Proteomic evidence of a paedomorphic evolutionary process within a marine snail species: a strategy for adapting to extreme ecological conditions?


Correspondence: Angel P. Diz, Department of Biochemistry, Genetics and Immunology, Faculty of Biology, University of Vigo, 36310, Vigo, Spain. Tel.: +34 986813828; fax: +34 986813828; e-mail:


The exposed and sheltered ecotypes of the marine snail Littorina saxatilis from European rocky shores are considered a key model system to study adaptation and ecological speciation. Previous studies showed that two ecotypes (RB and SU) of this species in NW Spain have adapted differently to different shore levels and microhabitats. In order to understand how this divergent adaptive process has been accomplished, we followed a quantitative proteomic approach to investigate the proteome variation in a number of different biological factors, that is, ecotype, ontogeny and their interactions. This approach allowed testing the hypothesis that one of the ecotypes has evolved by paedomorphosis, and also whether or not the molecular mechanisms related to ecotype differentiation are set up in early developmental stages. Additionally, the identification of some candidate proteins using mass spectrometry provides some functional insights into these evolutionary processes. Results from this study provided evidence of higher ontogenetic differentiation at proteome level in the RB (metamorphic) than in SU (paedomorphic) ecotype that point to the possibility of juvenile stage retention in this latter ecotype. The level of protein expression (proteome) differences between ecotypes maintained nearly constant from late embryonic stages to adulthood, although some proteins involved in these changes considerably differed in embryonic compared to other ontogenetic stages. Paedomorphosis may be the evolutionary response of the SU ecotype of solving the trade-off during sexually immaturity that is caused by the evolution of small size arising from adaptation to the wave-exposed habitat. Some potential candidate genes of adaptation related to energetic metabolism have been identified, providing a promising baseline for future functional analyses.


When investigating the causes of speciation, and the origin of varieties, authors have usually emphasized the importance of knowing the genetic architecture of key traits contributing to reproductive isolation, the influence of historical and geographical scenarios or the evolutionary forces that model trait evolution when adapting to new environments or niches (Coyne & Orr, 2004; Rundle & Nosil, 2005; Schluter & Conte, 2009; Samdja & Butlin, 2011). However, speciation studies have hardly focused on the strategies of developmental control and timing of traits across ontogeny, even though it has been long accepted that the key role of these sorts of changes is to produce trait innovations, such as those occurring during several radiation processes (Minelli & Fusco, 2012). A model case to investigate the relationship between developmental processes and speciation is the existence of species that have evolved by paedomorphosis in response to new habitats, that is, retaining juvenile features for adult stages for some traits during evolution (Futuyma, 2009). A well-known example of this phenomenon is the retention of branchia (present in premetamorphic specimens) in some species of salamanders that have undergone adaptation to novel aquatic environments (Denoël et al., 2005). Paedomorphic evolution may be produced either by reducing growth rate of somatic tissues (neoteny) or by accelerating the maturation of sexual tissues (progenesis), although these two processes can be difficult to distinguish in practice. A pattern of evolution by paedomorphosis in several snake species has been recently suggested (Calvete et al., 2010 and references therein). In this study, they reported proteome variation in the venom composition across different ontogenetic stages (from newborn to adulthood) in a Central American snake subspecies (Crotalus spp.). Strikingly, the protein composition of venom from South American adult snake subspecies was less variable and more similar to what was found in the Central American newborn (juvenile) snake subspecies, suggesting a pattern of evolution by paedomorphosis. The authors discussed this finding in terms of population fitness, response to selective pressures and adaptive mechanisms, and even extended this observation to other snake species (Calvete et al., 2011). These latter studies represent a good example on how proteomic approaches can be used to test evolutionary questions, following a qualitative approach focusing on protein composition of venom for hypothesis testing, rather than on expression differences for particular proteins at different ontogenetic stages. Nevertheless, it would be also feasible to follow a quantitative approach to test the same hypothesis as presented here, that is, by comparing the protein expression levels at different ontogenetic stages (the paedomorphic taxa should show a lower level of differentiation than the metamorphic ones). Focusing on the proteome level may have some intrinsic advantages when undertaking evolutionary studies on the grounds that proteins are closer to the phenotypic level – the direct target of natural section – than genomic or transcriptomic levels (Khaitovich et al., 2004; Diz et al., 2012). Therefore, a proteomic approach might be more pertinent in order to infer the molecular basis of an adaptive process, something that is still poorly known (Bierne et al., 2011).

The striking polymorphism in the marine intertidal snail Littorina saxatilis (Olivi, 1792) from the Galician coasts (NW Spain) has been considered an interesting case in which to study adaptation and incipient ecological speciation (Rolán-Alvarez, 2007; Johannesson et al., 2010; Stapley et al., 2010). Two ecotypes, ridged and banded (RB) and smooth and unbanded (SU), are adapted to different shore levels (upper and lower shore, respectively) and microhabitats (barnacle and mussel belt, respectively). These ecotypes differ in a considerable number of phenotypic and genetic traits due to their differential survivorship at different ecological conditions (reviewed in Rolán-Alvarez, 2007). For example, the upper RB ecotype resists higher osmotic, heat and predation risks, whereas the lower SU ecotype resists higher risk of dislodgment caused by the strength of wave action characteristic to its micro-habitat. The larger RB ecotype resembles populations from other micro-areas in NW Spain (e.g. within estuary populations), whereas the smaller SU ecotype represents a specific adaptation to an extreme micro-habitat, and it is exclusively found in the most wave-exposed granite rocky shores (Rolán-Alvarez, 2007). These two ecotypes meet and mate at the mid-shore, where both microhabitats overlap, producing fertile hybrids, and it was actually observed that both ecotypes apparently freely interbreed under laboratory controlled conditions (Rolán-Alvarez et al., 2004; Saura et al., 2011). Interestingly, they have already developed a partial prezygotic reproductive isolation in the wild as a side effect of differential adaptation, due to differences in their micro-habitat choice and mating (size based) preferences (Rolán-Alvarez, 2007; Conde-Padín et al., 2008b).

