Genetics & Molecular Drivers

1. Cytogenetics: a primer

Three diagnostic platforms now routinely interrogate the leukaemic genome:

• Karyotype (G-banded metaphases) — pan-genome view at low resolution (~5 Mb). Detects translocations, deletions >5 Mb, ploidy changes, and clonal evolution. Requires dividing cells; ~5–7 day turnaround.
• FISH — targeted fluorescent probes; faster (~24 h), higher resolution, can be done on interphase cells. Standard panels: BCR-ABL1, KMT2A, RUNX1::RUNX1T1, CBFB::MYH11, PML::RARA, IGH break-apart for B-ALL, −5/del5q, −7, +8, del(17p), del(11q), +12, del(13q) for CLL.
• Molecular (NGS / RT-PCR) — targeted-panel DNA sequencing of ~30–500 leukaemia-relevant genes; RNA-seq for fusion detection (especially Ph-like ALL).

Cytogenetic notation is concise. 46,XX,t(9;22)(q34;q11) means a female karyotype with a balanced translocation between chromosomes 9 and 22 at the indicated cytobands. 47,XY,+8 means trisomy 8 in a male. der(7) is a derivative chromosome 7. idic(X)(p11) is an isodicentric. complex karyotype in AML is ≥3 unrelated abnormalities (excluding favourable changes); monosomal karyotype is ≥2 monosomies or 1 monosomy + 1 structural abnormality. Both confer adverse risk.

2. The Philadelphia Chromosome — First of its Kind

In 1960 Peter Nowell (a haematologist at the University of Pennsylvania) and David Hungerford (a graduate student at the Fox Chase Cancer Center) independently observed an abnormally small chromosome in CML cells — the Philadelphia chromosome, named for the city in which they worked. Their joint paper (Science 1960) was the first description of a consistent cytogenetic abnormality in any human cancer.

For 13 years the Ph chromosome was assumed to be a deletion. In 1973 Janet Rowley, using the new Q-banding technique, showed it was the result of a balanced reciprocal translocation t(9;22)(q34;q11). Through the 1980s, several groups (Heisterkamp, de Klein, Groffen, Witte) cloned BCR and ABL1 and showed the fusion transcript and protein. The cancer field had its first molecularly-defined oncogene and a template for everything to come.

3. Translocations & Fusion Proteins

Recurrent translocations dominate paediatric and core-binding-factor leukaemias. Three mechanisms are recurrent:

• Constitutively active kinase — BCR::ABL1, ETV6::RUNX1, NUP214::ABL1, EBF1::PDGFRB, KIF5B::FLT3.
• Aberrant transcription factor — PML::RARA (locks differentiation), RUNX1::RUNX1T1 (dominant-negative core-binding-factor), CBFB::MYH11, KMT2A::AFF1, KMT2A::MLLT3, ETV6::RUNX1, TCF3::PBX1, DEK::NUP214, NUP98 fusions.
• Cytokine-receptor / immunoglobulin enhancer juxtaposition — IGH::CRLF2, IGH::MYC (Burkitt), IGH::CCND1 (mantle-cell), IGH::BCL2 (follicular).

The major recurrent fusions and their diseases:

Single fusions are rarely sufficient. Mouse models repeatedly show that PML-RARA, AML1-ETO, and BCR-ABL alone produce a chronic, slow-onset disease; cooperating mutations (FLT3, KIT, NRAS, KMT2D) drive transformation to acute leukaemia. The two-hit model of leukaemogenesis — class I (signalling) plus class II (differentiation/transcription) — was articulated by Gilliland (Curr Opin Hematol 2002) and remains a useful schematic for AML.

4. Mutational Landscapes by Disease

Whole-genome and whole-exome sequencing across >10,000 leukaemia samples (TCGA, Beat AML, MDS-WG) has produced detailed maps. The recurrent driver gene catalogues by disease:

Background point-mutation rates in leukaemias are 0.5–2 per Mb — lower than most solid tumours (which are 5–50/Mb). The corollary: leukaemias are largely driven by a small number of high-impact lesions, and immunotherapy strategies that rely on neoantigen burden (checkpoint inhibitors) work less well than in mutation-rich cancers. CAR-T and bispecific T-cell engagers, which exploit lineage antigens regardless of mutational burden, are dominant.

5. Epigenetic Drivers — the DNA-Methylation Machinery

An astonishing fraction of myeloid neoplasms is driven by mutations in the DNA-methylation and chromatin-modification machinery:

• DNMT3A — encodes a de novo DNA methyltransferase. Hot-spot R882H mutation impairs catalysis; ~25% of AML; one of the earliest CHIP drivers (decades pre-leukaemia).
• TET2 — α-KG-dependent dioxygenase that hydroxylates 5-methylcytosine to 5-hydroxymethylcytosine, the first step of active demethylation. ~10–20% of AML, ~50% of CMML; pure CHIP driver.
• IDH1, IDH2 — arginine-substitution mutations gain a neomorphic activity producing 2-hydroxyglutarate, which competitively inhibits TET2 and other α-KG-dependent dioxygenases. Inhibitors (ivosidenib R½CONSOLE; enasidenib) reverse this.
• ASXL1 — PRC2-associated chromatin regulator; truncating mutations confer poor risk in AML, MDS, CMML.
• EZH2 — PRC2 catalytic subunit (H3K27 methylation); loss-of-function in MDS, T-ALL.
• KMT2A (MLL1) — H3K4 methyltransferase. Translocations (rather than mutations) drive AML and infant ALL; menin-MLL inhibitors (revumenib FDA 2024) target the menin-KMT2A interaction.
• EZH2, BCOR, KDM6A — further chromatin-related drivers.

