Role of HMGB1 in Chemotherapy-Induced Peripheral Neuropathy

Chemotherapy-induced peripheral neuropathy (CIPN), one of major dose-limiting side effects of first-line chemotherapeutic agents such as paclitaxel, oxaliplatin, vincristine, and bortezomib is resistant to most of existing medicines. The molecular mechanisms of CIPN have not been fully understood. High mobility group box 1 (HMGB1), a nuclear protein, is a damage-associated molecular pattern protein now considered to function as a pro-nociceptive mediator once released to the extracellular space. Most interestingly, HMGB1 plays a key role in the development of CIPN. Soluble thrombomodulin (TMα), known to degrade HMGB1 in a thrombin-dependent manner, prevents CIPN in rodents treated with paclitaxel, oxaliplatin, or vincristine and in patients with colorectal cancer undergoing oxaliplatin-based chemotherapy. In this review, we describe the role of HMGB1 and its upstream/downstream mechanisms in the development of CIPN and show drug candidates that inhibit the HMGB1 pathway, possibly useful for prevention of CIPN.


Introduction
Chemotherapy-induced peripheral neuropathy (CIPN) is one of the major doselimiting side effects of first-line chemotherapeutic agents such as paclitaxel, oxaliplatin, vincristine, and bortezomib. In terms of evidence-based medicine, no drugs are recommended for the prevention of CIPN, and duloxetine is the only agent that has limited benefit in treating established CIPN [1]. The mechanisms for development and maintenance of CIPN have not been fully understood, although preclinical studies have provided evidence for some possible mechanisms for CIPN [2], such as neuroimmune interactions [3], mitochondrial dysfunction [4], reactive oxygen species (ROS) accumulation [5], and transcriptional or functional upregulation of cation channels [6,7] in the spinal cord, dorsal root ganglion (DRG), and peripheral sensory neurons. Our preclinical studies have demonstrated the involvement of high mobility group box 1 (HMGB1) in the development and maintenance of CIPN [8][9][10][11]. HMGB1 (also known as amphoterin) is a non-histone nuclear protein that has highly conserved amino acid sequences during evolution and is ubiquitously expressed in most mammalian cell types [12]. HMGB1 is essential for life, because HMGB1-deficient mice die within 24 h of birth due to hypoglycemia [13]. In the nucleus, HMGB1 contributes to nucleosome stability and sliding, DNA replication and repair, gene transcription, etc. On the other hand, HMGB1 is actively secreted by activated immune cells and passively released from necrotic cells in pathological conditions including inflammation. Extracellular HMGB1 acts as a damage-associated molecular pattern (DAMP) protein via activation of several pattern recognition receptors (PRRs) including Toll-like receptors (TLRs), receptors for advance glycosylation end products (RAGE), C-X-C motif chemokine receptor 4 (CXCR4), etc., leading to acceleration of inflammation and pain [11,12,[14][15][16][17][18][19][20]. There are plenty of papers showing that an anti-HMGB1-neutralizing antibody (HMGB1-nAb) [21,22] strongly suppresses somatic or visceral pathological pain with inflammatory and/or neuropathic components [23][24][25][26][27][28][29][30][31][32][33][34]. We have shown that the HMGB1-nAb strongly prevents the development of CIPN in rodents [8][9][10]. Similarly, recombinant human soluble thrombomodulin (TM) [thrombomodulin alfa (TMα), ART-123, recomodulin ® ], capable of promoting thrombin-dependent HMGB1 degradation [11,35,36], also prevents the development of CIPN in rodent models [8][9][10]. Intriguingly, the efficacy of TMα in preventing CIPN in humans has been confirmed by a placebo-controlled, randomized, double-blind phase IIa study [37]. Thus, targeting extracellular HMGB1 is considered a promising strategy to prevent CIPN. In this review, we focus on the role of HMGB1 in the development of CIPN, and describe large or small molecules that deactivate HMGB1 or inhibit the upstream/downstream signaling molecules of HMGB1, which may be useful for prevention of CIPN.

Inactivation of HMGB1 by the Thrombin and TM System
Thrombin at high concentrations (5 U/mL or higher) inactivates HMGB1 by cleaving a peptide bond between Arg10 and Gly11 of HMGB1 ( Figure 1A). Endothelial TM, known to accelerate proteolytic activation of protein C (PC) and thrombin-activatable fibrinolysis inhibitor (TAFI) by thrombin, causes HMGB1 degradation in the presence of thrombin at low or minimal concentrations (2 U/mL or lower), an effect mimicked by TMα, a recombinant human soluble TM. TM, as well as TMα, sequesters HMGB1 and accelerates thrombininduced HMGB1 degradation ( Figure 1D) [11,35,36,88,89]. The endothelial TM/thrombin axis may thus play a role in decomposing excessive HMGB1 in the blood stream under pathological conditions [11].