These ecotypes differ in a number of morphological, physiological or even behavioural traits (Rolán-Alvarez, 2007). They also differ in about 3% of the genome studied by AFLPs (Galindo et al., 2009) and 7–16% of the proteome studied by two-dimensional electrophoresis (2-DE; Martínez-Fernández et al., 2008, 2010). Moreover, ecotype phenotypic differentiation at both shell shape and proteome levels showed a strong genetic determination; that is, about 70% of the ecotype differences found at these two levels were able to be attributed to a genetic origin after following an experimental design specifically designed to answer this question (Conde-Padín et al., 2007, 2009; Martínez-Fernández et al., 2010). However, it is unknown how these differences in proteomic expression vary across ontogeny in each ecotype and such information could be useful in two complementary contexts. First, we have observed an apparent retention of the juvenile stage in the SU compared to RB shell shape according to a quick inspection of shells (Fig. 1). This might suggest a paedomorphic evolutionary pattern in the SU ecotype, although this hypothesis has never been formulated or tested in this organism. From another point of view, the relative importance of regulatory vs. structural genes when contributing to ontogenetic and adaptive changes has also been discussed (Bertossa, 2011). For example, one hypothesis would be that a few regulatory genes are activated early in the ontogeny of an organism and that this would trigger a cascade of secondary effects at the molecular level arising from the regulation of expression levels of many other genes during late developmental stages of this organism. Therefore, this proteomic study could be useful in simultaneously checking how ecotype differences at proteome level vary across ontogeny.

Figure 1.

Experimental design and pictures from Littorina saxatilis taken from the two ecotypes (RB and SU) and three ontogenetic stages (adults, juveniles and embryos) analysed in this study. Pooled samples were made with 10 individuals each, for adults and juveniles. For embryo samples, RB and SU females were firstly isolated and then all embryos extracted from the brood pouch. Ten different female individuals were used to create the pooled embryo sample for each ecotype. Three pooled samples (biological replicates) were prepared for each group studied (see text). Therefore, a total of 24 biological replicates were analysed, that is, 12 for adults (totalling three for each ecotype and sex), six for juveniles and six for embryos (three for each ecotype). To keep a balanced design, the anova analysis was repeated twice either using male or female samples in adults for each protein spot.

Here, we present a proteomic study in the marine snail L. saxatilis comparing protein expression levels across different ontogenetic stages (embryos, juveniles and adults) and ecotypes (RB and SU) to address the following evolutionary questions: (i) Whether or not a pattern of adaptive evolution by paedomorphosis that was preliminarily inferred from morphological observations can be also supported by proteomic data. If such a paedomorphic pattern exists, an expected outcome in this new analysis would be that proteomic differentiation across ontogeny would be significantly larger in RB than in SU ecotypes. (ii) To test whether or not the proteome differences previously observed in this marine snail are set up in early developmental stages and remain similar in the adulthood. Finally, candidate proteins identified by mass spectrometry (MS) will aid in shedding some light on molecular mechanisms involved in the ecotype adaptation of this marine snail.

Materials and methods

Sampling and experimental design

Littorina saxatilis individuals from the two different ecotypes (RB and SU) were collected in May 2009 from an intertidal rocky shore area in Silleiro Cape, NW Spain (42°06′15″N; 8°53′56″O). Individual samples from RB and SU ecotypes were collected in the same journey at the upper and lower rocky shore, respectively, and rapidly transported to ECIMAT, a marine science laboratory, to keep them in aquarium tanks during 10 days under controlled conditions, including the use of fresh prefiltered seawater from an open circuit. Briefly, the seawater was at 14.2 °C, with a salinity of 3.63% and an oxygen level of 7.6 mg L−1. The system was also provided with 14/10 h photoperiod (daylight/darkness) supplied through fluorescent lighting. Further technical details of the breeding system in this species are given in previous work (Conde-Padín et al., 2007, 2009). This design ensures that all analysed individuals shared the same environmental conditions during 10 days, aiming to minimize the environmental over genetic effects on the protein expression levels. After this period, samples were immediately snapped frozen in liquid nitrogen, labelled and stored at −80 °C prior to further analysis.

A pooling sample strategy was followed after protein extraction, an appropriate strategy in terms of reducing the biological variation and increasing the statistical power (Kendziorski et al., 2005; Diz et al., 2009a). This has been successfully applied in two previous studies in the same marine organism (Martínez-Fernández et al., 2008, 2010). Pooled samples were made with 10 individuals each, for adults and juveniles. Three pooled samples (biological replicates) were prepared for each group studied (Fig. 1): Adult-Male RB, Adult-Female RB, Adult-Male SU, Adult-Female SU, Juvenile RB, Juvenile SU, Embryo RB and Embryo SU. Therefore, we analysed 24 biological replicates in total, that is, 12 for adults (being three for each ecotype and sex), six for juveniles and six for embryos (half for each ecotype) (Fig. 1). In L. saxatilis, a direct developer, the females carry the embryos in a brood pouch during their development and then release small crawl ways. For embryo samples, we firstly isolated RB and SU females and then extracted from each female all the embryos contained in the brood pouch. Ten different female individuals were used to create the embryo-pooled sample for each ecotype.

Protein extraction

Proteins were extracted from whole individuals and solubilized in lysis buffer (7 m urea, 2 m thiourea, 4% CHAPS, 1% DTT and 1% IPG), following approximately a constant ratio of 30 mg of tissue in 1 mL of lysis buffer, with a sonicator (Branson Digital Sonicator 250; Danbury, CT, USA) using five blasts of 30% amplitude and 10 s each, with 30-s breaks, on ice to avoid protein burning. After centrifugation for 15 min at 21 000 g, 10 °C, the pellet was discarded and protein supernatant stored at −80 °C until electrophoresis. Protein concentration was measured with a modification of Bradford method (Ramagli & Rodriguez, 1985).

