HVJ-E links Apolipoprotein d to antitumor effects

Abstract

Background

Virotherapy eradicates tumors by directly killing cancer cells and causing adjuvant effects. However, the mechanism by which non-replicating virotherapy exerts anti-tumor effects is unclear.

Methods

In this study, we investigated the genes that mediate the anti-tumor effects of ultraviolet (UV)-irradiated Hemagglutinating Virus of Japan envelope (HVJ-E) using RNA sequencing, gene knockout, and a drug-inducible gene expression system. We examined the antitumor effects of Apolipoprotein d (Apod) using genome-wide CRISPR library screening, in situ biotinylation combined with mass spectrometry, flow cytometry, biochemistry, and tumor-bearing mouse models.

Results

Here, we show that HVJ-E represses tumor growth via Irf7-induced Apod expression in tumor cells in vivo. Irf7 in B16F10 cells is a pivotal transcription factor for HVJ-E-induced anti-tumor effects. Apod substantially suppresses tumor growth even in HVJ-E-insensitive tumors. Apod is required to increase NKG2D-ligand genes in HVJ-E-treated tumors. Genome-wide CRISPR library screening and in situ biotinylation of Apod reveal an association of Apod with ERK2. Mechanistically, Apod prevents the nuclear translocation of ERK2 and Importin7, increasing NKG2D-ligands expression in B16F10 cells and attenuating tumor growth. Treating a local tumor with a combination therapy of Apod with the anti-OX40, T cell costimulatory molecule, antibody substantially repressed tumor growth in target and non-target lesions alongside T cell activation.

Conclusion

Our findings provide insights into the molecular mechanisms of how HVJ-E induces anti-tumor effects and can aid the development of therapeutic strategies for eliciting anti-tumor immunity.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Non-replicating virotherapy induces anti-tumor effects. However, the mechanisms are not fully understood.

WHAT THIS STUDY ADDS

• We demonstrate that the non-replicating virotherapy HVJ-E induces anti-tumor effects via Apod. Apod inhibits the nuclear translocation of the ERK1/2-Importin 7 complex, leading to an upregulation of NKG2D-ligands in cancer cells. Furthermore, Apod enhances anti-tumor immune responses when combined with an OX40 agonist antibody in a tumor-bearing mouse model.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our findings provide insights into the development of therapeutic strategies for activating anti-tumor immunity.

Introduction

Several recent clinical trials showed suppression of tumor progression by viruses, including DNA viruses such as herpes simplex virus (HSV), adenovirus, vaccinia virus, poxviruses, and parvovirus, and RNA viruses such as vesicular stomatitis virus, reovirus, coxsackievirus, measles virus, Newcastle disease virus, and hemagglutinating virus of Japan (HVJ).1,4 Talimogene laherparepvec, a genetically-modified HSV-1, can effectively control tumor growth in patients with melanoma.5,7 The antitumor effects of oncolytic viruses rely on the lytic ability of cancer cells and their pleiotropic immune-stimulating properties through stimulating the interferon (IFN) signaling pathway by recognizing virus-derived nucleic acids and activating immune cells against tumors and the injected virus per se. However, viral replication in cancer cells is not always necessary to cause tumor regression, since several replication-incompetent viruses show antitumor effects.8,11 The mechanism of how a non-replicating virus induces antitumor effects is not fully understood.

HVJ-E, a Ultraviolet (UV)-irradiated form of HVJ envelope, has almost completely lost its replicating ability and has strong antitumor effects after intratumor injection in several mouse tumor models.^{11 12} We recently performed a first-in-human phase I study to evaluate the safety of GEN0101 (ie, clinically applied version of HVJ-E) in patients with advanced melanoma and castration-resistant prostate cancer and demonstrated the safety of GEN0101 in these patients.^{3 4 13} HVJ-E induces cell death in several cancer cell lines via pro-apoptotic gene expression in vitro¹¹ and activates dendritic cells (DCs) to promote T-cell cytotoxic activity against cancer cells.¹² HVJ-E transduces fragments of its viral RNA genome into cancer cells, thereby activating the cytoplasmic RNA receptor retinoic acid-inducible gene-I (RIG-I). This activation subsequently triggers the downstream mitochondrial antiviral signaling, leading to the activation of several inflammatory transcription factors, including interferon regulatory factor (IRF) 3 and 7.^{11 14} Pro-apoptotic factor NOXA and TNF-related apoptosis-inducing ligand are downstream factors for these transcription factors in the in vitro setting.¹¹ HVJ-E can also activate systemic antitumor effects with a combination of T-cell co-stimulatory molecule stimulation through promoting interaction between NKG2D and NKG2D-ligands (NKG2D-L).¹⁵ These data suggest that changes in the gene expression profile are critical for the HVJ-E-induced antitumor effects. However, it remains unclear how HVJ-E represses tumor growth in vivo.

In this study, we demonstrated that changing gene expression in cancer cells is critical for repressing tumor growth by HVJ-E. Irf7 elicited the HVJ-E-mediated antitumor effects in vivo. Apolipoprotein d (Apod) was a crucial downstream factor of Irf7 and induced antitumor effects by preventing the nuclear translocation of ERK2. Notably, local treatment combining Apod with anti-OX40, a T-cell costimulatory factor expressed by T cells, agonist antibody (OX40 antibody) induced significant systemic antitumor effects, accompanied by T-cell activation. Our findings will lead to the development of an antitumor therapeutic method based on these mechanisms.

Results

HVJ-E-induced antitumor effects require Irf7 expression in B16F10 tumors

To examine the mechanisms by which HVJ-E represses tumor growth, we first examined the contribution of lymphocytes to HVJ-E antitumor effects using a syngeneic melanoma model. We found that a total of three intratumor injections of HVJ-E with 2,000 hemagglutination units (HAU) per injection every other day suppressed the growth of B16F10 mouse melanoma cells (p=0.0004, online supplemental figure S1a). Furthermore, HVJ-E repressed tumor growth in MC38 mouse colon cancer cells injected into C57BL/6N mice and 4T1 mouse breast cancer cells and CT26 mouse colon cancer cells injected into BALB/c mice (online supplemental figure S1b). HVJ-E administration also showed a comparable antitumor effect on B16F10 tumors in SCID mice lacking functional T and B cells (p<0.0001, online supplemental figure S1c). Surprisingly, HVJ-E administration repressed tumor growth of B16F10 and CT26 mouse colon cancer cells in NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice lacking functional T, B, and natural killer (NK) cells (p<0.0001; online supplemental figure S1d,e) without amplification of HVJ RNA genome (online supplemental figure S1f). These results suggest that lymphocyte-independent mechanisms are primarily responsible for the HVJ-E-induced antitumor effects.

We hypothesized that HVJ-E causes a cell-intrinsic response in tumor cells to repress tumor growth. We first analyzed HVJ-E-induced gene expression in B16F10 cells in vivo. To avoid host-cell contamination in in vivo analysis, we isolated a Green fluorescent protein (GFP)-negative population (ie, tumor cells) in GFP-ubiquitously-expressing mice after HVJ-E administration. Comparing the gene expression profile of B16F10 tumors after three doses of HVJ-E or phosphate-buffered saline (PBS), we identified 27 significantly upregulated genes, including two transcription factors, Irf7 and Batf2 (p=5.8e−10 for Irf7, p=4.6e−7 for Batf2, figure 1a). To identify genes that play a pivotal role in HVJ-E-induced tumor regression, we generated nine knockout (KO) B16F10 cells for genes that are upregulated by HVJ-E or are pivotal for type-I IFN pathway (online supplemental figure S2). Of the two transcription factors, Irf7 plays a role in the activation of virus-inducible cellular genes, and Batf2 is involved in the development of dendritic cells.¹⁶ Irf7, but not Batf2, KO significantly attenuated HVJ-E-induced tumor growth suppression (Irf7 KO #1, p=0.0298 and Irf7 KO #2, p=0.0146; figure 1b and online supplemental figure S3). RIG-I-like receptors, such as Rig-I, Mda5, and Lgp2, detect intracellular viral RNA and activate Irf 3/7, thus leading to a type-I IFN-mediated antiviral response.¹⁷ HVJ-E transfers its fragmented negative-sense single-stranded RNA genome into the infected cell in a sialic acid-dependent manner.^{14 18} We therefore examined whether RIG-I-like receptors play a role in responding to HVJ-E infection in cancer cells. RIG-I-like receptors, including Ddx58, Lgp2, and Mda5, KO slightly attenuated the antitumor effects of HVJ-E as the previous finding,¹¹ but were still susceptible to HVJ-E-induced tumor growth suppression (figure 1b and online supplemental figure S3). We next investigated whether type-I IFN and its receptors interferon alpha and beta receptor subunit (Ifnar)1/2 are involved in HVJ-E-induced antitumor effects. Neither Ifnar1 nor Ifnar2 KO prevented HVJ-E-induced tumor growth suppression (figure 1b and online supplemental figure S3), suggesting that type-I IFN signaling is not the main pathway underlying HVJ-E-induced antitumor effects. HVJ-E intratumor injection also increased the expression of three apolipoprotein genes, Apod, apolipoprotein L9a (Apol9a), and apolipoprotein L9b (Apol9b) (figure 1b). We generated Apod and Apol9a/Apol9b KO B16F10 cells because Apol9a and Apol9b have almost identical sequences. Apod, but not Apol9a/Apol9b, KO significantly attenuated HVJ-E-induced tumor growth suppression (figure 1b and online supplemental figure S3). These data indicate that Irf7 and Apod in B16F10 cells are pivotal in HVJ-E-induced antitumor effects in vivo.

