Metastasis

1. Metastasis is the Killer

The single most important fact in clinical oncology is this: a tumour confined to its tissue of origin is, in most cases, curable. A tumour that has seeded distant organs is, in most cases, not. Roughly 90% of cancer deaths are caused by metastases — not by the primary tumour itself. The primary lesion in the breast, lung, prostate or colon is rarely lethal in isolation; what kills the patient is liver failure from hepatic deposits, respiratory failure from miliary lung spread, fractures and hypercalcaemia from bone disease, or neurological collapse from brain metastases.

And yet metastasis remains the least molecularly tractable aspect of the disease. The hallmark activating invasion and metastasis (Hanahan & Weinberg, 2000;Part I) is the only one for which no targeted therapy has yet entered routine clinical use. The reason is that metastasis is not a single biochemical event but a sequence of improbable steps stitched together — a multi-stage cell-biological process whose individual rate constants we are only beginning to measure.

This Part traces the cascade biologically — cell-state transitions, cytoskeletal and ECM remodelling, vascular biology, immune interaction, organ-specific niches, and the dormant solitary cell — with an eye to where therapy might one day intervene.

2. The Invasion-Metastasis Cascade

The classical decomposition (Fidler 1970s; refined by Chambers, Groom & MacDonald 2002; Massagué & Obenauf, Nature 2016) breaks the journey from primary tumour to clinical metastasis into six obligatory steps:

The product rule applies. If each step has a probability \( p_i \) of success, the overall probability of completing the cascade is

\[ P_{\text{cascade}} \;=\; \prod_{i=1}^{6} p_i \]

Even if no individual step is rare, the product of six independent inefficient events is vanishingly small. Step 6 (colonisation) is by far the most stringent — orders of magnitude smaller than any other — and is where the most interesting and least understood biology lives. Most disseminated cells survive but never grow.

3. Epithelial-to-Mesenchymal Transition (EMT)

Carcinomas (cancers of epithelial origin — ~85% of human cancers) face a fundamental geometric problem. Epithelial cells are immobile by design: they form polarised sheets locked together by adherens junctions (E-cadherin), tight junctions (claudins, ZO-1), and desmosomes (desmogleins, plakoglobin), and they sit on a basement membrane they cannot cross. To invade, the cancer cell must dismantle this architecture and acquire mesenchymal traits — front-back polarity, motility, contractility, and the ability to navigate ECM — without entirely abandoning its epithelial identity. This developmental program is the Epithelial-to-Mesenchymal Transition (EMT), first described in gastrulation by Greenburg and Hay (1982), recognised in cancer by Thiery, Kalluri, Weinberg and others.

Molecular markers — the cadherin switch

The single most diagnostic event of EMT is the E-cadherin to N-cadherin switch: transcriptional silencing of CDH1 (E-cadherin) and induction of CDH2 (N-cadherin, normally a neural/mesenchymal cadherin). E-cadherin is the keystone of the adherens junction; its loss disassembles the junction, releases β-catenin into the cytoplasm (where it can enter the nucleus and act as a Wnt-pathway transcriptional co-activator), and licenses the cell to leave the epithelium.

EMT-inducing transcription factors

Three families of zinc-finger and bHLH transcription factors orchestrate the program. Each represses CDH1 at its E-box elements and activates a mesenchymal gene network. They are induced by TGF-β, Wnt, Notch, hypoxia (HIF-1α), and inflammatory cytokines from the tumour microenvironment (see Part VI).

Partial EMT — the hybrid E/M state

The textbook description of EMT as a binary switch was always an idealisation. Recent single-cell and lineage-tracing work (Pastushenko & Blanpain, Nature 2018; Yu et al. Science 2013; Aiello et al. 2018) has made it clear that tumour cells in the act of invading are typically in a hybrid epithelial/mesenchymal (E/M) state — co-expressing epithelial and mesenchymal markers, retaining some cell-cell cohesion while gaining motility, often migrating in collective clusters rather than as isolated cells.

The functional importance is striking. CTC clusters in patient blood are ~50× more efficient at seeding metastasis than single CTCs (Aceto et al., Cell 2014). Many of these clusters are cohesive groups of hybrid E/M cells, sometimes accompanied by neutrophils. The partial-EMT cell — not the fully mesenchymal cell — appears to be the most metastatically competent entity, plausibly because it retains the proliferative and colonising potential of an epithelial cell while acquiring just enough mesenchymal plasticity to invade and survive.

