Regulated cell death in cisplatin-induced AKI: relevance of myo-inositol metabolism

Abstract

Regulated cell death (RCD), distinct from accidental cell death, refers to a process of well-controlled programmed cell death with well-defined pathological mechanisms. In the past few decades, various terms for RCDs were coined, and some of them have been implicated in the pathogenesis of various types of acute kidney injury (AKI). Cisplatin is widely used as a chemotherapeutic drug for a broad spectrum of cancers, but its usage was hampered because of being highly nephrotoxic. Cisplatin-induced AKI is commonly seen clinically, and it also serves as a well-established prototypic model for laboratory investigations relevant to acute nephropathy affecting especially the tubular compartment. Literature reports over a period of three decades have indicated that there are multiple types of RCDs, including apoptosis, necroptosis, pyroptosis, ferroptosis, and mitochondrial permeability transition-mediated necrosis, and some of them are pertinent to the pathogenesis of cisplatin-induced AKI. Interestingly, myo-inositol metabolism, a vital biological process that is largely restricted to the kidney, seems to be relevant to the pathogenesis of certain forms of RCDs. A comprehensive understanding of RCDs in cisplatin-induced AKI and their relevance to myo-inositol homeostasis may yield novel therapeutic targets for the amelioration of cisplatin-related nephropathy.

INTRODUCTION

Cisplatin, also known as cis-platinum, cis-diamminedichloroplatinum (II), and platinol, is a commonly used chemotherapeutic agent to treat various types of malignancies, including germline cancers, neoplasm of the prostate, bladder, and lung, and lymphoma (1–4). Interestingly, since its discovery, over a century’s trials and efforts were required to refine its clinical use. Cisplatin was initially discovered by an Italian chemist, M. Peyrone, in 1844 (described as Peyrone’ chloride), and its molecular configuration was thoroughly documented by a Swiss chemist, Alfred Werner, in 1893 (1). However, it took another seven decades to gain the recognition of its therapeutic implications. In 1965, Rosenberg et al. (5) noted that Escherichia coli cultured in the presence of an electrolysis product of platinum electrodes underwent morphological filamentation and growth inhibition, which led to its amenability for therapeutic usage afterward (6). Finally, cisplatin was approved for the treatment of ovarian and testicular neoplasms by the US Food and Drug Administration in 1978, and since then it has been recommended as a first line of chemotherapeutic agent to treat various malignancies.

The clinical use of cisplatin is largely limited by its high degree of nephrotoxicity. It is estimated that ∼30% of cisplatin-treated patients (a single dose at 50–100 mg/m²) exhibit notable nephrotoxic manifestations (7). In the laboratory settings, cisplatin nephropathy serves as a classical prototype for acute kidney injury (AKI)-related investigations due to the readily discernible severity of renal damage. However, it should be kept in perspective that the murine models of cisplatin-induced renal injury may not recapitulate the clinical events since the dose for mice (10–30 mg/kg) is much higher than what is used for its clinical usage (8). In this regard, repeated low-dose administration of cisplatin to mice is readily accepted to recapitulate the clinical settings (9, 10). Besides, mice used in laboratory investigations are usually healthy and young with good organ functions to handle other associated potential nephrotoxins, which may not be case in the clinical settings where patients may already be in a debilitated state (8). The selective organ (kidney) damage pattern of cisplatin may be partly related to its drug metabolism and pharmacokinetics. Cisplatin uptake experiments with kidney slices revealed that its concentration in renal tissues was five times higher than that of the culture medium following ∼4 h of exposure (11). Cisplatin, after freely traversing across the glomerular filter, is handled by certain efflux transporters localized at the apical surfaces of renal tubular epithelia (12, 13). Most cisplatin in the glomerular ultrafiltrate is reabsorbed by proximal tubules, among which the S3 segment tubules efficiently handle majority of the resorptive load, and such a localized buildup would likely contribute toward nephrotoxicity (14). Organic cation transporter 2 (OCT2) and copper transporter 1 (Ctr1) are the major transporters participating in the reabsorption of cisplatin in renal tubules, and genetic or pharmaceutical inhibition of these two transporters could provide protection against cisplatin-induced nephrotoxicity (15–17).

It is now well established that cisplatin [PtCl₂(NH₃)₂] is a square planar inorganic complex (1). The molecular structure of cisplatin is characterized by a centralized platinum ion encircled by two amine groups on one side and two chloride radicals on the other side (Fig. 1A). Conceivably, the interactions between the platinum ion and two chloride radicals are unstable in an intracellular environment, which apparently allows the sequential replacement of chloride radicals by water molecules (Fig. 1A) (18). Hydrated cisplatin gets activated and crosslinks with nucleophilic N7 sites of the purine bases in the DNA, followed by the generation of DNA:DNA adducts, mostly via intrastrand crosslinking (Fig. 1B) (1, 19). These crosslinks lead to DNA damage by dampening DNA synthesis and replication, which ultimately results in cellular demise (20). This DNA damage has been implicated as the major cause of cisplatin being endowed with anticancer properties and unfortunately also the side effects, such as nephrotoxicity (20, 21). However, questions have been raised regarding its role in predominant DNA damage in cisplatin-induced AKI, since only ∼1% of intracellular cisplatin is translocated into the nuclear compartment (22). Besides, renal proximal tubular cells, the primary target of cisplatin-related nephrotoxicity, are the quiescent cells that are devoid of active DNA replication. Other mechanisms, such as oxidant stress, inflammation, vascular dysfunction, and mitochondrial damage, are probably of much greater significance in the pathogenesis of cisplatin-induced AKI (23–26). In recent years, various forms of regulated cell death (RCD) have been described in the literature, and some of them have been implicated in cisplatin-induced AKI (Fig. 2). In addition, the pathogenesis of RCDs has been reported to be conceivably modulated by the metabolism of myo-inositol, a unique carbocyclic sugar. Here, we discuss the available information in the previous investigations on various forms of RCDs in cisplatin-induced AKI and their relevance to myo-inositol metabolism, which may provide unique insights into the novel pathogenetic mechanisms, and of course would eventually facilitate in the development of therapies for cisplatin-induced AKI.

Figure 1.

The structure, hydration, and DNA binding of cisplatin. A: cisplatin is a square-planar complex with a platinum ion in the center; the two chloride ligands can be replaced by H₂O in two successive steps. B: cisplatin binds with N7 sites of purine bases (A or G) in DNA, leading to DNA damage. [Partially adapted from Ref. (19).]

