CEPAEA NEMORALIS ( L . ) IN GÖTEBORG , S . W . SWEDEN : VARIATION IN A RECENT URBAN INVADER

Although present since the mid 19th century, the introduced snail species Cepaea nemoralis has spread extensively in Göteborg, mainly in the last two decades. Early samples show the predominance of yellow-shelled morphs, but in modern populations pink shells predominate. There is great variation in morph frequencies among modern populations with high values of FST, and with no relationships to habitat. The patterns shown resemble those seen in Sheffield (UK), a city also colonised over a similar time period, but differ from those seen in regions where the species has been established for much longer. The combination of early and recent records suggests not only that founder effects play a large part in determining morph frequencies, but that present populations derive from multiple colonisations from different sources. key woRDs: Cepaea nemoralis, polymorphism, invasion, Göteborg


INTRODUCTION
The numerous studies on the shell colour and banding polymorphism of the European land snail Cepaea nemoralis (L.) have revealed a great variety of patterns of variation (Jones et al. 1977, Cook 1998).While there is no doubt that natural selection influences these patterns in many cases even if the agencies involved are unknown (OżgO 2008, silveRtown et al. 2011), there is also evidence that founder effects and the balance between genetic drift and gene flow also affect such patterns (COOk 1998).This is especially true where populations have been small and isolated in the past (CAmeRon & Dillon 1984), or where C. nemoralis has recently colonised previously unoccupied areas, usually assisted by accidental transport by humans (CAmeRon et al. 2009).Even in such cases, however, evidence of rapid selection in relation to the shade or darkness of the habitat or to a measurable change in macroclimate has sometimes been found (OżgO 2005, 2011, OżgO & kinnisOn 2008, OżgO & sChilthuizen 2011, silveRtown et al. 2011, CAmeRon et al. 2013).The known genetic basis for most of the polymorphism makes the species a good model organism for studying the mechanics of evolutionary processes over a range of time-spans (OżgO 2008(OżgO , 2014)).
While C. nemoralis has become less frequent and widely-distributed in parts of its range (Cowie & JOnes 1987), it has also colonised previously unoccupied areas, both within its assumed natural range and in places where it has been introduced beyond it, especially towards the eastern half of Europe (OżgO 2005, PeltanOvá et al. 2012).The species is well-adapted to human-made habitats such as hedges, waste ground and gardens, and it has spread rapidly in many cities.CAmeRon et al. (2009) examined variation in two cities, Sheffield in central England and Wrocław in S.W. Poland.Although Sheffield is within the natural range of the species, its occupa-tion of the city has been very recent, following great reductions in industrial pollution.Wrocław was colonised earlier; it is questionable whether it lies within the natural range.There were marked differences in the pattern of variation between the two cities; while neither showed any effect of habitat on morph frequencies, variation among populations was much less in Wrocław and populations were more polymorphic.This was attributed to a longer time present in the city, and greater continuity and therefore gene flow among populations.
The evidence from Holocene deposits indicates that C. nemoralis is an introduced species in Sweden (Waldén 1986).While fossil records from the Middle Ages suggest that it has been established for several centuries in the extreme south (Skåne) of the country, it has spread recently over much of Sweden south of ca.59°N, and has very recently been found even further north (T.von PRosChwitz, unpublished data).The oldest record of C. nemoralis in Göteborg (57.7°N) is from the beginning of the 1850s (mAlm 1851).At that time the species must have been a very rare element, and only one further occurrence (from a cemetery about 2 km from the original site) is known from the 19th century.In 1921-1922 the land snail fauna in 148 sites in the city and its immedi-ate surroundings was investigated and showed only three localities for C. nemo ralis (lohmAnDeR 1923).The series consisted both of natural and strongly man-influenced / man-made habitats.The same sites were re-investigated in the 1970s, when the number of localities holding C. nemoralis had increased only moderately to seven (wAlDén 1992).The polymorphism in a population detected in 1967 in Askim, ca. 12 km south of the centre of the city, was described by Meeuse (1968).In an investigation of 37 manmade and strongly man-influenced sites in 1978-1986, all of which must be considered as suitable for C. nemoralis, it was found in ten (von PRosChwitz 1988).Today the species occurs in many of these sites.There is thus good reason to believe that the present common occurrence of C. nemoralis in Göteborg is a result of increased spread during a rather short period, mainly in the last two decades.
The city has a metropolitan area and human population very similar to those of Sheffield and Wrocław.In this paper, we analyse the variation of C. nemoralis populations in Göteborg.Apart from comparisons with the other two cities using recent records, the earlier records enable us, with reservations, to track changes in frequencies with time, and relate these to the process of spread.