On arrival at a distant site, the cell typically undergoes the reverse transition — MET (mesenchymal-to-epithelial) — which is required for outgrowth into a clinically relevant secondary tumour (Korpal et al. 2011; Tsai & Yang 2013). EMT and MET are thus complementary halves of metastasis, not separate processes.

4. Invadopodia and ECM Degradation

EMT prepares a cell to leave the epithelium, but it must still cross the basement membrane — a dense sheet of laminin, type-IV collagen, nidogen and perlecan, ~50–100 nm thick. The specialised structures that perform this breach are invadopodia (Chen 1989; Linder, Wiesner & Himmel 2011) — ventral, actin-rich membrane protrusions that physically push into the matrix and locally concentrate proteases.

Anatomy of an invadopodium

• Actin core — branched (Arp2/3, cortactin, N-WASP) and bundled (fascin) F-actin generates the protrusive force; cofilin treadmilling supplies turnover.
• Tks5/Tks4 scaffolds — PX-domain proteins required to position invadopodia at sites of ECM contact (Seals et al. 2005).
• Adhesion ring — integrins (αvβ3, α6β1) and signalling kinases (Src, FAK, PYK2) anchor the structure.
• Protease delivery — vesicular trafficking of MT1-MMP (MMP-14), the master invadopodial protease, to the tip; secretion of soluble MMP-2 and MMP-9, and activation of MMP-2 by MT1-MMP/TIMP-2.
• Mechanical sensing — substrate stiffness biases formation; durotaxis along stiffness gradients (Levental, Weaver).

The matrix metalloproteinases (MMPs) are zinc-dependent endopeptidases. MMP-2 (gelatinase A) and MMP-9 (gelatinase B) cleave denatured collagen (gelatin) and type-IV collagen — both basement-membrane substrates. MT1-MMP additionally cleaves fibrillar collagen-I and activates pro-MMP-2. Because MMPs would otherwise digest normal tissue indiscriminately, they are tightly regulated: secreted as inactive zymogens (cysteine-switch), activated by furin or other MMPs, and inhibited by endogenous TIMPs (1–4).

5. Intravasation

Once across the basement membrane and through the stroma, the invading cell must enter the circulation. Most carcinomas favour blood vessels (haematogenous spread); some (notably breast, head & neck, melanoma) seed lymph nodes via lymphatics first.

Why intravasation is easier in tumours

• Aberrant tumour vasculature — tumour vessels are tortuous, dilated, leaky, and lack normal pericyte coverage. Endothelial junctions are loose; gaps of 100–1000 nm permit transmigration of cell-sized objects.
• Perivascular invadopodia (TMEM) — Tumour MicroEnvironment of Metastasis (Robinson et al., Sci Transl Med 2009) is a tripartite structure: tumour cell + perivascular macrophage + endothelial cell, where macrophage-derived VEGF locally permeabilises the vessel and the tumour cell intravasates.
• Macrophage chaperonage — tumour-associated macrophages (TAMs; Pollard, Nat Rev Cancer 2004) provide chemotactic EGF gradients along which tumour cells travel toward vessels (the EGF/CSF1 paracrine loop).
• Lymphatic intravasation — lymphatic capillaries lack tight junctions and a continuous basement membrane; access is mechanically easier. VEGF-C-driven lymphangiogenesis (Skobe, Alitalo) expands the entry surface.

The intravasation rate per unit time is enormous in absolute terms. Estimates from CTC enumeration (CellSearch and microfluidic platforms) and from kinetic modelling suggest a 1-cm³ tumour can shed 10⁴–10⁶ cells per gram per day into the blood. The bottleneck is not entry — it is everything that follows.

6. Survival in Circulation

Once in the bloodstream the tumour cell is in mortal danger. Three threats kill the vast majority of CTCs within hours:

Platelet cloaking — the trojan horse

Within seconds of entering the blood, tumour cells are coated by platelets attracted by tissue-factor-driven local thrombin generation (Trousseau’s syndrome — recognised since 1865 — is the clinical correlate). The platelet cloak:

• Physically shields the tumour cell from NK-cell recognition (the cloaked cell looks like a clot, not a foreign cell).
• Transfers MHC-I from platelets to tumour cell surface, providing “self” markers (Placke et al., Cancer Res 2012).
• Releases TGF-β from α-granules, reinforcing the EMT phenotype.
• Promotes arrest at the distant capillary bed by enlarging the cell-platelet aggregate beyond the vessel diameter.