Two-dimensional electrophoresis

Two-dimensional electrophoresis (2-DE) coupled with mass spectrometry (MS) is one of the available and commonly used methods in quantitative proteomics, where protein expression profiles of different samples from an organism, tissue or cell at a specific time and condition are quantitatively analysed and proteins of interests identified. It is important to note that this type of quantitative analysis can be regarded as relative rather than as absolute (reviewed in Diz et al., 2012). Proteins from all 24 biological replicates and six technical replicates (adult samples from both ecotypes) were analysed using 2-DE in batches of six gels/samples per run following a randomizing block design to control the well-described run-to-run variation in 2-DE (see Diz & Skibinski, 2007). Biological replicates allow accounting for biological variation, which is intrinsic to all natural organisms, whereas technical replicates, the same sample run in different days, allow for assessing the experimental error. Approximately 75 μg (analytical gels, for quantitative analyses) or 150 μg (preparative gels, for protein identification by mass spectrometry) of total protein was used for each gel. The first dimension, where proteins are separated according to their isoelectric point through isoelectric focusing (IEF), was carried out on immobilized pH gradient strips (pH 5–8/17 cm, Bio-Rad Hercules, CA, USA) with a horizontal electrophoresis apparatus Protean IF System (Bio-Rad) following manufacturer's instructions, including an active strip rehydration step. After two steps of strip equilibration (2 × 15′) with DTT and iodoacetamide respectively, the second dimension of gel electrophoresis was carried out with laboratory-cast 12.5% polyacrylamide gels (22 × 27 × 0.1 cm3) using an EttanDaltsix electrophoresis system (GE Healthcare, Little Chalfont, UK) that allows loading a batch of six gels/samples per run. Electrophoresis was carried out at 20 °C, at constant current of 12 W gel−1 for approximately 5 h until the bromophenol blue front reached the bottom of the gel. Protein spots were visualized by a slightly modified silver staining method fully compatible with MS analysis (see Shevchenko et al., 1996). Co-migrating broad-range standards (Bio-Rad) were used in the second dimension to allow estimation of molecular mass.

Analysis of 2-DE patterns

Silver-stained gels were scanned using an Image Scanner (Bio-Rad) and saved as TIFF files. The SameSpots v3.3 software (Nonlinear Dynamics Ltd, Newcastle upon Tyne, UK) was used for semi-automatic alignment of gels, protein spot detection and volume measurements. A final step of spot filtering, incorporating visual checking, was carried out. Spots resulting from artefacts or showing anomalies were discarded and manually edited where this was required. The number of detected spots per gel that were finally chosen for further statistical analysis strongly relied on the stringency level applied in the different protein spot filtering steps. We followed a quite stringent criterion that included a very exhaustive checking across all individual ‘spots’ automatically detected by the software at first instance. Up to 70% of these initial spot detections were discarded for statistical analysis. Absolute spot volumes provided by 2-DE software were normalized for each gel and then transformed to logarithmic scale (see Diz & Skibinski, 2007). It is important to note that despite the lower linear range (but higher sensitivity) of silver staining for protein quantification (up two orders of magnitude) compared to other stains (see Miller et al., 2006), our final data set did not include absolute spot volumes differing more than two orders of magnitude. Moreover, those saturated spots (doughnut effect) due to silver staining were removed from final data set and statistical analyses.

Statistical analyses were carried over the normalized protein spot volumes. Pearson's correlation coefficient (r) and coefficient of variation (CV) calculations were carried out using the whole protein spot data set from technical replicates as explained in Diz & Skibinski, 2007; aiming to assess the experimental reproducibility. Analyses of variance (anova) using the volume of each protein spot (dependent variable) were carried out to test for differences in protein expression patterns due to different factors and interactions. The variables were verified for homoscedasticity and normal distribution of residuals, and deviations from the model assumptions described accordingly. These anovas were, however, applied on different subgroups of biological replicates because the treatments were not orthogonal in all cases (juveniles or embryos cannot be separated into males and females). The analysis included three biological replicates from all three ontogenetic (embryos, juveniles and adults) and ecotype (RB and SU) classes (18 biological replicates in total), therefore including the fixed factors Ecotype and Ontogeny by a two-way anova (Fig. 1). This analysis was repeated twice using either male or female adults to fit to a balanced anova design (the exploratory study investigating the factors Sex and Ecotype in adults is presented elsewhere). All these analyses were accomplished on each spot separately. An important clarification is the fact that in these types of studies what is quantitatively compared (relative quantification; see more details in Diz et al., 2012) are the volumes of each protein spot across different 2-DE gels/samples and not the volumes from different spots within the same 2-DE gel/sample. All P-values from a priori tests obtained from the above analyses were corrected to account for the multiple hypothesis testing problems (reviewed in Diz et al., 2011). We applied the SGoF multitest adjustment at both 5% and 1% significance levels (Carvajal-Rodríguez et al., 2009) and complementarily calculated the false discovery rate (q-value; Storey, 2003) using the software SGoF+ v7.0 (Carvajal-Rodríguez & de Uña-Alvarez, 2011). In addition, the level ecotype co-expression across different ontogenetic stages was investigated. To undertake this analysis, the fold change of expression level was first calculated for every protein spot (n = 393) between ecotypes (RB/SU) at different ontogenetic stages. The Pearson correlation coefficient was calculated for each pair-comparison of the different ontogenetic stages (adult vs. juvenile; adult vs. embryo; juvenile vs. embryo) using fold change data in order to assess the level of similarity of ecotype differences across ontogeny.

