Host immunodeficiency level does not affect 8B tumour formation, whereas 9G tumour formation is affected
To analyse the requirement for tumour initiation followed by stable growth in immunocompetent individuals, we used the mouse inducible glioblastoma cell lines 8B and 9G [15], which were previously established by overexpression of an oncogenic H-RasL61 in p53-deficient neural stem cells on the C57BL/6 background. As Hide et al. [15] previously described, 8B and 9G lines showed similar growth profiles in vitro (ref. 15 and Fig. 1A). However, these two lines showed a different phenotype when the cells were inoculated into nude mouse brains; 8B cells stably grew in the inoculated mouse brain and eventually killed the mice, but 9G cells did not stably grow in the brain and all mice survived during the observation period [15]. Hide et al. had concluded the difference in tumorigenicity between 8B and 9G was tumour-cell-intrinsic. However, we wondered whether this difference in tumorigenicity of both cell lines was not only due to cell-intrinsic properties but also to active immunity. To address this, we further performed experiments using C57BL/6 mice which are immunocompetent and syngeneic to the 8B and 9G. In the C57BL/6 mice, 85% of the animals did not survive inoculation with 8B cells while after injection with 9G cells only ~40% of the mice died (P = 0.00333, Log-rank test) (Fig. 1B), which are similar data as previously observed using nude mice [15]. Next, we tested tumour-formation of 8B and 9G cells in severe immunodeficient NOD/SCID mice that have a deficiency in T and B cells and reduced macrophage (Mφ) and NK cell activity. Interestingly, all the NOD/SCID mice died not only after 8B tumour cell inoculation but also due to 9G cell tumour formation, albeit at a slower pace as it took nearly double the time for all mice to be dead (NS; P ≥ 0.0083 (i.e., 0.05/6), Bonferroni correction, Log-rank test) (Fig. 1B). These results suggested that the tumour-initiation and following stable growth were regulated at least partly by host immunity. Similar to the observation by Quintana et al. [14], the tumour forming capacity of 8B and 9G cells was significantly influenced by immune deficiency at the host level.
Although CD133 is a well-known marker of glioblastoma-initiating cells [19] [7], its expression was detected neither on 8B nor 9G (Supplementary Fig. 1).
Mφs have overwhelmingly infiltrated the 8B- and 9G-tumour microenvironment
Tumour cells arise from normal cells whose genome has been genetically/epigenetically altered by various factors, such as mutagens or radiation, and are characterised by their ability to multiply indefinitely. Many tumour cells proliferate indefinitely in culture dishes. However, not all such tumour cells form tumours in vivo. As shown in Fig. 1A, 8B and 9G cells grow stably in culture dishes. However, the situation is different in vivo: 9G cells grow stably in immunocompetent individuals, but have difficulty growing in immunocompetent individuals. On the other hand, 8B cells continue to proliferate stably in immunocompromised individuals and even in immunocompetent individuals and eventually kill the individual. These results suggest that whether de novo tumour cells eventually lead to stable tumour formation in an individual may be influenced by immunological factors in that individual. We therefore analysed why potential tumour-initiating cells such as 8B cells stably form tumours in immunocompetent individuals, but not 9G cells, focusing on differences in the interaction of 8B and 9G with immune cells.
To analyse what interaction between potential tumour initiating/forming cells and immune cells in the early period of tumour stable formation, we injected only a thousand 8B or 9G glioblastoma cells orthotopically as a potentially tumour-initiating population into syngeneic immunocompetent animals. We analysed what kind of immune cells infiltrated the tumour area. Immunohistochemical analysis showed a remarkable infiltration of CD11b+, F4/80+ cells, a small population of CD11c+ cells; as well as a small number of CD3+ T cells and several CD19+ cells or Ly6g+ cells were observed in both 8B and 9G tumour tissues (Supplementary Fig. 2). This result in the glioblastoma model suggests that immune cell infiltration into the tumour occurs and interaction is likely.
In these glioma tissues, CD11b+ cells overwhelmingly infiltrated. In the brain, not only macrophages but also microglia are known as CD11b expressing cells [20], so we questioned whether microglia had infiltrated into the glioblastoma tissues. To elucidate this, we stained the tumour tissue with the microglia specific P2RY12 antibody [21] [22] to distinguish microglia from peripherally infiltrating Mφs. Co-staining using P2RY12 and F4/80 antibodies indicated that F4/80+ Mφs, rather than P2RY12+ microglia, selectively infiltrated into the glioblastoma tissue (Supplementary Fig. 3A and 3B). Therefore, we hereafter focused on peripherally infiltrating Mφs rather than microglia for further in vitro experiments.
