Raman spectroscopic insights into rare blue and green shell colouration in bivalve molluscs

Animal colours serve several important functions including visual roles in crypsis¹, reproductive signalling² and aposematism (warning)³, and non-visual functions, for example, thermoregulation⁴, photoprotection⁵ and structural strengthening⁶. The mechanisms producing these colours involve chemical pigments that absorb light, or photonic nanostructures, which reflect light to produce structural colour⁷, or a combination of the two⁸.

Across the palette of animal colours, blue and green are surprisingly rare^9,10 and are often structural rather than pigmentary in origin^11,12. These colours are similarly rare amongst molluscs, but they appear across divergent groups and somewhat unusually, are produced by both pigmentary and structural mechanisms^13,14, making them a good model group to study how blue and green have evolved.

To date, molluscan studies of colour have focused on pigments, identifying four commonly occurring classes: ommochromes (yellow, red, brown) found mainly in cephalopods (e.g^15,16.), and melanins (red, yellow, brown, black) (e.g^17,18.), carotenoids (yellow, orange, red, blue) (e.g¹⁹.) and tetrapyrroles (red, orange, green, blue, violet, yellow or brown (e.g^17,20.), which occur commonly in molluscan shells. As in most other animal taxa, molluscan green and blue pigments are rare²¹, but have been detected in several gastropod shells, including a number of neritids^22,23 and vetigastropods, for example, the blue abalone, Haliotis discus Reeve, 1846²⁴, Lithopoma tectum ([Lightfoot], 1786)²⁵ and green turban snail, Turbo marmoratus Linnaeus, 1758²⁶.

Less well studied are blue and green pigments in bivalves. Bivalvia comprises the largest class of molluscs, including clams, scallops, mussels and oysters, and exhibits an enormous diversity of shell colours²¹. Currently there are no known green pigments in this class. However, a blue-green pigment called marennine occurs in the soft tissues of oysters²⁷.

In contrast to pigmentary colours, instances of structural colour (other than iridescence due to nacre) are rare in molluscs¹⁴, apart from cephalopods²⁸. In most cases, including cephalopods, structural colour is blue or green, and is produced by a multilayer reflector, consisting of alternate layers of high and low refractive index material whose thicknesses are less than a micron²⁹. These photonic structures frequently occur in soft tissue, for example, the proteinaceous layers of reflectin found in specialised iridocyte cells of cephalopod skin, which produce dynamic displays of iridescence³⁰, and the guanine plates³¹ and spherical vesicular bodies in the iridophores of sea slugs³². However, there are also rare examples of photonic structures in molluscan calcareous shell. For example, the iridescent stripes of blue and green found on the shells of the gastropods, Patella pellucida Linnaeus, 1758³³, Cymbula granatina (Linnaeus, 1758)³⁴ and Helcion pruinosus (Krauss, 1848)³⁵, all of which generate colour via reflecting layers of calcite or aragonite. In Patella pellucida, randomly arranged, light-absorbing colloidal particles beneath the calcite lamellae enhance the appearance of the blue stripes³³.

Within bivalves, structural colour has been identified from the blue eyes of the scallop, Argopecten irradians (Lamarck, 1819), which contain proteinaceous nanospheres³⁶, and from iridocytes in the mantle of the giant clam, Tridacna maxima (Röding, 1798), which possess guanine platelets³⁷. In addition to these structural colours, bivalves have evolved a unique type of photonic structure in the periostracum, an organic layer overlying the calcareous shell (Fig. 1). This has been identified in only two species of Perna (^38,39) and is comprised of proteinaceous nanofibres, which diffract incident light to produce green iridescence.

The fact that blue and green colours are rare in molluscs and other animals, might be explained because a particular colour serves no useful function, or because the colour is difficult to synthesise¹⁰. Grant & Williams’s²¹ phylogenetic study of the distribution of shell colour across bivalves suggests the latter may be the case in this class and that blue and green colours may provide some selective advantage since they have arisen multiple times in independent lineages. It is not known whether the blue and green pigments found in gastropods are also found in bivalves, since similar shell colours can arise from different pigments and are not necessarily homologous¹⁷ and most surveys have focussed on more common shell colours (e.g. red, orange, brown, yellow, purple, etc.)^{17,20,23,40,41,42}.

