The most commonly expressed genes in the Laternula dataset comprise various functional classes, which is reflected in the overall GO classifications (Figure 2). As stated previously, the edge of the mantle comprises three folds and the periostracum with the tissue for this transcriptome analysis taken from a cross section across all layers. BLAST sequence similarity searches revealed a wide range of diverse functions among the most commonly expressed genes (contigs comprising over 300 individual sequences) (Table 1) reflecting the complex contractile and secretory nature of this organ.

The mantle, whilst not a muscle per se, is contractile and hence many of the highly expressed sequences consist of structural or muscle-related genes, such as actin, collagen, troponin, calponin, adipose differentiation-related protein and myosin 55, although some e.g. collagen, may also be involved in shell synthesis 56. Interestingly, the most commonly expressed sequence is that of a MAP kinase interacting serine threonine protein kinase (Mnk1). This gene is a transcriptional and translational regulator of mRNA, in particular acting via the phosphorylation of the elongation initiation factor (EIF4E), which is an important modulator of cell growth and proliferation 57. Studies in Aplysia have shown Mnk1 to be a negative regulator of cap-dependant translation in neurons 58, whilst in other species it has also been shown to bind stress activated p38 and may play a role in response to environmental stress 59. The role of this gene in cell growth links with the identification of the B cell translocation gene (also involved in cell differentiation) and the Y-box factor homologue (a transcriptional and translational regulator of mRNA) 60, indicating that the mantle is an area of continual growth.

From the above, the mantle is clearly a metabolically and transcriptionally active tissue. This is further exemplified by the presence of ATP synthases, an ADP/ATP translocase, NADH-ubiquinone oxidase, genes from the glycolysis pathway, ribosomal RNAs and arginine kinase. The latter is a phosphagen kinase and these enzymes are prevalent in systems with fluctuating energy demands, acting as an energy buffering system 61 and also as an energy shuttle delivering ATP generated by mitochondria to high energy requiring processes, such as membrane turnover and potentially shell deposition 62. Whilst the phosphagen kinases are a multigene family, arginine kinase is the only form of this gene in arthropods and molluscs 63. It also has other functions such as buffering intracellular pH which would be important in the extrapallial space with the supersaturation of shell matrix components, including calcium ions.

The secretory nature of the mantle tissue requires a number of membrane transport proteins, represented in our limited identifications by a component of a voltage gated potassium channel complex, a V-type ATP synthase, which may transport solutes and lower pH in organelles, prosaposin and the endoplasmic reticulum transport protein SEC61 α sub-unit. The latter protein has been functionally studied in yeast and shown to play a crucial role in translocation of secretory polypeptides across the endoplasmic reticulum membrane 64. Protein alignments of SEC61 α sub-unit from cold (Polar) and temperate fish species identified a number of putative cold adaptive amino acid substitutions 64. The L. elliptica data contained the full length sequence of this gene and alignment with other species, specifically the fish forms 64 showed that L. elliptica does not have the proposed teleost cold water-specific amino acid modifications at positions 327, 328 and 339 in the loop between transmembrane regions 7 and 8 (Figure 3). One hypothesis, at the time, was that these changes were not adaptive, but inherited from a common fish ancestor and our data would appear to substantiate the latter hypothesis. Indeed over the stretch of 120 amino acids shown (Figure 3), the addition of L. ellipitica, the temperate bivalve L. gigantea and three insects to the fish alignment 64 indicates that within this stretch alone, there are 11 putative invertebrate-specific substitutions, 3 substitutions specific to the insects and one restricted to the 2 mollusc species. The implications of these changes cannot be quantified without functional studies, however, our Antarctic invertebrate dataset provides a significant resource for further investigation of gene and protein evolution in cold adapted metazoan species.

