Targeting monoamine oxidase A-regulated TAM polarization for cancer immunotherapy


 Targeting tumour-associated macrophages (TAMs) is a promising strategy to modify the immunosuppressive tumour microenvironment and improve cancer immunotherapy. Monoamine oxidase A (MAO-A) is an enzyme best known for its function in the brain; small molecule MAO inhibitors (MAOIs) are clinically used for treating neurological disorders. Here we observed MAO-A induction in mouse and human TAMs. MAO-A-deficient mice exhibited decreased TAM immunosuppressive functions corresponding with enhanced antitumour immunity. MAOI treatment induced TAM reprogramming and suppressed tumour growth in preclinical mouse syngeneic and human xenograft tumour models. Combining MAOI and anti-PD-1 treatments resulted in synergistic tumour suppression. Clinical data correlation studies associated high intratumoural MAOA expression with poor patient survival in a broad range of cancers. We further demonstrated that MAO-A promotes TAM immunosuppressive polarization via upregulating oxidative stress. Together, these data identify MAO-A as a critical regulator of TAMs and support repurposing MAOIs for TAM reprogramming to improve cancer immunotherapy.


Introduction
Over the past decade, cancer immunotherapy has achieved signi cant breakthroughs. In particular, immune checkpoint blockade (ICB) therapy has yielded remarkable clinical responses and revolutionized the treatment of many cancers 1 . So far, the FDA has approved cytotoxic T-lymphocyte antigen 4 (CTLA-4) and programmed cell death protein 1/ligand 1 (PD-1/PD-L1) blockade therapies for treating more than 10 different malignancies 2, 3 ; however, only a small fraction of cancer patients respond to these therapies 4,5,6 . Most ICB therapies work through enhancing antitumour CD8 + T cell responses, which can be greatly limited by the immunosuppressive tumour microenvironment (TME) 7 . Tumour-associated macrophages (TAMs), a key component of the immunosuppressive TME, dampen T cell antitumour reactivity in the majority of solid tumours 8,9,10,11,12,13,14,15 . Growing evidence suggests that TAMs are responsible for inhibiting antitumour T cell reactivity and limiting the ICB therapy e cacy, making TAMs potential targets for reversing the immunosuppressive TME and improving cancer immunotherapy 16,17,18 .
TAMs mature from bone marrow-derived circulating monocytes. These monocytes are recruited to the tumour sites, exposed to chemokines and growth factors in the TME, and subsequently differentiate into TAMs 19,20,21,22 . Depending on the surrounding immune environment, macrophages can be polarized toward an immunostimulatory phenotype by pro-in ammatory stimuli (e.g., IFN-γ) or toward an immunosuppressive phenotype by anti-in ammatory stimuli (e.g., IL-4 and IL-13) 23 . Although a binary polarization system is commonly used in macrophage studies, in most large-scale transcriptome analyses, TAMs showed a continuum of phenotypes expressing both immunostimulatory and immunosuppressive markers in addition to the extreme ends of polarization 23,24,25 . These mixed phenotypes and polarization states suggest the complexity of the TME and the residential TAM functionality. As a tumour develops, the enrichment of IL-4 and IL-13 produced by tumour cells and CD4 + T cells in the TME results in the polarization of TAMs towards an immunosuppressive phenotype, that promotes tumour growth, malignancy, and metastasis 23,26,27,28,29,30 . In established solid tumours, TAMs predominately exhibit an immunosuppressive phenotype, evidenced by their production of antiin ammatory cytokines and arginase-1 (Arg1), as well as their expression of mannose receptor (CD206) and scavenger receptors 31,32,33 . Through metabolizing L-arginine via Arg1, TAMs can directly suppress cytotoxic CD8 + T cell responses 34,35 . Mannose receptor (CD206) expressed by TAMs can impair cytotoxicity of CD8 + T cells by suppressing CD45 phosphatase activity 36 . In addition, TAMs can inhibit T cell activities through immune checkpoint engagement by expressing the ligands of the inhibitory receptors PD-1 and CTLA-4. For example, PD-L1 and PD-L2 expressed on TAMs interact with PD-1 of T cells to directly inhibit TCR signaling, cytotoxic function, and proliferation of CD8 + T cells 31 . These characteristics of TAMs make them potential targets for reversing the immunosuppressive TME to augment antitumour immunity.
Although the predominant phenotype of TAMs in established solid tumours is immunosuppressive, polarization is not xed. Plasticity, one of the key features of TAMs, enables TAMs to change their phenotype in solid tumours and thereby providing a therapeutic window 37, 38. Repolarizing/reprogramming TAMs from an immunosuppressive and tumour-promoting phenotype toward an immunostimulatory and tumouricidal phenotype has thus become an attractive strategy in immunotherapy 27 . Preclinical and clinical studies are ongoing, evaluating TAM-repolarizing reagents (e.g., CD40 agonists, HDAC inhibitors, PI3Kγ inhibitors, creatine, etc.) for improving ICB therapy; certain e cacies have been reported 17,29,31,39,40,41,42 . Therefore, the search for new molecules regulating TAM polarization and the development of new combination treatments targeting TAM reprogramming are an active direction of current cancer immunotherapy studies.
