Environment as a methylation layer read from structure

Lactase persistence is not a difference in the lactase gene's material: the LCT promoter reads the same γ in human and mouse, with |Δγ| = 0.034. The difference is a methylation substrate read from structure, a local CpG hotspot the single scalar hides, driving a discontinuous and hysteretic switch.

Environment is not an external scalar added to a switch; it is written onto DNA at specific CpG positions that are readable from sequence. At the lactase locus LCT is conserved (|Δγ| = 0.034, a negligible material difference), so persistence cannot be a γ difference; instead a local CpG hotspot (O/E 0.88, 2.6× the 2.5 kb mean) seeds a methylation layer M whose age-driven accumulation flips LCT off discontinuously. The silencing is hysteretic, matching the fact that human lactase is not re-inducible.

Persistence is a regulation question, not an inventory question

Every human carries the lactase gene LCT; the trait is only whether its switch stays on past weaning. The existing two-layer logic applies directly: the SET is conserved (LCT is present in human and mouse) and the material is conserved.

The LCT promoter reads γ = 1.315 in human and 1.350 in mouse, a difference |Δγ| = 0.034, far too small to carry the phenotype. So lactase persistence cannot live in the LCT material; it has to be set at the switch. This is the reading §4 already gives the snake ZRS, which is off at essentially the same γ as human (1.290 vs 1.264).

The difference lives in a layer the single scalar discards

Reducing a promoter to one number γ erases the dinucleotide CpG signal, which is where environment writes onto DNA. Read directly from structure, the LCT regulatory region holds a local CpG hotspot: a 200 bp window reaches O/E 0.88, which is 2.6× the 0.34 average across the surrounding 2.5 kb that a single scalar reports. The enrichment holds at 2.3-4.5× across 100-300 bp windows and peaks 46 bp from the LCT TSS, at the promoter itself.

The human locus also retains 1.8× more of this methylation substrate than mouse. Against a dinucleotide-preserving shuffle the bare magnitude is only modestly above chance (74th percentile), so what is structurally specific is the position, not the size: the hotspot sits at the promoter, where methylation gates the gene. This positional substrate is the designed exposure to the drive, read from structure rather than asserted.

The methylation layer M

γ stays read-only while a dynamic variable M, seeded by the measured CpG substrate and accumulating with age, lowers the switch's effective drive. When that drive crosses the γ-set spinodal the LCT state flips off discontinuously, the same bistability as every other switch but now driven by a structural, environment-coupled layer.

Two methylation trajectories on the same γ give opposite outcomes: a high age-rate silences LCT, while the persistence cis variant's low rate keeps it on. The material γ is blind to the difference; the methylation layer resolves it.

The silencing is hysteretic, and that is the honest payoff

Once methylation has flipped the switch off, lowering methylation does not turn it back on, because the baseline drive is itself sub-spinodal. This matches the biology: human lactase is not re-induced.

So the common observation that regular dairy improves tolerance is not lactase returning; it is a layer above the DNA switch, the colonic microbiome and physiological tolerance, which lies outside this model. The same machinery is trainable for a trait whose baseline drive clears the spinodal, so fixed versus trainable is a matter of sub- versus supra-spinodal baseline.

What stays open

Absolute methylation level, the absolute age of silencing, the rate magnitudes, and the above-DNA plasticity are Layer-2 open items, marked with their reasons. What is admissible from structure is which switch, the same-material check, the methylation substrate, and the direction of its tuning.