An Analysis of Bivalve Larval Shell Pigments Using Microraman

Research article

Received: 13 August 2013 Revised: 24 February 2014 Accepted: 6 March 2014 Published online in Wiley Online Library: 1 April 2014

(wileyonlinelibrary.com) DOI 10.1002/jrs.4470 An analysis of bivalve larval shell pigments using micro-Raman spectroscopy Christine M. Thompson,a* Elizabeth W. North,a Sheri N. Whiteb and Scott M. Gallagerb

Micro-Raman spectroscopy has been used on adult bivalve shells to investigate organic and inorganic shell components but has not yet been applied to bivalve larvae. It is known that the organic matrix of larval shells contains pigments, but less is known about the presence or source of these molecules in larvae. We investigated Raman spectra of seven species of bivalve larvae to assess the types of pigments present in shells of each species and how the ratio of inorganic : organic material changes in a dorso-ventral direction. In laboratory experiments, we reared larvae of three clam species in waters containing different organic signatures to determine if larvae incorporated compounds from source waters into their shells. We found differences in spectra and pigments between most species but found less intraspecific differences. A neural network classifier for Raman spectra classified five out of seven species with greater than 85% accuracy. There were slight differences between the amount and type of pigment present along the shell, with the prodissoconch I and shell margin areas being the most variable. Raman spectra of 1-day-old larvae were found to be differentiable when larvae were reared in waters with different organic signatures. With micro-Raman spectroscopy, it may be possible to identify some unknown species in the wild and trace their natal origins, which could enhance identification accuracy of bivalve larvae and ultimately aid management and restoration efforts. Copyright © 2014 John Wiley & Sons, Ltd. Additional supporting information may be found in the online version of this article at the publisher’s web site.

Keywords: Raman spectroscopy; polyenes; bivalve larvae; chemotaxonomy; classiﬁcation

Introduction each genera being the result of both biological and environmental forces.[6] Recent work has suggested that the organization of the Most research on bivalve shell formation has investigated adult organic matrix of larval shells may differ between species because shells, leaving a gap in research involving bivalve larval shell birefringence patterns that reflect the orientation of aragonite microstructure. Pigments, often polyenes similar to carotenoids crystals in the organic matrix appear to be species-specific.[9–11] or porphyrins, have been identified in adult bivalve shells and Studies of mineralization of larval bivalves have traditionally – in pearls mineralized from bivalves.[1 3] The types of pigments focused on the calcium-carbonate portions of the PDI and PDII which exist in larval shells, how they are distributed, and their using electron microscopy and infrared spectroscopy. This re- origins remain unknown. Knowledge of the role of pigments in search focuses on a novel method for studying the content of the larval shell could enhance our current understanding of larval the shells of bivalve larvae: Raman spectroscopy. Raman spec- shell formation as well as provide applications in larval ecology troscopy is a nondestructive method that provides qualitative research, particularly aiding in identification of larvae in the field. structural information on a mixture of organic molecules in a Much of our current understanding of the larval shell structure sample. By radiating a sample with focused laser light, a spectrum comes from comparisons with adult shells. Unlike adult shells, is produced on the basis of the light scattering from the excited which can be either calcite or aragonite or a mixture of the two, functional groups in organic molecules. Raman spectroscopy has all bivalve larval shells consist of aragonite, a calcium-carbonate been employed to study pigments in calcium-carbonate material polymorph, embedded in an organic matrix.[4] Bivalve larvae start in various mollusk shells,[1,12,13] pearls from bivalves,[2,3] and out by secreting a shell made of amorphous calcium-carbonate, corals.[14,15] Pigments that have been identified in mollusk shells which then becomes a crystalline aragonite after a few days.[5] include substituted polyenes such as carotenoids, unsubstituted Whereas adults can have different shell ultrastructure based on polyenes, and porphyrins.[12] Most commonly, Raman spectra indi- species groups, larval shell ultrastructure observed using electron cate the presence of all-trans-polygenic pigments containing microscopy is similar between species indicating that larval shell formation is a highly conserved evolutionary process.[4–6] Larval shells have two main components: the youngest part of the shell, * Correspondence to: Christine M. Thompson, University of Maryland Center for prodissoconch I (PDI), which is secreted <24 h after fertilization, Environmental Science, Horn Point Laboratory, Cambridge, MD 21613, USA. and the prodissoconch II (PDII), which is secreted approximately E-mail: [email protected] 40 h after fertilization and contains regions of mineral between a University of Maryland Center for Environmental Science, Horn Point Labora- darker growth bands.[7] Genera of adult bivalves have different [8] tory, Cambridge, MD, 21613, USA organic matrix organizations reflecting a unidirectional miner- 349 alogical evolution at the superfamily level with divergences in b Woods Hole Oceanographic Institution, Woods Hole, MA, 02543, USA

