The obese adipose tissue microenvironment in cancer development and progression

Abstract

Obesity is associated with both increased cancer incidence and progression in multiple tumour types, and is estimated to contribute to up to 20% of cancer-related deaths. These associations are driven, in part, by metabolic and inflammatory changes in adipose tissue that disrupt physiological homeostasis both within local tissues and systemically. However, the mechanisms underlying the obesity–cancer relationship are poorly understood. In this Review, we describe how the adipose tissue microenvironment (ATME) evolves during body-weight gain, and how these changes might influence tumour initiation and progression. We focus on multiple facets of ATME physiology, including inflammation, vascularity and fibrosis, and discuss therapeutic interventions that have the potential to normalize the ATME, which might be translationally relevant for cancer prevention and therapy. Given that the prevalence of obesity is increasing on an international scale, translational research initiatives are urgently needed to provide mechanistic explanations for the obesity–cancer relationship, and how to best identify high-risk individuals without relying on crude measures, such as BMI.

The obesity pandemic is growing globally, representing a considerable cost to public health and a clinically urgent issue for a substantial proportion of the population^1,2. On a global scale, more people are overweight or obese than underweight; in 2013, ~36% of men, ~38% of women and ~23% of children were overweight or obese^3,4. In the United States alone (one of the most obese countries worldwide³), more than one-third of the adult population is obese and an additional one-third is overweight⁵, requiring ~$190 billion in annual health-care expenditure⁶. If these trends continue, the incidence of global obesity, excluding the population who are overweight, is expected to reach ~20% by 2025 (REF³). Given its prevalence, a critical need exists to evaluate the contribution of this comorbidity to cancer and other diseases in both clinical and preclinical research.

Associations between obesity and several pathological conditions are well-established, the most widely recognized being hypercholesterolaemia, hypertension, cardiovascular disease and type 2 diabetes mellitus. However, the link between obesity and cancer is relatively underappreciated among the general population. A retrospective analysis of >1,000 epidemiological studies reported that excess body fatness is correlated with increased risk of 13 distinct types of cancer in adults, out of the 24 that were evaluated⁷. Of the remaining 11 types of cancer, correlations were not necessarily refuted as data were too limited to draw definitive conclusions. In addition to being a risk factor for numerous cancers, obesity is associated with worse outcomes for a subset of tumour types. Indeed, a prospective study following up 1 million adults over a 16-year period reported that obesity is associated with an increased relative risk of death across 10 types of cancer in men and 12 in women⁸, including but not limited to tumours that are in direct contact with adipose tissue. Given these statistics and the prevalence of obesity worldwide, it is not surprising that obesity now competes with smoking tobacco as the leading preventable risk factor for cancer⁹, and is estimated to be responsible for ~14% and ~20% of all cancer-related deaths in men and women, respectively^8,10. Thus, translational research initiatives are urgently needed to provide mechanistic explanations for these striking statistics in order to better serve this growing segment of the population¹¹.

Although these epidemiological studies are critically informative, they do not provide the whole picture. Most studies use crude measures of obesity, such as BMI or waist circumference, which do not capture the diverse biology of ‘fatness’ (Box 1). For instance, epidemiological studies cannot easily uncouple the relative contributions of diet versus adiposity to cancer risk, yet these factors could act alone or together to affect dysbiosis (pathologic imbalance in the gut microbiota), inflammation and other factors that influence malignancy. In addition, although the volume of adipose tissue is associated with risk of disease, the quality of adipose tissue (for example, the presence of inflammation, adipocyte hypertrophy or hypoxia) is an important factor that is not accounted for with simple weight measurements, and is a major driver of metabolic aberrancies including insulin resistance, metabolic syndrome, all-cause mortality and cancer death^12,13. The importance of adipose tissue quality is exemplified by metabolically obese normal-weight (MONW) individuals, who display metabolic abnormalities despite appearing lean. Finally, marked differences exist between the biology and physiological roles of different fat depots within the body (FIG. 1); both the type of adipose tissue (for example, thermogenically active brown fat versus energy-storing white fat)¹⁴ and the distribution of adipose tissue (such as visceral versus sub-cutaneous adipose tissue)¹⁵ differentially influence the risk of developing the metabolic syndrome. For example, during body-weight gain, accumulation of visceral white adipose tissue (WAT), particularly omental and mesenteric fat, is strongly associated with the development of insulin resistance and the metabolic syndrome^16–18, compared with accumulation of subcutaneous WAT. Given the complex biology underlying adipose tissue and obesity, properly controlled preclinical studies are required to understand how obesity affects tumour biology at the molecular and cellular levels, and to define the driving forces behind the obesity–cancer relationship.

Box 1 |. Weighing the value of BMI.

High BMI has been associated with reduced mortality following certain health events, including heart failure, heart attack and some cancers¹⁷³. This observation is known as the obesity paradox, and is debated widely owing to its clinical implications. Several competing explanations for the obesity paradox have been proposed.

Weight and the metabolic syndrome do not always correlate

BMI is used as a readily available cost-effective surrogate for predicting obesity comorbidities. Although this assumption is reasonable for population studies, ~25% of individuals who are obese are metabolically healthy¹⁷⁴ and ~20% of individuals who are lean are metabolically obese¹⁷⁵ (metabolically obese normal-weight; that is, they have qualitatively similar health risks as people who are obese despite having a healthy-range BMI).

BMi categories are not one-size-fits-all

BMI does not consider age or sex; therefore, certain demographics are frequently misclassified (for example, postmenopausal women)¹⁷⁶. Additionally, BMI categories are based on morbidity and mortality statistics from white European populations, and do not universally predict adverse health conditions. For example, a high number of Asian Indians with a healthy-range BMI have type 2 diabetes mellitus. Therefore, BMI categories need to be adjusted between populations¹⁷⁷.

BMi is not static

For some cancers, weight loss is the first symptom that prompts a medical evaluation. Although illness-induced weight loss is unlikely to push an individual into an entirely new BMI category, even modest weight loss can have biological consequences. For example, 10% acute weight loss in women who are morbidly obese leads to an improvement in some systemic markers of the metabolic syndrome and inflammation¹⁵⁶, and reduced systemic predictors of metastasis based on preclinical models¹²⁴.

A biological basis for the obesity paradox

Although excess body fat might increase the risk of cancer, it probably affects other biological processes that could suppress the progression of some malignancies (for example, immune function and clonal selection). Defining the basis for the obesity paradox is an active area of ongoing investigation.

Fig. 1 |. Major adipose depots and anatomical locations in adult humans and mice.

There are two major types of adipose tissue, Lipid-rich white adipose tissue (WAT; energy storing) and mitochondria-rich brown adipose tissue (BAT; energy burning). Adipocytes from BAT and WAT emerge from distinct cell fate lineages; however, WAT can convert to metabolically active fat through the process of browning. WAT is found in many anatomical locations. The largest WAT depots are subcutaneous (under the skin; for example, inguinal, gluteal and femoral) and visceral (within the abdominal cavity, between the organs; for example, omental, mesenteric and epicardial). Smaller WAT depots are found around blood vessels (perivascular, which regulate vascular tone¹⁸⁹), within the bone marrow (which regulate bone remodelling¹⁹⁰), or as ectopic depots within specific organs (for example, non-alcoholic fatty liver disease, fatty pancreas or intramuscular fat). These adipose depots have unique molecular features; therefore, weight distribution has important biological consequences. For example, most visceral adipocytes have slightly higher mitochondrial density, are more lipolytic, are less sensitive to insulin, produce less leptin, and contribute more circulating free fatty acids compared to subcutaneous adipocytes¹⁵, and therefore have a stronger association with the metabolic syndrome. Part a depicts the general distribution of fat in humans from a forward view and part b depicts types of visceral fat depots from a side view.

Part c depicts the general distribution of fat in mice from a supine view.