Animals living in constant cold temperatures could initially be thought to be more vulnerable to damage by reactive oxygen species, due to slow cell and protein turnover rates and the consequent accumulation of oxidised proteins 1265. However, L. elliptica is being used as a model for ageing studies as this organism appears to have uncoupled the ageing process and antioxidant production. It has a higher antioxidant capacity compared to shorter lived temperate clams, with a constant requirement for stable antioxidant status until old age 1112. Enzyme assays show that catalase, in particular, remains at a constant level throughout its lifetime and this is exemplified by the expression of catalase in the genome of adults of this species. The stress associated endoplasmic reticulum protein (SERP2) is also highly expressed and this may be linked to antioxidant capacity or it may protect unfolded target proteins against degradation and facilitate correct glycosylation. As a member of the RAMP4 family (ribosome associated membrane proteins) it may also be involved in the stabilisation of membranes in response to stress. It is known that there are problems folding proteins at low temperatures 66 and to date, of the few Antarctic marine species investigated, the majority do not exhibit the classical heat shock protein (HSP) stress response. Indeed, several species express the inducible form of HSP70 permanently, possibly as a measure towards more efficient folding of proteins at low temperatures 21 and the expression of SERP2 may contribute towards this "extra" required function and form part of a "preparative defence" strategy against the cold 67.

The final category of highly expressed genes comprises those involved in immune function, e.g. thymosin β 68, α macroglobulin and a mannan-binding lectin associated serine protease, which is a complement control module. The reason for this up-regulation may be slightly more complex than it initially appears. The mantle edge is in constant contact with the external environment and hence there will be permanent challenges to the immune system. This is compounded in the Rothera population of L. elliptica by a significant amount of physical damage (Harper, pers comm). These almost certainly will require the enhanced expression of immune-related genes, as the external protection of the shell is compromised, along with matrix deposition for shell repair. However, physical damage is a stressor in its own right which along with the effect of the cold environment (another potential stressor) may induce changes in the immune system as stress and the immune system have been found in many species to be inextricably linked 69.

The formation of the skeleton in animals is well conserved and frequently involves calcification of a macromolecular network of proteins, lipids and polysaccharides. In molluscs the mantle is the source of matrix proteins and other secreted factors which promote the extracellular assembly of the shell. Relatively few matrix proteins contributing to the shell in molluscs have been identified and most of the studies so far have focused on single proteins such as Asprich, lustrin A, perlustrin and calconectin, whilst other proteins involved in calcium deposition include carbonic anhydrase 36383940707172. In a recent study, 331 randomly selected clones from a cDNA library of the juvenile mantle of tropical abalone (Haliotis asinina, Linnaeus) were sequenced 73. The authors reported that 26% of the genes encoded secreted proteins and of the 106 unigenes identified 15 were involved in trafficking and mineral binding, mechanisms which they suggested probably contribute to construction of the shell. In the present study a conservative estimate using the GO cellular component annotation of known genes suggests 40% of the transcripts are likely to be secreted proteins. A comparison of the transcriptome of the mantle from adult L. elliptica with the cDNA isolated from juvenile tropical abalone mantle 73 revealed relatively poor conservation, with only 31 of the Haliotis sequences sharing significant sequence similarity with the Laternula transcripts. This may be due to either the disparity in sample sizes or maturity stage of the animals, rather than evolutionary distance, as BLAST sequence similarity searching of all 6778 Haliotis asinina sequences in GenBank produced a higher match with 728 Laternula contigs matching 1435 Haliotis sequences (21%). Indeed there were relatively few matches to ESTs from libraries generated specifically to study nacre building gene sets in Haliotis asinina and the bivalve Pinctada maxima (6,122 and 6,737 ESTs respectively) indicating the divergence in biomineralisation processes between these two different molluscs 56. This was further highlighted in the Haliotis/Pinctada study, where there was very little overlap between even the most highly expressed genes and addition of the results from the Laternula and M. galloprovincialis datasets substantiate this (Table 2). Hence there is a requirement to understand shell deposition in a variety of molluscs and not just work on a single model species, particularly where there is a requirement to understand environmental effects.