AREA STUDIED
Recent and historic records of C. nemoralis in the vicinity of Göteborg include a number from rural locations some distance from the city, and from small islands to the west.Most, however, come from urban environments within the city or its neighbouring towns.To make comparisons with the entirely urban populations sampled in Sheffield and Wrocław, we have confined our analyses to sites lying within the rectangle defined as between 11.8 and 12.2°E and 57.6 and 57.8°N, an area of ca.540 km 2 (Fig. 1).This area excludes offshore islands and the rural hinterland of the city.The topography is one of small hills and valleys on a granite platform.The city receives ca.780 mm of precipitation per year, with significant rainfall in all months.January mean temperature is ca.−1.5°C and for July ca.16°C.Comparable data for Sheffield are 800 mm, 3.8 and 16.2°C, and for Wrocław 570 mm, −0.5 and 19.9°C.

MATERIAL AND METHODS
The samples of C. nemoralis have three sources.The first is the lots of specimens held in the Göteborg Natural History Museum.These were collected by a number of people, but were all scored for shell variation by teD von PRosChwitz (see below).The earliest sample comes from a site described as Dahlin's garden or meadow, made in the 1850s.This site was sampled on a number of occasions in the 19th century, but it changed during that century as a result of development, and now holds apartment blocks.There are a few samples from the first 70 years of the 20th century, but most consist of single or only a few shells.From 1975, and especially since 1990, there are more and larger samples available.
The second source is a single sample reported by Meeuse (1968) at Askim, south-west of the city centre (Fig. 1), scored by Meeuse himself and accompanied by extensive ecological notes.He distinguished between "red", "reddish yellow" and "yellow" shells.All the "reddish yellow" shells were banded, and we have assumed that all these and those in the "red" category were pink in the conventional scoring system outlined in Jones et al. (1977); banding usually dilutes the strength of the shell colour in pink shells.
The third source is samples made by RhonA Cox and RobeRt and AlexAnDeR CAmeRon in 2012 and 2013.These samples were scored in the field, and live snails were returned to the habitat in which they were found.The habitats in which these samples were made were classified as shaded (usually in urban woodland) or open, where the site was not overshadowed by trees.This classification was necessarily imprecise, though made by the same observers; urban habitats are frequently very heterogeneous even within very small areas.The presence or absence of the closely-related Cepaea hortensis (Müller) was noted for each site.Only adult snails or large juveniles in which the dark colour of the lip was emerging were scored to avoid any confusion with that species.A few dissections of white-lipped adults were carried out.All white-lipped snails were C. hortensis; these were much smaller than C. nemoralis.Samples from the first and second sources were located by reference to streets or addresses; those from the third by GPS on site.Site co-ordinates have been standardised to decimal degrees.
Scoring of colour and banding morphs basically followed the system outlined by Jones et al. (1977).Scoring of banding variation was done in more detail for the samples held in the Natural History Museum than for those scored in the field, with precise notes on the fusion of bands and of the absence of individual bands.These details and the whole data set are available as a digital archive from the Göteborg Natural History Museum at www.gnm.se/cepaea.For the analysis, three colours: yellow, pink and brown, and four banding categories: unbanded, mid-banded (00300), trifasciate (00345), and many-banded (12345) were recognised (Jones et al. 1977), with minor variants being allocated to the most appropriate class.Thus 00045 shells were allocated to the trifasciate class, and 10345 to the many-banded class.The genetics of the classes recognised is well-known; that of fusions of bands is not, and it is difficult to  -pre-1923, red squares -1951-1989, blue triangles -1990-1999, yellow circles -2000 onwards.Built up areas shown in grey standardise recording among observers.Only samples containing 10 or more scoreable shells were used in formal analyses.
The distinction between pink and brown shells is not always easy.Most brown shells sampled were dark and unbanded, but in a few cases banded shells may have been misallocated.A more serious problem concerns the colour of old specimens, as colours fade even in specimens kept in the dark.As shown later, there is a highly significant decline in the proportion of yellow shells over time, with early (pre-1990) samples containing mostly shells scored as yellow.Great care was taken in scoring these early samples.For the main analysis, however, we have considered only samples made since 1990; in any case the primary aim was to examine current patterns of variation.
In all analyses, we have used morph frequencies, not estimated allele frequencies.For midbanded the frequencies are those within the banded shells; for trifasciate they are those within shells with more than one band, reflecting the dominance hierarchy at these loci (Jones et al. 1977).Where regression or least squares correlation has been used all these frequencies have been arcsine transformed.Besides tests of association of morphs with each other and with position and habitat, we have examined the linkage disequilibrium between shell colour and banding, taking simply the proportion of unbanded shells within each colour class and other associations among colour and banding classes.
Following CAmeRon et al. (2009), spatial pattern has been examined using Moran's I for the frequencies of yellow, brown, unbanded, mid-banded in banded and trifasciate in many-banded.Overall genetic similarity has been analysed by Principal Components Analysis (PCA) using the same frequencies and Moran's I has been calculated also for the site scores at the first two PCA axes.Variation within and among populations has been estimated by the proportion of samples visibly monomorphic at each locus, by the proportions of samples with different numbers of morphs present, and by estimating F ST based on morph frequencies (CAmeRon et al. 2009).We estimated means and standard deviations of these F ST values using a bootstrapping procedure with 10,000 permutations.The software used for the analyses comprised: SAM (Rangel et al. 2006) for Moran's I and R (R CoRe teAm 2012) for all the others.