Several excellent reviews have been published that highlight various aspects of a growing appreciation for the link between obesity and cancer^9,19–23. This Review describes how local changes within the obese adipose tissue microenvironment (ATME) affect tumorigenesis, with a focus on WAT, and how the quality of adipose tissue influences cancer incidence and/or progression. We first consolidate preclinical findings that focus on different components of the ATME, such as inflammation, vascularity and fibrosis, to gain an overall picture of how these local changes might mechanistically influence cancer. We then focus on translational implications of these mechanisms, by discussing therapeutic intervention strategies that might be useful to normalize the obese ATME. Finally, we discuss some important considerations for future management of the obesity–cancer relationship, including how to best identify high-risk individuals.

Fuelling cancer with the obese ATME

In healthy individuals, WAT can comprise approximately a quarter of body mass^24,25 and is a major physiological fuel source (FIG. 1). WAT consists of a variety of different cell types (for example, immune cells, fibroblasts and endothelial cells) that support tissue homeostasis and insulin sensitivity. Studies in mice and humans show that under normal-weight conditions (BMI of 18.5–24.9), the ATME is rich in type 2 (anti-inflammatory) cytokines, and as a consequence has an anti-inflammatory immune cell landscape²⁶. However, during weight gain, adipocytes become hypertrophic and die, which is a key event during the evolution of the obese ATME. Adipocyte death triggers innate immune responses²⁷, thus pivoting the immune repertoire towards a type 1 (pro-inflammatory) state. This shift is characterized by increased production and release of numerous cytokines into the microenvironment (for example, TNF, IFNγ, IL-1β and IL-6) and remodelling of the immune cell landscape²⁶. These inflammatory changes coincide with chronic fibrosis and vascular inflammation, which feed-forward to constantly disrupt tissue homeostasis²⁸. Of note, immune regulation of metabolic homeostasis has been covered by several excellent reviews^26,28–31 and will not be the focus here. Instead, in the following section we discuss the ancillary consequences of these heterotypic relationships; namely, how local changes in the obese ATME affect tumour biology (FIGS 2,3), with a focus on tumour types that have direct exposure to WAT, most notably breast cancer.

Fig. 2 |. Evolution of the adipose tissue microenvironment during obesity.

During healthy body-weight conditions (metabolic homeostasis), the adipose tissue microenvironment (ATME) is well-vascularized and rich in anti-inflammatory cytokines (such as IL-4, IL-10 and IL-13), and as a consequence, hosts a variety of type 2 immune cells, including alternatively activated (M2-like) macrophages, group 2 innate lymphoid cells, type 2 T helper (T_H2) cells and IL-4-producing eosinophils. In response to body-weight gain or metabolic obesity, adipocytes undergo hyperplasia and hypertrophy; as the vascular supply becomes limited, these cells become stressed or die. This releases damage-associated molecular patterns (DAMP) into the microenvironment, which trigger the infiltration and activation of innate immunecells (for example, dendritic cells, macrophages and granulocytes). These effects promote the development of crown-like structures (CLS) and type 1 (pro-inflammatory) immune responses. This response includes accumulation of type 1 cytokines (for example TNF, IFNγ, IL-1β and IL-6), and pro-inflammatory immune cells, including various granulocytes, group 1 innate lymphoid cells, B cells and CD8⁺T cells, which perpetuate chronic inflammatory responses. Macrophages are highly diverse within the obese ATME; those associated with CLS (CLS-associated macrophage, CAMφ) proliferate and express the cell surface markers CD11c and CD9, while those that are further away (non-CAMφ) express lymphocyte antigen 6C (Ly6C). These inflammatory changes coincide with chronic fibrosis and vascular inflammation, which feed-forward to sustain inflammation. IL-10^hi T_reg cells, T regulatory cells producing high levels of IL-10; ILC1, group 1 innate lymphoid cells; MDSC, myeloid-derived suppressor cells.

Fig. 3 |. Interactions between tumour cells and cells within the obese adipose tissue microenvironment.

Tumour cell biology is affected by multiple cellular players in the adipose tissue microenvironment (ATME). The enzyme neutrophil elastase cleaves insulin receptor substrate 1 (IRS1), leading to insulin resistance and elevated levels of free insulin, which can enhance phosphoinositide 3-kinase (PI3K) signalling within tumour cells. Mast cells produce various proteases such as cathepsin S, which can promote cancer progression and therapeutic resistance to cytotoxic therapies. Adipose stromal cells and myofibroblasts promote extracellular matrix (ECM) remodelling and facilitate tumour angiogenesis through vascular endothelial growth factor (VEGF) production. However, angiogenic vessels are dysfunctional, and when combined with adipocyte hypertrophy this triggers innate immune responses by macrophages and dendritic cells. Macrophages form crown-like structures (CLS) that not only serve as biomarkers for the metabolic syndrome, but also produce high levels of cytokines (promote vascular permeability) and cause desmoplasia, which together support tumour progression. At the same time, dendritic cells become chronically activated through the suppression of tolerogenic signalling pathways such as the canonical WNT and peroxisome proliferator-activated receptor-γ (PPARγ) pathways, eventually leading to T cell exhaustion. T cell activation is further suppressed by myeloid-derived suppressor cells (MDSCs), which engage co-inhibitory checkpoint molecules on T cells (such as programmed cell death 1; PD-1). Together, each of these cell types facilitates evolution of the tumour microenvironment and disease progression. MAPK, mitogen-activated protein kinase; TGFβ, transforming growth factor-β.

Crown-like structures and adipose tissue macrophages.

During weight gain, adipocytes accumulate lipids, become hypertrophic and eventually die. The mechanism of adipocyte death is still an open question in the field, although it could be through pyroptosis³² or necrosis³³. Both processes involve rupture of the cell membrane and release of cellular contents into the microenvironment, such as lipids, cytokines and damage-associated molecular patterns (for example, fatty acids, ATP, reactive oxygen species, cholesterol and nucleic acids^34–39). These signals are major triggers for the accumulation of phagocytic macrophages⁴⁰ through enhanced recruitment (following peripheral myelopoiesis and monocytosis)⁴¹, as well as in situ proliferation^42,43 (which is, paradoxically, a hallmark feature of type 2 inflammation^43,44). Up to 90% of macrophages in the obese ATME are directly associated with adipocyte hypertrophy³³, where they encircle the dying adipocyte to form crown-like structures (CLS), scavenge lipids and cellular debris, and sometimes evolve into multinucleated giant cells or foam cells⁴⁵ (FIGS. 2,3). The outcomes of these interactions include activation of macrophage pattern recognition receptors (such as nucleotide-binding oligomerization domain (NOD)-like receptors (NLRs) or Toll-like receptors (TLRs)) and downstream signalling through inflammasome activation^37,41,46, inhibitor of nuclear factor-κB (NF-κB) kinase subunit-β (IKKβ)-NF-κB⁴⁷ and/or c-Jun N-terminal kinase 1 (JNK1; also known as MAPK8)⁴⁸. Indeed, targeting these pathways using genetic or pharmacological approaches is sufficient to protect against insulin resistance during diet-induced obesity^47,48. Therefore, the formation of CLS in the obese ATME is a principle inflammatory lesion and biomarker of adipose tissue inflammation, and is correlated with insulin resistance and metabolic syndrome. Furthermore, in individuals who have normal body weight, the presence of CLS in WAT correlates with metabolic dysfunction⁴⁹, suggesting that these structures are a stronger indicator of metabo-inflammation and metabolic risk than BMI (Box 1).