chains of conjugated C=C bonds (i.e., polyenes). These pigment We compared the spectra from the PDI region to the later larval bands exhibit strong absorption due to coupling of electronic and shell, and we compared spectra between growth band and vibrational translations and can be detected at low concentrations mineral areas. The third part of the study tested whether PDI in the shells even among other biological materials.[3,16] Although shells of early larvae incorporate pigments from their surround- structurally similar to β-carotene and other natural carotenoids, ing environment. We reared three clam species in the laboratory these polyenes showed different spectral bands when compared in waters with different organic signatures to test if pigment with those of pure β-carotene, indicating that they have different peaks in Raman spectra differed between treatments and chain lengths, lack methyl groups (unsubstituted), and occur in thereby determine if early larvae that are not feeding might multiple combinations.[1–3,12,15,17] In bivalves and pearls, different incorporate pigment from surrounding waters into their shells. pigment peaks were seen in different colored regions of the specimens,[1,2] indicating that they are responsible for coloration. Different chain lengths, combinations, and relative proportions of Methods pigments can be responsible for coloration, but pigments can be detected in low concentrations due to resonance even if they do We carried out micro-Raman analysis on seven species of lab- not play a role in coloration.[3] In mollusks, polyenes likely form spawned or hatchery-spawned larvae and two species which were complexes with organic matrix proteins[3] and may play additional grown in experimental conditions. This analysis targeted the pig- roles stabilizing the aragonite crystals and mediating crystal- ment regions of the spectra to determine (1) whether pigments lization.[2] In corals, C=C bands of polyenes shifted to lower in the larval shell differed between species, (2) how pigment com- wavenumbers after demineralization, indicating interactions with position differed within a shell, and (3) if early PDI larval shells incor- the mineral constituents.[18] A Raman mapping study on an adult porate pigments from their surrounding environment. gastropod shell revealed thin peaks of polyenes associated with growth lines.[13] Similar molecules are present in bivalve larvae Differences between species and may have comparable roles (S.M. Gallager, unpub. data). Raman spectroscopy of bivalve larval shells could have useful We obtained 7-day-old (7 d) Argopecten irradians (bay scallop), applications for identification or tracing origins of bivalve larvae. Mercenaria mercenaria (hard clam), and Crassostrea virginica Although Raman spectroscopy has been applied as a chemotax- (eastern oyster) larvae from the Aquaculture Research Corpora- onomic method to distinguish microbial cells[19,20] and species of tion in Dennis, MA, in 2011 and 6–8 days Mulinia lateralis (coot lichens and other fungi,[21,22] it has not been tested for bivalve clam), Tagelus plebeius (razor clam), Ischadium recurvum (hooked larvae. In addition, it is not known if the pigment material in larval mussel), and Rangia cuneata (common rangia) larvae from shells could be associated with waters of their natal origin. If adults spawned in the laboratory at University of Maryland these pigments are associated with other organic molecules Center for Environmental Science, Horn Point Laboratory, in known to accumulate in growth bands,[23] we might see more 2009. Larval samples had been stored in ethanol or formalde- presence of pigments in this region. Raman spectra of fish oto- hyde prior to analysis. liths have shown higher organic signals in dark bands; however, Before the micro-Raman analysis, it was necessary to pretreat these were signals from proteinaceous material and not from larvae to remove excess material that may be on the shell and tis- pigments.[24] It has been speculated that pigments are modified sue. Because larval shells are clear, the fluorescence from tissue from food, and therefore, different colors and pigments found overwhelms the Raman signal. First, larvae were in rinsed in de- in the same species of mollusk may be due to different ionized water. Next, larvae were soaked in vials with 40% sodium habitats.[12] Although the source of pigments in larval shells is hypochlorite (common bleach) for 15 min or up to an hour for unknown, if larvae incorporate pigments via algal food (metabolic) larger shells. Vials were lightly shaken to break up the shells sources, pigments from the early PDI shell may differ from those in and remove tissue. The bleach was rinsed off with deionized wa- the later shell because the larva does not feed during PDI ter and set out on a glass slide to air-dry overnight. This proce- formation. Recent isotopic studies have revealed that in early dure follows that of Weiss et al., in which larval shells were larvae, shell carbon is formed from dissolved inorganic carbon, prepared for Raman analysis in this manner.[5] but tissue material arises from maternal carbon.[25] During this We recorded spectra using a Kaiser Optical Systems, Inc. laser period, pigments may be incorporated via another mechanism Raman spectrometer. The system includes the Kaiser NRxn spec- such as through dissolved organic carbon (DOC) uptake. If the trograph with a thermoelectric-cooled charge-coupled device DOC contains pigment molecules (most likely plant-based), this (CCD) camera (2048 × 512 pixels) fiber-optically coupled to a could potentially enable Raman spectra to reflect the pigment Leica DM LP microscope. We used a 100 mW Nd:YAG laser as signature of a larva’s natal waters. the excitation source operating at 532 nm. The laser was focused We used micro-Raman spectroscopy to investigate fundamen- using a 50× objective at 10–15 mW. Spectra were recorded in the tal questions regarding the role of pigments in bivalve larval range of 100–4400 cmÀ1 of Raman shift with a spectral resolution shells: what pigments are present, where they are present, and of ~5 cmÀ1. A single spectrum was obtained from an average of if they are influenced by organic material in source waters. First, 10 accumulations, and exposure times were adjusted on the basis we determined how pigment composition may vary between of the fluorescence in the sample to avoid saturation of the CCD. larvae from seven different bivalve species. We hypothesized that Wavelength calibration was performed using a neon light source, different species would contain unique pigment peaks due to and the system was calibrated daily using the 520 cmÀ1 peak of a differences in coloration and organic matrix organization be- polished silicon wafer. tween species. Next, we analyzed how pigments were distributed Raman spectra were collected on a total of three to five shells within a larval shell by sampling spectra along dorsal-ventral per species. To record a spectrum, the laser was focused on a