This is why low-dose aspirin reduces metastatic incidence and cancer-specific mortality in long-term cohorts (Rothwell et al., Lancet 2010, 2012); platelet inhibition disrupts the cloak. Median CTC half-life in circulation is 1–2 hours, but the small subpopulation cloaked, clustered, and possibly stem-like is sufficient for the cascade to continue.

7. Arrest and Extravasation

The cell must now leave the circulation. Two not-mutually-exclusive mechanisms drive arrest:

• Mechanical trapping — tumour cells are 15–25 µm; pulmonary and hepatic capillaries narrow to 5–10 µm. The first downstream capillary bed mechanically halts the cell. This explains the dominant pattern of bowel-cancer liver metastases (drained by portal vein) and the prevalence of lung metastases from many primaries (drained into the systemic venous return).
• Active adhesion — selectins (P-selectin on activated platelets/endothelium; E-selectin on cytokine-stimulated endothelium) capture sialyl-Lewis-x glycans on tumour cells, a leukocyte-mimicking strategy. Integrin α4β1/VCAM-1, αvβ3/von-Willebrand-factor, and chemokine-receptor engagement (CXCR4-CXCL12, CCR7-CCL21) lock the arrest.

Once arrested, extravasation resembles a hijacked version of leukocyte transendothelial migration. The tumour cell extends invadopodia between or through endothelial cells, contracts endothelial junctions (VE-cadherin disassembly via Src signalling), degrades the underlying basal lamina with MT1-MMP and MMPs, and emerges into the parenchyma. In some organs (lung, brain) the cell may proliferate within the vessel briefly before extravasating; in others (liver) it may extravasate within minutes.

8. Organ Tropism — Seed and Soil

Not every primary tumour metastasises to every organ with equal probability. Breast cancer favours bone, lung, liver, brain. Prostate cancer favours bone (especially axial skeleton). Colorectal cancer favours liver. Uveal melanoma is almost pathognomonic for liver. Pancreatic cancer favours liver and lung. These patterns are not explained by blood-flow anatomy alone.

In 1889, the London surgeon Stephen Paget, analysing autopsy data from 735 women who had died of breast cancer, made the founding observation of metastatic biology:

“When a plant goes to seed, its seeds are carried in all directions; but they can only live and grow if they fall on congenial soil.” — Paget, The Lancet, 1889.

Paget’s seed-and-soil hypothesis held that tumour cells (seeds) reach many organs but only colonise tissues whose microenvironment (soil) supports their growth. This was challenged by Ewing in 1928 with a purely mechanical hypothesis (vascular drainage), but Paget’s view has been overwhelmingly vindicated by modern data.

Molecular basis of the soil

• Chemokine gradients — bone marrow, lung, and liver constitutively produce CXCL12 (SDF-1); breast-cancer cells expressing CXCR4 home to these sites (Müller et al., Nature 2001). CCR7 mediates lymph-node homing.
• Integrin–ECM matching — Hoshino et al. (Nature 2015) showed that tumour-derived exosomes carrying specific integrins (α6β4, α6β1 → lung; αvβ5 → liver) directly determine organ tropism by preconditioning the niche.
• Bone tropism — the “vicious cycle” of breast and prostate metastasis (Mundy 2002; Weilbaecher, Guise & McCauley, Nat Rev Cancer 2011). Tumour-secreted PTHrP and IL-6 stimulate osteoblasts → RANKL → osteoclast activation → matrix resorption → release of TGF-β and IGF-1 from bone matrix → further tumour growth. Targeted by bisphosphonates and denosumab.
• Liver as default sink — the high blood-filtering rate, fenestrated sinusoids, and Kupffer-cell-mediated “alarm” signals make the liver a uniquely permissive harbour for portal-system tumours.
• Tropism atlases — Massagué’s lab derived organ-specific gene-expression signatures by serial in vivo selection (Kang et al. 2003 bone; Minn et al. 2005 lung; Bos et al. 2009 brain) — identifying genes that govern site-specific colonisation.