Figure 2.

Top: denomination. Bottom: timeline of the events for the five regulated cell deaths related to cisplatin-induced acute kidney injury (AKI). MPT, mitochondrial permeability transition.

APOPTOSIS: THE CLASSIC FORM OF RCD

The term apoptosis was first coined by Kerr et al. (27) in the early 1970s to differentiate it from necrosis (Fig. 2). It is the most extensively investigated RCD in various forms of AKI. Mechanistically, apoptosis is caspase dependent, and its executionary phase mainly relies on the activation of effector caspases, including caspase-3, caspase-6, and caspase-7 (Fig. 3) (28, 29). Apparently, the caspase-3, caspase-6, and caspase-7 complex, and especially caspase-3, then activates caspase-activated DNase (CAD) by proteolytically cleaving the inhibitor of CAD (ICAD) (30). Normally, CAD is bound to ICAD and remains detached with genomic DNA (30). However, in an apoptotic scenario, ICAD gets cleaved and is detached from CAD, which is followed by the dimerization and activation of CAD (30, 31). Activated CAD then binds to genomic double-stranded DNA, culminating in DNA fragmentation (30). In doing so, the apoptotic bodies are generated during the late stages of apoptosis, and they are phagocytosed by contiguous cells without causing obvious inflammation and damage to the adjacent cells (32), a process that is distinct from necrosis. Interestingly, the process of apoptosis in cisplatin-induced AKI was initially identified by two independent groups of investigators in 1996 (33, 34) (Fig. 2). Lieberthal et al. (33) reported cisplatin-induced apoptosis in proximal tubular epithelial cells with morphological evidence of cellular damage and DNA fragmentation, whereas Takeda et al. (34) presented evidence of considerable DNA fragmentation in the S3 segment of proximal tubules and outer medullary collecting tubular cells following cisplatin treatment. Overall, it seems that protein levels of activated caspase-3 are now considered as one of the major markers for apoptosis in cisplatin-induced AKI (35–37). However, the role of other executionary caspases, i.e., caspase-6/7, in cisplatin-induced AKI has not been adequately investigated.

Figure 3.

Cellular events leading to apoptosis. The initiation of apoptosis includes three pathways. Extrinsic pathway apoptosis is initiated by the binding between death receptor and death ligand, followed by sequential activation of Fas-associated death domain (FADD)/tumor necrosis factor receptor-associated death domain (TRADD), caspase-8/10, and caspase-3/6/7. Intrinsic pathway apoptosis is triggered by the mitochondrial leakage of cytochrome c (Cyt C), which binds to apoptotic protease activating factor 1 (APAF1) in the cytoplasm. APAF1 then leads to consequential activation of caspase-9 and caspase-3/6/7. The leakage of Cyt C is strictly controlled by proapoptotic proteins (BAX, BAK1, and P53) and antiapoptotic proteins (BCL-2 and BCL-XL). Endoplasmic reticulum (ER) stress pathway-related apoptosis is mediated by ER stress sensors. Inositol-requiring protein 1 (IRE1) promotes apoptosis by activating successively caspase-12/4, caspase-9, and caspase-3/6/7. Protein kinase R-like ER kinase (PERK) and activating transcription factor 6 (ATF6) drive apoptosis through C/EBP homologous protein (CHOP), which potentiates death receptor activation and mitochondrial permeability. Apoptosis is executed by caspase-3/6/7, among which caspase-3 cleaves inhibitor of caspase-activated DNase (ICAD) to revitalize caspase-activated DNase (CAD), followed by DNA damage.

Apoptosis can be induced using three different routes, i.e., the extrinsic pathway, the intrinsic pathway, and the endoplasmic reticulum (ER) stress pathway (Fig. 3). Extrinsic pathway apoptosis is initiated when death ligands (such as Fas ligand and TNF-α), secreted by immune cells or tubular cells, bind with the corresponding death receptors [such as Fas and TNF receptor (TNFR)] on the plasma membrane of the targeted cells (38–41). This interaction leads to the activation of two initiator caspases (caspase-8 and caspase-10) via Fas-associated death domain or TNFR-associated death domain, which then activates effector caspases (38). Earlier investigations indicated that there was an increased expression of Fas and Fas ligand following cisplatin treatment in both in vivo and in vitro systems (42). Interestingly, studies have shown that Fas mutation attenuates cisplatin-induced proximal tubular cell death and morphological cellular disruption (43). Also, in support of this notion are studies where TNF-α was reported to be upregulated in cisplatin-treated tubular cells, whereas its inhibition dampened the tubular damage (44). Specifically, genetic ablation of TNF-α conferred the mice with resistance to cisplatin-induced cellular injury (44). However, in the initial studies, it was not clear if the upregulated TNF-α by itself modulates cisplatin-induced tubular injury or in conjunction with the activation of a particular receptor, the interaction with which leads to a calamitous inflammatory response and ultimately acute renal failure (45). In this regard, Ramesh and Reeves (45, 46) contemporaneously identified that it is TNFR2 and not TNFR1 that dominantly mediates TNF-α-induced apoptosis. However, other investigators reported that TNFR1 deletion also alleviated cisplatin-induced AKI (43). Of note, one of the death receptors, i.e., death receptor 3 (DR3), and its ligand TNF-like ligand 1A have been reported to mitigate cisplatin-induced AKI, via alleviating proapoptotic signals (47). Other death receptors, such as TNF-related apoptosis inducing ligand (TRAIL)-R1 and TRAIL-R2, and their ligands have not been investigated in cisplatin-induced AKI. Similarly, the role of caspase-10 in such a cellular damage has not been examined, although the role of death receptors in the activation of caspase-8 leading to tubular damage has been well implicated in cisplatin-induced AKI (43).