Figure 1. HVJ-E-induced antitumor effects require Irf7 expression in B16F10 tumors. (a) Heatmap of gene expression data in B16F10 tumors, showing genes with log2 fold change >4, expression >8 in the HVJ-E group, and p<0.01. (b) Box plot of relative tumor volume of two independent clones in each KO B16F10 cell line. Relative tumor volume showing the ratio of HVJ-E-treated tumor volume to PBS-treated tumor volume at day 14. P value was calculated by Steel’s test compared with the wild type group. Gray dots show each relative HVJ-E-treated tumor volume. The black dot indicates an outlier. Number of samples is shown in online supplemental figure S3. 2,000 hemagglutination units HVJ-E were intratumorally injected three times every 2 days. Apod, apolipoprotein d; Apol9a, apolipoprotein L9a; Apol9b, apolipoprotein L9b; HVJ, hemagglutinating virus of Japan; IFNR, Interferon Alpha And Beta Receptor; IRF, interferon regulatory factor; KO, knockout; PBS, phosphate buffered saline; RLR, RIG-I-like receptors; WT, wild-type.

To confirm the HVJ-E-induced Irf7 expression in B16F10 cells in vivo, we generated endogenous Irf7-monitoring B16F10 cells by homozygous knock-in of TY1 tag and three tandem mClover3 sequences just before the stop codon of endogenous Irf7 gene (Irf7-TY1-Clover B16F10 cells, online supplemental figure S4a). Intratumor HVJ-E injection induced mClover3 expression in~30–50% of the B16F10 tumor population, indicating that intratumor HVJ-E injection activated Irf7 expression in B16F10 cells (online supplemental figure S4b,c). These results suggest the contribution of alterations in gene expression in tumor cells to HVJ-E-induced antitumor effects.

Irf7 expression represses B16F10 tumor growth through Apod expression

Because Irf7 is necessary for HVJ-E-induced antitumor effects in B16F10 tumors, we sought to examine whether ectopic Irf7 expression in B16F10 cells is sufficient to suppress tumor growth in vivo. We engineered B16F10 cells with tetracycline-dependent GFP (control) and GFP-Irf7 expression by knocking in an expression cassette at the Gt(Rosa)26Sor locus (figure 2a). Doxycycline-dependent gene expression was confirmed by GFP expression in~100% of cells in vitro and in~40% of cells in vivo (online supplemental figure S5a,b). B16F10 tumor growth was suppressed by doxycycline-induced GFP-Irf7, but not by GFP expression (p<0.0001 for GFP-Irf7, figure 2b).

Figure 2. Irf7 expression represses B16F10 tumor growth through Apod expression. (a) A schema of the tetracycline-inducible expression DNA cassette in the Gt(Rosa)26Sor locus. rtTA, reverse tetracycline-controlled TransActivator; TRE3GS, Tet Response Element sequence; pause site, transcription block sequence. The TRE3GS promoter drives GFP and GFP-Irf7 expression. (b) The growth curve of B16F10 cells containing tetracycline-inducible GFP (Tet GFP) and Tet GFP-Irf7. 200 µL of 2 mg/mL doxycycline was orally administrated on the indicated days. P values were calculated using Student’s t-test. (c) Volcano plot of gene expression in HVJ-E-treated wild-type (WT) B16F10 cells and Irf7 KO B16F10 cells in vivo. The red dots show genes with significantly altered expression in the HVJ-E-treated B16F10 WT cells compared with PBS-treated B16F10 WT. Gene names are shown for genes with log2 fold change >2 between the HVJ-E-treated WT B16F10 and the HVJ-E-treated Irf7 KO B16F10 cells. x-axis, log2 fold change between the HVJ-E-treated WT B16F10 and the HVJ-E-treated Irf7 KO B16F10; y-axis, −log10 (p value). (d) Dot plot of relative Apod expression in the indicated plasmids-transfected cells, normalized to 18S (n=3). P values were calculated using the Tukey HSD. Error bars show the SD. (e) Dot plot of gene expression data in Apod KO B16F10 tumors. Tumor cells were collected 1 day after three doses of 2,000 hemagglutination units HVJ-E or PBS from GFP mice, and immune cells were depleted from the tumor cells using GFP by flow cytometry. The red dots show genes with significantly changed expression. y and x axes show log2 fold change. (f) Heatmap of NKG2D-ligands and major histocompatibility complex class I/II gene expression in PBS or HVJ-E-treated WT and Apod KO B16F10 tumors. (n=3). Apod, apolipoprotein d; GFP, Green fluorescent protein; HSD, honestly significant difference; HVJ, hemagglutinating virus of Japan; IRF, interferon regulatory factor; KO, knockout; PBS, phosphate buffered saline; RLR, RIG-I-like receptors.

We next sought to identify genes induced by HVJ-E in an Irf7-dependent manner. Comparing the gene expression profiles between HVJ-E-treated wild-type and HVJ-E-treated Irf7 KO B16F10 tumors identified Apod as an Irf7-dependent candidate gene (figure 2c). As consistent with the result, Irf7, but not Stat1, an inflammatory transcription factor, expression induced Apod expression (p<0.0001, figure 2d). Apod belongs to the apolipoprotein multigene family, which is related to lipid transfer. Apod is also involved in the regulation of cell proliferation.¹⁹ Notably, Apod KO suppressed HVJ-E-induced antitumor effects (figure 1b and online supplemental figure S2). Apod KO significantly attenuated changes in gene expression profile by HVJ-E, even though Irf7 was still upregulated in HVJ-E-treated tumors (log2 fold change of Irf7 expression (HVJ-E/PBS) 2.96) (figure 2e). Although HVJ-E-treated B16F10 tumors increased the expression of NKG2D-L, including H60b and Raet1a, and major histocompatibility complex (MHC)-class I/II genes, HVJ-E-treated Apod KO B16F10 tumors did not show the response (figure 2f). These data indicate that Irf7 and Apod in B16F10 cells play a pivotal role in HVJ-E-induced antitumor effects in vivo.

Apod is associated with tumor growth repression

We tested doxycycline-inducible GFP-Apod to examine whether Apod expression is sufficient to suppress tumor growth in vivo, similar to Irf7 (figure 2a,b). Apod expression diminished tumor growth in vivo (p=0.0014, GFP-Apod; figure 3a, online supplemental figure S6a–d), suggesting the possibility that Apod intrinsically induces antitumor effects.

Figure 3. Apod is associated with tumor growth repression. (a) Tumor growth curve of B16F10 cells containing tetracycline-inducible GFP-Apod in C57BL/6N mice. (b) Gene set enrichment distribution in doxycycline-treated B16F10 cells with Tet GFP-Apod in comparison to control treatment in vivo. The representative gene ontologies with significant differences (adjusted p value<0.05) in Tet GFP-Apod are shown. (c) CBB-stained SDS-PAGE gel of recombinant Apod protein purified from Expi293 cells using HiTrap DEAE and HisTrap HP columns. The protein size was the same as the expected size. (d) Tumor growth curve of B16F10 cells after murine Apod protein intratumoral injection every other day. P values were calculated using Turkey’s HSD test. (e) Tumor growth curve of LL/2 cells. 2,000 hemagglutination units HVJ-E were intratumorally injected on days 0, 2, and 4. (f) Dot graph of the percentage of sequencing reads assigned to the HVJ genome sequence in vivo. (n=3). (g) Tumor growth curve of LL/2 cells after intratumoral Apod protein injection every other day. (h) Kaplan-Meier curves of overall survival in the TCGA-SKCM dataset with APOD high or low expression divided at the median. TCGA-SKCM samples, including only cutaneous melanoma, were filtered to single data for a patient, metastasis, and no prior prognosis, resulting in 337 samples. P values were calculated using the log-rank test. (i) Tumor growth curve of MeWo cells with intratumoral injection of human APOD protein on the indicated days. (j) Tumor growth curve of serous adenocarcinoma patient-derived xenograft after intratumoral human APOD protein injection on the indicated days. P values were calculated using the Wilcoxon rank-sum test (i, j) and Welch’s t-test (a, e, g), according to the normality and variance of the data. Error bars show the SD. Apod, apolipoprotein d; CBB, Coomassie Brilliant Blue; GFP, Green fluorescent protein; HSD, honestly significant difference; HVJ, hemagglutinating virus of Japan; PBS, phosphate buffered saline; SDS-PAGE, sodium dodecyl sulfate–polyacrylamide gel electrophoresis; TCGA-SKCM, The Cancer Genome Atlas-skin cutaneous melanoma.

