Cancer-derived cholesterol sulfate is a key mediator to prevent tumor infiltration by effector T cells

Abstract Effective tumor immunotherapy requires physical contact of T cells with cancer cells. However, tumors often constitute a specialized microenvironment that excludes T cells from the vicinity of cancer cells, and its underlying mechanisms are still poorly understood. DOCK2 is a Rac activator critical for migration and activation of lymphocytes. We herein show that cancer-derived cholesterol sulfate (CS), a lipid product of the sulfotransferase SULT2B1b, acts as a DOCK2 inhibitor and prevents tumor infiltration by effector T cells. Using clinical samples, we found that CS was abundantly produced in certain types of human cancers such as colon cancers. Functionally, CS-producing cancer cells exhibited resistance to cancer-specific T-cell transfer and immune checkpoint blockade. Although SULT2B1b is known to sulfate oxysterols and inactivate their tumor-promoting activity, the expression levels of cholesterol hydroxylases, which mediate oxysterol production, are low in SULT2B1b-expressing cancers. Therefore, SULT2B1b inhibition could be a therapeutic strategy to disrupt tumor immune evasion in oxysterol-non-producing cancers. Thus, our findings define a previously unknown mechanism for tumor immune evasion and provide a novel insight into the development of effective immunotherapies.


Introduction
Although immunotherapies, such as immune checkpoint blockade, have revolutionized cancer treatment, many patients still do not derive clinical benefits (1). This resistance to immunotherapy could be caused by several tumor-intrinsic mechanisms that affect the generation and function of cancer-specific T cells. They include a lack of antigenic mutations, disruption of the antigen-processing machinery, loss of major histocompatibility complex (MHC) expression, and defects in interferon-γ (IFNγ)-or tumor necrosis factor (TNF)-mediated signaling pathway that acts to increase tumor immune susceptibility (2)(3)(4)(5). On the other hand, tumors often constitute specialized microenvironments that exclude T cells from the vicinity of cancer cells, even though functional cancer-specific T cells exist in the draining lymph nodes (6,7). As effective tumor immunotherapy requires physical contact of T cells with cancer cells, elucidation of the mechanisms is required for the development of better therapies. So far, it has been proposed that epigenetic silencing of the genes encoding chemokines in the tumor microenvironments and oncogenic pathways in tumors such as WNT/βcatenin signaling are associated with T-cell exclusion (4,(8)(9)(10)(11). However, other mechanisms underlying the prevention of tumor infiltration by T cells are largely unknown.
DOCK2 is a mammalian homolog of Caenorhabditis elegans CED-5 and Drosophila melanogaster Myoblast City, and is predominantly expressed in hematopoietic cells (12,13). Although DOCK2 does not contain the Dbl homology (DH) domain that is typically found in guanine nucleotide exchange factors (GEFs), DOCK2 mediates the GTP-GDP exchange reaction for Rac via its DOCK homology region (DHR)-2 domain (14,15). Accumulating evidence in mice indicates that DOCK2 is a major Rac GEF acting downstream of chemoattractant receptors and antigen receptors in lymphocytes and plays key roles in their migration and activation (13,(16)(17)(18)(19). In addition, it has been shown that DOCK2 mutations cause severe immunodeficiency in humans (20). Thus, DOCK2 is essential for immune surveillance mechanisms in humans and mice.
Sulfation plays an important role in the regulation of the biological activity of many endogenous molecules. We earlier reported that cholesterol sulfate (CS), but not cholesterol, acts as an endogenous inhibitor of DOCK2 (21). Indeed, CS directly binds to the catalytic DHR-2 domain of DOCK2, and inhibits its Rac GEF activity (21). This inhibitory effect is specific for CS, and no such effect was seen for other cholesterol derivatives (21). In both humans and mice, sulfation of cholesterol is mediated by the sulfotransferases SULT2B1b and, to a lesser extent, SULT2B1a, which are produced from the same gene, SULT2B1, through alternative splicing (21)(22)(23)(24). However, SULT2B1b is also known to sulfate and inactivate oxysterols, cholesterol-oxidized products that have been shown to favor tumor growth by directly promoting tumor cell growth and indirectly by dampening anti-tumor immune responses (25)(26)(27). Therefore, the expression of SULT2B1b in tumors would affect anti-tumor responses differently in oxysterol-producing tumors and oxysterol-non-producing tumors.
In this study, we show that CS is abundantly produced in certain types of human cancer, such as colon cancer, and CS-producing cancer cells exhibit resistance to cancerspecific T-cell transfer and immune checkpoint blockade unless they produce oxysterols. Thus, our findings define a previously unknown mechanism of tumor immune evasion and provide a novel insight into the development of effective immunotherapies.