All former statistical analyses were carried out in spss for Windows version 14.0. The comparison of the frequency of significant spots across anova factors was accomplished by homogeneity maximum likelihood G test using the poptools application (Hood, 2008). This latter analysis allowed checking the distribution of significant cases (in the 393 spots) across factors using a nonparametric and statistically robust approach (Sokal & Rohlf, 1995).

Testing the hypothesis of paedomorphism

In addition to the former analyses, the existence of paedomorphosis in this marine species was specifically tested with proteomic data obtained in this study. Under this hypothesis, one would expect to find a significantly lower protein differentiation across ontogeny in SU (paedomorphic) than RB (metamorphic) ecotype (hereafter ‘test’ analysis). Moreover, it is feasible with the same proteomic data to carry out an analysis that serves as a ‘control’ of the former relationship (hereafter ‘control’ analysis). This ‘control’ analysis is based on the assumption that under the paedomorphic hypothesis one would expect that the paedomorphic ecotype (SU) would have retained a juvenile stage in its adult stage; hence, protein expression patterns would more closely resemble juvenile stages rather than the adult stages of the RB ecotype (metamorphic). Therefore, in the ‘control’ analysis similar results in terms of ontogenetic differentiation by re-analysing proteomic data between ‘embryos’ and ‘juveniles’ in the RB ecotype and between ‘embryos’ and ‘adults’ in the SU ecotype would be expected. One-way anova analysis for each protein spot was accomplished to detect significant differences across ontogeny within each ecotype. The coefficient η2 was additionally calculated in the former anova to quantify the contribution of ontogeny to protein variation (Pierce et al., 2004). Furthermore, we accomplished the former ‘test’ (all ontogenetic stages between ecotypes) and ‘control’ (embryos and juveniles for RB vs. embryos and adults for SU ecotype) analyses by two complementary approaches. Firstly, a qualitative approach was followed, where the frequency of significant protein spots differing across ontogeny between ecotypes was investigated, and therefore, the comparison between ecotypes (within both ‘test’ and ‘control’ analyses) was made by homogeneity G test (see above). Secondly, a quantitative approach was followed, where the mean η2 (representing the percentage of variance explained across ontogeny) between ecotypes (both within test and control analyses) was compared using a one-way anova.

Protein identifications and mass spectrometry analysis

The protein spots of interest were excised using the automatic spot picker Proteineer spII (Bruker Daltonics, Billerica, MA, USA) into a microplate with mQ H2O. Trypsin digestion was carried out using the automatic digester Proteineer dp (Bruker Daltonics), using a DP Chemical 96 kit (Bruker Daltonics) following the manufacturer's instructions. Briefly, the excised spots were distained using 15 mm potassium ferricyanide/50 mm sodium thiosulphate, and the proteins were reduced by 10 mm DTT (dithiothreitol) for 45 min and alkylated by 55 mm IAA (iodoacetamide) for 30 min. The trypsin digestion was performed by adding 8 μL of trypsin (20 ng μL−1) to each spot and incubating at 37 °C for 4 h. Tryptic peptides were analysed using matrix-assisted laser desorption/ionization-time of flight tandem mass spectrometry (MALDI-TOF/TOF MS) with an Autoflex III smartbeam (Bruker Daltonics). MALDI sample preparation was automatically carried out by a Proteineer dp device (Bruker Daltonics), applying 3 μL of digests onto a disposable Anchorchip, MTP-sized MALDI targets, prespotted with HCCA (α-cyano-4-hydroxycinnamic acid) matrix for 96 samples and standards (1–4 kDa) for 24 close external calibration points. The data were acquired and analysed automatically using the flexControl and flexAnalysis v3.0 software (Bruker Daltonics), respectively. The major peaks obtained by MALDI-TOF to be further characterized by TOF/TOF analyses were selected. Data (MS and MS/MS spectra) were generated in PKL file format and were submitted for protein identification through database search against SwissProt 56.6 (405506 sequences; 146166984 residues) or NCBInr 20070216 (4626804 sequences; 1596079197 residues) databases using BioTools v3.0 software (Bruker Daltonics) and version 2.2.04 of MASCOT search engine (Matrix Sciences Ltd, London, UK). Search parameters were set as follows: taxonomy = all entries/metazoa/other metazoa; enzyme = trypsin; missed cleavage = one; fixed modification = carbamidomethyl of cysteine; variable modification = oxidation of methionine; mass values = monoisotopic; mass tolerance for precursor ions = 100 ppm; mass tolerance for fragment ions = 0.5 Da; protein mass = unrestricted. Protein identification was considered positive when the MASCOT score for a specific protein match was statistically significant (< 0.05). The whole process was repeated several times by excising each protein spot from different 2-DE gels (both biological and technical replicates) to increase the confidence of protein identifications. Additionally, the procedure of protein spot identification was repeated again following a LC-MS/MS approach. Briefly, peptides were eluted using a linear gradient starting at 98% A (0.1% formic acid in water) and ending at 50% B (0.1% formic acid in acetonitrile) in 60 min. Mass analyses were carried out in an Apex-Qe7T FT-ICR (Bruker Daltonics), a mass analyser with a higher resolution in an attempt to overcome the limited success in the number of protein identifications obtained with the previous MS method. This latter approach provided a few extra high-quality spectra which allowed some protein identifications following a de novo sequencing approach (see Steen & Mann, 2004).


