Temporal stability and directional change in a color cline of a marine snail from NW Spain

Abstract

The evolution and maintenance of color clines is a classic topic of research in evolutionary ecology. However, studies analyzing the temporal dynamics of such clines are much less frequent, due to the difficulty of obtaining reliable data about past color distributions along environmental gradients. In this article, we describe a case of decades-long temporal stability and directional change in a color cline of the marine snail Littorina saxatilis along the coastal inlet of the Ría de Vigo (NW Spain). L. saxatilis from this area shows a clear color cline with 3 distinct areas from the innermost to the more wave-exposed localities of the Ría: the inner, protected localities show an abundance of fawn-like individuals; the intermediate localities show a high diversity of colors; and the outer, wave-exposed localities show populations with a high frequency of a black and lineated morph. We compare data from the 1970s and 2022 in the same localities, showing that the cline has kept relatively stable for at least over half a century, except for some directional change and local variability in the frequency of certain morphs. Multiple regression analyses and biodiversity measures are presented to provide clues into the selective pressures that might be involved in the maintenance of this color cline. Future research avenues to properly test the explanatory power of these selective agents as well as the possible origins of the cline are discussed.

Genetically dependent intraspecific color variation, or color polymorphism, is a common phenomenon in a wide variety of animal groups, including insects (Dearn 1990; Suárez-Tovar et al. 2022), arachnids (Oxford and Gillespie 1998), anurans (Hoffman and Blouin 2000), squamates (Abalos 2021), birds (Galeotti et al. 2003), bivalves (Cook 2017), or gastropods (Gefaell et al. 2023). A special case of color polymorphism is clines, in which there is intraspecific and geographical gradual color variation along an environmental gradient (Mayr 1963; Endler 1977; Smith and Smith 2012). Color clines can either consist of gradual changes in the intensity of a continuous color variation (e.g., going from brighter to less bright individuals), or gradual changes in the frequencies of discrete color morphs (see Davison et al. 2019 for a discussion of the continuous vs. discrete variation controversy in gastropod color polymorphism). From an evolutionary point of view, clines are normally caused by selection or neutral diffusion via interbreeding and dispersal (Sætre and Ravinet 2019), and they can be of interest to evolutionary biologists inasmuch as they can provide clues into the putative adaptive significance of color variation. Color clines have also been described in several terrestrial and marine taxa (e.g., Dearn 1981; Phifer-Rixey et al. 2008; Antoniazza et al. 2010; Cooper 2010; Takahashi et al. 2011; Köhler et al. 2017; Magnúsdóttir et al. 2018).

A particularly understudied topic in the evolutionary ecology of color clines is its temporal stability and change across prolonged periods (i.e., decades-long). So far, not many studies have systematically explored this issue, in part because it is usually hard to obtain reliable data on past dynamics of color variation (de Jong and Brakefield 1998; Cook et al. 2002; Mullen and Hoekstra 2008; Saccheri et al. 2008; Brakefield and de Jong 2011; Evans et al. 2018; Hantak et al. 2021). Particularly important in this realm are the studies conducted in the model organism Cepaea nemoralis (e.g., Wolda 1969; Cook and Pettitt 1998; Cook et al. 1999; Ožgo and Schlthuizen 2012; Cameron et al. 2013; Ramos-Gonzalez and Davison 2021). This kind of investigation is of uttermost importance not only because of the insights it can provide on how evolutionary forces shape biodiversity in the medium-to-long run but also because it can yield interesting information on how recent anthropogenic processes such as climate change can affect these evolutionary dynamics (e.g., de Jong and Brakefield 1998; Brakefield and de Jong 2011; Ožgo and Schlthuizen 2012; Roulin 2014; Consentino et al. 2017; Evans et al. 2018).

The rough periwinkle (Littorina saxatilis, Olivi 1792) is a marine gastropod species known for its high phenotypic diversity and its role as a model organism for evolutionary ecology (e.g., Johannesson 2015; Rolán-Alvarez et al. 2015). L. saxatilis lives in the rocky intertidal ecosystems all over the American and European coasts of the North Atlantic Ocean, where it feeds on microalgae that grow on the surface of rocks (Reid 1996). In part due to its ovoviviparous mode of reproduction (that significantly limits its dispersal capability), as well as to the heterogeneity of the intertidal ecosystem, L. saxatilis shows a high tendency to evolve local adaptations and to present an enormous intraspecific phenotypic variation (reviewed in Rolán-Alvarez et al. 2015). Indeed, several different ecotypes have been described in L. saxatilis, among which the Crab and Wave ecotypes stand out. These ecotypes—present at least in the Swedish, British, and Spanish coasts—show specific morphological and behavioral adaptations to the different parts of the intertidal ecosystem, and molecular data show that there exists partial reproductive isolation between them (Rolán-Alvarez 2007; Johannesson 2015; Johannesson et al. 2017).

