The Tumour Microenvironment

1. The Tumour as an Ecosystem — Beyond the Tumour Cell

Histologists looking at a solid tumour at low power see something striking: the cancer cells are a minority component. In pancreatic ductal adenocarcinoma they may comprise as little as 10% of the tumour mass; the bulk is fibrotic stroma. In breast and colon cancer, 30–70% of the cells in a typical biopsy are not malignant at all. Hanahan and Coussens (Cell 2012) made this the explicit conceptual revision of the original Hallmarks paper: the tumour is a heterocellular tissue.

The principal non-malignant cell populations of the tumour microenvironment (TME) are:

2. Hypoxia and HIF — the Master TME Signal

Solid tumours outgrow their blood supply. Cancer cells more than ~150 µm from a functional capillary fall below the diffusion limit of O₂ and become hypoxic (typically 0.1–1% O₂; physiological tissue is 4–7%). Hypoxia is not a passive endpoint — it is the central regulator of the TME, transcriptionally reprogramming both tumour and stromal cells through the hypoxia-inducible factor (HIF) family.

The HIF axis — an oxygen sensor

HIF is a heterodimer of an oxygen-labile α-subunit (HIF-1α, HIF-2α/EPAS1, or HIF-3α) and a constitutively expressed β-subunit (HIF-1β / ARNT). Under normoxia, two conserved proline residues in HIF-α (P402 and P564 in HIF-1α) are hydroxylated by the prolyl-hydroxylase domain enzymes (PHD1/2/3, EGLN1-3), a reaction that requires O₂ and α-ketoglutarate as cosubstrates. The hydroxyproline is recognised by the von Hippel-Lindau (VHL) E3 ubiquitin ligase complex, which polyubiquitinates HIF-α and targets it for proteasomal destruction. Half-life: about 5 minutes.

\[ \text{HIF-}\alpha + \text{O}_2 + \alpha\text{KG} \;\xrightarrow{\text{PHD2, Fe}^{2+}}\; \text{HIF-}\alpha\text{(OH)} + \text{succinate} + \text{CO}_2 \]

Under hypoxia the hydroxylation reaction cannot proceed, HIF-α is stabilised, translocates to the nucleus, dimerises with HIF-1β, recruits p300/CBP, and binds hypoxia-response elements (HREs) bearing the consensus 5′-RCGTG-3′. Several hundred genes are induced, including:

• Angiogenesis: VEGFA, ANGPT2, PDGFB.
• Glycolysis: GLUT1, HK2, PDK1, LDHA, MCT4 — the entire transcriptional program of the Warburg effect (Part V).
• pH regulation: CA9 (carbonic anhydrase IX), NHE1.
• EMT and invasion: SNAI1, TWIST1, LOX.
• Erythropoiesis (HIF-2α): EPO — the systemic response that earned the 2019 Nobel.
• Stem-cell self-renewal: OCT4, NANOG (HIF-2α).

VHL — the substrate-recognition hub

The von Hippel-Lindau tumour-suppressor protein (pVHL) is the substrate receptor of an elongin-B/elongin-C/Cul2/Rbx1 cullin-RING E3 ligase. It binds hydroxylated HIF-α via its β-domain and presents it to the ubiquitination machinery via its α-domain. Bi-allelic inactivation of VHL is the founding genetic event of clear-cell renal-cell carcinoma (ccRCC) and of haemangioblastomas in von Hippel-Lindau disease — in both, HIF-2α (and to a lesser extent HIF-1α) is stabilised even in normoxia, driving constitutive angiogenic and metabolic programs that explain the famously vascular phenotype of these tumours.

3. Angiogenesis — the Angiogenic Switch Revisited

Judah Folkman’s 1971 hypothesis — that solid tumours cannot grow beyond ~1–2 mm without recruiting their own blood supply — introduced angiogenesis to oncology. We covered the canonical pro/anti-angiogenic balance in Part I; here we examine the quality of the resulting vasculature, why it fails, and the therapeutic consequences.

VEGF and the angiogenic cascade

Vascular endothelial growth factor A (VEGFA) is a HIF-induced ~45 kDa disulphide-linked homodimer secreted by hypoxic tumour cells, CAFs, and TAMs. It signals primarily through VEGFR2 (KDR) on endothelial cells, triggering: tip-cell selection (via Notch / DLL4 lateral inhibition), basement-membrane degradation (MMPs), endothelial proliferation, and lumen formation. Other pro-angiogenic factors: bFGF, PDGF-BB, angiopoietin-2, IL-8.

