Measuring mass: variation among 3,161 species of Canadian Coleoptera and the prospects of a mass registry for all insects

BINs as a species proxy

DNA barcoding employs sequence variation in a 658 base pair segment of the cytochrome c oxidase subunit I gene (COI) as a basis for specimen identification and species discovery in animals (Hebert et al., 2003). The BIN system clusters these COI sequences into molecular operational taxonomic units (Floyd et al., 2002) that correspond closely with Linnaean species. For example, about 90% of all species in the well-studied European Coleoptera fauna show perfect correspondence with BINs (Pentinsaari, Hebert & Mutanen, 2014; Hendrich et al., 2015). Divergent BINs associated with the same Linnaean species name often show morphological differences upon closer inspection, while some closely related species share BINs or even identical barcode haplotypes (Pentinsaari, Hebert & Mutanen, 2014; Hendrich et al., 2015). Each BIN is assigned a unique alphanumeric identifier that serves as a species proxy (Ratnasingham & Hebert, 2013). Because they provide a taxonomic assignment for undescribed species (Brunke et al., 2019; Pentinsaari et al., 2019; Brunke, Pentinsaari & Klimaszewski, 2021), this study employed BINs to structure data collection. While substantial efforts were also made to assign each BIN to a Linnaean species, this was not always possible because of both the lack of taxonomic specialists for some families and difficulties in resolving synonymies and cryptic species. As a result, we employ the BIN count as the best estimator of the number of species examined in this study.

Body mass data

Specimens were available for 3,161 BINs of Canadian Coleoptera. They represented 1,100 of the 2,008 genera (54.8%) and 96 of the 111 beetle families (86.5%) known from Canada (Bousquet et al., 2013). Most were morphologically identified to a genus (3,156 BINs) and many to a species (2,719 BINs to 2,389 recognized species). The specimens were obtained from sampling programs coordinated by the Centre for Biodiversity Genomics (CBG) at the University of Guelph and are stored in its voucher collection. Specimens missing major body segments (head, abdomen) were not weighed. In total, 3,744 specimens were analyzed, meaning a single specimen was weighed for most BINs. However, 2–8 specimens were weighed for 334 BINs, and for these taxa, a mean mass was calculated. Standard deviation and coefficient of variation were calculated for all BINs represented by at least three specimens in order to explore the extent of mass variation within BINs.

Specimens fell into three main curatorial categories (ethanol, pointed, pinned). Specimens from 70% ethanol were first air-dried and then weighed repeatedly until the mass measurement stabilized. Specimens on points were unmounted using 70% ethanol, dried, and weighed. Pinned specimens were weighed on their pin and the pin mass was subtracted. When mass variation among pins of the size used on a specimen exceeded 12.5% of its overall weight (Gilbert, 2011), it was unpinned and weighed directly.

The mass of small specimens (~2,800 representing 2,500 BINs) was quantified to the nearest 0.0001 mg using a high-precision balance (Mettler Toledo™ XP6U), while the ~950 larger specimens (>10 mg) were weighed to the nearest 0.1 mg using a less sensitive instrument (Mettler Toledo™ MS104S). Up to three significant figures were recorded. The BIN, taxonomic assignment, and mass of each analyzed specimen are provided in a supplementary document (Data S1).

In order to examine the relationship between fresh and dry body mass, 73 specimens of Coleoptera representing a range of body sizes were freshly collected, killed, and weighed within 24 h. The specimens were then stored dry at room temperature and weighed every 60 days until their mass stabilized, and the final weight was recorded as their dry mass. A linear regression was performed to assess the correlation between fresh and dry mass.

Data description and distribution analyses

All analyses were performed in R version 3.6.3 (R Core Team, 2020). Mass values were log-transformed before further analysis, and the interquartile range (IQR) was used as a measure of variance. A two-sided D’Agostino test was employed to evaluate skewness in the data, and an Anscombe-Glynn test to assess kurtosis. One-way ANOVAs using family, subfamily, and generic assignments as variables were used to assess mass variation at different levels in the taxonomic hierarchy using respective groups containing two or more quantified BINs. A nested ANOVA was also used to examine variation partitioning in the 50 families and 449 genera that were nested with two or more subgroups. Because taxonomic groups with low species diversity tend to show less variation in mass (Chown & Gaston, 2010), a separate analysis was performed to ascertain how variation in mass was partitioned in the most diverse groups of Canadian Coleoptera. In particular, a one-way ANOVA examined 65 genera with mass data for 10 or more BINs, while a nested ANOVA examined the six families with mass data for >100 BINs. To quantify the extent of mass that could be partitioned at each taxonomic level, ω² values were calculated for all ANOVAs.

