Intravital Imaging Identifies the VEGF–TXA₂ Axis as a Critical Promoter of PGE₂ Secretion from Tumor Cells and Immune Evasion

Tumor cell-derived prostaglandin E₂ (PGE₂), a cyclooxygenase (COX) metabolite of arachidonic acid, promotes tumor progression by a number of mechanisms that modulate cell growth, invasion, migration, angiogenesis, immune evasion, and so on (1–3). Currently, it is believed that the concentration of PGE₂ within tumor tissues is regulated mostly by transcriptional regulation of proteins that constitute the pathways for the synthesis, transport, and degradation of PGE₂ (4).

PGE₂ synthesis starts from the activation of phospholipase A2 (PLA2), which liberates arachidonic acid from cell membrane phospholipids (5). Arachidonic acid is then oxygenated by cyclooxygenases COX1 and COX2 to yield PGH₂, which is the substrate of prostaglandin E synthases. The product, PGE₂, is secreted extracellularly and then binds to the G protein–coupled receptors (GPCR), designated as EP1, EP2, EP3, and EP4 (6, 7). Through their association with G_s, EP2, and EP4 connect PGE₂ to the protein kinase A (PKA) pathway. The activity of the arachidonic acid cascade is regulated by PLA2s, particularly cytoplasmic cPLA2α (5, 8). cPLA2α is activated primarily by the elevation of cytoplasmic Ca²⁺ concentration (8, 9). Curiously, despite its importance in tumorigenesis, the regulation of PGE₂ production and secretion from tumor cells in the tumor microenvironment (TME) remains largely elusive.

Thromboxane A₂ (TXA₂) synthesis also starts from the COX metabolite, PGH₂, which serves as the substrate of thromboxane A synthase (7). Similarly to PGE₂, TXA₂ has been implicated in many facets of tumorigenesis (3). TXA₂ is generated primarily by platelets but also by other cell types, including endothelial cells (10). Low-dose aspirin, which preferentially inhibits COX1 to reduce TXA₂ production in platelets, decreases the risk of not only myocardial infarction, but also metastatic cancer (11). However, our knowledge of the effect of TXA₂ is limited mostly to endothelial cells and hematopoietic cells, leaving unanswered the question of whether TXA₂ can directly influence tumor cells.

Using intravital imaging with genetically encoded calcium indicator, we explored the regulatory mechanisms of PGE₂ secretion, and identified the VEGF–TXA₂ axis as a potentially targetable regulator of PGE₂ secretion and immune evasion.

Plasmids encoding GCaMP6s (12) and tdTomato (13) were obtained from Addgene (plasmid #40753) and Takara Bio (#632533), respectively. The plasmid encoding AKAR3EV was reported previously (14). The plasmid encoding hM3D DREADD fused with mCherry (Addgene plasmid #50460) and Venus-Akaluc was provided by Dai Watanabe (Kyoto University Kyoto, Japan; ref. 15) and Atsushi Miyawaki (RIKEN Brain Science Institute, Tokyo, Japan; ref. 16), respectively. For in situ hybridization, partial cDNA of mouse Txas1 was synthesized and subcloned into pCR-BluntII-TOPO (Thermo Fisher Scientific). The details of recombinant plasmids construction are summarized in supplementary methods.

The Braf^V600E melanoma cell line was provided by Reis e Sousa at the Francis Crick Institute, London, United Kingdom (17). The breast cancer cell line 4T1 and human acute monocytic leukemia cell line THP-1 was purchased from ATCC. Madin–Darby canine kidney (MDCK) cells were purchased from the RIKEN BioResource Center (no. RCB0995). Human umbilical vein endothelial cells (HUVEC) were purchased from the Lonza Group, Ltd. The targeting sequences for CRISPR/Cas9-mediated gene knockout was summarized in Supplementary Table S1. The details of establishment of cell lines are summarized in Supplementary Materials and Methods.