SITE DISTRIBUTION AND VARIATION OVER TIME
Appendix Table 1a gives details of all samples held in the Göteborg Natural History Museum and that of Meeuse (1968) in date order.Appendix Table 1b gives details of all samples scored in the field in 2012/13.The more detailed scores and locality details for samples in Appendix Table 1a are available as a digital archive from the Göteborg Natural History Museum (see above). Figure 1 shows the distribution of sampling sites within the study area.There are very few sites that have been sampled on more than one occasion, and with the exception of the first site at which C. nemoralis was found (Dahlin's garden), such a repeat sample was made within only a few years of the original and we have combined the samples to give an aggregate score.We cannot follow any changes with time on a site-by-site basis.Any apparent change in mean or median frequencies over time could thus result simply from the sampling of different sites; even within 21st century samples, the range of frequencies recorded is great (see below).The number of samples in some ranges of dates is also small.Nevertheless, inspection of the data demonstrates a substantial decline in the frequency of yellow shells over time, perhaps complicated by the difficulty of accurately scoring shells stored for several decades.
In the period from 1850 to 1922 there are records from only five sites, one of which is not precisely located.Five samples come from the first recorded locality (Dahlin's garden), and were made over a period of around 40 years.The earliest, containing only two shells, was made in the 1850s.All but one of the shells collected are yellow, and all fall into the trifasciate or many-banded categories (Table 1).There are significant variations in the proportions of these two forms between sampling occasions, but no consistent trend with time.No unbanded or midbanded shells were found.In the remaining samples from this period no pink shells were recorded, but a few unbanded and mid-banded shells were present.
No samples are held in the Göteborg Natural History Museum for the period 1923-1950.The frequencies of yellow shells in later samples are shown in Table 2.The small samples made between 1951 and 1989 are predominantly yellow.In the first part of this period, all samples are small, but from 1967 to 1985 there are ten samples with 10 or more shells.The frequency of yellow shells varies greatly among them.While the median is 62.3%, it is noticeable that these samples fall into two categories: those with less than 25% yellow shells (4) and those with more than 60% (6) (Appendix).There is no geographical pattern; samples with very different frequencies lie close together.Collectively, samples from this period contain some shells in all colour and banding combi-nations except that brown banded shells are missing (Appendix).
Between 1990 and 1999 there are only five samples with 10 or more shells.The median frequency of yellow shells is just below 50%, and there is relatively little variation among them.Just less than 40% of shells in the 17 smaller samples are yellow.Among 21st century records medians are also below 40%, but the range of variation among samples is large.While this decline in the frequency of yellow shells with time is of interest, investigation of relationships with position or habitat, and estimates of the amount of variation among samples, needs to exclude systemic changes over time.For analysis of such patterns we have therefore confined formal analysis to samples of 10 or more shells made since 1990.There are no significant differences in median values of yellow among large samples from different periods after this date.