Human tissue samples
Human cancer tissue samples and normal colon tissue samples were obtained during surgery from patients in compliance with Institutional Review Board protocols of Kyushu University Hospital, Keio University Hospital or Saitama Medical University Hospital. Immediately after resection during surgery, the specimens were placed in liquid nitrogen and were stored for analyses. The study protocol was approved by the institutional review board at each Cholesterol sufate mediates tumor immune evasion 279 participating institution. Informed consent was obtained by an opt-out/opt-in approach, according to each participating institution's policy.

Mice and treatments
C57BL/6 mice and BALB/c nude mice were purchased from Japan Clea (Tokyo, Japan). Sult2b1 -/mice were obtained from the Jackson Laboratory (stock no. 018773, Bar Harbor, ME, USA), which had been backcrossed with C57BL/6 mice more than seven generations before use (21). DOCK2 -/mice on a C57BL/6 genetic background have been described previously (13) and were used in some experiments after crossing with OTII T-cell receptor (TCR) Tg mice. Mice were maintained under specific pathogen-free conditions in the animal facility of Kyushu University. The animal experiment protocol was approved by the committee of Ethics of Animal Experiments, Kyushu University. When indicated, 200 µl of anti-mouse PD-L1 (programmed cell death ligand 1) antibody (10F.9G2, 1 mg ml −1 , BioXCell, West Lebanon, NH, USA) or isotype-matched control (LTF-2, 1 mg ml −1 , BioXCell) was injected intra-peritoneally into tumor-bearing mice. For T-cell transfer, activated OTII CD4 + T (2 × 10 6 ) and OTI CD8 + T cells (5 × 10 5 ) were injected intravenously into tumor-bearing mice.
For the expression of PD-L1, the pBJ-neo vector encoding mouse PD-L1 was used. To express the I-A b /OVA complex, cells were transfected with the pBJ-neo vector encoding mouse I-A b α along with the pCAG vector encoding I-A b β covalently bound to OVA peptide (OVA323-339: ISQAVHAAHAEINEAGR) at a 1:5 ratio using polyethylenimine. After selection with G418 (1 mg ml −1 ), clones with the comparable expression of PD-L1 or the I-A b /OVA complex were selected and used. For expression of OVA protein, the pBJneo vector encoding OVA with a FLAG tag was transfected into E0771-control or E0771-SULT cells. Clones were isolated by limiting dilution, following selection with 1 mg ml −1 of G418.

Knockout of cholesterol 25-hydroxylase or Sult2b1
To genetically inactivate the expression of cholesterol 25-hydroxylase (CH25H) or SULT2B1b in MC38 or Pan02 cells, respectively, the CRISPR-Cas9 genome editing system was used in combination with gene targeting by homologous recombination (HR).
The amplified DNAs were inserted into the EcoR I-Bgl II and BamH I-Sal I sites of the HR410PA-1 targeting vector (System Biosciences), which carries the GFP expression cassette and puromycin resistance gene between the homology arms. Pan02 cells were transfected with 1.25 µg of sgRNA vector and 1.25 µg of HR targeting vector using 10 µl of Lipofectamine 2000 (Invitrogen) in 60 mm culture dishes. Forty-eight hours after transfection, cells were selected with puromycin (2 µg ml −1 ) and cloned by limiting dilution. Cells with genetically inactivated SULT2B1b expression were identified by western blotting.

RT-qPCR
Total RNA was isolated using ISOGEN (Nippon Gene, Tokyo, Japan Comparison of SULT2B1b expression among human colon cancer tissues and mouse cell lines was performed by quantitative PCR using the plasmid DNA encoding human or murine SULT2B1b as a standard (21). Briefly, a series of a 1:4 dilution of the plasmid DNA was amplified to obtain a linear standard curve for Cq (threshold cycle) to M[Log (-2)] (R 2 > 0.97) using the primer pairs 5′-ATGACATCTCGGAAATCAG CCA-3′ and 5′-GCACATCTTGGGTGTTCTCCG-3′ for human SULT2B1b, and 5′-CCCTGTGGAGCTCGTCTGAG-3′ and 5′-GACTGAGGCTCTCCGGTGAG-3′ for mouse Sult2B1b.
The concentration of target genes in each sample was determined using the standard curve, and normalized by dividing with the amount of total cDNA synthesized.