C57BL/6NCrSlc mice, BALB/cCrSlc mice, and BALB/c nu/nu mice (nude mice) were purchased from SHIMIZU Laboratory Supplies. NOD/Shi-scid, IL-2RγKO Jic mice (NOG mice) were purchased from In-Vivo Science International Inc. Mice were housed in a specific pathogen-free facility and received a routine chow diet and water ad libitum. Female mice at the age of 7 to 11 weeks were used. The animal protocols were reviewed and approved by the Animal Care and Use Committee of Kyoto University Graduate School of Medicine (MedKyo20081).

Cells were harvested by trypsinization, washed three times with PBS, and injected subcutaneously into the flank of recipient mice at 2 × 10⁵ cells in 100 μL of 50% Matrigel (Corning) in PBS.

Tumors were cut into pieces and digested with Collagenase Type IV (200 U/mL; Worthington Biochemical Corporation) and DNase I (10 U/mL; Roche) for 30 minutes at 37°C. After staining, cells were analyzed with a FACS Aria IIu cell sorter (Becton Dickinson). Data were analyzed using FlowJo software (Tree Star). Dead cells were excluded by a Live/Dead Fixable Near-IR Dead Cell Stain Kit (Invitrogen). The details of antibodies are summarized in Supplementary Materials and Methods.

PGE₂ and TXB₂ concentrations in tumor tissues were quantified by ELISA according to the manufacturer's protocol (Enzo Life Sciences). Tumor lysate were prepared essentially as described previously (the details are summarized in Supplementary Materials and Methods; ref. 18). VEGF concentrations in tumor tissues were also quantified by ELISA according to the manufacturer's protocol (R&D Systems). The list of mice employed for ELISA analysis was summarized in Supplementary Table S2. TXA₂ secretion from HUVECs and THP-1 were quantified by ELISA as described previously (19).

Intracellular Ca²⁺ concentrations in Braf^V600E melanoma cells were visualized with a genetically encoded calcium indicator, GCaMP6s. The GCaMP intensity was quantified as the ratio of the fluorescence intensity at each time point to the fluorescence intensity at the minimum intensity projection over 10 minutes. To stimulate TP, 10 to 500 nmol/L I-BOP (Cayman Chemical Company) were added to the media after more than 24 hours serum starvation. For analysis of the PKA activity in MDCK cells, Braf^V600E melanoma cells expressing hM3D receptor and MDCK cells expressing AKAR3EV were seeded in a 96-well plate (1:400). The details of instrumentation settings and data processing are summarized in Supplementary Materials and Methods.

Intravital imaging was performed as previously described with some modifications (20). In brief, mice were anesthetized with 2% isoflurane (FUJIFILM WAKO Pure Chemical Corporation) inhalation (O₂ and air gas ratio was 80:20) and placed in the prone position on an electric heating pad. The body temperature was maintained at 36.5°C. The skin flap was then placed on a cover-glass. To obtain the proportion of cells experienced Ca²⁺ transients, a fluorescent intensity ratio image depicting the ratio of the maximum intensity projection to the minimum intensity projection over 10 minutes was prepared to represent the fold increase in GCaMP intensity. The ratio images were presented in the intensity-modulated display mode (IMD), with eight colors from red to blue representing the fold increase in GCaMP intensity and 32 grades of color intensity representing the fluorescence intensity according to the color scale shown in the respective figure. The proportion of Ca²⁺ transient-positive cells was calculated as the proportion of pixel areas whose fluorescence intensity increased at least 4.5-fold. The details of instrumentation settings, data processing, and in vivo cell depletion are summarized in Supplementary Materials and Methods.

After the administration (intraperitoneally) of 100 μL of 5 mmol/L AkaLumine-HCl, bioluminescent images were acquired every 1 minute. The maximum bioluminescent intensity during the imaging was adopted in each mouse. Image acquisition and analysis were carried out with MetaMorph software. The details of instrumentation settings are summarized in Supplementary Materials and Methods.

