Fig. 3: Individual transcriptomes quantify caste canalization and its regulation by the JH signalling pathway.

a, Developmental potential scores (−1 < Δ<1) reflecting caste commitment for gyne (positive score) and worker (negative score) phenotypes across developmental stages of M. pharaonis as body length increases, corresponding to the representative images towards the right (courtesy L.P.). Δ values are based on normalized difference of transcriptomic distances between target individuals and an average gyne (calculated as the mean transcriptome across gynes within the same stage) and an average worker in the subsequent stage. A positive Δ value means that transcriptomic distance between a target individual and the average gyne in the next stage is less than between the target individual and the average worker in the next stage, so the target individual has a gyne-biased developmental potential. Caste identities were known from morphological characters, except for first instar larvae where they were inferred via BPA. b, Caste- and body length-specific expression of JH and ecdysone (20E) pathways, coloured according to gyne-biased (red) or worker-biased (blue) expression in third instar larvae (upper panel). The expression of jhamt and E93 showed both caste-specific (qualitative) and body size-specific (quantitative) thresholds (vertical dashed lines; estimated with a threshold regression model⁵⁷ (Methods)) (lower panel). Note that the downregulation of jhamt (terminating the JH biosynthesis) preceded the upregulation of E93 to actively initiate metamorphosis (see Extended Data Fig. 5a for expression patterns of other genes in this pathway). c, In third instar larvae of intermediate body length, feeding with JHA increased worker pupal body length (two-sided t-test; P = 0.0004; d.f. = 16.38; t = 4.44; n = 16 for both control and treatment groups), while precocene I decreased gyne pupal body length (two-sided t-test; P = 0.0001; d.f. = 50.70; t = −4.13; n = 30 and 28 for control and treatment groups, respectively). The box plot shows the median (centre line), 25% and 75% quartiles (boxes), outermost values (whiskers) and data points (jitter plot overlapping with boxes and whiskers).