Transplantation and in vivo tumor growth
After washing with phosphate-buffered saline (PBS), the following numbers of cancer cells were re-suspended in   . Acquired data were uploaded to a Cytobank web server 8.0 (Cytobank, Santa Clara, CA, USA) for data processing and gating out of dead cells, doublets and normalization beads. The antibody information is listed in Supplementary Table S1.

Matrix-assisted laser desorption/ionization imaging mass spectrometry
Matrix-assisted laser desorption/ionization (MALDI) imaging analyses were performed as described previously (21), using an Ultraflextreme MALDI-TOF/TOF mass spectrometer and 7T FT-ICR-MS (Solarix Bruker Daltonik) equipped with an Nd:YAG laser. Data were acquired in the negative reflectron mode with raster scanning by a pitch distance of 50 µm. Each spectrum was the result of 300 laser shots at each data point.
In the TOF/TOF measurement, signals between m/z 50 and 1000 were collected. In FT-ICR-MS imaging, signals between m/z 300 and 500 were collected by continuous accumulation of selected ion mode. Image reconstruction for both was performed using the FlexImaging 4.1 software program (Bruker Daltonics). Molecular identification was performed using FT-ICR-MS data; the high mass accuracy provided by FT-ICR-MS allowed selective ion signals for the metabolites to be obtained within a mass window of 5 p.p.m., enabling the identification of the specific elemental composition of compounds by querying the highly accurate masses against database.

Quantification of CS and oxysterols by mass spectrometry
The amounts of CS in human cancer tissue samples were quantified as described previously (21). To monitor the production of CS and/or oxysterols in E0771, 3LL, E.G7-OVA and MC38 cells with or without the expression of SULT2B1b, cells were seeded at 5 × 10 5  The MRM mode and a dwell time of 2 ms per channel were used. The absolute contents of CS and oxysterols were calculated based on the calibration curves. The MRM calibration curves were generated from the triplicate analyses of these standard solutions using the chromatographic peak area of each analyte to that of the IS.

Quantification and statistical analysis
Statistical analyses were performed using Prism 7.0 (GraphPad Software, San Diego, CA, USA). The data were initially tested for normality using the Kolmogorov-Smirnov test. Parametric data were analyzed using a two-tailed unpaired Student's t-test when two groups were compared. Nonparametric data were analyzed with a two-tailed Mann-Whitney test when two groups were compared. Statistical differences between more than two experimental groups were evaluated using an analysis of variance (ANOVA) with Dunnett's post hoc test. The P-value of Kaplan-Meier curves was determined using the log-rank test. P-values less than 0.05 were considered statistically significant.

CS production and CD8 + T-cell infiltration are inversely correlated in human colon cancer tissues
The public database of the Human Protein Atlas (https://www. proteinatlas.org/ENSG00000088002-SULT2B1/pathology) indicates that many types of tumors, including colon cancers, highly express SULT2B1 (Fig. 1A). In contrast, the expression of DOCK2 is very low in all samples tested, and there is no association between expression levels of SULT2B1 and DOCK2 (Supplementary Figure 1), suggesting that high levels of SULT2B1 expression are mainly attributed to its expression in cancerous tissues. When the clinical course was compared between colon cancer patients with high and low SULT2B1 expression, SULT2B1-high patients showed poor prognosis (Fig. 1B). Quantitative mass spectrometric analyses also revealed that CS was abundantly produced in colon cancer tissues, compared with that in normal colon tissues (Fig. 1C). Importantly, the CS levels in the same specimen from the colon cancer tissues (see CS-high region and CS-low region) were inversely correlated with infiltration of CD8 + T cells (Fig.  1D). From these results, we hypothesized that cancer-derived CS may be involved in tumor immune evasion.