Antisense and sense digoxigenin (DIG)-labeled RNA probe for Tbxas1 was synthesized by using the RNA Labeling Kit (Roche). Sections of tumor tissues resected 4 days after inoculation were hybridized with antisense or sense RNA probes. Pictures were taken with a SZX16 stereo microscope (Olympus). The details of experimental procedure are summarized in Supplementary Materials and Methods.

Tumor tissues were resected 2 days after inoculation. Staining used the following primary and secondary antibodies: anti-CD31 (clone Mec13.3; BD Bioscience) and Alexa flour 488 goat anti-rat IgG antibody (Molecular Probes). Hoechst 33342 (Thermo Fisher Scientific) was used for nuclear staining. Stained sections were analyzed on a Fluoview FV1000 confocal microscope (Olympus). Image analysis was performed using MetaMorph software on maximum projections of Z-plane sections. The details of experimental procedure are summarized in Supplementary Materials and Methods.

Graphing and statistical analysis were performed with GraphPad Prism Software (GraphPad Software). The P values were assessed by unpaired Student two-sample t test.

How can we monitor the secretion of PGE₂? The rate-limiting step in PGE₂ secretion is Ca²⁺-induced activation of cytosolic phospholipase A2 (cPLA2; refs. 9, 21). Therefore, we reasoned that Ca²⁺ transients might be used as a surrogate marker for PGE₂ secretion. To investigate this possibility, we first visualized PGE₂ secretion from Braf^V600E melanoma cells by artificially increasing intracellular Ca²⁺ concentrations using the Designer Receptors Exclusively Activated by Designer Drug (DREADD) method (15). Braf^V600E melanoma cells expressing a Gq-coupled artificial GPCR were stimulated with the agonist clozapine-N-oxide (CNO) to elevate the cytoplasmic Ca²⁺ concentration. PGE₂ secretion was monitored by using Madin–Darby canine kidney (MDCK) cells expressing AKAR3EV, a Förster resonance energy transfer (FRET)-based biosensor for PKA (14). Secreted PGE₂ will bind to Gs-coupled EP2 and EP4 on the MDCK cells, culminating in PKA activation. As anticipated, CNO induced PKA activation in MDCK cells surrounding the Braf^V600E melanoma cells 30 seconds after CNO administration, and the activation ceased within 4 minutes (Fig. 1A; Video 1). Pretreatment with inhibitors against EP2 and EP4, PF-04418948 and ONO-AE3–208, respectively, abolished the CNO-induced PKA activation (Fig. 1A; Video 1). These results provided the basis for the use of Ca²⁺ transients as a surrogate marker of PGE₂ secretion from Braf^V600E melanoma cells.

The Ca²⁺ transients were monitored in Braf^V600E melanoma cells expressing a genetically encoded calcium sensor, GCaMP6s (12). After implantation into the subcutaneous tissue, we observed Ca²⁺ transients in the Braf^V600E melanoma cells under a two-photon excitation microscope (Fig. 1B; Video 2). When the threshold was set to a 4.5-fold increase in the fluorescence intensity of GCaMP6s, we found that the proportion of Ca²⁺ transient-positive cells reached its zenith 2 days after implantation and return to the basal level 10 days after implantation (Supplementary Fig. S1). Thus, we focused on the early phase of tumorigenesis thereafter. During 3 to 5 days after implantation, Ca²⁺ transients were observed in 5.3 ± 1.9% of Braf^V600E melanoma cells during a 10-minute observation period (data are from 20 mice). In stark contrast to the Braf^V600E melanoma cells in vivo, Ca²⁺ transients were rarely observed in vitro (Fig. 1C; Video 2). These observations suggest that Ca²⁺ transients are induced by ligand(s) supplied from the TME.