In addition to the Crab and Wave ecotypes, of particular interest within the L. saxatilis system is the high diversity of color morphs that it can present (see Figure 1A). These color morphs can be found either within the same populations or between populations in geographically adjacent areas, sometimes giving rise to color clines (Reid 1996 and references therein; Gefaell et al. 2023). A region where a high diversity of color morphs of L. saxatilis exists is the Rías Baixas (Galicia, NW Spain). The Rías Baixas constitute a series of 4 coastal inlets that originated through the immersion of river valleys that are open to the sea (Méndez and Vilas 2005). In at least 3 out of 4 of these Rías, L. saxatilis presents a purported color cline, described decades ago but not verified since then (Fischer-Piette and Gaillard 1961; Fischer-Piette et al. 1961; Sacchi 1979; Sacchi and Malcevschi 1983; Reid 1996). More specifically, such a color cline would have 3 relatively well-differentiated zones: an interior zone, comprising the innermost localities closest to the mouth of the main river of each Ría, and dominated by fawn-like colors (“fulva”); an intermediate zone, corresponding to the midway localities of the Rías and in which high color diversity is found in terms of the number of color morphs; and an exterior zone, corresponding to the most wave exposed localities and mostly dominated by a lineated (“lineata”) morph (see Figure 1B).

Figure 1.

(A) Color categories used by Sacchi (1979). Tessellata and sanguinea colors are not shown either because they are not representative of the composition of populations (due to low frequencies; tessellata) or are absent from the Ría de Vigo (sanguinea). (B) Color cline hypothesis for the Ría de Vigo as derived from Sacchi (1979; see also Reid 1996). According to this hypothesis, 3 different zones exist along the Ría de Vigo, each with its own color distribution: an interior zone, with high frequencies of fawn-like (“fulva”) individuals; an intermediate zone, with high color diversity; and an exterior zone, with high frequencies of a lineated (“lineata”) morph. Dots and numbers represent the different localities explored by Sacchi in 1974–1976 and re-sampled by us in 2022. Note that in the north there is a leap from 8_N to 10_N, as no snails were found in the 9_N location (“Moaña, entro il porto” in Sacchi 1979; see the main text). The light pink area represents the approximate urban surface of the city of Vigo. The maps were taken and modified from d-maps (https://www.d-maps.com).

Of these 4 Rías, the Ría de Vigo (the southernmost one; with around 41 km of coast on the north and 57 km on the south) has been the most studied in the past. Sacchi (1979) analyzed the frequency of the different color morphs in several localities across this Ría, corroborating the existence of the above-mentioned color cline, previously described by Fischer-Piette and Gaillard (1961) and Fischer-Piette et al. (1961). Color-frequency data provided by Sacchi (1979), readily available within that article, constitute a unique opportunity to analyze the patterns of temporal stability and variation of a color cline in an intertidal organism. In this article, we attempt to resurvey exactly the same populations as Sacchi surveyed in the 1970s to (1) corroborate the existence of such cline; (2) determine to what extent it has changed or remained stable for the last few decades; and (3) provide a preliminary examination of the potential evolutionary mechanisms behind the medium-to-long-term dynamics of such a color cline, to come up with future ways to experimentally test their explanatory power.

Material and Methods

Sample collection and color assignment

In this study, 2 distinct samples were used: “1979” and “2022.” The “1979” data were taken from Sacchi’s original publication (Sacchi 1979: Tables 1–5). These localities were collectively called “1979” due to the date of Sacchi’s article, but data on such populations were collected in August–September 1974 and 1976 (Sacchi 1979). Sacchi took samples from 42 different localities across the Ría de Vigo (average N = 1020), avoiding the urban area of the city of Vigo (see light pink area in Figure 1B). For our analyses, from these original 42 populations, 5 corresponding to Cíes Islands (an archipelago of 3 islands at the mouth of the Ría de Vigo) were excluded, as these populations do not show geographical continuity with the remaining populations. Moreover, many of these populations are associated with the Cíes Lagoon, in which a different L. saxatilis ecotype exists (Galindo et al. 2020), and this could confound the cline analysis. This exclusion of localities from Cíes Islands yielded a total of 37 valid localities along the Ría of Vigo for the “1979” sample, ranging from the innermost areas (close to the mouth of the Verdugo River, the Ría de Vigo’s main river) to the outermost parts (Cabo Home, N; and Baiona, S). The average distance between localities was 2.61 km (SD = 1.99 km). Based on previous results on genetic differentiation in L. saxatilis from this area (Rolán-Alvarez et al. 2004; Tirado et al. 2016; Rivas et al. 2018), these distances are big enough to consider populations from different localities as partially genetically isolated.

Samples for the 2022 data were collected in 5 nonconsecutive days between February and March 2022. Based on Sacchi’s naming of the different localities and their positions as shown in Figure 1 of that article (Sacchi 1979:9), the coordinates of these localities were inferred (see Table 1). Snails were collected in 36 out of 37 localities (see locality dots and numbers in Figures 1B and 2). The remaining population, corresponding with Sacchi’s “Moaña, entro il porto” (Moaña, within the port; 9_N in our coding system; located between 8_N and 10_N in Figures 1B and 2), was also visited, but no L. saxatilis were found there, probably due to the expansion of the port since Sacchi’s samplings. Although L. saxatilis does not constitute an endangered species and it is very common in intertidal ecosystems throughout the Ría of Vigo, we followed a “reduction” strategy for our sample size (see Russell and Burch 1959), thus significantly reducing the number of snails used (average N = 57) in comparison with Sacchi.

