Oxidized high mobility group B‐1 enhances metastability of colorectal cancer via modification of mesenchymal stem/stromal cells

Abstract High mobility group box‐1 (HMGB1) is known to be a chemotactic factor for mesenchymal stem/stromal cells (MSCs), but the effect of post‐translational modification on its function is not clear. In this study, we hypothesized that differences in the oxidation state of HMGB1 would lead to differences in the function of MSCs in cancer. In human colorectal cancer, MSCs infiltrating into the stroma were correlated with liver metastasis and serum HMGB1. In animal models, oxidized HMGB1 mobilized three‐fold fewer MSCs to subcutaneous tumors compared with reduced HMGB1. Reduced HMGB1 inhibited the proliferation of mouse bone marrow MSCs (BM‐MSCs) and induced differentiation into osteoblasts and vascular pericytes, whereas oxidized HMGB1 promoted proliferation and increased stemness, and no differentiation was observed. When BM‐MSCs pretreated with oxidized HMGB1 were co‐cultured with syngeneic cancer cells, cell proliferation and stemness of cancer cells were increased, and tumorigenesis and drug resistance were promoted. In contrast, co‐culture with reduced HMGB1‐pretreated BM‐MSCs did not enhance stemness. In an animal orthotopic transplantation colorectal cancer model, oxidized HMGB1, but not reduced HMGB1, promoted liver metastasis with intratumoral MSC chemotaxis. Therefore, oxidized HMGB1 reprograms MSCs and promotes cancer malignancy. The oxidized HMGB1–MSC axis may be an important target for cancer therapy.


| INTRODUC TI ON
Colorectal cancer (CRC) is the leading cause of cancer death worldwide, and its incidence has been increasing in recent years. 1 In Japan, it is the most common cancer, and the second leading cause of cancer death. 2 The average 5-year survival rate for CRC is 72.6%, but the 5-year survival rate is only 18.8% for stage 4 cases with distant metastasis. 2 One-fourth of advanced cases are associated with liver metastasis, 3 which is a life-threatening event accounting for 30% of CRC deaths. 4,5 We have studied the role of diabetes in the progression of liver metastasis. 6,7 High mobility group box-1 (HMGB1) is overexpressed on a diabetic background, which promotes proliferation and invasion of CRC cells via its receptor, receptor for advanced glycation end-products (RAGE), suppresses host antitumor immunity by inducing cell death in monocytic cells, and subsequently promotes liver metastasis. [8][9][10][11][12] Mesenchymal stem/stromal cells (MSCs) are self-renewing mesenchymal cells that differentiate into multiple lineages. 13 In nontumor tissues, MSCs play a role in regeneration during inflammation and other disorders. 14 MSCs have been isolated from many cancerous tissues, and influence the development and progression of cancer. 15 HMGB1 is known as a chemotactic factor of MSCs. 16 Other chemotactic factors of MSCs include platelet-derived growth factor, insulin growth factor, hepatocyte growth factor, fibroblast growth factor, transforming growth factor (TGF)β, and chemokines, including stromal cell-derived factor-1. [17][18][19] In contrast with these factors, HMGB1 is released from necrotic tissues and is thought to be involved in tissue regeneration in necroinflammatory reactions. 20 HMGB1 is also secreted by cancer cells and promotes cancer cell proliferation, invasion, and metastasis through RAGE expressed on cancer cells. 10 The recruitment of MSCs via HMGB1 in cancer is thought to promote cancer progression, but the mechanism is unclear.
The action of HMGB1 is altered by post-translational modifications. HMGB1 is known to be post-translationally modified by acetylation, phosphorylation, and oxidation. 21 Acetylation and phosphorylation alter DNA binding and bending properties of the HMGB1 protein and subsequently affect its secretion. 22,23 Furthermore, HMGB1 contains three conserved redox-sensitive cysteines (C23, C45, and C106), and the modification of these cysteines determines the bioactivity of extracellular HMGB1. 24,25 The disulfide bond between C23 and C45 in the HMGB1 molecule and the reduction of C106 (disulfide HMGB1; oxidized HMGB1) results in proinflammatory cytokine-stimulating activity, whereas the reduction of all cysteine residues (reduced HMGB1) results in chemotaxis mediator activity. 24,[26][27][28] When all cysteine residues are oxidized, HMGB1 is inactivated. 29 Oxidized HMGB1 is reduced by superoxide dismutase (SOD), catalase, and peroxidase, to its reduced form. 29 In this study, we investigated the effect of HMGB1 on bone marrow MSCs (BM-MSCs), in particular the role of HMGB1-pretreated BM-MSCs in cancer.