There are three major classes of cell surface receptors that trigger Ca²⁺ transients: GqPCR, calcium channels (CC), and transmembrane receptors directly or indirectly associated with tyrosine kinase (TK) activity (Fig. 1D). The contribution of these pathways was examined by using specific inhibitors. Among them, only the Gq inhibitor significantly suppressed Ca²⁺ transients in Braf^V600E melanoma cells (Fig. 1D). In line with this result, CRISPR/Cas9-mediated gene knockout of Gnaq, which encodes guanine nucleotide-binding protein G(q) subunit alpha, abolished Ca²⁺ transients (Fig. 1E; Video 3). Re-introduction of the Gnaq gene into Gnaq^−/− cells restored the Ca²⁺ transients to a level similar to that of parental cells (Fig. 1E; Video 3). Collectively, these results indicate that the GqPCR signaling pathway is responsible for triggering Ca²⁺ transients in Braf^V600E melanoma cells.

Autocrine binding of PGE₂ to EP2 and EP4 will trigger cAMP production in Braf^V600E melanoma cells. Therefore, to monitor the PGE₂ secretion in vivo, we examined the transcriptional activity of a cAMP response element (cAMP-RE)-driven promoter. The transcriptional activity of cAMP-RE was assessed 4 days after implantation (Fig. 2A). As expected, the transcriptional activity of cAMP-RE was markedly suppressed in Gnaq^−/− as well as Ptgs1/Ptgs2^−/− Braf^V600E melanoma cells (Fig. 2B, left). Comparable levels of bioluminescence were obtained among parental, Gnaq^−/− and Ptgs1/Ptgs2^−/− Braf^V600E melanoma cells when the ubiquitous promoter CAG was used, negating the possibility that the difference in cell growth rate biased the results (Fig. 2B, right). Of note, these cells grew at similar rates in vitro (Supplementary Fig. S2A). In agreement with the decrease of cAMP-RE transcriptional activity, PGE₂ in Gnaq^−/− tumors was markedly reduced as well as in Ptgs1/Ptgs2^−/− tumors (Fig. 2C). Collectively, these results demonstrate that the GqPCR-Ca²⁺ signaling pathway plays a major role in the high PGE₂ concentration within tumor tissues.

Do GqPCR-mediated Ca²⁺ transients cause immune evasion? As anticipated, Gnaq^−/− tumors started to regress around 8 days after implantation as Ptgs1/Ptgs2^−/− tumors did (Fig. 2D). Re-expression of Gnaq restored the immune evasion of Gnaq^−/− tumors (Fig. 2D). Pharmacologic inhibition of Gq or COX also decreased the growth rate of tumors, albeit less efficiently (Fig. 2E). We confirmed that the tumor regression was caused by antitumor immunity by using immuno-deficient mice; Gnaq^−/− Braf^V600E melanoma cells were able to grow in NOD/Shi-scid, IL-2RγKO Jic (NOG) mice as efficiently as parental Braf^V600E melanoma cells did (Fig. 2F). Moreover, the mice that had already rejected Gnaq^−/− Braf^V600E melanoma were resistant to the subsequent challenge with parental Braf^V600E melanoma cells, indicating the development of antitumor immunity by the preceding challenge with Gnaq^−/− melanoma cells (Fig. 2G). To further investigate the involvement of antitumor immunity, we examined the accumulation of CD8⁺ T cells and cDC1, both of them are key for subsequent anti-tumor immunity (Fig. 2H). Gnaq^−/− Braf^V600E melanoma showed greater accumulation of CD8⁺ T cells and cDC1 in total numbers as Ptgs1/Ptgs2^−/− Braf^V600E melanoma did (Fig. 2I). These results further support that GqPCR-Ca²⁺ signaling pathway plays a major role in PGE₂-dependent immune evasion. We confirmed that Cas9 expression alone has limited influence on antitumor immunity by using the cells expressing Cas9 alone (Supplementary Fig. S2B).