Curiously, previous observational studies have shown that in bivalves, rare blue and green shell colours are more commonly associated with the periostracum than the calcareous shell itself²¹ (Fig. 1). This led to the suggestion that the characteristics of the molecular structure of some blue and green bivalve pigments may mean some are more easily incorporated into organic than inorganic material²¹.

An alternate suggestion to explain the distribution of rare colours, arises from previous Raman studies. Raman spectroscopy is frequently used with success to identify shell pigments in molluscs (e.g^{22,23,43,44,45}.). One study⁴³, however, reported that green organic layers interleaved within the calcareous shell of three species, are Raman inactive, meaning that no pigments were detected⁴³. The finding by Wade and colleagues⁴³ is the first report of Raman inactive colouration in molluscs. The authors suggested this finding may be the result of novel green pigment(s) with weak or no Raman activity or the presence of structural colour, or a combination of both.

Conversely, previous chemical analyses of the calcareous shell layer of green nerite gastropods and turban snail shells have identified green pigments as polyenes and bilins respectively^{22,23,24,25,26}, both of which are Raman active and can be detected by Raman spectroscopy. The explanation for this disparity may be due to the long evolutionary divergence between bivalves and gastropods and as previously suggested, in the location of pigments: in three green bivalves tested by Wade and colleagues⁴³, colour occurred only in organic material. Identifying rare bivalve pigments and determining how they are produced would be the first step towards understanding how and why these unusual, and possibly metabolically costly, pigments have evolved.

To this end, this study will use Raman spectroscopy to explore pigments producing rare blue and green colours across a taxonomically diverse group of bivalve species. There are currently around 10,000 accepted bivalve species⁴⁶ and the class is thought to have diverged into six major clades (Archiheterodonta, Protobranchia, Pteriomorphia, Palaeoheterodonta, Anomalodesmata and Imparadentia) in the Ordovician, 485 Mya⁴⁷. We focus on 15 bivalve species that exhibit rare green and blue shell colours, and represent five of the six major clades, choosing species to reflect (1) the diversity of shell structures that exhibit blue or green colouration (periostracum, organic layers in calcareous shell or calcareous shell), (2) taxonomic diversity, and (3) the different mechanisms that produce colour by including at least one species with known structural colour.

Optical analyses

In every shell with external green colouration (Fig. 2a-d, j-n), a single green layer was observed in the middle layer or basal layer (closest to the calcareous shell) of the periostracum, but never in the upper layer (Table 1; Fig. 3a-d, j-n). Green colour was entirely absent from the calcareous shell.

Almost all species with internal green shell colouration displayed multiple green organic layers interleaved between calcareous shell layers (Table 1; Figs. 2f-i and 3f-i). A greater number of green layers corresponded with a deeper visible shell colour.

Chama croceata Lamarck, 1819 differed from all other shells with internal green colouration in that discrete organic layers were not present, either in the periostracum or calcareous shell. Instead, a green region was identified in the base of the outer calcareous shell, which could be observed on the inside of the shell (Fig. 3i and inset). Diatoms, possibly from the genus Navicula, were observed in recesses within the bioeroded, external side of the shell. In addition to diatoms, an indistinct green layer was observed, thought to be a filamentous endolithic green alga, similar to Ostreobium.

In some of the shells with green periostracal colour, an additional yellow area (Fig. 3a-Arcuatula arcuatula (Hanley, 1844), b–Septifer bilocularis (Linnaeus, 1758), d – Mytella guyanensis (Lamarck, 1819)) or brown area (Fig. 3c–Perna viridis) was observed in the underlying calcareous shell.

The two species with external blue shell colour, Septifer bilocularis and Mutela legumen Graf & Cummings, 2006, had areas of blue colour in the nacre of the calcareous shell (Fig. 3e and o, respectively). In the case of the blue Septifer bilocularis, which was missing its periostracum, there were additional brown streaks in the calcareous shell (Fig. 3e). Blue and brown areas in calcareous shell (as seen in slide sections) corresponded to externally visible areas of the same colour (see Figs. 2e and 3e). In Mutela legumen an additional single distinct blue layer in the base of the periostracum was also present (Table 1; Fig. 3o).