350 positions of the larval shell for four species and comparing speciﬁc region of the larval shell. On all shells, spectra were taken organic pigment peaks to inorganic calcium carbonate peaks. at positions starting at the PDI region and moving in a ventral

wileyonlinelibrary.com/journal/jrs Copyright © 2014 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2014, 45, 349–358 Analysis of bivalve larval shell pigments using micro-Raman spectroscopy direction. For one shell of each species, three spectra were taken calculating peak areas and ratios. For each spectrum, we calcu- from each of seven regions of the shell (Fig. 1): the PDI, the PD I/II lated the integrated area for each organic peak and compared À line, the first growth band (GB1), the GB1 line, the second growth it as a ratio to the aragonite 1085 cm 1 peak. For C. virginica,we band, the second growth band line, and the shell margin (SM). chose other locations between 1200 and 1600, where nona- For the remaining shells, only one spectrum was taken from ragonite peaks were present. These bands are likely associated each of the seven regions. with biomolecules similar to those seen in spectra from other Spectra were analyzed using a freely available integrated soft- biological samples.[28] These ratios were used in a multivariate ware system for processing Raman spectra[26] implemented in analysis of variance (MANOVA) test using R (v. 2.14.0) for each MATLAB (v. R2011a). This system allowed for preprocessing, species to compare the differences in inorganic : organic material plotting, averaging, and classifying spectra. All spectra were across the shell. To test if the ratio of inorganic : organic material preprocessed by baseline correction with intensities standardized was a function of shell area in the regions between bands, we to a scale of 0–1 based on the highest peak. Descriptive plots of calculated the area of each region in pixels using images of each averaged spectra were created to visually compare spectral shell and the image analysis toolbox in MATLAB. A regression characteristics between each species. An average spectrum was with replication was performed using the calculated ratios for calculated by averaging the values from individual spectra at each species against region area. each Raman shift. To determine if the spectra of species were different from one Pigment origins another, a principal component analysis (PCA) was performed and classification tests were conducted using the PCA scores. Laboratory experiments were performed with larvae from three The PCA was performed using all spectra from each species for clam species reared in waters with different organic signatures 1cmÀ1 bins along the 600–1800 cmÀ1 region where pigment to test if early PDI larval shells incorporate pigments from their bands were present. Between 14 and 37 spectra were used for each surrounding environment. Water for the two organic signature species. Three classification methods were employed: a discrim- treatments was collected from (1) the Choptank River at the Horn inant function analysis, a neural network, and a support vector Point dock and (2) the Little Choptank River, another subestuary machine (SVM)[27] to classify the PCA scores from each species. of the Chesapeake Bay that is south of the Choptank River. A leave-one-out cross-validation method was used to calculate Salinities at both sites ranged from seven to eight. Water was classification accuracy. This works by training the classifier each collected on the same day that the bivalves were spawned time, leaving out one spectrum and classifying the left-out (described in the succeeding text) and was filtered in the field spectrum into the category with the highest probability. For using a battery-operated pump and polypropylene cartridge sys- the purposes of this study, successful groups were considered tem with a 1-μm filter. A pigment analysis using high-performance to have >85% correct classifications. liquid chromatography was performed on 30–60 ml of water collected from each site at the start of each experiment to detect presence of pigments in waters. Analysis was performed by the Within-shell differences Horn Point Laboratory Analytical Services Department. In order to determine how pigment composition differed within Three species of larvae were reared in the water from both river each shell, we used spectra that were taken in replicates of three systems. Ripe adult broodstock of M. lateralis (family: Mactridae), along the dorsal-ventral direction of one shell (Fig. 1) for A. Macoma mitchelli (family: Tellinidae), and Mytilopsis leucophaeata irradians, M. mercenaria, R. cuneata, and C. virginica. In this analy- (family: Dreissenidae) were collected from the Choptank River in sis, we used the MATLAB routine for identifying peaks and June 2012. Three experiments were conducted, one for each clam species. Larvae were spawned in the laboratory from June to August by subjecting adults to alternating hot and cold temperatures to induce spawning.[29] Fertilized eggs were divided between the two water treatments and placed in 3-l glass À containers at starting densities of 23–30 eggs ml 1.After24h, approximately 200 larvae were sampled and preserved in ethanol. The remaining larvae were fed a mixture of Isochrysis galbana (strain C-ISO) and Thalassiosira pseudonana (strain 3H) at a density of 5 × 105 cells mlÀ1 and kept for another 24 h. After this period, all 2-day-old larvae were preserved in ethanol. Raman spectra on these shells were taken with an XploRA con- focal Raman microscope by Horiba Jobin Yvon, Inc. The system includes a flat field spectrograph with a multichannel air cooled CCD and color camera optically coupled to an Olympus BX41 microscope. We used a 532 nm, 25 mW solid state laser as the excitation source through a 100× objective using a 1200 H grating and hole and slit size of 300 and 100, respectively. Spec- tra were recorded in the range of 200–2000 cmÀ1 with a spectral Figure 1. Locations for Raman sampling on a 7-day-old prodissoconch II resolution of ~2 cmÀ1. Spectra were acquired from 20 larval shells larva of Crassostrea virginica. Spectra were sampled from seven regions of from the PD1 region for each shell by averaging three accumula- the shell in an anterior–posterior transect. The regions are the following: prodissoconch I (PDI), prodissoconch I/II line (PDIL), first growth band tions with an exposure time of 10 s. Spectral acquisition was (GB1), first growth band line (GB1L), second growth band (GB2), second controlled using the LabSpec software (version 6). Wavelength 351 growth band line (GB2L), and shell margin (SM). Scale bar = 15 μm. calibration was performed on the XplorRA system using a neon