2-DE and protein co-expression patterns across ontogeny

Results from the reproducibility experiment, using protein data set from technical replicates, showed, on average, a coefficient of variation (CV) = 19.9 ± 2.9 and a Pearson's correlation coefficient (r) = 0.94 ± 0.019, these values being in the range of those reported in similar studies (see Diz & Skibinski, 2007). anova residuals were typically normally distributed but heteroscedasticity affected an average of 15% of the studied protein spots. Results from the two-way anova analysis (< 0.05) for each spot (n = 393) on samples from adults, juveniles and embryos are summarized in Table 1 (Supporting information, Table S1 for results using a < 0.01). This analysis allowed investigating the protein expression pattern in Ecotype and Ontogeny factors (and corresponding Interaction). The analysis was performed independently for male and female adults. In both cases, similar results were obtained, that is, a significantly higher percentage of protein spots differing significantly in their expression levels across the different ontogenetic levels (Ontogeny factor) than those observed between ecotypes (Ecotype factor) or even for the interaction of both factors (Interaction term). Note that Ecotype differentiation reached up to 30% of the protein spots studied in this analysis. Apparently, however, Ecotype differentiation in protein expression did not change dramatically across the ontogeny (Fig. 2; Fig. S1 shows 2-DE gel examples across ontogeny), suggesting that proteome differences between ecotypes retain similarity along the different ontogenetic stages. However, the correlation between fold changes in the protein expression levels between ecotypes (RB/SU) rendered somewhat different results as the correlation was substantially higher between male adult and juveniles (= 0.64, N = 393; < 0.0001) than between male adult and embryos (= 0.38, N = 393; < 0.0001) or juvenile and embryos (= 0.36, N = 393; < 0.0001). These results demonstrate that differences in expression between ecotypes involved some different genes in early (embryonic) stages compared to juvenile and adult stages, although the overall rate (percentage) of proteome differentiation is quite similar across ontogeny (Fig. 2).

Figure 2.

Summary results from different anova analyses. On the y-axis is shown the percentage (%) of protein spots showing significant differences in expression between ecotypes for each ontogenetic stage. Notice that adult samples were split in males and females to maintain the same sample size in all anova analyses carried out for each ontogenetic stage (i.e. three biological replicates per ecotype and ontogenetic stage). This thus allows a visual comparison to be made of results obtained for different ontogenetic stages.

Table 1. Percentage of protein spots for each factor (Ecotype, Ontogeny and Interaction) in a two-way anova that showed significant differences at 5% (significance level) after multitest correction by the SGoF method. DF, degrees of freedom per factor in the anova (the error term in the anova had 18 d.f.). A likelihood G test was used to check for homogeneity of frequencies of significant protein spots across factors (and interaction) and also between males and females of adult samples obtained from anova analyses. Notice that in ‘Adults’, ‘Males’ and ‘Females’ columns show the results from the same anova analyses but either using males or females (from adult samples) in order to keep a balanced design across factors. Notice that d.f. from the G test varies between 1 (comparing sexes) and 2 (comparing factors from anova)
Factord.f.AdultsG test
  1. ns, not significant.

  2. a

    < 0.001.

G test 75a64a 

Paedomorphism hypothesis

The results from the ‘test’ and ‘control’ analyses checking for paedomorphosis, following both a qualitative and quantitative approach (see 'Materials and methods'), are presented in Table 2. The percentage of significant protein spots that differed across different ontogenetic stages in the ‘test’ analysis was different in the two ecotypes (on average 35.5% in RB and 27% in SU; Table 2a). From an evolutionary point of view, it might actually suggest that the percentage of proteins significantly different in expression from embryo to adulthood stage could have been reduced in the SU compared to RB ecotype, a pattern which resembles a mechanism of evolution by paedomorphosis (see 'Discussion'). The ‘control’ analysis was specifically carried out in order to confirm this hypothesis as it would be expected that the level (%) of protein expression differences across different ontogenetic stages (Ontogeny factor) should be as large in RB (including only embryo and juvenile samples) as in SU (including only embryo and adult samples), a result that was in fact observed after doing this analysis (RB = 26% and SU = 25%). A further convincing result would be to find the same trend in the quantitative comparisons by using η2 coefficient in order to quantify the degree of variation in the protein expression levels explained by ontogenetic differences (Table 2b). In this analysis, both significant and nonsignificant protein spots were included. The percentage of variance explained across different ontogenetic stages in the ‘test’ analysis was significantly different between ecotypes (on average 53% in RB and 46% in SU), while such differentiation disappeared in the ‘control’ analysis (on average 46% in RB and 45% in SU).

Table 2. ‘Test’ and ‘control’ analyses carried out using protein expression data in order to test the hypothesis of paedomorphism. ‘Test’ analysis compared protein expression levels between ecotypes across all ontogenetic stages, whereas ‘control’ analysis used only data from embryos and juveniles from RB-ecotype samples and embryos and adults from SU-ecotype samples (see 'Materials and methods'). (A) Qualitative analyses. The frequencies of significant protein spots involved in ontogenetic differentiation were compared between ecotypes. (B) Quantitative analysis. The η2 coefficient estimated the magnitude of ontogenetic differentiation in each spot (see M&M). The quantitative analysis compared the mean η2 coefficient between ecotypes. In both analyses n = 393 protein spots were analysed. Notice that adult samples were split in males and females (the above analyses repeated twice) to keep a balanced design across ontogenetic stages
A setTestControl
% Signif. G P % Signif. G P
FemalesRB34.9  26.5a  
  17.80.0001 1.80.1845
SU27.2  22.4  
MalesRB36.1  26.4a  
  7.20.0072 0.30.5759
SU27.2  28.2  
B setTestControl
Mean η2 F P Mean η2 F P
  1. a

    Notice that for this ecotype we compared only embryos and juveniles in the ‘control’ analysis (see text).

