• Marie-Laure Guillemin,

    Corresponding author
    1. Evolution et Génétique des Populations Marines (EGPM), UMR 7144 CNRS-Université Pierre et Marie Curie, Station Biologique de Roscoff, BP 74, Place Georges Teissier, 29680 Roscoff, Cedex, France. Laboratoire International Associé“Dispersal and Adaptation in Marine Species” (LIA DIAMS) PUC, Chile and CNRS-UPMC, France
    Search for more papers by this author
  • Sylvain Faugeron,

    1. Departamento de Ecología and Center for Advanced Studies in Ecology and Biodiversity, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, Casilla 114-D, Santiago, Chile. Laboratoire International Associé“Dispersal and Adaptation in Marine Species” (LIA DIAMS) PUC, Chile and CNRS-UPMC, France
    2. E-mail:
    Search for more papers by this author
  • Christophe Destombe,

    1. Evolution et Génétique des Populations Marines (EGPM), UMR 7144 CNRS-Université Pierre et Marie Curie, Station Biologique de Roscoff, BP 74, Place Georges Teissier, 29680 Roscoff, Cedex, France. Laboratoire International Associé“Dispersal and Adaptation in Marine Species” (LIA DIAMS) PUC, Chile and CNRS-UPMC, France
    2. E-mail:
    Search for more papers by this author
  • Frederique Viard,

    1. Evolution et Génétique des Populations Marines (EGPM), UMR 7144 CNRS-Université Pierre et Marie Curie, Station Biologique de Roscoff, BP 74, Place Georges Teissier, 29680 Roscoff, Cedex, France. Laboratoire International Associé“Dispersal and Adaptation in Marine Species” (LIA DIAMS) PUC, Chile and CNRS-UPMC, France
    2. E-mail:
    Search for more papers by this author
  • Juan A. Correa,

    1. Departamento de Ecología and Center for Advanced Studies in Ecology and Biodiversity, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, Casilla 114-D, Santiago, Chile. Laboratoire International Associé“Dispersal and Adaptation in Marine Species” (LIA DIAMS) PUC, Chile and CNRS-UPMC, France
    2. E-mail:
    Search for more papers by this author
  • Myriam Valero

    1. Evolution et Génétique des Populations Marines (EGPM), UMR 7144 CNRS-Université Pierre et Marie Curie, Station Biologique de Roscoff, BP 74, Place Georges Teissier, 29680 Roscoff, Cedex, France. Laboratoire International Associé“Dispersal and Adaptation in Marine Species” (LIA DIAMS) PUC, Chile and CNRS-UPMC, France
    2. E-mail:
    Search for more papers by this author

Current address: Instituto de Ecologia y Evolucion, Facultad de Ciencias, Universidad Austral de Chile, Casilla 567, Valdivia, Chile. E-mail:


The extent of changes in genetic diversity and life-history traits associated with farming was investigated in the haploid–diploid red alga, Gracilaria chilensis, cultivated in Chile. This alga belongs to one of the most frequently cultivated seaweed genera around the world. Fifteen farmed populations, 11 wild populations, and two subspontaneous populations were sampled along the Chilean coast. The frequency of reproductive versus vegetative individuals and of haploid versus diploid individuals was checked in each population. In addition, the distribution of genetic variation in wild and cultivated populations was analyzed using six microsatellite markers. Our results first demonstrated that farmed populations are maintained almost exclusively by vegetative propagation. Moreover, the predominance of diploid individuals in farms showed that farming practices had significantly modified life-history traits as compared to wild populations. Second, the expected reduction in genetic diversity due to a cultivation bottleneck and subsequent clonal propagation was detected in farms. Finally, our study suggested that cultural practices in the southern part of the country contributed to the spread of selected genotypes at a local scale. Altogether, these results document for the first time that involuntary selection could operate during the first step of domestication in a marine plant.

Recently, Duarte and collaborators (2007) highlighted the growing interest for “domestication of the oceans” as a possible way to face the Malthusian paradigm of producing food on land. Based on the needs of human society, and considering the diversity of renewable marine biotic resources still unexploited, it has been suggested that advances in domestication of the marine biota could be one of the next most important developments in human history (Duarte et al. 2007). However, domestication of seaweeds (i.e., macroalgae) is in its infancy compared to that of terrestrial plants. Domestication of land plants arose at the emergence of agriculture, about 13,000 years ago (Diamond 2002), whereas domestication of seaweeds is a contemporary phenomenon. Indeed, fewer than 20 species of macroalgae have been domesticated during the last two centuries, seven of which reached this status just in the past decade (Duarte et al. 2007). The expansion of seaweed farming depends on their domestication, which requires a good understanding of the life cycle of the target species to control reproduction and propagation.