Green

In contrast to other colours, Raman spectra for all organic green layers were very noisy and no clear pigment peaks could be identified at either 532 or 785 nm excitation wavelengths for Yoldia limatula, Alectryonella plicatula, Arcuatula arcuatula, Monia zelandica, Musculus discors, Mytella guyanensis, Perna viridis, Septifer bilocularis, Cleidothaerus albidus or Glauconome virens (Fig. 4a and b respectively, see Supplementary Figure S1 for non-normalised spectra, Fig. 5 for SNR graphs, see Supplementary Tables S1 and S2 for peak positions at 532 and 785 nm excitation, respectively). As such, we refer to these species as being Raman inactive, meaning that the constituent molecules have no vibrational modes readily detectable to Raman spectroscopy and that no pigments were detected.

In contrast, novel, non-carotenoid-based pigment peaks were identified in three species, Corbicula fluminea (O. F. Müller, 1774), Anodonta anatina (Linnaeus, 1758) and Chama croceata, using 785 nm excitation only (Fig. 4b, Supplementary Table S2). Four peaks identified in these species were shared (640 cm⁻¹, 1002 cm⁻¹, 1115 cm⁻¹ and 1609 cm⁻¹; locations of these peaks are highlighted with green bars in Fig. 4b). Several of the observed non-carotenoid peaks are weak and difficult to consider unambiguous detections of real Raman-scattering phenomena, with signal: noise ratios (SNRs) of less than 10:1 (Fig. 5). Corbicula fluminea had three peaks that exceeded the conservative 10:1 threshold for unambiguous detection, at 640 cm⁻¹, 1115 cm⁻¹ and 1609 cm⁻¹ (indicated by asterisks in Fig. 4b), while the 1002 cm⁻¹ peak was close to the threshold (Fig. 5a). An unambiguous peak was also detected at 1115 cm⁻¹ in Chama croceata (indicated by an asterisk in Fig. 4b, Supplementary Table S2). In Chama croceata only, a further two unique pigment peaks were observed at 1312 and 1340 cm⁻¹, of which only the 1340 cm⁻¹ peak was unambiguous (indicated by an asterisk in Fig. 4b, Supplementary Table S2).

Two blue specimens were examined. In Mutela legumen, two peaks corresponding to carotenoid-based pigments were observed at 1128 and 1514 cm⁻¹ in both the organic periostracum and calcareous shell under 532 nm excitation (Fig. 4a, Supplementary Table S1). The peaks were identical in wavenumber, varying only in intensity: only the calcareous shell peaks were unambiguous, with SNRs around 11:1 (Fig. 5a and b).

In the second species, Septifer bilocularis, carotenoid-based pigments were also detected, but these differed from those found in Mutela legumen, with different peaks at 1090 and 1478 cm⁻¹ in the blue calcareous shell using 785 nm excitation (Fig. 4b, Supplementary Table S2). These peaks were unambiguous with SNRs of approximately 100:1 (Fig. 5b).

Additional yellow and brown calcareous shell colours demonstrated carotenoid-based pigment peaks (Fig. 4a and b). Under 532 nm excitation, two major peaks (v₁, v₂) associated with yellow carotenoid-based pigments were observed at 1091 cm⁻¹ and 1481 cm⁻¹ in Arcuatula arcuatula, at 1090 cm⁻¹ and 1481 cm⁻¹ in Mutela guyanensis, and at 1134 cm⁻¹ and 1522 cm⁻¹ in the green specimen of Septifer bilocularis (Fig. 4a, Supplementary Table S1). These peaks were also observed under 785 nm with more consistent positions of 1093–1094 cm⁻¹ and 1475–1476 cm⁻¹ (Fig. 4b, Supplementary Table S2). Two minor carotenoid-based pigment peaks (v₃: at ~ 1010 cm⁻¹, v₄: ~1300 cm⁻¹) were only observed under 532 nm excitation (Fig. 4a, Supplementary Table S1). All peaks were unambiguously detected using this excitation, with SNRs between 11:1 and 400:1 (Fig. 5a and b). Carotenoid-based pigment peaks for the yellow areas of the green Septifer bilocularis and blue areas of the blue Septifer bilocularis shells were non-identical.