Mφ depletion prolong survival of 8B inoculated mouse
To define the importance of Mφs in tumour initiation followed by stable growth in immunocompetent mice, we depleted Mφs in vivo. In mouse glioblastomas, the timing of tumour initiation and stable growth itself is difficult to monitor and was therefore assessed by mouse death.
We observed survival/death of 8B-transplanted mice for a long enough period to be able to evaluate the Mφ contribution in tumour-initiation followed by stable growth: in mice orthotopically implanted with 8B, the time to final death was 266 days (N=30) (Fig. 1A). Therefore, day 365 was set as a sufficient period to observe the final survival/death of mice transplanted with 8B, and the contribution of Mφ in tumour development was assessed with the mice alive or dead at 365 days.
Since the F4/80 antibody do not deplete F4/80+ cells in vivo, we instead administered an anti-CD11b antibody (M1/70 proven for CD11b+ cells depletion [23]), as the majority of F4/80+ cells are CD11b positive (data not shown). We further used the RB6-8C5 (anti-Gr-1) antibody [24] which depletes other types of myeloid lineage cells mainly granulocytes, and rat IgG as experimental controls. In the Mφ-depleted group (anti-CD11b group), the ratio of mice alive tended to be higher, although not significantly so, than in the control IgG group (Chi-square test (Pearson), χ2 (1) =4.286, p=0.0384) (The p values <0.05/3=0.0167 (Bonferroni correction) was considered as statistically significant) (Supplementary Fig. 4A and B), suggesting that the role of Mφs was important during the tumour-initiation and following growth.
8B cells made resistant to macrophage proliferation
To analyse whether tumour cells regulate the activity of immune cells, we performed a co-culture experiment of tumour cells and spleen cells. Splenocytes include various kinds of immune cells, including macrophages. In the analysis, Mφs depicted by F4/80+ or CD11b+ efficiently proliferated in the 9G co-culture, although their proliferation was less in the 8B co-culture compared with the 9G co-culture (Fig. 2A). The reduced proliferation was also observed when tumour supernatant was used instead of tumour cells (Data not shown). These results suggest that one or more soluble factor(s) secreted by these tumour cells affect Mφ proliferation.
After 14 days culturing F4/80+ cells with 8B or 9G tumour cell supernatant (8B-Mφ or 9G-Mφ, respectively), the morphology of the resulting Mφs showed clear differences (Fig. 2B). A large proportion of 8B-Mφs showed a flattened and enlarged appearance compared to the 9G-Mφs (Fig. 2B). This difference in cell size was also observed by flow cytometry (Fig. 2C).
We also performed morphological analysis of the Mφs after splenocytes were cultured with the respective tumour supernatants. The Mφs those that did not proliferate (Fig. 2D, gate Ⅰ) and those that did proliferate and expressed Ly6c (Fig. 2D, gate Ⅱ) appeared round-shape, just like the senescent cells in both the 8B-sup and 9G-sup cultures. Interestingly, the CFSE-reduced, i.e. proliferated, Ly6c-negative Mφ populations of 9G-sup cultures (Fig. 2D, gate III) clearly had pseudopodia, suggesting that these Mφ were not in the senescent state. On the other hand, the corresponding Mφ population in 8B-sup cultures showed a round shape (Fig. 2D, gate III), suggesting that they may have entered the senescence-like state after proliferation. These morphological observations indicate that some fractions of 9G-Mφs were in a non-senescence-like state, but most fractions of 8B-Mφs were in a senescence-like state.
8B induced Mφs into a senescence-like state
These observations suggested that the appearance of Mφs induced by 8B tumours was similar to that of cells in a senescence-like state. Therefore, we explored whether the Mφs were in such a state using a senescence-associated β-galactosidase (SA-b-Gal) assay. Flattened and enlarged Mφs showed β-galactosidase activity. The proportion of β-galactosidase-positive Mφs was about 1.6-times larger in 8B-Mφs than in 9G-Mφs (Fig. 2E). Additionally, we performed SA-b-Gal staining in brain tumours that were established after injection of 8B and 9G cells. Positive b-gal activity was observed around the tumour region in the 8B-transplanted brain, whereas no significant positive staining was detected in the 9G-transplanted brain (Fig. 2F). Most of the cells showing SA-β-gal activity (senescent cells) in the 8B tumour were, although not all, F4/80 positive (Fig. 2G). The proportion of the SA-β-Gal positive area within the F4/80+ population was approximately 1.8-fold higher in 8B than in 9G tumours (Fig. 2H). Furthermore, we analysed the series of cellular senescence-related genes expression in Mφs. The analysis revealed that 8B-Mφs highly expressed p21 and Glb1 (Fig. 2I) compared to 9G-Mφs among six tested genes which are known to correlate with cellular senescence [25].