In patients with cancer, CLS have been associated with worse outcomes, and interest is growing in using these structures as a prognostic biomarker. In breast cancer, histology reveals CLS of the breast in ~50% of women. A positive status for CLS of the breast is associated with higher BMI and systemic markers of the metabolic syndrome compared with negative CLS status, as well as increased breast cancer risk, distant recurrence and mortality^13,50,51. Consistent with preclinical observations that overexpression of aromatase (an enzyme that converts androgens to oestrogens) in the mammary gland stimulates tumorigenesis⁵², CLS of the breast in humans are associated with elevated levels of aromatase and increased risk of breast cancer in women with a history of benign breast disease⁵¹. These relationships are similar in prostate cancer; in a prospective study of 169 men with newly diagnosed prostate cancer, CLS within periprostatic fat were found in 49.7% of men in association with higher BMI and tumour grade, concomitant with elevated circulating levels of insulin, triglycerides, leptin and other factors⁵³. Of note, the association between CLS and tumorigenesis is not specific to hormone-driven cancers with close proximity to large fat depots; CLS are even found in the tongue, where they are correlated with increased BMI and reduced disease-specific and overall survival in patients with early-stage squamous cell carcinoma of the tongue⁵⁴. These data provide clinical evidence that CLS can be used as a biomarker and might be an important biological driver of the obesity–cancer relationship, and suggest that the consequences of CLS are probably multifactorial.

Macrophages within CLS and surrounding adipose tissue are diverse, and probably affect tumour biology in different ways. In a variety of cancer types, macrophages are often the most abundant cell type in the tumour microenvironment, and their accumulation is frequently associated with worse outcomes and resistance to cytotoxic therapy⁵⁵. Macrophages are highly plastic cells; compartmentalization of macrophages within the ‘M1–M2’ linear paradigm (in which macrophages are polarized towards a classical (M1 or pro-inflammatory) or alternative (M2 or anti-inflammatory) activation state) has been largely dismissed, as it is now known that macrophages can adopt myriad activation states that are much more diverse. Indeed, macrophages in the ATME are phenotypically and spatially heterogeneous. For instance, obese ATME macrophages can be functionally subdivided by expression of cluster of differentiation molecule 11c (CD11c; also known as integrin alpha X), a classical dendritic cell marker, whereby the CD11c⁺ fraction exclusively promotes insulin resistance in mouse models of diet-induced obesity^56,57 and is frequently associated with CLS⁵⁸. Specific ablation of CD11c⁺ cells normalizes insulin sensitivity, even when the population of CD11c⁻ macrophages remains intact^56,57. Interestingly, in preclinical models of breast cancer, tumour-associated macrophages similarly express high levels of canonical dendritic cell markers including MHC class II and CD11c, distinguishing them from normal mammary tissue macrophages⁵⁹. The population of CD11c⁺ macrophages accelerates breast cancer growth by promoting T cell exhaustion, through progressive expansion of programmed cell death 1 (PD-1)⁺ granzyme B⁻ CD8⁺ T cells⁵⁹. It is conceivable that the increase in the number of CD11c⁺ macrophages in the ATME during obesity contributes to their availability in the mammary gland during breast tumour progression, thus contributing to an immunosuppressive microenvironment.

Functional heterogeneity in subpopulations of ATME macrophages has also been identified based on their spatial relationship with CLS. ATME macrophages that are not directly associated with CLS express the monocyte marker lymphocyte antigen 6C (Ly6C), whereas CLs-associated macrophages (CAMφ) lose the expression of Ly6C and instead express the canonical exosome marker CD9 (REF⁶⁰). Both tumour-derived and stroma-derived exosomes have received widespread attention for their role in cancer progression and metastasis^55,61. CD9⁺ CAMφ secrete exosomes, which could, in principle, facilitate direct communication between CLS and tumours; however, the content of these exosomes is unknown. Another unique feature of CAMφ in the ATME is that they proliferate⁴², which is typically a hallmark feature of type 2 inflammatory responses⁴⁴, thus challenging the dogma that ATME macrophages are type 1 cells. Paralleling a growing appreciation for more complex macrophage diversity in tumour biology, it is conceivable that ATME macrophages are similarly capable of co-expressing both M1 and M2 markers leading to a blended phenotype, emphasizing the importance of moving away from the M1–M2 paradigm in this field. Supporting this notion, transcriptional profiling showed no statistically significant enrichment for traditional M1 or M2 signatures in either CD11c⁻ or CD11c⁺ ATME macrophages in response to obesity⁶², including specifically in CD11c⁺ CAMφ⁵⁸, and proteomics analyses have confirmed that obesity induces ‘metabolic activation’ in ATME macrophages that is phenotypically distinct from both M1 and M2 activation⁶³.

Adipose tissue leukocytosis—inflammatory reactions beyond macrophages.

Once CLS develop, the innate immune system is engaged, the mandate of which is to rapidly defend the body while enabling the slower developing adaptive immune response. In addition to the formation of CLS, obesity is associated with a variety of other inflammatory changes (FIG. 3). Within the myeloid compartment, even though the frequency of eosinophils is decreased⁶⁴, this reduction is compensated by several other granulocytic cell types that become activated. Neutrophils, which have been heavily implicated in cancer⁶⁵, are rapidly increased in WAT after just 3 days of a high-fat diet, where they produce the serine protease neutrophil elastase that cleaves insulin receptor substrate 1 (IRS1)⁶⁶. This cleavage event prevents IRS1 from binding to phosphoinositide 3-kinase (PI3K), thus leading to insulin resistance⁶⁶, which could in turn promote the proliferation of epithelial cells. Although activating mutations in the PI3K pathway represent one of the most common oncogenic events in cancer, obesity might act as a surrogate mechanism of PI3K activation in wild-type tumours via its inductive effects on insulin. The frequency of mast cells is also increased, which respond to IL-6 and IFNγ by producing cathepsin S⁶⁷, a member of the cysteine cathepsin protease family with roles in cancer progression and therapeutic resistance⁶⁸. Granulocytic myeloid-derived suppressor cells also become elevated as a defence mechanism in numerous tissues including the obese ATME, where they attempt to counter the development of inflammation and insulin resistance⁶⁹. In breast cancer models, the immune checkpoint molecule programmed death-ligand 1 (PD-L1) is upregulated in myeloid-derived suppressor cells during obesity by IFNγ released from tumour cells, causing impaired CD8⁺ T cell surveillance, and accelerated tumour onset and progression⁷⁰. Even though each of these granulocyte dynamics might initially serve as an attempt to keep inflammation in check during bodyweight gain, in the context of cancer, this effect might lower the barrier to malignant transformation of pre-cancerous cells by impairing immune surveillance, and ultimately supporting tumour proliferation and progression.

Supplementing these granulocytic responses, antigen-presenting myeloid cells such as classic dendritic cells (cDC) become involved in the obese ATME. In response to over-nutrition, adipocyte hyperplasia suppresses tolerogenic (anti-inflammatory) canonical WNT-β-catenin and peroxisome proliferator-activated receptor-γ (PPARγ) signalling pathways in cDC1 and cDC2 cells, respectively⁷¹. This effect enables the activation of cDC and T cell-mediated responses within visceral WAT, thus bridging innate and adaptive immunity to facilitate immune surveillance. However, perpetual antigen presentation by cDCs during sustained over-nutrition might eventually lead to T cell exhaustion and chronic inflammation. Indeed, obesity accelerates thymic ageing in conjunction with reduced frequency of peripheral naive T cells⁷², compromised T cell receptor diversity^72,73 and increased expression of checkpoint molecules and molecular markers of exhaustion⁷⁴. However, upon PD-1 blockade, obesity facilitates strong T cell effector functions⁷⁴, which suggests that this process is partially reversible, and might explain the high efficacy of checkpoint inhibitors (for example, PD-1-blocking or PD-L1-blocking antibodies) in men who are obese and have melanoma⁷⁵.