FemalesRB53.3  45.9a  
 (0.014)  (0.016)  
  24.30.0001 2.70.1023
SU44.0  42.3  
 (0.021)  (0.021)  
MalesRB53.6  45.9a  
 (0.014)  (0.016)  
  7.50.0065 0.60.4328
SU48.3  47.7  
 (0.021)  (0.021)  

Protein identifications

The protein spots identified by mass spectrometry (MS) are shown in Table 3, and statistical analysis testing for expression changes across factors for these identified proteins are shown in Table S2. Two clearly different groups of protein spots can be identified. One group is based on protein spots that did not show any statistically difference in expression for any of the factors. These proteins seem to be constitutively expressed in this organism, possibly playing a cellular housekeeping role. A second group of protein spots showed statistically significant differences either between ecotypes or across ontogeny (most of these spots were not significant for any interaction effect between factors). The first group of protein spots, even though of limited use in this context, provided direct evidence of expression for some genes in this organism, whereas the second group might represent proteins that are candidates for involvement in the adaptive process of ecotype divergence and/or changes involved during the ontogeny of this species. Actins, ATP synthases, tropomyosins and tubulines are highly conserved proteins, and it is assumed that they could mostly play key (housekeeping) roles in the cellular function. Thus, it is not surprising to find that these proteins are highly abundant in the protein map (Fig. 3). This fact, together with a high homology of these proteins across different organisms, could be one of the explanations for their greater identification success by MS (Diz et al., 2009b). Exceptions to this trend are protein spots 310, 352 and 444 identified as actins, which showed significant changes in their expression levels between ecotypes (this was only found in adults) and across ontogeny. Remarkably, the protein spot 444 was one of the few spots that also showed differences between sexes (data not shown).

Figure 3.

Representative two-dimensional electrophoresis (2-DE) protein map from Littorina saxatilis (RB ecotype from adult sample) showing those protein spots that have been successfully indentified by mass spectrometry (MS). Please see Table 3 for more details about these protein spot identifications. 2-DE is an analytical technique that allows the separation of a complex set of proteins according to two independent properties: the isoelectric point (pI) and molecular weight (MW). Proteins were visualized after silver staining (see text for more details).

Table 3. Summary table for protein spot identifications by mass spectrometry (MS) using peptide mass fingerprinting-PMF (MS), peptide fragmentation (MS/MS) or both methods (see Table S3 for more details). Note that some few protein spots were identified by de novo sequencing analysis from MS/MS spectra combined to BLAST search against nrNCBI database (see more details about peptide stretches obtained after de novo analysis in Table S3). Protein names are the definitions returned by MASCOT database search for protein spot identification through BioTools 3.0 software. Gene Ontology (GO) terms concerning biological function and cellular component for each protein were provided by UniProt. The protein expression ratios (RB/SU) were calculated from proteomic results of samples from all ontogenetic stages. Statistical details are provided in Table S2
Spot codeMSMS/MSNameBiological functionCellular componentRatio RB/SU
  1. ns, not significant; ***< 0.001; **< 0.01.

Act1-8 ActinMuscle contraction (cellular motility)Cytoskeleton
310 Actin  ns
352 Actin  ns
444 Actin  ns
463 Aconitate hydrataseMetabolism (Tricarboxylic acid cycle)Mitochondrion1.44***
510  de novo EnolaseMetabolism (Glycolysis)Cytoplasm1.19**
455 Arginine kinaseMetabolism (ATP maintenance)Cytoplasm2.02***
456 Arginine kinase  ns
212Arginine kinase  2.62***
Syn1  ATP synthase subunit betaMetabolism (ATP production)Mitochondrion
497 ATP synthase subunit alpha  ns
Rep1  DNA repair protein radC homologDNA replication, recombination, and repairNucleus
202  de novo Fructose-bisphosphate aldolaseMetabolism (glycolysis)Cytoplasm1.51***
203  de novo Fructose-bisphosphate aldolase  0.92**
206  de novo Fructose-bisphosphate aldolase  1.27**
515  de novo Strombine/Alanopine dehydrogenaseMetabolism (related to anaerobic glycolysis)Cytoplasmns
381Mitochondrial import receptor subunit TOM34Chaperone/protein transportationMitochondrionns
Rib1  50S ribosomal protein L31 type BRibonucleoprotein/ Ribosomal protein/RNA bindingRibosome
356 40S ribosomal protein S18Ribonucleoprotein/ribosomal protein/RNA bindingRibosomens
Exo1  RNA exonuclease 4  
Tro1-2 TropomyosinMuscle contraction (cellular motility)Cytoskeleton
Tub1 Tubulin alpha/beta chainStructural element (cellular motility/microtubules)Cytoskeleton
Tub2 Tubulin alpha/beta chainStructural element (cellular motility/microtubules)Cytoskeleton

Regarding the second group of proteins with a potential evolutionary interest (e.g. adaptation), we will mainly focus on two enzymes, that is, arginine kinase (AK) and fructose-bisphosphate aldolase (FBA). The identification of two protein spots as AK and FBA and their possible role in the adaptive process of the two divergent ecotypes in L. saxatilis had been already described and discussed in a previous proteomic study (Martínez-Fernández et al., 2008). The present study confirms the same finding but also extends it, by allocating six different protein spots in the protein map which have been identified as AK (spots 455, 456 and 212) and FBA (spots 202, 203 and 206). These new findings show that the functional picture is more complicated than previously thought and one needs to be cautious when providing a functional rationalization related to the biological factor under study (see Discussion and box 1 in Diz et al., 2012). There are also some other protein spots (310, 352, 444, 463 and 510) that showed ecotype or ontogenetic differences in expression (see Table S2). Another interesting result, which can be relevant to undertake future studies in this species, came from the fact that four protein spots were identified by a sequencing de novo approach which provides new and direct information on a few short stretches of peptide sequences from Littorina (amino acids in blue mean that partial amino acid sequence was directly inferred by de novo sequencing approach; see Table S3). These short sequence tags could be, for example, used in further confirmatory studies to elaborate a set of primers to amplify and characterize the genes that codify for these proteins.