light source with daily calibrations using a silicon wafer. Mul- different species (Fig. 2b–d). For C. virginica, no strong peaks in tivariate analysis using an SVM was performed using the Raman these regions were observed, so the maxima in each pigment processing program with spectra of larvae from each water treat- region were plotted in Fig. S1. C. virginica, I. recurvum, and A. ment grouped as categories. irradians had the most distinguishable peaks between species. Some bands, such as those exhibited for I. recurvum, clearly show presence of multiple pigment peaks within the ν2 and ν1 regions Results (Fig. 2c–d). Classification of PCA scores showed that Raman spectra were Differences between species distinguishable between most species. Both the neural network The spectra from four to six individuals of each of the seven and SVM classifiers on the full spectrum had the best classifica- bivalve species were averaged to examine general differences tion success, with spectra from five of the seven species classified between species (Fig. 2a). All spectra showed the sharp character- to their correct category greater than 85% of the time (Fig. 3). istic bands from aragonite at 705 and 1085 cmÀ1.[30,31] No mineral This confirms that the spectral differences among most of the controls were used as it is known that bivalve larvae contain species may be unique; however, T. plebeius had very low aragonite as opposed to calcite,[4] and the Raman spectra from accuracies in all tests. all shells in our samples verified this. The greatest differences occur in the regions where signals from organic molecules are Differences within shells present, around 600–1800 cmÀ1 (Fig. 2b–c). The strongest organic bands, commonly occurring around 1100–1130 cmÀ1 For four species, we used PDII larval shells to investigate how À1 and 1500–1530 cm , correspond to the ν2 and ν1 stretching spectral patterns differed at points along a dorsal-ventral direc- vibrations of carbon–carbon single (C–C) and double bonds tion and between regions where growth bands are present (C=C), respectively. Smaller peaks around 1000–1020 cmÀ1 likely (Fig. 1). We observed differences in peak wavenumber and correspond to CH out-of-plane wagging mode (ν4) in the inorganic : organic material from M. mercenaria, A. irradians, and pigments as demonstrated for other bivalves[1,12] but could also R. cuneata spectra at single bond, double bond, and CH peaks. correspond to methyl groups substituted along the carbon M. mercenaria had the smallest amount of differences, with the chains.[2] These three regions are likely associated with the pres- other two species having greater outliers in the PDI and SM ence of polyenes, which are commonly found in Raman spectra regions. For C. virginica, we chose other organic peaks indicated from adult mollusks,[1] as indicated by the lower wavenumber by wavenumber. Three-dimensional scatter plots of peak wave- shift of the C–C bands indicating nonmethylation.[16] numbers show most points for spectra clustered together with Within the three regions, each species had different locations a few outliers (Fig. 4). This indicates that the same polyenes or of peaks in the Raman spectrum, suggesting that different com- groups of polyenes are present throughout the shell. It was binations of polyenes make up the pigments in larval shells of often the region associated with the CH group that had the

À Figure 2. Standardized mean Raman spectra for seven bivalve species for (a) the region of 200–2300 cm 1 and (b–d) regions corresponding to pigments. Spectra were preprocessed and standardized before averaging. Focused pigment regions correspond to (b) CH out of plane wagging (ν4), (c) 352 single-bond stretching (ν2), and (d) double-bond stretching (ν1). Species key: AI = Argopecten irradians,CV=Crassostrea virginica,IR=Ischadium recurvum, ML = Mulinia lateralis,MM=Mercenaria mercenaria,RC=Rangia cuneata,TP=Tagelus plebeius.

For M. mercenaria, ratios for all three functional groups were lower at the PDI and SM and were significantly different from the other shell locations (C–C, p = 0.019; C=C, p = 0.0002; CH, p = 0.001) but not different from each other. Although R. cuneata showed similar patterns to M. mercenaria, no summary ANOVAs were significant despite a significant MANOVA. Finally, all peak ratios for C. virginica were significantly different across the shell (1031, p = 0.0002; 1134, p = 0; 1316, p = 0.001; 1530, p = 0.002), with the GB1 and the SM having the most variation. We performed regression analyses for each species to determine if the total area between growth bands described a signifi- Figure 3. Classification accuracies using discriminant function analysis cant amount of variability in the inorganic : organic material ratios (DFA), neural network (NN), and support vector machine (SVM) classifica- fi as seen in the Raman spectra. Only M. mercenaria had a strong tions for Raman spectra of seven species of bivalve larvae. Classi cation – 2 accuracies refer to the percentage of spectra that were correctly classified positive regression for C C and C=C positions (R = 0.79 and into the correct species category in a leave-one-out cross-validation proce- 0.76, respectively). This indicates that there is a higher ratio of dure. Dashed line denotes 85% accuracy (our criterion for a successful clas- aragonite to pigments in regions with larger surface area in sification). Each species group contained between 14 and 37 spectra for a M. mercenaria, such as the PDI region, but no other species total of 190 spectra. Species key: AI = Argopecten irradians,CV=Crassostrea showed this pattern. virginica,IR=Ischadium recurvum,ML=Mulinia lateralis,MM=Mercenaria mercenaria,RC=Rangia cuneata,TP=Tagelus plebeius. Differences in water sources most variation in peak wavenumber. No clear patterns emerged We tested whether Raman spectra of the shells of unfed 1-day-old with the peaks in the C. virginica scatter plot. larvae of three clam species produced different peaks when We tested whether the ratio of inorganic : organic material reared in waters with different organic signatures. Pigment changes across the shell using intensities for the same peak analysis of the treatment waters revealed that for the M. wavenumbers pictured in Fig. 4. For each species, a MANOVA lateralis (Fig. 6 a–b) and M. leucophaeata experiments (Fig. 6 indicated that there were significant differences in the ratios e–f), the Little Choptank River water had higher pigment con- (alpha = 0.05) across the shell, and post hoc comparison tests centrations with different compositions of photoprotective showed that the wavenumbers with significantly different ratios and photosynthetic carotenoids than the Choptank River water. were not the same between species (Fig. 5). For A. irradians, the For the M. mitchelli experiment (Fig. 6 c–d), pigment concentra- only significant difference was found between the CH group tions were lower than the previous experiments, and the com- between the SM and PDI line (p = 0.04), but these two groups positions of individual pigments were more similar between were not significantly different from the other shell locations. the Choptank and Little Choptank Rivers.