For brown calcareous shell colour, the blue Septifer bilocularis showed carotenoid-based pigment peaks at 1095 cm⁻¹ and 1482 cm⁻¹. The brown shell region of Perna viridis showed peaks at 1086 cm⁻¹ and 1482 cm⁻¹. All peaks were unambiguous under 532 nm excitation, with SNRs between 50:1 and 100:1 (Fig. 5b).

As in previous studies, carotenoid-based pigments were easier to detect using 532 nm excitation than 785 nm, the former producing strong Raman peaks that were non-symmetric in shape and varied significantly in position (~ 43 cm⁻¹) between species (suggesting non-identical pigments) due to resonance enhancement of vibrations associated with the carotenoid chromophore at this excitation wavelength (Fig. 4a). By comparison, under non-resonant 785 nm excitation the carotenoid peaks were harder to detect, more symmetric, and showed less variation in positions between species (< 5 cm⁻¹) (Fig. 4b).

Calcium carbonate

The calcium carbonate of the calcareous shell was also apparent in Raman spectra of several species, having peak positions consistent with either calcite (280, 712, 1085 cm⁻¹) or aragonite (208, 704, 1085 cm⁻¹) (Fig. 4a and b, Supplementary Tables S1 and S2). Calcite was only detected in Chama croceata and Monia zelandica (J. E. Gray, 1843).

Aragonite was detected in Anodonta anatina, Arcuatula arcuatula, Alectryonella plicatula (Gmelin, 1791), Cleidothaerus albidus, Corbicula fluminea, Glauconome virens (Linnaeus, 1767), Mutela legumen, Mytella guyanensis, Perna viridis, Septifer bilocularis (green and blue specimens) and Yoldia limatula (Say, 1831). In Musculus laevigata (Linnaeus, 1767), only the 1085 cm⁻¹ (v₁ carbonate mode) peak common to both calcite and aragonite was observed. Negative calcite peaks at 1085 cm⁻¹ in Glauconome virens, Perna viridis and Yoldia limatula (all 785 nm) are artefacts of the background oscillation subtraction (the removal of periodic semi-sinusoidal patterns from a raw Raman spectrum to isolate true Raman peaks) (See Supplementary Figure S3). The oscillation spectrum used for subtraction was averaged from spectra with minimal (but not necessarily zero) Raman signal, which can lead to an artefact when the scale of the background oscillation is large, because the intensity of the subtracted calcite peak can be greater than the true signal, leading to an apparent negative peak for some calcite peaks.

This section is descriptive in nature and reflects observations across a limited sample of 15 species. Traits were mapped onto a simplified bivalve phylogeny (Fig. 6) and showed that non-carotenoid-based green pigments occurred only in species sampled from Palaeoheterodonta and Imparadentia, and within these clades, in two freshwater species and marine Chama croceata.

Pigments were not detected from any green Pteriomorphia specimens examined, nor any species of Protobranchia or Anomalodesmata. All species examined in this study from these clades are marine, except Arcuatula arcuatula, which occurs in brackish water.

Two different carotenoid-based blue pigments were identified in two species with nacreous shells, one each from the Pteriomorphia and Palaeoheterodonta.

Green colour across all clades was observed only in the periostracum or in organic layers within the calcareous shell, supporting the idea proposed by Grant & Williams²¹ that rare green shell colour occurs more often in organic than calcareous material. No significant carotenoid-based pigments were detected in any green-coloured shell. This result was also reported by Wade et al. ⁴⁵ for a smaller number of samples. We have identified several weak Raman peaks consistent with an unidentified (non-carotenoid) pigment in three species (Corbicula fluminea, Anodonta anatina, and Chama croceata). Four peaks were identified in Corbicula fluminea that exceeded the 10:1 signal: noise ratio (SNR) required for unambiguous detection. These peaks were also observed in Anodonta anatina and Chama croceata, albeit not all met the 10:1 threshold. However, the fact that four peaks were seen across three species with similar positions and widths suggests that their appearance is due to the occurrence of the same organic compound in all three species. This compound most likely contains an aromatic chromophore with ring breathing modes at ~ 640 and ~ 1000 cm⁻¹ and a ring stretching mode at ~ 1610 cm⁻¹. The presence of an aromatic chromophore suggests this compound may play a role in pigmentation, but the observed peak positions are not consistent with reported Raman spectra for any known pigments in mollusc shells, such as carotenoids, melanins, or tetrapyrroles^48,49,50,51, nor is it a good match to protein, lipids, or other common biological macromolecules⁵². The existence of novel green bivalve shell pigments with weak Raman activity was first proposed by Wade et al⁴³. but this is the first direct detection of such pigments in mollusc shells and thus warrants further investigation.

