Data for: Evolutionary transition from surface to subterranean living in Australian water beetles (Coleoptera, Dytiscidae) through adaptive and relaxed selection

Summary of contents:Zip Folder "R Scripts" containing the R code for the analysesThree scripts called: antenna_legs.R; dorsal.R; triplets.RZip Folder "raw data" containing the measurement data (linear and landmarks)Zip folder "ZYX Dytiscids Photographs - beetles dorsal views" containing the photographs from which measurements were taken - Folder is 5.64GB (9.3GB uncompresed)Zip folder "ZYX Dytiscids Photographs - beetles linear measuring photos" containing the photographs from which measurements were taken - Folder is 21.12GB (38.9GB uncompresed) Samples overview and ecologyA total of 362 specimens were measured from the South Australian Museum (SAMA) Entomology collection, plus 13 published drawings (Watts & Humphreys, 2003, 2004, 2006, 2009). This comprised 121 species (108 from SAMA and 13 from camera-lucida drawings) in Dytiscidae (Coleoptera), including 31 ‘surface’ aquatic diving species, 88 ‘subterranean’ aquifer-inhabiting species, and four ‘interstitial’ species (Figure 1). These classifications refer to their ecology: surface refers to diving dytiscid species that inhabit aboveground stream- or pond-related habitats; subterranean refers to species that live in underground aquifers (calcrete/ fractured rock); interstitial species reside in ephemeral water bodies such as seasonal drying streams. A species list, including authorities for each species, is included in Supplementary File 1. The selected dytiscid species are distributed across genera Limbodessus Guignot, 1939, Paroster Sharp, 1882, Allodessus Guignot, 1953, Gibbidessus Watts 1978, Neobidessodes Hendrich & Balke, 2009 and Uvarus Guignot, 1939, in Dytiscidae, with over 90% of species distributed in genera Limbodessus and Paroster. Among the 13 species sampled from drawings, five are in Limbodessus, five in Paroster, two in Neobidessodes, one in Exocelina Broun, 1886. Species sampled from drawings were used for the aquifer triplet analysis only, detailed below. Digital imagingPhotographs of the dorsal view of the whole animal and close-up images of the antennae and the legs were taken with one Auto Montage system, which used algorithms to merge multiple photos in different focuses into one high-quality image. The imaging system was a Leica M205C microscope on a vertical track operated by Leica Application Suite v3.8.0 attached to a Leica DFC500 camera. From the series of images with incremental focus, a stacked montage image was automatically produced in the Leica Application Suite v3.8.0 (Figure 2).To maintain consistency and reduce the impact of specimen presentation orientations, all specimens were imaged with the scutellum or the anterior end of the middle suture of the elytra being the uppermost point. No colour calibration was used. No optical control was used except an automatic exposure time regulator at its default value. The specimens were presented as flat to the plane of the camera as possible to avoid horizontal tilts.To capture as many details of the specimen as possible, we used optimising steps for the multifocus within Leica Application Suite v3.8.0. A step was defined as each focus plane. Each step varied between 0.01mm to 0.05mm. The outcome was high-resolution (4080*3072 pixels) TIFF image files with scale bars. MorphometricsTwo morphometric methods were used to capture the morphological variation among specimens. We quantified the morphology in the body outline of the specimens by using geometric morphometrics (landmark-based measurements). As the landmarking required a dorsal view of the beetles, we used 255 specimens of 99 species that were dorsally orientated and presented. Morphology of the three pairs of limbs and antennae were quantified using linear measurements for 107 species, each represented by one specimen. The body shape dataset used both fixed homologous landmarks identifiable on all specimens and semi-landmarks that constitute a homologous curve (Zelditch et al., 2004). Since male dytiscids were slightly larger than females, male specimens were chosen where possible to remove bias caused by potential allometry. Multiple specimens were selected to represent their species. However, some species were represented by a single specimen. For example, Paroster extraordinarius, the first and only known stygobiotic dytiscid species found in South Australia (Leys et al., 2010), was represented by one holotype specimen. In summary, a set of 71 landmarks were chosen to capture the outline shape of the dytiscids, illustrated in Figure 2 with definition described in Table 1. We used the software tpsDIG2 v.2.32 (Rohlf, 2021) to manually place the fixed landmarks and semi-landmarks on the photographs (Figure 2). The landmarks, represented by two-dimensional coordinates, were exported in the .TPS file format.The body length, and dimensions of the limbs and antennae comprised a total of 17 measurements (in mm) recorded from the photographs with scale bars (Figure 2, Table 2). When photographing, the length of the body, antennae and tarsi were recorded on-site using the default measuring tool implemented within Leica Application Suite v3.8.0. The length and width of legs were later measured using the measuring tool in the GIMP v2.10.32 software (The GIMP Development Team, 2022). Both dorsally and ventrally presented specimens were used to capture the size of the features. As the antennae were usually curled out of the plane of the camera, up to three images were taken at different angles. The length of the antennae was recorded by the sum of the length of the eleven segments – the scape, the pedicel and nine flagellomeres. As the coxae of dytiscids are fused medially to the sternum and the trochanters are not always presented in the same orientation with the rest of the legs, both coxae and trochanters were not recorded, leaving only femur, tibia, and tarsus (Figure 2, Table 2).

