Type I IFNs (IFN-I) are important for tumor immune surveillance and contribute to the therapeutic responses for numerous treatment regimens. Nevertheless, certain protumoral activities by IFN-I have been increasingly recognized. Indeed, our recent work showed that systemic poly(I:C)/IFN treatment can undesirably trigger high arginase (ARG1) expression within the tumor-associated monocyte/macrophage compartment. Using a line of CRISPR-generated Arg1-YFP reporter knock-in mice, we have determined that a subset of tumor-associated macrophages represent the major Arg1-expressing cell type following poly(I:C)/IFN stimulation. More detailed analyses from in vitro and in vivo models demonstrate a surprising IFN–to–IL-4 cytokine axis in transitional monocytes, which can subsequently stimulate IL-4 target genes, including Arg1, in macrophages. Intriguingly, IFN stimulation of transitional monocytes yielded concurrent M2 (YFP+)- and M1 (YFP–)-skewed macrophage subsets, correlated with an inhibitory crosstalk between IFN-I and IL-4. Genetic abrogation of IL-4 signaling in mice diminished poly(I:C)/IFN-induced ARG1 in tumors, leading to enhanced activation of CD8+ T cells and an improved therapeutic effect. The present work uncovered a monocyte-orchestrated macrophage phenotype conversion mechanism that may have broad implications.
Recent progress in cancer immunotherapies has established the essential roles by the adaptive immunity, particularly the tumor-specific T cells, in the therapeutic success against cancers (1, 2). As innate immunity contributes to the initiation, amplification, and the effector actions of the adaptive immunity, to properly program the innate immune system also holds great promise to improve the management of cancer patients (3).
It has been shown that engagement of nucleic acid–induced innate immune response, mimicking one after viral infection, can promote antitumor adaptive immunity (4, 5). A converged downstream event by various nucleic acid–elicited innate immune signaling is the production of type I IFNs (IFN-I) (6, 7). IFN-Is function through a common receptor expressed by most cell types, including nearly all lineages of immune cells (8). As potent immune-regulatory cytokines, IFN-Is function to enhance Ag presentation, promote NK cell activities, and directly improve T cell responses, contributing to a generally positive role on antitumor immunity (7, 9, 10). Nevertheless, the immune-regulatory activities by IFN-I are dichotomous, including some immunosuppressive effects that can be protumoral (11, 12). Therefore, improved understandings to the latter aspects of IFN-I are key to optimize the antitumor potential of innate immunity.
Within the tumors, tumor-associated macrophages (TAMs) are the most abundant immune cells (13). A major origin for TAMs are the infiltrating circulation-borne monocytes that undergo further intratumoral differentiation (14, 15). Moreover, some elegant works have shown that embryonically derived, self-renewing tissue-resident macrophages can also significantly contribute to the TAM pool (16). Owing to their high responsiveness to local inflammatory milieu and physiochemical cues, monocytes and the TAMs are subjected to dynamic regulations (15, 17, 18). The functional plasticity of macrophages has been well established (19, 20). An older view in which the macrophages can mainly polarize toward an M1 or M2 (corresponding to stimulation by either the Th1 or Th2 cytokines) phenotype has been lately modified to a more generalized notion that these cells exhibit heterogeneity at many different levels and may adopt a wide spectrum of phenotypic states (21–23). Functionally, the tumor-associated monocytes and TAMs play major proneoplastic and immune-suppressive roles during tumor progression (13, 15, 24). Additionally, under therapeutic contexts, the dynamic behaviors of these cells also strongly influence the treatment effects (13, 25, 26).
Given the complexity and functional significance of tumor-associated monocytes and TAMs, we had recently studied their regulation by IFN-I in a mouse model. IFN-I was found to engage a previously unsuspected protumoral axis in tumor-associated differentiating monocytes, leading to their high arginase 1 (ARG1) expression and subsequent inhibition of antitumor immunity (27). In the current study, we continued to explore the mechanisms underlying such an IFN–ARG1 regulatory axis. Our data demonstrate an unanticipated activity by IFN-I to induce IL-4 production in monocytes. This forms a feedforward, paracrine cytokine axis that shapes a subset of M2-skewed TAMs to curtail the treatment effects of IFN-I. The current work has provided some new insights to IFN-I–mediated regulation of antitumor immunity.
Materials and Methods
The animal care and use protocols were in strict accordance to the Regulation for Management of Laboratory Animals (1988) and Guidelines for Care and Use of Laboratory Animals (2006) issued both by the Ministry of Science and Technology of People’s Republic of China. The animal experiments were approved by the Institutional Animal Care and Use Committee of Model Animal Research Center of Nanjing University. The mice were housed and experimented in an Association for Assessment and Accreditation of Laboratory Animal Care–accredited facility at the Model Animal Research Center of Nanjing University. All mice used are of C57/BL6J background. Wild-type (WT), Ifnar1−/−, and Il4ra−/− mice (CRISPR lines, catalog no. T007665 and T005980, respectively) were purchased from GemPharmatech (Nanjing, China). Arg1-YFP knock-in reporter mice were generated using a CRISPR-powered strategy outlined in Supplemental Fig. 1A. In the modified allele, an internal ribosomal entry site (IRES)-eYFP follows the Arg1 STOP codon within the last exon (no. 8), similar to the design in a previously constructed Arg1 reporter line (28). To expand this mouse model’s potential range of applications, two individual LoxP sites were designed either upstream of exon 5 or downstream of eYFP. Although no attempts have been made in the current study to use these LoxP sites, such a design is conducive for future generation of Arg1 conditional knockout. For construction of this line, a sequence corresponding to the portion of the modified Arg1 allele (LoxP-exon 5–8 (5′)-IRES-YFP–exon 8 (3′)-LoxP, ∼3 kb in size) flanked by 5′ and 3′ homologous arms was first cloned into a donor vector. Moreover, two single guide RNAs were designed to target upstream of exon 5 and downstream of the stop codon, respectively. Next, the fertilized eggs from the C57BL/6 mice were injected with Cas9/single guide RNA ribonucleoproteins together with the donor vector to allow WT allele cleavage and subsequent homologous recombination. The resultant founders were screened for ones that had undergone the correct genome editing. Tail DNA samples were analyzed by PCR. The sequences of different primers and expected product sizes are listed in Supplemental Fig. 4D.