We next examined whether changing gene expression profiles mediates apolipoprotein-mediated antitumor effects by analyzing gene expression profiles in apolipoprotein-expressing tumors. Gene Set Enrichment Analysis revealed that Apod induces expression of genes related to responses to IFN-beta (p=0.0003) and suppressed gene expression related to cell adhesion (p=0.001), cell migration (p=0.0007), and epithelial cell proliferation (p=0.03) (figure 3b). The results suggest that apolipoproteins generate the virus-infected status and repress cell proliferation without virus infection in B16F10 cells in vivo.

To investigate if Apod expression in cancer cells per se is critical for the antitumor effects or whether the enrichment of extracellular Apod in the tumor microenvironment induces antitumor effects, we generated a recombinant murine Apod protein using the Expi293 cell system (figure 3c). Intratumoral injection of murine Apod protein repressed B16F10 tumor growth in a dose-dependent manner, with the antitumor effects reaching a plateau at 40 µg/injection (figure 3d and online supplemental figure S6e). B16F10 tumor contained~12 ng/tumor (g) 5 days after inoculation (online supplemental figure S6f), suggesting an increase of Apod from 0.0026 ng/mm³ to 880.1 ng/mm³ by the intratumoral injection of 40 µg Apod. HVJ-E increased in Apod in tumors under the GolgiStop condition (online supplemental figure S6g). The intratumoral injection of Apod increased cleaved PARP (online supplemental figure S6h,i), and the increase in HVJ-E-induced cleaved PARP was dependent on Apod (online supplemental figure S6j). Tumors derived from mouse LL/2 Lewis lung carcinoma cells were HVJ-E-resistant tumors since HVJ-E did not efficiently infect LL/2 cells in vivo (figure 3e,f). Notably, intratumoral injection of murine Apod significantly suppressed HVJ-E-resistant LL/2 tumor growth (p<0.0001, figure 3g). Moreover, high APOD expression was correlated with improved overall survival in The Cancer Genome Atlas-skin cutaneous melanoma (TCGA-SKCM) data²⁰ (figure 3h). The intratumoral injection of APOD protein repressed tumor growth in MeWo human melanoma cells and serous adenocarcinoma patient-derived xenograft (PDX) (p=0.0129 and p=0.0453, figure 3i,j), suggesting that the Apod-induced antitumor effects are not limited to melanoma. The PDX contained APOD at concentrations~200 ng/tumor (g), estimating a~20,000-fold increase of APOD in the tumors by the intratumoral injection of 40 µg APOD. Thus, Apod protein in the tumor microenvironment induces antitumor effects, including HVJ-E-resistant tumors.

Apod increases NKG2D-L expression through preventing the nuclear translocation of ERK2

We sought the mechanisms of Apod to modulate tumor growth. To identify genes related to Apod-mediated repression of tumor growth, we inoculated the mixture of CRISPR library-containing and Tet GFP-Apod-containing B16F10 cells since the extracellular Apod may enter cells using a receptor and inhibit a protein related to cell proliferation. Doxycycline-controlled Apod expression repressed CRISPR library-containing and Tet GFP-Apod-containing B16F10 tumor growth (p=0.0382, figure 4a). In contrast, GFP expression did not modulate tumor growth (p=0.667, figure 4a). Analyzing the composition of gRNAs in the tumors, we detected the enrichment of gRNAs against Mapk1, specifically in Apod-expressed tumors, compared with doxycycline-negative and GFP-expressed tumors (figure 4b). We next sought to identify proteins in close proximity to GFP-Apod proteins using in situ biotinylation in Tet GFP-Apod B16F10 cells. GFP-Apod was in close proximity to proteins related to GOs of Signaling by MET, Signaling to ERKs, and Membrane Trafficking (figure 4c and online supplemental table S4). However, we did not detect conventional apolipoprotein co-factors for carrying lipids because the cancer cells have limited expression of these genes. These data suggest that Apod is associated with ERK2 encoded in Mapk1.

Figure 4. CRISPR screening and in situ biotinylation of Apod reveal an association of Apod with ERK2. (a) Schema of CRISPR screening of genes related to Apod-mediated repression of tumor growth. Tumor growth curve of B16F10 cells harboring CRISPR library mixed with B16F10 cells containing Tet GFP or Tet GFP-Apod in C57BL/6N mice. 200 µL of 2 mg/mL doxycycline was orally administrated. The number in parentheses indicates the number of samples. Error bars show the SD. (b) Dot plot of p value of the enriched gRNAs with messenger RNA expression level comparing Dox+ and Dox− in Tet GFP and Tet GFP-Apod cells (n=5). The red dot shows p<0.00001. (c) Table of enriched GOs in the data of in situ biotinylation of GFP-Apod using anti-GFP antibody. We considered proteins with fivefold enrichment in Dox-treated Tet GFP-Apod cells, comparing both proteins enriched with control IgG antibody in Dox-treated Tet GFP-Apod cells and anti-GFP antibody in Dox-treated Tet GFP cells (n=3). The enrichment of GOs was calculated with Metascape. (d) Western blotting of ERK1/2 in WT and Mapk1 KO B16F10 cells. H2B is a loading control. (e) Tumor growth curve of WT and Mapk1 KO B16F10 cells in C57BL/6N mice. The number in parentheses indicates the number of samples. Error bars show the SD. P values were calculated using the Tukey HSD. (f) Tumor growth curve of Mapk1 KO B16F10 cells in C57BL/6N mice. Apod (40 µg) was intratumorally injected. The number in parentheses indicates the number of samples. Error bars show the SD. (g) Dot plot showing NKG2D-L median expression in WT and Mapk1 KO B16F10 cells in PBS and Apod-treated tumors. The number in parentheses indicates the number of samples. (h) Immuno-staining of ERK1/2 in Apod-treated or HVJ-E-treated B16F10 tumors. Apod (40 µg) or HVJ-E (2,000 hemagglutination units) was intratumorally injected on days 0, 2, and 4. Tumors were analyzed 1 day after the final treatment. (i) Dot plot showing relative nuclear ERK signals compared with the whole cell area. The number in parentheses indicates the number of samples. P values were calculated using Welch’s t-test. Apod, apolipoprotein d; GO, gene ontology; GFP, Green fluorescent protein; HSD, honestly significant difference; HVJ, hemagglutinating virus of Japan; KO, knockout; PBS, phosphate buffered saline; WT, wild-type.

To demonstrate whether the Apod function is related to ERK2, we generated Mapk1 KO B16F10 cells (figure 4d). Mapk1 KO attenuated tumor growth similar to Apod treatment (#1, p=0.0463; #2, p=0.0032; figures4e 3d). Apod treatment did not repress tumor growth of Mapk1 KO B16F10 cells (p=0.9162, figure 4f, online supplemental figure S7a), suggesting a requirement of Mapk1 for the Apod-induced antitumor effects. We examined NKG2D-L expression levels since HVJ-E increased NKG2D-L-related genes, including H60b and Raet1a, expression in an Apod-dependent manner (figure 2f). NKG2D-L is a critical factor for the response to virus infection and tumor immunity.²¹ Indeed, systemic antitumor effects induced by the combination of HVJ-E with OX40 antibody are dependent on the interaction between NKG2D and NKG2D-L.¹⁵ Apod treatment increased NKG2D-L in wild-type B16F10 tumors (p=0.0403) but not in the Mapk1 KO tumors (p=0.1177, figure 4g, online supplemental figure S7b). Immuno-staining of ERK1/2 indicated that Apod and HVJ-E treatment decreases ERK1/2 in the nucleus, compared with PBS treatment in B16F10 tumors (p<0.0001, figure 4h,i). The HVJ-E-induced decrease in nuclear ERK1/2 was dependent on Apod (p=0.2304, online supplemental figure S7c). These data suggest that Apod is involved in increasing NKG2D-L and preventing the nuclear localization of ERK1/2.