Several of the most highly expressed genes in our dataset are almost certainly involved in shell deposition, including tyrosinase. The periostracum is secreted as a soluble precursor (the periostracin) and this is then cross-linked by o-diphenols and tyrosinase (or phenoloxidases) to form an insoluble periostracum 7475. Tyrosinase can also be involved in pigment formation in the prismatic layer and evidence from the pearl oyster demonstrates several different paralogues of tyrosinase which are involved in these different functions 7677. However, in order to discover genes within our dataset that are likely to play a role in shell deposition and calcium regulation, we searched the literature to generate an in-house database of proteins involved in extracellular matrix (ECM) formation and calcium homeostasis in metazoans (Supplemental Tables 1 and 2). Numerous transcripts were identified; hence the following section will give only a brief outline of the putative role of the more abundant transcripts.

The presence of putative transcripts for carbonic anhydrase in L. elliptica mantle is unsurprising as this protein was first identified in the shell in 1948 78 and it has subsequently been implicated in matrix mineralisation by generating an acidic environment through the conversion of respiratory CO₂ into HCO₃ in the presence of water 38. Putative transcripts for the matricellular glycoprotein, secreted protein acidic rich in cystein (SPARC, a basal membrane component) were also identified. This trimodular protein promotes proper assembly and maturation of the matrix scaffold and is highly conserved in animal phyla 79. In vertebrates the latter is achieved in part through the interaction of SPARC with fibril forming collagens (I, II, III and V) 8081 and although it is necessary to conduct further work to better characterize these transcripts, orthologues of collagen I, II and V were identified.

Additional transcripts identified in the L. elliptica mantle transcriptome potentially implicated in ECM formation/turnover in metazoans include the thrombospondins, which are a family of large, secreted, multi-modular, calcium-binding glycoproteins which appear to interact with collagens and integrins and have been implicated in skeletal disorders in mammals 82. Transcripts for tenascin, a large glycoprotein containing several fibronectin III type repeats and implicated in cell adhesion in chordates was identified in L. Elliptica 83. Interestingly, several transcripts for metalloproteinase 1 (collagenase 1) which are important in extracellular matrix turnover were also identified. It is apparent even from this brief consideration that numerous homologues of genes identified in the ECM of the vertebrate skeleton are also present in the mantle transcriptome. Future work will permit a more precise characterization of the localization and function of these mantle transcripts and hence provide a better understanding of shell formation. This will be essential for ecophysiological studies.

The L. elliptica contig11573 (283 bp) and contig14182 (252 bp) nucleotide reads share the highest sequence similarity for the N-terminal region and TM2 to TM3 of the metazoan parathyroid hormone receptor (PTHR) and calcitonin/calcitonin Gene-related peptide (CLR/CGRPR) receptors, repectively (Table 3). In vertebrates these receptors are important mediators of the action of the calcitropic factors, calcitonin (CT) and parathyroid hormone (PTH) which stimulate respectively, calcium uptake and bone formation and calcium release for serum and bone turnover. In invertebrates, putative protostome CLR/CGRPR transcripts that remain to be functionally characterised have previously been identified 4784. Several scaffolds were identified in the Lottia genome assembly which shared high sequence similarity (e^-18 - e^-13) to the L. elliptica contigs similar to PTHR/CLR. In molluscs, a CLR/CGRP, expressed in the mantle of the eastern oyster Crassostrea virginica (JC8022) 85 has been isolated and found to be functionally conserved with the vertebrate orthologues. Despite the recent identification of a prototype of vertebrate CT/CGRP peptide ligand in the ascidian Ciona intestinalis 86 no transcripts which are orthologues of CT or PTH have been identified in Laternula and the functional significance of the Antarctic bivalve receptors remains to be established.