Historically, many classifications of Mφs have been proposed [26], the major being M1/M2 [27]. Tumour-associated Mφs are commonly classified as M2 Mφs [26] [28]. Although 8B cells with tumour-initiating capacity are expected to induce Mφs with an M2 phenotype, 8B-Mφs did neither express most of the M2-Mφs-related genes such as Irf4 or Retnla, except for Arg1 (Fig. 2J), nor did 8B-Mφs M1-related genes. Furthermore, 9G-Mφs do also not express the typical gene profile to fit the M1/M2 classification (Fig. 2J). Thus, these data indicated that 8B- and 9G-Mφs cannot be classified as M1 or M2 based on their gene expression pattern.
8B tumour infiltrating CD3+ T cells downregulate CD3ζ
Despite the infiltration of CD3+ T cells in the 8B-tumour environment in immunocompetent mice (Supplementary Fig. 2), tumour growth was persistent (Fig. 1B), suggesting suppression of T cell activity in the tumour tissue. Arginase-1 is known to be a strong immunosuppressive molecule that downregulates the expression of CD3ζ, a key intracellular molecule in T cell signalling, and inducing hyporesponsiveness of these cells [29]. We observed an expression of Arg1 in 8B-Mφs but not in 9G-Mφs (Fig. 2J). When we analysed CD3ζ expression in T cells from 8B and whole splenocyte co-cultures, including T cells and macrophages, 17A2 antibody staining positive CD3ε/γ/δ+ T cells, i.e. surface CD3 positive T cells lost their CD3ζ expression. In contrast, surface CD3+ T cells retained CD3ζ expression in the 9G co-culture (Fig. 2K). In vivo analysis revealed that CD3ζ expression was observed in ~75% of surface CD3+ T cells in 9G tumours, but only in ~30% of surface CD3+ cells in 8B-tumour tissue (Fig. 2L). To further understand the activity of tumour-specific T cells, we analysed T cells harvested from 8B tumour-immunized C57BL/6 mice. As expected, IFN-γ production was approximately ten-fold lower in CD3ζ-negative T cells compared to CD3ζ-positive T cells (Fig. 2M). Thus, although T cells infiltrated heavily into the 8B tumour, most of these T cells lost their CD3ζ expression and suggested became dysfunctional. It is therefore suggested that 8B tumours induce dysfunction of infiltrating T cells by downregulating their CD3ζ expression and that resulted in the development of tumours in immunocompetent mice.
8B-derived IL-6 induce Mφs into a senescence-like state
Next, we focused on the unknown factor(s) that induced the senescence-like phenotype of Mφs. We analysed the tumour culture supernatant by cytokine/chemokine array and found interleukin-6 (IL-6), Ccl5, Cxcl1, and Cxcl10 were selectively secreted by 8B cells, where Ccl2 was secreted at a 2x higher amounts (Fig. 3A). Notably, some of these factors, such as IL-6 and Cxcl1, are known to be senescence-related cytokines/chemokines [30] [31]. To identify which of these cytokine(s) is/are important for inducing the senescence-like state in Mφs, we cultured splenocytes in the presence of the cytokines secreted by 8B or after withdrawal of each individual cytokine/chemokine from the five candidates (Fig. 3B). Accelerated senescence induction in Mφs was observed in the presence of the five cytokines compared to in their absence. After withdrawal of either Ccl2 or IL-6, the numbers of b-galactosidase-positive Mφs were reduced. Notably, SA-b-Gal-positive Mφs were reduced to the same level as in the negative control by withdrawing IL-6. This result, together with the observation that 9G sup also contained Ccl2, suggests that IL-6 is the responsible factor for inducing Mφs into the senescence-like state.
The p38 MAPK signalling pathway is responsible for IL-6 secretion in 8B
Next, we attempted to identify the signalling pathway that induces selective expression of IL-6 in 8B cells. We used the following inhibitors: U0126 (MEK1/2 inhibitor), LY294002 (PI3K inhibitor), BAY117082 (NFkB inhibitor), SB203580 (p38 inhibitor), SP600125 (JNK inhibitor), and PD0325901 (MEK1/2 inhibitor). Because BAY117082 was strongly cytotoxic for 8B at low concentrations (even at 1 mM), it was difficult to use it in this assay (Supplementary Fig. 5A). Although all inhibitors were relatively more toxic for the cells than vehicle alone, IL-6 secretion was maintained by MEK1/2 inhibition by U0126 and PI3K inhibition by LY294002. In contrast, JNK inhibition by SP600125 and MEK1/2 inhibition by PD0325901strikingly up-regulated IL-6 secretion, whereas p38 inhibition by SB203580 led to decrease of IL-6 secretion (Supplementary Fig. 5B); thus, p38 MAPK signalling is important for IL-6 secretion in 8B cells.