Figure 4. Scatterplot of peak values representing single bond (v2), double bond (v1), and CH (v4) peaks from Raman spectra along a transect of (a) Argopecten irradians, (b) Mercenaria mercenaria, (c) Rangia cuneata, and (d) Crassostrea virginica shells. Different locations were used for C. virginica because the spectra did not show peaks in the same regions as the other species. Transect positions were prodissoconch I (PDI), prodissoconch I/II line (PDIL), first growth band (GB1), first growth band line (GB1L), second growth band (GB2), second growth band line (GB2L), and shell margin (SM). Each 353 location is represented by a different color and shape (key in center of figure). This figure is available in colour online at wileyonlinelibrary.com/journal/jrs

Figure 5. Ratio of inorganic : organic material based on integrated peak areas for Raman spectra taken from locations along a shell transect of (a) Argopecten irradians, (b) Mercenaria mercenaria, (c) Rangia cuneata, and (d) Crassostrea virginica. For (a–c), peak areas were taken from peaks À1 representing carbon–carbon single bond (v2), double bonds (v1), and a methyl group (v4), and used with the area of the aragonite 1085 cm to create the inorganic : organic ratio. For (d), for C. virginica, organic peaks were chosen on the basis of peaks thought to be associated with organic molecules. Asterisks indicate if the ratio of material was significantly different between one or more regions across the shell (p < 0.05). Shell locations: PDI = prodissoconch I, PDIL = prodissoconch I/II line, GB1 = first growth band, GB1L = first growth band line, GB2 = second growth band, GB2L = second growth band line (GB2L), SM = shell margin.

Descriptive analysis of the Raman peaks for each species that were previously unknown and brings forth potential applica- showed that there are differences in spectra between water treat- tions for using Raman data in ecological studies. ments and larval age that is species-specific (Fig. 7). The spectra Raman spectra of the seven different species studied showed from 1-day-old (1 d) M. lateralis raised in water from the Little distinct patterns in spectra that arose from aragonite and Choptank (Fig. 7a) had pigment peaks at a slightly different pigment molecules embedded within the organic matrix. With wavenumber and intensity compared with spectra of the same the exception of C. virginica, the larval shells contained pigments À age larvae raised in Choptank River water (see ~1530 cm 1). M. similar to those found in adult shells.[1] These pigments are likely mitchelli samples showed more similar spectra between 1-day- nonmethylated (unsubstituted) polyenes with chain lengths old larvae in terms of both intensity and wavenumbers of peaks, between eight and 15 conjugated double bonds based on the and M. leucophaeata did not show any strong pigment signals in wavenumber shift of the ν2 bands and comparisons with Raman its spectra. analysis of carotenoids and protein-carotenoid complexes.[16,17] Classification analysis using an SVM indicated Raman spectra These molecules are structurally different than those present in were different between water treatments for the two clams periostracum, which are mainly polyphenols and quinones, and that had pigment peaks (Fig. S2). M. lateralis had 100% classi- also have coloring properties.[7] Recent studies have suggested fication accuracy between the day 1 Choptank and Little that peaks associated with polyenes in biominerals are likely Choptank River shells, indicating that these spectra were very composed of a variety of pigment molecules of different lengths distinct. M. mitchelli had 90% classification accuracy also indi- that can be dissected using full with at half-maximum data from cating distinct spectra. The spectra of larvae reared in Choptank peaks. On the basis of the width of the bands observed, it is likely River water were not strongly different between 1- and 2-day-old that the spectral bands from these species contain different larvae for any species, but those reared in the Little Choptank River groups of pigments in different orientations within the matrix,[2,3] water had distinct spectra between 1- and 2-day-old larvae for all most notably the spectra of I. recurvum present multiple peaks, À1 three species. particularly around 1500 cm (ν1). In terms of coloration, although most larvae appear clear, these pigments may contribute to color in the later shell. It is possible that the same pigments Discussion may absorb differently in different genera because of their orientation within the organic matrix,[3] and similar colors of pearls This study has demonstrated that Raman spectroscopy can be a from different families have different pigment mixtures.[16] If the useful tool for investigating fundamental questions of the roles role of pigments in the organic matrix is to mediate crystallization of pigments in the larval shell matrix. We were able to determine of aragonite,[2] then slight differences in matrix structure as the distribution of pigments within the larval shell, how they indicated by polarized light microscopy of different species[9–11]

354 differ between species, and speculate on how they are formed. may result in different pigment organizations that can be seen in This study demonstrates important aspects of bivalve larval shells Raman spectra.