Compared with extrinsic apoptosis, intrinsic apoptosis has drawn much more attention in cisplatin-induced AKI. Intrinsic apoptosis is closely associated with outer mitochondrial membrane permeabilization and leakage of mitochondrial proteins, such as cytochrome c, apoptosis-inducing factor, somatic cytochrome c, Smac/DIABLO, endonuclease G, etc. (Fig. 3) (48, 49). This process is strictly modulated by proapoptotic factors (Bax and Bak1) and antiapoptotic factors (Bcl-2 and Bcl-xL) (50). Released cytochrome c binds to apoptogenic protease activating factor 1 in the cytoplasm, and this intricate interaction leads to the assembling of the multimeric apoptosome, where procaspase-9 is recruited, cleaved, and activated (51). Following the activation of caspase-9, there is activation of effector system caspases, i.e., caspase-3/6/7 (51). After the first report regarding BAX upregulation and membrane translocation in cisplatin-treated tubular cells (52), the process of intrinsic apoptosis was extensively investigated during the last two decades. Dong and coworker’s (53) laboratory reported that Bax expression levels were increased in cisplatin-treated mice, and mice with genetic ablation of Bax were partially resistant to cisplatin-induced tubular damage. Likewise, pharmacological inhibition of Bax by nutlin-3 alleviated cytochrome c leakage and cellular disruption following cisplatin treatment (54). Interestingly, both Bax and Bak were upregulated, and they were oligomerized and activated in mitochondria following cisplatin exposure (54). On the other hand, Bcl-2 and Bcl-xL have been reported to be downregulated in cisplatin-treated mice or cells, whereas overexpression of Bcl-2 renders renal tubular cells to be resistant to cisplatin (55–58). Also, p53, a tumor suppressor, exacerbates cisplatin-induced AKI via inhibition of antiapoptotic proteins (Bcl-2 and Bcl-xL) (39).

The third route, i.e, ER stress-related apoptosis, has been poorly investigated in cisplatin-induced AKI. Under basal conditions, the ER is the organelle where proteins are synthesized and processed, in other words, properly folded and packaged (59). However, in pathological states, excessive unfolded/misfolded proteins accumulate in the ER lumen due to its functional disruption, and that leads to an accentuated unfolded protein response (UPR) and thereby causes ER stress (60). When ER stress persists for a long time, glucose-regulated protein 78/binding protein is dissociated from the three key ER stress sensors, i.e., inositol-requiring protein 1 (IRE1), protein kinase R-like ER kinase (PERK), and activating transcription factor 6 (ATF6), and this process downstream facilitates apoptosis (Fig. 3) (59, 60). IRE1 accelerates apoptosis by activating murine caspase-12 or its human homolog caspase-4, followed by cleavage of procaspase-9 and procaspase-3 (60, 61). PERK and ATF6 drive apoptosis via increasing transcription of C/EBP homologous protein, which, in turn, participates in apoptosis via modulating mitochondrial permeability and death receptor activation (60). Overall, one can say that the literature reports, to a certain extent, support the view that these three ER stress sensors (IRE1, PERK, and ATF6) are conceivably activated and upregulated in cisplatin-induced AKI (58, 62, 63). Besides, heme oxygenase-1, an ER-resident enzyme, accelerates apoptosis in cisplatin-induced AKI (7). In addition, activation of downstream molecules, including cleaved caspase-12 and C/EBP homologous protein, were reported to be induced in cisplatin-treated kidney tissues (64, 65). However, all these publications provide somewhat weak evidence for the involvement of ER stress-related apoptosis in cisplatin-induced AKI due to the paucity of data regarding mice with relevant genetic deficiency or lack of specific pharmacological inhibitory agents for these key sensors. Further validation is needed to elucidate a definitive concrete role of the ER stress pathway and apoptosis in cisplatin-induced AKI.

NECROPTOSIS

The term “necroptosis” was initially coined by Yuan and coworkers (66) in 2005 while working on ischemic brain injury (Fig. 2). These authors characterized necroptosis as a process in which the Fas/TNFR family activates a nonapoptotic death pathway in the absence of apoptotic signaling. Later, several investigators noted that necroptosis is an essential process seen in many forms of AKI (67–72). The process of necroptosis includes an array of cascade of events, which culminate in the generation of the necrosome complex and phosphorylation of mixed lineage kinase domain-like (MLKL) pseudokinase (73–75). Phosphorylation of MLKL is mediated by receptor-interacting serine/threonine protein kinase (RIPK)1 and RIPK3 as well as death receptors (Fig. 4, top left) (73, 74). Upon activation of death receptors (notably TNFR1 and Fas), the kinase domain of RIPK1 is activated, which downstream leads to the phosphorylation of RIPK3 (73, 74). In this scenario, RIPK1 and RIPK3 are recruited and assembled into a necrosome complex, where RIPK1 and RIPK3 undergo auto- and transphosphorylation, followed by phosphorylation and activation of MLKL (73, 74). Phosphorylated MLKL gets homotrimerized and undergoes a conformational change, which results in the exposure of its NH₂-terminal 4-helix bundle (NB) domain and central brace region (76–78). Subsequently, MLKL translocates into the plasma membrane, where the central brace region and NB domain interact with polyphosphoinositides, resulting in pore formation in the plasma membrane and ultimately leading to cellular rupture, i.e., necroptosis (76–78).

Figure 4.

Molecular mechanisms pertaining to four types of nonapoptotic regulated cell deaths. Top left: necroptosis is triggered by receptor-interacting serine/threonine-protein kinase (RIPK)1 activation upon binding between death receptor and death ligand. RIPK1 then phosphorylates RIPK3, leading to the assembly of necrosome, where mixed lineage kinase domain-like (MLKL) is recruited and activated. Activated MLKL gets oligomerized and disrupts the cytomembrane, ultimately culminating in cell death. Top right: the initiation of pyroptosis includes two ways. Canonical pyroptosis is induced by the binding of the pattern recognition receptor (PPR) and its corresponding adaptor, which leads to the activation of caspase-1. Caspase-1 then cleaves gasdermin D (GSDMD), IL-1β, and IL-18. Noncanonical pyroptosis is mediated by lipopolysaccharide (LPS)-activated caspase-4/5/11, which also cleaves GSDMD. NH₂-terminal domain of GSDMD (GSDMD-N) gets oligomerized and generates pores in the cytomembrane, resulting in IL-1β and IL-18 leakage and cell death. Bottom left: ferroptosis is executed by overwhelming lipid peroxidation, which is catalyzed by ferritinophagy-induced free iron and lipoxygenase (LOX). The lethal lipid peroxidation can be inhibited by glutathione peroxidase 4 (GPX4) together with GSH. Bottom right: mitochondrial permeability transition (MPT)-mediated necrosis is characterized by generation of the MPT pore (MPTP), which disrupts the inner mitochondrial membrane (IMM) and allows the passage of small solutes [molecular weight (MW) < 1.5 kDa]. The formation of the MPTP is usually induced by increased calcium concentration or oxidant stress, and the MPTP can be inhibited by cyclosporine A (CsA), a potent antagonist of cyclophilin D (Cyp-D). NCOA4, nuclear receptor coactivator 4.