Mice (both male and female, aged 6–10 wk) were injected s.c. with 1 × 106 Lewis lung carcinoma (LLC) cells at the flanks. After the tumors became palpable, their sizes were measured every 2 d and calculated as follows: ([length × width2]/2). The i.p. treatments with polyinosinic–polycytidylic acid (poly(I:C)) (5 mg/kg) were initiated when tumors reaching ∼50 mm3 and were generally repeated on 2-d intervals.
Poly(I:C) (no. Tlrl-pic-5) was from InvivoGen. Mouse IFN-β (no. 12405–1) was from PBL Assay Science. Mouse M-CSF (no. 315–02), IFN-γ (no. 315-05), and IL-4 (no. 214-14) were from PeproTech. The ELISA kits for mM-CSF (no. MMC00), mouse TNF-α (no. RC052) were from R&D Systems. IL-4–neutralizing Abs were from Bio-X-Cell (no. BE0045) and BioLegend (no. 504122). IFNAR1 blocking Ab (no. 127302) was from BioLegend.
Abs for immunoblotting and flow cytometry
Primary Abs for Western blotting were purchased from Cell Signaling Technology (ARG1, no. 93668; pStat6-Y641, no. 565543; Stat6, no. 93623; CSF1R, no. 3152), Sango (Stat1, no. AB55186), and Genscript (actin, no. A00730). Abs for flow cytometry were purchased from BioLegend: (CD16/32, no. 101330; CD45, no. 103116; CD11b, no. 101228; Ly6G, no. 127618 and no. 127614; Ly6C, no. 128022 and no. 128037; F4/80, no. 123122 and no. 123132; MHC class II [MHC II] (I-A/I-E), no. 107618; CD206, no. 141712; IL4, no. 504118; IL4Ra, no. 144808; CD4, no. 100406; CD8, no. 100737; IFN-γ, no. 505812; CCR2, no. 150603; Siglec-F, no. 155505; CD64, no. 139323; AF647 Rat IgG2b, κ Isotype, no. 400626; AF647 Rat IgG2a, κ Isotype, no. 400526; PE/Cyanine7 Rat IgG1, κ Isotype, no. 400415 and allophycocyanin Rat IgG2a, κ Isotype, no. 400511). The viability of cells was determined using LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (no. L34965; Invitrogen).
Tumor samples and preparation of cells for flow cytometry
Processing of tumor samples and analyses of immune compartments were proceeded as previously described (27). Briefly, to prepare samples for flow cytometry, the tumors were dissociated with the help of enzymatic digestion, whereas the spleen and lung tissue was triturated using the end of a syringe plunger. Following erythrocyte lysis, the remaining cells were subjected to Ab staining (with Fc block) and subsequently subjected to flow cytometry (BD LSRFortessa). For establishing the gating strategies, controls of fluorescence minus one (FMO) staining were performed. For analyses of intracellular Ags, cells were first subjected to cell surface staining, followed by fixation/permeabilization (using BD Cytofix/Cytoperm Plus reagent) and the 2nd round of Ab staining.
BD FACSAria III high speed sorter was used for sorting experiments. Monocytes from tumor were identified as CD45+CD11b+Ly6G−Ly6C+F4/80− cells. Monocytes and macrophages from differentiated monocytes in vitro were identified and sorted based on characteristics of Ly6C+F4/80− and F4/80+, respectively. In some experiments, F4/80+ cells were separated further based on YFP signals.
The LLC cells (American Typer Culture Collection) were cultured in DMEM medium supplemented with 10% FBS, 100 U/ml penicillin, and 100 μg/ml streptomycin. The preparation of bone marrow (BM) mononuclear cells was performed via centrifugation through gradient of Histopaque-1077 (GE Healthcare), as previously described (27). In some cases, the Monocyte Isolation Kit (Miltenyi Biotec) was used to isolate BM monocytes with high purity. The mM-CSF was added to instruct BM cell differentiation toward macrophages w/or w/o cotreatment with IFN-β (100 U/ml) in the same base medium described above. In some experiments, the supernatant from minced LLC tumors [mock- or poly(I:C)-treated for 24 h] was used to drive macrophage differentiation (27). For flow cytometry, the differentiated macrophages were first dissociated from the plates with 0.02% EDTA in PBS.
Western blot and quantitative PCR analyses
Western blot analyses were performed according to standard procedures and protein bands were detected by chemiluminescence. Quantitative PCR (qPCR) analyses were based on the SYBR system and performed on the ABI Step One Plus platform. The housekeeping genes RPL19 and 36B4 were used as internal controls. The sequences for qPCR primers used are listed in Supplemental Fig. 4E.
After YFP-based cell sorting, total RNA from samples were isolated (TRIzol reagent). The RNA sequencing (RNAseq) and data analyses were performed by Annoroad Gene Technology (Beijing, China). Briefly, the RNA integrity score was first determined (Agilent 2100 RNA Nano 6000 Assay Kit). The mRNAs were enriched, purified, and fragmented to construct the libraries (for the Illumina platform) according to manufacture instructions. The DNA libraries were sequenced using Strategy PE150. The clean reads were mapped against Mus_musculus.GRCm38.91. Gene expression value was quantitatively estimated through fragments per kilobase per million mapped fragments. Genes with significantly differential expression between groups was determined by DESeq2, using a cutoff of fold change ≥ 2 and p < 0.05. Gene ontology enrichment analyses were performed using PANTHER classification system (29). The raw and processed data were deposited at Gene Expression Omnibus (GSE161428).
All data presented in this study are derived from at least two independent experiments. For quantitative results, average values from biological replicates were presented with error bars denoting SEM (n ≥ 3) or sometimes the data range (when n = 2). The p values were determined by Student t tests and are represented as follows: *p < 0.05, **p < 0.01, ****p < 0.0001.