No pigment peaks were detected in the remaining 10 species with green-coloured shells, which suggests that pigments are either not present or present at such low concentrations that they cannot be reliably detected. This threshold will depend on the optical properties of the sample and the measurement settings used⁵³but will generally be higher for samples with strong background fluorescence and lower for compounds under resonant excitation, where the laser is tuned to the molecule’s specific absorption wavelength, which can enhance Raman scattering by multiple orders of magnitude. However, the existence of pigments in these species is unlikely given that no pigment peaks were detected under 532 nm excitation (resonant with carotenoids and most other visible-light pigments) in both this study and the work reported by Wade et al⁴³.. While measurement settings used in this study were tuned to maximise SNR from each sample without saturating the detector (exposure times varied from 0.01 to 5 s), it is necessarily more difficult to speculate that no pigment was detected for samples where colour occurs in the periostracum, because of exceptionally strong background fluorescence / oscillations that can mask very weak Raman peaks.

The lack of detectable pigment peaks for 10 green-shelled species provides support for the speculative idea that structural colour may be more common than currently recognised. To date, structural colour in bivalve shells has only been reported in two species of Perna^38,39. In both cases, the authors were unable to explain how the green iridescence produces the non-iridescent, external green appearance of the shell, concluding that green pigments must also be involved. However, our results show that there are no detectable pigments in the green periostracal layer of Perna viridis, which would suggest that the green shell colour in this species is entirely due to the photonic structure.

Chama croceata was the only species examined to host algae in the shell. Two different types were identified from different shell regions. Diatoms were recorded from recesses within the bioeroded shell, similar to those reported from the shell surface of other bivalves⁵⁴, although in these species there was no reference to any alteration of overall shell colour. In addition to diatoms, an indistinct green layer was observed, thought to be a filamentous endolithic green alga, reported previously as contributing to the green colour of some corals⁵⁵. The green of both types of algal cells seen against the white shell of Chama croceataappear to be the source of green shell colour, though the weak Raman signal of the novel, non-carotenoid-based pigment did not correspond to known spectra for chlorophyll (e.g⁵⁰.). It is not clear whether the presence of the green, weakly Raman active pigment in this sample reflects the presence of an algal-derived pigment. If this is the case, the same pigment may be present in the other two species through accumulation as a result of dietary uptake or by de novo synthesis but is unlikely to be due to the presence of intact or degraded algae as none were observed with light or scanning electron microscopy (SEM).

Unlike green, blue was observed both as a layer in the organic periostracum but also as a diffuse region in the calcareous shell in the two species examined. Significant carotenoid-based pigment peaks were detected in both blue species, Mutela legumen and Septifer bilocularis, from calcareous shell with diffuse blue colour. Carotenoids can only be synthesised by plants and fungi and are therefore almost always acquired by animals, including bivalves, through their diet⁵⁶. Plant carotenoids are commonly yellow, orange and red, so we speculate that a carotenoid has been chemically or structurally modified with a protein in Mutela legumen and the blue Septifer bilocularis to produce a blue pigment, as has been reported previously in other invertebrates^57,58. Raman results recorded by Hedegaard²³ for blue shell of Mytilus edulis (1098 cm⁻¹, 1484 cm⁻¹), closely match those of Septifer bilocularis (1092 cm⁻¹, 1481 cm⁻¹), which belongs to the same family as Mytilus, but not Mutela legumen (1128 cm^–1, 1514 cm^–1).

