Genomic Context Rather than Counting Mode or Genotype Sets the Sign of Repeat-Layer Non-Additivity in Citrus

Whether the transposable-element layer responds non-additively to combined abiotic stress, and at what resolution that can be measured, has not been established. Whether it extends into the repetitive fraction is unknown, because a family-level effect may be an artefact of the repeat library that defines it. In a public citrus factorial, two counting rules discarding different information agreed closely (r = 0.961); resolution dependence proved a property of the library, not the plant. In one genotype, two-way non-additivity passed q < 0.05 and replicate consistency in 675 of 8106 tests (8.3%), sub-additive (391 against 284, p = 4.4 × 10−5); that excess has two opposite components: induced families lean sub-additive (93.5%), repressed families super-additive (91.5%). The gene set was non-additive in 19,508 of 61,836 (31.5%), but its direction was balanced once collapsed per gene (p = 0.52). The genotype modulated the repeat-layer interaction in 551 tests without a stable direction, and the gene-level interaction far more often (16,973 tests). The repeat layer therefore carries non-additivity but no reproducible direction: it loses its sign when restricted to intergenic insertions, while unique-only counting and the second genotype reduce its magnitude without changing its sign, so the sign is set by genomic context alone.