Most of our knowledge about plant domestication comes from terrestrial species. In seed-propagated cultivated plants, the domestication process generally leads to differences between domesticated and wild individuals in traits such as seed dormancy, seed dispersal mechanisms, and growth habit. This set of changes in life-history traits is known as “the domestication syndrome” (Hammer 1984; Koinange et al. 1996). This condition usually is reflected in reduced fitness of domesticated individuals in their primary natural environment when compared to their wild counterparts (Clement 1999; Gepts 2004) and reduced genetic diversity due to bottleneck effects developed during the domestication process (Brush et al. 1995; Buckler et al. 2001). For example, domesticated plants propagated by seeds, such as maize, rice, and wheat, display two-thirds of the genetic diversity found in their wild relatives (Buckler et al. 2001). In comparison, more than one “domestication syndrome” may characterize different vegetatively propagated plants (Elias et al. 2007) depending on the part of the plant that is selected for. For example, in cassava, a plant asexually propagated by stem segments, the selection operated by farmers produced a complex set of changes including branching pattern (Jennings 1995), root yield, morphology of stems and of seedlings (Pujol et al. 2005b; Elias et al. 2007). In contrast, in cultivated grapevines, also clonally propagated, the selection for larger and sweeter berries has produced changes in the morphology of the plant (cane diameter, internode length, and leaf size), a shift from dioecy to hermaphrodism, and a reduction of seed viability (Olmo 2000; Zohary 2004).

Studies on evolution under domestication concentrate on plants propagated by seeds, including grasses (Eyre-Walker et al. 1998; Wang et al. 1999; Sun et al. 2001), common bean (Papa et al. 2005; Zizumbo-Villarreal et al. 2005), and sunflower (Tang and Knapp 2003; Burke et al. 2005), whereas vegetatively propagated crops have received less attention. This lack of interest is probably due to the belief that domestication of such organisms is a simple process consisting in controlling the vegetative propagation of the species (Zohary 2004). Yet, the mode of reproduction—sexual versus asexual reproduction—has major outcomes on genetic response to selection. When recombination occurs, selection associated with domestication should reduce diversity in restricted regions of the genome: those that contain genes controlling traits of human interest, as in the case of the teosinte branched 1 gene that controls tillering and apical dominance in maize (Wang et al. 1999). Conversely, as clonal reproduction mimics a complete physical linkage over the entire genome, intense selection acting on genes controlling traits of human interest should affect the whole genome. In addition, domestication involves repeated bottlenecks that may affect vegetatively propagated plants particularly significantly, by leading to the fixation of a single genotype. For example, modern European cultivated potatoes arose from a limited number of introductions, and genetic studies have revealed their relatively narrow gene pool (Provan et al. 1999). Thus, bottlenecks and selection act jointly to reduce the diversity of the entire genome. The limited genetic diversity of crops jeopardizes the potential for sustained genetic improvement over the long term and renders them more vulnerable to disease and pest epidemics. In contrast to sex and recombination that generate new resistance variants effective against rapidly evolving parasites (Hamilton 1980), vegetative propagation does not produce variable progeny that may escape infection. This risk was brought sharply into focus in the 1840s with the potato late blight catastrophic epidemy following the virtual eradication of effective resistance in European potato crop (Provan et al. 1999), and in 1970 with the outbreak of Southern corn leaf blight that was attributed to extensive use of a single genetic male sterility factor that, unfortunately, was genetically linked to disease susceptibility (Tanksley and McCouch 1997).

In this context, recent studies have pointed out the importance of traditional management practices to compensate for genetic erosion in vegetatively propagated crops (Brush et al. 1995; Pujol et al. 2005a). These studies indeed showed that fields of crop plants propagated by stems or by tubers, such as cassava, potatoes, and yams, remain genetically diverse when cultivated traditionally (Quiros et al. 1992; Elias et al. 2000, 2001; Pujol et al. 2005a; Scarcelli et al. 2006). This unexpected diversity in clonal plants was reportedly related to a mixed reproductive regime, in which products of sexual reproduction were regularly and voluntarily incorporated by farmers into the vegetatively propagated stands (Quiros et al. 1992; Elias et al. 2000; Pujol et al. 2005a; Scarcelli et al. 2006). Interestingly, these cultivated populations generally revealed a high level of heterozygosity (Elias et al. 2000; Pujol et al. 2005a; Scarcelli et al. 2006), apparently because vegetative propagation preserves heterozygous genotypes against segregation.