In Mutela legumen blue colour is found in both the periostracum and shell, and Raman spectra suggest colour is due to identical pigments with a lower concentration of pigment in the periostracum. Another hypothesis is that the spectrum observed for the periostracum might reflect the proximity/overlap of the laser spot across both periostracum and shell, since the spot size was of a similar diameter (5 μm) to the periostracal thickness (6 μm), and the spot may be effectively enlarged by scattering of laser light within the solid matrix of the sample itself⁵⁹. If this is the case, the strong carotenoid response from the shell would be detected even if the laser spot was centred on the periostracum. If the periostracum spectrum is due to an ‘overlap’ with pigments found in the shell, the blue periostracal colour of Mutela legumen may be structural in origin. However, we did not find evidence for ‘overlap’ readings in another species with a thin periostracum and nearby carotenoid-based pigment (for example, see Arcuatula arcuatula spectrum in Fig. 4), which would suggest that Mutela legumen spectra correctly reflect the existence of a blue pigment in both periostracum and shell in this species. This is unusual compared to other taxa, where blue is often structural, for example, in blue gastropods³³ and cephalopods⁶⁰, and also in unrelated taxa including, blue birds⁶¹ and butterflies⁶².

Although our sample size is small, our results demonstrate that green shell colour can be produced by at least two different mechanisms (pigments vs. photonic structures) and suggests a possible relationship between habitat (marine vs. freshwater) and mechanism of colour production, which would be interesting to investigate with a larger sample size. Future work will aim to determine whether nanostructures are contributing to green colour in bivalves, and if so, to characterise photonic architectures, focussing on green pteriomorphs, where the most vivid colours are found and Raman evidence did not find evidence of pigments. The known examples of structural green colour in this group have been linked to the presence of a vacuolar layer in the periostracum^38,39. Consistent with this idea, this periostracal layer is absent in the freshwater palaeoheterodont clade (for example^63,64,), where we instead found pigments producing both blue and green colours. We speculate that structural green colour has evolved in pteriomorphs through changes in the size of the vacuoles.

The absence of detectable pigments in most green shells, points to a greater role for structural colour and other non-carotenoid chromophores than has previously been recognised. In contrast, blue shell colour was consistently associated with carotenoid-based pigments, highlighting distinct evolutionary and mechanistic pathways for colour production. Collectively, these findings reveal a greater diversity of colour-generating mechanisms in bivalve shells than previously appreciated and provide a framework for future studies investigating the evolution, chemistry and photonic architecture of shell colouration across the Bivalvia (Fig. 6).

Methods

Sample selection and preparation

Green and blue shells were selected from families representing five major clades: subclass Protobranchia: Yoldia limatula (Say, 1831), NHMUK 20230858; infraclass Pteriomorphia: Septifer bilocularis (Linnaeus, 1758) blue specimen NHMUK 20260050, Septifer bilocularis green specimen NHMUK 1953.1.29.174–179, Alectryonella plicatula (Gmelin, 1791) NHMUK 20230907/1, Arcuatula arcuatula (Hanley, 1844) NHMUK 20230896/1, Monia zelandica (J. E. Gray, 1843) NMNZ M.029110, Perna viridis (Linnaeus, 1758) NHMUK 20230895/1, Musculus discors (Linnaeus, 1767) NHMUK 1877.11.28.15, Mytella guyanensis (Lamarck, 1819) NHMUK 20230912; superorder Imparidentia: Corbicula fluminea (O. F. Müller, 1774) NHMUK 20050005, Glauconome virens (Linnaeus, 1767) NHMUK 20240252/1, Chama croceata Lamarck, 1819 NHMUK 1952.12.19 85; subterclass Palaeoheterodonta: Anodonta anatina (Linnaeus, 1758) NHMUK 20180293, Mutela legumen Graf & Cummings, 2006 NHMUK 1904.12.13-3-4; and superorder Anomalodesmata: Cleidothaerus albidus (Lamarck, 1819) NHMUK 20230910 (Fig. 2). No green or blue species have been identified in the Archiheterodonta²¹.

Observers can differ in their opinion of colour; therefore, shell colours of specimens were determined by comparison with coloured charts in the Munsell Book of Color⁷⁰. Intact shell valves were photographed alongside Munsell colour charts (Fig. 2 and Supplementary Figure S2). Names used hereafter for shell colours refer to green or blue.