Poly(I:C)-IFN drives a subset of ARG1high macrophages selectively in tumors
Activation of nucleic acid–induced innate immune signaling has shown promise in cancer therapy (4, 5). Nevertheless, our previous work on LLC model in mice uncovered that although i.p. poly(I:C) treatment caused inhibition of tumor progression, it also led to an induction of protumoral enzyme ARG1 in tumor-associated differentiating monocytes (27). To further our understandings to such an intriguing protumoral regulatory axis, we established (via CRISPR technology) a new mouse line in which a reporter (eYFP) was knocked-in at the 3′-UTR of Arg1 gene, led by an IRES (Supplemental Fig. 1A). As ARG1 expression are largely controlled at the levels of transcription (30), this reporter would allow convenient tracking of ARG1+ cells in complex tissues such as tumors. Consistent with the well-known hepatic expression of ARG1, strong YFP fluorescence was present in the hepatocytes of Arg1-YFP heterozygous mice (Supplemental Fig. 1B). Furthermore, in established s.c. LLC tumors, we observed via flow cytometry that a noticeable (∼7%) population of CD45+ stromal cells were positive for YFP (tumor samples from the WT nonreporter mice as negative control), and that most of these YFP+ cells exhibited features of TAMs (CD45+F4/80+) (Supplemental Fig. 1C), consistent with previous observations (31, 32).
Having validated our Arg1-YFP mouse model, we next tried to thoroughly establish the cellular compartment featuring increased Arg1 expression following in vivo poly(I:C) stimulation (27). Poly(I:C) was administered (i.p.) to the reporter mice bearing LLC tumors (repeated on 2-d intervals for four treatments). The treatment led to a robust increase (3-fold) in the relative abundance of YFP+ cells within the CD45+CD11b+Ly6G–F4/80+ TAM compartment (Fig. 1A, with FMO gating controls shown in the bottom box), and a paralleled enhancement of YFP fluorescent intensity in these cells. Such induction of YFP signals in the TAMs occurred despite an overall reduction in TAM numbers (∼40%, in conjunction with an marked increase of their precursor Ly6C+ monocytes), a previously characterized effect attributed to an IFN/mir-155 pathway–mediated negative regulation of M-CSF receptor (CSF1R) and monocyte maturation (27). Notably, the poly(I:C)-driven population of CD45+CD11b+Ly6G–F4/80+YFP+ (ARG1+) TAMs had largely downregulated levels of MHC II (Fig. 1B), consistent with the notion that these cells represent a protumoral subset (33). Both the YFP– and YFP+ subsets in the poly(I:C) group of tumor-derived CD45+CD11b+Ly6G–F4/80+ cells additionally showed noticeable expression of CD206 (Supplemental Fig. 1D), consistent with their TAM identities (13). In contrast to the situation in the tumors, the presence of CD45+YFP+ cells in the spleen was negligible (Supplemental Fig. 1E, red box). Furthermore, other tumor-associated immune cell compartments, including the CD45+CD11b– nonmyeloid cells, CD45+CD11b+Ly6G+ granulocytic cells, or CD45+CD11b+ Ly6G–F4/80–Ly6C+ monocytes, had minimal YFP+ cells, regardless of poly(I:C) treatment (Supplemental Fig. 1E). In confirmation, TAMs (CD45+F4/80+) similarly represented the majority of all intratumoral YFP+ cells with or without poly(I:C) treatment (Supplemental Fig. 1F). Kinetically, we found that the YFP+ TAM population (CD45+CD11b+Ly6G–F4/80+) accumulated over repeated administrations (Fig. 1C), consistent with the time-dependent trend of total ARG1 levels in tumor tissues (but not in lung) (Supplemental Fig. 1G). Taken together, extending our previous study (27), these analyses established that the TAM compartment was selectively responsible for systemic poly(I:C) treatment–induced ARG1 expression.
As one of the potent innate immune agonists (6), poly(I:C) delivered i.p. was shown to induce a systemic IFN-I response (34). Indeed, we previously had demonstrated poly(I:C) treatment–elicited induction of IFN-I activities in tumors (27). To formally confirm the role of IFN-I in poly(I:C)-triggered events, we adopted a CRISPR-generated IFN-I receptor (IFNAR1)–deficient mouse line (Fig. 1D). Consistent with IFN-I’s potent antitumoral functions (7, 9), poly(I:C) became ineffective in controlling LLC tumor growth in the IFNAR1-deficient mice (Fig. 1D). In the ensuing experiments, we examined the changes on TAMs. In agreement with the flow cytometry data on TAM numbers and YFP (Arg1) (see (Fig. 1A), the poly(I:C) group of tumor lysates from the WT mice showed downregulation of CSF1R, together with a reciprocal induction of ARG1 (Fig. 1E). Notably, such TAM-related alterations by poly(I:C) were jointly abrogated by IFNAR1 deficiency (Fig. 1E). These results establish that poly(I:C)-induced IFN-I is responsible for noncanonically inducing the protumoral ARG1+ TAMs, concurrently with its engagement of apparent antitumor effects.
IFN-stimulated monocytes release an ARG1-inducing soluble factor(s)
Given that IFN-I is mostly considered as an M1-type cytokine for macrophages (35–38), we sought to further understand the mechanism(s) underlying its induction of protumoral ARG1 in TAMs. We previously established an in vitro system that modeled IFN’s action upon monocytes-to-TAM transition, and largely recapitulate the observed IFN-dependent ARG1 induction in tumors (also see (Fig. 1) (27). Particularly, IFN treatment of differentiating (under M-CSF instruction) BM monocytes, or BM mononuclear cells as convenient proxies, could lead to strong ARG1 induction (illustrated in (Fig. 2A) (27). In comparison, simply adding IFN to BM-derived mature macrophages failed to induce Arg1 expression. These in vitro observations suggest that IFN-I’s impact is initiated on less-differentiated monocytes and that a subsequent induction of Arg1 is promoted by M-CSF–instructed macrophage maturation. In corroboration, the dependence of poly(I:C)–ARG1 axis in tumors in vivo on the concurrently recruited monocytes, as well as on M-CSF signaling were established by application of respective inhibitors (27). In this study, we further verified that the LLC tumors produce very high levels of M-CSF compared with normal tissues such as lung and liver (Supplemental Fig. 1H). Therefore, M-CSF ± IFN-treated BM monocytes serve as a relevant in vitro model to understand the IFN–ARG1 axis in tumors.