Why tumour vessels are abnormal

Compared with normal capillary beds, tumour vasculature exhibits:

• Disorganised hierarchy — loss of the orderly artery → arteriole → capillary → venule structure.
• Tortuous, dilated vessels with chaotic branching and blind ends.
• Leakiness — wide inter-endothelial gaps (up to ~400 nm), loose tight junctions; underlies the enhanced permeability and retention (EPR) effect exploited by nanoparticle drugs.
• Pericyte deficiency — only ~10–30% pericyte coverage vs >90% in normal tissue; vessels are mechanically unstable.
• High interstitial fluid pressure — impaired lymphatic drainage + leaky vessels → 10–40 mmHg pressure (vs ~0 in normal tissue), opposing convective drug delivery.
• Perfusion heterogeneity — oscillatory flow, transient stasis → cycling hypoxia and reoxygenation injury.

Jain’s vascular-normalisation hypothesis

Rakesh Jain (MGH/Harvard) proposed in the early 2000s that anti-angiogenic therapy works not by destroying the vasculature but by normalising it — transiently pruning the most aberrant vessels, restoring pericyte coverage, lowering interstitial pressure, and improving oxygen and drug delivery to the surviving tumour. Thenormalisation window is typically a 1–2 week period after initiating VEGF blockade during which chemotherapy, radiotherapy, and (critically) cytotoxic T-cell infiltration are all enhanced. Beyond the window, excessive pruning produces ischaemia, hypoxia, and selection for a more aggressive phenotype.

Clinically, anti-VEGF agents include bevacizumab (humanised anti-VEGFA mAb; approved in CRC, NSCLC, glioblastoma, ovarian, RCC; combined with chemo), ramucirumab (anti-VEGFR2 mAb), and the receptor tyrosine-kinase inhibitors sunitinib, sorafenib, pazopanib, axitinib, cabozantinib, lenvatinib. Survival benefits in unselected populations are typically modest (months) — consistent with the evolutionary escape predicted by Jain’s framework, and motivating combinations with immunotherapy (Section 10).

4. Cancer-Associated Fibroblasts (CAFs)

CAFs are the dominant non-malignant cell population in many carcinomas, particularly the desmoplastic ones (pancreatic, breast, head and neck). They are not a uniform population — modern single-cell RNA-seq has revealed substantial functional heterogeneity.

Origins

CAFs arise from:

• Resident tissue fibroblasts activated by tumour-derived TGF-β, PDGF, FGF.
• Pancreatic stellate cells (PSCs) — the dominant source in PDAC; trans-differentiate from quiescent vitamin-A-storing cells to myofibroblastic α-SMA⁺ CAFs.
• Hepatic stellate cells in HCC, mesothelial cells in peritoneal metastases, adipocytes (via dedifferentiation) in breast cancer.
• Bone-marrow-derived mesenchymal stromal cells recruited via CXCL12-CXCR4.
• Endothelial-to-mesenchymal transition (EndMT) — minor contribution.

Heterogeneity — myCAF, iCAF, apCAF

Tuveson and colleagues (Öhlund et al., J Exp Med 2017; Elyada et al., Cancer Discov 2019) defined three principal CAF transcriptional states in PDAC:

Paracrine support of tumour cells

Beyond ECM deposition, CAFs supply:

• TGF-β — reinforces CAF activation (autocrine), induces EMT in tumour cells, suppresses cytotoxic T cells, promotes Treg differentiation.
• HGF — activates MET on tumour cells, drives invasion and a major route of resistance to EGFR-targeted therapy in NSCLC.
• IL-6 — STAT3 activation in tumour cells; survival, stemness, metabolic reprogramming.
• CXCL12 (SDF-1) — chemoattractant for CXCR4⁺ tumour cells; also excludes T cells from the tumour bed (Feig et al., PNAS 2013).
• Metabolic exchange — Lisanti’s “reverse Warburg”: CAFs glycolyse and export lactate / alanine that fuels OXPHOS in tumour cells (Part V).

5. Extracellular Matrix Remodelling

The tumour ECM is not the bystander tissue scaffold of normal organs. It is densely cross-linked, mechanically stiff, anisotropic, and a bona fide signalling hub.

Compositional changes

Tumour stroma is enriched in type I collagen(deposited by myCAFs), fibronectin with alternative splicing of the EDA/EDB domains, tenascin-C, versican, and hyaluronan (HA — especially in PDAC, where HA accumulation generates osmotic pressure that physically collapses vessels).