To extend our findings to another mouse strain, we used 4T1 breast cancer cells having a BALB/c background. We were able to recapitulate Ca²⁺ transients in vivo, which was suppressed by a Gq inhibitor (Supplementary Fig. S3A). Genetic ablation of Gnaq (Gnaq^−/−) in 4T1 breast cancer cells also resulted in a significant suppression of Ca²⁺ transients (Supplementary Fig. S3A). As was observed in Braf^V600E melanoma cells, Gnaq^−/− 4T1 breast cancer cells also exhibited a T-cell-dependent, but not macrophage-dependent, rejection 10 days after implantation, while T-cell independent factors are likely to involved as well in the reduced tumor growth (Supplementary Figs. S3B and S3C). Collectively, these results demonstrated that the GqPCR pathway drives Ca²⁺ transients and thereby causes immune evasion in both Braf^V600E melanoma cells with a C57BL/6 background and 4T1 breast cancer cells with a BALB/c background.

Which GqPCR ligand causes Ca²⁺ transients in the Braf^V600E melanoma cells in vivo? During the course of our experiments, we noticed that the COX inhibitor was able to suppress Ca²⁺ transients (Supplementary Figs. S4A and S4B). Thus, we focused on the following three GqPCRs, the ligands of which are COX metabolites: the PGE₂ receptor EP1 subtype encoded by Ptger1 (EP1), TXA₂ receptor encoded by Tbxa2r (TP), and prostaglandin F₂α receptor encoded by Ptgfr (FP). Among them genetic ablation of Tbxa2r, but not the others, resulted in almost complete suppression of Ca²⁺ transients (Fig. 3A; Supplementary Fig. S5A) and marked decrease in PGE₂ without significant decrease in TXA₂ (Fig. 3B). Because of the short half-life, TXA₂ is typically monitored by measurement of TXB₂. In this study, the concentration of TXB₂ was considered as that of TXA₂. Accordingly, Tbxa2r^−/− tumors started to regress around 8 days after implantation as observed in Gnaq^−/− or Ptgs1/Ptgs2^−/− tumors (Fig. 3C). The effect of TXA₂ on Braf^V600E melanoma cells was confirmed by a TXA₂ mimetic I-BOP, which triggered Ca²⁺ transients in vitro (Supplementary Fig. S5B). These results clearly identified TXA₂-TP signaling as the primary pathway that dictates Ca²⁺ transients, thereby facilitating PGE₂ secretion and immune evasion.

TXA₂ was originally described as being released from platelets, but is now known to be released by a variety of cells, including myeloid-lineage cells and endothelial cells (22, 23). Which host cells produce TXA₂ to stimulate Braf^V600E melanoma cells? Or do tumor cells themselves secrete TXA₂? Knockout of thromboxane A synthase 1 in Braf^V600E melanoma cells (Tbxas1^−/−) did not have significant effect on Ca²⁺ transients, negating the autocrine stimulation of TP (Fig. 4A). To examine the source of TXA₂, we performed in situ RNA analysis for Tbxas1. We detected Tbxas1 signal in endothelial cells specifically, but not in tumor cells (Supplementary Fig. S6A). This result further supports the notion that host cells serve as the source of TXA₂. We also confirmed the infiltration of CD31⁺ endothelial cells as early as 2 days after implantation (Supplementary Fig. S6B). Cell depletion using specific antibodies targeting platelets or neutrophils, clodronate liposome targeting macrophages, and NOG mice did not suppress the Ca²⁺ transients (Supplementary Figs. S6C–S6E). Collectively, these results imply that endothelial cells might serve as the source of TXA₂.