PATTERNS OF VARIATION IN MODERN SAMPLES
In total, 84 samples of more than 10 shells made since 1990 were available for the analysis (Table 3).No morph was present in all samples, and the range of recorded frequencies was great in all.Among the samples collected in 2012-2013 and scored in the field, there are no significant differences in any morph frequencies between shaded and open habitats overall (Table 4), and some of the small differ-ences observed run counter to expectation; yellow has a higher median frequency in shaded than in open sites.In the classic Cain and Sheppard dia- gram (COOk 2008), where "effectively unbanded" represents the proportion of unbanded, midbanded and trifasciate shells in the total, samples lie mainly in the bottom right sector, and there is no clear pattern of difference between habitats (Fig. 2).
When each shaded site is compared with its nearest open neighbour in a paired sample comparison, the same result is obtained (Table 4); indeed, the excess of pairs in which yellow is more frequent in the shaded habitat borders on formal significance (χ 2 = 3.85).The lack of the expected trend is not a product of geographical differences in the distribution of habitats.We note that there is also no correlation in morph frequencies between members of each pair, even though the median distance separating each is only 0.6 km.Similarly, the recorded presence or absence of C. hor tensis has no significant effect on median morph frequencies (data not shown).Within the whole data set, there are also no significant correlations for any morph with latitude or longitude; values of R, the correlation coefficients (calculated on arcsine transformed values) are generally less than 0.1.With only one exception, there are no significant among-sample associations between morph frequencies at different loci.There is a weak association between the frequencies of brown and of unbanded shells; both morphs are absent from many samples, but where unbanded is present there is no difference in its median value between those samples that do or do not contain brown shells.Nearly all brown shells are unbanded (see below).
Associations between morphs at different loci within samples are similarly weak, with the excep-   tion of that between brown and unbanded (Table 5).
Of 22 samples containing brown shells, 18 have only brown unbanded shells, and in every case unbanded has a higher frequency in brown than in either yellow or pink shells.There is a barely significant trend for unbanded to be in excess in pink rather than yellow shells (the loci are linked), but there are no significant associations between colour and mid-banded or trifasciate.A simple summation of all shells shows the same pattern.
Positive spatial autocorrelation in morph frequencies as estimated by Moran's I is present over short distances in the proportions of unbanded, mid-banded and brown shells, but not in those of yellow or trifasciate shells (Table 6).Figure 3 shows a distinct cluster of sites in which brown shells occur, although these are not the only sites to hold the morph.Despite the formal significance of these autocorrelations, the associations are weak when contrasted with the maximum possible values of the index.In the PCA analysis (Fig. 4), the first axis is strongly correlated with the frequency of trifasciate and shows no spatial relationships.The second axis shows varying correlations with the other morphs, and also shows significant positive values of Moran's I at short distances.Site scores on this axis show a significant negative association with latitude (R = 0.405, df 82, P < 0.001), albeit with a wide scatter.No single morph showed a significant association with latitude (see above).
Values of F ST estimated from morph frequencies are given in Table 7, together with the equivalent values for Sheffield and Wrocław.Table 8 shows the extent of visible monomorphy at those loci comparable across all three cities and Figure 5 shows the frequency distribution of samples with varying numbers of the six morphs that can be compared across all three cities (yellow unbanded, mid-banded and many-banded (including trifasciate), and the same combinations for pink).In all cases, samples from Göteborg resemble those of Sheffield rather than those from Wrocław.Table 7 also shows estimates of F ST for a very limited set of populations from Skåne, an area thought to have been occupied for several centuries.These values resemble those of Wrocław.

