Metabolism of immune cells in cancer
Author: Robert D Leone ， Jonathan D Powell
雜誌：Nat Rev Cancer
IF: 53.03
摘要
Through the successes of checkpoint blockade and adoptive cellular therapy, immunotherapy has become an established treatment modality for cancer. Cellular metabolism has emerged as a critical determinant of the viability and function of both cancer cells and immune cells. In order to sustain prodigious anabolic needs, tumours employ a specialized metabolism that differs from untransformed somatic cells. This metabolism leads to a tumour microenvironment that is commonly acidic, hypoxic and/or depleted of critical nutrients required by immune cells. In this context, tumour metabolism itself is a checkpoint that can limit immune-mediated tumour destruction. Because our understanding of immune cell metabolism and cancer metabolism has grown significantly in the past decade, we are on the cusp of being able to unravel the interaction of cancer cell metabolism and immune metabolism in therapeutically meaningful ways. Although there are metabolic processes that are seemingly fundamental to both cancer and responding immune cells, metabolic heterogeneity and plasticity may serve to distinguish the two. As such, understanding the differential metabolic requirements of the diverse cells that comprise an immune response to cancer offers an opportunity to selectively regulate immune cell function. Such a nuanced evaluation of cancer and immune metabolism can uncover metabolic vulnerabilities and therapeutic windows upon which to intervene for enhanced immunotherapy
檢驗點阻斷和過繼細胞療法的成功使得免疫療法已經成為癌症的一種既定治療方式。細胞代謝已成為癌細胞和免疫細胞活力和功能的關鍵決定因素。為了維持巨大的合成代謝需求，腫瘤採用了不同於未轉化體細胞的特殊代謝。這種代謝導致腫瘤微環境通常呈酸性、缺氧和/或耗盡免疫細胞所需的關鍵營養物質。在這種情況下，腫瘤代謝本身是一個檢查點，可以限制免疫介導的腫瘤破壞。因為我們對免疫細胞代謝和癌症代謝的理解在過去十年中顯著增長，我們正處於能夠以治療上有意義的方式解開癌細胞代謝和免疫代謝相互作用的風口浪尖。儘管有代謝過程似乎是癌症和應答免疫細胞的基礎，但代謝異質性和可塑性可能有助於區分兩者。因此，了解組成對癌症免疫應答的不同細胞的不同代謝要求為選擇性調節免疫細胞功能提供了機會。對癌症和免疫代謝的這種細緻入微的評價可以揭示代謝的弱點和幹預增強免疫治療的治療窗
背景
Work over the past several decades has shown that activated immune cells employ many metabolic pathways attributed to cancer cells1–3 (Fig. 1). This convergence of metabolic adaptations creates a fundamental competition for nutrients required by cancer cells and immune cells within the tumour microenvironment (TME). However, we are coming to find fundamental differences between the metabolic programmes of cancer cells and immune cells, as well as between different immune cells. Understanding these differences can reveal specific metabolic vulnerabilities and, consequently, novel targets for therapeutic approaches aimed at metabolic programming in order to enhance cancer immunotherapy. Although the ability of cancer cells and tumour tissue to upregulate glycolytic catabolism of glucose to form lactate, even in oxygen-replete conditions (aerobic glycolysis), a process known as the 『Warburg effect』, has been considered a hallmark of malignancy, it has become increasingly clear that cancer metabolism is heterogeneous, and that cancer cells can engage in a broad range of metabolic programmes to meet the demands of growth and proliferation, and that in addition to aerobic glycolysis, mitochondrial respiration is fundamentally important in this regard4–7 . Predictably, highly metabolically active cancer cells (Fig. 1) impart profound effects on the TME, leading to nutrient depletion, hypoxia, acidity and the generation of metabolites that can be toxic at certain concentrations. A significant amount of glucose from the TME is metabolized through aerobic glycolysis, generating high amounts of lactate and H+, thereby lowering the intratumoural pH. That said, it is likely that the balance between lactate-generating glycolysis and oxidative phosphorylation (OXPHOS) is dependent on the degree of hypoxia, which can be both heterogeneous and wide ranging within the TME. It is instructive to note that in moderately hypoxic regions, CO2 derived from mitochondrial respiration is hydrated by extracellular carbonic anhydrase enzymes, forming HCO3 – and H+. Thus, oxidative metabolism can be a significant and often overlooked source of extracellular acidification within the TMEGiven the recent establishment of cancer immunotherapy, including the use of blocking antibodies against immune checkpoint pathways and adoptive cell