Further experiments showed that after IFN addition to BM monocytes, the output of induced ARG1 protein (mRNA alike) occurred with an evidently delayed kinetics (Supplemental Fig. 1I), and appeared to be correlated with M-CSF–driven maturation of monocytes toward macrophages. To better understand this process, we took advantage of the Arg1-YFP reporter line and applied M-CSF ± IFN treatment on BM mononuclear cells derived from these mice. IFN treatment did not apparently affect the viability of the cells under the experimental condition (Supplemental Fig. 1J). Remarkably, the YFP+ cells induced by IFN in this model only appeared as differentiated F4/80+ macrophages (Fig. 2B, Supplemental Fig. 1K), confirming the separation of the initial IFN-I signaling and the subsequent increase of Arg1 levels respectively into cells of two differentiation stages. Interestingly, in such IFN-induced, YFP+ monocyte-derived macrophages (CD45+CD11b+Ly6G–F4/80+), there existed significant coinduction of another M2-like macrophage marker CD206 (Fig. 2C), supporting the engagement of a coordinated phenotype-shifting program.
To probe the stage-wise requirement of IFN signaling for driving an ARG1high phenotype in monocyte-derived macrophages, we added a neutralizing Ab against IFN-I receptor after the initial 24 h of IFN treatment of transitional monocytes (Fig. 2D, 2E). Such late blockade of IFN-I signaling was sufficient to decrease the levels of a classical IFN-stimulated gene (ISG) (i.e., STAT1; (Fig. 2E). In stark contrast, neither the numbers of YFP+ population in the F4/80+ macrophages nor ARG1 levels were inhibited, but instead moderately increased upon the same treatment (Fig. 2D, 2E), strongly suggesting that IFN-I signaling was mainly required at an early stage in the IFN–ARG1 axis. These results, together with the fact that M-CSF + IFN treatment of mature macrophages did not induce Arg1 (27), support the notion that the IFN–ARG1 axis may first involve IFN-I–initiated early-phase events in monocytes, which in turn, engage subsequent late-phase events in monocyte-derived macrophages to drive their M2-like (i.e., ARG1+CD206+) phenotype.
Because monocytes/macrophages often shape inflammatory responses via secretion of soluble factors (39), we explored whether IFN signaling in monocytes may result in some secondary, secreted factor to next skew the differentiated macrophages toward an ARG1high phenotype. To this end, monocytes (modeled by BM mononuclear cells) or differentiated macrophages were treated with M-CSF ± IFN for 40 h. At harvest, the conditioned medium (CM) of the transitional monocytes were transferred to naive, differentiated macrophages. In the first part of the experiment, IFN-I expectedly induced ARG1 in the transitional monocyte culture, but not in the macrophage culture (Fig. 2F). Remarkably, in the ensuing part of experiment, the CM from IFN-treated transitional monocytes was able to trigger robust ARG1 induction (mRNA alike) in differentiated macrophages (Fig. 2F, Supplemental Fig. 1L), strongly suggesting the release of a secondary, phenotype-shifting ligand(s) (illustrated in (Fig. 2G).
An unanticipated monocyte-dependent IFN–to–IL-4 cytokine axis drives the coexisting macrophages toward an M2-skewed phenotype
Based on the above analyses, we decided to focus on characterizing the IFN-elicited secondary factor(s) from monocytes that can reprogram macrophages toward an M2-like phenotype. We reanalyzed the microarray data obtained previously from M-CSF + IFN-treated purified BM monocytes (27). Amid many expected antiviral and immune activation markers upregulated by IFN, a noticeable induction of Il4 (in association with Arg1) caught our attention (Fig. 3A). Although IL-4 is a classical driver cytokine for M2 macrophages, monocytes/macrophages are not generally considered as its major producers (40). Additionally, Il4 is not commonly known as an ISG. In contrast, inhibitory actions of IFN-I on IL-4–mediated response have been established, reflecting negative crosstalk between an M1-type proinflammatory and an M2-type prorepair signals (37, 41). Nevertheless, when our microarray results were compared with a published mRNA-profiling database from IL-4–stimulated, BM-derived macrophages (BMDM) (42), some overlap in induced genes (including Arg1) was apparent. Indeed, 6 out of the 12 such overlapped genes were among the top-10 most induced M2 genes from the published IL-4–dependent list (Fig. 3B). Furthermore, induction of these typical M2 markers were validated by qPCR in IFN-treated transitional monocytes in the presence of M-CSF (Supplemental Fig. 2A), supporting an induction of bioactive IL-4.
We confirmed the time-dependent induction of Il4 mRNA by IFN signaling in the M-CSF–cultured BM mononuclear cells. Within the time points analyzed, its induction precedes that of Arg1 (see Supplemental Fig. 1I) and peaked at 24 h, which was then down to a lesser magnitude at 48 h (Fig. 3C). In comparison, direct treatment of differentiated macrophages with IFN failed to induce Il4 (Fig. 3D), mirroring the regulatory pattern observed for Arg1 (27). Next, after M-CSF ± IFN stimulation of BM mononuclear cells, we sorted out the monocyte (Ly6C+F4/80–) and the macrophage (F4/80+) compartments to dissect their differential contributions to Il4 or Arg1 induction. Indeed, IFN-mediated Il4 induction showed a monocyte-preferential pattern, whereas Arg1 induction was expectedly featured in macrophages (Supplemental Fig. 2B, 2C). The latter macrophage-associated pattern of Arg1 induction is in line with previous observations that IL-4–mediated Arg1 induction positively correlates with macrophage maturation status (28). In contrast, consistent with the monocyte-specific Il4 induction pattern, we found that the macrophage-driving maturation signal M-CSF was not necessary for such an IFN-I–specific effect (no effect by IFN-γ or TNFα) in BM mononuclear cells (Supplemental Fig. 2D, 2E). Flow cytometry analyses of intracellular IL-4 within IFN-treated BM mononuclear cells further validated the monocyte-specific induction pattern (Fig. 3E, compare the Ly6G–Ly6C+ gate with the Ly6G+ gate of CD45+CD11b+ cells). Importantly, a monocyte-intrinsic feature of the IFN–Il4 axis was firmly demonstrated by results from treatment of magnetic beads-purified (>90% as Ly6C+) BM monocytes (Fig. 3F). It is interesting to note that in a seemingly cooperative manner, IFN signaling also triggered upregulation of IL-4 receptor (subunit α, IL-4RA) levels in monocytes (CD45+CD11b+Ly6G–F4/80–Ly6C+) and macrophages (CD45+CD11b+Ly6G-F4/80+) (Fig. 3G), but not in other less represented cell compartments within the M-CSF–instructed BM mononuclear culture (Supplemental Fig. 2F).