Table 1.

Localities analyzed by Sacchi (1979) and resampled again in 2022, with their corresponding codes and GPS coordinates. Data from the locality 9_N are not shown, as we did not find snails there (see main text). Distance (m) is the distance from each locality to the river mouth. N is the sample size of the color samplings (1979 or 2022). Size (mm) is the average size (± SD) in 15 representative specimens. H’ is the Shannon index and D_Roger is the divergence between studies in color frequencies by Roger’s distance. $\chi$² is the chi-squared test of color frequencies as measured by D_Roger between color samplings. “ns” indicates non-statistical significance, ^**P < 0.01, ^***P < 0.001

Locality	Code	GPS coordinates	Distance (m)	N 1979	N 2022	Size (mm)	H’ 2022	H’ 1979	D Roger	$\chi$ 2
Punta Cabalo	1N	42°19’36.64“N 8°38’10.68”W	6,960	462	58	9.1 ± 2.20	0.86	1.10	0.05	1.9ns
Punta Pereiró	2N	42°19’19.26“N 8°38’50.07”W	8,093	554	52	10.2 ± 1.97	0.80	0.62	0.08	5.9ns
San Adrián	3N	42°18’05.1“N 8°39’20.3”W	11,495	1,368	50	9.4 ± 3.70	1.32	0.90	0.26	395***
Costa San Adrián	4N	42°17’30.05“N 8°39’30.42”W	12,905	486	50	11.2 ± 3.80	1.01	1.36	0.17	53***
Punta Collo de Bestia	5N	42°17’26.07“N 8°40’14.29”W	14,040	1,028	53	11.3 ± 1.49	1.16	0.69	0.12	150***
Borna	6N	42°16’50.75“N 8°41’48.43”W	17,008	1,804	51	6.8 ± 0.94	1.28	1.60	0.25	141***
Punta Arroás	7N	42°16’40.42“N 8°42’25.56”W	18,066	852	57	8.4 ± 0.74	1.78	1.63	0.11	45***
Meira	8N	42°17’0.05“N 8°43’9.46”W	21,307	752	59	10.5 ± 1.94	1.67	1.40	0.30	132***
Praia de Con	10N	42°16’14.13“N 8°44’24.69”W	24,450	846	50	4.8 ± 0.78	2.03	2.11	0.20	101***
Punta Niño Corvo	11N	42°15’47.70“N 8°45’18.60”W	26,019	1,000	64	7.0 ± 1.06	1.19	2.23	0.28	40***
Punta Rodeira	12N	42°15’29.04“N 8°45’56.67”W	27,237	1,250	56	7.0 ± 1.28	0.79	1.62	0.20	57***
Punta Balea	13N	42°15’6.62“N 8°47’17.89”W	30,212	1,220	58	5.5 ± 0.96	0.89	1.78	0.25	62***
Punta Borneira	14N	42°14’48.60“N 8°47’35.73”W	31,018	1,174	48	5.0 ± 0.89	0.29	0.02	0.03	36***
Punta Salgueira	15N	42°15’3.12“N 8°48’15.04”W	32,368	1,064	58	5.5 ± 0.98	0.32	0.08	0.05	61***
Punta Igreixiña	16N	42°15’14.73“N 8°50’3.75”W	36,235	1,200	57	5.7 ± 0.79	0	0	0	-
Punta Subrido	17N	42°14’51.56“N 8°51’47.65”W	40,007	1,822	50	5.3 ± 1.01	0	0	0	-
Cabo Home	18N	42°15’2.18“N 8°52’18.21”W	41,632	1,704	58	5.4 ± 1.03	1.03	0.00	0.19	388***
Muelle Arcade	1S	42°20’17.73“N 8°36’48.03”W	858	1,252	59	9.6 ± 3.57	1.43	0.95	0.31	59***
Punta Puntal	2S	42°20’6.31“N 8°36’52.21”W	1,221	332	73	10.1 ± 1.15	0.71	0.79	0.04	2.9ns
Punta Pesqueira	3S	42°19’46.01“N 8°36’51.01”W	1,982	832	50	8.5 ± 2.95	1.10	0.94	0.11	4.7ns
Cesantes	4S	42°18’57.52“N 8°37’7.82”W	3,915	1,853	63	10.1 ± 1.15	0.93	1.25	0.17	10ns
Punta Soutelo	5S	42°17’17.01“N 8°38’17.26”W	11,631	495	57	10.5 ± 1.93	1.22	1.05	0.31	170***
Cabanas	6S	42°16’53.95“N 8°39’46.13”W	14,253	788	63	9.9 ± 1.71	2.07	1.79	0.26	104***
Punta Cubillo	7S	42°16’08.4“N 8°40’23.8”W	16,061	680	60	9.5 ± 1.23	1.52	1.46	0.53	212***
Areiño	8S	42°15’35.34“N 8°41’56.07”W	18,791	726	48	6.3 ± 1.07	1.62	1.42	0.65	592***
Alcabre	9S	42°13’25.85“N 8°46’19.56”W	28,311	2,118	61	6.7 ± 1.54	1.18	1.13	0.07	139***
Coruxo	10S	42°12’3.21“N 8°47’7.28”W	31,474	1,190	60	5.4 ± 0.98	1.12	0.94	0.06	47***
Canido	11S	42°11’24.75“N 8°48’35.26”W	34,158	852	63	6.0 ± 0.88	1.32	1.95	0.21	69***
Cabo Estai	12S	42°10’57.47“N 8°48’49.55”W	35,432	1,322	54	8.1 ± 1.50	1.53	1.28	0.48	761***
SE Monteferro	13S	42°9’13.61“N 8°50’4.35”W	39,827	696	59	5.7 ± 1.28	0.41	0.70	0.05	4.5ns
SW Monteferro	14S	42°8’52.07“N 8°49’56.00”W	43,863	718	60	5.4 ± 1.18	0.24	0.21	0.02	14**
Panxón	15S	42°8’32.82“N 8°49’33.62”W	44,911	1,030	61	6.8 ± 1.06	0.00	0.39	0.04	3.2ns
Praia América	16S	42°7’32.02“N 8°49’27.14”W	47,616	844	56	4.3 ± 0.56	1.14	1.54	0.34	118***
Santa Marta	17S	42°6’56.26“N 8°50’13.00”W	53,976	1,160	64	7.4 ± 2.10	1.84	1.01	0.18	161***
Punta Tenaza	18S	42°7’34.25“N 8°50’56.89”W	56,305	372	56	6.1 ± 0.88	0	0.72	0.12	8.7ns
Cuncheira	19S	42°7’23.57“N 8°51’6.54”W	57,405	786	53	7.3 ± 1.53	0	0	0	-
Average				1,020 ± 435	57 ± 6		1.0 ± 0.60	1.0 ± 0.64	0.2 ± 0.15