DISCUSSION
Genetic differentiation among C. nemoralis populations is heavily influenced by their slow movement and poor powers of active dispersal.While local differences in selection can create differences in genetic composition over distances of only a few tens of metres (OżgO 2008(OżgO , 2011)), there are other patterns of microgeographical variation that reflect founder effects and the common origin of adjacent populations (CAmeRon & Dillon 1984).Such patterns are also found in molecular variation (Arnaud et al. 1999, BelliDo et al. 2002, sChWeiger et al. 2004).They are reflected in distance decay in genetic similarity.Both selection and the sources of existing pop-ulations may be involved in particular cases (Jones et al. 1980, CAmeRon & PAnnett 1985, OżgO & kinnisOn 2008).
Such patterns can be detected even in areas with long-established and relatively continuous populations (CAmeRon & PAnnett 1985).However, they are most evident in areas where populations have been restricted to a few isolated refugia in the past and have subsequently spread (CAmeRon & Dillon 1984), or in areas that have recently been colonised, usually as a result of accidental transport by humans followed by limited local active dispersal (CAmeRon et al. 2009).In the former case areas of several km 2 Fig. 5. Frequency distribution of samples with given numbers of morphs (maximum 6) in Sheffield (black), Göteborg (grey) and Wrocław (white).See text for further details  Populations of C. nemoralis in Göteborg show a pattern of local differentiation that shows no evident relationship to habitat or to the presence or absence of a similar species.Although there is a latitudinal influence on overall genetic similarity, perhaps reflecting sources of colonisation, there are no patterns resembling "Area Effects", where large numbers of nearby populations have near-identical and often extreme morph frequencies.The pattern is mainly of very local trends for some but not all morphs to have similar frequencies in populations that are close to one another.The differentiation among populations is strong: many morph frequencies span nearly the whole range from 0-100% and values of F ST are high (see below).
In these respects, populations of C. nemoralis in Göteborg resemble those studied in Sheffield much more than those in Wrocław, the two cities sampled by CAmeRon et al. ( 2009).The degree of differentiation was less marked in Wrocław than in Sheffield and populations there were more polymorphic.These differences between the two could be related both to the history of the species in each and the degree of interconnectedness among populations; there was evidence that Wrocław had been occupied for longer, and had more continuous populations.In the case of Sheffield, the species had spread mostly in the previous twenty years.A priori, the pattern shown in Göteborg can be interpreted in the same way as for Sheffield: the very recent spread by accidental transport of small numbers by humans followed by some very limited active dispersal from the places in which snails were dropped.There is no doubt that the species has spread rapidly in the city in the last twenty years, and the contrast with areas where C. nemoralis has been established for many years (often centuries) is marked (CAmeRon et al. 2009); it seems to hold even in a comparison with populations from Skåne in the extreme south of Sweden (Table 7).
There is, however, a much better historical record of C. nemoralis in Göteborg than in Sheffield and Wrocław, where there were very few samples available other than those made by CAmeRon et al. (2009) during their studies from 2005 to 2008.This record indicates that the history of C. nemoralis in the city is not a simple one of spread from some initial point of introduction, with variation among populations generated by founder effects and genetic drift.The earliest records are characterised by very high frequen-cies of yellow shells; these decline with time and in recently sampled populations the median value is less than 50%.Leaving aside possible scoring problems, there has either been selection against yellow sustained over a long period, or recent populations originate from sources other than those involved in the early samples.Temporal trends in the frequency of yellow shells across the geographical range of the species are unclear; silveRtown et al. ( 2011) did not find any significant changes over time over the whole range.Cook (2014), however, detected an increase in yellow in some habitats within Great Britain and some local studies using accurately relocated sites have also shown increases over the last 50 years (CaMerOn & COOk 2013, CaMerOn et al. 2013).OżgO & sChilthuizen (2011) showed an increase in yellow effectively unbanded.These changes have been attributed to rising mean temperatures over the period.The trend, where it has been detected, is thus in the opposite direction to that seen in Göteborg.As silveRtown et al. ( 2011) have shown, the frequency of yellow shells decreases northwards over the range of the species and the modern median frequency in the city is within the range expected, whereas the early samples are discordant.This does suggest a different and perhaps less distant source or sources for modern populations.
One feature common to C. nemoralis populations in all three cities is the lack of variation with habitat.While this may reflect in part the difficulties in classifying the habitats, it is at first sight surprising, given that a number of studies have shown such variation even among recently established populations or habitat types (OżgO 2011, OżgO & BOguCki 2011, CAmeRon et al. 2013).In Wrocław, shaded habitats were mostly rather transitory, but both Sheffield and Göteborg contain some ancient woodland.Again, it suggests that colonisation of some habitats is very recent, and that adjacent populations in different habitats may have different origins.The lack of any correlation between members of a pair of shaded and open habitats (usually less than 1 km apart) supports this suggestion.The evidence from the studies mentioned above suggests that if populations survive for several generations, the effects of selection ought to become apparent.
This study of a third set of urban C. nemoralis populations reinforces the case made by Cook (1998) that this variation in C. nemoralis was strongly influenced by migration and gene flow, and in particular that the leptokurtic transport of individuals (some travelling much greater distances than possible by active dispersal) helped to maintain the polymorphism, which is found in nearly all populations.Similar patterns are seen in more rural areas where the evidence suggests that the species has spread recently (CAmeRon et al. 2011, PokRyszko et al. 2012).As OżgO (2011) has also pointed out, the species is hermaphrodite, has multiple matings and long term allosperm storage.A single multiply-mated individual can carry a considerable proportion of the population's genetic variation at these loci, even when some alleles are at low frequency.Founder Effects do not have to result in monomorphic populations, but they can generate great variation among populations within a small area.Appendix Table 1b continued