Monoamine oxidase A (MAO-A) is an outer mitochondrial membrane-bound enzyme encoded by the Xlinked MAOA gene. MAO-A is best known for its function in the brain, where it is involved in the degradation of a variety of monoamine neurotransmitters, including serotonin, dopamine, epinephrine, and norepinephrine. Through regulating the availability of serotonin, MAOA modulate neuronal activities thereby in uencing mood and behavior in humans 43,44,45,46,47 . Through regulating the availability of dopamine and the abundance of dopamine breakdown byproduct hydrogen peroxide (H 2 O 2 ; hence oxidative stress), MAO-A is involved in multiple neurodegenerative diseases, including Parkinson's disease (PD) 48,49 . FDA-approved small-molecule MAO inhibitors (MAOIs) are currently available for the treatment of neurological disorders, including depression and PD 47,49,50,51,52,53,54,55 . However, the functions of MAO-A outside of the brain are largely unknown. In this study, we investigated the role of MAO-A in regulating TAM polarization and evaluated the possibility of repurposing MAOIs for reprogramming TAMs and improving cancer immunotherapy, using knockout and transgenic mice, preclinical mouse syngeneic and human xenograft tumour models, as well as human TAM and clinical data correlation studies.

MAO-A-de cient mice show reduced tumour growth associated with altered TAM polarization
In a search for new molecules regulating TAM reprogramming, we inoculated C57BL/6J mice with syngeneic B16-OVA melanoma tumours, isolated TAMs, and assessed TAM gene expression pro les. Monocytes isolated from tumour-free and tumour-bearing mice were included as controls. In addition to changes in classical genes involved in regulating macrophage immune responses, we observed the induction of a Maoa gene in TAMs (Fig. 1a), suggesting that MAO-A may be involved in modulating TAM activities.
To study the role of MAO-A in antitumour immunity in vivo, we used MAO-A-de cient mice that carry a hypomorphic MAO-A mutant 56 . Although a degree of Maoa expression leakage in the brain had been previously reported in these mice 56 , analysis of their immune system showed a nearly complete ablation of MAO-A expression in major lymphoid organs including spleen and bone marrow (BM) ( Supplementary   Fig. 1a). Since we focused on immune cells in this study, we denote these mice as Maoa knockout (KO) mice. When challenged with B16-OVA melanoma cells (Fig. 1b), tumour growth in Maoa KO mice was signi cantly suppressed compared to that in Maoa wildtype (WT) mice (Fig. 1c,d). Although similar levels of TAMs (gated as CD45.2 + CD11b + Ly6G -Ly6C -/low F4/80 + cells) were detected in Maoa WT and Maoa KO mice ( Supplementary Fig. 1b,c), compared to their WT counterparts, TAMs isolated from Maoa KO mice exhibited a less immunosuppressive phenotype, indicated by their decreased expression of immunosuppressive markers (i.e., CD206; Fig. 1e), and their increased expression of immunostimulatory molecules (i.e., CD69, CD86, and MHC class II I-Ab; Fig. 1f-h). Further analysis showed that TAMs from Maoa KO mice expressed reduced levels of immunosuppression-associated genes (i.e., Mrc1, Chi3l3, and Arg1; Fig. 1i) and increased levels of pro-in ammatory cytokine genes (i.e., Il6, Tnfα, and Ccl2; Fig. 1j).
Corresponding to the altered TAM polarization in Maoa KO mice, tumour-in ltrating CD8 + T cells in these mice showed enhanced activation (i.e., increased production of Granzyme B; Supplementary Fig. 1d). Single-cell RNA sequencing (scRNAseq) analysis was performed on tumour in ltrating immune cells isolated from Maoa WT and Maoa KO mice ( Fig. 1k and Supplementary Fig. 1e,f). UMAP analysis of extracted TAMs showed a reduced immunosuppressive phenotype in Maoa KO mice, with an increased ratio of Mrc1 low Cd86 high cells to Mrc1 high Cd86 low cells ( Fig. 1l and Supplementary Fig. 1g). Gene expression pro le analysis con rmed a reduction of TAMs expressing immunosuppressive genes (i.e., Mrc1 and Chi3I3; Fig. 1m) and an enrichment of TAMs expressing immunostimulatory genes (i.e., Ccl2, Ccl7, Cd86, H2-Aa, and H2-Ab1; Fig. 1n) in Maoa KO mice. These data strongly indicate that MAO-A is involved in regulating TAM polarization thereby modulating antitumour immunity.