Cross-linking by lysyl oxidase (LOX)

LOX and the LOXL family (LOXL1–4) are copper-dependent amine oxidases that catalyse the oxidative deamination of lysine and hydroxylysine residues in collagen and elastin, producing reactive aldehydes that spontaneously condense into di-, tri-, and tetra-functional cross-links (pyridinoline, deoxypyridinoline). The result: insoluble, mechanically stiff fibres. LOX is a HIF target gene; hypoxic tumours secrete LOX into both the local ECM and (as we see in Section 9) into pre-metastatic organs.

\[ \text{collagen-Lys-NH}_2 + \text{O}_2 \;\xrightarrow{\text{LOX, Cu}^{2+}, \text{LTQ}}\; \text{collagen-Lys-CHO} + \text{NH}_3 + \text{H}_2\text{O}_2 \]

Matrix metalloproteinases (MMPs)

MMPs are zinc-dependent endopeptidases (24 in humans, including MMP-2 and MMP-9 gelatinases, MMP-1/8/13 collagenases, MMP-14/MT1-MMP membrane-type) that degrade ECM proteins. In tumours they enable:

• Basement-membrane breach and local invasion (MMP-2, MMP-9).
• Release of matrix-bound growth factors (latent TGF-β, VEGF, FGF).
• Generation of cryptic ECM fragments with biological activity (e.g. endostatin from collagen XVIII).
• Cleavage of cell-surface receptors (HER2, E-cadherin shedding → EMT).

Endogenous TIMPs (TIMP1–4) inhibit MMPs; the MMP/TIMP balance is pathologically shifted toward proteolysis. Broad-spectrum MMP inhibitors (marimastat, batimastat) failed in 1990s clinical trials — they are now understood to have inhibited ECM-protective MMPs alongside the pro-tumour ones.

Mechanical signalling — stiffness drives malignancy

Normal breast tissue has a Young’s modulus of ~150 Pa; invasive ductal carcinoma reaches 4–12 kPa — a 30–80× increase. Valerie Weaver and colleagues (Levental et al., Cell 2009; Paszek et al., Cancer Cell 2005) showed that ECM stiffening alone, in non-transformed mammary epithelial cells, activates integrin clustering, focal-adhesion assembly, FAK / Rho / ROCK / actomyosin contractility, and YAP/TAZ nuclear translocation — driving a malignant phenotype without any genetic alteration. Stiffness is acause of progression, not merely a consequence.

Aligned, tumour-associated collagen (TACS-3 in Provenzano’s classification) radiates outward from the tumour edge and provides “highways” for migrating cancer cells — visible by second-harmonic-generation microscopy and prognostic for outcome.

6. Immune Cells in the TME

The tumour immune infiltrate is heterogeneous, dynamic, and profoundly shaped by the non-immune compartments above. We summarise the principal players.

Effector cells

• CD8⁺ cytotoxic T cells (CTLs) — recognise tumour-derived peptide on MHC class I, kill via perforin / granzyme B and Fas/FasL. The presence of intratumoral CD8⁺ infiltrate (especially at the invasive margin, the “Immunoscore” of Galon et al., J Pathol 2014) is the single most powerful prognostic marker in colorectal cancer — outperforming TNM staging.
• CD4⁺ helper T cells — Th1 (IFN-γ, supports CTLs/M1 macrophages) is pro-tumour-rejection; Th2 (IL-4/13) and Th17 are context-dependent; T_FH support tertiary lymphoid structures.
• NK cells — kill MHC-I^lowtargets via missing-self recognition (KIR/NKG2A inhibitory receptors disengaged) and stress-ligand recognition (NKG2D for MICA/MICB, ULBPs). Crucial for picking up tumour cells that escape CTLs by HLA-I downregulation.
• Dendritic cells — cDC1s (XCR1⁺, BATF3-dependent) cross-present tumour antigen on MHC-I to CD8⁺ T cells in tumour-draining lymph nodes; their abundance correlates with checkpoint-blockade response.
• B cells & tertiary lymphoid structures (TLS) — intratumoral TLS predict immunotherapy response (Cabrita et al.Nature 2020; Helmink et al. Nature 2020).