Several additional lymphoid cell types within the obese ATME might also contribute to malignancy. B cells accumulate in WAT during body-weight gain, and preclinical models show that genetic depletion of B cells protects against the metabolic syndrome in response to a high-fat diet even though mice still gain weight⁷⁶, suggesting that B cells might regulate the development of CLS but not adipocyte hypertrophy or hyperplasia. Activated effector CD8⁺ T cells also infiltrate WAT in response to obesity; this event might even precede the increase in numbers of macrophages in the ATME as CD8-depletion prevents macrophage infiltration and improves insulin sensitivity⁷⁷. In lean mice, CD4⁺ forkhead box protein P3 (FOXP3)⁺ regulatory T cells (T_reg cells) are present in the ATME^78,79, where they interact closely with CAMφ (which are rare in lean mice) and produce high levels of anti-inflammatory stimuli, such as IL-10 (REF⁸⁰). These interactions might help keep CLS frequency low in lean adipose tissue. However, during obesity, these populations of T_reg cells are reduced relative to the expansion of adipose tissue^78–80, thus alleviating homeostatic safeguards. Despite the fact that T_reg cells usually enable tumour outgrowth owing to their immunosuppressive properties, the loss of T_reg cells in the obese ATME might enable the expansion of CLS and lead to a pro-tumorigenic microenvironment. Finally, IL-12-regulated group 1 innate lymphoid cells, with long-term residency in the ATME, proliferate and produce the type 1 cytokine IFNγ in response to obesity; however, these cells are distinct from both mature and immature natural killer cells in the ATME, which are principal members of the group 1 innate lymphoid cell family⁸¹. Similar to CAMφ, these studies collectively suggest an ATME-specific activation phenotype of these different populations of immune cells; how these unique populations affect tumour biology remains largely unclear.

Taken together, these data suggest that the obesity-inflammation axis is a central regulator of the metabolic syndrome that might also underlie many of the associated risks with cancer. Indeed, the link between chronic inflammatory conditions and cancer is widely appreciated⁸², both at the level of risk and progression; it is probable that obesity falls within this realm. Of note, although much of the discussion in this section is focused on solid tumours, the effect of obesity on haematological malignancies and the role of fat residing within the bone marrow is discussed in BOX 2.

Box 2 |. Obesity and haematological malignancies.

In older adults, bone marrow adipocytes comprise up to ~70% of bone marrow volume, and are therefore an important component of the microenvironment during haematopoiesis. Although adipocytes impair haematopoietic repopulation and expansion during infection^178–181, obesity is associated with chronically increased lymphopoietic and haematopoietic processes during steady-state conditions¹⁸². Given the effect of adipocytes on normal haematopoietic responses, does obesity predispose individuals to haematological malignancies? Leptin, a hormone produced predominantly by adipocytes to help control appetite, might facilitate communication between adipocytes and bone marrow progenitor cells. In preclinical models, downregulation of the leptin receptor in haematopoietic progenitor cells is required for the initiation of B cell and T cell acute lymphoblastic leukaemia (ALL)¹⁸³, and diet-induced obesity accelerates the onset of disease in association with elevated circulating levels of leptin¹⁸⁴. Countering these effects, fasting prevents the initiationof ALL by upregulating the expression of the leptin receptor, thus supporting differentiation¹⁸³. Within the bone marrow, acute myeloid leukaemia (AML) blast cells can metabolically reprogramme adipocytes to promote lipolysis, and this event in turn transfers fatty acids to tumour cells¹⁸⁵. This bidirectional communication and exchange of fatty acids mirrors that of other solid tumours in larger fat depots, including breast and prostate cancer, for which associations between obesity and cancer are apparent. Indeed, epidemiological studies show that obesity is associated with an increased risk of multiple myeloma⁷ and AML¹⁸⁶ in adults. In an epidemiological study of 5,420 children, obesity also increased the risk of relapse from ALL by 50%¹⁸⁷; however, similar analyses in adults are based on smaller cohort sizes, and the data are variable and inconclusive. Although the role of obesity in additional haematological malignanciesin patients has not been formally established, these associations might exist.

Vascular inflammation.

A hallmark feature of cancer is the ability to recruit a vascular supply and engage the endothelium to permit the passage of inflammatory cells, circulating factors and metastatic tumour cells. Given the well-established association between obesity and the cardiovascular system, defining how obesity affects vascular function is relevant to understanding the obesity-cancer relationship. Indeed, the vascular demand in growing adipose tissue mimics that of a growing tumour; although both tissues are highly vascularized, vessel function is aberrant and often insufficient to meet metabolic demand in both cases (FIG. 4). In the obese ATME, vascular dysfunction leads to areas of hypoxia and adipocyte death — a triggering event for CLS formation. In mouse models of obesity, several studies showed that in mice, adipocyte-specific deletion of hypoxia-regulated genes that are necessary to stimulate angiogenesis, such as Vegfa⁸³ or Hif1α⁸⁴, reduces WAT vascularity, increases CLS formation and leads to insulin resistance, compared with mice with wild-type adipocytes. In pre-clinical models of breast cancer, angiogenesis within the obese mammary fat pad primes the ATME during early stages of tumour initiation⁸⁵. However, vessels that are conditioned by obesity are highly dysregulated. Vascular inflammation resulting from adipokine exposure leadsto enhanced permeability (fig. 3), which might facilitate the robust inflammatory changes that ensue in the ATME. For example, the discordant interplay between lipopolysaccharide and adiponectin regulates vascular function; lipopolysaccharide causes increased production of leukocyte adhesion molecules and decreased endothelial adhesion molecules, whereas adiponectin reverses these effects⁸⁶ (fig. 4). Exposure of microvascular cells to the free fatty acid palmitate induces activation of the NOD-, LRR- and pyrin domain-containing 3 (NLRP3) inflammasome in endothelial cells, in concordance with weakened endothelial tight junctionsvia decreased expression of zonula occludens 1 (ZO1) and ZO2 (REf⁸⁷) (fig. 4). Notably, patients with breast cancer who are obese are less sensitive to anti-vascular endothelial growth factor (VEGF) therapy; preclinical models suggest that this effect is attributable to elevated levels of IL-6 production by adipocytes and myeloid cells in the breast ATME, such that blocking IL-6 is sufficient to improve the response to anti-VEGF therapy⁸⁸. Interestingly, treatment with the antidiabetic agent metformin could phenocopy these effects⁸⁸. Given that obesity affects multiple aspects of vascular biology, it is probable that vascular normalization strategies are also an important consideration for patients with cancer who are obese⁸⁹.

Fig. 4 |. Vascular inflammation and fibrosis in the obese adipose tissue microenvironment.

During obesity, adipose tissue expands rapidly, leading to high demand for a vascular supply. Similar to a growing tumour, this effect results in regions where vascular supply is insufficient, creating areas of hypoxia. Hypoxia triggers angiogenesis through induction of pro-angiogenic factors (such as vascular endothelial growth factor (VEGF) and hypoxia-inducible factor 1α (HIF1α)); however, resulting blood vessels are poorly functional, and thus low oxygen levels persist. Chronic hypoxia eventually contributes to adipocyte death, and supports the formation of crown-like structures (CLS). Vascular integrity is in part reduced by downregulation of endothelial adhesions (such as zonula occludens 1 (ZO1) and ZO2) in responseto obesity-derived factors (for example, lipopolysaccharide and palmitate), while leukocyte adhesion molecules are increased to facilitate infiltration of immune cells (such as T cells). Adiponectin can reverse these effects and improve vascular integrity. In addition, prospero homeobox protein 1 (PROX1)⁺ lymphatic endothelial cells use lipids to stimulate lymphangiogenesis through fatty acid oxidation (FAO) and production of acetyl-CoA. The extracellular matrix is also aberrant; CLS-associated macrophages (CAMφ) and myofibroblasts contribute to increased tissue stiffness in the obese adipose tissue microenvironment, and adipocytes produce high levels of collagen VI, which further support adipocyte hyperplasia.