therapy with chimeric antigen receptor T cells (CAR T cells), several recent studies have begun to establish the relationship of tumour-intrinsic metabolism to successful immunotherapy. For instance, it has been reported that increased glycolytic metabolism in melanoma cells is associated with resistance to adoptive T cell therapy and checkpoint blockade8,9 . Other studies have shown that signalling through immune checkpoint proteins on tumour cells, including PD1 and B7-H3, was responsible for increased glucose depletion within the TME10–12. Interestingly, some immunosuppressive checkpoint pathways are actually induced as a direct consequence of tumour acidification13. Further, immune checkpoint blockade can dampen glycolysis of tumour cells, restore glucose in the TME and permit T cell glycolysis and cytokine production14. Several recent studies have demonstrated that targeting specific aspects of tumour-intrinsic metabolism, such as the hexosamine biosynthesis pathway (HBP) or glutamine metabolism, could foster an immune response and sensitize tumours to checkpoint blockade15,1 Because of the emergence of immunotherapy as a pillar of oncologic therapy, it is increasingly vital to understand as much as possible about the metabolic interdependence of infiltrating immune cells and cancer. This Review aims to discuss the following fundamental questions: which metabolic programmes are critical for the function of specific cell subsets involved in the immune response to cancer; how these metabolic programmes might be perturbed within the TME; the implications of metabolic derangements in the TME for current immunotherapeutic paradigms; and how metabolic interventions might be leveraged to enhance the antitumour immune response.
The TME and immune contexture
Highly active metabolic pathways that are characteristic of cancer cells (Fig. 1) can create profound changes in the composition of nutrients and other small molecules within the TME. This can have critical effects on the immune response. The high metabolic activity of cancer cells and disorganized vasculature within the TME can contribute to a nutrient-depleted and hypoxic microenvironment, establishing metabolic competition between cancer cells and infiltrating immune cells14,17,18. Indeed, the glucose uptake and effector function of antitumour CD4+ T cells has been shown to be inversely proportional to glycolytic activity of cancer cells in mouse models18, and glucose availability in the TME allows for improved cytokine expression from antitumour CD8+ T cells14. Furthermore, transcriptomic analyses of patients with melanoma from The Cancer Genome Atlas revealed that effector T (Teff) cell genes, such as CD40lg and IFNG, are inversely correlated with HK2 expression, which encodes the rate-limiting enzyme in the glycolytic pathway18. Metabolic programmes active within cells of the TME can also lead to the generation of toxic concentrations of certain metabolites. Elevated levels of adenosine, kynurenine, ornithine, reactive oxygen species (ROS) and potassium, as well as increased acidosis, have all been reported in the TME, and each can have profound effects in suppressing the antitumour immune response. The immune contexture of the TME comprises a range of distinct cell types19 (Table 1). Effector cells perform functions aimed at cell killing and can arise from either the innate (non-specific) or adaptive (antigen-specific) arms of the immune system. Antitumour effector cells arising from the adaptive system include CD4+ and CD8+ Teff cells, which orchestrate and carry out antigen-specific killing of cancer cells, respectively. CD8+ Teff cells are critically important in direct tumour cell killing through the induction of apoptosis and cytokine secretion. CD4+ T cells comprise numerous subsets. Some of these subsets, the most well studied of which is the T helper 1 (TH1) subset, can also provide significant antitumour activity. These antitumour CD4+ T cells, collectively termed conventional CD4+ (CD4+ conv) T cells, are distinct from immunosuppressive, pro-tumorigenic CD4+ T cells known as regulatory T (Treg) cells. Although CD4+ conv cells may engage in direct tumour cell killing, they primarily contribute to antitumour immunity through cytokine secretion and assisting in CD8+ T cell activation. Antitumour CD4+ conv T cells share significant metabolic characteristics with CD8+ Teff cells. Although less well understood in terms of antitumour immunity, B cells may also perform effector roles in the TME20. Importantly, as part of the adaptive immune system, T cells and B cells can give rise to memory cell populations, which can persist long after the resolution of an infection or tumour response. CD8+ memory T (Tmem) cells are a crucial aspect of long-term tumour control. Innate cells, such as natural killer (NK) cells and inflammatory macrophages, perform critical antitumour effector functions as well. There are also immunosuppressive cell populations within the TME, including CD4+FOXP3+ Treg cells, myeloid-derived suppressor cells (MDSCs), anti-inflammatory macrophages and some B cell populations20. Through various mechanisms, including cytokine secretion and metabolic derangements, these cells can dampen or eliminate the effectiveness of antitumour effector cell populations. Lastly, antigen-presenting cells, such as intratumoural dendritic cells (DCs), have been shown to perform essential roles in maintaining active adaptive immune response within the TME21,22. Numerous excellent reviews can be referred to for more detailed discussions of tumour immunology and immunotherapy19,23–25.
The metabolism of the antitumour response
Glucose metabolism of antitumour effector T cells. CD4+ conv and CD8+ Teff cells form the critical effector compartment of the antitumour response. When naive CD4+ and CD8+ T cells, which are non-proliferative, recognize their cognate antigen in the context of co-stimulatory signalling, they become proliferative and enact metabolic features to support immense growth26–28. Although many early investigations highlighted the upregulation of aerobic glycolysis as a hallmark of T cell activation, it is now clear that upregulated tricarboxylic acid (TCA) cycle metabolism and OXPHOS are also a critical aspect of CD4+ conv and CD8+ T cell activation. Although TCA cycle metabolism is upregulated within 24 h post activation, upregulated aerobic glycolysis appears to be a more rapid event, occurring within 6 h after activation27–32. The transcriptional activity of MYC and hypoxia inducible factor 1 (HIF-1) are both upregulated in response to T cell activation and promote metabolic reprogramming26,29,33,34. Notably, although HIF-1 is well known to regulate metabolism in response to hypoxia, its activity is also induced in response to T cell activation in the absence of hypoxia. MYC and HIF-1 transcriptional activity leads to upregulation of genes encoding enzymes that promote glycolysis, such as pyruvate kinase (PKM1), hexokinase 2 (HK2) and GLUT1 (refs29,34,35). Pathways emanating from proximal metabolites in the glycolytic pathway are also integral components of T cell activation and function (Fig. 1). The pentose phosphate pathway (PPP) metabolizes glucose-6-phosphate to generate NADPH and ribose-5-phosphate36. Glucose shuttling into the PPP is significantly increased upon CD4+ T cell activation29. The PPP is the primary cellular source for NADPH, which is required for fatty acid and plasma membrane synthesis in newly activated CD8+ T cells37. NADPH is also critical for REDOX homeostasis in proliferating mammalian cells38–40. ROS Glucose carbons that are not metabolized to lactate or by proximal glycolytic pathways contribute significantly to the TCA cycle in Teff cells6 (Fig. 1). In highly proliferative cells, intermediates of the TCA cycle are rapidly consumed to serve as building blocks for a broad range of biomolecular syntheses, a process called cataplerosis45. For example, citrate can be exported to the cytoplasm to regenerate acetyl-CoA for use in lipid and cholesterol synthesis, both of which are critical for producing membranes in proliferative Teff cells. Other TCA cycle intermediates function as building blocks for biosynthesis of, for example, nucleotides and amino acids, which are in high demand during proliferation. Like cancer cells, Teff cells are highly proliferative and upregulate specific glycolytic programmes, including aerobic glycolysis, PPP, HBP and TCA cycle support, to allow massive cell division and effector functions.
T cells and glucose restriction in the TME.