Because Importin7 regulates the nuclear translocation of ERK1/2 proteins,²² we thought that Apod is associated with ERK1/2 and/or Importin7. Apod was associated with Importin7, compared with the association of Apod with ERK1/2 (figure 5a). Apod treatment in vitro decreased in Importin7 and ERK1/2 in the nucleus of B16F10 cells (figure 5b). In situ biotinylation analysis of GFP-Apod (figure 4c) and genistein, an endocytosis inhibitor, treatment (online supplemental figure S7d) suggested that endocytosis is involved in the regulation of nuclear translocation of ERK1/2 by Apod using ERK1/2 protein level in the nucleus as a marker for Apod function. Although we did not detect the association of Importin7 with ERK1/2 in the cytoplasm of B16F10 cells with GFP expression and the increase of NKG2D-L (figure 5c,d), GFP-Apod expression showed the association of Importin7 with ERK1/2 and the increase of NKG2D-L (figure 5e,f), suggesting that the Importin7 and ERK1/2 complex is retained in the cytoplasm in the Apod-expressing setting. We next examined whether the decrease of nuclear ERK1/2 is involved in the increase of NKG2D-L. PD98059, an MEK inhibitor, decreased ERK1/2 in the nucleus and increased in NKG2D-L (p<0.0001, figure 5g,h). These data suggest that an impairment of ERK signaling is one of the triggers to increase NKG2D-L. However, we did not detect any increase in NKG2D-L and NKG2D in splenic T cells by Apod and HVJ-E administration at the concentration increasing NKG2D-L in B16F10 cells in an in vitro setting (online supplemental figure S7e,f), suggesting that an increase of NKG2D in T cells in vivo is as a response to the increase in NKG2D-L in cancer cells. We found that ERK1, encoded in Mapk3, is not involved in the regulation of NKG2D-L (figure 5i,j, and online supplemental figure S7g). EGFP-Mapk1 restored the Apod-mediated increase of NKG2D-L suffered from Mapk1 KO (figure 5k,l). These results suggest that Apod increases NKG2D-L expression through preventing the nuclear translocation of ERK2.

Figure 5. Apod increases NKG2D-L expression through preventing the nuclear translocation of ERK2. (a) Western blotting of proteins precipitated with Apod proteins using the indicated antibodies. (b) Western blotting of proteins fractionated to nuclear and cytoplasm in Apod-treated B16F10 cells using the indicated antibodies. The numbers indicate the ratio of the quantity of protein in the nuclear to that in the cytoplasm. (c) Western blotting of proteins precipitated with anti-ERK1/2 antibody in Tet GFP B16F10 cells treated with Dox− or Dox+. (d) Dot plot showing NKG2D-L median expression in Tet GFP B16F10 cells treated with Dox- or Dox+. n=3. (e) Western blotting of proteins precipitated with anti-ERK1/2 antibody in Tet GFP-Apod B16F10 cells treated with Dox− or Dox+. (f) Dot plot showing NKG2D-L median expression in Tet GFP-Apod B16F10 cells treated with Dox− or Dox+. n=3. (g) Western blotting of proteins fractionated to nuclear and cytoplasm in PD98059-treated B16F10 cells using the indicated antibodies. The numbers indicate the ratio of the quantity of protein in the nuclear to that in the cytoplasm. (h) Dot plot showing NKG2D-L median expression in 100 nM PD98059-treated or DMSO-treated B16F10 cells. n=3. (i) Western blotting of ERK1/2 in WT and Mapk3 KO B16F10 cells. H2B is a loading control. (j) Dot plot showing NKG2D-L median expression in PBS or Apod-treated Mapk3 KO B16F10 cells. n=3. (k) Western blotting of ERK1/2 in WT, Mapk1 KO, Mapk1 KO+EGFP, and Mapk1 KO+EGFP-Mapk1 B16F10 cells. H2B is a loading control. (l) Dot plot showing fold increases in NKG2D-L in Apod-treated B16F10 cells, compared with PBS-treated cells. n=3. P values were calculated using the Tukey HSD (l) and Student’s or Welch’s t-test (d, f, h, j), according to the normality and variance of the data. Apod, apolipoprotein d; GFP, Green fluorescent protein; HSD, honestly significant difference; HVJ, hemagglutinating virus of Japan; KO, knockout; NKG2D-L, NKG2D-ligands; PBS, phosphate buffered saline; WT, wild-type.

Apod + OX40 antibody elicits abscopal antitumor effects by activating T cells

Combining HVJ-E with OX40 antibody activates CD4 and CD8 T cells in both target and non-target tumors, inducing systemic antitumor effects.¹⁵ Given that Apod was a causative factor in the HVJ-E-induced antitumor effects and that intratumor expression of Apod significantly suppresses tumor growth, we determined whether the systemic antitumor effects are elicited by a doxycycline-induced expression of Apod in B16F10 cells. However, only the Apod expression did not repress tumor growth at the non-target tumors (online supplemental figure S8a). We thus thought a combination of Apod with an immune-modulating antibody induces systemic antitumor effects since Apod induced NKG2D-L expression in cancer cells. Combining OX40 antibody with toll-like receptor 9 (TLR9) stimulation activates antitumor immunity,23,26 suggesting that the combination of Apod with OX40 antibody activates antitumor immunity since Apod plays critical roles in HVJ-E-induced antitumor effects. Injection of the OX40 antibody into the Apod-expressing B16F10 tumor remarkably suppressed target tumor (p=0.0005 and p=0.0234, Dox−/Ctrl vs Dox+/OX40 and Dox−/OX40 vs Dox+/OX40, respectively) and wild-type B16F10 non-target tumor (p<0.047, Dox−/Ctrl, Dox−/OX40, and Dox+/Ctrl vs Dox+/OX40; online supplemental figure S8a) growth. Consistent with this, Apod expression and OX40 antibody increased CD4 and CD8 T cells at the target and non-target lesions (online supplemental figure S8b). These data indicate that these two factors can induce systemic antitumor effects.

To determine whether extracellular Apod can induce systemic antitumor effects, we intratumorally injected the recombinant Apod protein with OX40 antibody in bilaterally B16F10 cells-inoculated mice. Target tumor (p<0.023, PBS/Ctrl, PBS/OX40, and Apod/Ctrl vs Apod/OX40) and non-target tumor (p=0.0036, Apod/Ctrl vs Apod/OX40; p=0.0346, PBS/Ctrl vs Apod/OX40; figure 6a) growth were significantly suppressed and CD4 and CD8 T cells were increased at the target and non-target lesions (figure 6b). We found a significant difference between Apod/Ctrl and PBS/Ctrl in target lesion on day 10 (p=0.0004) but not on day 14, because mice bearing two tumors showed slightly faster tumor growth and reduced therapeutic efficacy, compared with mice with one tumor. The combination therapy significantly induced CD94 and NKG2D expression compared with control groups (figure 6c), suggesting an increase in the association between cancer cells and T cells. Moreover, the activation status of CD4 and CD8 T cells was confirmed at the target and non-target lesions by IFN-γ expression (figure 6d, online supplemental figure S8c). These results indicate that alongside OX40 stimulation, Apod can increase T cells at the target and non-target lesions.

Figure 6. Apod+OX40 agonist antibody elicits abscopal antitumor effects by activating T cells. (a) Tumor growth curve of B16F10 cells in C57BL/6N mice. Apod (40 µg) was intratumorally injected with 10 µg anti-OX40 agonist or control antibody. The number in parentheses indicates the number of samples. Error bars show the SD. (b) Dot plot of the number of CD45/CD3/CD4 and CD45/CD3/CD8 T cells. PBS/Ctrl IgG, n=9; PBS/OX40 n=8; Apod/Ctrl IgG, n=10; Apod/OX40, n=9. (c) Percentage of CD94 and NKG2D expression in CD45/CD3/CD4 and CD45/CD3/CD8 T cells. Tumors were analyzed 14 days after treatment initiation. PBS/Ctrl IgG, n=9; PBS/OX40 n=8; Apod/Ctrl IgG, n=10; Apod/OX40, n=9. (d) Percentage of IFN-γ expression in CD45/CD3/CD4 and CD45/CD3/CD8 T cells after PMA ionomycin treatment. PBS/Ctrl IgG, n=8; PBS/OX40 n=8; Apod/Ctrl IgG, n=8; Apod/OX40, n=7. P values were calculated using the Tukey HSD (a), Steel-Dwass test (b), and Wilcoxon test (d), according to the normality and variance of the data. Apod, apolipoprotein d; HSD, honestly significant difference; IFN, interferon; PBS, phosphate buffered saline; PMA, phorbol myristate acetate; WT, wild-type.