One of the most interesting possibilities of proteomic studies is to explore and even provide evidences that support new evolutionary hypotheses (see Diz et al., 2012). This present proteomic study on Littorina saxatilis reports differences in the protein expression patterns across different ontogenetic stages in the two different ecotypes of this marine snail species. This result could be explained by a pattern of evolution by paedomorphosis in the SU ecotype. Evidence supporting this hypothesis came from the observation that the level of proteome differentiation between embryos and adults in SU ecotype resembles the level of differentiation between embryos and juveniles in the RB ecotype, somewhat expected after a rough morphological observation (Fig. 1; Table 2). Paedomorphic patterns of evolution inferred from proteome analysis have been already reported for venoms of different snake species (see Calvete et al., 2010, 2011). Likewise, paedomorphic patterns have also been described in several salamander taxa by using morphological and anatomical comparisons and their evolutionary causes and consequences extensively discussed (Denoël et al., 2005, 2009; Doyle & Whiteman, 2008; Wiens & Hoverman, 2008). Although with some caution due to the differences in life and phylogenetic histories between salamanders and marine snails, we will try to extend the mentioned discussion in order to understand better this evolutionary pattern in the context of L. saxatilis populations.

For example, it has been suggested that paedomorphosis may allow a more efficient use of the available resources in the habitat, allowing paedomorphic forms to adapt to a novel niche in a context of strong intraspecific competition (Denoël et al., 2005). Similarly, it can be hypothesized that the paedomorphic ecotype (SU) of L. saxatilis may represent a specialized form capable of exploiting a new and extreme habitat, that is, the lower rocky shore of highly wave-exposed localities (Rolán-Alvarez, 2007). Differences in size between ecotypes are striking and have been interpreted previously in adaptive terms, with a smaller size in the lower shore being necessary to resist wave-dislodging and larger size in the upper shore may represent a strategy for increasing survivorship under crab predation (Conde-Padín et al., 2007; Rolán-Alvarez, 2007). However, an obvious trade-off for surviving with a smaller size may be to prevent the achievement of minimum size/age for sexual maturation, and it is in such context when one can think of a selective pressure favouring a shift towards paedomorphosis in the SU ecotype, perhaps via progenesis. In fact, at least in salamanders, progenetic paedomorphs are usually smaller than metamorphic adults (Denoël et al., 2005). Another obvious trade-off to evolving preferentially with a smaller size is related to fecundity. In this ovoviviparous organism, female size is significantly correlated with the number of embryos being carried in the brood pouch and therefore in the number of embryos released by time unit (Conde-Padín et al., 2008b). However, natural selection probably favoured the evolution of an increase in the size of new born released snails in the SU compared to RB ecotype to maximize their survivorship when facing the extreme conditions of the lower shore habitat (Conde-Padín et al., 2007).

Paedomorphosis, as a source of intraspecific polymorphism and affecting indirectly different life history characteristics, has been proposed as a mechanism able to favour speciation in certain contexts. For example, in some salamander populations the paedomorphic form breeds several weeks earlier than the metamorphic one, causing a partial sexual isolation (Denoël et al., 2005). In L. saxatilis the partial reproductive isolation observed between ecotypes is caused by two main mechanisms: microhabitat choice and sexual isolation (Conde-Padín et al., 2008a). The former mechanism may have evolved independently from paedomorphism, but sexual isolation is indirectly caused by size differences between ecotypes, and these might also be a consequence of the paedomorphic change in the SU ecotype. Therefore, it is reasonable to think that this paedomorphic change might be indirectly the responsible mechanism of stabilizing the observed size differences between ecotypes.

The sheltered/exposed (RB/SU) L. saxatilis polymorphism from NW Spain is a good model and well-established case of divergent ecological adaptation to different shore level and microhabitats within the same species. Evidence supporting such differences between ecotypes has been provided by several studies that focussed on size, shell morphology, fecundity, dispersal ability, microhabitat choice and sexual behaviour, whereas a relatively high gene flow between ecotypes was found for neutral molecular markers (reviewed in Rolán-Alvarez, 2007; Johannesson et al., 2010). Two previous experimental studies had already shown that these ecotypes differ in a relatively high proportion of the proteome (7–16%; Martínez-Fernández et al., 2008, 2010). Our study extends such observed differences between ecotypes to 30% of the proteome studied. In addition, the degree of ecotype differentiation in protein expression estimated here was apparently similar across different ontogenetic stages providing evidence that support the hypothesis that differentiation between the ecotypes might be determined relatively soon during ontogeny. However, such a trend is in fact incomplete, as the bias in expression towards one ecotype in embryos is not maintained for some protein spots in older ontogenetic stages, suggesting that some genes are being preferentially activated at embryonic level. Interestingly, in other studies where different (early to late) embryonic stages were analysed in more detail, drastic changes in gene expression patterns were also reported, for example, in oocyte maturation in bovine cattle (Massicotte et al., 2006), early embryogenesis from chicken (Agudo et al., 2005), mouse brain (Hartl et al., 2008) and even gastropods (Sun et al., 2010). Interestingly the proteome and phosphoproteome of larval metamorphosis were also studied in bryozoans and barnacles, and important changes in the phosphoproteome rather than in the proteome itself across ontogeny were detected (Thiyagarajan et al., 2009). The developmental stages included in these studies represent earlier and better characterized embryological (or larval) stages compared with those samples from our experiment. Unlike the above studies, our study involved a wider range of ontogenetic stages that allowed a more comprehensive picture of the evolution of ecotype differentiation at the molecular level in this marine snail.