Glucose limitation within the TME can markedly affect the T cell response. For example, low-glucose conditions (0.1 mM) suppressed the generation of the glycolytic intermediate phosphoenolpyruvate (PEP) in T cells, which disrupted calcium-dependent NFAT signalling in vitro18. Compared with control, decreasing the glucose concentration in growth media has been shown to suppress the extracellular acidification rate (a measure of aerobic glycolysis), augment the oxygen consumption rate (a measure of OXPHOS), attenuate mTOR signalling and suppress the effector function of both CD4+ and CD8+ Teff cells46–48. Reduced mTOR complex 1 (mTORC1) signalling interfered with Teff cell differentiation and, in the case of CD4+ T cells, specifically favoured the development of immunosuppressive, pro-tumorigenic Treg cells49. Interestingly, in CD8+ T cells, mTOR blockade with rapamycin favoured differentiation of longlived Tmem cells, which may play an important role in sustaining antitumour responses49–51. Decreasing glucose availability in culture suppressed production of the critical effector molecules interferon-γ (INFγ), IL-17 and granzyme B in Teff cells compared with control growth media47,48,52,53. In activated CD4+ T cells cultured in glucose-free media containing the alternative sugar fuel galactose (which suppresses aerobic glycolysis), the glycolytic enzyme GAPDH assumed a moonlighting role, binding the 3′ untranslated region of Ifng mRNA and suppressing its translation and Teff cell function31.
Glucose restriction in media conditioned by primary ovarian cancer cells led to microRNA-mediated suppression of the histone methylase EZH2 (enhancer of zeste homologue 2), leading to decreased NOTCH signalling, suppressed cytokine production and decreased viability of Teff cells54.
Increasing the glycolytic capacity of mouse sarcoma cells through either pharmacologic treatment with the AKT activator 4-hydroxytamoxifen in co-culture experiments or overexpression of key glycolytic enzymes (for example, Glut1, Hk2 and Pdk1) in tumour cells followed by injection into mice led to suppression of CD8+ T cell effector function compared with vehicle-treated tumour cells or empty vector overexpression, respectively14.
Similarly, compared with wild-type tumours, implanted Hk2-overexpressing melanoma cells suppressed CD4+ T cell antitumour effector function and in vivo responses in mouse models18. Furthermore, expression of glycolysis-related genes in tumour samples from patients with melanoma and non-small-cell lung cancer was inversely correlated to T cell infiltration8. Tipping the metabolic balance can also be accomplished through directly manipulating T cell metabolism. For example, overexpression of the glycolytic enzyme PEP carboxykinase in tumour-specific CD4+ T cells improved antitumour responses compared with control vector-transfected T cells in an adoptive T cell model using melanoma-specific T cells18.
Amino acids and the antitumour T cell response.
Like cancer cells, highly proliferative immune cells, such as activated T cells, are reliant on amino acid metabolism to support protein and nucleotide synthesis. As such, amino acid transporters, including SLC7A5 (also known as LAT1)58, SLC38A1 (also known as SNAT1), SLC38A2 (also known as SNAT2)59 and SLC1A5 (also known as ASCT2)60, have been found to be highly upregulated during T cell activation compared with naive cells in in vitro human and mouse studies61. Essential amino acids must be obtained exogenously. For example, leucine was required for mTORC1 signalling, effector function and proper differentiation in effector CD8+ and CD4+ conv T cells. Interestingly, deletion of the leucine transporter, Slc7a5, in mouse models caused metabolic failure during in vitro activation and cytokine-directed differentiation of CD4+ (TH1, IL-17-producing TH17) and CD8+ Teff cells, but had no adverse effect on the differentiation of Treg cells26,62. Activated T cells also rapidly metabolize arginine, and exogenous arginine supplementation leads to improved T cell fitness and increased generation of central Tmem cells63. Serine, tryptophan and cysteine are also vital nutrients for T cell responses and, as such, are important mediators of antitumour immune responses44,64–66. Tryptophan is an essential amino acid and its availability within the TME is an important factor in determining strength and quality of the T cell response. Human T cell proliferation and activation were strongly suppressed in tryptophan-free media compared with normal growth media66,67. Cancer cells, tumourassociated macrophages (TAMs), MDSCs, suppressive DCs and cancer-associated fibroblasts can deplete tryptophan levels through enzymatic activity of indoleamine 2,3-dioxygenase (IDO)68, which can be expressed at high levels in these cells within the TME. Underlining the importance of this metabolic pathway for tumour growth, IDO expression has been correlated with poor outcomes in patients with several cancer types, including gastric cancer, colorectal cancer, non-small-cell lung cancer and melanoma.