Discussion

The antitumor effects of virotherapy depend on viruses directly killing cancer cells and an adjuvant effect that induces antitumor immunity. Our work suggests that HVJ-E represses tumor growth by activating the Irf7-Apod axis. Irf7 is a multifunctional transcription factor in several types of cells, including immune cells²⁷ and cancer cells.²⁸ Irf7 has a role in the maturation of DCs²⁹ and the activation of NK cells through IFN-β.³⁰ Our findings indicate that the induction of Irf7 expression in cancer cells represses tumor growth in vivo through increasing Apod expression. Apod is pivotal for response to HVJ-E infection since inducible Apod expression suppressed tumor growth in vivo and deficiency of Apod in B16F10 cells attenuated the HVJ-E-induced changes of gene expression profile, including NKG2D-L and MHC class I/II. These results suggest that Apod has a role in protecting the body from virus infections, such as HVJ. A previous study reports that amphipathic apolipoproteins, including Apob and Apoe, are crucial to the formation of hepatitis C virus particles with infectivity.²⁷ Although Apod is a member of apolipoproteins, its structure differs from amphipathic apolipoproteins. This distinction suggests unique roles for Apod, including its involvement in exerting antitumor effects, as revealed in this study. Other replicative oncolytic viruses might have a distinct mechanism to induce antitumor effects from HVJ-E since it is a non-replicative oncolytic virus. Our findings suggest that inducing Apod expression in tumors causes antitumor effects.

Inducing APOD expression may become a surrogate drug for virotherapy against various tumor types, including HVJ-E-resistant tumors. The primary role of apolipoproteins is the transportation of lipids in blood vessels. However, the function of APOD is not limited to lipid transport.³¹ APOD acts as a cardioprotective factor in myocardial infarction³² and represses the production of pro-inflammatory cytokines, probably by altering arachidonic acid level.³³ In Drosophila melanogaster, human APOD expression increases stress resistance and extends lifespan by reducing age-associated lipid peroxidation.³⁴ APOD ectopic expression in esophageal squamous cell carcinoma slightly modulates colony formation efficiency.¹⁹ We found the dose dependency of Apod to induce antitumor effects, suggesting that the concentration of Apod may be too low to cause the effects at the homeostatically expressed amount. Taken together, Apod is a pivotal factor in maintaining homeostasis.

Notably, we identified that ERK2 is a critical target of Apod to repress tumor growth and to increase NKG2D-L expression through preventing nuclear translocation of ERK2 in B16F10 cells using CRISPR screening and in situ biotinylation. These results are consistent with previous reports that APOD is involved in regulating the nuclear translocation of ERK1/2 in vascular smooth muscle cells³⁵ and that the ERK signaling pathway modulates NKG2D-L expression.³⁶ Blocking the interaction of Importin7 with ERK1/2 induces apoptosis in several cancer cells but not normal cells.³⁷ However, we found that ERK2 KO is not lethal in B16F10 cells and that ERK2 is critical for the Apod-mediating repression of tumor growth and modulation of NKG2D-L expression, suggesting that an impairment of ERK2 signaling is one of the triggers for the Apod-mediating antitumor effects. We found a possibility that Apod affects several proteins since in situ biotinylation of GFP-Apod detects close proximity of Apod to proteins belonging to cellular signaling GOs, including Signaling by MET, Signaling to ERKs, non-membrane-bounded organelle assembly, and ribonucleoprotein complex biogenesis (figure 4c). Moreover, Apod was associated with Importin7. Importin7 controls the nuclear translocation of several proteins tuning cellular responses, such as YAP and Smad3 other than ERK1/2.³⁸ These results suggest that Apod may also modulate the nuclear translocation of several proteins other than ERK1/2. Although our results suggest that Apod increases NKG2D-L through preventing nuclear translocation of ERK2, further studies are required to understand the complete Apod-induced signaling pathway for the antitumor effects.

The systemic antitumor effect is activated by the combination therapy with a TLR9 ligand, including unmethylated CG-enriched oligodeoxynucleotide, and OX40 antibody.²³ HVJ-E induces the expression of IFN-β and IFN-γ independently of the TLR signaling pathway.³⁹ Surprisingly, the combination of Apod, the primary component in the antitumor effect of HVJ-E, with OX40 stimulation activated T cells at the non-target lesion. The result is consistent with systemic antitumor effects by combining HVJ-E with OX40 antibody.¹⁵ These results suggest that two stimuli are necessary to activate systemic antitumor effects. One is the promotion of association between cancer cells and T cells by TCR-MHC and NKG2D-NKG2D-L axes. The other is the T-cell co-stimulation signal that activates T cells, including OX40. The requirement of these two stimuli for promoting systemic antitumor effects may be one reason why the abscopal-like effect is rarely seen in patients with cancer and why the efficacy of ICI differs from one case to another.

Although our findings revealed that the Apod+OX40 antibody induces T-cell activation, additional studies are required to elucidate how it alters the status of tumor microenvironment-forming cells. Our findings indicate that the combination of endogenous protein with T-cell co-stimulation leads to the activation of antitumor immunity without the need for adjuvants, neoantigens, or T-cell modification. Nevertheless, it is essential to demonstrate the efficacy of the combination of Apod with OX40 stimulation therapy in patients with malignant tumors. Apod expression induced in cancer cells showed substantial antitumor effects, compared with the intratumor injection of Apod protein, suggesting a compound and messenger RNA (mRNA)-inducing Apod expression may be a candidate for an anticancer drug. Overall, the study provides important insights into the mechanisms of how HVJ-E induces antitumor effects and the triggers to activate systemic antitumor immunity and forms a basis for the development of novel cancer therapies.

Methods

Cell lines and cell culture

Mouse B16F10 melanoma cells (CRL-6475), LL/2 Lewis lung carcinoma cells (CRL-1642), 4T1 mammary carcinoma cells (CRL-2539), CT26 colon carcinoma cells (CRL-2638), human MeWo melanoma cells (HTB-65) were purchased from the American Type Culture Collection. Mouse MC38 colon carcinoma cells (ENH204) were purchased from Kerafast. B16F10, LL/2, MC38, MeWo cells, and their derivatives were cultured in Dulbecco's Modified Eagle Medium (DMEM) medium (08458–45, Nacalai Tesque) containing 10% fetal bovine serum (FBS) (172012, Sigma), 100 U/mL penicillin, and 100 µg/mL streptomycin (26253–84, Nacalai Tesque). 4T1 and CT26 were cultured in Roswell Park Memorial Institute (RPMI) 1640 (30264-56, Nacalai Tesque) medium containing 10% FBS, 100 U/mL penicillin, and 100 µg/mL streptomycin. Genistein (NPI 031) was purchased from Selleck. The cells were cultured at 37°C in a humidified atmosphere of 95% and 5% CO₂.

Plasmids

Irf7, Batf2, and Apod complementary DNA (cDNA) were amplified from B16F10-derived cDNA using KOD FX Neo (KFX-201, TOYOBO) and the SuperScript III First-Strand Synthesis System for RT-PCR (18080–061, Thermo Fisher Scientific). The PCR product was introduced into the pENTR site of the following vector. Human APOD and GFP cDNA were introduced into pENTR. pMA-Irf7 Arm-3 X TY1-T2A-mClover3-T2A-mClover3-T2A-mClover3-P2A-DTR-PGK promoter-NeoR/BSD-pA was constructed to generate Irf7-TY1-Clover B16F10 cells using artificially synthesized DNA and restriction enzymes. pMA-ROSA-pTetOne-EGFP-gw-P2A-BSD and pMA-ROSA-pTetOne-gw-P2A-BSD were constructed using the synthesized DNA and restriction enzymes. Targeting vectors with 250 bp 5’ and 3’ targeting arms for Irf7 and Batf2 were generated using artificially synthesized DNA and integrated with the PGK promoter-NeoR-pA fragment from pGolden-Neo (#51422, Addgene) or PGK promoter-BSD-pA PCR fragment from PGKgb2-BSD with HindIII/NotI digestion. gRNA sequences were introduced into pX330 (#42230, Addgene). The gRNA sequence is shown in online supplemental table S1.