MAO-A directly regulates TAM polarization and in uences TAM-associated T cell antitumour reactivity In our Maoa KO mice tumour challenge study, MAO-A de ciency impacted both immune and non-immune cells (Fig. 1b). To determine whether MAO-A directly regulates immune cells, we conducted a BM transfer experiment wherein BM cells harvested from Maoa WT or KO mice were adoptively transferred into BoyJ (CD45.1) WT recipient mice followed by B16-OVA tumour challenge (Fig. 2a). In this experiment, MAO-A de ciency comparison was con ned to immune cells. MAO-A de ciency in immune cells resulted in suppressed tumour growth (Fig. 2b,c), altered TAM polarization (i.e., downregulation of immunosuppressive markers such as CD206, Fig. 2d; and upregulation of immunostimulatory markers such as CD69, CD86, and MHC class II I-Ab; Fig. 2e,f and Supplementary Fig. 2a), and enhanced tumourin ltrating CD8 + T cell activation (i.e., increased production of cytotoxic molecules such as Granzyme B; Supplementary Fig. 2b), indicating that MAO-A directly regulates immune cell antitumour activity, in particular TAM polarization and T cell antitumour reactivity.
To further study whether MAO-A acts as a macrophage autonomous factor directly regulating TAM polarization and thereby in uencing antitumour immunity, we performed a macrophage adoptive transfer tumour experiment. BM cells were harvested from Maoa WT and KO mice then cultured into bone marrow-derived macrophages (BMDMs). These Maoa WT or KO BMDMs were then mixed with B16-OVA melanoma cells and subcutaneously (s.c.) injected into BoyJ WT recipient mice to establish solid tumours (Fig. 2g). In this study, MAO-A de ciency comparison was con ned to TAMs. Suppressed tumour growth (Fig. 2h,i), downregulated expression of TAM immunosuppressive markers (i.e., CD206; Fig. 2j), upregulated expression of TAM immunostimulatory markers (i.e., CD69 and CD86; Fig. 2k,l), and enhanced tumour-in ltrating CD8 + T cell reactivity (i.e., increased production of Granzyme B; Fig. 2m) were observed in mice receiving Maoa KO BMDMs. Collectively, these in vivo studies demonstrate that MAO-A acts as an autonomous factor directly regulating TAM polarization, and thereby in uencing T cell antitumour reactivity and impacting tumour growth.
To verify whether MAO-A de ciency directly contributed to the alleviated immunosuppressive polarization of Maoa KO macrophages, we performed a rescue experiment. We constructed a MIG-Maoa retroviral vector, used this vector to transduce Maoa KO BMDMs, and achieved overexpression of MAO-A in these macrophages ( Fig. 3l-n, and Supplementary Fig. 3b). MAO-A overexpression signi cantly exacerbated the immunosuppressive phenotype of IL-4/IL-13-stimulated Maoa KO BMDMs (i.e., upregulation of immunosuppressive signature genes such as Chi3l3 and Arg1; Fig. 3o,p). Taken together, these results indicate that MAO-A acts as an autonomous factor promoting macrophage immunosuppressive polarization under anti-in ammatory stimuli.

MAO-A promotes macrophage immunosuppressive polarization via ROS upregulation
Next, we sought to investigate the molecular mechanisms regulating MAO-A promotion of macrophage immunosuppressive polarization. It has been reported that intracellular reactive oxygen species (ROS; hence, oxidative stress) elicit macrophage immunosuppressive features 57, 58, 59, 60, 61. MAO-A catalyzes the oxidative deamination of monoamines, thereby generating hydrogen peroxide (H 2 O 2 ) as a byproduct that can increase intracellular ROS levels. We therefore speculated that MAO-A might promote TAM immunosuppressive polarization in TME via upregulating ROS levels in TAMs (Fig. 4a).
To test this hypothesis, we directly measured ROS levels in TAMs isolated from Maoa WT and KO mice bearing B16-OVA tumours and detected signi cantly lower levels of ROS in Maoa KO TAMs (Fig. 4b,c). Measurement of ROS levels in in vitro-cultured Maoa WT and KO BMDMs also showed reduced levels of ROS in Maoa KO BMDMs, with or without IL-4/IL-13 stimulation, in agreement with the in vivo TAM results (Fig. 4d). Supplementing H 2 O 2 to IL-4/IL-13-stimulated Maoa WT and KO BMDMs elevated their intracellular ROS to similar levels ( Supplementary Fig. 4a,b) and eliminated their differences in expression of immunosuppressive markers (i.e., CD206; Fig. 4e) and signature genes (i.e., Chi3l3 and Arg1; Fig. 4f,g).
On the other hand, supplementation of tyramine, a substrate of MAO-A, increased ROS levels and upregulated the expression of immunosuppressive genes (i.e., Chi3l3 and Arg1) in Maoa WT BMDMs but not in Maoa KO BMDMs ( Fig. 4h-j). Taken together, these data indicate that MAO-A regulates macrophage immunosuppressive polarization via modulating macrophage intracellular ROS levels.
Collectively, these in vivo and in vitro data support a working model that MAO-A promotes TAM immunosuppressive polarization in TME, at least partly through upregulating TAM intracellular ROS levels and thereby enhancing the IL-4/IL-13-induced JAK-Stat6 signaling pathway.