In addition to understanding the effects of obesity on blood vessels, lymphatic vessels are also central regulators of inflammatory responses, and lipids are well-established mediators of lymphatic function⁹⁰. Indeed, elevated BMI is associated with increased lymphoedema in humans⁹¹, including after postoperative lymph node dissection in patients with breast cancer⁹², suggesting that obesity affects lymphatic vessel physiology. Lymphatic vessels are regulated and defined by the expression of prospero homeobox protein 1 (PROX1); in lymphatic endothelial cells, PROX1 mediates fatty acid oxidation leading to elevated production of acetyl-CoA, which promotes epigenetic modifications that are critical for lymphangiogenesis⁹³ (a key step during tumour metastasis⁹⁴) (fig. 4). This effect demonstrates that remodelling of the lymphatic endothelium relies on fat catabolism. Interestingly, Prox1^+/− mice develop late-onset obesity, characterized by abnormal lymphatic function and ruptured vessels⁹⁵, and restoration of Prox1 expression specifically in lymphatic endothelial cells is sufficient to reverse this phenotype⁹⁶. In the tumour microenvironment, lymphatic vessels not only provide a conduit for metastatic tumour cells, but also act as a mechano-biosensor for lymph flow and stiffness of the extracellular matrix (ECM), and have an active role in regulating the immune composition⁹⁷. How obesity affects these complex lymphatic processes is unclear, particularly for tumours that are enveloped by adipose tissue such as breast cancer. Of note, lifestyle modifications, including exercise⁹⁸ or diet-induced weight loss⁹⁹, rescue some of the deleterious effects of obesity on lymphatic function, suggesting that these effects are reversible through non-pharmacological approaches (which might be relevant for patients with cancer once clearer mechanisms are defined). For a more detailed summary of the relationship between obesity and vascularity, see refs^100,101.

Fibrosis and matrix composition.

With vascular inflammation comes changes in matrix composition within tissues (fig. 4). Remodelling of the ECM and stromal microenvironment, a process called desmoplasia, accompanies the disruption of tissue homeostasis including during the process of malignancy. Desmoplasia is regulated by myofibroblasts, which deposit stiff matrix components including fibrillar collagen and fibronectin, to alter the biomechanics and tensile properties of the microenvironment. Tumours tend to have stiffer matrices compared with normal tissues^102,103. During obesity, CAMφ promote myofibroblast activation and tissue fibrosis in WAT¹⁰⁴. In the mammary gland, the stroma also becomes enriched with myofibroblasts during obesity to promote ECM stiffness¹⁰⁵ (Fig. 4). Matrices isolated from obesity-associated stromal cells promote the aggressiveness of breast cancer cells by stimulating mechano-signalling, and this effect is reversible with body-weight loss through caloric restriction¹⁰⁵.

Adipocytes are also one of the highest sources of collagen VI, and deletion of Col6a1 in lean MMTV-PyMT mice (a mouse model of breast cancer) leads to reduced epithelial hyperplasia and breast tumour growth¹⁰⁶ (fig. 4), suggesting a pro-tumorigenic role for collagen VI deposition. During obesity, the production of several collagens is upregulated in adipose tissue, including collagen VI, reflecting a broad shift towards adipose tissue fibrosis¹⁰⁷, which can be further perpetuated by hypoxia¹⁰⁸. In pancreatic ductal adenocarcinoma, of which desmoplasia is a hallmark histological feature, obesity promotes crosstalk between adipocytes, tumour-associated neutrophils and pancreatic stellate cells to promote matrix remodelling and impaired vascular perfusion, leading to tumour progression and ineffective delivery of chemotherapy¹⁰⁹. Taken together, the robust changes to ECM composition that coincide with obesity provide a pro-tumorigenic niche by disrupting tissue mechano-homeostasis, and these changes probably contribute to some of the inflammatory effects within the obese ATME. In support of this notion, obese leptin-deficient ob/ob mice show improvements in ATME inflammation when Col6a1 is deleted, marked by reduced macrophage infiltration, adipocyte death and CLS formation¹⁰⁷.

Transmitting signals between tumour cells and adipocytes.

One component of the ATME that cannot be overlooked is the adipocyte — the conductor of microenvironmental evolution during obesity. Aside from having indirect effects on tumour cells through inflammation, vascularity and fibrosis as discussed previously, adipocytes can also have direct effects on tumour cells through paracrine signalling¹¹⁰. In melanoma, adipocytes in subcutaneous WAT transfer lipids to tumour cells to induce metabolic reprogramming, growth and invasion¹¹¹. In prostate cancer, adipocytes secrete CC-chemokine ligand 7 (CCL7), which stimulates the migration of CC-chemokine receptor 3 (CCR3)⁺ tumour cells to the periprostatic fat (a common region of cancer recurrence in patients¹¹²). In humans, CCR3 expression in prostate cancer cells is associated with more aggressive disease. Adipokines can also have autocrine effects that might indirectly affect tumour biology by regulating insulin resistance and inflammation. For example, adiponectin, which promotes ATME homeostasis, acts locally on adipocytes to regulate glucose uptake, adipogenesis and lipid storage^113,114, whereas leptin (which promotes ATME dysfunction) regulates adipocyte lipolysis by engaging the central nervous system¹¹⁵. Downregulation of adiponectin, and upregulation of leptin in sera are both significantly correlated with CLS formation in breast tissue of women with early-stage breast cancer¹³. These studies suggest that targeting these cytokine axes might hold therapeutic value to disrupt both direct and indirect interactions between adipocytes and tumour cells within the ATME.

Adipose stromal cells (ASCs; commonly referred to as pre-adipocytes) are multipotent mesenchymal progenitor cells within adipose tissue that can also exert local effects on tumour cells. ASCs are recruited from WAT into tumours through chemokine gradients. For example, prostate cancer cells produce growth-related α-protein (CXCL1) and IL-8, which attract CXC-chemokine receptor 1 (CXCR1)⁺ and CXCR2⁺ ASCs, respectively¹¹⁶. In patients who are obese, these cytokines are overexpressed by prostate tumour cells, and IL-8 in particular is associated with high-grade disease¹¹⁶. Upon arrival, ASCs promote various hallmark cancer phenotypes, most notably vascularization^117–119, and display phenotypic plasticity in response to cues from the tumour. For example, in breast cancer, the differentiation of ASCs into myofibroblasts can be induced by tumour extracellular vesicles through a transforming growth factor-β (TGFβ)–mitogen-activated protein kinase (MAPK) signalling axis to support ECM remodelling, increased VEGF production and angio-genesis¹²⁰ (flG. 3). ASCs can also act as a local source of leptin in the mammary fat pad to promote oestrogen receptor (ER)⁺ breast tumour progression during obesity¹²¹. Given their presence within WAT, it is probable that ASCs are important contributors to the obesity–cancer relationship. Indeed, syngeneic mouse models of melanoma, breast, prostate and lung cancer showed that specific depletion of ASCs using pro-apoptotic peptides inhibits tumour vascularization and proliferation, and induces necrosis¹²². Interestingly, this effect was more pronounced in obese mice compared with lean mice, where ASCs are more abundant.

In addition to acting locally, adipocytes are experts at mediating endocrine effects, which facilitate inter-organ crosstalk between the ATME and distant tissues¹²³. This ability has important implications for cancer, as it suggests that tumours with no direct contact with adipose tissue can still be affected by obesity, reflecting the epidemiological data that link obesity with diverse cancer types^7,8. For instance, the ATME produces IL-5, which stimulates lung neutrophilia and breast cancer metastasis by increasing circulating levels of granulocyte-macrophage colony-stimulating factor (GM-CSF)¹²⁴. The obese ATME also produces IL-6 and TNF, which regulate chronic inflammation within the liver and support malignancy by activating the oncogenic transcription factor, signal transducer and activator of transcription 3 (STAT3)¹²⁵. Given the relationship between obesity and diet, it is not surprising that the gastrointestinal tract also has a close relationship with adipose tissue, which can facilitate early changes associated with tumour initiation. In the pre-malignant colon epithelium, high adiposity, but not diet, triggers epigenetic changes that mirror early alterations associated with tumour initiation, including activation of PI3K and RAS¹²⁶. Obesity also exacerbates adipokine production by the ATME in response to Helicobacter felis infection in mice, marked by a statistically significant increase in the pro-inflammatory cytokines IL-6, CCL7 and leptin, which accelerate the onset of gastric cancer¹²⁷. Importantly, this effect is not observed in lean control mice, demonstrating that these effects are specifically driven by the obese ATME. Together, these data demonstrate that the inflammatory response of the obese ATME has far-reaching effects on a variety of tumour types that can exist throughout the host, thus providing a rationale for targeting canonical adipokines even for tumours that are more distant from major adipose depots.

