While the shell of the land snail Cepaea nemoralis is typically classed as yellow, pink or brown, the reality is that colour variation is continuously distributed. To further understand the origin of the continuous variation we used crosses of C. nemoralis to compare quantitative measures of the colour with the inferred genotype of the underlying supergene locus. We also used a recently developed linkage map to find quantitative trait loci (QTL) that may influence colour. The results show that the colour locus of the supergene – at around 31.385 cM on linkage group 11 – is involved in determining the quantitative chromatic differences that are perceptible to human vision. We also found some evidence that variation within colour classes may be due to allelic variation at or around the supergene. There are likely other unlinked loci involved in determining colour within classes, but confirmation will require greater statistical power. Although not investigated here, environmental factors, including diet, may also impact upon variation within colour types.

Shell reflectance measures:

The shell ground colour of the two parents and 73 offspring from cross C451 x C452, and the parents and 32 offspring from cross C102 x C103 was measured using an Ocean Optics spectrometer (model USB2000+UV-VIS-ES), with a Xenon light source (DT- MINI-2-GS UV-VIS-NIR), in the range of 300-700 nm. To establish the baseline light spectrum standard, a WS-1 diffuse white reflectance was used, with total darkness used to establish the standard for the dark spectrum (Taylor et al. 2016). The probe was held at a 45° incidence angle, about 2 mm from the shell (Davison et al. 2019), using the underside of the shell, because there is nearly always an unbanded region and it is usually least damaged. Ocean Optics SpectraSuite 2.0.162 was used to collect the data, with an integration time of 750 msec, a boscar width of 5, and an average of 10 scans. To account for experimental error and slight spatial differences in shell colour, the reflectance of each shell was measured six times, randomising the order, with the software light calibration updated every third measurement. Following collection, the unprocessed reflectance spectra data were binned and smoothed into 5 nm categories using Pavo version 2.7.0 (Maia et al. 2013; Maia 2019). 

Reflectance spectra were then plotted, deriving the 95% CI for the means of each group. Subsequent analyses were performed using R version 4.0.5 (31 March 2021), following the method of Delhey et al. (2015), and the same as previously (Davison et al., 2019). In brief, in order to estimate variation in shell colours, a psychophysical framework of avian colour vision developed by Delhey et al. (2015) and Vorobyev et al. (Vorobyev & Osorio, 1998; Vorobyev et al., 1998) was used to assess whether chromatic differences between reflectance spectra exceed a discrimination threshold, or “just noticeable difference” (JND), which can be perceived by a receiver, such as an avian predator. As previously, to analyze chromatic variation, the quantum catches for each cone type were converted into three chromatic coordinates (x, y, and z), where Euclidean distances between points reflect perceptual differences. Then, to identify the main axes of chromatic variation, we carried out a principal components analysis (PCA) on the chromatic coordinates (x, y, and z), using a covariance matrix rather than a correlation matrix because it preserves the perceptual distances (JNDs) (Delhey et al., 2015). The methods of Delhey were also used to assess achromatic variation, “brightness” or luminance variation, by computing achromatic contrast between each reflectance spectrum and a reference, corresponding to a dark spectrum, and using the same noise-to-signal ratio. Two-sample t-tests were subsequently used to compare the means of different sample sets (e.g., pink/yellow, or pink1/pink2).

Genotyping data:

A ~10x depth whole genome sequence was generated using Illumina sequencing of the parents and offspring (Johansen et al. 2023), here we reused the same data but for a different purpose. Previously, FASTP v0.23.2 software was used in pre-processing of fastq files, specifically to check the quality, and for adapter trimming and quality filtering of the reads. Then, fastq files were aligned to the C. nemoralis genome using the Burrows-Wheeler Alignment method (Li and Durbin 2009) in bwa v0.7.17-r1188, using default settings and marking low-quality alignments (-M). Samtools v1.11 was used to sort and compress the sam files to bam files, then bcftools v1.10 (Danecek et al. 2011) was used to call variants, outputting a variant call format (vcf) file for all individuals that were in the cross. To identify loci that were heterozygous in the yellow parent C452 (C^Y3C^Y4), we derived a vcf file for tig00045252, a contig that is perfectly associated with the colour locus in that cross (Johansen et al. 2023). Then, we identified all single nucleotide polymorphisms (SNPs) on tig00045252, only retaining those SNPs that were heterozygous in the yellow parent C452 and homozygous in the pink parent C451. By inspecting the offspring genotypes, we were able to infer whether individuals were either pink1/pink2 (C^P1C^Y3^^ / C^P1C^Y4), or yellow1/yellow2 (C^Y2C^Y3 / C^Y2C^Y4).