As described above, both the IL-6 producer 8B and the non-producer 9G originate from the same cell line [15]. The cause of their different IL-6 secretion is unclear. Ohsawa et al. [32] reported that constitutive activation of Ras and mitochondrial dysfunction in drosophila cells resulted in upd (an ortholog of human/mouse IL-6) secretion and tumour formation in vivo. As 8B and 9G cells both bear constitutive active Ras, we speculated that a difference in their respective mitochondrial dysfunction results in the different IL-6 production. As mitochondrial dysfunction results in the accumulation of reactive oxygen species (ROS) within cells, we compared ROS accumulation in 8B and 9G cells. In the presence of excess amounts of N-acetylcysteine (NAC) (5000 μM), ROS levels were detected almost equally in 8B and 9G. In contrast, higher ROS levels were recorded in 8B in the presence of 500 μM NAC or in the absence of NAC. (Supplementary Fig. 5C). Considering that elevated ROS levels induce p38 activation [33], activation of the ROS-p38 axis was suggested to be the possible mechanism of selective IL-6 secretion in 8B in relation to 9G.
8B derived IL-6, a senescence inducible factor of Mφs, is one of the factors responsible for tumorigenesis in immunocompetent C57BL/6 mice
Next, we focused on the impact of IL-6 secreted by 8B on senescence induction in Mφs and tumour-forming capacity in vivo. We created 8B-IL-6-knockout cells (8B-IL-6-KO) by CRISPR/Cas9 technology as well as mock transfectants with scrambled guide-RNA transfected 8B cells (8B-mock) (Fig. 3C). Both cell lines proliferated comparably in vitro (Fig. 3D). The flattened and enlarged appearance of Mφs was observed when cells were cultured in 8B-mock supernatant culture, whereas Mφs with pseudopodia were observed in the 8B-IL-6-KO supernatant culture (Fig. 3E). In 8B-IL-6-KO-Mφs, expression of Glb1, p16, p19, p21, and Arg1 was lower than in 8B-mock-Mφs (Fig. 3F). Thus, Mφs induced by the 8B-IL-6-KO supernatant lost their senescent features, suggesting that IL-6 secreted by 8B cells is indeed the factor responsible for senescence induction in Mφs.
We further analysed the impact of IL-6 on tumorigenicity (Fig. 3G). 8B-IL-6-KO cells, as well as 8B-mock cells, exerted tumorigenicity in NOD/SCID mice, and all mice died by 200 days after tumour inoculation. Interestingly, only 8B-mock cells still exerted tumorigenicity in immunocompetent C57BL/6 mice, and almost all the mice died in a similar time course as the NOD/SCID mice. In contrast, 8B-IL-6-KO cells showed remarkable reduced tumorigenicity in C57BL/6 mice as about 60% of the mice survived more than 400 days after the inoculation of 8B-IL-6-KO cells. This indicates that tumour cells that were not secreting IL-6 were not capable to form a tumour in immunocompetent mice, although they could do so in immunodeficient mice. In contrast, IL-6 secretion by tumour cells allowed tumour formation not only in immunodeficient mice but even in immunocompetent mice. Thus, IL-6 secretion by tumour cells defines their capacity of tumour-initiation and following growth in immunocompetent mice in this orthotopic glioblastoma transplantation model.
Senescence-like Mφs induced by 8B express high levels of CD38, and supplementation with its substrate, nicotinamide mononucleotide, suppresses the Mφs go into a senescence-like state
To identify a surface protein that was selectively expressed in senescence-like Mφs, we examined a series of cell surface molecules by flowcytometry. From the experiments, we found that 8B-sup induced senescence-like Mφs expressed CD38 at a higher level compared to 9G-sup induced Mφs (Fig. 4A and Supplementary Fig. 6). In addition, more than 70% of F4/80+ cells expressed CD38 in 8B-tumours. In contrast, CD38 expression was detected only on ~25% of the F4/80+ cells in 9G-tumours (Fig. 4B). CD38 is a membrane-bound multifunctional protein with NADase activity that hydrolyzes NAD+ to nicotinamide and cyclo-ADP-ribose, and its expression is known to increase with age and aging [34]. A recent analysis revealed that CD38 is the main enzyme involved in NAD precursor nicotinamide mononucleotide (NMN) degradation in vivo [35]. As the NAD level is reduced with age and is involved in age-related metabolic decline [35], exogenously supplying NAD may prevent senescence. Therefore, we predicted that adding NMN would cancel the senescence phenotype of Mφs and change the tumour microenvironment into an anti-tumorigenic state. We found that the addition of NMN to the 8B-Mφ culture resulted in Mφs with pseudopodia (Fig. 4C). Furthermore, the number of SA-b-Gal-positive senescence-like Mφs was reduced after addition of NMN (Fig. 4C), and Arginase-1 expression in 8B-Mφs cultured in the presence of NMN showed a three-fold reduction of Arginase-1 expression compared to the control culture (Fig. 4D).
