaging of granzyme B, which may be a predictive biomarker of immunotherapy response. 281 The already wide field of cell death imaging may thus expand even further in the near future.