Module 0: Evolution & Phylogenomics

The earliest unambiguous proboscidean, Eritherium azzouzorum from the Paleocene of Morocco, is only slightly younger than the K–Pg boundary. By the late Eocene (~37 Mya), Moeritherium— an amphibious, tapir-like animal roughly 230 kg — already exhibits the diagnostic second-incisor tusks and retracted nasal bones presaging the trunk. Stable isotope analyses of Moeritherium enamel (Liu et al. 2008) indicate an aquatic/marsh lifestyle, suggesting the trunk originated as a snorkel rather than as a feeding appendage.

The Oligocene and Miocene saw an explosive radiation of stem proboscideans: Deinotheriidae (with downward-curving mandibular tusks and a lineage persisting into the Pliocene), Gomphotheriidae (bunodont cheek teeth, four tusks, pan-continental distribution), and the Mammutidae (mastodons). These groups colonised Eurasia, North America, and — uniquely among terrestrial mammals of their size — repeatedly crossed between continents via Beringia and the Gomphothere Interchange of the Great American Biotic Interchange.

Elephantidae emerge

True elephantids — characterised by high-crowned (hypsodont) lamellar molars with horizontal tooth-replacement conveyor belts — appear in the late Miocene of Africa. The key transitional genus is Primelephas, whose fossils from the Lukeino Formation (Kenya, ~6 Mya) bracket the last common ancestor of the three extant elephant genera. From Primelephas three crown lineages diverge: Loxodonta (African bush and forest elephants), Elephas (Asian elephant), and the extinct Mammuthus (the mammoths).

Palaeoloxodon antiquus, the straight-tusked elephant of Pleistocene Europe and the Levant, is the largest terrestrial mammal with solid skeletal evidence — adult males exceeded 13 tonnes and 4 m shoulder height. Ancient DNA (Meyer et al. 2017) unexpectedly placed Palaeoloxodonnested within the African forest elephant clade (Loxodonta cyclotis), overturning previous morphology-based placements and revealing substantial inter-lineage gene flow before human arrival.

Shoshani & Tassy (1996) reviewed >400 proboscidean fossil taxa and proposed an elephantid topology based on morphology. The phylogenomic era began with Rohland et al. (2007, 2010), who sequenced nuclear loci from mammoth, mastodon, Asian and both African elephant lineages, demonstrating a deep genetic split between Loxodonta africana (savannah) and L. cyclotis (forest) of 2.6–5.6 Mya — comparable to the human/chimpanzee divergence and firmly establishing the forest elephant as a distinct species rather than a subspecies.

Roca et al. (2014) and Palkopoulou et al. (2018) subsequently performed genome-scale analyses across Loxodonta, Elephas, Mammuthus, Palaeoloxodon and mastodon, resolving the tree as:

Genome-wide introgression statistics (D, f₃, f₄-ratio) show that the straight-tusked elephant acquired up to ~40% of its genome from a forest-elephant-like source, a striking example of reticulated evolution within megaherbivores. Similar introgression was later detected between woolly and Columbian mammoths.

Coalescent theory for elephant TMRCA

Under the neutral Kingman coalescent, if a sample of n chromosomes is drawn from a panmictic diploid population of constant effective size Ne, the time to the most recent common ancestor (TMRCA) is a sum of independent exponential waiting times:

For elephants, generation time is long (\(\tau_g \approx 25\) years) and Ne estimates vary: Rohland (2010) inferred Ne_Loxodonta ≈ 40 000, Ne_Elephas ≈ 30 000 (with Pleistocene bottlenecks). The simulation below samples the joint coalescent for both lineages including a Late Pleistocene bottleneck for Elephas.

Every year roughly 20 000 African elephants are killed for their tusks. Wasser et al. (2015, Science) pioneered a stable-isotope and microsatellite-based geographic-origin mapping of seized ivory: large ivory shipments are typically drawn from two or three concentrated poaching hotspots rather than distributed continent-wide. The assignment framework uses a composite likelihood over carbon (\(\delta^{13}\mathrm{C}\)), nitrogen (\(\delta^{15}\mathrm{N}\)), strontium (\(^{87}\mathrm{Sr}/^{86}\mathrm{Sr}\)), and hydrogen (\(\delta^{2}\mathrm{H}\)) isotope ratios plus 16 microsatellite loci.

Hart et al. (2015) and Wasser et al. (2018) extended this approach using genome-wide SNP reference panels, achieving geographic assignment with\(\sim 300\) km precision for savannah elephants and finer resolution for forest elephants whose population structure is deeper. Elephantine ivory forensics is now admissible as evidence in U.S. federal wildlife-trafficking prosecutions.

TRPV1, the capsaicin and noxious-heat receptor (activated above ~43°C), normally occurs as a single gene in mammals. Weissenböck et al. (2010, 2012) showed that Loxodonta africanacarries five functional TRPV1 paralogs(TRPV1a–e), presumably duplicated in the early Loxodonta stem after the split from Elephas. Elephas maximus retains only the ancestral single copy.