The role of necroptosis in cisplatin-induced proximal tubular injury was originally reported almost a decade ago by Monte and coworkers (79) (Fig. 2). These authors demonstrated that necrostatin-1 (Nec-1; an inhibitor of RIPK1) alleviated cell death following cisplatin exposure to HK-2 cells. Following this seminal observation, several groups of investigators confirmed that the administration of Nec-1 indeed attenuates aberrant renal pathological alterations induced by cisplatin (80–83). Interestingly, the expression of molecules relevant to necroptosis, i.e., RIPK1, RIPK3, and MLKL, was noted to be upregulated in kidneys of mice that received cisplatin (83–86). Other RIPK1 inhibitors, i.e., wogonin and Cpd-71, or gene disruption of the molecules mentioned above also had an ameliorative effect in cisplatin-induced nephropathy (82, 83, 85). However, certain concerns were raised regarding the investigations related to RIPK1 inhibition in necroptosis. First, although Nec-1 is commonly used as a necroptosis inhibitor, it should be noted that it is also involved in the modulation of certain aspects of the immunoregulatory system, which raises the question regarding the specificity of the inhibitory effect (87–89). Second, RIPK1 participates in other biological processes, including inflammation, immunity, and cell survival, besides the downstream activation of RIPK3 (90, 91). In this regard, genetic interference of the downstream molecules, i.e., RIPK3 and MLKL, would be good potential targets for the inhibition of necroptosis. Indeed, Han and coworkers (83) observed that RIPK3 or MLKL knockout mice had a notable alleviation of tubular damage following cisplatin exposure compared with wild-type mice. In line with these observations, Meng and coworkers (92) recently reported that cisplatin-induced HK-2 cell death can be attenuated by the microRNA hsa-miR-500a-3P; the latter binds to the 3′-untranslated region of MLKL and suppresses its phosphorylation and plasma membrane translocation. Other points worth considering in the process of necroptosis include the dose and duration of exposure of the offending agent or inhibitors. In vitro studies have indicated that necroptosis usually occurs following the administration of a high dose of cisplatin or its prolonged exposure at low dosages (83). Coneivably, the prevailing notion is that the first low dose of cisplatin sensitizes mouse kidneys to necroptosis whereas the second dose renders them susceptible to cisplatin-induced injury (93, 94). Likewise, a recovery pattern of injury in necroptosis was observed in the mouse model of folic acid-induced AKI. In folic acid-induced injury, where the injury is usually milder compared with cisplatin, no notable protective effect of Nec-1 administration was observed at day 2 (95). However, in the state of prolonged folic acid-induced renal injury at day 4, the administration of Nec-1/Nec-1s or genetic ablation of RIPK3/MLKL had significant protection against tubular injury (71).

PYROPTOSIS

Pyroptosis, recently redefined as gasdermin-dependent cell death, is a caspase-mediated nonapoptotic RCD, and it was originally described by Cookson and Brennan and other investigators (96–98) in monocyte/macrophage death induced by mounting of an overwhelming bacterial assault (Fig. 2). Strikingly different from apoptosis, the affected cells undergoing pyroptosis exhibit a diffuse pattern of DNA fragmentation, disruption of membrane integrity, and triggering of inflammation, and the process is characterized by the absence of traditional caspase-3 activation (96, 97, 99). The onset of pyroptosis can be triggered by two different routes, which include a canonical inflammasome pathway and the other noncanonical inflammasome pathway (100). In canonical pyroptosis, exposure to different stimuli leads to the generation of various oligomeric inflammasome complexes. They are usually formed following activation of a pattern recognition receptor (PRR) and a downstream adaptor (101, 102). The PRRs sense specific stimuli and transmit the signals to the adaptor, leading to the activation of caspase-1, which is followed by cleavage of IL-1β, IL-18, and gasdermin D (GSDMD; a pyroptosis effector) (Fig. 4, top right) (98, 102). Of note, the PPRs usually undergo oligomerization to increase efficiency of caspase-1 activation (98). Interestingly, multiple types of inflammasomes having their own unique potencies have been identified to recognize a wide variety of stimuli (103). Noncanonical pyroptosis is mediated by murine caspase-11 or its human homolog caspase-4/5 (104–106). Upon exposure of lipopolysaccharide (LPS), caspase-4/5/11 binds to the lipid A motif in LPS, thereafter culminating into its activation (Fig. 4, top right) (105). Subsequently, activated caspase-4/5/11 is oligomerized and exerts its proteolytic effect to cleave the common substrate GSDMD, without involving the inflammatory cytokine IL-1β or IL-18 (98, 104–106).

Here, a brief description of the common substrate, i.e., GSDMD, of canonical and noncanonical pathways may be worth discussion. GSDMD, the executor of pyroptosis, is a member of the gasdermin family, which is characterized by having a highly conserved NH₂-terminal domain (98, 107, 108). To date, six members of gasdermin, i.e., GSDMA, GSDMB, GSDMC, GSDMD, GSDME (DFNA5), and DFNB59, have been identified in humans, whereas there are 12 members in mice (107, 108). GSDMD is the only member that harbors the cleavage site for caspase-1/4/5/11 (108). The NH₂-terminal domain of GSDMD (GSDMD-N) gets oligomerized and binds to phosphoinositides and cardiolipin while being inserted into the plasma membrane (109, 110), which is followed by pore formation, membrane disruption, and cellular lysis. Interestingly, GSDMD only disrupts the plasma membrane from the inside since phosphoinositides are localized in the cytoplasmic leaflet, and not the exoplasmic leaflet, of the plasmalemma (109, 110). The membrane pore formed by oligomerized GSDMD has an inner diameter of 10–14 nm, which readily allows the secretion of IL-1β (4.5 nm) or IL-18 (7.5 nm) (98, 109). In this regard, IL-1β and IL-18, processed by caspase-1 and secreted through pores formed by GSDMD-N, are regarded as sensitive markers for pyroptosis (111–113). Other gasdermins also generate pyroptotic alterations after activation even if no caspase cleavage sites are present, suggesting that there are other potential routes for pyroptosis to ensue (109).