MAO-A blockade for cancer immunotherapy-syngeneic mouse tumour model studies
The identi cation of MAO-A as a key regulator of TAM immunosuppressive polarization makes MAO-A a promising new drug target for cancer immunotherapy. Because of the known functions of MAO-A in the brain, small molecule MAOIs have been developed and clinically utilized for treating various neurological disorders, making it a highly feasible and attractive approach to repurpose these established MAOI drugs for cancer immunotherapy 51, 72 . In an in vitro WT BMDM IL-4/IL-13-induced polarization culture (Fig. 5a), addition of multiple MAOIs e ciently reduced ROS levels in BMDMs (Fig. 5b) and suppressed their immunosuppressive polarization, evidenced by their decreased expression of immunosuppressive markers (i.e., CD206; Fig. 5c) and immunosuppressive genes (i.e., Chi3l3 and Arg1; Fig. 5d,e). Notably, the MAOIs that we tested include phenelzine, clorgyline, mocolobemide, and pirlindole, covering the major categories of established MAOIs classi ed on the basis of whether they are nonselective or selective for MAO-A, and whether their effect is reversible (Fig. 5a) 51, 54, 73 . Among these MAOIs, phenelzine (trade name: Nardil) is clinically available in the United States 72 . In the following studies, we chose phenelzine as a representative to study the possibility of repurposing MAOIs for cancer immunotherapy, using two syngeneic mouse tumour models: a B16-OVA melanoma model and a MC38 colon cancer model 74 .
First, we studied the therapeutic potential of phenelzine in a B16-OVA tumour prevention model (Fig. 5f). Phenelzine treatment effectively suppressed B16-OVA tumour growth in B6 wildtype mice (Fig. 5g,h). No tumour growth difference was observed when we depleted TAMs in experimental mice via a clodronate liposome treatment, indicating that phenelzine suppressed tumour growth via modulating TAMs (Fig.  5g,h and Supplementary Fig. 5a). Correspondingly, TAMs isolated from phenelzine-treated mice displayed a less immunosuppressive phenotype (i.e., decreased expression of CD206; Fig. 5i) that was correlated with an enhanced antitumour reactivity of tumour-in ltrating CD8 + T cells (i.e., increased production of Granzyme B; Fig. 5j) in these mice. Further studies showed that phenelzine treatment also effectively suppressed the progression of pre-established solid tumours in both B16-OVA and MC38 models (Supplementary Fig. 5b-f).
Next, we evaluated the potential of phenelzine for combination therapy, in particular combining with other ICB therapies such as PD-1/PD-L1 blockade therapy (Fig. 5k). Although most ICB therapies target CD8 + T cells, these cells are in fact closely regulated by TAMs in the TME, making targeting TAMs another potential avenue for immunotherapy 14,39 . In both B16-OVA and MC38 tumour models, phenelzine treatment signi cantly suppressed the progression of pre-established solid tumours at a level comparable to the anti-PD-1 treatment; importantly, the combination of phenelzine and anti-PD-1 treatments yielded synergistic tumour suppression e cacy ( Fig. 5l-o). These tumour suppression effects of phenelzine were due to immunomodulation but not direct tumour inhibition, because phenelzine treatment did not suppress the growth of B16-OVA and MC38 tumours in immunode cient NSG mice (Supplementary Fig.  5g-k).
Collectively, these syngeneic mouse tumour model studies provided proof-of-principle evidence for the cancer immunotherapy potential of MAOIs via targeting TAM reprogramming and thereby enhancing antitumour T cell responses.

MAO-A blockade for cancer immunotherapy-human TAM and clinical data correlation studies
To explore the translational potential of MAO-A blockade therapy, we rst studied MAO-A regulation of human macrophage polarization. Using a Tumour Immune Dysfunction and Exclusion (TIDE) computational method 75  Collectively, these in vitro data suggest that MAO-A is highly expressed in human macrophagesespecially during their immunosuppressive polarization, and that MAO-A blockade has the potential to reprogram human macrophage polarization.
To directly evaluate whether MAO-A blockade could reprogram human TAM polarization in vivo, we established a human tumour/TAM xenograft NSG mouse model. A375 human melanoma cells were mixed with monocytes sorted from healthy donor peripheral blood mononuclear cells (PBMCs), and s.c. injected into NSG mice to form solid tumours, with or without phenelzine treatment after inoculation (Fig.  6h). Phenelzine treatment effectively suppressed immunosuppressive polarization of human TAMs (gated as hCD45 + hCD11b + hCD14 + ; Supplementary Fig. 6b), supported by their decreased expression of immunosuppressive markers (i.e., CD206 and CD273; Fig. 6i,j).