CS production in tumors inhibits infiltration by effector T cells
To functionally examine the role of cancer-derived CS in tumor immune evasion, we retrovirally expressed SULT2B1b in a mouse breast cancer cell line E0771, which lacks its endogenous expression (Supplementary Figure 2A). In comparison to E0771-MOCK expressing the vector alone (pMX-IRES-GFP), this cell line (designated E0771-SULT) not only produced, but also secreted, large amounts of CS ( Fig. 2A). In 'trans-cancer' migration assays, the presence of E0771-SULT cells in Matrigel significantly reduced CCL21-induced T-cell migration as compared with that of E0771-MOCK (Fig. 2B). Similar results were obtained when E0771-MOCK cells were treated with CS (Fig. 2B), indicating that CS mediates T-cell exclusion. Although the expression of SULT2B1b did not affect the growth of E0771 cells in vitro (Fig. 2C), the growth of E0771-SULT cells was much faster than E0771-MOCK cells when transplanted into syngeneic C57BL/6 mice (Fig. 2D). Similar results were obtained when Sult2b1-deficient (Sult2b1 -/-) C57BL/6 mice were used as recipients (Fig. 2E)

. A CyTOF
Cholesterol sufate mediates tumor immune evasion 283 analysis revealed that infiltrating CD8 + T cells and their granzyme expression were reduced in the case of E0771-SULT cells (Fig. 2F and G), suggesting that cancer-derived CS suppresses anti-tumor T-cell responses, most likely by inhibiting the Rac GEF activity of DOCK2 (21). In support of this, the difference in tumor growth between E0771-SULT and E0771-MOCK disappeared when transplanted into nude mice lacking T cells (Fig. 2H) or DOCK2 -/mice (Fig. 2I). Although the expression of Sult2b1b in E0771-SULT cells was higher than that in human colon cancer tissues, a mouse pancreas cancer cell line Pan02 showed comparable expression (Supplementary Figure 3). We also found that genetic deletion of Sult2b1 in Pan02 (Pan02-ΔSULT) suppressed tumor growth in C57BL/6 mice ( Fig. 2J; Supplementary Figure 2B), with an increased percentage of tumor-infiltrating CD8 + T cells (Fig. 2K).

CS production renders tumor cells resistant to immune checkpoint blockade
The interaction between PD-1 and PD-L1 is one of the most prominent targets for immune checkpoint blockade (1).  CD11b + antigen-presenting cells was unchanged between C57BL/6 mice bearing E0771-SULT and those bearing E0771-MOCK ( Fig. 3A and B). In addition, PD-1 and Tim3, markers for exhausted T cells, were comparably expressed on tumor-infiltrating T cells, splenic T cells and lymph node T cells from these mice (Supplementary Figure 4). The expression of PD-L1 was low on E0771-MOCK cells, even after treatment with IFN-γ (Supplementary Figure 5), and anti-PD-L1 antibody treatment only marginally affected growth of E0771-MOCK in C57BL/6 mice (Supplementary Figure 6). To clearly examine the effect of CS production on immune checkpoint blockade, we forced the expression of PD-L1 on E0771-SULT and E0771-MOCK cells (E0771-SULT-PD-L1 and E0771-MOCK-PD-L1), transplanted them into C57BL/6 mice, and started the antibody therapy when the tumor size reached around 150 mm 3 (Fig. 3C). As a result, in comparison to the E0771-SULT and E0771-MOCK. (G) The percentages of CD8 + T cells and NK cells in infiltrating leukocytes and their expression of granzyme B were compared between E0771-SULT and E0771-MOCK (n = 9). Data are presented as the mean ± SD. *P < 0.05 (two-tailed unpaired Student's t-test). (H and I) After transplantation, tumor growth in BALB/c nude mice (H; n = 7) or DOCK2 -/mice (I; n = 6) was compared between E0771-SULT and E0771-MOCK. (J) After transplantation, tumor growth in C57BL/6 mice was compared between Pan02-control and Pan02-ΔSULT (n = 10, 8). Data are presented as the mean + SD. **P < 0.01 (two-tailed Mann-Whitney test). (K) The percentages of CD8 + T cells in infiltrating leukocytes were compared between Pan02-control and Pan02-ΔSULT (n = 9). Data are presented as the mean ± SD. **P < 0.01 (two-tailed unpaired Student's t-test). isotype-matched control, anti-PD-L1 antibody significantly inhibited in vivo growth of E0771-MOCK-PD-L1 cells in C57BL/6 mice (Fig. 3D). In contrast, E0771-SULT-PD-L1 cells did not show any response to anti-PD-L1 antibody therapy (Fig. 3E) in spite of the comparable expression of PD-L1 on their surface (Supplementary Figure 5). Thus, CS production renders cancer cells resistant to immune checkpoint blockade.