To functionally analyze the released IL-4 activity, we next applied a neutralizing Ab against IL-4 to the CM from IFN-stimulated transitional monocytes. Such blockade of IL-4 abolished the ARG1-inducing activity by the CM on naive BMDMs, confirming IFN-mediated release of active IL-4 from monocytes (Fig. 3H). In contrast, a blocking Ab against IFN-I receptor moderately potentiated the CM-mediated ARG1 induction (Fig. 3H), which together with our earlier observations on late IFN-I blockade (see (Fig. 2D, 2E), are in line with the well-known antagonizing effect of IFN-I on IL-4 signaling and function (37, 41). In the following experiments, the IL-4–blocking Ab was added with IFN to the transitional monocyte culture. The Ab eliminated IFN-mediated induction of ARG1 together with that of pSTAT6 (Supplemental Fig. 2G), a classical component of IL-4 signaling (19). In transitional monocytes from Arg1-YFP mice, a similar Ab-dependent inhibition of IFN-induced YFP+ population within F4/80+ macrophages was also observed (Supplemental Fig. 2H).
We further adopted a CRISPR-generated Il4ra (encoding IL-4RA) knockout mouse model. As expected, BMDMs from the Il4ra−/− mice were completely refractory to IL-4–mediated Arg1 induction (Supplemental Fig. 2I). Importantly, IFN-mediated induction of ARG1 (Fig. 3I), as well as numerous other M2 macrophage markers were inhibited in monocyte-to-macrophage transitional culture genetically deficient in IL-4RA (Supplemental Fig. 2J, 2K). In contrast, expression of a classical ISG (Irf7) and that of Il4 was apparently not affected by IL-4RA deficiency (Fig. 3I). Together with the Ab-mediated IL-4 blockade, these results from Il4ra−/− culture model establish that IFN-induced IL-4 in transitional monocytes may act as a “feedforward,” function-switching signal to initiate an M2-skewing phenotypic conversion in concurrently differentiated macrophages.
To corroborate such findings, we adopted a more relevant in vitro system to mimic the subsequent changes in tumor-associated monocytes and macrophages upon in vivo poly(I:C) treatment (i.p.) (27). To this end, the cell-free supernatants of the minced tumor tissues [from mock- or poly(I:C)-treated mice] were used to treat BM mononuclear cells (illustrated in Supplemental Fig. 2L) (43). The poly(I:C)-group of supernatant samples were previously shown to contain abundant IFN-I, which could in turn upregulate Arg1 in BM-derived, transitional monocytes (27). In this study, we further demonstrated substantial induction of Fizz1, another M2 macrophage marker, along with Arg1, in cells treated with the poly(I:C)-group of supernatants (Fig. 3J). Indeed, the upregulation of Arg1 and Fizz1 mRNAs was largely abrogated by a neutralizing Ab against IL-4 (Fig. 3K). These results demonstrate that the IFN–IL-4 axis in monocytes can be readily engaged within a tumor-associated milieu of soluble factors, supporting the relevance of earlier results from the simplified, IFN/M-CSF treatment model (shown in (Figs. 2 and (3). Interestingly, some M1 macrophage markers such as H2-d1 and Ccr2 (Fig. 3J), but not Nos2, Il1b and Tnfa (Supplemental Fig. 2M), also showed moderate upregulation, consistent with a common perception of IFN-I as M1-type signal.
To gain some mechanistic insights regarding IFN-mediated Il4 induction, we pretreated BM mononuclear cells with a protein synthesis inhibitor (cycloheximide [CHX], 100 ng/ml). In agreement with the direct mode of action by the IFN–STAT regulatory pathway on the transcription of classical ISGs, CHX treatment did not affect IFN-mediated induction of Isg15. In stark contrast, Il4 induction by IFN was largely abrogated by CHX treatment (Fig. 3L). Ionomycin, which is a calcium ionophore capable of inducing Il4 expression in various immune cell types (44, 45), was used as another control. Indeed, ionomycin potently induced Il4 mRNA levels (∼40-fold in 12 h) in BM mononuclear cells. However, different from its inhibitory effect on the IFN–IL-4 axis, CHX instead enhanced ionomycin-triggered induction of Il4 for more than 4-fold (not shown). These results, together with the observation of a relatively delayed, IFN-dependent Il4 induction pattern (compare (Fig. 3C and Isg15 in Supplemental Fig. 1I, and a further delayed Arg1 serving as another reference), support that the mechanisms underlying Il4 induction are most likely to extend beyond the canonical IFN signaling.
IFN-I triggers IL-4 induction in normal and tumor-associated monocytes in vivo
We next sought to corroborate the in vitro characterized, monocyte-dependent IFN–IL-4 cytokine axis in mice. Poly(I:C)-treated tumor-bearing mice was first used for analyses. In agreement with the in vitro results (see (Fig. 3), poly(I:C) treatment for 24 h could lead to increased Il4 mRNA and protein levels in CD45+CD11b+Ly6G–F4/80–Ly6C+ tumor-associated monocytes (Supplemental Fig. 2N and (Fig. 4A, left). In contrast, no such induction of IL-4 was noted in the granulocyte or macrophage compartments (Fig. 4A, middle and right). Moreover, a similar induction of IL-4 in CD45+CD11b+Ly6G–Ly6C+ monocytes from the spleen was observed (Supplemental Fig. 2O). Importantly, using IFNAR1-deficient mice, we firmly established that poly(I:C)-induced IL-4 in tumor-associated monocytes (CD45+CD11b+Ly6G–F4/80–Ly6C+) was dependent on IFN signaling (Fig. 4B). Therefore, although TLR signaling in macrophages under certain context was previously shown to induce IL-4 (46), IFN-I is apparently the direct driver of monocytic IL-4 expression in our model of poly(I:C) treatment (i.p.). Indeed, our previous study also documented a lack of direct effect by poly(I:C) on BM monocytes (27), likely owing to their low expression of TLR3 (47).