Next, we studied whether MAO-A blockade-induced human TAM reprogramming could impact human T cell antitumour reactivity, using a 3D human tumour/TAM/T cell organoid culture (Fig. 6k). NY-ESO-1, a well-recognized tumour antigen commonly expressed in a large variety of human cancers 77 , was chosen as the model tumour antigen. An A375 human melanoma cell line was engineered to co-express NY-ESO-1 as well as its matching MHC molecule, HLA-A2, to serve as the human tumour target (denoted as A375-A2-ESO; Supplementary Fig. 6c,d). NY-ESO-1-speci c human CD8 + T cells were generated by transducing healthy donor peripheral blood CD8 + T cells with a Retro/ESO-TCR retroviral vector encoding an NY-ESO-1 speci c TCR (clone 3A1; denoted as ESO-TCR); the resulting T cells, denoted as ESO-T cells, expressed ESO-TCRs and speci cally targeted A375-A2-ESO tumour cells, thereby modeling the tumour-speci c human CD8 + T cells ( Supplementary Fig. 6e,f). Human MDMs were cultured from healthy donor PBMCs, followed by IL-4/IL-13 stimulation to induce immunosuppressive polarization in the presence or absence of phenelzine treatment (Fig. 6k). The A375-A2-ESO human melanoma cells,ESO-T cells, and IL-4/IL-13polarized MDMs were mixed at a 2:2:1 ratio and placed in a 3D tumour organoid culture mimicking TME (Fig. 6k). IL-4/IL-13-polarized MDMs effectively suppressed ESO-T cell-mediated killing of A375-A2-ESO tumour cells; this immunosuppressive effect was largely alleviated by phenelzine treatment during MDM polarization (Fig. 6l). Accordingly, ESO-T cells co-cultured with phenelzine-treated MDMs, compared to those co-cultured with non-phenelzine-treated MDMs, showed an enhancement in T cell activation (i.e., increased cell number, increased CD25 expression, and decreased CD62L expression; Fig. 6m and Supplementary Fig. 6g). Collectively, these data suggest that MAO-A blockade-induced human TAM reprogramming has the potential to improve antitumour T cell responses.
To study MAOA gene expression in primary human TAMs, we collected fresh ovarian cancer tumour samples from patients, isolated TAMs (sorted as DAPI -hCD45 + hCD11b + hTCRαβ -hCD14 + cells; Supplementary Fig. 6h), and assessed their MAOA gene expression. Primary human monocytes isolated from health donor PBMCs (sorted as DAPI -hCD45 + hCD11b + hTCRαβ -hCD14 + cells; Supplementary Fig. 6i) were included as controls. Like mouse TAMs, human TAMs expressed high levels of MAOA gene, con rming MAO-A as a valid drug target in human TAMs ( Fig. 1a and Fig. 6n).
Lastly, we conducted clinical data correlation studies to investigate whether intratumoural MAOA gene expression is correlated with clinical outcomes in cancer patients, using the TIDE computational method Intratumoural MAOA expression level was negatively correlated with patient survival in multiple cancer patient cohorts spanning ovarian cancer (Fig. 6o) 78 , lymphoma (Fig. 6p) 79 , and breast cancer (Fig. 6q) 80 . Moreover, analysis of a melanoma patient cohort receiving anti-PD-1 treatment showed that high levels of intratumoural MAOA expression largely abrogated the survival bene t offered by the PD-1 treatment, suggesting that combining MAO-A blockade therapy with PD-1/PD-L1 blockade therapy may provide synergistic therapeutic bene ts through modulating TAM polarization and thereby changing the immunosuppressive TME and improving antitumour immunity (Fig. 6r) 81 . Of note, these whole tumour lysate transcriptome data analyses could not localize the MAOA expression to a speci c cell type (e.g., TAMs); future studies of quality transcriptome data generated from single cells or sorted TAMs are needed to obtain such information. Nonetheless, the present clinical data correlation studies identi ed MAO-A as a possible negative regulator of survival in a broad range of cancer patients, including those receiving existing ICB therapies, suggesting MAO-A blockade as a promising avenue for developing new forms cancer therapy and combination therapy.
Taken together, these human TAM and clinical correlation studies con rmed MAO-A as a promising drug target in human TAMs and support the translational potential of MAO-A blockade for cancer immunotherapy through targeting TAM reprogramming.

Discussion
Based on our ndings, we propose an "intratumoural MAO-A-ROS axis" model to elucidate the role of MAO-A in regulating TAM immunosuppressive polarization ( Supplementary Fig. 7). Analogous to the wellcharacterized MAO-A-ROS axis in the brain, where MAO-A controls ROS levels in neurons and thereby modulates neuron degeneration via regulating neuron oxidative stress, the MAO-A-ROS axis in a solid tumour controls ROS levels in TAMs and thereby modulates TAM immunosuppressive polarization via sensitizing the IL-4/IL-13-induced JAK-Stat6 signaling pathway ( Supplementary Fig. 7). The resemblance between these mechanisms is intriguing: from an evolutionary point of view, it makes sense that some critical molecular regulatory pathways are preserved between the nervous and immune systems, considering that both systems are evolved to defend a living organism by sensing and reacting to environmental danger and stress. Indeed, neurons and immune cells share a broad collection of surface receptors, secretory molecules, and signal transducers 82 . In particular, many neurotransmitters/neuropeptides and their synthesis/degradation machineries traditionally considered speci c for neurons are expressed in immune cells, although their functions in the immune system are to a large extent still unknown 83,84 . Studying these molecules and their regulatory mechanisms may provide new perspectives in tumour immunology and identifying new drug targets for cancer immunotherapies, as exempli ed by our current nding of this "MAO-A-ROS axis" regulation of TAM polarization in the TME.