Fig. 3 .
Fig. 3.The distribution of samples with brown shells (solid diamonds) and of those without brown shells (open diamonds)

Fig. 4 .
Fig. 4. The scatter of samples on the first two principal components of the PCA.The magnitude and direction of the relationship of each morph with the components is shown.B -brown

Table 1 .
Variation in C. nemoralis made before 1923.In this and later tables: P -pink, Y -yellow, 0 -unbanded, 3mid-banded, tri -00345 and variants, 5 -many banded with at least one of the top two bands (1 and 2) present

Table 2 .
The number of samples and shells, the percentage of yellow shells among them arranged by date and size (smallless than 10 shells, large -10 or more).The percentage of yellow shells in total is given for small samples, and the range and median values for sets of large samples.The last row represents samples scored in the fields (C -samples made by RhonA Cox and RobeRt & AlexAnDeR CAmeRon).The remainder, with the exception of the sample of Meeuse (1968) were scored from lots held in the Göteborg Natural History Museum

Table 6 .
Values of Moran's I for morphs and the first two PCA axes in each of four distance classes.Significant values of I are given in bold font; no adjustment for number of tests has been made

Table 7 .
Values of F ST estimated from morph frequencies and corrected for sampling error, together with the values estimated by bootstrapping and the associated standard deviations.Values of F ST for Sheffield and Wrocław are taken from CAmeRon et al. (2009).* -this figure relates to punctate shells, present in Sheffield but not elsewhere in any numbers.Trifasciate shells are similarly very rare in Sheffield.Data for Skåne, based on only eight samples, comes from unpublished material of R. Cox and R. CAmeRon

Table 8 .
The proportions (%) of samples visibly monomorphic at each locus considered in this study and in Sheffield and Wrocław (data from CAmeRon et al. 2009).All three indicates samples monomorphic for colour, presence or absence of bands and for midbanded

APPENDIX Appendix Table 1a .
Locations and scores of polymorphism for all samples of Cepaea nemoralis used in this study and held in the Göteborg Natural History Museum and that ofMeeuse (1968, asterisked)samples in the museum were scored by teD von PRosChwitz in 2014.Fuller details of locations and minor banding variants are available as a digital archive from the Göteborg Natural History Museum at www.gnm.se/cepaea.The samples are arranged by date.See text for shell colour and banding categories.Note that only samples of ten or more shells made from 1990 onwards were used in formal analyses.

Table 1b .
Details of location and morphs of Cepaea nemoralis for samples of 10 or more shells made in 2012 and 2013 and scored in the field.C: 1 -Cepaea hortensis present, 0 -absent; H: 1 -woods and shaded habitats, 2 -open habitats.See text for shell colour and banding categories.