Despite the relatively short sequence of the family 2 B1 G-protein coupled receptor transcripts in L. elliptica it was possible to identify conserved amino acid motifs implicated in receptor conformation and ligand affinity in metazoan orthologues (Figures 4 and 5) 84. The characteristic N-terminal motif for interacting with receptor activity-modifying proteins (RAMPs) was identified in contig11573 87. RAMPs are known to modulate GPCRs and in vertebrates they define the specificity of the CLR by modifying its affinity for the ligand. Putative RAMP transcripts were not identified in the present study and it remains to be established if the bivalve family 2 B1 GPCRs are modulated in a similar way to those in vertebrates.

All animals used in experimental work were collected at Rothera Research Station, Adelaide Island, Antarctic Peninsula (67° 4' 07" S, 68° 07' 30" W) by SCUBA divers during the austral summer at depths of 10-15 m. The animals were immediately returned to the laboratory where they were maintained in a through-flow aquarium with a temperature of 0.6 ± 0.3°C, under a simulated natural light:dark cycle. All animals were mature adults, with a range of shell sizes between 50.1-83.5 mm. As shell length is related to animal age: surface aging estimates using growth rings produced an mean age of 8.3 years (SE mean 0.207) with a range from 6-14 years and a median of 8 years (S. Morley pers comm). Mantle tissue was dissected from the animals and cross sections comprising all 3 folds and the periostracum were immediately flash frozen in liquid nitrogen for later RNA extraction.

Mantle RNA was extracted from 24 animals using a modified TRI reagent protocol. After homogenization in Tri Reagent (Sigma) and chloroform extraction, the samples were subjected to a lithium chloride precipitation step. RNA was precipitated using a 1:1 isopropanol:saline solution (0.8 M sodium citrate and 1.2 M NaCl) and after resuspension, the RNA was subjected to a further precipitation using 250 μl 7.5 M LiCl. The extractions were further cleaned using RNeasy mini kit columns (Qiagen, Crawley, Sussex, UK) following manufacturer instructions in order to eliminate LiCl and salt residues. 5 μg of RNA was PCR amplified using the protocol described in 88 prior to preparation for the 454 run. Samples were nebulised at 30psi for one minute and subsequently purified with Ampure (Agencourt) to produce fragments 300 bp and above. The ends were polished and the 454 titananium adapters containing specific MID sequences were attached. Fragments containing both a and b adapters were selected and quantified. Libraries were amplified by emulsion PCR, beads recovered and enriched and placed on a picotiter plate for sequencing by the 454 procedure.

The raw data comprised 1,034,155 reads. Crossmatch (P. Green, unpublished) was then applied to screen for adaptor sequences and other artifacts of the pyrosequencing procedure and also vector sequences using the UniVec database http://www.ncbi.nlm.nih.gov/VecScreen/UniVec.html. Stripping the masked sequence from the ends and removing reads with masked sequence in the middle resulted in 778,629 sequences that were entered into the Newbler program 51 for assembly. This resulted in 18,290 contigs. All singletons were discarded. Files containing the reads have been submitted to the National Center for Biotechnology Information Short Read Archive (accession number SRA011054). The mapping facility of Newbler was applied to the assembly to determine the number of SNPs, and Phobos 89 was used for microsatellite discovery. The contigs were then searched for sequence similarity using BLAST 90 against the genbank non-redundant database 53 and unannotated data from other bivalve species: the gastropod snail: Lottia gigantea http://genome.jgi-psf.org/Lotgi1/Lotgi1.home.html and the Mytilus 454 mantle-specific datasets (4442949.3: M. galloprovincialis mantle unassembled and 4442954.3 M. edulis mantle unassembled) 54 lodged under the MG-RAST database: Meta Genome Rapid Annotation using Sub-system Technology http://metagenomics.nmpdr.org/) 91. The Gene Ontology (GO) 92 mappings were determined by an in-house database on all Swissprot and Trembl 93 BLAST scores below a threshold of 1e-10. Sequence manipulation was carried out using the EMBOSS suite of programmes 94. Sequences were clustered using ClustalW 95 and the alignments displayed using BoxShade v3.21 96.