Considering the importance of TAMs in regulating antitumour immunity, there has been considerable efforts in developing cancer therapeutic strategies targeting TAMs. These strategies can be roughly divided into two categories: 1) those which deplete TAMs, and 2) those which alter TAM immunosuppressive activities 39 . The rst category includes strategies targeting TAM recruitment and survival, such as blocking the CCL2-CCR2 axis thereby preventing monocyte mobilization from the bone marrow and recruitment into in ammatory sites, or blocking the CSF1-CSF1R axis thereby inducing apoptosis of TAMs, or blocking the CXCL12-CXCR4 and angiopoietin 2 (ANG2)-TIE2 axes thereby depleting TIE2 + macrophages that are critical for tumour angiogenesis 19,39,85 . However, an intrinsic downside of depleting TAMs is the loss of their innate immunostimulatory role as the primary phagocytes and professional antigen-presenting cells (APCs) in solid tumours. Reprogramming or repolarizing immunosuppressive TAMs towards an immunostimulatory phenotype therefore can be an attractive direction; this second category of TAM-repolarizing strategies includes those reprogramming TAMs via CD40 agonists, HDAC inhibitors, PI3Kg inhibitors, and creatine 39,40,86,87,88 . Many of these TAM reprogramming strategies are currently under active clinical evaluation 39 . Notably, CD40 agonists work through activating CD40L-downstream NF-kB pathway 87,89 ; HDAC inhibitors work through altering histone modi cations 86,90,91 ; PI3Kγ inhibitors work through stimulating NF-κB activation while inhibiting C/EBPβ activation 88,92,93 ; and creatine uptake works through regulating cytokine responses 40 . Our discovery of MAO-A as a critical regulator of TAM polarization through modulating oxidative stress provides a new drug target and a new mechanism of action (MOA) for expanding TAM-repolarizing strategies.
Compared to many new therapeutic candidates, MAO-A is unique in that it is already an established drug target due to its known functions in the brain 72 . In fact, small molecule MAOIs have been developed to block MAO-A enzymatic activity in the brain and are clinically used for treating various neurological disorders 72 . Notably, some MAOIs cross-inhibit the MAO-A isoenzyme MAO-B, that co-expresses with MAO-A in the brain ( Supplementary Fig. 7) 51 . However, in human macrophages, especially in M2-like immunosuppressive macrophages, MAO-A is the dominant form (i.e., the expression of MAOA was about 40-fold higher than that of MAOB in M2-like human macrophages; Supplementary Fig. 8) 76 . In our studies, we tested multiple clinically approved MAOIs (phenelzine, clorgyline, moclobemide, and pirlindole) and demonstrated their e cacy in regulating macrophage ROS levels and immunosuppressive polarization, pointing to the possibility of repurposing these drugs for cancer immunotherapy (Fig. 5 and  6). Developing new cancer drugs is extremely costly and time-consuming; drug repurposing offers an economic and speedy pathway to novel cancer therapies because approved drugs have known safety pro les and modes of actions and thus can enter the clinic quickly 94 .
MAOIs had been used extensively over two decades after their introduction in the 1950s, but since then their use has declined because of reported side effects and the introduction of other classes of antidepressant drugs 72 . However, these MAOIs side effects were vastly overstated and should be revisited 72 . For instance, a claimed major side effect of MAOIs is the risk of triggering tyramine-induced hypertensive crisis when patients eat tyramine-rich foods such as aged cheese (hence, "cheese effects"); this concern led to cumbersome food restrictions that are now considered largely unnecessary 72 . A transdermal delivery system (Emsam) has also been developed to deliver MAOIs that can largely avoid potential food restrictions 95 . Therefore, interest in MAOIs as a major class of antidepressants is reviving, and repurposing MAOIs for cancer immunotherapy can be an attractive new application of these potent drugs 72 . Moreover, many cancer patients suffer from depression and anxiety; these overwhelming emotional changes can negatively interfere with the quality of life and cancer treatment e cacy of cancer patients 96 . Repurposing MAOIs for cancer immunotherapy thus may provide cancer patients with anti-depression and antitumour dual bene ts, making this therapeutic strategy particularly attractive.
Because preclinical evidence largely supports combinatorial approaches being necessary to achieve signi cant antitumour e cacy, most TAM-targeting strategies currently under clinical evaluation are tested in combination with standard chemotherapy or radiation therapy, or in combination with T cell-directed ICB therapies such as PD-1 or/and PD-L1 blockade therapy 39 . In our study, we found that MAOI treatment synergized with anti-PD-1 treatment in suppressing syngeneic mouse tumour growth (Fig. 5ko), and that intratumoural MAOA gene expression levels dictated poor patient survival in melanoma patients receiving anti-PD-1 therapy (Fig. 6r). These data highlight the promise of MAOI treatment as a valuable component for combination cancer therapies.
In summary, here we identi ed MAO-A as a critical molecule regulating TAM immunosuppressive polarization and thereby modulating antitumour immunity, and demonstrated the potential of and BMDMs (5 x 10 6 cells per mouse) were mixed and s.c. injected into BoyJ mice to form solid tumours.
Tumour growth was monitored twice per week by measuring tumour size using a Fisherbrand TM Traceable TM digital caliper; tumour volumes were calculated by formula 1/2 x L x W 2 .At the end of an experiment, tumours were collected and tumour-in ltrating immune cells were isolated for analysis using ow cytometry.