CS-producing cancer cells exhibit resistance to cancerspecific T-cell transfer
To examine the role of CS in cancer-specific CD8 + T-cell responses, we used E.G7-OVA (29), which was developed by introducing the gene encoding full-length OVA into a mouse T-cell lymphoma cell line, EL4. In E.G7-OVA, through antigen processing of OVA protein, OVA257-264 peptide is presented by the MHC class I H-2K b molecule on their surface (29), which is recognized by CD8 + T cells from OTI TCR transgenic mice (30). When CD8 + T cells from OTI TCR transgenic mice were activated in vitro and injected intravenously into C57BL/6 mice on day 10 after transplantation of E.G7-OVA-MOCK, tumor progression was suppressed (Fig. 4A). However, E.G7-OVA cells expressing SULT2B1b (E.G7-OVA-SULT) exhibited resistance to this therapy, with a lower percentage of tumor-infiltrating OTI CD8 + T cells ( Fig.  4A and Supplementary Figure 2A). Similar results were obtained when the same protocol was applied to E0771 cells expressing OVA protein with or without the expression of SULT2B1b (E0771-OVA-SULT and E0771-OVA-control) (Fig. 4B). Recent evidence indicates that CD4 + T cells can target tumor cells in various ways, either directly by eliminating tumor cells or indirectly by modulating the tumor microenvironment (31,32). To examine the role of cancer-derived CS on tumor-specific CD4 + T-cell responses, we expressed the I-A b molecule covalently bound to OVA323-339 (I-A b /OVA complex; Fig. 4C) as an 'artificial tumor-specific antigen' on E0771 and 3LL with or without Sult2B1b expression (E0771-SULT-I-A b /OVA; 3LL-SULT-I-A b /OVA; E0771-control-I-A b / OVA; 3LL-control-I-A b /OVA) ( Supplementary Figures 2A and  7). OTII TCR recognizes OVA peptide (OVA323-339) presented by MHC class II I-A b molecules (33). When CD4 + T cells from OTII TCR transgenic mice were activated in vitro and injected intravenously into C57BL/6 mice on day 9 after transplantation of E0771-control-I-A b /OVA or 3LL-control-I-A b / OVA, tumor growth was markedly suppressed (Fig. 4D and  E). However, the adoptive transfer of OTII CD4 + T cells failed to inhibit tumor progression when they expressed SULT2B1b ( Fig. 4D and E). Similarly, this inhibitory effect was lost when DOCK2-deficient (DOCK2 -/-) OTII CD4 + T cells were administered (Fig. 4F). Collectively, these results indicate that CS production by tumor cells dampens anti-tumor T-cell responses by inhibiting the function of DOCK2.

The opposite effect of the expression of SULT2B1b on anti-tumor responses of oxysterol-producing tumors and oxysterol-non-producing tumors
Oxysterols such as 25-HC are oxidized derivatives of cholesterol that regulate the expression of a wide range of metabolic and inflammatory genes through liver X receptor (LXR) signaling (34)(35)(36). In addition, oxysterols have been shown to favor tumor growth directly, by promoting tumor cell growth, and indirectly, by dampening anti-tumor immune responses (25)(26)(27). Unlike E0771, Pan02, E.G7-OVA and 3LL, the mouse colon cancer cell line MC38 highly expressed CH25H (Fig. 5A), which is responsible for the conversion of cholesterol to 25-HC (37). Although CH25H is up-regulated in immune cells that are exposed to inflammatory mediators (38), the expression levels of CH25H in tumor-infiltrating macrophages and T cells were much lower in comparison to the level in MC38 cells (Supplementary Figure 8). Interestingly, we found that the expression of SULT2B1b in MC38 cells (MC38-SULT) did not promote-rather it inhibited-tumor progression in vivo (Fig. 5B). It is known that SULT2B1b sulfates oxysterols and inactivates them (24)(25)(26)(27)36). Consistent with this, quantitative mass spectrometric analyses revealed that in MC38-SULT cells, the production of 25-HC was lost, but the level of 25-HCS markedly increased (Fig. 5C). To examine whether 25-HC affects tumor growth, we treated MC38 cells with culture supernatants from various MC38-derivatives. Although the supernatant from wild-type MC38 cells promoted the cell growth of MC38 cells in vitro, this effect was not seen for those from MC38-SULT and MC38-ΔCH25H lacking the expression of CH25H (Fig. 5D), suggesting the relevance of 25-HC to tumor growth. Consistent with this, direct stimulation of MC38-ΔCH25H with 25-HC, but not 25-HCS, promoted tumor growth in vitro (Supplementary Figure  9). Importantly, we found that the expression of SULT2B1b in MC38-ΔCH25H cells significantly augmented tumor progression when transplanted into C57BL/6 mice (Fig. 5E). In addition, the Human Protein Atlas also indicates that the expression of cholesterol hydroxylases is relatively limited to certain tumor types, such as hepatocellular carcinoma and melanoma, and is low in Sult2b1-expressing cancers (Fig. 5F). Thus, SULT2B1b inhibition could be a therapeutic strategy to disrupt tumor immune evasion in many cancers.