Figure 2.

Pie charts comparing the color frequencies of the different localities along the Ría de Vigo in 1979 and 2022 samples. The map was taken and modified from d-maps (https://www.d-maps.com).

The sample was collected using a sampling quadrat (diameter = 60 cm) and following a stratified method. For this, a characterization of the different microhabitats in the localities was performed based on rock type and intertidal zone. Then, the sampling quadrat was placed randomly at different points of the representative littorinid microhabitat, usually the supralittoral zone (most of the times irrespective of rock type). Sampled individuals were picked up from the sampling quadrat, using as many replicates as were necessary until reaching an average number of ~50 snails. The samples were taken alive from the field to the laboratory and then stored at −20 °C. The determination of the color of the individuals was based on Dautzenberg and Fischer (1912), as shown in Sacchi (1979; see Figure 1A). According to this classification, 10 different color categories would exist; from these, 9 would be present in the Ria of Vigo: albida (white individuals), fulva (a wide class of colorations, from grayish-hazel to yellow-straw—or even light brown), lutea (yellow, from citrine to golden), aurantia (orange), fusca (brown to blackish), nigra (brilliant-black), bande (shells with 2 large dark main bands on a lighter background), tessellata (reticulated or checkered tint), and lineata (dark lines coinciding with the grooves of the sculpture). The remaining category, sanguinea, described as “blood red,” does not exist in the Ría de Vigo. The available evidence for the species in the genus Littorina suggests that shell color (both background, bands, and line patterns) is genetically based, with different background colors and ornamentations coded by different alleles in Mendelian unlinked loci (Ekendahl and Johannesson 1997; Kozminsky 2011; Johannesson and Butlin 2017; Gefaell et al. 2023 and references therein). Genomic analyses have revealed that color traits in the Swedish system of L. saxatilis are influenced by several chromosome inversions (Westram et al. 2018; Koch et al. 2021), although the details and reach of such influences are yet not known.

Out of the previous qualitative descriptions, RGB categories for each color category were created to aid the assignment process (except for bande, tessellata, and lineata, which are not unicolor and can be diagnosed based on observed ornamentations) (see Sacchi 1979:12). To avoid potential biases derived from frequency expectations for the different 1979 localities, the 2022 samples were classified without knowing the precise frequencies and distributions of colors provided by Sacchi (1979).

Ecological data sampling

Ecological data for each locality were collected to provide clues on potential environmental factors associated with the color cline. These data were taken on 4 consecutive days during October 2022 using a sampling quadrat (diameter = 60 cm) and following a semiquantitative procedure similar to the Ballantine scale (used to estimate wave exposure; Ballantine 1961; see also Raffaelli and Hawkins 1996; Little et al. 2009). For this, semiquantitative scales and presence/absence classes for key ecological parameters of the intertidal ecosystem were created based on relevant parameters for the Rías Baixas as derived from key published sources (Rolán 1990; Méndez and Vilas 2005; Muñoz Sobrino et al. 2012). Semiquantitative procedures such as the Ballantine scale constitute a useful approach when dealing with large geographical scales (such as the Ría de Vigo; ~41 km bordering the northern coast, and ~57 km the southern coast), and can provide insights into the main ecological trends at play in intertidal ecosystems (Raffaelli and Hawkins 1996; Burrows et al. 2008; Little et al. 2009).