The metabolism of immunologic memory.
Unlike Teff cells, CD8+ Tmem cells preferentially rely on OXPHOS78–81. Compared with CD8+ Teff cells, enhanced spare respiratory capacity, a parameter indicative of the ability of cells to upregulate OXPHOS, is also highly characteristic of Tmem cells78. Initial studies using etomoxir as an inhibitor of carnitine palmitoyl transferase 1A (CPT1A), a mitochondrial transporter responsible for the import of long-chain fatty acids destined for fatty acid β-oxidation (FAO), implicated FAO as the primary fuel for OXPHOS in Tmem cells. However, more recent work using T cell-specific Cpt1a knockout models has called this into question and demonstrated that offtarget effects of high-dose etomoxir (200 μM) are likely responsible for the earlier findings82. This should not be taken to imply that Tmem cells do not use FAO in support of OXPHOS and spare respiratory capacity but, rather, that FAO is not the sole pathway responsible for this metabolic phenotype. Indeed, the expression of CPT1A is consistently upregulated in CD8+ Tmem cells compared with Teff cells. Furthermore, a CD8+ T cell subset known as tissue-resident memory cells were specifically dependent on fatty acid binding protein 4 (FABP4) and FABP5 to import extracellular fatty acids for FAO and for maintenance of a long-term memory phenotype83.
Hypoxia and the antitumour T cell response.
Although tumours are highly heterogeneous, high levels of metabolic activity and associated oxygen consumption, as well as disorganized, poorly functioning vasculature, can generate hypoxic regions with median oxygen saturation levels <2% (compared with a median of about 5% in normal tissues)89,90. The effect of hypoxia on Teff cells is not straightforward. Complicating this area of study is the fact that HIF-1 transcriptional activity is upregulated in response to T cell activation in normoxic conditions34, so it is challenging to understand the effect of hypoxia on further augmenting HIF-1 activity while also evaluating HIF-1-independent effects. Early in vitro studies of CD8+ Teff cell activation, differentiation and function showed that whereas proliferation and the expression of some cytokines were suppressed in hypoxia, the lytic capacity, activation markers and survival were improved91.
Subsequent in vivo studies showed that CD4+ and CD8+ splenic T cells were more poorly activated after concanavalin A challenge in mice exposed to subatmospheric O2 tension (8%) compared with mice exposed to ambient O2 tension (20%)92. Other studies showed that in vitro hypoxic exposure causes intracellular accumulation of the metabolite (S)-2-hydroxyglutarate (S-2-HG), which profoundly alters CD8+ T cell activation and differentiation, suppressing cytokine secretion and cytolytic capacity, but, interestingly, augmenting proliferation, long-term survival and antitumour response after in vivo transfer in mouse models84. Contrary to previous findings demonstrating the necessity of oxidative metabolism and oxidative metabolic capacity in forming long-lived memory CD8+ T cells, glycolytic activity enforced through constitutive HIF-1α activity (achieved through conditional knockout of the HIF-1 regulator.