| Patients
We obtained frozen tissue samples from 16 patients with CRC with serosal invasion (pT3) and 1-3 regional lymph node metastases (pN1), who were diagnosed at the Department of Molecular Pathology, Nara Medical University, from 2014-2019 (Table S1). As written informed consent was not obtained from the patients for their participation in the present study, all identity-related information was removed from patient samples prior to their analysis to ensure strict privacy protection (unlinkable anonymization). All procedures were performed in accordance with the Ethical Guidelines for Human Genome/Gene Research enacted by the Japanese Government and with the approval of the Ethics Committee of Nara Medical University (approval number: 937, 2014/10/20).
Immunohistochemical staining for CD133 was performed using the immunoperoxidase technique 6 with a CD133 antibody (0.5 μg/ ml, Abcam). The number of positive cells was counted, and the mean value was calculated from the microscopic observation of 30 highpower fields of view.

| MSC preparation
BALB/c mice (5-week-old, male; SLC Japan, Shizuoka, Japan) or F344 rats (6-week-old, male, SLC) were euthanized, and bone marrow cells were harvested by flushing the bone marrow from the femur with regular DMEM (WAKO). After centrifugation at 1500 rpm for 5 min, the pellet was suspended in PBS (WAKO). After lysis of red blood cells, centrifugation was repeated at 1500 rpm for 5 min, and the pellet was resuspended and cultured in MSC culture medium (MesenCult-ACF Plus; Veritas, Tokyo, Japan) for 3 days. Floating cells were carefully removed by PBS washing.

| Sphere formation assay
Cells were mixed with syngeneic BM-MSCs, at the indicated cell ratios, with a total cell number of 1000 cells, and were seeded onto

| Orthotopic tumor model
KM12C and KM12SM colon cancer cells (1 × 10 6 ) were inoculated into the cecal submucosa of nude mice. After euthanasia at 4 weeks, livers were excised and sectioned into 2-mm-thick slices, and metastatic foci were counted using a stereomicroscope (Nikon). 32 For evaluation of serum HMGB1, cardiac blood was collected at euthanasia.

| Subcutaneous tumor model
CT26 cells were pretreated with siRNA for HMGB1 (10 nM) or siMix (10 nM). CT26 cells (1 × 10 7 ) were inoculated into the scapular subcutaneous tissues of six BALB/c mice; three were inoculated with siHMGB1-treated CT26 cells, and another three were inoculated with siC-treated cells. After euthanasia at 4 weeks for histological examination.

| Subcutaneous spongel model
Spongel® (5 mm × 5 mm, LTL Pharma, Tokyo, Japan) was soaked with recombinant human HMGB1 (hrHMGB1, 10 μg/ml, R&D) or PBS (WAKO) and inserted into the subcutaneous tissue on the back of the BALB/c mice. Spongel was removed at 1 week after insertion and fixed in 10% formalin at 4°C for 24 h for histological examination. After euthanasia at 2 weeks, tumors were excised for histological examination by frozen section.

| Mouse tumorigenesis model
CT26 cells were mixed with none, reduced HMGB1-or oxidized HMGB1-treated mouse BM-MSCs (1% of CT26 cells). Cell mixtures were inoculated into the subcutaneus of BALB/c mice (five mice each group). At 3 weeks after inoculation, tumor formation was assessed macroscopically.

| Depletion of extracellular vesicles (EV)
Cultured medium of mouse MSC cells (5 ml from 5 × 10 6 cells) was mixed with ExoQuick-TC (1 ml; System Bioscience, Palo Alto, CA, USA) at 4°C for 12 h. The mixture was centrifuged (1500 g, 30 min) to precipitate EV as pellets. The supernatant was used for sphere formation assay as the EV-depleted culture medium. For control culture medium, the EV-depleted culture medium, in which EV pellets were redissolved, was used.

| Liver metastasis model
For establishment of liver metastasis, CRC cells were pretreated with siHMGB1 (10 nM). CRC cells (1 × 10 6 ) were mixed with syngeneic BM-MSCs (1 × 10 4 ) or none and were inoculated into the spleen of syngeneic rodents (BALB/c mice or Fisher F344 rats). Each group contained five rodents. After euthanasia at 4 weeks, livers were excised and sectioned into 2-mm-thick slices, to count metastatic foci using a stereomicroscope (Nikon). 32

| RT-PCR
To assess murine mRNA expression, RT-PCR was performed and PCR products images were measured using NIH ImageJ software (version 1.52; Bethesda, MD, USA). 30 The primer sets are listed in Table 1.

| Western blotting
Whole cell lysates were prepared according to our previous report. 33,34 The Minute Cytoplasmic and Nuclear Extraction Kit

| ELISA
Levels of HMGB1 was measured using an ELISA kit (Shino-Test, Corp., Tokyo, Japan) according to the manufacturer's instructions.