Pin variation and reproducibility assessment

A pinned insect is not easily separated from its pin, creating a complexity because the pin can outweigh the specimen. Gilbert (2011) proposed a workaround that involves estimating pin mass from key parameters (material, shape, size) before subtracting this value from the total weight to produce a mass estimate for the insect. Because the CBG employs insect pins from a single supplier, this source of variation was readily assessed. The mean and standard deviation in both diameter and mass was determined for 100 pins of each size. Because there was no overlap in diameter among different pin sizes, the size associated with each specimen could be determined, allowing its mass to be subtracted.

The consistency in determinations of pin size and of body mass was assessed by comparing mass values for 112 specimens weighed in 2014 and again in 2018. The congruence in net mass values was examined using a paired-sample t-test. To evaluate short-term variation in mass, 18 specimens were weighed daily on four consecutive days when variation in humidity was pronounced.

Since the specimens studied had been DNA barcoded before weighing, many lacked a leg or the part of it (e.g., tarsus) that was used for DNA extraction. As part of a set of pilot tests performed before commencing the weighing of specimens on a large scale, 50 specimens were examined to determine the reduction in mass caused by the loss of a leg.

Mass variation in Canadian Coleoptera

Measurements of the 3,744 beetles representing 3,161 BINs revealed their mass varied by more than five orders of magnitude (0.0024–797 mg) (Fig. 1). Among BINs with a species assignment, Ptiliola kunzei (Ptiliidae, BOLD:ACI8875) and Ptiliolum fuscum (Ptiliidae, BOLD:AAM7677) possessed the lowest mass (0.0056 mg). However, six BINs in the same subfamily (Ptiliinae) weighed less, and a specimen identified to the genus Nanosella (BOLD:ADH5266) had the lowest mass (0.0024 mg). The largest species was Hydrophilus triangularis (Hydrophilidae, BOLD:AAQ2470) at 797 mg. The median mass of all species was ~1.3 mg, represented in the data by species such as Bembidion nitidum (Carabidae, BOLD:AAD2752) and Dichelotarsus piniphilus (Cantharidae, BOLD:AAH0933). Considering all BINs, the mass distribution approximated a lognormal distribution with strong kurtosis (z = −7.39, p = 1.49⁻¹³) but insignificant skew (z = 1.96, p = 0.051). The coefficient of variation for the 196 BINs with a minimum sample size of three specimens ranged from 0.02 to 0.975 (Data S1). The fresh and dry body mass of 73 beetle specimens freshly collected from the field showed a strong linear relationship (R² = 0.97), with dry mass being approximately 49% of the fresh mass.

Much of the variation in mass among species was linked to their higher taxonomic assignments (family, subfamily, genus) (Table 1. a–c). In fact, ω² values indicated that 90% of the mass variation could be explained by higher taxonomic placement with 55% of the variation at the family level, 20% at the subfamily level, and 15% at the genus level. Because of these relationships, variation in mass among congeneric species was typically limited (Fig. 2). Considering all 518 genera where two or more BINs were examined, the variance, measured by IQR, had a median of 0.163 log₁₀ (mg), which is a 1.4-fold difference and translates to +/− 20% divergence from the median for the genus. Cases of extreme variation where the larger members of the genus were as twice as massive as the median (IQR > 0.6) were only observed in nine of the 518 genera, including Anotylus (Staphylinidae), Dyschirius (Carabidae) and Ilybius (Dytiscidae) (Data S1). Genera with ten or more measured BINs showing low IQR (<0.15) included e.g., Acrotrichis (Ptiliidae), Anaspis (Scraptiidae), and Paria (Chrysomelidae) (Data S1).

Reliability of specimen mass measurements

Diameter measurements allowed the discrimination of each pin size as differences among pins of a particular size were an order of magnitude (±0.005 mm) less than the diameter difference (0.05 mm) between adjacent pin sizes. Pins of one size did vary in mass (±0.2–0.9 mg) with this variation increasing with larger pins, but it usually represented a small component of the overall mass. In the few cases where variation in pin mass represented >12.5% of the total weight, the specimen was unmounted and weighed directly.

High humidity slowed analysis as the balances required longer to stabilize, but changes in mass linked to variation in temperature and humidity were small. For example, the mean standard error based on measurements of 18 specimens weighed every 24 h over four days was 1–2% of their mean (Data S1). Comparison of specimen weights between 2014 and 2018 further indicated that differences between paired measurements were within ±3% of their average in 109 of 112 cases while the others were within ±5%. A paired t-test demonstrated that mean mass increased by 0.3% (t = 2.0657, df = 111, p = 0.02) over the interval, likely reflecting higher humidity when the second measurements were made.

Based on a pilot test conducted before weighing specimens for analyses (full data not available), the loss of an appendage had a small impact on mass. For example, a leg typically represented 1–2% of the specimen’s mass, while the tibia plus tarsus were around 0.5%. The loss of a major body segment (head, abdomen) had much larger impacts as they comprised 12–50% of the total mass.