The elevation in serum levels of adipokines in response to obesity and the metabolic syndrome coincides with a spike in circulating levels of cholesterol, and hypercholesterolaemia is a hallmark consequence of obesity. In preclinical breast cancer models, the cholesterol metabolite 27-hydroxycholesterol can act as a selective ER modulator, and studies have shown that it can promote the growth of ER⁺ breast tumours^128,129. Similar to the effects of the IL-5-GM-CSF axis¹²⁴, 27-hydroxycholesterol can also affect lung inflammation and facilitate breast cancer metastasis by promoting lung neutrophilia¹³⁰. Cholesterol metabolism also leads to the generation of the oncometabolite, 6-oxocholestan-3β,5α-diol. Unlike 27-hydroxycholesterol, 6-oxo-cholestan-3β,5α-diol has no effect on ER⁺ breast cancers, and instead stimulates growth of tumour cells within the mammary fat pad by acting through the glucocorticoid receptor¹³¹. These data suggest that controlling indirect consequences of obesity, for example, through the use of statins, might have beneficial effects on both cancer incidence and prognosis.

Finally, adipocytes sequester and catabolize chemotherapeutic agents, which might reduce the therapeutic efficacy of these compounds. In acute lymphoblastic leukaemia, adipocytes seem to protect tumour cells from vincristine¹³² and daunorubicin¹³³. In addition to being able to absorb daunorubicin, adipocytes express carbonyl reductases and aldo-keto reductases, which metabolize daunorubicin to daunorubicinol (its largely inactive metabolite). These findings have potential implications not only for patients with cancer who are obese, but also for those patients who are lean with tumours that are in direct contact with adipose tissue, such as breast cancer. Of note, in vitro functional screens have shown that both cytotoxic and targeted therapies that effectively kill tumour cells in monoculture are often less effective when tumour cells are co-cultured with various stromal cell types¹³⁴. One such cell type is macrophages, which are abundant in the obese ATME and are well-characterized for their ability to protect tumour cells from the cytotoxic effects of chemotherapy¹³⁵. Adipocytes and macrophages in CLS might act synergistically to reduce the efficacy of cancer therapies during obesity, although this effect has not been formally tested. Taken together, although reduced efficacy of chemotherapy in patients who are obese has often been attributed to under-dosing, these data suggest that more complex adipose tissue biology might be involved.

For instance, adipose tissue biology and metabolic activity are strongly affected by circadian rhythms, which synchronize light-dark, fasting-feeding and sleep-wake cycles, leading to optimized host energy balance. Circadian rhythms regulate microbiome composition¹³⁶, fatty acid and glucose metabolism and insulin release¹³⁷; unsurprisingly, interest in the role of circadian rhythms during obesity and the metabolic syndrome is growing. Chronobiology studies in mice show that mutations in clock genes, such as Clock, Bmal1 (REF.¹³⁸) or Per3 (REF.¹³⁹), lead to enhanced development of obesity and the metabolic syndrome, and suppress adipogenesis. Interestingly, time-restricted feeding of mice without changing overall caloric intake or food content can attenuate metabolic aberrancies associated with a high-fat diet, including obesity, adipocyte hypertrophy and WAT inflammation¹⁴⁰. How these anti-inflammatory effects might translate to cancer biology within the ATME is unknown. In cancer, pre-clinical studies have shown that agonists of REV-ERBα-related receptor (REV-ERBα; also known as NR1D1) or REV-ERBβ (also known as NR1D2), essential components of circadian rhythms and lipid metabolism, are selectively lethal in both malignant (for example, glioblastoma) and benign neoplasms (such as nevi), but not normal cells, by regulating autophagy and de novo lipogenesis¹⁴¹. Therefore, in light of these findings, it is plausible that ATME composition within humans can be fine-tuned by chronic behaviours that challenge our natural light–dark cycle, such as sleep disturbances and/or irregular eating behaviours (for example, night feeding).

Therapies against ATME dysfunction

In addition to conventional therapies that largely target tumour cells, targeting the tumour microenvironment is an effective strategy to treat cancer¹⁴². With growing interest in immunotherapy, many of these strategies are focused on targeting immune cells within the tumour microenvironment. We would similarly argue that therapeutic interventions that target the dysregulated ATME might be an effective strategy to improve the obesity-cancer relationship, not only for patients who are obese, but potentially also for those who are over-weight or MONW. In the following section, we review different intervention strategies that aim to normalize the obese ATME, which might be relevant to both cancer initiation and metastatic disease. A summary of ongoing and completed clinical trials can be found in TABLE 1. In BOX 3, we also discuss the possibility that some effects of obesity might be imprinted, and therefore might not be reversible.

Table 1 |.

Preclinical and clinical interventions that relate to the connection between obesity and cancer

Trial type	Intervention	Details	Clinical trial IDs	Refs
Effects on cancer and related end points
Clinical	Telephone-based consultation	Breast Cancer Weight Loss (BWEL) study Effect of telephone-based weight loss coaching in adjuvant treatment of women who are overweight and obese with breast cancer Phase III	NCT02750826 (recruiting)	154
Clinical	Mediterranean-style diet	Effect of Mediterranean-style diet compared with low-fat diet recommendations on breast cancer incidence Retrospectively registered trial	ISRCTN35739639	162,191
Clinical	Antidiabetic (metformin)	Effects of metformin on cancer progression, recurrence and/or response to therapy Phase III trial (IDs are shown) Additional phase I/II trials in prostate cancer (9), breast cancer (9) and >30 others are also ongoing	NCT01697566 (endometrial cancer) NCT01101438 (breast cancer) NCT02065687 (endometrial cancer) NCT02040376 (medulloblastoma)	192–195
Clinical	Antidiabetic (pioglitazone)	Effects of pioglitazone on cancer progression, recurrence and/or response to therapy Phase II trial (IDs are shown) >5 additional phase I trials are also ongoing	NCT01655719 (thyroid cancer) NCT00780234 (lung cancer) NCT01838317 (pancreatic cancer) NCT00099021 (head and neck cancer) NCT00427999 (prostate cancer)	196–200
Clinical	Exercise	Yale Exercise and Survivorship study Effect of aerobic exercise on levels of insulin and insulin-like growth factors in survivors of breast cancer	NA	201
Clinical	Exercise	Effects of exercise on biomarkers of the metabolic syndrome and sarcopenic obesity in survivors of breast cancer who are overweight and obese	NA	168
Clinical	Exercise	Effects of exercise on insulin resistance and circulating biomarkers in postmenopausal survivors of breast cancer	NA	202
Clinical	Bariatric surgery	Swedish Obese Subjects study Effects of bariatric surgery on long-term weight loss and cause-specific mortality (including cancer)	NCT01479452	152
Preclinical	Caloric restriction mimetics	Effects of caloric restriction mimetics on anticancer immune surveillance	NA	146
Preclinical	Fasting-mimicking diet	Effects of fasting-mimicking diet in combination with chemotherapy in breast cancer and melanoma models	NA	147
Preclinical	Dietary restriction	Effects of dietary restriction (60% of normal daily food intake) on PI3K wild-type versus mutant tumour models.	NA	159
Effects on the ATME and related end points
Clinical	VLCD	Effect of VLCD on inflammatory transcriptional profile of the ATME	NA	155
Clinical	VLCD	Effects of rapid weight loss on systemic and adipose tissue inflammation and metabolism in postmenopausal women who are obese	NA	156
Clinical	Mediterranean-style diet	Effect of Mediterranean-style diet on endothelial function in patients with the metabolic syndrome	NA	160,161
Clinical	Low-carbohydrate versus low-fat diet	Effect of a low-carbohydrate versus low-fat diet in patients with severe obesity	NCT00160108	163–165
Clinical	Antidiabetic (pioglitazone)	Effect of pioglitazone on ATME physiology, inflammation and lipid distribution in people who are obese Phase IV	NCT00108615 NCT00470262 NCT00579813	203
Preclinical	Caloric restriction	Effect of 30% caloric restriction on inflammation in the mammary fat pad of obese mice	NA	158
Preclinical	Antidiabetic (pioglitazone)	Effect of pioglitazone on periprostatic WAT inflammation in obese mice	NA	143
Preclinical	Oestrogen supplementation	Effect of oestrogen on mammary WAT inflammation in obese mice	NA	149

The list of clinical trials is not exhaustive; only trials categorized as “Active, not recruiting” or “Completed” by ClinicalTrials.gov are shown, unless otherwise indicated. ATME, adipose tissue microenvironment; IDs, identifiers; NA, not applicable; PI3K, phosphoinositide 3-kinase; WAT, white adipose tissue; VLCD, very-low-calorie diet.