Establishment of cell lines

pX330 encoding Cas9 and gRNA plasmids, linearized plasmids, and single-strand DNAs were introduced into B16F10 cells using a Neon Electroporation System (Thermo Fisher) with voltages 1,100/width 30/pulse 1. Irf7, Batf2, Ifnar2, Mapk1, and Mapk3 were knocked out using the targeting vector encoding the neomycin-resistant gene. Ifnar1 was knocked out using the two targeting vectors encoding neomycin-resistant and blasticidin S-resistant genes. Ddx58, Ifih1, Dhx58, Apod, Apol9a, and Apol9b were knocked out using SUCCESS.⁴⁰ B16F10 cells were selected in a medium containing 2 mg/mL G418 (165–1294, Nacalai Tesque) and 5 mg/mL blasticidin S (029–18701, Wako). After 5 days of electroporation, ~3000 cells were re-seeded to form single colonies. KO of B16F10 cells was established by eliminating almost the entire coding region using the CRISPR/Cas9 system. The PGK-neomycin resistance and PGK-blasticidin resistance gene cassettes were used to select KO clones. TetOn-GFP, GFP-Irf7, GFP-Batf2, and GFP-Apod constructs were introduced into the ROSA26 locus to avoid unexpected gene silencing. The inducible expression of GFP, GFP-Irf7, and GFP-Apod was confirmed by treatment with 2 µg/mL doxycycline hydrochloride n-hydrate (045–31123, Wako) for 2 days. The TY1 tag and 3×mClover3 construct were knocked in just before the stop codon of the Irf7 gene to fuse the TY1 tag with endogenous Irf7 and to monitor endogenous Irf7 expression by flow cytometry. PCR confirmed recombination between the target genomic sequence and the selection cassette. PCR confirmed the deletion of the target coding region. The primers are listed in online supplemental table S1.

HVJ-E production

HVJ (VR-105 parainfluenza1 Sendai/52, Z strain)-E was produced using virus-free chicken eggs and inactivated by UV irradiation, as previously described.¹²

Tumor mouse model

The Osaka University Animal Experiments Committee approved all mouse experiments, which were performed in accordance with the guidelines. The experimental endpoint of tumor growth was defined as a tumor diameter of 2 cm in any direction, measured using a digital caliper. All cancer cells were confirmed to be negative for Mycoplasma contamination by PCR (6601, TaKaRa). 0.5×10⁶ B16F10, MC38, or LL/2 cancer cells in 50 µL PBS were intradermally injected into 6–8 weeks old female C57BL/6N mice or 0.5×10⁶ 4T1 and CT26 cells were intradermally injected into 6–8 weeks old female BALB/cA mice. Mice were maintained under specific pathogen-free conditions. C57BL/6N, NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG), and NOD.CB17-Prkdc^scid/J (NOD-SCID) mice were purchased from Charles River. CB17/IcrJcl-Prkdc^scid (SCID) and BALB/cAJcl mice were purchased from CLEA Japan. Transgenic GFP-expressing C57BL/6N mice were kindly provided by Dr Masaru Okabe. HVJ-E treatment was started when the inoculated tumors reached a diameter of 4.5–5.5 mm, after 4–6 days of inoculation. 2,000 HAU HVJ-E in 50 µL of PBS was intratumorally injected three times every other day. To induce gene expression driven by the tet-inducible promoter in B16F10 cells, 200 µL of 2 mg/mL doxycycline in ultrapure water was orally administered to mice at the indicated time points. HVJ-E/PBS treatment was started on the day when the tumor size reached 8 mm to collect the tumors 1 day after three doses of administration. B16F10 cells with tetracycline-promoter cassette tumors were collected after 1 day from two doses of doxycycline. Treatment was started on the day when the tumor size reached 8 mm. Tumor size was measured every other day. Tumor volumes were calculated according to the following formula: tumor volume (mm³)=length×(width)²/2. GFP mice were used to isolate cancer cells. The GFP-negative population was isolated using an FACSAria II (BD) or FACSAria IIIu flow cytometer (BD). GFP expression in cells from GFP mice (~95%) was confirmed by flow cytometry.

Cancer cells (0.5×10⁶) were bilaterally inoculated in the flanks of 6–8 weeks old female C57BL/6N mice for B16F10. Next, 40 µg Apod protein and 10 µg of OX40 agonist antibody (Ultra-LEAF Purified anti-mouse CD134 (OX-40) antibody, OX86, 119431; BioLegend) or control antibody (IgG from Rat Serum, I4131; Sigma-Aldrich) were intratumorally injected into the left side tumor three times every other day. Doxycycline (200 µL, 2 mg/mL in ultrapure water) was orally administered to mice to activate the tetracycline promoter in B16F10 cells.

Patient-derived xenograft model

The collected melanomas were cut into 5 mm cubes. Two or three cubes were subcutaneously inoculated into NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice (Charles River). Then, patient-derived tumors were transferred into NSG mice. Then, two to three 5 mm cubes of the grown patient-derived tumors were subcutaneously inoculated into NOD.CB17-Prkdc^scid/J (NOD-SCID) mice, which have NK cell activity (Charles River). Apod treatment was initiated when the transferred tumors started proliferating. Apod was injected intratumorally every other day. Tumor sizes were measured every 4 days.

Detection of HVJ-E RNA

B16F10 cells (0.5×10⁶) were inoculated in C57BL/6N and NSG mice. 2,000 HAU HVJ-E was intratumorally injected 5 days after inoculation. Tumors were collected in 7 or 14 days after the HVJ-E treatment. Neutrophils were obtained from the spleen using the EasySep Mouse Neutrophil Enrichment Kit (#19762, STEMCELL). RNA was collected from 1×10⁵ tumor cells using ISOGEN (NIPPONGENE). 1×10⁵ neutrophils in 700 µL RPMI with 10% FBS were incubated with 500 HAU HVJ-E in a 12-well plate for 1 day. After washing the cells with PBS, RNA was collected using ISOGEN (NIPPONGENE). cDNA was generated using SuperScript III (18080400, Thermo Fisher). Semi-quantitative Polymerase Chain Reaction (qPCR) was performed using Thunderbird (QPS-101, TOYOBO) with HVJ_10 769F (GGCATAGAAGGTTACTGCCAGAA), HVJ_10 897R (TGTCACGGCTATAGCTTGATTGTC), 18S_rRNA_F (TCAAGAACGAAAGTCGGAGGTT), and 18S_rRNA_R (GGACATCTAAGGGCATCACAG).

Flow cytometry analysis and cell sorting

The collected tumor tissues were finely cut with scissors and incubated in 2 mL 0.5% collagenase and 2% FBS in PBS pre-warmed at 37°C. The finely cut tissues were incubated for 45–60 min, with pipetting every 15 min at 37°C. The dispersed cells were diluted with 8 mL 2% FBS/PBS and then passed through a 70 µm cell strainer (352350, Falcon). The cell strainer was washed with 10 mL 2% FBS/PBS. After centrifuging the cells at 1,500 rpm at 4°C for 5 min, the cell pellets were treated with 1–2 mL hemolysis buffer (0.17 M NH₄Cl, 0.01 M KHCO₃, 0.082 mM EDTA, pH 7.3) for just 5 min with gentle shaking. The hemolysis reaction was stopped by adding a 10-fold volume of 2% FBS/PBS. The hemolysis-treated cells were filtered using a 40 µm cell strainer (352340, Falcon).

To isolate murine tumor cells, tumor tissues were collected from the cancer cell-bearing GFP mice. GFP-negative tumor cells were isolated using a flow cytometer. All samples were filtered through a Cell-Strainer Cap (352235, FALCON) and analyzed using an FACSCanto II (BD). FACSAria II (BD) and FACSAria IIIu (BD) instruments were used for sorting cells without fixation.

Extracellular proteins were stained with the indicated fluorescent-labeled antibodies in 50 µL of 2% FBS in PBS for 30 min on ice, followed by washing two times with 2% FBS in PBS and centrifuging at 600×g and 4°C for 3 min. The stained cells were resuspended in 1% paraformaldehyde (PFA) in PBS. We used the following antibodies against murine or human proteins and recombinant proteins: anti-CD45 (BioLegend), anti-CD3 (BioLegend), anti-CD4 (BioLegend), anti-CD8 (BioLegend), anti-OX-40 (BioLegend), anti-PD-1 (BioLegend), anti-CD94 (BioLegend), anti-NKG2D (BioLegend), NKG2D recombinant protein Fc-chimera (R&D Systems), and anti-human IgG Fc (BioLegend). Further information on the antibodies is listed in online supplemental table S1. For staining intracellular proteins, the cells that were stained with extracellular proteins and fixed in 4% PFA in PBS were resuspended in 1× permeabilization buffer (2106783; Invitrogen), followed by centrifugation at 600×g and 4°C for 3 min. Intracellular proteins were stained with the indicated fluorescent-labeled antibodies in 50 µL of 1× permeabilization buffer for 30 min, followed by washing once with 1× permeabilization buffer and centrifuging at 600×g and 4°C for 3 min. We used the following antibodies against murine proteins: anti-Foxp3 (BioLegend), anti-granzyme A (BioLegend), anti-granzyme B (BioLegend), anti-IFN-γ (BioLegend), and anti-Ki67 (BioLegend); further information is shown in online supplemental table S1. For cleaved-PARP staining, the cells were fixed with 70% ice-cold ethanol at 4°C for 30 min, following centrifugation with 600×g at 4°C for 3 min. Cleaved-PARP was stained by anti-cleaved PARP (BD Bioscience) in 50 µL 2% FBS in PBS on ice for 30 min, followed by washing two times with 2% FBS in PBS and centrifuging with 600×g at 4 °C for 3 min. To analyze T-cell cytotoxicity, proliferation, and production of cytokines, a single-cell suspension of tumor tissues was stimulated for 6 hours with 20 ng/mL phorbol myristate acetate (PMA) (162–23591; FUJIFILM) and 2 µg/mL ionomycin (095–05831; FUJIFILM) in the presence of 20 µg/mL (+)-Brefeldin A (022–15991; FUJIFILM) as the GolgiPlug. Stimulated cells were stained as described above. Splenic T cells were collected using the EasySep Mouse T Cell Isolation Kit (19851A, STEMCELL Technologies). All samples were filtered with a Cell-Strainer Cap (352235; FALCON) and analyzed using FACSCanto II (BD Biosciences) and Attune NxT Acoustic Focusing Cytometer (Thermo Fisher). FACSAria II and FACSAria IIIu (BD Biosciences) were used for sorting cells stained as described above without fixation.