The expansion is hypothesised to be an adaptive response to sustained savannah heat stress: higher copy number allows tissue-specific regulation of heat and pain sensation, particularly in the trunk tip and pharyngeal mucosa. The elephant skin-cooling system (covered in detail in M3) relies on exquisite thermal discrimination, and TRPV1 dosage may tune behavioural responses such as mud-wallowing and dust-bathing thresholds.

Peto (1977) noted that if cancer were simply a stochastic accumulation of somatic mutations, then large long-lived animals — with orders of magnitude more cells and cell divisions per lifetime — should suffer vastly higher cancer rates than small mammals. Empirically they do not. This discrepancy is Peto’s paradox.

Under the Armitage-Doll multistage model, a cell requires k independent driver mutations to transform. With mutation rate \(\mu\) per division and D divisions in a lifetime, the probability a given cell becomes malignant is approximately

and the organismal lifetime cancer probability is\(P_{\mathrm{org}} = 1 - (1 - p_{\mathrm{cell}})^{N_{\mathrm{cells}}}\). A 6-tonne elephant has ~6 × 10¹⁵ cells vs. 3 × 10¹³for a human — 200× more. Without compensation, elephant cancer incidence should be essentially 100%.

Abegglen et al. (2015, JAMA) resolved the paradox by demonstrating that Loxodonta africana carries ~20 retrotransposed copies of TP53(“LA-TP53” retrogenes) compared with a single copy in humans. Elephant lymphocytes exposed to ionising radiation or doxorubicin show roughly twice the apoptotic response of human cells, effectively culling incipient cancer cells before they proliferate. The retrogenes are transcribed, translated, and functionally active, though several lack introns and appear to act dominantly to sensitise the apoptotic machinery.

With n = 20 the per-cell cancer probability drops by a factor\(\eta^{19} \approx 4 \times 10^{-3}\), more than compensating for the 200× increase in cell count. Cancer incidence in zoo elephant necropsies is ~5% lifetime — lower than the 11%–25% reported in humans.

The ~20 TP53 retrogenes in Loxodonta (often called LA-TP53, for “L. africana TP53”) arose via LINE-1-mediated retrotransposition of the ancestral TP53 mRNA. Sulak et al. (2016) reconstructed the timing of these retrogene births by dating synonymous-site divergence between paralogs: the wave of duplications coincided with the onset of gigantism in the proboscidean lineage during the mid-Miocene (~15 Mya), consistent with strong positive selection for tumour suppression under increasing body-size pressure.

Biochemically, the retrogene products dimerise with the canonical full-length TP53, priming the apoptotic “trigger” at lower DNA-damage thresholds. Vazquez et al. (2018) further showed that a subset of the retrogenes encode a shortened “LIF6” (leukaemia inhibitory factor 6) transcript that localises to mitochondria and directly permeabilises the outer membrane in response to TP53 activation, providing an auxiliary apoptotic pathway absent in smaller mammals.

The net phenotype is that elephant cells preferentially undergo apoptosis rather than repair in response to DNA damage — a risk-averse strategy favoured when mutation-bearing clones carry the catastrophic risk of metastatic cancer in an individual with \(10^{15}\) cells. Mice and humans instead rely more heavily on repair pathways; the evolutionary trade-off mirrors the classical “disposable soma” model of senescence.

Because elephants live in strict matriarchal family units, mitochondrial DNA shows deep geographic structure: females rarely disperse, so mtDNA lineages are largely local while nuclear markers (via male dispersal) homogenise over longer timescales. Debruyne et al. (2003, 2007) identified two deeply divergent mtDNA clades (F and S) within Loxodonta africana with coalescence\(>\) 1 Mya — older than the nuclear split — likely reflecting incomplete lineage sorting and ancient introgression between forest and savannah populations.

Ancient DNA from mammoth specimens (Lynch et al. 2015; van der Valk 2021) revealed that the Pleistocene woolly mammoth (M. primigenius) underwent multiple severe bottlenecks, with Wrangel Island’s final refugium population (< 500 individuals) accumulating deleterious mutations indicative of terminal genomic meltdown before extinction ~4000 years ago.

The coalescent waiting time for mitochondrial lineages is a factor of four smaller than for nuclear autosomal lineages (haploid and maternally inherited, Ne,mt = Ne/4), making mtDNA a high-resolution probe of recent demographic history and an ideal tracer for matriline reconstruction within extant herds.

Current IUCN Red List assessments (2021) separate Loxodonta africana(savannah elephant, Endangered) and Loxodonta cyclotis (forest elephant, Critically Endangered) in recognition of the deep genetic divergence first documented by Rohland et al. Savannah elephants now number ~350 000 individuals; forest elephants ~100 000, with both populations declining ~30&percnt; and ~86&percnt; respectively over recent decades (Thouless et al. 2016; Poulsen et al. 2017).