Seaweed aquaculture has been developed primarily in Asia, particularly in Japan, Korea, and China, with new countries, including Chile, building up important seaweed farming operations only recently. Several species are intensively cultivated for food and colloid production. Worldwide, more than 90% of farming activities are concentrated on four taxa (Zemke-White and Ohno 1999): the brown kelps Laminaria (kombu) and Undaria (wakame) and the red algae Porphyra (nori) and Gracilaria. Depending on the species, propagation by both asexual and sexual means (thallus fragmentation and cultivation from spores, respectively) is used in seaweed farming. Little is known concerning the relationship between natural and cultivated populations of marine macroalgae (van der Meer 1983). Furthermore, and in spite of the wide variety of life cycles found in seaweeds, limited attention has been paid to possible changes in life-history traits in relation to cultural practices, as well as to the consequences of using sexual versus asexual propagation for farming.

In Gracilaria chilensis, a mixture of sexual and vegetative reproduction occurs in natural populations. Like most red algae, G. chilensis has a complex haploid–diploid isomorphic life cycle (Fig. 1), where both haploid and diploid individuals consist of an erect system of cylindrical and branched thalli growing from a perennial holdfast attached to the substratum. In addition, haploid and diploid individuals are able to reproduce vegetatively by fragmentation of erect thalli which can recruit in soft bottom areas by sediment burial, yet are unable to form new holdfasts and thus to be reattached to hard substrata. Free-floating thalli detached from the holdfast grow indefinitely and propagate naturally. Contrary to hard substrata populations, spore recruitment in soft bottom populations is unlikely, and reproduction is mainly vegetative (Causey et al. 1946; Stokke 1957; Simonetti et al. 1970). Soft bottom populations form extended beds, which were intensively harvested along the Chilean coast for about 20 years, until the collapse of 1985 (Buschmann et al. 2001) likely caused by overexploitation (Santelices and Ugarte 1987; Vasquez and Westermeier 1993; Norambuena 1996). Subsequently, G. chilensis was intensively planted and farmed by taking cuttings of bigger fronds in sandy bays scattered among natural stands within its area of origin (i.e., between 30°S and 45°S in Chile, according to Bird et al. 1986) and introduced outside of its natural range of distribution in northern Chile. In the early 1980s, up to 500 farms were created to sustain the production of this agarophyte (Buschmann et al. 2001). At the beginning, no regulations controlled the transport of algal material across the country, and it was a common practice of artisan fishermen to establish new farms (A. Pizarro, pers. com.). Farming success relied on a straightforward vegetative propagation of thallus fragments planted in soft bottoms. This method is especially useful in intertidal mud flats, where farmers can easily shove the thalli into the sand or mud during low tides (Santelices and Doty 1989). Seaweeds are repeatedly harvested by hand, and the largest thalli are retained to be replanted.

Figure 1.

Haploid–diploid life cycle of G. chilensis. Meiosis occurs on the reproductive diploid individuals (tetrasporophyte) to produce haploid spores (tetraspores). When liberated, tetraspores attach to the substratum, develop perennial holdfasts, and grow into haploid dioecious gametophytic individuals (male and female in similar proportions). Male gametes are liberated in the water column whereas female gametes are retained on the female thallus. Fertilization occurs on the female individuals and the zygote develops as a third stage called a cystocarp, a structure growing directly on female gametophyte thallus and producing several thousands of diploid spores (carpospores) by repeated mitosis from the single zygote. When liberated, carpospores attached to the substratum develop into perennial holdfasts from which grows a tetrasporophyte. Vegetative propagation has only been reported in terms of thallus breakage that can occur both in haploid and diploid phases and generates subsequent reestablishment of thalli in the soft bottom habitat through natural embedding or human-assisted embedding. Solid lines: haploid–diploid sexual life cycle; dotted lines: putative asexual reproduction.

Overall, the process of domestication of G. chilensis, based exclusively on vegetative propagation, started 25 years ago and might have included some (voluntary or involuntary) selection for the highest growth rates. This model species offers the opportunity to investigate potential changes in genetic diversity and evolution of life-history traits (i.e., mating system and life cycle) associated with farming practices. To address this issue, we compared the frequency of reproductive versus vegetative individuals and the frequency of haploid versus diploid individuals between farmed and natural populations. In addition, based on six microsatellite markers (Guillemin et al. 2005), we analyzed the effects of cultivation on genetic variation, as well as the effect of stock management on the pattern of genetic differentiation between farmed and wild populations.