The fact that hypomethylating agents (azacitidine, decitabine) are useful in MDS and elderly AML — despite weak single-agent activity in solid tumours — reflects the centrality of DNA methylation to the myeloid programme. Their synergy with venetoclax (VIALE-A; Pollyea, NEJM 2020) gave us the standard-of-care for unfit elderly AML patients.

6. Splicing-Factor and Cohesin Complex Mutations

An entirely new class of cancer driver was discovered in 2011 in MDS: recurrent, mutually exclusive, hot-spot mutations in SRSF2, SF3B1, U2AF1, and ZRSR2 — all components of the spliceosome (Yoshida et al., Nature 2011; Papaemmanuil et al., NEJM 2011).

• SF3B1 K700E — defines MDS with ring sideroblasts (MDS-RS); causes use of cryptic 3′ splice sites in target genes (notably ABCB7).
• SRSF2 P95H/L/R — alters exonic splicing-enhancer recognition; common in CMML and MDS-AML.
• U2AF1 S34F/Y, Q157 — alters 3′ splice-site selection; found in MDS, AML.
• ZRSR2 — X-linked; alters minor (U12) spliceosome.

All of these are heterozygous, mutually exclusive, and confer dependency on the residual wild-type splicing machinery — making spliceosome modulators (H3B-8800, E7107) a candidate therapeutic strategy.

Cohesin-complex mutations (STAG2, RAD21, SMC1A, SMC3) occur in ~10% of AML/MDS and disrupt sister-chromatid cohesion and 3D-genome organisation. They define an MDS-related AML subgroup with adverse risk in the 2022 ELN scheme.

7. NOTCH1 and the Logic of T-ALL

Activating mutations in NOTCH1 are present in >55% of T-ALL — the dominant signalling driver of the disease (Weng, Ferrando, Look, Aster, Science 2004). They cluster in two domains:

• Heterodimerisation domain (HD) — mutations destabilise the autoinhibited form of NOTCH1, allowing γ-secretase cleavage in the absence of ligand.
• PEST domain (C-terminal) — loss-of-function mutations remove the FBXW7 docking site, prolonging the half-life of activated NOTCH1 intracellular domain (NICD).

Loss-of-function mutations in FBXW7 (the ubiquitin ligase that targets NICD) phenocopy PEST-domain NOTCH1 mutations. The two are partly mutually exclusive.

γ-secretase inhibitors (GSIs) showed early promise for NOTCH-dependent T-ALL but ran into severe gut toxicity (goblet-cell metaplasia from on-target effects on intestinal-stem-cell NOTCH). They are still in development but have not become standard. The take-away: knowing the driver does not always yield a tractable drug.

8. CHIP, CCUS, and Pre-leukaemic Clones

Three overlapping pre-leukaemic states sit on the path to overt myeloid neoplasia:

Mathematically, the probability that a healthy adult harbours a clone with a given driver scales with HSC count, mutation rate, and time:

$$\;\Pr(\text{leukaemia} \mid \text{driver}) \;\propto\; \binom{N_{HSC}}{1}\,\mu\,t\;,$$

This is essentially the explanation for why CHIP rises so steeply with age. Once a clone exists, its further expansion is governed by its selective coefficient $s$ — the per-division advantage over wild-type HSCs. The clone’s size grows approximately as

$$\;\frac{dN}{dt} = (s)\,N\;\;\Rightarrow\;\;N(t) = N_0\,e^{s t}\,.\;$$

Single-cell phylogenies estimate $s$ at ~5–15%/yr for DNMT3A, ~5%/yr for TET2, and substantially higher (~20–30%/yr) for splicing-factor and TP53 mutations — explaining their over-representation in MDS/AML relative to their frequency in CHIP.

9. Clonal Architecture & Evolution

At single-cell resolution, AML and MDS are not monoclonal but oligoclonal: a founder clone with “type-1” mutations (DNMT3A, TET2, ASXL1) gives rise to subclones with cooperating drivers (FLT3, NPM1, NRAS). Patterns recur:

• Linear evolution — serial acquisition along one trunk; common in CML chronic-phase progression.
• Branching — multiple subclones from a common ancestor; AML and ALL relapse often arise from a minor subclone present at diagnosis (Mullighan, Science 2008; Ding, Nature 2012).
• Clonal sweep — one subclone becomes dominant under selection (e.g. TP53 mutant under chemotherapy).
• Therapy-driven evolution — novel mutations emerge under selection: BTK-C481S under ibrutinib, ABL-T315I under imatinib, BCL2-G101V under venetoclax (Tausch, Blood 2019).

Practically: a single time-point snapshot of mutational status is incomplete. Modern MRD strategies use NGS to track multiple subclonal mutations simultaneously; the re-emergence of any subclone above a defined threshold heralds relapse weeks to months before frank cytopenias. We will see the practical readouts in Part VI and Part VIII.

A Closing Structure: ABL T315I

The single most consequential point mutation in the leukaemia molecular clinic is ABL T315I, the gatekeeper mutation that renders BCR-ABL refractory to all four first/second-generation TKIs. The structure of ABL kinase containing T315I (PDB 3IK3, with ponatinib bound) shows how the isoleucine fills the gatekeeper pocket and how ponatinib’s alkyne linker accommodates it.