内容概述：
"R脚本"压缩文件夹，内含用于分析的R代码，包含3个脚本文件：antenna_legs.R、dorsal.R及triplets.R。
"原始数据"压缩文件夹，内含测量数据（线性测量数据与标志点数据）。
"ZYX龙虱科甲虫背面观照片"压缩文件夹，内含用于获取测量数据的照片，该文件夹大小为5.64GB（解压后为9.3GB）。
"ZYX龙虱科甲虫线性测量照片"压缩文件夹，内含用于获取测量数据的照片，该文件夹大小为21.12GB（解压后为38.9GB）。

样本概述与生态学：
本数据集共测量了南澳大利亚博物馆（South Australian Museum, SAMA）昆虫标本馆藏的362号标本，外加13幅已发表的绘图（Watts & Humphreys, 2003, 2004, 2006, 2009）。样本涵盖龙虱科（Dytiscidae，鞘翅目Coleoptera）的121个物种：其中108个物种来自SAMA馆藏，13个物种来自明视野绘图（camera-lucida drawings），包含31种"地表"水生潜水龙虱、88种"地下"含水层栖息龙虱以及4种"间隙"龙虱（图1）。
上述分类依据物种的生态位划分：地表类群指栖息于地表溪流或池塘生境的水生龙虱；地下类群指栖息于地下含水层（钙结层/裂隙岩石）的物种；间隙类群则栖息于季节性干涸溪流等临时水体中。物种名录（含各物种的命名人信息）详见补充文件1。
本研究选取的龙虱科物种分属于6个属：Limbodessus Guignot, 1939、Paroster Sharp, 1882、Allodessus Guignot, 1953、Gibbidessus Watts 1978、Neobidessodes Hendrich & Balke, 2009及Uvarus Guignot, 1939，其中超过90%的物种隶属于Limbodessus和Paroster两个属。在13幅绘图对应的物种中，5种隶属于Limbodessus属，5种隶属于Paroster属，2种隶属于Neobidessus属，1种隶属于Exocelina Broun, 1886。绘图对应的物种仅用于下文详述的含水层三联体分析。

数字成像：
使用一套自动拼接系统（Auto Montage system）拍摄甲虫整体背面观照片以及触角、足的特写照片，该系统通过算法将不同焦平面的多张照片合并为一张高质量图像。成像系统为搭载在垂直导轨上的徕卡M205C体视显微镜，搭配徕卡DFC500相机，通过徕卡应用套件v3.8.0（Leica Application Suite v3.8.0）操控。通过一系列递增焦平面的照片，可在徕卡应用套件v3.8.0中自动生成堆叠拼接图像（图2）。
为保证成像一致性并降低标本摆放角度的影响，所有标本均以小盾片或鞘翅中缝线前端作为最高点进行拍摄。未进行色彩校准，除使用默认参数的自动曝光调节器外，未进行其他光学参数调节。标本尽可能与相机平面保持平行，以避免水平倾斜。
为尽可能捕捉标本细节，我们在徕卡应用套件v3.8.0中对多焦平面成像进行了优化设置：每个步长对应一个焦平面，步长范围为0.01mm至0.05mm。最终生成带比例尺的高分辨率（4080×3072像素）TIFF（标记图像文件格式，Tagged Image File Format）图像文件。

形态测量学：
本研究采用两种形态测量方法以捕捉标本间的形态变异。
其一为基于标志点的几何形态测量学（geometric morphometrics）方法，用于量化标本的身体轮廓形态。由于标志点标注需要甲虫的背面观图像，因此仅选取了99个物种共255号背面朝向且可用于成像的标本。其二为线性测量方法，用于量化107个物种（每个物种对应1号标本）的三对附肢及触角的形态。
身体形状数据集同时包含可在所有标本上识别的同源固定标志点以及构成同源曲线的半标志点（semi-landmarks）（Zelditch et al., 2004）。由于雄性龙虱体型略大于雌性，研究尽可能选取雄性标本以消除潜在异速生长（allometry）带来的偏差。每个物种均选取多个标本进行测量，但部分物种仅能获取1号标本：例如Paroster extraordinarius，这是南澳大利亚发现的首个且目前唯一已知的地下栖息龙虱物种（Leys et al., 2010），仅能获取1号正模标本（holotype）。
综上，本研究选取了71个标志点以捕捉龙虱的轮廓形态，如图2所示，标志点的定义详见表1。我们使用tpsDIG2 v.2.32（Rohlf, 2021）软件在照片上手动标注固定标志点与半标志点（图2），标注得到的二维坐标标志点以.TPS文件格式导出。
通过带比例尺的照片，共记录了17项以毫米为单位的测量数据，涵盖体长以及附肢、触角的尺寸（图2，表2）。拍摄时，使用徕卡应用套件v3.8.0内置的默认测量工具现场记录了体长、触角及跗节的长度。足的长度与宽度则后续使用GIMP v2.10.32软件（The GIMP Development Team, 2022）的测量工具完成。
本研究同时使用背面观与腹面观的标本以获取各结构的尺寸信息。由于触角通常会卷曲出相机平面，因此最多会从三个不同角度拍摄触角照片。触角长度通过11个节段的总长度计算得到：包括柄节、梗节以及9个鞭小节。由于龙虱的基节（coxae）与胸骨内侧愈合，且转节（trochanters）的朝向通常与足的其余部分不一致，因此未记录基节与转节的尺寸，仅测量腿节（femur）、胫节（tibia）与跗节（tarsus）（图2，表2）。