Xenograft human tumour-TAM co-inoculation model

Tumour-in ltrating immune cell (TII) isolation and analysis
Solid tumours were collected from experimental mice at the termination of a tumour experiment. Tumours were cut into small pieces and smashed against a 70-µm cell strainer (Corning, 07-201-431) to prepare single cells. Immune cells were enriched through gradient centrifugation with 45% Percoll (Sigma-Aldrich, P4937) at 800 g for 30 mins at 25 °C without braking, followed by treatment with Tris-buffered ammonium chloride buffer to lyse red blood cells according to a standard protocol (Cold Spring Harbor Protocols). The resulting TII isolates were then used for further analysis.
In some experiments, TII isolates were sorted via FACS using a FACSAria II ow cytometer (BD Biosciences) to purify immune cells (sorted as DAPI -CD45.2 + cells), which were then subjected to scRNASeq analysis of gene expression pro ling of TIIs.

Mouse bone marrow-derived macrophages (BMDM) culture and polarization
To generate BMDMs, BM cells were collected from femurs and tibias of MaoaWT mice and Maoa KO mice, and were cultured in C10 medium containing with 20% of L929-conditional medium in a 10-cm dish Human NY-ESO-1-speci c TCR-engineered CD8+ T (ESO-T) cells The Retro/ESO-TCR vector was constructed by inserting into the parental pMSGV vector a synthetic gene encoding an HLA-A2-restricted, NY-ESO-1 tumour antigen-speci c human CD8 TCR (clone 3A1) 98 . Vsv-gpseudotyped Retro/ESO-TCR retroviruses were generated by transfecting HEK 293T cells following a standard calcium precipitation protocol and an ultracentrifugation concentration protocol 102 ; the viruses were then used to transduce PG13 cells to generate a stable retroviral packaging cell line producing GALV-pseudotyped Retro/ESO-TCR retroviruses (denoted as the PG13-ESO-TCR cell line). For virus production, the PG13-ESO-TCR cells were seeded at a density of 0.8 x 10 6 cells per ml in D10 medium, and cultured in a 15-cm dish (30 ml per dish) for 2 days; virus supernatants were then harvested and stored at -80 °C for future use.
3D human tumour/TAM/T cell organoid culture A375-A2-ESO human melanoma cell line was generated by engineering the parental A375 cell line to overexpress an NY-ESO-1 tumour antigen as well as its matching HLA-A2 molecule 98 .
Human MDMs were generated from healthy donor PBMCs and polarized with IL-4/IL-13 in the presence or absence of phenelzine treatment. ESO-T cells were generated by engineering healthy donor PBMC CD8 + T cells to express an NY-ESO-1-speci c TCR (clone 3A1). The A375-A2-ESO tumour cells, MDMs, and ESO-T cells were mixed at a 2:1:2 ratio. Mixed cells were centrifuged and resuspended in C10 medium at 1 x 10 5 cells per μl medium. The cell slurry was >adjusted to 5 ul per aggregate and was gently transferred onto a microporous membrane cell insert (Millicell, PICM0RG50) using a 20-µl pipet to form a 3D human tumour/TAM/T cell organoid. Prior to cell transfer, cell inserts were placed in a 6-well plate immersed with 1 ml C10 medium. Two days later, the organoids were dissociated by P1000 pipet tip and disrupted through a 70-µm nylon strainer to generate single cell suspensions for further analysis.
In vivo PD-1 blocking antibody (clone RMP1-14) and its isotype control (rat IgG2a) were purchased from BioXCell. In vivo TAM depletion clodronate liposomes and their control vehicle liposomes were purchased from Clodrosome.