Flow cytometry data were analyzed using FlowJo V.10.10.0 (BD Biosciences). The median value of NKG2D-L detected with NKG2D recombinant protein was calculated using the Statistics function of FlowJo. The fold increase in NKG2D-L was calculated by dividing the median value of the Apod group by the average median value of the PBS group.

RNA sequencing

Treatment was started on the day the tumor size reached 8 mm in GFP mice for collecting tumor cells. After 1 day from three doses of HVJ-E or two doses of doxycycline every 2 days, tumors were collected. RNA was extracted using ISOGEN (NIPPONGENE) from GFP-negative tumor cells in B16F10-inoculated GFP mice, sorted by FACSAria II (BD) and FACSAria IIIu (BD). Sequencing libraries from at least two biological replicate RNA samples were generated using the NEBNext Poly(A) mRNA Magnetic Isolation Module (#E7490, New England BioLabs) and NEBNext Ultra RNA Library Prep Kits for Illumina (#E7530, NEW ENGLAND BioLabs) as previously described.⁴¹ Sequencing libraries were analyzed using a HiSeq X instrument (Illumina).

CRISPR screening

CRISPR screening was performed as previously described.^{42 43} Briefly, B16F10-Cas9 cells were established with lentivirus encoding Cas9 (lentiCas9-Blast, #52962, Addgene). The Cas9 activity was evaluated using a lentivirus encoding GFP and BFP (pKLV2-U6gRNA5(gGFP)-PGKBFP2AGFP-W, #67980, addgene). The lentivirus encoding genome-wide gRNAs were transduced to the B16F10-Cas9 cells at a multiplicity of infection of 0.3 and the transduced cells were selected with 0.75 µg/mL puromycin. 3×10⁵ B16F10 cells with the library of gRNAs and 2×10⁵ B16F10 cells with Tet-GFP or Tet-GFP-ApoD were intradermally injected into 6 weeks old female C57BL/6N mice. Doxycycline (200 µL, 2 mg/mL in ultrapure water) was orally administered to mice to activate the tetracycline promoter in B16F10 cells. Tumors were harvested 10 days after inoculation of B16F10 cells since the cellular composition of B16F10 cells can be maintained within this period.⁴⁴ The gRNA sequence was amplified from the genomic DNA collected from each tumor using Q5 Hot Start High-Fidelity 2× Master Mix (M0494, NEB) and indexed using NEBNext Q5 Hot Start HiFi PCR Master Mix (#0543, NEB). The sequencing libraries were analyzed using the HiSeq X instrument (Illumina).

Protein expression and purification

Expi293 cells (A14527, Thermo Fisher) were cultured in Expi293 Expression Medium (A1435101, Thermo Fisher) at 37°C and 8% CO₂ at 120 rpm on a shaker (2-1987-01, AS ONE) in an incubator (SCA-165D, ASTEC). When viable cell density and viability reached 3.5–4.5×10⁶ cells/mL and ≥95% confluence, 150 µg plasmid encoding mouse Apod or human APOD with 6×His Tag at the C-terminus (CAGIP-mApod or hAPOD-6×His) was transfected into 2×10⁸ Expi293 Cells suspended in 8 mL Expi293 Expression Medium with 0.1% PLURONIC F-68 (24040–032, Gibco) using 500 µL polyethylenimine (PEI MAX-transfection grade linear polyethylenimine hydrochloride (MW 40,000), Polysciences) transfection reagent and cultured for 3 hours at 37°C, 8% CO₂, 120 rpm. After 3 hours, 100 mL Expi293 Expression Medium with 840 µL 0.5 M valproic acid was added to the transfected cell suspension and cultured for 7 days in a 500 mL Erlenmeyer flask (431145, Corning) at 37°C, 8% CO₂, 120 rpm. After 7 days of transfection, the floating cells were centrifuged at 1,000 rpm at 4°C for 5 min. After collecting the supernatant, it was centrifuged again at 10,000 rpm at 4°C for 5 min. The supernatant was filtered through a 0.22 µm low protein-binding Durapore (PVDF) membrane (SLGVR33RS, Millex). The recombinant proteins were purified on a HiTrap DEAE FF column (17505501, Cytiva) using a NaCl gradient consisting of 20 mM Tris-HCl pH 8.0 (Strat Buffer), 20 mM Tris-HCl, 0.5 M NaCl pH 8.0 (Elution buffer). Then the proteins were further purified using a HisTrap HP column (17524701, Cytiva) with an imidazole gradient consisting of 20 mM sodium phosphate (pH 7.4), 0.5 M NaCl, 20 mM imidazole (Start Buffer), and 20 mM sodium phosphate (pH 7.4), NaCl (0.5 M), and imidazole (500 mM) (Elution Buffer), using an AKTA FPLC system. The eluted proteins were purified with PBS using a HiTrap Desalting column (17140801, Cytiva) and Acrodisc Mustang E (365–06161, Pall Corporation) to remove lipopolysaccharide (LPS). The purity of the recombinant proteins was confirmed using sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). LPS negativity (<0.25 EU/mL) was confirmed using a Limulus Color KY Test Wako (291–53101, FUJIFILM) and a Toxinometer ET-6000 (FUJIFILM). The quantity of the proteins was measured using a DC Protein Assay (5000111JA, BIO-RAD). Bovine serum albumin (23209, Thermo Fisher) was used to generate a standard curve.

Western blotting

Western blotting was performed as previously described.⁴¹ The antibodies used for western blotting are shown in supplemental data. Cell fractionation was performed as previously described.⁴⁵ 30 µg Apod conjugated to 30 µL protein A-agarose (sc-2001, Santa Cruz) with anti-Apod antibody (sc-166612, Santa Cruz) was used for precipitation of Apod-associated proteins to 1 mL cytoplasm fraction from 2×10⁷ cells after preclearing with IgG-agarose (PM035-8, MBL). Co-immunoprecipitation was performed using 30 µL protein A-agarose (sc-2001, Santa Cruz) with 6 µL anti-ERK1/2 antibody (ab184699, abcam) or control IgG (ab46540, abcam) to 500 µL cytoplasm fraction from 2.5×10⁶ cells after preclearing with IgG-agarose (PM035-8, MBL). Western blotting for the immunoprecipitated samples was performed using EasyBlot (GTX225826-01, GeneTex).

ELISA assay

Tumors were isolated from B16F10 tumor-bearing mice 5 days after inoculation. For the GolgiStop setting, 50 µL of 2 mg/mL (+)-Brefeldin A (022–15991, FUJIFILM WAKO) was injected intratumorally 1 hour before tumor harvest, 23 hours after HVJ-E treatment. Tumor tissues were dissolved in 100 µL RIPA buffer (50 mM Tris‐HCl pH7.6, 150 mM NaCl, 1 w/v% Nonidet P40, 0.5 w/v% sodium deoxycholate, 0.1 w/v% SDS, complete EDTA-free protease inhibitor cocktail (11873580001, Roche)). Samples were centrifuged at 4°C, 15,000 rpm for 10 min, and the supernatant was collected. ELISA assay was performed using the Mouse Apolipoprotein D (APOD) ELISA kit (CSB-EL001935MO, CUSABIO TECHNOLOGY LLC). The result of OD 450 nm was measured by Mithras LB 940 Multimode Microplate Reader (BERTHOLD TECHNOLOGIES GmbH & Co KG). The analysis of data was performed with CurveExpert V.1.4.