Figure 6. Pigment analysis of ﬁltered river water used in experiments with (a–b) Mulinia lateralis,(c–d) Macoma mitchelli, and (e–f) Mytilopsis leucophaeata larvae. Sums of pigment concentrations are shown on the left, and proportions of individual pigments are represented on the right. KEY: TChl = total cholorphyll, PPC = photoprotective carotenoids (Allo, Diad, Diato, Zea, and Caro), PSC = photosynthetic carotenoids (But, Fuco, Hex, and Peri), Allo= alloxanthin, Diad = diadinoxanthin, Diato = diatoxanthin, Zea = zeaxanthin, Caro = carotenoids, Fuco = Fucoxanthin, He = hexanoyloxyfucoxanthin, Peri =peridinin. This ﬁgure is available in colour online at wileyonlinelibrary.com/journal/jrs

C. virginica did not show any peak characteristic of the spectra I. recurvum showed the most significant clustering on the ba- of polyene molecules. Although peaks were present in the sis of the classification analysis, whereas M. lateralis, R. cuneata, À 1200–1600 cm 1 region, these peaks were lower and broader T. plebeius and M. mercenaria did not cluster as well and were than the peaks from the other species. Weiss et al.[5] failed to pro- often misclassified. This could represent similarities between duce spectra of the closely related Crassostrea gigas larval shell growth conditions, and thus pigments present, although growth due to intrinsic fluorescence with Raman. Their study used a conditions differed for these species (M. mercenaria specimens 780-nm laser, and longer wavelengths may not excite some came from a Cape Cod hatchery, and the other three were raised pigments as well as green/blue wavelengths.[2] C. virginica may in the Horn Point Laboratory). These four species belong to the contain fewer or different pigments than other species (as adult veneriod order of bivalves and could have similar shell formation C. virginica shells are white), but it is possible a shorter wavelength processes requiring a particular suite of pigments inside the may excite these pigments assuming they are different. White organic matrix, whereas the other three species studied represent pearls showed both polyene pigment peaks when excited with orders Pterioida (A. irradians and C. virginica)andMytiloidoa a 487.8-nm laser, but peaks were all below detection limits for (I. recurvum). Similarities of spectra among some species lasers with longer wavelengths[2] and appeared similar to our may reflect evolutionarily patterns of shell formation between C. virginica spectra. It may be that our 532-nm wavelength was closely related species. Further investigations of these species using too long to excite pigment molecules in larval C. virginica speci- different wavelengths may be able to parse out these similarities mens. The smaller peaks in the 1200–1600 cmÀ1 regions may be using resonance effects. associated with organic material, possibly proteins or porphyrins, Although we expected the presence of pigments to be higher as spectra from other biological materials have shown small peaks in dark bands of the larval shell, this was not observed in our shell in this region.[28,31–33] analysis. Growth lines were not significantly different from their

We used three classiﬁcation methods to determine how neighboring regions and did not seem to contain consistently 355 spectra differed between species. A. irradians, C. virginica,and more or less inorganic material in relation to pigment. It is

Figure 7. Mean Raman spectra for bivalve larvae of (a) Mulinia lateralis, (b) Macoma mitchelli, and (c) Mytilopsis leucophaeata for day 1 (unfed, D1) and day 2 (fed, D2) treatments reared in ﬁltered Choptank River (CR) and Little Choptank River (LC) waters. Spectra were preprocessed and standardized before averaging.

possible they may contain increased organic material that was require a different suite of pigments. Additionally, it is known that not visible in the Raman spectra, such as organic matrix proteins metabolic carbon can contribute to shell carbon in larval bivalves which may not appear as strong, resonating Raman peaks. in later larval stages,[25,36] so it may be possible that the organic Most of the variability along the length of the shell occurred in material in the shell may have similar sources. If bivalves incorpo- the PDI region or the SM region. The SM is where shell is actively rate pigments from food via metabolic processes,[2] then the PDI forming and thus may not have the full pigment composition may have a different composition than is present in the rest of present in the new shell. There are currently two models of shell the shell. The larva does not feed during PDI formation and formation for bivalves: an extracellular matrix model and a cellu- may incorporate pigments either from maternal sources or lar-mediated model involving hemocytes.[34,35] It is unknown if surrounding water. Raman spectra of bivalve eggs contain peaks the cellular model applies to larvae, but it has been suggested in similar regions to the pigments in larval shells (C. Thompson, that due to energy demands on developing larvae, energetically unpub. result), which may provide a source for them in the expensive strategies involving organic molecules and proton larval shell. pumping are not likely employed, and the larvae likely favor We reared three species of veneroid clams in waters with simple dissolved inorganic carbon incorporation.[25] It has been different organic signatures to test whether early shell formation

