Miniaturisation profoundly alters animal morphology, particularly locomotory structures like insect wings. Larger insects possess membranous wings optimised for flight dominated by inertial forces, while microinsects have highly derived bristled wings with a reduced membrane, adapted to viscous interactions. Distantly related microinsects share striking similarities in some aspects of wing architecture, such as elongated bristles or narrowed wing blades, while features such as venation or proportion of bristled wing area vary widely. The basis of these differences remains unknown. Although insect wing morphology is largely shaped by evolutionary history, the role of evolutionary constraints in macro-to-micro wing transition has not been examined. For the first time, we combined morphological analyses with evolutionary modelling to explore how selection for wing optimisation during miniaturisation is constrained by evolutionary inertia in key wing features. Analysing 39 bark beetle species, ranging greatly in size, we found that some modifications, like bristle elongation or wing narrowing, exhibit very low evolutionary constraints, enabling rapid adaptation to miniaturisation. In contrast, traits like venation development or bristled area proportion were highly constrained, requiring longer evolutionary timescales to adapt. Our findings provide novel insights into the origins of wing‑architecture diversity in microinsects, emphasising the role of evolutionary constraints in modulating the transition from macro‑ to micro‑wings.

Studied taxa

We analysed 39 species of bark beetles and pinhole borers (Curculionidae: Scolytinae, Platypodinae), representing a broad range of body sizes, from 8.7 mm to 1.2 mm in length. Whenever possible, both wings from at least two males and two females per species were examined. However, the count of marginal bristles along the wing edge was based on only two specimens per species. Specimens were randomly selected from the insect collection at the Department of Forest Ecosystem Protection, Faculty of Forestry, University of Agriculture in Krakow.

Sample preparation and morphological measurements

Specimens were identified to species and sex [19–21] before analysis. Each individual was photographed using the Keyence VHX-7000 4K high-accuracy digital microscope (Keyence, Japan). Wings were carefully removed from each specimen using microsurgical tools, washed in pure ethanol, placed on a microscope slide, gently straightened in a drop of ethanol using a soft microbrush, and preparations were made for imaging. Wing preparations were subsequently photographed using the VHX-7000 microscope. Body length (from the top of the pronotum to the apex of the elytra), wing length, and wing width (see Fig. 1a for details) were measured based on the obtained body and wing images. Subsequently, seven parameters of wing architecture previously considered to be strongly affected by miniaturisation were analysed:

 1)     Bristled area – The percentage of the bristled area relative to the total wing area (see Fig. 1b).

2)     Bristle length – The mean length of marginal bristles at the central region of the wing (zones 2/4 and 3/4, see Fig. 1b). Although bristle length was measured in four wing zones (see Fig. 1b), zones 2/4 and 3/4 were chosen for analysis as they were the most representative for the analyzed groups. Bristles occur in this region even in larger species, allowing for the observation of gradual changes with decreasing size.

3)     Bristle number – The total number of bristles divided by the wing length (to obtain a size-independent relative variable).

4)     Bristled region percentage – The percentage of the bottom wing edge (trailing edge) covered with bristles relative to the total length of the bottom wing edge (see Fig. 1b).

5)     Wing aspect ratio – The ratio of wing length to wing width (see Fig. 1a), see [22].

6)     Veins divergence – Measured as the distance between the medial posterior and cubital veins (including the cubitoanal strut), calculated by subtracting distance A from distance B (see Fig. 1a). Both distances A and B were divided by wing length to obtain a size-independent relative variable.

7 Veins length – Measured as the relative sum of vein length (sum of vein length divided by wing length), see Fig. 1a.

Measurements were conducted using Digimizer v. 6.3.0 software (MedCalc Software Ltd, Belgium).

Phylogenetic reconstruction

We started with a genus-level phylogenetic topology from the literature [23], which was based on five genes (COI, EF-1a, 28S, CAD, ArgK). All genera not represented in our data set were pruned from this phylogeny before missing genera were coded in based on other published phylogenies [24]. Genera with only one representative species in our data were assigned the appropriate species names. Genera* *with two species representatives were added as sister species; while the relationship between species from genera with more than two species representatives was obtained from the literature [25–29], before coding these manually into our phylogeny.