Suppressor / pro-tumour cells

• Tumour-associated macrophages (TAMs) — the dominant myeloid population in many tumours. The classical M1/M2 dichotomy (Mantovani):
  • M1 (IFN-γ + LPS-activated): anti-tumour, iNOS⁺, IL-12, TNFα.
  • M2 (IL-4/IL-13/IL-10-activated): tumour-promoting, Arg1⁺, CD163, CD206; secrete VEGF, MMP-9, TGF-β; suppress T cells.
  In reality TAMs occupy a continuum; most tumours are dominated by M2-leaning phenotypes. High TAM infiltration correlates with poor prognosis in most epithelial cancers (exception: CRC, where it is variable).
• Myeloid-derived suppressor cells (MDSCs) — immature myeloid cells (granulocytic, PMN-MDSC; or monocytic, M-MDSC) that suppress T cells via Arg1, iNOS, ROS, and PD-L1; expanded by tumour-derived G-CSF, GM-CSF, IL-6.
• Regulatory T cells (Tregs) — FOXP3⁺ CD4⁺ CD25^high; suppress effector T cells via IL-10, TGF-β, CTLA-4, IL-2 sink, perforin, adenosine (CD39/CD73). Recruited by CCL22 from M2 TAMs. High intratumoral Treg/CD8 ratios predict poor outcome (e.g. ovarian, Curiel et al., Nat Med 2004).
• Tumour-associated neutrophils (TANs) — N1 (anti-tumour) vs N2 (pro-tumour, TGF-β-driven); release neutrophil extracellular traps (NETs) that capture circulating tumour cells and accelerate metastasis.

Hot vs cold tumours

Clinical immuno-oncology classifies tumours by the spatial pattern of CD8⁺T-cell infiltration: inflamed (“hot”)tumours have abundant intratumoral CTLs and respond to checkpoint blockade; immune-excluded tumours have CTLs trapped at the periphery (often by stromal CXCL12); immune-desert (“cold”) tumours have essentially no CD8 infiltrate and are largely refractory.

7. Cancer Immunoediting — Elimination, Equilibrium, Escape

The 19th-century observation by William Coley — that erysipelas infections could cause sarcoma regression — suggested that the immune system saw cancer. In 1957 Burnet and Thomas formalised this as the cancer immunosurveillance hypothesis. For decades the idea languished; it was resurrected in modern form by Robert Schreiber, Lloyd Old, and Mark Smythin their three-Es framework (Schreiber, Old & Smyth, Science 2011; original statement Dunn et al., Nat Immunol 2002): cancer immunoediting consists of three sequential phases.

8. Mechanisms of Immunosuppression in the TME

The escape phase deploys a coordinated set of biochemical mechanisms; modern immuno-oncology drug development targets each axis.

A. The PD-1 / PD-L1 axis — adaptive resistance

PD-1 (PDCD1) is an inhibitory receptor expressed on activated T cells. Its ligands PD-L1 (CD274) and PD-L2 (PDCD1LG2) are expressed on antigen-presenting cells, but also crucially upregulated on tumour cells by IFN-γ — an “adaptive” resistance mechanism: the very Th1 response that should kill the tumour induces its own brake. When PD-L1 engages PD-1 on the CTL, the cytoplasmic ITSM motif of PD-1 recruits SHP-2, which dephosphorylates CD28 and TCR signalling components, attenuating activation. Antibodies blocking this axis (pembrolizumab, nivolumab anti-PD-1; atezolizumab, durvalumab, avelumab anti-PD-L1) release the brake.

B. CTLA-4 — the priming-phase checkpoint

CTLA-4 (CD152) outcompetes CD28 for the costimulatory ligands B7-1 (CD80) and B7-2 (CD86) on dendritic cells, dampening priming in lymph nodes (and constitutively expressed on Tregs). Ipilimumab (anti-CTLA-4 IgG1) was the first checkpoint inhibitor approved (2011) for metastatic melanoma — a 10-year survival of ~20% in previously near-uniformly fatal disease. Mechanism includes both checkpoint blockade and Fc-mediated Treg depletion.

C. Tryptophan catabolism — IDO and TDO

Indoleamine-2,3-dioxygenase (IDO1) converts tryptophan to N-formyl-kynurenine. Local tryptophan depletion activates GCN2 in T cells (amino-acid starvation response → T-cell anergy); kynurenine is itself a ligand for the aryl-hydrocarbon receptor (AhR), driving Treg differentiation. IDO is expressed by tumour cells and DCs, induced by IFN-γ. Despite strong preclinical rationale, the phase III ECHO-301 trial of the IDO inhibitor epacadostat with pembrolizumab in melanoma was negative (2018) — a cautionary tale on TME drug development.

D. TGF-β — the master suppressor

TGF-β secreted by CAFs, Tregs, and tumour cells suppresses CTL function (downregulates perforin, granzyme), drives T-cell exclusion (Tauriello et al., Nature 2018; Mariathasan et al., Nature 2018), promotes EMT, and induces Tregs. Bintrafusp alfa (a PD-L1/TGF-βRII bifunctional fusion) and small-molecule TGF-βRI inhibitors (galunisertib) are in clinical development.