To gain an insight into the origin of TXA₂, we examined contribution of VEGF, which has been implicated in induction of TXA₂ synthesis in vascular endothelial cells (19). We first confirmed that VEGF is already detectable in Braf^V600E melanoma 2 days after implantation (Supplementary Fig. S6F). Human umbilical vein endothelial cells (HUVEC) secreted TXA₂ by VEGF-administration, even though differentiated macrophages (THP-1) also did (Supplementary Fig. S6G). We next examined the effect of a VEGFR inhibitor, motesanib, on TXA₂-TP mediated PGE₂ secretion from Braf^V600E melanoma cells. Intravital imaging revealed that Ca²⁺ transients in Braf^V600E melanoma cells were robustly suppressed within 30 minutes after intravenous administration of motesanib (Fig. 4B; Video 4). Accordingly, TXA₂ and PGE₂ concentrations were decreased in tumor tissues 4 days after inoculation (Fig. 4C), but not in tumor tissues 14 days after inoculation (Supplementary Fig. S7). Similar results were obtained with anti-VEGF antibody treatment (Fig. 4D and E). Collectively, these results strongly suggest that VEGF increases intra-tumoral TXA₂ and, thereby, drives Braf^V600E melanoma cells to secrete PGE₂. Altogether, we clarified the indispensable role of VEGF receptor signaling in PGE₂ secretion mediated by host endothelial cell-derived TXA₂, providing a molecular basis underlying the immunomodulatory effect of anti-VEGF therapies.

During the course of our experiments, we noticed that genetic ablation of COX1 (Ptgs1^−/−) alone also decreased the fraction of the Ca²⁺ transient-positive cells (Fig. 5A and B). This observation was unexpected, because we considered PGE₂ secretion as a downstream event of Ca²⁺ transients in tumor cells. These results suggest that tumor cell-derived COX1 metabolites is required in the early phase of TME to induce Ca²⁺ transients for PGE₂ secretion, as a positive feedback loop.

Finally, we examined the possibility that COX metabolites can systemically modulate the TME to induce Ca²⁺ transients by using a bilateral tumor burden model. When GCaMP6s-expressing parental cells were inoculated contralaterally, Ptgs1/Ptgs2^−/− cells carrying the Ca²⁺ indicator did not show Ca²⁺ transients (Fig. 5C). However, when Ptgs1/Ptgs2^−/− cells were mixed with parental cells without GCaMP6s and inoculated ipsilaterally, the Ptgs1/Ptgs2^−/− cells exhibited Ca²⁺ transients (Fig. 5C). To confirm that the effect of COX metabolites is limited to local Braf^V600E melanoma cells, we examined the transcriptional activity of the Ca²⁺-responsive NFAT promoter (Fig. 5D). The transcriptional activity of the NFAT promoter in Ptgs1/Ptgs2^−/− cells was markedly lower than that of parental cells injected contralaterally (Fig. 5D). The NFAT activity in parental cells was also suppressed by daily intraperitoneal injection of a COX inhibitor. A similar result was obtained by using Gnaq^−/− cells (Fig. 5D). Further, contralateral inoculation of parental cells did not result in any enhanced growth of Ptgs1/Ptgs2^−/− cells or Gnaq^−/− cells (Fig. 5E). Collectively, these results indicate that COX1 metabolites, probably PGE₂, acts locally to induce Ca²⁺ transients and promote immune evasion.

Here, we have provided evidence that VEGF receptor signaling plays a pivotal role in tumor cell-derived PGE₂ secretion mediated by host endothelial cell-derived TXA₂, thereby promoting immune evasion in the early stage of tumorigenesis (Fig. 6).

VEGF is now recognized as one of the most potent promoters of various aspects of tumorigenesis, including angiogenesis, invasiveness, metastasis, and recurrence (18). Moreover, previous studies have also demonstrated that anti-VEGF therapy activates antitumor immunity in animal models (24–26). In line with these animal experiments, several clinical trials have successfully evaluated the combination of immune checkpoint inhibitors (ICI) with VEGF/VEGFR blockade (27). Indeed, ICIs in combination with VEGFR inhibitors have become a new standard of care in treatment-naïve patients with advanced renal cell carcinoma (28). The new pathway revealed in this study provides a molecular basis underlying the effect of anti-VEGF therapy on the augmentation of tumor immunity.