Roles of Macrophage-Derived HMGB1 and Its Upstream/Downstream Molecules in CIPN Caused by Paclitaxel
Macrophages appear to be the main source of HMGB1 involved in CIPN caused by paclitaxel ( Figure 2A) [9]. Depletion of macrophages by liposomal clodronate or treatment with minocycline and ethyl pyruvate, capable of inhibiting HMGB1 release from macrophages, completely abolishes the development of CIPN in mice treated with paclitaxel (Table 1) [9]. Actually, macrophages accumulate in the sciatic nerve and dorsal root ganglion (DRG) from the CIPN mice treated with paclitaxel ( Figure 2A) [9,93]. In macrophage-like RAW264.7 cells, paclitaxel causes cytoplasmic translocation and extracellular secretion of nuclear HMGB1, which involves the activation of the ROS/p38MAPK/NF-κB pathway and subsequent upregulation of histone acetyltransferases (HATs), possibly essential for acetylation of HMGB1 (Figure 2A) [9]. Deletion of ROS by N-acetylcysteine (NAC), an antioxidant, prevents the development of CIPN in mice treated with paclitaxel in vivo (Table 1), in agreement with the in vitro experiments using macrophages [9]. Coculture experiments using macrophage-like RAW264.7 cells and neuron-like NG108-15 cells suggest that "unknown humoral factors" derived from neurons accelerate HMGB1 secretion by macrophages in response to paclitaxel (Figure 2A) [9]. Thus, a crosstalk between macrophages and neurons mediated by HMGB1 and unknown humoral factors may play a key role in CIPN development after paclitaxel treatment.  Antagonists of RAGE or CXCR4, membrane receptors targeted by at-HMGB1, prevent the development of CIPN caused by paclitaxel (Figure 2A and Table 1) [9]. On the other hand, the role of TLR4 in the paclitaxel-induced CIPN appears to be different among species or strains. TAK-242, a TLR4 antagonist capable of penetrating into the CNS, prevents CIPN after paclitaxel treatment in C57BL/6 mice and Sprague-Dawley rats, but not in ddY mice (Table 1) [9,94]. It is to note that intraperitoneal (i.p.) administration of LPS-RS, a peripherally preferential TLR4 antagonist, does not prevent the CIPN in C57BL/6 mice (Table 1) [9], while intrathecal administration of LPS-RS blocks rat CIPN caused by paclitaxel (Table 1) [95]. Therefore, the CIPN caused by paclitaxel appears to involve central, but not peripheral, TLR4 in rats and C57BL/6 mice. Interestingly, there is a report showing that macrophage TLR9, an intracellular compartment receptor for HMGB1, participates in the CIPN caused by paclitaxel in male, but not female, mice, suggesting the existence of sex dimorphism in the role of TLR9 in the development of CIPN [96]. The same study has shown the involvement of TLR9 in the paclitaxel-induced release of TNF and CXCL1 from male, but not female, macrophages [96]. Similar sex dimorphisms have also been described in the role of TLR4 in pain processing [80,97]. Collectively, activation of RAGE and CXCR4 possibly by macrophage-derived at-HMGB1 may play a major part in CIPN due to paclitaxel.

Role of HMGB1 and Its Upstream/Downstream Molecules in CIPN Caused by Oxaliplatin
We have shown that the inactivation of HMGB1 with HMGB1-nAb or TMα and pharmacological blockade of RAGE, CXCR4, or TLR4 prevent CIPN in rodents treated with oxaliplatin ( Figure 2B and Table 1) [10]. Peripheral TLR4 appears to be involved in the CIPN caused by oxaliplatin, because systemic administration of LPS-RS as well as TAK-242 exhibited a preventive effect in this CIPN model ( Figure 2B and Table 1). Most surprisingly, macrophages do not play a major role in the CIPN caused by oxaliplatin, since depletion of macrophages by liposomal clodronate as well as inhibition of macrophagederived HMGB1 by minocycline or ethyl pyruvate had no effect on the CIPN in mice treated with oxaliplatin [10]. An independent group [98] has suggested the involvement of macrophages in the same CIPN model, being apparently inconsistent with our results [10]. However, their study showed that liposomal clodronate only slightly (by 30%) reduced the oxaliplatin-induced allodynia, suggesting that the participation of macrophage-derived HMGB1 in the CIPN caused by oxaliplatin should be very minor, if any [99]. This notion is further supported by our findings that the number of macrophages in the sciatic nerve did not increase in mice treated with oxaliplatin. Thus, the origin of HMGB1 involved in CIPN caused by oxaliplatin is still open to question ( Figure 2B). Alternatively, it is likely that HMGB1 derived from multiple cells including macrophages contributes to the development of CIPN following oxaliplatin treatment ( Figure 2B), because oxaliplatin at relatively high concentrations (3-10 µM) causes HMGB1 release from Schwann cells and macrophage-like RAW264.7 cells in vitro [10]. Our study to explore other origins that secrete HMGB1 in response to oxaliplatin is now in progress.

