(e) Analysis of cuticle thickness

Beetle elytron comprises a thick elytral dorsal cuticle (EDC) and a thin elytral ventral cuticle (EVC) separated by pillar-like trabeculae [27]. The EDC comprises a thin epicuticle underlain by a thicker exocuticle and, in turn, a thick endocuticle. For H. axyridis, EDC epicuticle, exocuticle and endocuticle are 0.2–0.4, 1.9–3.2 and 18.4–23.6 µm thick, respectively. The epicuticle and exocuticle comprise fine laminae (each ca 80–200 nm thick); the endocuticle comprises an orthogonal array of bundles of chitin fibres. Cuticular melanin is concentrated in the exocuticle [27,28].

Cuticle thickness was calculated as the sum of the thickness of the EDC and EVC. For each specimen, the complete longitudinal section was imaged (i.e. from the base to the apex) as a mosaic and the images were merged using Adobe Photoshop. Melanin-rich and melanin-poor cuticle regions were located on the SEM images by comparison with microphotographs of untreated elytra (electronic supplementary material, figure S2e). Cuticle (and, where possible, exocuticle) thickness was measured from the same location in all specimens matured at temperatures less than or equal to 350°C for each taxon to minimize measurement error, since cuticle thickness varies from the base to the apex of the elytron, especially in P. javana.

Differences in cuticle thickness were tested for significance using PAST (PAleontological STatistics v. 4.0) [29] and RStudio for the following sample pairs: (a) melanin-rich and melanin-poor regions of the same elytron; (b) untreated and matured elytra of the same taxon; and (c) elytra (of the same taxon) matured at different temperatures. Note that for (b) and (c), data for melanin-rich and melanin-poor cuticular regions were tested separately. The data for each test pair were tested for normality using the Shapiro–Wilk test. For normal data, differences in variance and mean value between samples were assessed using a two-tailed F-test and the t-test, respectively. An unequal variance t-test, also known as the Welch test, was used in place of the basic t-test where variances are significantly different. For non-normal data, a two-tailed Mann–Whitney U test was used to compare the medians for each test pair.