The node heights of the phylogeny were inferred by analysing COI sequences downloaded from BOLD in BEAST v2.7.6 [30]. Cryphalus intermedius sequences, of which none were published in BOLD, were obtained from the authors of the publication of genus Cryphalus taxonomy [27]. Only the node heights were sampled during the BEAST analysis – the topology, which was obtained from the literature, remained fixed. Based on analysis of our alignment in ModelTest-NG [31], we used a GTR site model, with gamma category count set to four, and without estimation of invariant sites. The analysis was performed with a Yule tree prior and a relaxed molecular clock.

Phylogenetic comparative analyses

For our phylogenetic comparative analysis of wing traits, we applied the SLOUCH (Stochastic Linear Ornstein-Uhlenbeck Models for Comparative Hypotheses) method [32], implemented in R (4.4.1) [33]. SLOUCH generates two models:

The null model (intercept-only model): This model does not include any predictor variables and assesses the phylogenetic effect on a trait. It estimates how long it takes for a trait to lose half of its correlations with the ancestral trait value. If the intercept-only half-life (measured in tree length units) is zero, the trait is not influenced by ancestral conditions. An intercept-only half-life greater than zero indicates that ancestral influences persist, with the strength of this effect proportional to the intercept-only half-life value. For example, a half-life of 0.5 implies that half of the evolutionary history of the group is needed for the trait to lose half of the correlation with the ancestral trait value. If the intercept-only half-life exceeds 30 times the phylogeny length, the phylogenetic effect is so strong that the model approximates Brownian motion (diffusion with negligible pull towards an optimum).

The regression model: This model includes a predictor variable (selection driver), assuming the trait (response variable) evolves toward an adaptive optimum under selection. The adaptive optimum is the state where the trait achieves perfect adaptation to the predictor (selection driver; in our case, it is the body size regime), ancestral constraints have disappeared, and selection pressure from this particular driver drops to zero. In our specific case, the adaptive optimum for a given wing parameter (e.g., bristle length) is the state in which that parameter is perfectly adapted (i.e., bristles have the optimal length) to the prevailing body-size regime (with body size acting as the selection driver), so that no further selection for optimisation of bristle length to body size occurs. Any delay in reaching this adaptive optimum is measured by the phylogenetic half-life, indicating how long it takes for the trait to evolve halfway toward adaptation. A phylogenetic half-life of zero means the trait adjusts instantly to the given predictor. A phylogenetic half-life of 1 implies that the full evolutionary history of the group is required for halfway adaptation to the adaptive optimum. The larger the half-life estimate, the stronger the phylogenetic inertia. The regression model produces an evolutionary and optimal regression slope. The evolutionary regression slope shows the observed relationship between the predictor and response variables, while the optimal slope shows the expected relationship in a scenario with no constraints on adaptation, i.e., the optimal adaptive values in the response variable as a function of the predictor.

Since the half-life parameter in SLOUCH is expressed as a proportion of phylogeny age, we set the root depth of our phylogeny to 1. This means that all estimates of half-life can be read as a percentage of the group’s evolutionary history. We also supply half-life parameters in Myr; the root age of our phylogeny was set to 105 Myr, which is based on the best available estimate for the split between Scolytinae and Platypodinae [34]. The root age used corresponds closely with a previously published phylogenetic study, which suggests ~100 Myr [35].

The null model and the regression model, for each of the seven wing traits (with body size [body length] as the singular predictor variable, i.e., the selection driver), were compared using δAICc (differences in the Akaike Information Criterion corrected for small sample size). When a regression model had an AICc score >2 points smaller than the corresponding null model, the regression model was considered statistically significant [36].

References:

19.      Vega FE, Hofstetter RW. 2015 Bark Beetles: Biology and Ecology of Native and Invasive Species. Oxford: Elsevier. (doi:10.1016/C2013-0-04752-4)

20.      Nunberg M. 1981 Klucze do oznaczania owadów Polskich Część XIX Chrząszcze - Coleoptera. Korniki – Scolytidae, Wyrynniki – Platypodidae. Warszawa: Polskie Towarzystwo Entomologiczne.