In 2012, Yu and coworkers (114) initially investigated the relevance of pyroptosis in renal injury in a murine model of unilateral ureteral obstructive nephropathy, where they observed increased expression of key pyroptosis markers, i.e., caspase-1 and IL-1β. Subsequently, the pathogenetic role of pyroptosis was implicated in a number of experimental renal injury models, including ischemia-reperfusion injury (IRI)-, contrast-, sepsis-, cadmium-, diabetes-, and calcium oxalate-induced nephropathy (115–120). However, the evidence presented in these reports was somewhat weak and indirect, since they lack the use of specific pyroptosis inhibitors and due to a paucity of pyroptosis-related gene disruption in in vivo studies in various experimental models. However, recently, two independent group of investigators demonstrated a direct pathogenetic role of pyroptosis in cisplatin-induced AKI using GSDMD knockout mice (121, 122). In this regard, Lu and coworkers (122) presented evidence that gene deletion of caspase-11 or GSDMD apparently protect mice against cisplatin- and IRI-induced AKI, suggesting the participation of noncanonical pyroptosis. In addition, Jia and coworkers (121) reported that GSDMD gene ablation is beneficial, whereas GSDMD-N overexpression is detrimental, to the kidney in cisplatin-induced nephropathy, as assessed by aberrant alterations in renal functional and morphological parameters. Also, the anomalous changes observed in IL-1β and IL-18 levels following cisplatin exposure were partially restored following GSDMD gene disruption (121, 122). Along these lines, Edelstein and coworkers (123) observed that caspase-1 knockout mice are resistant to cisplatin-induced tubular injury, suggesting the involvement of canonical pyroptosis, although the relevance of pyroptosis was not alluded to in this publication.

FERROPTOSIS

Almost a decade ago, ferroptosis was described as an erastin-triggered cell death by Dixon et al. (124). Incidentally, erastin is a small molecule that inhibits cystine-glutamate antiporter system Xc−. Ferroptosis is a unique type of iron-dependent nonapoptotic RCD (Fig. 2). Furthermore, Stockwell and coworkers (124, 125) suggested that the executionary phase of ferroptosis is characterized by iron catalyzed-lipid hydroperoxidation, indicating disruption in iron metabolism and associated with various redox perturbations. In fact, the process of ferroptosis encompasses an array of metabolic disorders involving lipids, iron, and amino acids (Fig. 4, bottom left) (125). Lipid peroxidate, the executor of ferroptosis, can be synthesized via Fenton reaction or the lipoxygenase (LOX) pathway (126, 127). Fenton reaction refers to a nonenzymatic catalysis between peroxides and ferrous iron, leading to the generation of highly reactive oxidant radicals (128). These radicals then target bis-allylic hydrogen atoms of polyunsaturated fatty acids (PUFAs), thus resulting in deleterious lipid peroxidation (126). Of note, PUFAs need to be esterified into membrane phospholipids before lipid peroxidation (125, 129). This may mean that a large number of cellular components, including plasmalemma, mitochondria, the ER, and lysosomes would be affected due to having PUFA in their membranes. LOX, a family of the iron-containing enzymes, mediates ferroptosis via catalyzing PUFA-containing phospholipids (130). Many of the LOXs, including LOX5, LOX12, and LOX15, are intimately associated with ferroptosis in a variety of pathobiological processes (131–133). The notion regarding the participation of iron metabolism is well supported by the fact that both Fenton reaction and LOX-mediated enzyme catalysis are iron dependent (126). Along these lines, it is worth mentioning that the onset of ferroptosis induced by various chemicals can be prevented by the administration of iron chelators (125, 134), whereas iron supplementation by ferric ammonium citrate facilitates the induction of ferroptosis, as alluded to in vitro studies (135). In addition, ferritinophagy, a selective autophagy of ferritin guided by its specific cargo receptor nuclear receptor coactivator 4, leads to an increased availability of free iron (136), which is essential for the Fenton reaction. Other iron metabolic sensors, including transferrin and transferrin receptor (iron import), ferroportin (iron export), ferritin and heme (iron storage), iron-responsive element-binding protein 2, heat shock protein-β1, and mitoNEET, iron metabolism regulator/transporter in mitochondria, have been also implicated in various pathobiological processes relevant to ferroptosis (137–143). The iron-catalyzed overwhelming lipid peroxidation damages the cells by membrane disruption and toxicity of lipid peroxides decomposition products (126). Nonetheless, the deleterious process of ferroptosis can be terminated by an enzyme, glutathione peroxidase 4 (GPX4), whereby noxious lipid peroxidates are converted into harmless lipid alcohols in the presence of glutathione (144). Here, it should be noted that two amino acids, i.e., cysteine and glutamate, of the antiporter system X_c⁻ that are pertinent to glutathione synthesis, are relevant to the induction of ferroptosis (145). In fact, the inhibitors of antiporter system X_c⁻ and GPX4 have been widely used as inducers of ferroptosis. Other agents, including NADPH, nuclear factor erythroid 2-related factor 2 (NRF2), heme oxygenase-1, p53, and coenzyme Q₁₀, also have been reported to be relevant in the onset of ferroptosis (131, 144, 146).

In 2014, Linkermann et al. (147) described the role of ferroptosis in AKI in an excellent seminal report, and the authors indicated that ferroptosis, rather than necroptosis and apoptosis, mediated renal parenchymal cellular necrosis following ischemia-reperfusion-induced AKI. Also, other investigators discovered that ferroptosis is also operative in oxalate-, rhabdomyolysis-, and folic acid-induced AKI (95, 147, 148). In a recent report, with convincing in vivo and in vitro evidence, we described the role of ferroptosis in cisplatin-induced AKI (149). In this report, we also indicated that cisplatin treatment leads to excessive lipid peroxidation, ferritinophagy-mediated free iron release, glutathione depletion, and a decrease in the activity of GPX4 (149). Here, it is encouraging to note that our findings were immediately validated by other group of investigators (150, 151). The link between ferroptosis and cisplatin-induced AKI has been only recently appreciated; nevertheless, the investigators of decades-old earlier publications should be credited for pointing out this unique association, although the evidence was somewhat indirect. In 1991, two groups of investigators noted that cisplatin-induced lipid peroxidation in rat kidneys was iron dependent (152, 153). A few years later, Baliga et al. (154) described that cisplatin treatment leads to excessive free iron load in both LLC-PK1 cells (pig proximal tubular cells) and kidneys and that iron chelators protected the tubular cells against cisplatin-induced cellular damage. Afterward, a flurry of articles suggesting perturbations in the iron metabolism induced by cisplatin were reported in the literature by several investigators (155–157).