The recorded variables and their corresponding scales for the ecological sampling are presented in Supplementary Table A1 (see online material available). These tried to capture either physical properties (e.g., type of rock, rock size) or ecological parameters like absence/presence or relative abundance of different intertidal species (Melarhaphe neritoides, Monodonta lineata, Littorina fabalis/obtusata, L. littorea, barnacle species—mostly Semibalanus balanoides and Chthamalus sp.—, Enteromorpha sp., and Lichina pygmaea). These species are unevenly distributed across the Ria, varying mostly according to the degree of wave exposure (Rolán-Alvarez, personal observation). The distance of each locality to the river mouth, as well as the distance between localities, was also calculated using Google Earth Pro (v. 7.3.6.9285) bordering the coast from the mouth of the Verdugo River to each locality. Given the linearity of the Ría de Vigo, this parameter can be used as an indicator of the degree of wave exposure.

Additionally, the salinity of each locality was estimated using a refractometer (averaged over 2 technical replicates), as previous studies show a clear gradient for this variable from the river mouth to the outermost parts of the Ría de Vigo (Torres López et al. 2001; Souto et al. 2003; Ibánhez et al. 2021). The average shell size (i.e., columella length) was also measured to the nearest 0.1 mm using vernier calipers in a set of 15 specimens selected at random from the 2022 data. Even though Sacchi didn’t measure such parameter, size was chosen as a potential indicator of local phenotypic adaptation of Littorina species to distinct microhabitats (either by natural selection or phenotypic plasticity; reviewed in Rolán-Alvarez et al. 2015). Previous studies suggest that size might constitute an important trait that affects mortality and is involved in ecotypic evolution (e.g., Boulding et al. 2017). Raw data on color frequencies and ecological parameters are available online (https://doi.org/10.6084/m9.figshare.21771152.v1).

Statistical analyses

Variation in shell color along the Ría de Vigo as well as its potential association with environmental parameters was studied both by comparing shell color frequencies between 1979 and 2022 samples through classical chi-squared tests and by using a stepwise multiple linear regression of physical and ecological variables on frequencies of the 2 most representative colors of the cline (fulva and lineata), as well as on the average shell size of L. saxatilis for each population. Early in situ inspection revealed that the fulva and lineata morphs represent the dominant morphs in the innermost and outermost localities of the Ria (respectively), so they can be useful in characterizing the shape of the cline. The stepwise multiple regression used different criteria to introduce (P ≤ 0.05) or exclude (P ≥ 0.10) variables into the model.

The Shannon index (H’; Shannon 1948; Jain et al. 1975) was used to calculate color-frequency variation at each locality separately for the 1979 and 2022 samples. Shannon index describes the species (or color morph) richness of a community (or a population) as,

$${\displaystyle \begin{array}{l}{\displaystyle H{}^{\prime }={\displaystyle \underset{i=1}{{\displaystyle \sum ^S}}}Pi\ast Lo{g}_{2}Pi}\end{array}}$$

where S stands for the number of different color morphs, and P_i is the relative frequency of each color morph. The greater the number of colors at relatively intermediate frequencies, the higher the Shannon index (H’). Conversely, monomorphic populations (all individuals showing the same color morph) have H’ = 0. Between-year differences for H’ were analyzed using a fixed-factor ANOVA.

The color-frequency divergence over time (1979 vs. 2022) was analyzed using Roger’s genetic distance (D_Roger; Rogers 1972; Nei 1987). D_Roger can also be used to measure discrete variation in a given phenotypic trait, such as color. Roger’s distance (D_Roger) measures color divergence between populations as,

$${\displaystyle \begin{array}{l}{\displaystyle D_{\mathrm{R}\mathrm{o}\mathrm{g}\mathrm{e}\mathrm{r}=}\frac{1}{\mathrm{L}}\sqrt{\frac{{\displaystyle \sum _{\mathrm{u}}}{({\mathrm{X}}_{\mathrm{u}}-{\mathrm{Y}}_{\mathrm{u}})}^2}{2}}}\end{array}}$$

where L stands for the total number of loci-traits (assuming one trait), X and Y are the analyzed populations, and u is the frequency of a given color. D_Roger = 0 is obtained when there is an identical color frequency between populations, and D_Roger = 1 when there exists a complete frequency divergence between them. The significance of D_Roger was evaluated through chi-squared tests (Sokal and Rohlf 1995). The contribution of each particular color to the former overall distance was estimated using the particular quadratic difference of the previous D_Roger algorithm per locality and averaged across localities (non-normalized). Differences between colors across the 36 localities were evaluated by one-way ANOVA and the Student–Newman–Keuls (SNK) a posteriori test. All the ecological variables were summarized by means of a principal component analysis (PCA) to test whether they could be interpreted as proxies of wave exposure by latter correlating them with the distance to the river mouth. Statistical analyses were carried out with SPSS v.24 (IBM Corp., USA), and plots were made using R version 4.2.0, with ggplot2 version 3.4.0 for data visualization (Wickham 2016; R Core Team 2022).