Monoamine oxidase inhibitors (MAOIs), including phenelzine, moclobimide, and clorgyline, were purchased from Sigma-Aldrich. Pirlindole was purchased from R&D systems.  To study cell surface marker expression, cells were stained with Fixable Viability Dye followed by Fc blocking and surface marker staining, following a standard procedure as described previously 101 . To study T cell intracellular cytotoxicity molecule production, intracellular staining of Granzyme B was performed using the BD Cyto x/CytopermTM Fixation/Permeabilization Kit (BD Biosciences, 55474) following the manufacturer's instructions. These cells were co-stained with surface markers to identify CD8 + T cells (gated as TCRβ + CD8 + cells in vitro or CD45.2 + TCRβ + CD8 + cells in vivo).. Primary antibodies against mouse Stat6, p-Stat6 (Tyr641), JAK1, p-JAK1 (Tyr1034/1035), JAK2, p-JAK2 (Tyr1008), JAK3, p-JAK3 (Tyr980/981), HRP-labeled anti-rabbit secondary antibodies, and HRP-labeled anti-mouse secondary antibodies were purchased from Cell Signaling Technology. MAO-A antibody was purchased from Abcam (Clone EPR7101). Primary antibodies against -actin (Santa Cruz Biotechnology) were used as an internal control for total protein extracts. Signals were visualized using a ChemiDoc Image System (Bio-Rad). Data were analyzed using an Image J software (Bio-Rad). Table 2. Quantitative real-time PCR (QPCR) Total RNA was extracted from cells using TRIzol reagent (Invitrogen, Thermo Fisher Scienti c, 15596018) following the manufacturer's instructions. SuperScript III First-strand (Thermo Fisher Scienti c, 18080051) was used for reverse transcription. QPCR was performed using a KAPA SYBR FAST qPCR Master Mix (Kapa Biosystems) and a 7500 Real-time PCR System (Applied Biosystems) according to the manufacturers' instructions. Housekeeping gene Ube2d2 was used as an internal control for mouse immune cells and ACTB was used as an internal control for human immune cells. The relative expression of a target gene was calculated using the ΔΔCT method. All primers used in this study are listed in Supplementary Table 3 performed using a Cellranger Software Suite (10X Genomics). BCL les were extracted from the sequencer and used as inputs for the cellranger pipeline to generate the digital expression matrix for each sample. Then cellranger aggr command was used to aggregate the two samples into one digital expression matrix. The matrix was analyzed using Seurat, an R package designed for single cell RNA sequencing. Speci cally, cells were rst ltered to have at least 300 UMIs (unique molecular identi ers), at least 100 genes and at most 50% mitochondrial gene expression; only 1 cell did not pass the lter. The ltered matrix was normalized using the Seurat function NormalizeData. Variable genes were found using the Seurat function FindVariableGenes. The matrix was scaled to regress out the sequencing depth for each cell. Variable genes that had been previously identi ed were used in principle component analysis (PCA) to reduce the dimensions of the data. Following this, 13 PCs were used in UMAP to further reduce the dimensions to 2. The same 13 PCs were also used to group the cells into different clusters by the Seurat function FindClusters. Next, marker genes were found for each cluster and used to de ne the cell types. Subsequently, 2 clusters of TAMs (identi ed by co-expression of Mrc1 and Cd86 signature genes) were extracted and compared between the Maoa WT and Maoa KO samples. Expression distribution of immunosuppressive and immunostimulatory signature genes in Maoa WT and Maoa KO TAMs were compared and presented in violin plots.

Detailed reagent information is provided in Supplementary
Tumour immune dysfunction and exclusion (TIDE) computational method TIDE analyses were conducted as previously described (http://tide.dfci.harvard.edu) 75 . Two functions of the TIDE computational method were used: 1) the prioritization function and 2) the survival correlation function.
The prioritization function of TIDE was used to rank a target gene by its immune dysfunction/risk score, that for TAMs was calculated as its gene expression log-fold change of M2-like/M1-like MDMs 75 . A transcriptome data set (GSE35449) was used, which was generated by microarray analysis of the gene expression pro ling of in vitro polarized M1-like or M2-like human MDMs 76 . A score higher than 1 indicates the preferential expression of a gene in M2-like compared to M1-like human macrophages. The higher a score is, the more "prioritized" a gene is in relating to TAM immunosuppressive polarization.
The survival correlation function of TIDE was used to study the clinical data correlation between the intratumoural MAOA gene expression and patient survival. Four patient cohorts were analyzed: ovarian cancer (GSE26712) 78 , lymphoma (GSE10846) 79 , breast cancer (GSE9893) 80 , and melanoma (PRJEB23709) 81 . For each patient cohort, tumour samples were divided into two groups: MAOA-high (samples with MAOA expression one standard deviation above the average) and MAOA-low (remaining samples) groups. The association between the intratumoural MAOA gene expression levels and patient overall survival (OS) was computed through the two-sided Wald test in the Cox-PH regression and presented in Kaplan-Meier plots. P value indicates the comparison between the MAOA-low and MAOAhigh groups, and was calculated by two-sided Wald test in a Cox-PH regression.

Statistical analysis
GraphPad Prism 6 (GraphPad Software) was used for the graphic representation and statistical analysis of the data. All data were presented as the mean ± standard error of the mean (SEM). A 2-tailed Student's t test was used for comparison between groups. Multiple comparisons were performed using an ordinary 1-way ANOVA followed by Tukey's multiple comparisons test, or using a 2-way ANOVA followed by Sidak's multiple comparisons test. P < 0.05 was considered statistically signi cant. ns, not signi cant; *P < 0.05; **P < 0.01; ***P < 0.001. For scRNAseq data analysis, Wilcoxon-rank sum test was utilized to determine the P value between two groups.
Benjamini-Hochberg Procedure was used to adjust the P value to reduce the false positive rate.
For the Kaplan-Meier plot of the overall patient survival for ovarian cancer, lymphoma, breast cancer, and melanoma with different MAOA levels, the P value was calculated by two-sided Wald test in a Cox-PH regression.

Reporting Summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Declarations Data availability
All data associated with this study are presented in the article or Supplemental Information. The genomics data generated during this study will be available at the public repository GEO when the manuscript is published. Further information and requests for new reagents generated in this study may be directed to and will be ful lled by the Lead Contact, Lili Yang (liliyang@ucla.edu).   Representative of 3 experiments. All data are presented as the mean ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001, by Student's t test.