MITOCHONDRIAL PERMEABILITY TRANSITION-MEDIATED NECROSIS

Mitochondrial permeability transition (MPT)-mediated necrosis refers to a variant of RCD induced by the formation of the MPT pore (MPTP), which disrupts the inner mitochondrial membrane and facilitates the traffic of solutes (molecular weight < 1,500 Da) across the inner mitochondrial membrane (Fig. 4, bottom right) (158–160). These aberrant changes lead to perturbations in the mitochondrial membrane potential, mitochondrial swelling and dysfunction, and eventually cell death (161). Inexplicably, well-defined molecular components of the MPTP remain elusive. Previous publications indicated several candidates for MPTP components, including adenine nucleotide translocase, voltage-dependent anion channel, c subunit of F_oATP synthase, phosphate carrier, and cyclophilin D (CypD) (162–165). However, some of the notions regarding these potential candidates were challenged by other investigators following the observations made in genetically modified mice (166–169). Nevertheless, it is well accepted that CypD, a mitochondrial molecular chaperone, is the key regulator participating in the functionality of the MPTP (170). The idea of possible involvement of the MPTP in necrosis was initially marshalled by the experiments in which cyclosporine A, a CypD inhibitor, was noted to inhibit Ca²⁺-dependent MPT and cell death (171–174). However, this has been somewhat controversial because of lack of specificity since cyclosporine A has the tendency to interact with other cyclophilins and calcineurin, and, moreover, overexpression of CypD has opposite effects on apoptosis and necrosis (172, 175). However, Li et al. (175) provided evidence that CypD-overexpressing cells were susceptible to sodium nitroprusside (nitric oxide donor)-induced necrosis and that their mitochondria were prone to undergo permeability transition (Fig. 2). Interestingly, convincing proof of the role of CypD was obtained by gene deletion approaches in mice by three different groups of investigators. Their elegant work indicated that CypD-deficient cells were more resistant to necrotic cell death induced by oxidant stress or Ca²⁺ overload (159, 176, 177). In addition, they noted that CypD knockout mice were less susceptible to neuronal or cardiac IRI. Of note, as mentioned above, CypD is considered as an essential regulator rather than an integral component of the MPTP, since the functionality of the MPTP was still preserved under higher Ca²⁺ concentration and oxidant stress even under conditions of CypD deficiency (160). MPT-mediated necrosis has been implicated in the pathogenesis of various injury models in several organ systems, including the pancreas, lung, kidney, and liver (87, 178–180). However, due to the limited knowledge, especially related to the constitutional component of the MPTP, the precise mechanisms of MPT-mediated necrosis remain still somewhat elusive.

The role of MPT-mediated necrosis in the kidney was initially implicated in renal IRI, which indicated beneficial effects of CypD deficiency in IRI-induced AKI (181, 182). The research literature related to MPT-mediated necrosis and cisplatin-induced AKI is limited. In this regard, a few years ago, Linkermann et al. (87) observed that CypD-deficient mice had prolonged survival following cisplatin treatment compared with wild-type mice, and these investigators documented that the pathogenesis of MPT-mediated necrosis is distinct from RIPK1/RIPK3-mediated necroptosis. Interestingly, some of the Chinese herbs, such as Panax notoginseng saponins and Astragalus polysaccharide, were noted to alleviate cisplatin-induced AKI by inhibition of the MPTP (183, 184). Of note, before MPT-mediated necrosis was fully recognized, cisplatin-induced AKI was reported to be relieved by cyclosporine A more than two decades ago (185), and the authors suggested a conceivable role of Ca²⁺-dependent MPT-mediated necrosis in this mouse model.

METABOLISM OF MYO-INOSITOL AND ITS RELATED DERIVATIVES AND THEIR RELEVANCE IN THE PATHOBIOLOGY OF RCDs

Myo-inositol (C₆H₁₂O₆, a carbocyclic sugar) is the predominant isomer (among the nine isomers) of inositol present in mammalian cells, and it participates in various biological processes either as a free molecule or as inositol-containing phospholipids and inositol phosphate derivatives (186). Myo-inositol was once termed as vitamin B₈, but this conception has been forsaken due to its extensive endogenous synthesis from glucose (186). It is estimated that the two kidneys, the dominant sites of myo-inositol metabolism in human, produce ∼4 g of myo-inositol daily, which substantially exceeds myo-inositol intake (∼1 g/day) from the regular diet (186, 187). Of note, other organs, including the brain, testis, and liver, are also involved in the endogenous synthesis of myo-inositol (188–190). The synthesis of myo-inositol involves three steps: the initial generation of glucose-6-phosphate via hexokinase-mediated phosphorylation of glucose, and following the conversion of glucose-6-phosphate to myo-inositol-1-phosphate by 1-d-myo-inositol phosphate synthase, it is then converted into myo-inositol by the action of inositol monophosphatase, which apparently dephosphorylates myo-inositol-1-phosphate (Fig. 5) (186). Dietary myo-inositol intake is absorbed from the gastrointestinal tract in free form or as phosphatidylinositol (PI) (186). Once inside the bloodstream, myo-inositol (∼30 μM in normal human plasma) (191) is transported into various organs via its specific transporters, including Na⁺/myo-inositol transporter 1/2 (SMIT1/2) and H⁺/myo-inositol transporter (HMIT) (Fig. 5) (192, 193). In addition, the dephosphorylation of inositol phosphates following the decomposition of inositol-containing membrane phospholipids also participates in the generation of free myo-inositol (Fig. 5) (194).

Figure 5.

The structure, source, and catabolism of myo-inositol. Intracellular myo-inositol is derived in three different ways: de novo synthesis, H⁺/myo-inositol transporter (HMIT)/Na⁺/myo-inositol transporter 1/2 (SMIT1/2)-mediated absorption, and dephosphorylation of inositol phosphates. De novo synthesis of myo-inositol from glucose is catalyzed by hexokinase, 1-d-myo-inositolphosphate synthase (MIPS), and inositol monophosphatase (IMPase) sequentially. Myo-inositol is catabolized by myo-inositol oxygenase (MIOX), a renal proximal tubule-specific enzyme, to generate d-glucuronic acid, which is further converted into xylitol via the glucuronate-xylulose (G-X) pathway. Xylitol is catabolized into d-xylulose-5-phosphate, which is involved in the pentose phosphate cycle. ALR1, aldehyde reductase; DGDC, dehydro-l-gulonate decarboxylase; DXR, d-xylulose reductase; GDH, l-gulonate dedydrogenase; LXR, l-xylulose reductase. [Partially adapted from Ref. (197).]