Possible existence of senescent Mφ in human glioblastoma tissue
Several gene expressions in human glioblastoma tissues were analysed using GEPIA, a web server for cancer and normal gene expression profiling and interactive analyses based on TCGA and GTEx data (Supplementary Fig. 7A and B) [18]. High expression of CD11B and CD14, those mainly observed in monocytes and macrophage, was identified in human glioblastoma tissues than in normal brain tissues, suggesting a macrophage infiltration in glioblastoma tissue. IL-6, defined as Mφ senescence inducible cytokine in the mouse glioblastoma model (Fig. 3A-G), and CD38, that detected in senescent Mφs (Fig. 4A), were also detected with higher levels in glioblastoma tissues than in normal tissues. These results suggested a possibility that senescent Mφ were infiltrated in human glioblastoma tissues. Further, pairwise Pearson correlation between CD38 and ARG1 expression in human glioblastoma tissues were analysed using GEPIA. There was a significant correlation between them, suggesting a possibility of senescent Mφ expressing ARG1.
Exogenous supplementation of NMN converts Mφs from an immunosuppressive to an immunogenic state
We analysed whether the addition of NMN contributes to the improvement of immune activating capacity of senescence-like macrophages. The addition of NMN to the 8B-Mφ culture resulted in Mφs with pseudopodia, resembling dendrites of dendritic cells (Fig. 4C). From this result, we speculated NMN treatment might improve the immune activating capacity of 8B-Mφ. Therefore, we further measured the effect of NMN treatment on the immune activating capacity of 8B-Mφs by using an allogeneic T cell response assay as one of immune activating capacity evaluation (Fig. 4E). Although CD8+ T cells proliferated efficiently in an M-CSF induced Mφ co-culture, their proliferation ratio was ~3-fold reduced in saline-treated 8B-Mφ co-cultures, to the same extend as the NMN-treated culture. CD4+ T cells showed a ~2-fold reduced proliferation in the saline-treated 8B-Mφ co-culture compared to the M-CSF induced Mφ co-culture. Notably, CD4+ T cell proliferation in the NMN-treated 8B-Mφ co-culture was restored to the level as was obtained in the M-CSF induced Mφ co-culture. Thus, NMN treatment improves the immune activating capacity of 8B-Mφs.
NMN supplementation therapy prevents tumour initiation in 8B-inoculated immunocompetent mice
Finally, we analysed the effectiveness of NMN treatment in preventing 8B-tumour initiation using the orthotopic tumour transplantation model in immunocompetent C57BL/6 mice (Fig. 4F). In the saline treated group, ~70% of mice showed tumour initiation and died as a result. In contrast, in the NMN treated group, tumour initiation was observed in only ~20% of mice, and survival was improved compared to that in the saline group (P = 0.076, Log-rank test).
In the glioblastoma orthotopic transplantation model, it is difficult to precisely determine the date of tumour initiation and formation. To examine the precise timing, we attempted to analyse subcutaneous tumour formation. However, 8B and 9G cells did not form any tumour when subcutaneously inoculated (data not shown). Therefore, we chose CT26 subcutaneous engraftable syngeneic tumour model to analyse the effect of NMN on tumour formation. As the CT26 mouse colorectal cancer cells secrete IL-6[36], Mφ senescence induction via IL-6 in the tumour microenvironment is expected. We therefore injected CT26 cells subcutaneously into syngeneic BALB/c immunocompetent mice and evaluated the effect of NMN treatment on tumour formation. The NMN treatment significantly delayed the tumour initiation (P < 0.05, Log-rank test) (Fig. 4G) and efficiently inhibited the tumour growth (Fig. 4H). Collectively, these results suggested that NMN treatment suppressed the senescence of Mφs that induces T cell dysfunction by expressing Arginase-1, and eventually but temporarily, prevented tumour initiation in immunocompetent mice.