How does VEGF induce TXA₂ within tumor tissues? Upon motesanib administration, it took only 20 minutes until the suppression of Ca²⁺ transients (Fig. 4B). Therefore, we can speculate that VEGF drives TXA₂ secretion through the activation of cPLA2 without gene expression change. Alternatively, VEGF may cause leakage of intravascular TXA₂ by maintaining high vascular permeability in the tumor tissue. Our previous study revealed that VEGFR inhibition reduces vascular permeability in endothelial cells (20). Further studies are needed to clarify the precise role of VEGFR signaling in increasing TXA₂ concentration, promoting PGE₂ secretion and immune evasion.

Intravital imaging of mice is one of the cutting-edge techniques to untangle the complex intercellular communications within the TME, as it provides information on cell dynamics at the single-cell level (29, 30). Importantly, such cell-to-cell communication cannot be recapitulated in vitro. In fact, we rarely observe PGE₂ secretion from Braf^V600E melanoma cells in vitro in the absence of stimulation. Only intravital imaging has enabled us to uncover PGE₂ secretion from a small fraction of tumor cells (Fig. 1C). This raises a question: Why do not all tumor cells exhibit Ca²⁺ transients? One possibility is the difference in responsiveness. In fact, even under in vitro conditions, we observed that only fraction of cells exhibited Ca²⁺ transients upon the stimulation with TP agonist (Supplementary Fig. S5B). Another possibility is that tumor cells in close proximity to TXA₂-secreting cells exhibit Ca²⁺ transients. Further analysis with high resolution intravital imaging in combination with techniques to visualize the activation status of host cells would lead to an answer.

One of the most surprising results of our study was that COX ablation, especially COX1 ablation, abolished Ca²⁺ transients (Fig. 5A and B). Considering the results of genetic ablation in TXA₂ synthase (Fig. 4A), our data support the notion that PGE₂ secreted from Braf^V600E melanoma cells promotes PGE₂ secretion via enhancement of the TXA₂ secretion from host endothelial cells (Fig. 6). This kind of positive feedback for PGE₂ secretion has rarely been observed but is still suggested to exist under physiological and pathological conditions including tumorigenesis (31, 32).

In this study, we have shown that TXA₂ released from host cells stimulates PGE₂ secretion and immune evasion. The TXA₂-TP signaling and/or VEGR receptor signaling would be promising targetable component of the PGE₂ secretion machinery, and thus could give rise to novel strategies for the prevention and/or treatment of multiple types of malignancies, especially in combination with immunomodulatory agents.

Y. Konishi reports grants from The Japan Society of the Promotion of Science (JSPS) during the conduct of the study. A. Takaori-Kondo reports grants from Celgene, Ono Pharmaceutical, grants, and personal fees from Bristol-Myers Squibb, Novartis, personal fees from MSD, and personal fees from Kyowa Kirin outside the submitted work. K. Terai reports grants from JSPS KAKENHI during the conduct of the study. No disclosures were reported by the other authors.

Y. Konishi: Conceptualization, resources, data curation, formal analysis, funding acquisition, validation, investigation, methodology, writing–original draft. H. Ichise: Investigation, methodology. T. Watabe: Investigation, methodology. C. Oki: Resources, methodology. S. Tsukiji: Resources, methodology, writing–review and editing. Y. Hamazaki: Methodology, writing–review and editing. Y. Murakawa: Writing–review and editing. A. Takaori-Kondo: Writing–review and editing. K. Terai: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, methodology, project administration, writing–review and editing. M. Matsuda: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, methodology, project administration, writing–review and editing.

The authors are grateful to Atsushi Miyawaki for technical suggestions on Akaluc bioluminescence imaging, Dean Thumkeo for technical suggestions on the quantification of PGE2, the members of the Matsuda Laboratory for their helpful input, and to the Medical Research Support Center of Kyoto University for DNA sequence analysis. This work was supported by the Kyoto University Live Imaging Center. Financial support was provided by JSPS KAKENHI grant nos. 16J09066, 19K23915, 20K17400, 20J01623 (to Y. Konishi), 18K07066, 21H02715 (to K. Terai), JST Moonshot R&D no. JPMJPS2022 (to M. Matsuda), and JST CREST no. JPMJCR1654 (to M. Matsuda).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.