Unlike the multiple sites of myo-inositol synthesis, its catabolism is restricted only in the kidneys, and it is mediated by the proximal tubular specific enzyme, myo-inositol oxygenase (MIOX) (Fig. 5) (195). The pioneering work by Howard and Anderson revealed that bilateral nephrectomized rats were unable to degrade [2-¹⁴C]myo-inositol into ¹⁴CO₂, indicating an essential role of the kidneys in myo-inositol catabolism (196). It is noteworthy that the urinary excretion of myo-inositol is marginal in normal states, which gets accelerated in the diabetic state with depletion of plasma myo-inositol levels (186). It was later on discovered that myo-inositol undergoes oxidative cleavage of the cyclic ring by MIOX in renal proximal tubular cells, yielding d-glucuronic acid (195, 197). Incidentally, it has been reported in the literature that genetic deletion of MIOX leads to elevated myo-inositol levels in murine serum and renal tissues (197). d-Glucuronic acid is then converted into xylitol via the glucuronate-xylulose pathway, and xylitol is further catabolized into d-xylulose-5-phosphate, which enters the pentose phosphate cycle (Fig. 5) (195). d-Chiro-inositol, an isomer of myo-inositol involved in insulin signaling transduction, is also a cleavage substrate of MIOX (195). Previous investigations have revealed that the catalytic activity of MIOX consumes NADPH and drives renal redox-related injuries in various pathological states (198–200). In addition, our recent work has implicated MIOX overexpression in the induction of apoptosis and ferroptosis in cisplatin-induced AKI (149, 198).

Myo-inositol is present mainly in its free form or as myo-inositol phospholipid in the kidneys, among which free myo-inositol is the predominant form (187). It is worth mentioning here that the renal concentration of myo-inositol also increases in mice following dietary myo-inositol supplementation (201). Although the beneficial effects of myo-inositol supplementation have been reported in various disease states, its role in renal pathology has been poorly investigated. Previous in vitro studies have revealed that myo-inositol treatment alleviates high-glucose-induced collagen synthesis by renal proximal tubular cells (202), whereas diabetic glomerulopathy was unaffected by myo-inositol administration (203). Besides, myo-inositol treatment alleviates the acute renal failure induced by methylene-myo-inositol (a SMIT inhibitor), which may possibly be related to the perturbations in osmoregulation by renal medullary tubules (204). Along these lines, more recent studies have indicated that myo-inositol is protective against cadmium-induced AKI via attenuating oxidant stress and apoptosis (205), although the precise mechanism(s) in alleviation of renal injury remains elusive.

Compared with the limited research on free myo-inositol relevant to AKI, the role of myo-inositol phospholipids in AKI has been extensively investigated. The intricacies and interconversion of various inositol phospholipids have been alluded to earlier (186, 187), and thus only RCD and AKI-related mechanisms are discussed here. Among the various forms of inositol phospholipids, PI, localized to the ER, is the most common form of inositol in mammalian cells (187, 206). Synthesis of PI relies on catalysis by PI synthase, which allows the reaction between myo-inositol and cytidine diphosphate-diacylglycerol (DAG) to yield PI (Figs. 6 and 7) (186, 207). PI is readily phosphorylated into PI phosphate and PI 4,5-bisphosphate [PI(4,5)P₂] by PI kinase and PI phosphate kinase, respectively, in various biological settings (Figs. 6 and 7) (186). Under the catalysis of phospholipase C, PI(4,5)P₂ is further hydrolyzed into DAG and inositol 1,4,5-trisphosphate (IP₃) (Figs. 6 and 7), and both participate in various biological processes as second messengers but initiate distinct downstream signaling (208). DAG remains on the plasma membrane and facilitates the plasmalemmal translocation of protein kinase C (PKC) and its subsequent activation (209). On the other hand, IP₃ is diffused into the cytosol and binds to its receptor localized on the ER Ca²⁺ channel, culminating in an increased concentration of cytosolic Ca²⁺ (Fig. 6D), which also leads to PKC activation (210). Various PKCs have been implicated in the pathogenesis of AKI in different murine models (211–214). PKC inhibition by Go6976 alleviates cisplatin-induced apoptosis in renal proximal tubular cells, which is apparently related to the reduced mitochondrial cytochrome c release and caspase-3 activation (215). Of note, the role of PKC in apoptosis and downstream signaling may be cell specific, and it can decelerate apoptosis in certain states. Also, there are reports that have indicated that PKC can drive necroptosis, pyroptosis, ferroptosis, and MPT-mediated necrosis in other non-AKI settings (214, 216–218). Another inositol phospholipid PI 3,4,5-trisphosphate [PI(3,4,5)P₃], which is synthesized by class I phosphoinositol 3-kinase (PI3K)-mediated phosphorylation of PI(4,5)P₂, also modulates RCDs via the phosphoinositide-dependent kinase 1 (PDK1)/AKT pathway (Figs. 6 and 7) (186). PI(3,4,5)P₃ resides on the plasma membrane and recruits PDK1 and AKT, and both contain PH domains that are essential for interactions (219). The interaction between PI(3,4,5)P₃ and PDK1 leads to activation of PDK1, which then phosphorylates and actives AKT (Fig. 7) (219). PI3K inhibition promotes cisplatin-induced proximal tubular apoptosis via hindering AKT-mediated Bcl-2-associated death promoter phosphorylation, which restores the antiapoptosis capacity of Bcl-XL/Bcl-2 (220). Besides, PI3K/AKT signaling also protects against other forms of AKIs via activation of NRF2 (221, 222), with the latter being a vital antioxidant in the pathogenesis of ferroptosis. Also, the role of PI3K/AKT signaling has also been implicated in necroptosis, pyroptosis, and MPT-mediated necrosis in non-AKI states (217, 223, 224); however, the precise mechanism(s) in these processes has not been thoroughly investigated. Apart from IP₃/DAG- and PI3K-mediated intracellular events, multiple types of polyphosphoinositides per se participate in necroptosis and pyroptosis in terminal stages via binding with phosphorylated MLKL and GSDMD-N, which apparently leads to membrane disruption and cell death (76, 78, 110).