In situ biotinylation of Apod proximity followed by mass spectrometry

In situ biotinylation of Apod proximity using anti-GFP (sc-9996, Santa Cruz Biotechnology) antibody was performed with Dox-inducible GFP-Apod expressing mouse B16F10 cells. B16F10 cells in a 60 mm culture dish were fixed with 4% PFA and permeabilized with 0.5% Triton X-100-containing PBS. The cells were blocked with Bovine Serum Albumin (BSA) and horse serum, then incubated with the anti-GFP antibody for 1 hour at room temperature. The cells were washed with 0.5% Triton X-100-containing PBS and incubated with HRP-conjugated anti-mouse IgG (#7076, Cell Signaling Technology). After washing, a biotinylation reaction was performed by incubating the cells with 200 µM biotin-tyramide and 0.0015% H₂O₂ in PBS for 1 min. The biotinylation reaction was stopped by washing cells with PBS immediately. The cells were harvested by scraping in Lysis buffer (150 mM NaCl, 1% Triton X-100, 0.5% Sodium deoxycholate (Doc), 1% SDS, 50 mM Tris-HCl (pH 8.0), and protease inhibitor (#03969, Nacalai Tesque)), and sheared by sonication using Bioruptor Sonicator (Diagenode). The sheared lysate was boiled for 20 min at 95°C, debris was removed by centrifugation, and then the lysate was diluted into 0.5% SDS concentration, followed by avidin pull-down using SoftLink Soft Release Avidin Resin (V2011, Promega). The avidin resin was stringently washed with 0.5% SDS RIPA buffer (150 mM NaCl, 1% Triton X-100, 0.5% Doc, 0.5% SDS, and 50 mM Tris-HCl (pH 8.0)), 0.5 M NaCl RIPA buffer (0.5 M NaCl, 1% Triton X-100, 0.5% Doc, 0.1% SDS and 50 mM Tris-HCl (pH 8.0)), 1.2 M NaCl RIPA buffer (1.2 M NaCl, 1% Triton X-100, 0.5% Doc, 0.1% SDS and 50 mM Tris-HCl (pH 8.0)), and 0.5% SDS RIPA buffer again.

For peptide sample preparation for mass spectrometry, the avidin resin was rinsed with distilled water three times and resuspended with 2 M urea in 50 mM NH₄HCO₃. After incubating the resin with 100 mM DTT for 30 min at 37°C and then with 250 mM IAA for 15 min at room temperature, add 0.5 µg Trypsin in 50 mM NH₄HCO₃ and incubate overnight at 37°C for peptide digestion. The trypsin-digested peptides were purified using C18 and SDB filters.

Proteins were identified by Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS) analysis using Orbitrap Elite, Hybrid Ion Trap-Orbitrap Mass Spectrometer (Thermo Fisher) coupled with an UltiMate 3000 HPLC system (Thermo Fisher). Peptide samples were loaded on a trap column (100 µm × 20 mm, C18, 5 µm, 100 Å, Thermo Fisher) and separated on a Nano HPLC capillary column (75 µm × 18 cm, C18, 3 µm, Nikkyo Technos) at a flow rate of 300 nL/min. Solvent A was 0.1% formic acid in 2% acetonitrile, while solvent B was 0.1% formic acid in 95% acetonitrile. The peptide samples were eluted using a gradient beginning with 2% B for 0–5 min, then 2–33% B for 5–120 min, followed by 90% B for 10 min, and finally equilibration with 2% B for 20 min. The data were acquired using a survey scan performed in a mass range from 380 to 1,500 m/z. The top 15 peaks were selected for fragmentation. Peptide identification, protein identification, and label-free quantification were performed using MaxQuant software (V.1.6.7.0) against mouse protein sequences in the UniProt Knowledgebase (UniProtKB/SwissProt). Search parameters were as follows: enzyme, Trypsin; variable modifications, oxidation (M), Acetyl (Protein N-term); fixed modifications, carbamidomethyl (C). A false discovery rate of less than 5% was adopted as the acceptance criteria for identifications.

Histopathological analysis

For immunostaining of ERK1/2 in B16F10 tumors, B16F10 tumors were collected 1 day after the final 40 µg Apod intratumor injection of three doses. The tumor tissues were fixed with 4% PFA/PBS, then dehydrated with 15% and 30% sucrose/PBS. The tissues were sectioned at 5 µm in OCT compound. The sectioned tissues on MAS-coated glass slides (TF1006M, MATSUNAMI) were treated with 0.2% Triton X-100/PBS for 10 min at room temperature, TrueBlack Lipofuscin Autofluorescence Quencher (23007, BTI) for 30 s, and 10% goat serum (ab7481, abcam)/PBS for 1 hour. ERK1/2 was stained using anti-ERK1/2 antibody (#4695, CST) at 1:250 in PBS for one night at 4°C. Alexa 488 anti-rabbit IgG (A-11034, Thermo Fisher) was used for detecting the first antibody. DNA was counter-stained with DAPI (19178–91, Nacalai Tesque). FV1200 (OLYMPUS) for figure 4h and LSM880 (ZEISS) for online supplemental figure S7c were used to take images. ImageJ was used for the calculation of ERK signals according to https://theolb.readthedocs.io/en/latest/imaging/measuring-cell-fluorescence-using-imagej.html.

Statistics

Shapiro-Wilk test was used to determine data normality. A two-tailed f-test was used to determine equal variance between two samples. For parametric data, Student’s two-tailed t-test was used to compare two samples. Welch’s t-test was used to compare normally distributed, heteroscedastic data. For non-parametric data, the Wilcoxon rank-sum test was used to compare the two samples. One-way analysis of variance with Tukey’s honestly significant difference (HSD) test was used to compare multiple groups. Steel tests were used to compare multiple non-parametric samples. Error bars indicate the SD. JMP Pro V.13 software was used for the calculations.

Bioinformatics

Softwares used in the study

bedtools V.2.26.0⁴⁶

bowtie2 V.2.2.3⁴⁷

clusterProfiler R package V.3.14.3⁴⁸

DESeq2 V.1.26.0⁴⁹

DOSE V.3.12.0⁵⁰

enrichplot R package V.1.6.1 (https://yulab-smu.top/biomedical-knowledge-mining-book/)

FastQC V.0.11.5 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/)

ggplot2 R package V.3.3.3⁵¹ (http://ggplot2.org)

ggrepel R package V.0.9.1 (https://cran.r-project.org/web/packages/ggrepel/index.html)

gplots R package V.3.1.1 (https://cran.r-project.org/web/packages/gplots/index.html)

IGV V.2.3.91^{52 53}

IGVtools V.2.3.91⁵³

Mageck V.0.5.2⁵⁴

org.Mm.eg.db V.3.10.0 (https://bioconductor.org/packages/release/data/annotation/html/org.Mm.eg.db.html)

R V.3.6.3 (https://www.r-project.org)

RStudio V.1.0.44 (https://www.rstudio.com)

RStudio Server V.1.4.1103 (https://www.rstudio.com/products/rstudio/download-server/)

Samtools V.0.1.17⁵⁵

STAR V.2.5.3a⁵⁶

Stringtie V.1.3.4b⁵⁷

RNA sequencing data analysis

Paired-end reads were mapped to the mouse reference genome mm9 or the human reference genome hg19 using the STAR aligner with the default settings after checking the read quality using FastQC. Gene expression levels were merged and calculated using StringTie and DESeq2. Correlation of gene expression between replicates was calculated by R. Sequencing tracks were generated by wig to tdf format conversion by IGVtools with the option -z seven and were rendered by IGV, as previously described.⁴¹ A gene expression heatmap was generated using heatmap.2 included in the gplots R package. The log2 fold change of gene expression was analyzed to calculate Gene Set Enrichment Analysis (GSEA) using clusterProfiler, enrichplot, DOSE, and org.Mm.eg.db. Biological process was used as the Gene Ontology term.

HVJ-E RNA detection

HVJ-derived RNA was detected using the RNA sequencing data. Paired-end reads were mapped to the mouse reference genome mm9 and HVJ complete genome M30202 by STAR using the following options: --outSAMattributes NH HI AS nM NM XS, --twopassMode Basic, --outFilterMatchNmin 3, --outFilterScoreMinOverLread 0.6, --outFilterMatchNminOverLread 0.6, according to the setting for viral track scanning.⁵⁸

Overall survival analysis

TCGA overall survival data were analyzed using cBioportal.⁵⁹ TCGA-SKCM data were filtered to avoid duplication, metastasis, and no prior prognosis, giving 337 samples, since the primary samples were less than metastasis. The patients were divided into two groups according to median gene expression.

CRISPR library analysis

The number of gRNAs in the sequencing libraries was calculated with Mageck as previously described.⁴³

Sex as a biological variable

Our study examined female C57BL/6 and BALB/c and male and female NSG and NOD-SCID mice. We did not consider sex as a biological variable.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.