As the LLC cells model lung tumor, we next seek whether the IFN–IL-4 cytokine axis can be engaged in a more relevant tissue environment. To this end, normal mice were administered (i.p.) with a single-dose of poly(I:C), and the dissociated cell suspensions of lung tissues were subjected to flow cytometry. A modified gating strategy (from that in (Fig. 1A) was used in this study in reference to established standards (Fig. 4C) (48, 49). The well-established resident alveolar macrophages were readily recognized as CD45+CD11b–Siglec-F+ cells (blue gate). In contrast, few interstitial macrophages with a CD45+CD11b+Siglec-F–Ly6G–Ly6C–CD64+ phenotype were present in the control mice, in general agreement with their low abundance at homeostasis shown previously (48). Although cells of this phenotype were found to greatly accumulate in lung tumors (48), our analyses of lungs from poly(I:C)-treated control mice did not find apparent induction of this subset. Interestingly, poly(I:C) led to a marked increase of CD64 staining in CD45+CD11b+Siglec-F–Ly6G–Ly6C+ monocytes, conceivably attributed to the induction of this surface marker by IFN (50). When the abundances of different myeloid cell types were normalized to total CD45+ cells, only the numbers of CD45+CD11b+Siglec-F–Ly6G–Ly6C+ monocytes (black gate) were found to be significantly expanded following 24 h of poly(I:C) treatment (Fig. 4C, bar graph on the right), suggesting their specific recruitment to the lung. Importantly, these Ly6C+ monocytes exhibited apparent poly(I:C)-dependent IL-4 induction (Fig. 4C, IL-4 panels), with the CD64– and the emerging CD64+ population in the poly(I:C) group showing very similar IL-4 levels (inset). It is noteworthy that the alveolar macrophages (blue gate) from the treated mice also showed apparent induction of IL-4 expression. In contrast, such effects were not observed in CD45+CD11b+Siglec-F–Ly6G–Ly6C– monocytes (gray gate) and CD45+CD11b+Siglec-F–Ly6G+ granulocytes (not shown). These results extend IFN-dependent, IL-4–producing cells to Ly6C+ monocytes and the resident alveolar macrophages in lung, implicating their potential regulation of TAM responses in orthotopic lung tumors. Interestingly, different from earlier observation in TAMs (see (Fig. 1A), little induction of YFP was observed in alveolar macrophages of the normal lung following repeated poly(I:C) administration (not shown), pointing to important roles of tissue context and/or ontogeny of macrophages in shaping their behaviors.
The IFN–IL-4 cytokine axis can drive the coexisting macrophages into different subsets
In many factors that may shape macrophage phenotypic heterogeneity, the complex and dynamic cytokine milieu represents a major determinant (22, 51). In this regard, IFN stimulation of BM-derived transitional monocytes provide a suitable model to investigate such an aspect under a shifting cytokine environment. In this model, we noted that IFN treatment only drove a portion of the subsequently differentiated macrophages to express Arg1 (YFP) (see (Fig. 2B–D), suggesting that variable YFP levels may mark distinct macrophage subsets under the influences of an IFN–IL-4 cytokine axis. Therefore, we next probed the overall differences between the YFP+ and YFP– macrophages following their transition from monocytes under M-CSF ± IFN treatment. Based on treatment as well as YFP intensities, three group of macrophages, i.e., control (the predominant YFP– cells selected), IFN/YFP– and IFN/YFP+, were sorted out via flow cytometry (Supplemental Fig. 3A). The samples were subjected to RNAseq analyses. At the global level, both subsets of IFN-stimulated, monocyte-derived macrophages showed similarly substantial mRNA expression changes compared with the control, with the majorities of alterations being gene inductions (Fig. 5A, Supplemental Fig. 3B). As expected, the most enriched biological functions for the upregulated genes in the treated groups relate to immune activation, antiviral responses and the IFN pathway (Supplemental Fig. 3C). Nevertheless, a smaller group of genes also showed differential expression between IFN/YFP+ and IFN/YFP– macrophages, with the numbers of genes regulated in either direction being more balanced (Fig. 5A). Intriguingly, classical IFN downstream genes did not display differential expression between the two subsets (Fig. 5B), implying that the two subsets differ in more specific functionalities linked to IFN.
Among the list of 84 genes with higher expression in IFN/YFP+ than the IFN/YFP– subsets, only a relatively smaller portion of genes appeared to have higher expression in the control compared with the IFN/YFP– samples (Supplemental Fig. 3D). Most of them instead exhibited patterns of moderate upregulation in the IFN/YFP– samples compared with the control, and a notably greater induction in the IFN/YFP+ samples (Fig. 5C). It is important to note that, besides Arg1, several other M2 macrophage markers including Fizz1, Chil4, and Ccl24 belong to this category (Fig. 5C), consistent with a productive response to monocyte-derived IL-4. Overall, the 84 genes within the “YFP+ above YFP–” category show enrichment for association with extracelluar matrix (Supplemental Fig. 3E), indicating that the YFP+ macrophages have acquired the “wound repair”–type, M2-skewed functionalities (19). We also validated the greater induction of a series of M2 markers in the YFP+ population by qPCR (Fig. 5D).
In contrast, compared with the IFN/YFP– cells, their YFP+ counterparts also featured numerous downregulated genes (Fig. 5A). Within such a list of 98 genes, most indeed showed induction from the control to the IFN/YFP– group (Fig. 5E), including some M1-like macrophage markers such as Ccr2, Ciita, and H2-Aa (52, 53). The higher levels of Ccr2 and Ciita mRNAs in the YFP– than the YFP+ cells were validated by qPCR (Supplemental Fig. 3F). Genes downregulated in the YFP+ group also include two other minor clusters that show little induction or even downregulation in IFN/YFP– samples in comparison with the control (Supplemental Fig. 3G). Overall, functions associated with immune activation, leukocyte adhesion and cytokine response are enriched in these 98 genes of “YFP– above YFP+” category (Supplemental Fig. 3H). Collectively, our results from a simple, defined in vitro model show that IFN–IL-4 cytokine axis in transitional monocytes can shape the coexisting differentiated macrophages into either M2- or M1-skewed subsets, readily distinguishable by their Arg1 expression.