Figure 6.

Key events of myo-inositol derivatives. A: synthesis of phosphatidylinositol (PI). B: synthesis of phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] and phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P₃]. C: decomposition of PI(4,5)P₂ into diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP₃). D: IP₃ binds with its receptor on the endoplasmic reticulum Ca²⁺ channel, resulting in endoplasmic retriculum Ca²⁺ release. CDP-DAG, cytidine diphosphate-diacylglycerol; CMP, cytidine monophosphate; PI3K, phosphatidylinositol 3-kinase; PIP, phosphatidylinositol phosphate; PLC, phospholipase C.

Figure 7.

Connections between myo-inositol metabolism and regulated cell deaths in acute kidney injury (AKI). Phosphatidylinositol (PI), synthesized in the endoplasmic reticulum (ER), is converted into phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] after a series of events. PI(4,5)P₂ is hydrolyzed into inositol 1,4,5-trisphosphate (IP₃) and diacylglycerol (DAG) by phospholipase C (PLC). DAG and IP₃-mediated increased cytoplasmic Ca²⁺ bind with protein kinase C (PKC), leading to its activation. PKC promotes cisplatin-induced AKI by facilitating cytochrome c (Cyt C) release-mediated apoptosis. On the other hand, PI(4,5)P₂ can be further phosphorylated by phosphatidylinositol 3-kinase (PI3K) into phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P₃], which binds with phosphoinositide-dependent kinase 1 (PDK1) to activate AKT. AKT attenuates apoptosis in cisplatin-induced AKI via BAD-restored BCL-2/BCL-XL potency. AKT also protects against other AKIs through nuclear factor erythroid 2-related factor 2 (NRF2), a potent antioxidant of ferroptosis. Of note, polyphosphoinositides present as the binding substrates of mixed lineage kinase domain-like (MLKL) and NH₂-terminal domain of gasdermin D (GSDMD-N), indicating their involvement in necroptosis and pyroptosis. PIP, phosphatidylinositol phosphate.

PERSPECTIVES, CONFLICTS, AND CONCLUSIONS

Over a period of the past two decades, more than 10 variants of RCDs have been described, among which only five types (related to cisplatin-induced AKI) are discussed in this review. Nevertheless, it should be stressed here that other forms of RCDs may also be involved in cisplatin-induced AKI. Due to the insufficient period of investigation following their discovery, the pathological mechanisms of most of the RCDs are poorly investigated, and thus the research in cisplatin-induced AKI has not been fully explored. Limited past investigatory efforts and the diversity of RCDs provide ample research opportunities for AKI-related investigations, since some key factors or relevant mechanism(s) in other forms of RCDs have been investigated before their discoveries. For example, autosis, an autophagic machinery-mediated RCD (49), has never been investigated in cisplatin-induced AKI, even if the role of autophagy has been well documented in the literature (225). In addition, parthanatos, driven by poly(ADP-ribose) polymerase 1 hyperactivation, can be kindled by excessive oxidative stress, hypoxia, or inflammation (226), the very biological processes that are notably observed in cisplatin-induced AKI. Further research is needed to explore the full profile of various RCDs in cisplatin-related nephropathy. Due to the lack of specific markers, it is noteworthy that the evaluation of ferroptosis in vivo by pathological staining is somewhat limited, which is different from the detection of apoptosis by its specific indicators, i.e., cleaved caspase-3, BAX, Bcl-2, or others. More convincing investigations are required to uncover the unique indicators for ferroptosis. The interaction between RCD and myo-inositol metabolism also remains to be fully investigated. For instance, free myo-inositol supplement protects against AKI via alleviating apoptosis (205), while its role in other RCDs has not been investigated. Besides, the PI(4,5)P₂-mediated PKC pathway and PI(3,4,5)P₃-related PI3K/AKT signaling are involved in some of the nonapoptotic RCDs, but the exact mechanisms have been poorly explored.

It is noteworthy that the conclusions in different publications seem to be controversial in certain aspects. For instance, our recent investigation revealed that ferroptosis and apoptosis, rather than necroptosis, are essential in cisplatin-induced AKI (149), and these results are similar to the seminal observations made by Linkermann et al. (147). However, other investigations using genetic modification approaches revealed that “necroptosis” is a vital component in cisplatin-induced tubular cell death (79, 83). A comprehensive examination of these publications revealed that the cisplatin concentration in our experiments was 20 μM, far lower than that used by Monte and Han (100 − 200 μM) (79, 83, 149). These unique observations may suggest that various RCDs are initiated at different stages of injury following cisplatin treatment in in vitro studies. In another setting, the research work by Han and coworkers et al. (83) on MLKL or RIPK3 knockout mice provided convincing evidence for the pathogenetic role of necroptosis in cisplatin-induced AKI, whereas other investigators believed that necroptosis seems to modulate renal vascular effects rather than renal tubular injuries (147). It seems that caspase-11-mediated noncanonical pyroptosis is conceivably operative in cisplatin-induced AKI (122). The pathogenetic role of caspase-11 in this scenario may be somewhat debatable since no exogenous LPS was introduced into the kidneys to activate caspase-11, indicating that other potential mechanisms may be operational in the induction of pyroptosis.

Finally, it should be borne in mind that cisplatin-induced renal cell death is rather complex, involving various types of RCDs that are initiated at different stages of injury. Besides, various RCDs per se are interconnected, and myo-inositol metabolism may be participating in the pathogenesis of RCDs at different stages/levels. The involvement of RCDs at different stages/levels may yield the opportunity for exploration of novel drugs with differing properties for the inhibition of RCDs or small inhibitory molecules relevant to myo-inositol metabolism may provide potential avenues to battle against cisplatin-induced AKI.

GRANTS

The work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK60635, Visiting Scholarships from China Scholar Council Grant 201706370164, National Natural Science Foundation of China Grant 81470925, and Science and Technology program of Guangdong Scientific Committee Grant 2019A1515012116.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

F.D., Y.D., and Y.S.K. conceived and designed research; X.Z. prepared figures; F.D. drafted manuscript; X.Z., I.S., Y.W., and Y.S.K. edited and revised manuscript; Y.D. and Y.S.K. approved final version of manuscript.