356 observed that the PDI may form differently than the rest of the incorporated pigments from water sources during the stage shell by ﬁrst secreting amorphous aragonite[2] and thus may when larvae do not feed. Water from the Little Chopank River

wileyonlinelibrary.com/journal/jrs Copyright © 2014 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2014, 45, 349–358 Analysis of bivalve larval shell pigments using micro-Raman spectroscopy had higher concentrations of pigments than that from the for their assistance with the larval rearing experiments. The Aqua- Choptank River in the M. lateralis and M. leucophaeata experi- culture Research Corporation in Dennis, MA, and the Horn Point ments. The differences in pigment concentrations could explain Laboratory Oyster Hatchery provided oyster larvae used in this the divergence between the spectra of 1-day-old M. lateralis study. S. Alexander, E. Vlahovich, and L. Guy at the Horn Point reared in water from the Little Choptank and that of 1-day-old M. hatchery provided useful advice and assistance with the larval lateralis reared in water from the Choptank River and support the experiments. This is UMCES-HPL contribution number 4884. idea that DOC could be directly incorporated into the shell. It is unlikely, however, that the actual pigments from the source waters are incorporated because the spectra reveal Raman signatures for polyenes and not the carotenoids isolated by References high-performance liquid chromatography. In corals, for instance, [1] C. Hedegaard, J. Molluscan Stud. 2006, 72, 157. algal pigments such as peridinin and diadinoxanthin were found [2] A. L. Soldati, D. E. Jacob, U. Wehrmeister, T. Hager, W. Hofmeister, in tissue but not in the skeleton (sclerites), which only had J. Raman Spectrosc. 2008, 39, 525. polyene bands.[18] M. lateralis pigment peaks from different [3] S. Karampelas, E. Fritsch, J.-Y. Mevellec, S. Sklavounos, T. Soldatos, Eur. J. Mineral. 2009, 21, 85. wavenumbers between 1- and 2-day-old larvae reared in the [4] M. Carriker, R. Palmer, Proc. Natl. Shellfish Assoc. 1979, 69, 103. Little Choptank River water suggest that a different suite of pig- [5] I. M. Weiss, N. Tuross, L. Addadi, S. Weiner, J. Exp. Zool. 2002, ments were present in the 2-day-old shells, which may have 293,478. resulted when the 2-day-old larvae began to feed and changed [6] J. Carter, in Skeletal Growth of Aquatic Organisms, (Eds: D. Rhoades, – from incorporating DOC or maternal sources to pigments derived R. Lutz), Plenum, New York, 1980, pp. 69 113. [7] M. Carriker, in The Eastern Oyster Crassostrea virginica (Eds: V. from food. No difference was observed in the M. leucophaeata Kennedy, R. I. E. Newell, A. Eble), Maryland Sea Grant College, College spectra, most likely because this species did not show pigment Park, MD, 1996, pp. 75–168. peaks in the Raman spectra. The spectra of shells of 1-day-old M. [8] J. Taylor, W. Kennedy, A. Hall, Bull. Br. Museum Nat. Hist. 1969, 3,1. mitchelli also were distinct between water treatments using [9] S. Tiwari, S. M. Gallager, Proceeding 2003 IEEE Conf. Image Process. fi Barcelona, Spain, 2003, 1061. classi cation methods; however, the pigment analysis did not [10] S. Tiwari, S. M. Gallager, Proc. Ninth IEEE Int. Conf. Comput. Vision, reveal strong differences in pigment concentrations between Nice, Fr. Oct. 14-17, 2003,1. river systems. So it remains inconclusive as to whether differ- [11] C. M. Thompson, M. P. Hare, S. M. Gallager, Limnol. Oceanogr. ences in spectra of M. mitchelli are a result of DOC incorporation Methods 2012, 10, 538. or maternal sources in early shell formation. For both M. lateralis [12] W. Barnard, D. de Waal, J. Raman Spectrosc. 2006, 37, 342. [13] G. Nehrke, J. Nouet, Biogeosciences Discuss. 2011, 8, 5563. and M. mitchelli, no difference was observed in spectra for [14] B. Kaczorowska, A. Hacura, T. Kupka, R. Wrzalik, E. Talik, G. Pasterny, Choptank River larvae, but it may be due to lower concentrations A. Matuszewska, Anal. Bioanal. Chem. 2003, 377, 1032. of DOC. Additionally, no difference was observed for the PDI shell [15] T. Kupka, H. M. Lin, L. Stobiński, C.-H. Chen, W.-J. Liou, R. Wrzalik, for 2-day-old M. lateralis or M. mitchelli between river systems, Z. Flisak, J. Raman Spectrosc. 2009, 41, 651. [16] E. Fritsch, B. Rondeau, T. Hainschwang, S. Karampelas, in Applica- which suggests that 1 day of no feeding may not be enough to tions of Raman Spectroscopy to Earth Sciences and Cultural reflect a strong signature in the early PDI shell. Although we Heritage, (Eds: J. Dubessy, F. Rull, M. C. Caumon), European Mineral- cannot say with certainty what was causing the observed differ- ogical Union and Mineralogical Society of Great Britain & Ireland, ences between and within species, evidence presented here may Great Britain, Vol. 12, 2012, pp. 455–489. suggest that environmental conditions could affect the pigment [17] J. C. Merlin, Pure Appl. Chem. 1985, 57, 785. [18] L. F. Maia, V. E. de Oliveira, M. E. R. Oliveira, F. D. Reis, B. G. Fleury, H. organization in early shells of some species. If this is the case, then G. M. Edwards, L. F. C. de Oliveira, J. Raman Spectrosc. 2013, 44, 560. pigment patterns in shells of larvae could be used to determine [19] W. E. Huang, R. I. Griffiths, I. P. Thompson, M. J. Bailey, A. S. Whiteley, whether larvae originated in different river systems, at least for Anal. Chem. 2004, 76, 4452. those species that accumulate pigment in their early shells. [20] P. Rosch, M. Harz, M. Schmitt, K. Peschke, O. Ronneberger, H. fi Burkhardt, H. Motzkus, M. Lankers, S. Hofer, H. Thiele, Appl. Environ. In conclusion, these ndings support the potential for using Microbiol. 2005, 71, 1626. Raman spectra to identify some species of bivalves in field sam- [21] H. G. M. Edwards, E. Newton, D. Wynn-Williams, R. Lewis-Smith, ples, particularly for those species with spectra that differ from Spectrochim. Acta Part A 2003, 59, 2301. those of co-occurring groups. In this study, C. virginica presented [22] K. De Gussem, P. Vandenabeele, A. Verbeken, L. Moens, Anal. spectra that were different from six other species found in the Bioanal. Chem. 2007, 387, 2823. [23] R. A. Lutz, D. C. Rhoads, Science 1977, 198, 1222. Choptank River, so it is possible that spectra could be used to [24] A. Jolivet, J.-F. Bardeau; R. Fablet; Y.-M. Paulet, H. de Pontual, Anal. be used to identify C. virginica in this system. Raman spectral Bioanal. Chem. 2008, 392, 551. methods may improve identification accuracies for certain spe- [25] G. G. Waldbusser, E. L. Brunner, B. A. Haley, B. Hales, C. J. Langdon, cies and could be used to augment evolving technologies such F. G. Prahl, Geophys. Res. Lett. 2013, 40, 2171. fi [11] [26] L. A. Reisner, A. Cao, A. K. Pandya, Chemom. Intell. Lab. Syst. 2011, as polarized light identi cation techniques. More studies are 105, 83. currently being conducted to test the ability of this method to [27] C. Chang, C. Lin, ACM Trans Intell Syst Technol 2001, 2,1. distinguish spectra of larval bivalves using multiple laser wave- [28] V. E. de Oliveira, H. V. Castro, H. G. M. Edwards, L. F. C. de Oliveira, J. lengths, with consideration for possible differences present in Raman Spectrosc. 2009, 41, 642. early shells based on growing conditions. [29] M. M. Helm, N. Bourne, A. Lovatelli, Hatchery culture of bivalves, Rome, 2004, pp. 1–92. [30] J. Urmos, S. Sharma, F. Mackenzie, Am. Mineral. 1991, 76, 641. Acknowledegments [31] S. N. White, Chem. Geol. 2009, 259, 240. [32] P. A. Tarantilis, A. Beljebbar, M. Manfait, M. Polissiou, Spectrochim. This research was funded by The Bailey Wildlife Foundation and Acta Part A 1998, 54, 651. [33] W. R. B. Chowdry, J. Silver, H. G. M. Edwards, Spectrochim. Acta Part A the National Science Foundation (OCE-0829512, OCE-1240266). 2003, 59, 2207. We would like to thank J. D. Goodwin, K. Hinson, K. Hanrahan, R. [34] A. S. Mount, A. P. Wheeler, R. P. Paradkar, D. Snider, Science 2004, 357 Woodward,T.Wazniak,J.Thalmann, A. J. Schlenger, and V. Kennedy 304, 297.