It is worth noting that IFN/YFP– and IFN/YFP+ monocyte-derived macrophages had similarly induced levels of pSTAT6 (Fig. 5F), implying that mechanism(s) other than direct IL-4/STAT6 signaling may define the YFP+/YFP– subsets in the present system. Antagonistic effects between two cytokines may possibly contribute to shaping distinct macrophage subsets (54, 55). The antagonism of IFN-I on IL-4–dependent responses has been previously reported (37, 41), and is suggested by our own results where IFN signaling blockade after its induction of IL-4 moderately enhances ARG1 expression in macrophages (Figs. 2D, 2E, 3H). Likewise, coaddition of IFN-I and IL-4 to differentiated BMDMs expectedly led to significant inhibition of IL-4–dependent ARG1 induction (Fig. 5G). Interestingly, IL-4–induced pSTAT6 levels also did not appear to be affected by IFN addition (Fig. 5G), correlated with the earlier YFP+/YFP– cell analyses (Fig. 5F).
In contrast, how IL-4 would regulate IFN-I–dependent responses had been under studied. Consistent with the results from IL-4RA–deficient cells (see (Fig. 3I), we found that IFN-I–driven expression of classical ISGs are not affected by exogenous IL-4 signaling (Supplemental Fig. 3I, J). We next turn our attention to the several M1-type genes (i.e., Ccr2, Ciita, H2-Aa) specifically induced in the IFN/YFP–, M1-skewed group of macrophages (see (Fig. 5E). Within these markers tested, only Ccr2 were consistently upregulated by IFN-I in differentiated BMDMs (not shown). Importantly, Ccr2 induction by IFN can be blunted by the coaddition of IL-4 (Supplemental Fig. 3J). In the reverse experiment, Ab blockade of IL-4 in IFN-stimulated, monocyte-derived macrophages enhanced the mRNA and protein expression of CCR2, whereas the levels of M2 marker Ccl24 was concomitantly inhibited (Fig. 5H, Supplemental Fig. 3K). These results therefore support that IL-4 can also antagonize some selective IFN-I–dependent effect(s) such as induction of CCR2 expression. Taken together, our results suggest that under the context of mixed IFN-I and IL-4 cytokine cues, antagonisms between IFN-I and IL-4 on certain regulatory programs may contribute to shaping the balance between M2- and M1-skewed macrophage subsets (Fig. 5I).
To examine whether IFN-I signaling could have a similar “diversifying” effect on TAMs, we sorted out YFP+ and YFP– TAMs (CD45+CD11b+Ly6G–F4/80+) from poly(I:C)-treated mice (Supplemental Fig. 3L). Indeed, we found that consistent with the Arg1 expression pattern, the YFP+ TAMs generally express higher M2 markers (e.g., Ccl24, Arg2), whereas the YFP– TAMs apparently feature more abundant M1 markers (e.g., H2-Aa, Cxcl10, and Ccr2). One interesting exception is at a classical M1 marker Nos2, which showed enrichment in YFP+ cells. The presence of certain “mixed” activation profiles in these YFP+ or YFP– TAMs is indeed consistent with the macrophage phenotypic spectrum model (21–23). Collectively, these results demonstrate that poly(I:C)/IFN can also drive “diversification” of TAMs into M2- and M1-skewed subsets.
IL-4RA deficiency limits poly(I:C) treatment–induced M2-skewed TAMs, leading to enhanced antitumor effects
To establish the role of IFN–IL-4 cytokine axis in driving M2-skewed TAMs in vivo, we applied poly(I:C) treatment on WT or IL-4RA–deficient tumor-bearing mice. We noted that similar to the results in IFN-stimulated transitional monocytes in vitro (see (Fig. 3G), poly(I:C) treatment in vivo also led to substantial upregulation of IL-4RA expression in tumor-associated monocytes and macrophages from the WT mice (Supplemental Fig. 4A). The lack of IL-4RA surface staining in cells from the mutant mice was validated in parallel analyses (Supplemental Fig. 4A). In the mock-treated group, IL-4RA deficiency did not affect the overall compositions of tumor-associated myeloid compartments (Fig. 6A and Supplemental Fig. 4B). However, it was noted that tumors from poly(I:C)-treated Il4ra−/− mice featured a moderate decrease and increase in macrophages and granulocytes, respectively, with otherwise minimal changes in the monocytes (Fig. 6A and Supplemental Fig. 4B). Importantly, poly(I:C) treatment–induced expression of ARG1 in tumors was largely abrogated by IL-4RA deficiency, consistent with the role of IFN–IL-4 cytokine axis in driving M2-skewed TAMs (Fig. 6B).
Although not apparently affecting tumor growth on its own, IL-4RA deficiency in mice enhanced poly(I:C)-mediated antitumor effects (Fig. 6C). Because phenotypic alterations of TAMs may play a role in shaping T cell–mediated antitumor immunity (13, 26), the changes in tumor T cell compartments were analyzed in poly(I:C)-treated WT and Il4ra−/− mice (Supplemental Fig. 4C). Poly(I:C) led to a strong decrease in CD4+ T cell population in tumors, which appeared to be further reduced in the treated Il4ra−/− group. However, the numbers for intratumoral CD8+ cytotoxic T cells were not apparently changed by either poly(I:C) or the IL-4RA status (Fig. 6D). Remarkably, when the activation status of tumor-associated CD8+ T cells were analyzed by intracellular staining of IFN-γ, we found that IL-4RA deficiency significantly elevated the numbers of IFN-γ+ CD8+ T cells under poly(I:C)-treated condition (Fig. 6E). Collectively, these results demonstrate that specific inactivation of IL-4 signaling blunts IFN-I’s undesirable effect on instigation of M2-skewed TAMs, and subsequently enhances antitumor immunity.