NK cells are particularly adept at cell killing during major histocompatibility complex class I (MHC-I) downregulation, a common evasion strategy of cancer cells, they form a critical effector component of the innate response. Metabolically, aerobic glycolysis and OXPHOS were upregulated after in vitro cytokine stimulation (IL-12 and IL-15) of NK cells112. Interestingly, SREBP transcription factors were required for these cytokineinduced metabolic changes during in vitro NK cell stimulation113. Pharmacologic inhibition of SREBP activity suppressed metabolic reprogramming, cytokine production and cytotoxicity in vitro and curtailed antitumour response in an adoptive NK cell mouse model. Interestingly, it has been reported that endogenous SREBP inhibitors, such as 27-hydroxycholesterol, can be increased within the TME and thus may be a mechanism of NK cell suppression112,114–118. Lung cancer progression in mice and tumour-associated transforming growth factor-β (TGFβ) are correlated with increased fructose-1,6-bisphosphatase (FBP1) expression in tumour-associated NK cells119. FBP1 is a key enzyme in gluconeogenesis, which, when activated, strongly suppressed glycolysis in NK cells, leading to dysfunction and diminished viability. Interestingly, pharmacologic inhibition of FBP1 was sufficient to re-establish glycolytic metabolism, as well as cytokine production and cytotoxicity in vitro, and improve antitumour response in adoptive cell therapy mouse models119. These studies showed that rescuing NK function through FBP1 inhibition was dependent on restoration of glucose metabolism, as blocking glucose metabolism with 2-deoxyglucose (2-DG) prevented the rescue caused by FBP1 inhibition. 2-DG by itself also led to NK cell dysfunction119, implying that inhibition of glucose metabolism could have profound effects on NK cell antitumour response. Other metabolic derangements within the TME are likely to affect NK cell function as well. For instance, low arginine levels can impair NK cell proliferation and IFNγ production120,121, and hypoxia can suppress cytolytic activity122–124.
結論
Although much of the foundation of immunometabolism has been informed by observations of cancer metabolism, it is clear that there are distinct differences between cancer and immunologic metabolic reprogramming. These differences provide opportunities to target metabolism as a means of enhancing the efficacy of immunotherapy (Fig. 3). Such an approach can be achieved through numerous different strategies. These include targeting tumour metabolic programmes to inhibit growth and alter the TME, targeting the metabolism of suppressive immune cells to inhibit their function and targeting effector cell metabolism to enhance tumour killing. Likewise, ex vivo pharmacologic or genetic reprogramming of T cell metabolic pathways prior to adoptive cellular therapy offers an opportunity to dramatically engineer enhanced features, which may include longevity or enhanced effector function (Box 2). Future work should begin to focus on the metabolic interdependence of immune cells and cancer cells within the TME. In addition to nutrient depletion and the generation of metabolites that can suppress the immune response at certain concentrations, cancer cells can engage in metabolic crosstalk with other cells within the TME, wherein metabolic programmes can be induced and co-opted to benefit malignant progression. It has been reported that pancreatic stellate cells can provide alanine to cancer cells and, thus, fuel proliferation184, and bone marrow stromal cells have been reported to provide cysteine to promote survival of chronic lymphocytic leukaemia cells185. In another report, ammonia from cancer cell glutamine metabolism diffused through the TME and triggered autophagy in cancer-associated fibroblasts, which in turn provided protein breakdown products, such as glutamine itself, to further support cancer cell metabolism186. It will be important to understand whether and by what mechanism immune-evading cancers may be co-opting the metabolic machinery of immune cells and benefitting from their remarkable metabolic flexibility.
未來的工作應該開始關注TME內免疫細胞和癌細胞的代謝相互依存。除了營養耗竭和生成代謝物（在某些濃度下可抑制免疫應答）外，癌細胞還可與TME內的其他細胞發生代謝串擾，其中可誘導並選擇代謝程序，從而有益於惡性進展。據報導，胰腺星狀細胞可以為癌細胞提供丙氨酸，因此，燃料增殖184，骨髓基質細胞被報導提供半胱氨酸促進慢性淋巴細胞白血病細胞的存活185。在另一份報告中，來自癌細胞穀氨醯胺代謝的氨通過TME擴散，並觸發癌症相關成纖維細胞的自噬，進而提供蛋白質分解產物，如穀氨醯胺本身，以進一步支持癌細胞代謝186。了解免疫逃避癌症是否以及通過何種機制可能選擇免疫細胞的代謝機制並從其顯著的代謝靈活性中獲益將是非常重要的。