Results

Color frequency, for both 1979 and 2022 samples, shows a clear clinal pattern along the Ría de Vigo from the innermost to the outermost localities (Figure 2). The innermost localities (sheltered and with low salinity) show a high frequency of fulva individuals, with frequencies > 50% in most of them. The outer localities (wave-exposed and relatively high salinity; around 35–36 g of salt per liter/kg of water) show a high frequency of the lineata morph, even reaching monomorphism (100%) in many of them. Finally, intermediate localities show high color diversity both in terms of the number of color morphs present at each locality (including albida, lutea, aurantia, or nigra) and in terms of the relative frequency of each of these morphs. A significant Pearson’ r correlation exists between the color frequencies of fulva and lineata morphs and each locality’s distance to the river mouth (used as a simple proxy for the degree of wave exposure), irrespective of the data series (1979 vs. 2022): positive for lineata 1979 (r = 0.83; df = 35; P < 0.001) and lineata 2022 (r = 0.83; df = 35; P < 0.001), and negative for fulva 1979 (r = −0.82; df = 35; P < 0.001) and fulva 2022 (r = −0.73; df = 35; P < 0.001) (Figure 3). Taken together, these results both support the existence of a color cline in L. saxatilis, with 3 different color zones across the Ría de Vigo, as well as its overall maintenance over time.

Figure 3.

Frequencies of fulva (up) and lineata (down) for the 1979 (left) and 2022 (right) samples in relation to the distance from the mouth of the Verdugo River. R² coefficients and P-values for the linear regression corresponding to each figure: for fulva 1979 (R² = 0.67; P < 0.001), fulva 2022 (R² = 0.53; P < 0.001), lineata 1979 (R² = 0.68; P < 0.001), and lineata 2022 (R² = 0.69; P < 0.001).

The mean color biodiversity (H’) is also similar between both samples (1979 vs. 2022; Table 1; F_ANOVA = 0.1; df = 1/70; P = 0.875). In addition, there is a high correlation across localities between data samples (r = 0.74; df = 35; P < 0.001). However, putting aside the global pattern of similarity (i.e., stability), some differences also exist between samples, as revealed by D_Roger and chi-squared tests (Table 1). Averaged divergence per locality between periods is about 20% of the maximum possible (D_Roger = 0.20 ± 0.15), and the frequencies differ significantly in most chi-squared comparisons (25 out of 36; Table 1). The contribution of each color to the D_Roger distance suggests that this divergence between samples affects preferentially the main clinal colors (i.e., fulva and lineata; Figure 4), as there were significant differences between colors by ANOVA (F = 5.08; df = 8/315; P < 0.0001). The SNK test identifies 2 significant groups (alpha = 0.05): fulva and lineata vs. the rest of the colors (Figure 4). In turn, this could be interpreted as an indication that, despite the stability of the overall pattern, the cline has changed during the studied period, resulting in an increased frequency of lineata along several intermediate localities (mostly southern) while decreasing that of fulva (see Figure 2).

Figure 4.

The contribution of each color to D_Roger, or the average squared difference in color frequency between the 1979 and 2022 samples for each single color morph. The bars represent standard errors.

Several potential predictive variables of the clinal color distribution (estimated through fulva and lineata frequencies) and the average shell size (a characteristic adaptive trait in this species; e.g., Rolán-Alvarez et al. 2015; Boulding et al. 2017) were explored using the stepwise linear multiple regression (Supplementary Table A2, see online material available) to gain insight into potential causal factors. This analysis shows that the 2 colors (both in northern and southern localities) exhibit a strong association with several variables that capture different physical and/or ecological characteristics of the environment (from 1 to 4 variables). We can interpret such variables as proxies of the degree of wave exposure and position along the Ría, as a PCA including all of them yields a PC1 that explains 41% of the variance and that shows a significant correlation with the river mouth (r = 0.79, P = 0.002; see Burrows et al. 2008 for an analogous use of PCA). In this sense, multiple regression analysis shows that lineata frequency is almost fully explained (R² = 0.97) by 3 variables in the north (including the presence of Monodonta lineata and the abundance of Lichina pygmaea) but only one (presence of Melarhaphe neritoides; R² = 0.69) in the south. On the other hand, fulva frequency is fully explained (R² = 0.98) by 4 variables in the north (absence of Monodonta lineata, mean salinity, average rock size, and Enteromorpha sp. abundance), while in the south a large portion of variance (R² = 0.83) is explained by (low) salinity. Interestingly, a trait typically associated with local adaptation—i.e., size—is fully explained (R² = 0.97) by 2 variables (presence of Melarhaphe neritoides and Enteromorpha sp. abundance) in the north, and explained (R² = 0.83) by 2 variables in the south (presence of Melarhaphe neritoides and L. littorea). All explanatory variables in multiple regression analyses show a clear environmental gradient along the Ría de Vigo, corresponding to the degree of wave exposure.