The few candidate proteins identified in this study allow us to infer a preliminary functional principle regarding ecotype differences. First, we observed that the identified proteins cover a wide range of protein functions, suggesting that differences between ecotypes may be a consequence of general rather than punctual changes in a wide range of cellular functions. Similarly, a proteomic study on domesticated honey bees populations adapted to distinct climates suggests that adaptation processes may involve a relatively large proportion of proteins and metabolic pathways, from biosynthesis to degradation (Parker et al., 2010). In addition, venom from snakes of the genus Crotalus showed large qualitative differences in the proteome analysed in newborn compared to adult specimens from Central American species, which directly affects neurotoxicity to rodents, whereas a paedomorphic pattern was observed in the venom composition of adult specimens from South American (Calvete et al., 2010). Second, this study provides further support for the involvement of at least two candidate proteins in the molecular mechanism underlying ecological adaptation of the two L. saxatilis ecotypes in different ontogenetic stages: arginine kinase (AK) and fructose-bisphosphate aldolase (FBA) (Table 3 and Table S3). These two proteins had been already identified as possible candidates in a previous proteomic study using adult samples directly collected from field and analysed using 2-DE + MS approach (Martínez-Fernández et al., 2008). Moreover, strong clinal variation was found for an intron sequence of arginine kinase (AK) in two different ecotypes of Littorina fabalis (Kemppainen et al., 2011). However, the expression biased towards one particular ecotype observed in our study is not necessarily informative unless we are be able to determine whether the particular set of identified spots as, for example, AK, represent the product from different genes, alleles or the same protein (amino acid sequence) but post-translationally modified (see further Discussion/Box 2 in Diz et al., 2012). Interestingly some of the identified spots are quite close in the proteome map (Fig. 3 and Table 3), which suggests that they can be either isoenzymes or even the same enzyme (i.e. same amino acid sequence) but post-translationally modified (see Jensen, 2006).

For instance, spots 455 and 212 (identified as AK) appear with a slight shifting in the horizontal axis (pI) of the 2-D proteome map, that is, showing a slight change in their pI. This slight shifting could be, for example, compatible with a post-transcriptional modification such as phosphorylation. The same argument stands also for spots 202, 203 and 206, which were identified as FBA. On the other hand, spot 456 varies slightly in its molecular weight (MW) compared to spot 455 (both identified as AK), which might be compatible, for example, with a glycosylation. These two post-translational modifications, or any other of those described so far, could have profound consequences in the protein function (see Jensen, 2006). AK spots seem to be preferentially highly expressed in RB ecotype and the expression difference maintained across ontogeny (but see Diz et al., 2012). FBA spots (202 and 206) typically showed a higher expression in RB ecotype, but the protein spot 203 (identified also as FBA) showed an up-regulated expression level in the SU ecotype, at least for some ontogenetic stages (see supplementary Table 2; see also Martínez-Fernández et al., 2008). As stated above, the functional interpretation of AK/FBA variants is presently difficult. These two enzymes play key roles in the energetic metabolism. For example, AK catalyses the reversible transfer of high-energy phosphate from arginine phosphate to ADP to form ATP, providing a new energy supply in situation of environmental stress (Ellington, 2001). FBA, or aldolase, is a key metabolic enzyme that catalyses the cleavage of fructose 1,6-bisphosphate into glyceraldehyde 3-phosphate and dihydroxyacetone phosphate in the glycolytic pathway (Brooks & Storey, 1988; Lu et al., 2004). Energy demand, however, may be originated in the SU ecotype when using its highly developed muscular foot for attaching the snail to the substrate during wave impacts or either in the RB ecotype due to a direct response to heat stress, as it was described for example in mussels (Gracey et al., 2008). Thus, this new evidence strengthen the candidature of these two enzymes, and the functional pathways where they are both involved, as good candidate targets for future confirmatory studies focussed on unravelling the functional consequences of the observed changes and their relation to the adaptive process in this two ecotypes.

Concluding remarks

The results from this study support a strong differentiation in gene expression between ecotypes, from embryonic to adulthood stages, which suggests that ecotype differences at a molecular level are set up early in the marine snail development and remain at similar levels (but partially involving different proteins) until late stages of marine snail maturity. Ontogenetic differentiation significantly differed between ecotypes, being higher in RB than in SU ecotype, suggesting that SU ecotype may have evolved by paedomorphosis from a common ancestor of both ecotypes. This can be interpreted in ecological terms as an advantage of SU ecotype for surviving in the extreme environment of lower rocky shore where having a small size can be a clear advantage in the face of strong wave pressure. In this circumstance, the smaller (juvenile) size may produce an evolutionary trade-off, impeding the successful reproduction of those specimens. It is in such context that the evolution of paedomorphosis favours the overall fitness of SU specimens. Finally, the results from ecotype differentiation in protein expression and candidate protein identifications suggest that adaptation may affect (either directly or indirectly) the expression of a relatively large number of genes involved in wide range of biological functions. It is important to note at this point that all results from this study relate to the proteome analysed, as an unknown proportion of the proteome remains not accessible using this methodology (merits and drawbacks of 2-DE were reviewed in Rabilloud et al., 2010; Diz et al., 2012). In summary, the comparison of protein expression levels among different biological factors of a particular species (ecotype and ontogeny) can be used to test some specific hypotheses, to infer general/adaptive patterns and their underlying molecular mechanisms, and can be further used to check novel evolutionary hypotheses that have not been previously shown based on classical morphometric, ecological and genetic studies.


We would like to thank the ECIMAT institution for providing marine laboratory facilities, especially Raquel Sampedro and Nerea González-Lavín who provided technical help during the sampling and maintenance of specimens, and A. Caballero, J. Galindo, R. Fallon, H. Quesada, D. Skibinski, J. Calvete and an anonymous reviewer for their helpful comments on this manuscript. We are also grateful to Manuel Marcos, Beatriz Paz and Paula Álvarez from the Proteomics service (CACTI) at University of Vigo for mass spectrometry analyses including ‘de novo’ sequencing approach, and Nieves Santamaría for administrative contribution. This work has been funded by following institutions: Ministerio de Ciencia e Innovación (MCI) (project codes CGL2008-00135/BOS and BFU2011-22599), Fondos Feder and Xunta de Galicia (INCITE09 310 006 PR and ‘Grupos de Referencia Competitiva’, codes 2010/80 and CN 2011/024). A.P. Diz is currently supported by an ‘I. Parga Pondal’ fellowship from Xunta de Galicia. The authors declare no conflict of interest.