Thrombin-Dependent Suppression of CIPN by Endogenous and Exogenous TM
TM, an endothelial membrane protein, consists of five domains: D1 (lectin-like domain), D2 (EGF-like domain), D3 (O-glycosylated serine-threonine-rich domain), D4 (transmembrane domain), and D5 (cytoplasmic domain). Thrombin binding to the D2 proteolytically converts protein C and TAFI into activated protein C (APC) and the activated form of TAFI (TAFIa), respectively ( Figure 1D) [100]. TMα, consisting of three extracellular regions of TM (D1, D2, and D3), is generated by the protein expression system using Chinese hamster ovary (CHO) cells, and preserves the efficacy of TM in accelerating the thrombin-dependent activation of protein C and TAFI ( Figure 1D) [11,35,36]. TMα as well as TM sequesters HMGB1 at the D1 and accelerates its degradation by thrombin binding to the D2 ( Figure 1D and Table 1). Therefore, TMα inhibits the mechanical allodynia following intraplantar injection of at-HMGB1 or ds-HMGB1 in an endogenous thrombin-dependent manner [11,18,89]. TMα also dramatically reduces the endogenous HMGB1-dependent inflammatory hyperalgesia following intraplantar LPS [28], cystitisrelated bladder pain [19,30], and pancreatitis-related pain [20,32]. Similarly, TMα prevents the development of CIPN in rodents treated with paclitaxel, vincristine, or oxaliplatin ( Figure 2 and Table 1) [8][9][10]. A placebo-controlled, randomized, double-blind phase IIa study has confirmed the clinical usefulness of TMα in preventing CIPN in patients with colorectal cancer undergoing oxaliplatin-based chemotherapy (Table 1) [37]. It is important to note that patients undergoing anticoagulant therapy should be excluded in future clinical trials for evaluation of the anti-CIPN effect of TMα, considering our preclinical study showing that anti-coagulants cancelled the preventive effect of TMα on CIPN in mice ( Figure 2B) [10,11]. Furthermore, our study has demonstrated that repeated administration of different anticoagulants aggravates CIPN and increases plasma HMGB1 levels in mice treated with a sub-effective dose of oxaliplatin [10], suggesting that the endothelial TM/thrombin system functions to reduce CIPN by degrading excessive HMGB1 released in response to oxaliplatin ( Figure 2B) [11]. It has also been reported that the anti-CIPN effect of TMα in mice treated with oxaliplatin involves the thrombin-dependent production of APC and TAFIa, in addition to degradation of HMGB1, by TMα ( Figure 2B and Table 1) [101]. The molecular mechanisms by which APC and TAFIa reduce the CIPN remain to be clarified. Apart from CIPN, TMα also reduces HMGB1-dependent symptoms in acute cystitis and pancreatitis accompanied by visceral pain [30,32] as well as disseminated intravascular coagulation (DIC) [102], and our ongoing studies focus on the effects of anticoagulants on HMGB1 levels in the blood stream in order to clarify the role of HMGB1 degradation by the endogenous TM/thrombin axis in those pathological conditions.