Slide-mounted cross-sections of a single valve from each of the shells were prepared, enabling the use of light microscopy to provide a reference image and to precisely locate and investigate colour in the calcareous shell or the periostracum using Raman spectroscopy. Valves were embedded in KEPT epoxy resin (Kemet International Ltd, Maidstone, UK), before being cut in half from the centre of the umbo to the shell margin. The exposed surface was then polished to remove the rock saw marks using silicon carbide papers (2500 grit), reducing the thickness of the samples to 130 μm. Samples were then further polished using diamond and aluminium pastes of different grades to produce a 100 μm slide-mounted section. Colour images of shell cross-sections were taken in reflection on a white stage insert, using a Hirox HRX-01 digital light microscope (Fig. 3).

Raman spectroscopy

Using the cross-sectional images as a guide, Raman measurements were taken from all coloured areas, with a focus on green and blue regions of colour in the periostracum and from calcareous shell (see Fig. 3 for location of Raman measurements).

Measurements were taken using a Renishaw Virsa instrument at the Natural History Museum with 532 and 785 nm excitation in back-scatter configuration with a 50x objective lens linked by fibre optic cables to the spectrometer, producing a laser spot of ~ 5 μm diameter. Measurements using 532 nm excitation were acquired using a laser power of 0.1 mW (set using a variable angle shutter) and were summed over 20 accumulations of 0.1 s. Measurements using 785 nm excitation were acquired using a laser power of 5 mW and summed over 5 accumulations of 10 s. Exposure times were adjusted on a sample-by-sample basis to avoid signal saturation due to intense sample fluorescence. A 2400 lines/mm grating was used to produce a spectral range of 100–3200 cm⁻¹ measured over multiple acquisitions in stepped mode, producing a spectral bin size of ~ 1.3 cm⁻¹ (532 nm) and ~ 0.9 cm⁻¹ (785 nm) per pixel. Cosmic rays were removed using the Renishaw WiRE software, all further Raman data processing (background subtraction, peak detection, peak fitting, data visualisation) was done using a Jupyter Notebook running custom Python code⁷¹ that utilises the following packages: Numpy⁷², SciPy^73,74, and LMFIT⁷⁵. Initial background subtraction was done by fitting a polynomial function to regions with no Raman signal and subtracting it from the data. A second background subtraction was then done to reduce the magnitude of background oscillations by fitting and subtracting a background reference spectrum, averaged from 9 (532 nm) and 11 (785 nm) spectra that exhibited a maximum absolute first derivative of < 0.021 between 200 and 1800 cm^[–1(see Supplementary Figure S3). An automatic peak detection algorithm was used to identify local maxima with signal: noise ratios exceeding 10:1 and a relative intensity higher than 5% vs. the maximum observed signal. Automatic peak fitting was done using pseudo-Voigt functions to represent each detected Raman peak. The Jupyter Notebook and underlying code are available on GitHub and archived at⁷¹.

Trait mapping

Basic qualitative traits were mapped onto a simplified phylogeny (modified from Bieler et al.⁶⁴). The following traits were recorded. Species habitat (marine, brackish or freshwater) with information sourced from MolluscaBase⁴⁶ and MUSSELp⁶⁵. Colour and location of shell colour (calcareous shell, periostracum or organic layers in the calcareous shell) were scored based on light microscopy of shell sections. Additional calcareous shell characters were included to examine the previously unexplored possible links with colour in this inorganic region. Data on calcium carbonate form (calcite or aragonite) originated from the Raman results in this study. Shell microstructure was sourced from the literature ^{[74 75,68}. Three periostracal characters were recorded, including (i) position of coloured layer(s), (ii) total number of periostracal layers and (iii) thickness of the periostracum. Position of coloured layer(s) was based on light microscopy of shell sections. Total number of periostracal layers were recorded using a combination of light microscopy and the authors own SEM images. Periostracal thickness was taken from values reported in the literature⁶⁹ or measurements taken by the authors from their own SEM images. Presence of Raman-active pigment peaks was noted, and based on this observation a decision was made on the likely mechanism of colour production.