Box 3 |. Is weight loss the simple answer?

Given that excess body fat is associated with an increased risk of numerous malignancies as well as a worse clinical course for some cancers, it is reasonable to posit that body-weight loss might be beneficial. Although we have discussed several studies that suggest that body-weight loss can reduce the growth of experimental tumours and/or suppress metastasis, other preclinical studies have shown that weight normalization might be insufficient to reverse the effects of chronic obesity on haematopoietic function¹⁷⁹, or inflammatory signals and epigenetic reprogramming in the microenvironment that are associated with tumour progression¹⁸⁸. Currently, there are more clinically relevant questions than answers: will weight loss protect against the development of some or all of the obesity-related cancers? Will it reduce the risk of cancer regardless of the duration that a person has been obese? In patients who are obese and have cancer, how much weight loss will be necessary to improve prognosis? Will weight loss alter the efficacy of either targeted therapies or immunotherapy? If weight loss proves to be beneficial in reducing the risk of cancer or improving prognosis, can we identify blood biomarkers such as fasting levels of insulin or leptin? To begin addressing some of these questions, the Breast Cancer WEight Loss Study (BWEL Study) is a prospective phase III clinical trial testing whether weight loss in patients who are overweight or obese who have breast cancer will lead to a lower rate of recurrence compared with women who do not take part in the weight loss programme¹⁷⁰. Given the importance of the link between obesity and cancer, an urgent need exists for additional studies to determine in what context body-weight loss will be clinically beneficial.

Pharmacological interventions to normalize the ATME.

Given the role of CLS in obesity-associated inflammation and insulin resistance, targeting these inflammatory structures during obesity is of considerable interest, for example, through specific macrophage-targeted therapies or vascular normalization strategies to optimize oxygenation. Although these treatment approaches have not been explored in great detail, pharmacological strategies against the metabolic syndrome have been tested. For example, treatment of obese mice with pioglitazone, a PPARγ ligand used to treat diabetes mellitus, induces adiponectin and reduces the expression of TNF, TGFβ and monocyte chemoattractant protein 1 (MCP1; also known as CCL2) in periprostatic adipose tissue, leading to reduced numbers of CLS¹⁴³. Given the potential role of CLS within periprostatic fat in prostate cancer, whether this approach would be beneficial in patients is an important open question. Metformin is another antidiabetic agent that has received broad attention for its application in obesity-associated cancer (see TABLE 1). In addition to inducing modest weight loss, which can have anti-inflammatory effects, metformin can reduce the elevated levels of insulin that occur in association with inflamed adipose tissue¹⁴⁴. Finally, gliptins are another class of hypoglycaemic agents that block dipeptidyl peptidase 4 (DPP4) and are prescribed for type 2 diabetes mellitus. One preclinical study showed that obesity-associated inflammatory pathways in ATME macrophages are regulated by liver-derived DPP4, such that targeting Dpp4 expression in hepatocytes, or its downstream effectors in adipose tissue macrophages, is sufficient to prevent WAT inflammation and insulin resistance¹⁴⁵. However, treatment of mice with the oral DPP4 inhibitor sitagliptin did not reproduce these effects, suggesting that some of these mechanisms might not be reversible with pharmacological interventions.

Additional pharmacological interventions for obesity that have been explored in cancer include compounds that induce weight loss or control appetite, which might be more amenable to patient compliance compared with a low-calorie diet. Preclinical studies in lung cancer showed that caloric restriction mimetics that recapitulate the biochemical effects of nutrient deprivation, boost anticancer immune surveillance by depleting immune-suppressive T_reg cells from the tumour microenvironment¹⁴⁶. Similar effects can be achieved without pharmacological agents, by using a fasting-mimicking diet. This diet stimulates an increase in the infiltration of CD8⁺ cytotoxic T cells into the microenvironment of breast and melanoma tumours, leading to enhanced immune surveillance and delayed progression¹⁴⁷. Although yet to be formally tested, it has been proposed that FDA-approved sympathomimetic compounds that suppress appetite (for example, phentermine or lorcaserin)²⁷ or that activate thermogenesis (such as mira-begron)¹⁴⁸ might likewise be valuable to reduce weight, WAT inflammation and cancer risk. Notably, these interventions might be relevant not only for obesity-associated cancers, but also for tumours that are particularly immunogenic, such as lung cancer or melanoma, which might benefit from improved anti-tumour immunity. The effect of combining these agents with immunotherapies in these tumours should be explored.

In addition to using behavioural or pharmacological approaches to target the more obvious culprits (for example, calorie consumption), other methods have been proposed to disrupt the obesity-cancer relationship. Preclinical models using oestrogen replacement therapy in obese ovariectomized mice have shown that treatment with 17β-oestradiol is sufficient to mitigate weight gain, CLS formation and production of several inflammatory markers including cyclooxygenase 2, TNF, IL-1β and aromatase¹⁴⁹. Whether oestrogen can be used as a monotherapy in carefully selected postmenopausal women to reduce WAT inflammation and potentially reduce the risk of breast cancer has yet to be explored.

Surgical intervention.

Bariatric surgical interventions such as gastric bypass, vertical-banded gastroplasty or banding are a method to rapidly induce weight loss, improve metabolic syndrome and normalize the obese ATME. In patients who are obese, gastric bypass surgery significantly reduces the number of infiltrating macrophages and the number of CLS in sub-cutaneous WAT, in concordance with downregulation of pro-inflammatory genes and upregulation of anti-inflammatory genes in the stromal-vascular fraction¹⁵⁰. Consistent with these changes, a retrospective study of 15,850 participants (7,925 had undergone gastric bypass compared with 7,925 matched controls), followed up fora mean period of 7.1 years, reported that gastric bypass surgery significantly reduced all-cause mortality by 40% in individuals who were obese (HR = 0.60) and reduced cause-specific mortality resulting from coronary artery disease (56%), diabetes mellitus (92%) and cancer (60%)¹⁵¹. Similarly, the Swedish Obese Subjects study, which prospectively assessed 4,047 people who were obese (2,010 had undergone various forms of bariatric surgery compared with 2,037 matched controls), followed up for a mean period of 10.9 years, reported that bariatric surgery significantly reduced all-cause mortality by 29% (HR = 0.71) in people who were obese after adjustment for sex, age and risk factors¹⁵². For cause-specific mortality, cancer-related mortality was reduced by 37% in the surgery group versus control group (relative risk (RR) = 0.63), almost double that of cardiac-related mortality (19% reduced risk; RR = 0.81)¹⁵². Importantly, it is not always clear from these studies whether bariatric surgery leads to reduced cancer-related mortality as a consequence of reduced cancer incidence or progression (whether a patient has an undiagnosed tumour at the time of surgery cannot be ruled out). However, a multisite retrospective study of 88,625 patients (22,198 patients who had undergone surgery and 66,427 participants who had not undergone surgery) with follow up between 2–9 years reported a 33% reduction in cancer diagnosis following bariatric surgery (HR = 0.67)¹⁵³, suggesting a probable role in cancer prevention.

Opportunities for lifestyle intervention.