9. Dormancy and Reawakening

The most clinically tragic feature of metastasis is its latency. Patients can remain disease-free for 5, 10, 20 years after curative resection of an apparently localised primary, then present with overt metastatic disease. Breast cancer relapses up to 30 years out; uveal melanoma recurs decades later in the liver; prostate cancer recurs in bone after long latency. The biological basis is tumour dormancy: disseminated cancer cells that survive in distant organs but do not proliferate, sometimes for years.

Two flavours of dormancy

The niche regulates the switch

Disseminated cells survive in specialised niches: the perivascular niche in bone marrow (Ghajar et al., Nat Cell Biol 2013) where endothelial-derived thrombospondin-1 enforces dormancy and TGF-β1 from sprouting vessels reawakens cells; the perisinusoidal hepatic niche; the brain perivascular space along laminin-rich basement membranes. Reawakening is triggered by:

• Inflammation — surgical wounding, infection, and lung injury awaken DTCs (Krall et al., Sci Transl Med 2018; Albrengues et al., Science 2018 — NETosis and neutrophil extracellular traps re-mobilise dormant cells).
• Vascular remodelling — new sprouting vessels deliver pro-proliferative cues (TGF-β1, periostin).
• Immune-checkpoint disequilibrium — immunosuppression releases the equilibrium phase of immunoediting.
• Hormonal cues — pregnancy, menopause, glucocorticoids modulate tropism and reactivation.

Adjuvant therapy in early-stage cancer is a wager that micrometastatic cells are still proliferating enough to be cytotoxic-sensitive. Once cells go dormant, they become radioresistant and chemoresistant (cycling-cell-targeting agents fail), and that is why decades-later relapses occur.

10. The Premetastatic Niche

One of the most surprising discoveries of 21st-century cancer biology is that the primary tumour begins preparing distant organs for metastatic seeding before any tumour cell arrives. Kaplan, Lyden and colleagues (Nature 2005) showed that VEGFR1⁺ bone-marrow-derived cells (BMDCs) accumulate in the future metastatic site — recruited by tumour-secreted factors travelling upstream through the circulation — and form clusters that subsequently recruit incoming tumour cells. Without this priming, tumour cells arriving at the site fail to colonise.

Components of the premetastatic niche

• Tumour-derived exosomes — ~30–150 nm extracellular vesicles carrying proteins, lipids, mRNA, miRNA. Hoshino et al. (Nature 2015) showed exosomal integrins are taken up by organ-specific resident cells (lung fibroblasts, liver Kupffer cells), inducing pro-inflammatory S100 proteins and remodelling the local stroma.
• BMDC recruitment — VEGFR1⁺ haematopoietic progenitors (Lyden); CD11b⁺Gr-1⁺ MDSCs; macrophage progenitors. Together these generate an immunosuppressed, ECM-remodelled microenvironment.
• Lysyl oxidase (LOX) — secreted by hypoxic primary tumours (Erler & Giaccia, Nature 2006); cross-links collagen-IV at distant sites, creating a sticky scaffold that captures BMDCs and incoming tumour cells.
• S100A8/S100A9 — alarmin cytokines induced in the future niche, attracting tumour cells and amplifying inflammation.
• Periostin and tenascin-C — stromal matricellular proteins required for metastatic colonisation (Malanchi et al., Nature 2011; Oskarsson et al., Nat Med 2011 — both Joan Massagué’s lab).

The premetastatic niche reframes metastasis as a systemic disease from very early on: by the time a tumour is clinically detectable, it has already been broadcasting signals altering the bone marrow, lung, and liver. This integrates elegantly with the immune and stromal reprogramming covered in Part VI: Microenvironment — the niche is a microenvironment that the tumour builds at a distance.

11. Therapeutic Implications

Metastasis is the deadliest aspect of cancer and the least therapeutically tractable. Why? Because most current cytotoxic and targeted agents were developed against the biology of the primary tumour — proliferating cells in a defined organ. Dormant disseminated cells in the bone marrow are a fundamentally different target. Several strategies are emerging:

For the broader therapeutic landscape, see Part VIII: Therapy.