Elephas maximus is sub-divided into three subspecies — E. m. maximus (Sri Lankan), E. m. indicus (mainland Asian), and E. m. sumatranus (Sumatran) — plus a reduced Bornean population of debated taxonomic status. Fernando et al. (2003) identified two mtDNA clades (α and β) with coalescence ~ 1.5 Mya, reflecting Pleistocene vicariance across the Sundaland corridor. Genome-wide STRUCTURE analyses confirm that mainland and Sri Lankan populations are genetically distinguishable at\(F_{ST} \approx 0.15\).

Population viability analyses (PVAs) for the smallest remnant populations — e.g. the Sumatran Way Kambas herd (~250 individuals) — project extinction probabilities above 50&percnt; within 100 years in the absence of gene flow augmentation. The long generation time (\(\tau_g = 25\) yr) and low reproductive rate (one calf every 4–5 years) compound the vulnerability.

Ancient-DNA sequencing from permafrost-preserved tissue has produced high-coverage woolly mammoth (Mammuthus primigenius) genomes (Palkopoulou et al. 2015; van der Valk 2021). The comparison with extant Asian elephant genomes reveals ~1.4 million single-nucleotide differences — of which roughly 2000 are candidate functional substitutions concentrated in genes for keratin, hair follicle development, circadian rhythm, insulin signalling, and lipid metabolism. Lynch et al. (2015) identified a suite of cold-adaptation fixations including EP300, PRPF8, and the haemoglobin β-chain variant that shifts the oxygen-binding P₅₀ to lower temperatures.

The haemoglobin adaptation is particularly elegant: Campbell et al. (2010) showed that three amino-acid substitutions in mammoth β-globin reduce the enthalpy of oxygen binding (\(\Delta H\)) so that erythrocytes release oxygen efficiently even at near-freezing peripheral temperatures. Re-expressing the mammoth haemoglobin in vitro confirmed the biochemistry.

The Wrangel Island population, the last mammoths to survive until ~4000 years before present, shows genomic hallmarks of inbreeding collapse: excess non-synonymous variants, loss of olfactory receptor function, and heterozygosity one-tenth that of mainland populations. This terminal “meltdown” provides a cautionary template for modern small populations of African forest elephant.

Node-age estimation in the elephant phylogeny relies on fossil-calibrated relaxed molecular clocks. The canonical calibration points are:

• Paenungulata crown (Proboscidea + Sirenia + Hyracoidea): 65 Mya, soft upper bound.
• Proboscidea stem: Eritherium, 60 Mya.
• Elephantidae crown (mastodon split): 25–30 Mya.
• Mammoth-Asian elephant split: 6.0 Mya (Rohland 2010).
• Loxodonta africana / cyclotis split: 2.6–5.6 Mya.

Using a strict molecular clock on autosomal intergenic SNPs, the per-site substitution rate in elephantids is\(\mu_{\text{auto}} \approx 1.2 \times 10^{-9}\) per site per year — roughly half the rodent rate, reflecting long generation time. The mitochondrial rate is about 20× higher, making mtDNA the workhorse for resolving sub-million-year splits.

Bayesian dating packages (MCMCtree, BEAST) implementing the above with relaxed clocks and fossil priors now routinely yield 95&percnt; HPD intervals below ±10&percnt; of point estimates for Miocene-bounded elephantid nodes.

Proboscidean evolution is a textbook example of iterative gigantism: across the 60 Myr record, many independent lineages — deinotheres, gomphotheres, mastodons, elephantids — converged on body masses above 5 tonnes. Each successful wave encountered the same physiological frontier: heat dissipation (Module 3), bone loading (Module 1), and reproductive constraints (Module 7). The recurring evolution of lamellar molars in Elephantidae is interpreted as an adaptation to C₄ grass expansion during late-Miocene aridification.

Key outstanding questions include (i) the precise timing of TP53 retrogene dosage increases relative to body-mass thresholds along the proboscidean stem; (ii) how much nuclear vs. mitochondrial introgression between Loxodonta, Palaeoloxodon and Mammuthus persisted through the Pleistocene; (iii) whether Elephas experienced a cryptic third species-level radiation on Sundaland; and (iv) the biomechanical and neurological consequences of the forest/savannah split for the divergent ear, tusk, and trunk-tip morphologies of the two Loxodonta lineages.

The simulations below make these questions quantitative: the coalescent framework allows formal hypothesis tests on TMRCA given any proposed Ne trajectory, and the TP53 dose-response model permits prediction of cancer incidence under hypothetical gene-editing interventions — whether for elephant welfare or, more controversially, for de-extinction of woolly mammoth-derived cold-adapted elephants.