Discussion

Our study corroborates the existence of a decades-long color cline in L. saxatilis from the Ría de Vigo (Sacchi 1979; Reid 1996). Lab cultures suggest that the Swedish populations of L. saxatilis become sexually mature around 6–8 months (Johannesson and Butlin 2017), and although no reliable data exists for the Galician populations, previous simulation studies assume that L. saxatilis from this area has a generation time of around 6 months (Fernández-Meirama et al. 2022). This means that it is reasonable to assume that the 46–48 years between both samples analyzed in this study (1979 vs. 2022) are equivalent to around 92–94 generations of L. saxatilis. Taken into account that Fischer-Piette and collaborators (Fischer-Piette and Gaillard 1961; Fischer-Piette et al. 1961) already described this color cline in the early 1960s (although they do not provide actual data for comparison with the time series presented in this study; see also Reid 1996), the temporal stability shown in these populations may have lasted for more than 60 years, equivalent to around 120 generations of L. saxatilis. Previous studies analyzing the temporal dynamics of color clines suggest that this number of generations might constitute time enough for evolutionary processes to act on populations. For instance, Brakefield and de Jong (2011) show a significant decay in the steepness of a color cline in the 2-spot ladybird (Adalia bipunctata) over the course of 50 generations, presumably driven by a genetic response to thermal changes induced by climate change along a 115 km transect in the Netherlands. Cameron et al. (2013) found that in about half the number of these generations, banding morphs of Cepaea nemoralis did not change over a cline, while the frequency of the yellow morph increased all over it. Therefore, depending on whether the putative selective pressures acting on a cline vary or not over a given timespan, and whether drift and migration are strong enough to influence the evolutionary outcomes, we might expect changes or stability in the shape of clines across generations.

The other remarkable phenomenon revealed in this study is the variability observed in some populations, despite the overall trend remaining stable over the past decades. We can distinguish between 2 different types of variability: what might be called “local variability” and “directional change.” By local variability we mean the divergence in color frequencies found in specific populations, such as 10_N or 12_S (see Table 1 and Figure 2). Conversely, directional change refers to predictable and recurrent patterns of variation in the frequency of certain morphs. When we compare data from 1979 and 2022, a systematic and recurrent pattern of variation in the lineata and fulva morphs is observed (lineata typically displaces fulva in several intermediate-to-inner localities; Figures 2 and 4).

Occasional sampling errors (errors in the geographical correspondence of the sampling sites studied in 1979 and 2022, since Sacchi’s article does not provide any geographic coordinate system to identify exact sampling points), assignation errors, or genetic drift could potentially explain local variability. However, although it cannot be known whether the 2022 sampling points exactly correspond to those of 1979, or whether the assignation process was done using the exact same criteria, we believe that the temporal stability and directional change observed in this cline between different time series cannot be simply attributed to either sampling or assignation errors, or genetic drift. As for assignation errors, it is virtually impossible to confuse lineata and fulva ecotypes (the morphs mostly involved in both the temporal stability and directional change of the cline). As for sampling errors or genetic drift, such factors would preferentially affect rare colors (due to the fact that random errors tend to affect lower frequencies most), while the contribution of individual colors to Roger’s distance (D_Roger) shows the opposite pattern (more pronounced in lineata and fulva, the most frequent morphs in the outer and inner localities, respectively; Figure 4). Another possible caveat with respect to the observed patterns has to do with potential seasonality effects. Our 2022 data were taken in different seasons (autumn–winter) than these of 1979 (summer). There are only a few studies so far that have analyzed the seasonal changes in the frequency of snails from the genus Littorina, yielding negative results. Ekendahl (1994, 1995) didn’t find any seasonal changes in the frequency of color morphs of the Swedish populations of L. fabalis. This researcher found, however, seasonal changes in microhabitat choice. Nevertheless, Carballo et al. (2005) found no seasonal changes in the microdistribution of the Crab and Wave ecotypes of L. saxatilis in a 3-year study close to the Ria de Vigo. We cannot formally rule out that seasonal microhabitat changes might have affected the clinal color frequencies compared in this study, but given the above-mentioned study, we believe this is highly unlikely. Therefore, the evidence presented in this study suggests that both the temporal stability observed in the cline and the directional changes detected in certain populations are real, demanding mechanistic explanations.

Regarding the directional change, the lineata morph is mostly associated with ecological parameters related to wave exposure and relatively high salinity (in a way similar—but not restricted—to the Crab ecotype; Rolán-Alvarez et al. 2007). The expansion of this color morph to the intermediate zone of the Ría (with a corresponding decrease of fulva) could indicate a change in any of such parameters from the outer to the mid localities over the past 46–48 years associated with an expansion of Vigo’s main city of Ría de Vigo. However, no systematic analyses of this possibility have yet been carried out, so future studies should systematically investigate this and related scenarios.

As for the temporal stability of the color cline of L. saxatilis in the Ría de Vigo, 2 main hypotheses could be invoked as potential explanatory factors: either the combination of migration and genetic drift or some kind of natural selection along one or more ecological gradients (e.g., Mayr 1963; Endler 1977; Ridley 2004; Sætre and Ravinet 2019). We believe that such stability in a clinal pattern is unlikely to be explained through the combination of migration and genetic drift, as such mechanisms predict a much greater degree of stochasticity. Moreover, the cline described in this study shows not only temporal stability but also spatial replication. Considering the low dispersal capacity of L. saxatilis (Rolán-Alvarez et al. 2015), it is highly unlikely that there is strong genetic connectivity between its northern and southern populations. This suggests that both clinal trends (north and south) are maintained independently through non-stochastic forces, and may qualify as separate spatial replicas. Furthermore, such a clinal pattern has been described in other Rías from the Rías Baixas (Reid 1996; Fischer-Piette and Gaillard 1961; Fischer-Piette et al. 1961; Rolán-Alvarez, personal observation). Although the genetic structure and connectivity between the northern and southern populations of L. saxatilis from the Ría de Vigo should be studied through different genetic markers to formally rule out the migration plus drift hypothesis (e.g., as in Westram et al. 2018), we believe that this is an a priori less convincing scenario.