| Small interfering RNA
Silence® siRNA targeting mouse Hmgb1 (ID:158974 and 158,975) and rat Hmgb1 (ID:197413 and 48,868) were purchased from Thermo Fisher. AllStars Negative Control siRNA was used as a control (Qiagen; Valencia, USA). The cells were transfected with 10 nM siRNA using Lipofectamine 3000 (Thermo Fisher) according to the manufacturer's recommendations.

| Statistical analysis
Statistical significance was calculated using a two-tailed Fisher's exact test and ordinary ANOVA using the InStat software (GraphPad, Los Angeles, CA, USA). A two-sided p-value of <0.05 was considered statistically significant.

| Effect of HMGB1 on human CRC liver metastasis
To investigate the effect of MSCs on metastasis of CRC in liver, we detected OCT3 + /CD73 + MSCs in the primary tumor in 16 cases of serous invasion (pT3) and lymph node metastasis (pN1) ( Figure 1A, B). MSCs were found in the stroma of the primary tumor, as shown in Figure 1B; the number of MSCs was fivefold higher in liver metastasis-positive cases in comparison with non-metastasized cases ( Figure 1C). HMGB1 is known to be a chemotactic factor for MSCs. 16 When the serum concentration of HMGB1 and the number of MSCs in the primary tumor were compared in CRC cases, a clear correlation was observed between the two ( Figure 1D). This correlation was observed in both metastasis-positive and metastasis-negative cases, but both HMGB1 levels and number of MSCs were higher in positive cases, Serum HMGB1 was three-fold higher in KM12SM than in KM12C.

TA B L E 1 Primer sets
Therefore, number of MSCs and HMGB1 levels were shown to be strongly associated with liver metastasis of CRC.

| Differential effect of reduced HMGB1 and oxidized HMGB1 on translocation of BM-MSCs
While HMGB1 is known to be a chemotactic factor for MSCs, its action is altered by oxidative modification; oxidized HMGB1 inhibits migration. 35 We compared the effects of reduced and oxidized HMGB1 on the migration of BM-MSCs into tumors in an animal model ( Figure 2). Subcutaneous inoculation of CT26 mouse colon cancer cells knocked down for HMGB1 into syngeneic BALB/c mice showed that the number of CD73 + /OCT3 + MSCs in the tumor was reduced by three-fold compared with that in HMGB1-expressing CT26 cells (Figure 2A).
Next, in a model of subcutaneous spongel implantation in BALB/c mice, reduced or oxidized HMGB1 was imbibed into the spongel ( Figure 2B). Both CD73 + MSCs and CD11b + myeloid cells were induced to migrate into the spongel by reduced HMGB1. In contrast, with oxidized HMGB1, the induction of migration of MSCs was reduced by three-fold, and the induction of myeloid cells migration was reduced by six-fold.
We generated a mouse model in which bone marrow cells were

| Differential effect of reduced HMGB1 and oxidized HMGB1 on stemness and differentiation of MSCs
Next, we compared the effects of reduced and oxidized HMGB1 on stemness and differentiation of BM-MSCs ( Figure 3). As shown in Figure 3A, BM-MSCs from BALB/c mice were treated with reduced or oxidized HMGB1. With reduced HMGB1 treatment, CD73 expression was maintained, but expression of stem cell markers CD44 and Gnl3 decreased. In contrast, the expression of the osteoblast differentiation marker osterix (Osx) and the vascular pericyte marker Acta2 was induced. With oxidized HMGB1 treatment, CD73 expression was maintained, but CD44 and Gnl3 expression increased.
Reduced HMGB1 inhibited BM-MSC proliferation compared with vehicle (PBS) ( Figure 3B). In contrast, oxidized HMGB1 enhanced BM-MSC proliferation three-fold. Examination of the proliferative signals showed that expression of PCNA, a marker of proliferation, was decreased, and pERK1/2 and p38 phosphorylation was increased by reduced HMGB1. In contrast, with oxidized HMGB1, levels of PCNA and phosphorylated ERK1/2 increased, whereas that of phosphorylated p38 remained unchanged ( Figure 3C). Therefore,

| Differential effect of reduced HMGB1 and oxidized HMGB1 on stemness and differentiation of CRC cells
Next, we compared the effects of BM-MSCs exposed to reduced or oxidized HMGB1 on syngeneic CRC cells in a co-culture system (1% of CRC cell population) (Figure 4). In contrast, tumor formation was observed with 1 × 10 3 CT26 cells when oxidized HMGB1-pretreated BM-MSCs were added.
As cancer stem cells are known to confer anticancer drug resistance, 36 we examined the effect of BM-MSCs on anticancer drug resistance in CRC cells ( Figure 4D). In CRC cells without BM-MSCs or with BM-MSCs pretreated with reduced HMGB1, there was no difference in 5-FU sensitivity. In contrast, the IC50 was seven-and six-fold higher in CT26 and IEC18SRC cells, respectively, with BM-MSCs pretreated with oxidized HMGB1.
Therefore, MSCs treated with oxidized HMGB1 increased the stemness of cancer cells and induced tumorigenesis and anticancer drug resistance.