Other Candidates that Directly Inactivate HMGB1
Glycyrrhizin (glycyrrhizic acid), a component of licorice root, possesses anti-inflamm atory and antiviral activities [103]. Direct binding of glycyrrhizin to both box A and box B of HMGB1 (Kd =~150 µM) has been demonstrated by nuclear magnetic resonance (NMR) and fluorescence studies (Table 1), and the IC 50 of glycyrrhizin is 50 µM in inhibiting cell migration of 3T3 fibroblasts stimulated with HMGB1 at 1 nM [104].
Methotrexate, a folic acid antagonist, is used in chemotherapy of tumors and autoimmune diseases including rheumatoid arthritis. Direct binding of methotrexate to two independent sites of HMGB1 has been demonstrated by surface plasmon resonance (SPR) analysis and electrophoretic mobility shift assay (EMSA) ( Table 1) [111]. Methotrexate appears to inhibit the interaction of HMGB1 to RAGE, but not TLR4 [111]. It would be interesting to investigate the effect of methotrexate on CIPN.
Metformin, a biguanide derivative, is the first-line drug in the treatment of type 2 diabetes, and has also an anti-inflammatory activity. There is evidence that metformin directly binds to the C-terminal acidic tail of HMGB1, as demonstrated by a pull-down assay using full-length and C-terminal acidic tail-lacking HMGB1 (Table 1), an effect contributing to its anti-inflammatory effects [112]. Metformin inhibits the high glucoseinduced upregulation of RAGE and HMGB1 in rat ventricular myocytes [113] and LPSinduced HMGB1 secretion in rabbit annulus fibrosus stem cells [114]. Most interestingly, preclinical studies have shown that metformin prevents CIPN in mice and rats treated with cisplatin and oxaliplatin, respectively (Table 1) [115,116].
(−)-Epigalocatechin-3-gllate (EGCG), a major effective component of green tea, is associated with many health benefits against multiple inflammatory diseases including rheumatoid arthritis [117]. EGCG is internalized into HMGB1-containing LC3-positive cytoplasmic vesicles (likely autophagosomes) in macrophages stimulated with LPS, leading to HMGB1 aggregation and inhibition of upregulation and extracellular release of HMGB1 [118]. A computational modeling study has shown that EGCG firmly binds to a region around C106 of HMGB1, leading to aggregation of HMGB1 (Table 1) [119]. Salicylic acid, a deacetylated form of aspirin, binds to the HMG-box domains of HMGB1, as assessed by NMR spectroscopic analysis (Table 1), and suppresses the chemoattractant activity of at-HMGB1 and the upregulation of proinflammatory cytokines and COX-2 induced by ds-HMGB1 [120]. The effects of EGCG and salicylic acid on CIPN have yet to be tested.
Low molecular weight heparin (LMWH, parnaparin, MW: 4500~6500), an anticoagulant that preferentially inhibits factor Xa rather than factor IIa (thrombin), binds to RAGE at Kd value of 17 nM, as determined by SPR assay [122], although LMWH also has some sensitivity to HMGB1 itself [123]. LMWH strongly inhibits the mechanical allodynia following intraplantar administration of at-HMGB1 capable of stimulating RAGE, but not of ds-HMGB1 capable of stimulating TLR4, suggesting a possible contribution of RAGE blockade, but not HMGB1 inactivation, in the anti-allodynic effect of LMWH [16]. LMWH also prevents endogenous HMGB1-dependent pain, including cystitis-related bladder pain [30] and neuropathic pain, caused by surgical injury of the spinal nerve [76]. As does FPS-ZM1, LMWH prevents the development of CIPN in rodents treated with paclitaxel [9] or oxaliplatin [10] (Table 1).
Azeliragon (also called PF-04494700 or TTP488), an orally bioavailable small molecule antagonist of RAGE that can penetrate BBB, is now being evaluated for efficacy and safety in patients with Alzheimer's disease, because stimulation of RAGE by amyloid β is involved in neurodegeneration (Table 1) [124][125][126]. Azeliragon blocks the interaction of RAGE with amyloid β, S100B, or HMGB1, as determined by a fluorescent polarization assay [67,[124][125][126]. Given plenty of clinical evidence for the safety, azeliragon is one of the most promising candidates for an anti-CIPN agent, although neither preclinical nor clinical evidence for the effect of azeliragon on pain is available.

TLR Antagonists
Lipopolysaccharide of Rhodobacter sphaeroides (LPS-RS), one of the best-known TLR4 antagonists [127] (Table 1), suppresses the LPS-induced inflammatory responses both in vitro [128] and in vivo [129]. Chemical modification of the lipid A, a membraneanchoring moiety of LPS, is well-established as an approach to develop TLR4-sensitive compounds [130]. Eritoran (E5564), a synthetic analogue of the lipid A, also antagonizes TLR4 (Table 1) [127] and suppresses the biological effects of LPS [131]. Both LPS-RS and eritoran antagonize TLR4 by targeting MD-2, a co-receptor of TLR4 [127]. FP7, a synthetic monosaccharide lipid A mimetic, also selectively blocks TLR4 signaling by binding to MD-2 and CD14 (Table 1) [132,133]. TAK-242 (also called CLI-095), a BBB-permeable small molecule, blocks TLR4 through covalent binding to Cys747 in the intercellular domain of TLR4 (Table 1) [134], thereby exhibiting anti-inflammatory activity [135]. Preclinical studies have shown the anti-CIPN effect of TAK-242 and LPS-RS in some animal models for CIPN (Table 1) [9,10,94,95]. TAK-242 is considered one of the best candidates for clinical application or as a seed in drug development, considering the chemical structure.

Conclusions
HMGB1 is now considered a key molecule responsible for the development of pathological pain including CIPN. HMGB1 and its upstream/downstream molecules are thus considered promising targets to develop agents to prevent CIPN. In particular, the clinical effectiveness of TMα, capable of inactivating HMGB1, in preventing the CIPN due to oxaliplatin in humans [37] is encouraging, because many compounds targeting other molecules, which were effective in preclinical studies using animal models for CIPN, did not show significant efficacy or enough safety in clinical trials [1].