21.      Grüne S. 1979 Brief illustrated key to European bark beetles. Hannover: Verlag M. & H. Schaper. 182 pp.

22.      Bhat SS, Zhao J, Sheridan J, Hourigan K, Thompson M. 2019 Aspect ratio studies on insect wings. Phys. Fluids 31, 121301. (doi:10.1063/1.5129191).

23.      Gohli J, Kirkendall LR, Smith SM, Cognato AI, Hulcr J, Jordal BH. 2017 Biological factors contributing to bark and ambrosia beetle species diversification. Evolution 71, 1258–1272. (doi:10.1111/EVO.13219)

24.      Pistone D, Gohli J, Jordal BH. 2018 Molecular phylogeny of bark and ambrosia beetles (Curculionidae: Scolytinae) based on 18 molecular markers. Syst. Entomol. 43, 387–406. (doi:10.1111/SYEN.12281)

25.      Cognato AI. 2000 Phylogenetic analysis reveals a new genus of ipini bark beetle (Scolytidae). Ann. Entomol. Soc. Am. 93, 362–366. (doi:10.1603/0013-8746(2000)093[0362:PARNGO]2.0.CO;2)

26.      Smith SM, Cognato AI. 2014 A taxonomic monograph of Nearctic Scolytus Geoffroy (Coleoptera, Curculionidae, Scolytinae). *ZooKeys *450, 1–182. (doi:10.3897/ZOOKEYS.450.7452)

27.      Justesen MJ, Hansen AK, Knížek M, Lindelöw Å, Solodovnikov A, Ravn HP. 2023 Taxonomic reappraisal of the European fauna of the bark beetle genus *Cryphalus *(Coleoptera, Curculionidae, Scolytinae). *ZooKeys *1179, 63–105. (doi:10.3897/ZOOKEYS.1179.101388)

28.      Pernek M, Avtzis DN, Hrašovec B, Diminic D, Wegensteiner R, Stauffer C, Cognato AI. 2008 Novel morphological and genetic markers for the discrimination of three European *Pityokteines *(Coleoptera: Curculionidae: Scolytinae) species. Period. Biol. 110, 329–334.

29.      Jordal BH, Knížek M. 2007 Resurrection of Crypturgus subcribrosus Eggers 1933 stat. n., and its close phylogenetic relationship to Nearctic Crypturgus (Coleoptera, Scolytinae). Zootaxa 1606, 41–50. (doi:10.11646/ZOOTAXA.1606.1.3)

30.      Bouckaert R, Vaughan TG, Barido-Sottani J, Duchêne S, Fourment M, Gavryushkina A, Heled J, Jones G, Kühnert D, De Maio N, Matschiner M, Mendes FK, Müller NF, Ogilvie HA, du Plessis L, Popinga A, Rambaut A, Rasmussen D, Siveroni I, Suchard MA, Wu CH, Xie D, Zhang C, Stadler T, Drummond AJ. 2024 BEAST 2.7: An advanced software platform for Bayesian evolutionary analysis. Zenodo. (doi:10.5281/zenodo.10717165)

31.      Darriba D, Posada D, Kozlov AM, Stamatakis A, Morel B, Flouri T. 2020 ModelTest-NG: A New and Scalable Tool for the Selection of DNA and Protein Evolutionary Models. Mol. Biol. Evol. 37, 291–294. (doi:10.1093/MOLBEV/MSZ189)

32.      Hansen TF, Pienaar J, Orzack SH. 2008 A comparative method for studying adaptation to a randomly evolving environment. Evolution 62, 1965–1977. (doi:10.1111/J.1558-5646.2008.00412.X)

33.      R Core Team. 2024 R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available at: https://www.r-project.org/

34.      Shin S et al. 2018 Phylogenomic data yield new and robust insights into the phylogeny and evolution of weevils. Mol. Biol. Evol. 35, 823–836. (doi:10.1093/MOLBEV/MSX324)

35.      Jordal BH, Sequeira AS, Cognato AI. 2011 The age and phylogeny of wood-boring weevils and the origin of subsociality. Mol Phylogenet Evol. 59, 708–724. (doi:10.1016/j.ympev.2011.03.016)

36.      Burnham KP, Anderson DR. 2004 Multimodel inference: Understanding AIC and BIC in model selection. Sociol. Methods Res. 33, 261–304. (doi:10.1177/0049124104268644)