Discussion
An important question in tumor immunology is why many patients exhibit resistance to immunotherapies. Although accumulating evidence indicates that resistance to immunotherapy is associated with a decrease in T-cell infiltration into tumors, the underlying mechanisms are still poorly understood. We identified cancer-derived CS as a key mediator of T-cell exclusion in the tumor microenvironment. By analyzing clinical samples, we found that the levels of CS were inversely correlated with the infiltration of CD8 + T cells in human colon cancer tissues. CS is a cellactive inhibitor of DOCK2, a Rac GEF critical for migration and activation of lymphocytes (21). Consistent with this, treatment of E0771-MOCK with CS inhibited CCL21-induced T-cell migration in vitro in trans-cancer migration assays. In addition, the cancer cell lines E0771, 3LL and E.G7-OVA engineered to produce CS exhibited resistance to cancer-specific T-cell transfer and immune checkpoint blockade in vivo, with a reduction of infiltrating T cells. Thus, CS constitutes a 'chemical barrier' to prevent tumor infiltration by effector T cells in several SULT2B1b-expressing tumor models.
Paradoxically, however, the expression of SULT2B1b in MC38 cells did not promote-rather it inhibited-tumor progression. Since the tumor-promoting activity of SULT2B1b could be recovered in MC38-ΔCH25H, it is clear that the production of 25-HC is involved in this process. SULT2B1b mediates CS production, but its other known substrates are oxysterols, such as 25-HC (24)(25)(26)(27)36). In this study, we have shown that 25-HC promoted growth of MC38 cells in vitro, whereas 25-HCS failed to do so. In addition, it has been reported that 25-HC, but not 25-HCS, impairs T-cell priming by inhibiting migration of dendritic cells (DCs) to the draining lymph nodes (26). Therefore, in tumors expressing cholesterol hydroxylases, the expression of SULT2B1b would lead to anti-tumor responses by inactivating oxysterols, whereas in tumors lacking cholesterol hydroxylases, the expression of SULT2B1b would dampen anti-tumor T-cell response by inducing tumor immune evasion. Importantly, in humans, the expression of cholesterol hydroxylases is relatively limited to certain tumor types and is low in Sult2b1-expressing cancers. Therefore, SULT2B1b could be a therapeutic target for promoting anti-tumor immunity in many cancers.

Cholesterol sufate mediates tumor immune evasion 289
In conclusion, we have shown that CS-producing cancer cells exhibit resistance to cancer-specific T-cell transfer and immune checkpoint blockade unless they produce oxysterols. Our findings define a previously unknown mechanism of tumor immune evasion and provide a novel insight into the development of effective immunotherapies.

Funding
This work was supported by the Leading Advanced Projects for Medical Innovation (LEAP; JP19gm0010001 to Y.F.) and Project for Cancer Research and Therapeutic Evolution (P-CAEATE; JP20cm0106362 to Y.F.) from Japan Agency for Medical Research and Development (AMED); Grants-in-Aid for Scientific Research A from the Japan Society for the Promotion of Science (JSPS; 19H00983 to Y.F.). Infrastructure of imaging metabolomics was partly supported by JST (Japan Science and Technology Agency) ERATO Suematsu Gas Biology Project.