E. Adenosine — CD39 / CD73

Hypoxic dying cells release ATP; CD39 (ENTPD1, on Tregs and tumour cells) hydrolyses ATP → AMP; CD73 (NT5E) hydrolyses AMP → adenosine. Adenosine binding the A2A receptor on T cells raises cAMP and suppresses effector function. CD73 / A2A inhibitors are in clinical trials.

F. IL-10 — the canonical anti-inflammatory cytokine

Secreted by Tregs and M2 TAMs; suppresses MHC-II expression on APCs and Th1 polarisation. Pegylated IL-10 (pegilodecakin) showed early promise but failed in phase III pancreatic.

G. Loss of antigen presentation — revisited

As covered in Section 7, B2M loss, HLA LOH, and TAP/PRC1 deficiency are recurrent modes of escape, particularly under checkpoint-blockade pressure. Strategies to restore presentation include hypomethylating agents (azacitidine), HDAC inhibitors, and IFN combinations — some now in trials.

9. The Premetastatic Niche

In 1889 Stephen Paget proposed the “seed and soil” hypothesis to explain why metastases home to particular organs. The seed is the disseminated tumour cell; the soil is the receptive distal organ. For more than a century the molecular nature of soil preparation was opaque. In 2005, Rosandra Kaplan, Bethan Psaila, David Lydenand colleagues (Kaplan et al., Nature 2005) showed that the primary tumour actively conditions distant sites before any tumour cell arrives — the premetastatic niche.

The Lyden mechanism

In B16 melanoma and Lewis lung carcinoma mouse models, Lyden’s group observed:

• The primary tumour secretes systemic factors (VEGFA, PlGF, TGF-β, TNFα, lysyl oxidase, S100A8/A9) that act at distant target organs.
• VEGFR1⁺ bone-marrow-derived haematopoietic progenitor cells (BMDCs) are mobilised and home to specific pre-metastatic sites, where they form clusters even before tumour cells arrive.
• These BMDC clusters secrete fibronectin, chemokines (SDF-1/CXCL12), and induce stromal cells to upregulate adhesion molecules, generating a “landing pad” that subsequently attracts and retains circulating tumour cells.
• Antibody depletion of VEGFR1⁺ BMDCs prior to dissemination abolished metastatic colonisation — demonstrating that the niche is necessary, not merely permissive.

Tumour-derived exosomes — the organotropism code

Hoshino, Lyden, and colleagues (Nature 2015) showed that tumour-secreted exosomes (~30–150 nm extracellular vesicles) display a tissue-specific integrin code that determines organotropism:

• Exosomal α6β4 / α6β1 → lung tropism (binds laminin-rich alveolar epithelium and S100A4⁺ fibroblasts).
• Exosomal αvβ5 → liver tropism (binds Kupffer cells via fibronectin).
• Exosomal αvβ3 → brain tropism (vitronectin-rich neural microenvironment).

The exosomes are taken up by resident cells in the destination organ, transfer their cargo (mRNA, miRNA, protein), reprogramme local signalling, recruit BMDCs, and prepare the niche.

LOX in the lung pre-metastatic niche

Erler, Giaccia and colleagues (Cancer Cell 2009) showed that hypoxic primary breast tumours secrete LOX into the systemic circulation. LOX accumulates in the lung, where it cross-links basement-membrane collagen IV, recruits CD11b⁺ myeloid cells that release MMP-2 (further matrix remodelling), and creates a stiffened landing site for circulating breast-cancer cells. Anti-LOX antibodies blocked metastasis in this model.

Organotropism — the empirical pattern

Different cancers metastasise to characteristic organs:

• Breast → bone, lung, liver, brain (with subtype-specific patterns: ER⁺ → bone-dominant; HER2⁺ and triple-negative → brain).
• Lung adeno → brain, adrenal, contralateral lung, bone.
• Prostate → bone (osteoblastic).
• Colorectal → liver via portal vein, then lung.
• Pancreatic → liver, peritoneum.
• Renal → lung (cannonball metastases), bone.
• Melanoma → everything — lung, liver, brain, GI tract; the most promiscuous.

The pattern reflects circulatory anatomy (first-pass capillary capture: gut → liver, breast → lung), molecular compatibility (CXCR4 on tumour, CXCL12 in destination organ), and now — we know — pre-conditioning by exosomes and BMDCs. Part VII picks up the seed-side biology of metastasis.

10. Therapeutic Implications

The TME is now the dominant locus of cancer drug development. Strategies fall into several rational categories.