A much more plausible explanatory mechanism might involve some form of natural selection. In addition to the temporal stability and spatial replication of the cline (which by its own right might already suggest the action of selective forces; Endler 1986), the main reason for this is that there are strong environmental gradients (e.g., temperature, salinity, CO₂, and wave exposure) along the Ría de Vigo that go parallel to the color cline (Ruiz-Villareal et al. 2002; Gago et al. 2003; Muñoz Sobrino et al. 2012; Ibánhez et al. 2021). These environmental gradients might give rise to differential selective pressures that end up maintaining the cline in a relatively stable form.

Furthermore, multiple regression analyses show that several key ecological variables can explain the observed color distribution, thereby strengthening the case for natural selection. The lineata and fulva color clines are clearly explained by a few ecological or physical variables that show gradients from inner to outer localities. For example, Monodonta lineata, Melarhaphe neritoides, and Lichina pygmaea tend to be associated with moderate to highly wave-exposed rocky shore areas (Ballantine 1961; Rolán 1990; Burrows et al. 2008), and they show concomitantly positive regression coefficients when explaining the lineata clines, whereas, for instance, Monodonta lineata shows a negative regression coefficient when explaining the fulva ones. The degree of wave exposure might be linked to differences in temperature or salinity, which in turn might constitute powerful selective agents. In the case of temperature, for instance, Phifer-Rixey et al. (2008) provide experimental evidence that thermal stress tolerance might be driving a color cline in L. obtusata from the Gulf of Maine. As for salinity, which also shows a concentration gradient over the Ría (e.g., Muñoz Sobrino et al. 2012; this study), its positive regression coefficients with lineata frequency and negative with fulva frequency suggest that it constitutes another selective pressure operating along the cline. In this sense, previous experimental evidence exists that the different color morphs of L. saxatilis from the White Sea have different saline-stress tolerance (Sokolova and Berger 2000), although the molecular mechanisms for such association are not known yet (see also Minton and Gunderson 2001 for a similar case in a different gastropod species). If that would be the case, then shell color would not be the primary target of selection, but only a correlated trait genetically associated with saline-stress resistance (or other biotic or abiotic factors) genes. This genetic association would not be surprising, as genomic data suggest that in L. saxatilis chromosomic inversions may contain several loci involved in adaptive traits, thereby contributing to low recombination between them (Koch et al. 2021).

As for the explanation of cline formation, 2 different mechanisms have been put forward: primary differentiation and secondary contact (Endler 1977). Spatial heterogeneity of a formerly homogenous environment leads to cline formation by primary differentiation. A similar primary differentiation scenario can be achieved through the colonization of a novel environment with distinctive characteristics by a species hitherto confined to a homogeneous area. In our case, the latter primary differentiation scenario is a priori supported by the historical geomorphology of the Ría de Vigo: before the early Holocene sea-level rise (12,000–7,000 years ago), sea level at this location was about 30–35 m below the current level, with the Ría de Vigo forming a river valley and the coast being located to the west of the Cíes Islands (Costas et al. 2009). As the sea level rose, marine water inundated the river valley, thereby gradually giving rise to the different zones (exterior, intermediate, and interior; in that order) of the Ría. It is possible, therefore, that from that point on populations of L. saxatilis began to colonize the intermediate and inner zones of the Ría, thereby being exposed to different selective pressures. An alternative hypothesis on the formation of the color cline observed in this study is viewing it as the product of a secondary contact between 2 distinct ecotypes (lineata and fulva) that differ in shell color and a suit of related traits, and that are locally adapted to the different ends of a temperature, salinity, and/or wave exposure gradient. Under this scenario, both ecotypes would show at least partial reproductive isolation between them, and the intermediate zone of the Ría would constitute a hybrid zone with a more or less steep cline in traits and allele frequencies, depending on the strength of divergent selection (as in the Crab and Wave ecotypes; see Westram et al. 2018). Although this latter scenario fits worst with historical geomorphology than the former, future studies on DNA markers and experimental settings should be put in place to reveal which of them is the correct one and which selective agents might constitute the causal mechanisms of this decades-long clinal pattern in a marine snail.

Supplementary Material

Supplementary material can be found at https://academic.oup.com/cz.

Funding

This work has received financial support from the Ministerio de Ciencia e Innovación (PID2021-124930NB-I00), Xunta de Galicia (GRC, ED431C 2020-05), Centro singular de Investigación de Galicia accreditation 2019–2022, and the European Union (European Regional Development Fund—ERDF). Juan Galindo was funded by a JIN project (Jóvenes Investigadores, Ministerio de Ciencia, Innovación y Universidades, RTI2018-101274-J-I00). Juan Gefaell was funded by a Xunta de Galicia Predoctoral Research Contract (ED481A-2021/274).

Conflict of Interest

The authors declare that they have no conflict of interest.