The activation of an immune response needs to be closely coordinated with immune-suppressive and prorepair mechanisms for an overall organismal fitness (56). Macrophages represent a critical player involved in such coordination due to their potent capabilities to drive both the activating and tolerant arms of immunity (20, 22, 57). However, in certain disease states best exemplified by cancers, the tolerant arm of functions in tumor-associated monocytes and TAMs often show dominance and promote tumor progression (13, 17). This may also contribute to the tumors’ adaptive resistance to various treatments (13, 25). By exploring an innate immunity/IFN-I–based tumor treatment model, we have recently identified an IFN-to-arginase immunosuppressive axis in tumor-associated monocytes/macrophages that thwarts the treatment effects (27). This ensuing work aimed to seek further mechanistic insights. Aided by an Arg1-YFP reporter mice and related in vitro model, we first associated the IFN–ARG1 axis with a secondary, paracrine activity by IFN-stimulated monocytes (Fig. 1, 2). Additional investigations led to the unexpected identification of a monocyte-dependent IFN–IL-4 cytokine axis, which is responsible for driving an M2-skewed subset in the coexisting macrophages (Figs. 3, 4, 5). The functional importance of the IFN–IL-4 cytokine axis is subsequently demonstrated by the enhanced poly(I:C)/IFN-I treatment effects on tumors in IL-4RA–deficient mice (Fig. 6). Overall, these results have suggested some previously unrecognized mechanisms for IFN-I–mediated immune regulation.
Dynamic macrophage phenotype conversion from an M1-type, proinflammatory to an M2-like, prorepair form is commonly observed upon immune stimulation in vivo (22, 57). Nevertheless, increasing diversities among prorepair macrophages and their defining signals other than the classical M2 cytokines have been characterized (21, 57). Our results therefore provide a remarkable scenario where monocytes directly convert the M1-type IFN-I signal into M2 cytokine IL-4. These results point to a critical instructive role by monocytes in coordination of IFN-I–associated inflammation and repair, especially considering that recent works have demonstrated the monocytes as a major IFN-responsive cell type in vivo (58–60). Because of the dominant role of IFN-I in antiviral response, it is tempting to speculate that the observed connection between IFN-I–stimulated monocytes and M2-skewed macrophage phenotype conversion may serve the adaptive purposes to instruct tissue repair responses during virus-induced inflammation. It is of note that a previous study has shown IL-4/IL-13 induction in RSV-infected murine macrophages to control the extent of lung injury (61). Our study suggests that monocyte-derived IL-4 may also play a role in such a viral infection model.
The present study further reveals that the IFN–IL-4 cytokine axis can be engaged in Ly6C+ monocytes within several different anatomical locations (BM, spleen, lung), as well as in those associated with LLC tumors (see (Figs. 3, 4). In contrast, no such responses are observed in mature BMDMs, or TAMs. Because of the mobile nature of monocytes and pleotropic activities by IL-4, IFN-induced monocytic IL-4 may potentially play a broader immune-regulatory role dependent on the inflammatory and anatomical contexts. Interestingly, the IFN–IL-4 cytokine axis can also be engaged in alveolar macrophages, suggesting that the pathway is subjected to regulation by differentiation and/or environmental cues. Given that monocytes do not generally serve as the major cellular source of IL-4 (40), future explorations of the mechanisms underlying the IFN–IL-4 axis are highly warranted. Intriguingly, our results suggest that the IFN–IL-4 axis requires new protein synthesis, therefore operating with a mechanism apparently different from those for classical ISGs (Fig. 3L). A scenario where an IFN-I–induced gene product(s) acts to subsequently drive Il4 induction appears plausible.
Despite the general engagement of IFN–IL-4 cytokine axis in monocytes, data from the current model demonstrate that a subsequent M2-skewed phenotypic shift occurred in LLC TAMs, but not in spleen- or lung-resident macrophages (see (Fig. 1A and Supplemental Fig. 1E). Such tumor-restricted response is likely to be shaped by the intratumoral abundance of M-CSF that is known to promote macrophage M2-skewing (62). From a broader perspective, our results highlight the importance of tissue contexts and possibly the ontogeny of macrophages in shaping their behaviors. Therefore, the results from the s.c. LLC model would be suitable to deduce behaviors of monocyte-derived TAMs under a strong influence of M-CSF. In contrast, given the recent elucidation of distinct origins and functional properties for TAMs in orthotopic lung tumors (48), future works exploring such more relevant models for IFN-I–based therapies are highly warranted.
One interesting observation revealed by application of the Arg1-YFP reporter is the additional heterogeneity in monocyte-derived macrophages under even the same influence by IFN–IL-4 cytokine axis (Fig. 2B, 2C). It is unlikely that such heterogeneity within macrophages arise from differentiation states, as few marker genes associated with macrophage differentiation showed significant variations between the YFP– and YFP+ subsets (Fig. 5C, 5E). Using ARG1 and CCR2 as respective M2- and M1-type macrophage markers, our data suggest that certain antagonizing programs between IFN-I and IL-4 contribute to the heterogeneity in responding macrophages (Fig. 2D, 2E and (Fig. 5H). Such a program of cytokine antagonism is cell type-specific. For instance, although IL-4 was shown to inhibit IFN-induced IL-6 in dendritic cells (63), Il6 mRNA induction in IFN-treated transitional monocytes was not altered by IL-4 blockade (Supplemental Fig. 3K). We speculate that such macrophage-specific, “conflicting signal”–based cellular heterogeneity represent an important mechanism to balance IFN-I–associated inflammation and tissue repair.
As critical players mediating interactions between innate immune system and tumors, IFN-Is serve as a convergence point of host responses under the contexts ranging from antitumor surveillance to multiple therapeutic interventions (7, 10). Our results suggest the need for combinatorial targeting to harness IFN’s action on tumors. Future examinations of how our findings may translate to dynamics of human TAMs would have both basic and translational implications.
We thank members of J.L.’s laboratory for stimulating discussions. We thank GemPharmatech Inc. for helpful mouse services.
This work is supported by grants from the National Key Research and Development Program of China (2019YFA0802802), the National Natural Science Foundation of China (31771574), and the Natural Science Foundation of Jiangsu Province.
The RNAseq data presented in this article have been submitted to the Gene Expression Omnibus under accession number GSE161428.
The online version of this article contains supplemental material.
Abbreviations used in this article
The authors have no financial conflicts of interest.