The benefits of pharmacological or surgical interventions raise the possibility that lifestyle interventions might have similar effects. Ongoing clinical studies are testing the effects of body-weight loss on breast cancer recurrence through telephone-based coaching, which includes recommendations for calorie consumption and physical activity¹⁵⁴. At the molecular level, additional clinical studies have evaluated the effect of weight loss on obesity-associated metabo-inflammation, commonly through diet adjustments such as caloric restriction. In patients, a very-low-calorie diet leads to transcriptional changes in subcutaneous WAT that are indicative of reduced inflammation¹⁵⁵, as well as reduced levels of systemic biomarkers of inflammation and the metabolic syndrome¹⁵⁶. However, in individuals who were morbidly obese and underwent acute body-weight loss, CLS formation in subcutaneous WAT increased¹⁵⁶, suggesting inflammatory remodelling of the ATME. Of note, adipose inflammation is a physiological response to healthy adipose tissue remodelling¹⁵⁷, which simply becomes dysregulated during obesity, suggesting that this increase in CLS might be an important intermediary during rapid body-weight loss in order to establish tissue homeostasis. Supporting this notion, in preclinical models in which long-term caloric restriction is associated with weight stabilization, CLS in the mammary gland are reduced¹⁵⁸. How these effects influence cancer remains unknown; however, given that caloric restriction affects breast WAT, there are probably implications for tumour biology. Furthermore, whether all tumours will be equally responsive is unclear (BOX 3); for example, tumours might differentially respond to dietary restriction depending on the mutational status of the PI3K pathway¹⁵⁹.

Of course, adherence to a very-low-calorie diet is not sustainable (or recommended) outside of a controlled trial setting. Therefore, more manageable forms of dietary intervention have been explored, such as the Mediterranean diet, which encourages healthier food choices, such as olive oil over butter, herbs over salt, fish and poultry over red meat, and plenty of plant-based foods. In prospective studies, the Mediterranean diet improved indicators of the metabolic syndrome in patients who were obese, including reduced circulating inflammatory factors, improved insulin sensitivity and improved vascular function, with durable responses of up to 2 years^160,161. Importantly, in humans, a Mediterranean diet supplemented with extra-virgin olive oil led to reduced incidence of breast cancer compared with a control diet¹⁶². Whether the Mediterranean diet affects the composition of the breast ATME is unknown. Another sustainable dietary intervention that has been investigated is a low-carbohydrate diet compared with a low-fat diet. Several prospective clinical trials have reported that a low-carbohydrate diet is superior to induce weight loss compared with a low-fat diet, in association with improved lipid profiles and insulin sensitivity^163–165. Whether these dietary interventions are sufficient to improve cancer outcomes remains unknown.

To complement these dietary interventions, physical activity is another important lifestyle factor that seems to affect cancer¹⁶⁶. An analysis of pooled data from12 prospective trials, including a total of 1.44 million participants, demonstrated that self-reported leisure-time physical activity is associated with a reduced risk of 13 out of 26 different types of cancer, and in some cases these associations were independent of BMI¹⁶⁷. In patients who are overweight or obese who have survived breast cancer, moderate-to-difficult exercise (65–85% maximum heart rate, 3 times weekly for 16 weeks) reduced systemic biomarkers of the metabolic syndrome, including reduced levels of insulin, insulin-like growth factor 1 (IGF1) and leptin, and increased adiponectin, and this effect was sustained for the 3 month follow-up period¹⁶⁸. These data suggest that exercise might be beneficial not only for the general population, but also specifically for patients with breast cancer, in whom ≥5% weight gain post-diagnosis is associated with increased mortality¹⁶⁹. Another appealing strategy is to combine a weight loss intervention with increased physical activity to stimulate body-weight loss while maintaining muscle mass. However, whether these solutions will be sustainable, or whether lifestyle interventions will be sufficient to completely erase the effects of obesity, remains unclear (BOX 3). Studies are underway to determine whether lifestyle interventions can improve cancer outcomes¹⁷⁰.

Conclusions and future considerations

Given the striking associations between obesity, metabo-inflammation and cancer, preclinical and clinical studies that mechanically dissect these relationships are essential. We propose that patients who are obese will probably require special treatment considerations, such as diet and lifestyle counselling that is evidence-based, or adjuvant pharmacological interventions that target inflammation and the metabolic syndrome. These strategies will also be relevant to individuals who are not obese (BMI <30) who still display metabolic aberrancies, including adipose tissue dysfunction. Of note, an important consideration will be to develop cost-efficient, effective ways to identify individuals with excess body fat to supplement BMI measurements in routine medical practice, in order to more accurately identify those at risk. In particular, incorporating blood-based tests or body composition analyses in periodic medical examinations might help pinpoint individuals with inflamed adipose tissue who are within normal-range BMI¹⁷¹. Finally, disease management and therapeutic efficacy remain major obstacles for patients who are obese and have cancer, and experimental data suggest that this hurdle is not purely a dosing problem. This problem is particularly relevant for cancer immunotherapy, for which pre-existing chronic inflammatory conditions might trigger adverse effects. Although clinical studies in patients have reported an increased response to immuno-therapy in men who are obese and have melanoma⁷⁵, preclinical work has suggested that adverse effects can occur, particularly when obesity is compounded by ageing¹⁷².

Obesity could account for up to 20% of cancer-related deaths, or 1.6 million deaths annually on an international scale. If this link is in fact causal, we have an urgent responsibility as a scientific community to respond to these statistics by increasing research initiatives focused on metabo-inflammation.

Key points.

• Obesity is associated with increased cancer incidence and mortality.

• Substantial changes occur within the adipose tissue microenvironment (ATME) with body-weight gain.

• Metabolic and inflammatory changes related to the obese ATME contribute to cancer development and progression.

• Targeting adipose tissue dysfunction through pharmacological or lifestyle interventions might be useful for the prevention and treatment of cancer.

• Given the limitations of BMI as a measurement of adiposity, finding novel ways to identify individuals who are metabolically unhealthy with excess adipose tissue will be critical to pinpoint those at risk who might benefit from weight loss or other personalized interventions.

Glossary

Adipocyte hypertrophy

Enlargement of adipocytes, which often occurs in association with obesity and increased numbers of crown-like structures in adipose tissue.

Metabolically obese normal-weight

(MoNW). individuals within a normal-range BMi category (18.5–24.9), yet with a high body fat composition, leading to qualitatively similar health risks as individuals who are obese.

Adipose tissue microenvironment

(ATME). The cellular and structural compartment of adipose tissue, including but not limited to the adipocyte.

Pyroptosis

Inflammatory programmed cell death, in which an immune cell bursts to release intracellular contents into the microenvironment to trigger a rapid immune response.

Myelopoiesis

Differentiation of haematopoetic progenitor cells within the bone marrow towards a myeloid lineage.

Monocytosis

Expansion of monocytes within the peripheral blood, which are precursors for macrophages and dendritic cells.

Crown-like structures

(CLS). A dying or dead adipocyte surrounded by a ‘crown’ of macrophages within adipose tissue; this structure is a histological biomarker of obesity-associated inflammation and the metabolic syndrome.

Metabo-inflammation

inflammation within the adipose tissue microenvironment that has metabolic consequences irrespective of BMi status.

Benign breast disease

Heterogeneous group of lesions within the breast that might increase the risk of developing breast cancer.

CLS-associated macrophages

(CAMφ). Macrophages that are directly associated with crown-like structures (CLs) in inflamed adipose tissue; these cells are phenotypically and transcriptionally distinct from other macrophages within the adipose tissue microenvironment.

Myofibroblasts

Contractile cells of the mesenchymal-fibroblast lineage that synthesize extracellular matrix and mediate tissue remodelling.

Adipose stromal cells

(AsCs). Multipotent mesenchymal progenitor cells found in adipose tissue that can differentiate into mesoderm lineages (such as adipocytes, myofibroblasts, chondrocytes and osteoblasts); several terms have been used in the literature to refer to these cells (for example, adipose-derived stem cells, pre-adipocytes, adipose mesenchymal stem cells or lipoblasts).

Neutrophilia

High number of mature neutrophils within the peripheral blood or within tissues, resulting from neutrophil leukocytosis.