[35] G. Zhang, X. Fang, X. Guo, L. Li, R. Luo, F. Xu, P. Yang, L. Zhang, N. Li, L. Qian, G. Zhang, Y. Li, H. Yang, X. Liu, J. Wang, Y. Yin, J. Wang, X. Wang, H. Qi, Z. Xiong, H. Que, Y. Xie, P. W. H. Holland, J. Paps, Nature 2012, 490, 49. Y. Zhu, F. Wu, Y. Chen, J. Wang, C. Peng, J. Meng, L. Yang, J. Liu, [36] E. F. Owen, A. D. Wanamaker, S. C. Feindel, B. R. Schöne, P. D. Rawson, B. Wen, N. Zhang, Z. Huang, Q. Zhu, Y. Feng, A. Mount, D. Hedgecock, Geochim. Cosmochim. Acta 2008, 72, 4687. Z. Xu, Y. Liu, T. Domazet-Lošo, Y. Du, X. Sun, S. Zhang, B. Liu, P. Cheng, X. Jiang, J. Li, D. Fan, W. Wang, W. Fu, T. Wang, B. Wang, J. Zhang, Z. Peng, Y. Li, N. Li, J. Wang, M. Chen, Y. He, F. Tan, X. Song, Q. Zheng, Supporting information R. Huang, H. Yang, X. Du, L. Chen, M. Yang, P. M. Gaffney, S. Wang, L. Luo, Z. She, Y. Ming, W. Huang, S. Zhang, B. Huang, Y. Zhang, Additional supporting information may be found in the online T. Qu, P. Ni, G. Miao, J. Wang, Q. Wang, C. E. W. Steinberg, H. Wang, version of this article at the publisher’s web site. 358