| Effect of extracellular vesicles from MSCs
As shown in Figure 4E, to evaluate the effect of the EV from MSCs on cancer stemness, we compared sphere-forming ability of CT26 cells among CT26 cells alone, contact co-culture with oxHMGB1-treated MSCs and CT26 cells, non-contact co-culture with oxHMGB1treated MSCs and CT26 cells. Only the contact co-culture of both cells showed an increased sphere-forming ability. Moreover, when we compared sphere-forming ability of CT26 cells between treatment with MSC cultured medium and treatment with EV-depleted MSC cultured medium, no difference was observed between these two conditions. These findings suggest that effects of MSCs might be depend on cell-cell contact.

| DISCUSS ION
In this study, we showed that the effects of HMGB1 on MSCs were different between the reduced and oxidized forms of HMGB1. tastasis. This suggests that oxidized HMGB1 may be a factor that reprograms MSCs to be tumor promoting.
According to our data, oxidized HMGB1 induced stemness in colon cancer cells. MSCs are known to promote cancer cell proliferation, motility, and invasion in the cancer microenvironment. 47 (Figure 2B inset).
How is oxidized HMGB1 produced in cancer is an interesting question that needs to be addressed. HMGB1 has no signal peptide and is secreted into the extracellular space by a variety of noncanonical secretory mechanisms. 52 HMGB1 that forms disulfide bonds binds to the nuclear exportin chromosomal region maintenance 1, which preferentially transports it out of the nucleus, resulting in its secretion. 53 Previously, we have reported that nuclear oxidation by deoxycholic acid decreases nuclear HMGB1 levels. 54 Peroxiredoxins I and II induce the formation of intramolecular disulfide bonds of HMGB1 in the nucleus. 55 As a result, nuclear export and even extracellular secretion of HMGB1 is induced. 53 In cancer, the expression of extracellularly localized SOD3 is epigenetically suppressed, 56 increasing oxidative stress in the cancer microenvironment. In such an oxidative environment, HMGB1 may be modified to an oxidized form in the cancer microenvironment, and its modification may be maintained. In our study, extracellularly secreted HMGB1 was mainly in the oxidized form. 30 We have previously shown that the co-expression of HMGB1 and RAGE in cancer cells correlates with malignancy potential of various cancers, including CRC. 10 The effect of reduced HMGB1 on MSCs in cancer is due to the expression of RAGE as a receptor. 35 In our previous study, phosphorylation of RAGE and AKT and nuclear translocation of NF-κB p65 are at low levels by reduced HMGB1 and high levels by oxidized HMGB1. 30 Therefore, oxidized HMGB1 was considered to be a highly functional ligand for RAGE in cancer. In contrast, in MSCs, RAGE is induced and activated by HMGB1 to promote the expression of CXCR4, 57 TGFβ, and proinflammatory cytokines. 58,59 However, reports analyzing the modification of HMGB1 and its function in MSCs are scarce. The differential action of the oxidized and reduced forms in this study may be important for future targeting of MSCs. Recently, it has been noticed that exosomes secreted by MSCs in tumors play a major role in the alteration of the cancer cell phenotype. 60 As we compared co-culture with cellcell contact and non-contact condition, the contact was essential for the pro-stemness effect of MSCs. Moreover, the culture medium of MSCs did not affect the sphere-forming ability of CT26 cells with or without EVs. However, it is still important to examine the differential effect of oxidized or reduced HMGB1 on MSC-derived exosomes in further studies.
This study suggested that oxidized HMGB1 recruits BM-MSCs into tumors and reprograms them, thereby increasing cancer stemness and promoting liver metastasis. Targeting of the posttranslational modifications of HMGB1 is expected to inhibit tumorspecific MSCs and suppress cancer stemness, 61 and could be an important therapeutic option for CRC in the future.

AUTH O R CO NTR I B UTI O N S
Study concept and design: HK. Acquisition of data: SK, RFT, RS.
Analysis and interpretation of data: SK, RFT, SM, HO, YM. Drafting and editing of the manuscript: SK, RFT. Critical revision of the manuscript: TS, IK. All authors gave final approval of the version to be published, and agree to be accountable for all aspects of the work.

ACK N OWLED G M ENTS
The authors thank Ms. Tomomi Masutani for expert assistance with the preparation of this manuscript.

D I SCLOS U R E
The authors disclose no potential conflicts of interest.

E TH I C A L A PPROVA L
As written informed consent was not obtained from the patients for their participation in the present study, all identity-related information was removed from patient samples prior to their analysis to ensure strict privacy protection (unlinkable anonymization). All pro-