Material and Methods


A total of 2653 individuals from 11 natural, 15 farmed, and two subspontaneous populations (i.e., escaped from farmed populations) were sampled between March 2002 and February 2006 along the Chilean coast (Fig. 2). The distinction between farmed and natural populations was based on whether G. chilensis was actively planted or not. All but one of the populations located in sandy or muddy bays were managed by fishermen and were thus considered as farmed populations (farmed status was confirmed by historical records). Two types of farmed populations were recorded: artisanal farmed populations (ART, Table 1) corresponding to cultivated areas managed by numerous artisan fishermen without a concerted strategy, and large industrial farmed populations (IND, Table 1) operated by single companies with planned stand management. Castro (F-CAS) was the only soft bottom population apparently not actively planted, although no precise information could invalidate its classification. On the other hand, we considered all stands on hard substrata as natural populations; in such places recruitment is possible only through spore settlement. Finally, we recognized as subspontaneous populations those small stands composed of individuals growing on artificial hard substrata and located within farmed populations (SP-BAI and SP-CUV, Table 1). These two subspontaneous populations were analyzed separately as they likely originated directly from spores released by farmed individuals. Most sampled populations were located within the natural distribution range of G. chilensis (Fig. 2, Table 1). In addition, three farmed populations and one subspontaneous population were sampled at the northern limit of the natural distribution range (F-LHE, F-PCH) or further north (F-CMA, SP-BAI, Fig. 2, Table 1). In the localities of Lenga, Tubul, Ancud, Metri, and Raul Marin, farmed and natural populations were separated by distances of only a few hundred meters. Table 1 summarizes the information available for each study site.

Figure 2.

Sampling location of the 15 farmed populations (F-), the 11 natural populations (P-) and the two subspontaneous populations (SP-) sampled in eight regions along the Chilean coast. The names of the different regions are in italics.

Table 1.  Characteristics of the sampled farmed (F-), natural (P-), and subspontaneous (SP-) populations. Total number of samples = 2653 individuals.
RegionLocalityPopulations abbreviations*GeopositionSampling dateNo. of quadrats
  1. *IND, large industrial farmed populations; ART, small artisanal farmed populations.

  CalderaCultivos MarinosF-CMA IND27° 04′ S, 70° 50′ W01 August 20024
  CoquimboLa HerraduraF-LHE IND29° 58′ S, 71° 21′ W26 October 20024
Playa ChangasF-PCH IND29° 57′ S, 71° 20′ W25 October 20024
  ConcepcionLengaF-LEN IND36° 45′ S, 73° 11′ W03 July 20044
TubulF-TUB IND37° 15′ S, 73° 26′ W01 September 20044
  Seno ReloncaviPiedra AzulF-PIA ART41° 30′ S, 72° 47′ W04 December 20024
QuillapeF-QUI IND41° 32′ S, 72° 44′ W01 September 20024
ChaicaF-CHA ART41° 38′ S, 72° 39′ W24 March 20044
MetriF-MET ART41° 36′ S, 72° 42′ W24 March 20041
  Canal de ChacaoAncudF-ANC IND41° 52′ S, 73° 48′ W01 November 20021
  Chiloe IslandCuraco de VelezF-CUV ART42° 26′ S, 73° 36′ W30 October 20024
CastroF-CAS ART42° 30′ S, 73° 47′ W01 November 20024
QuiquelF-QQL ART42° 21′ S, 73° 35′ W25 March 20044
TenaunF-TEN ART42° 20′ S, 73° 23′ W25 March 20044
  Chiloe ContinentalRaul MarinF-RMA IND43° 46′ S, 72° 57′ W01 April 20034
  ConcepcionLengaP-LEN36° 45′ S, 73° 11′ W03 July 20041
DichatoP-DIC36° 32′ S, 72° 56′ W01 September 20043
TubulP-TUB37° 15′ S, 73° 26′ W01 September 20044
  ValdiviaMolinosP-MOL39° 50′ S, 73° 23′ W08 March 20042
NieblaP-NIE39° 52′ S, 73° 23′ W07 March 20043
  Seno ReloncaviMetriP-MET41° 36′ S, 72° 42′ W24 March 20041
  Channel of ChacaoMaullinP-MAU41° 37′ S, 73° 35′ W24 March 20043
AncudP-ANC41° 52′ S, 73° 48′ W01 November 20021
  Chiloe ContinentalHornopirenP-HOR41° 58′ S, 72° 28′ W28 February 20064
ChaitenP-CTE42° 55′ S, 72° 42′ W18 January 20063
Raul MarinP-RMA43° 46′ S, 72° 57′ W01 April 20032
  CalderaBahia InglesaSP-BAI27° 09′S, 70° 53′W15 March 20021
  Chiloe IslandCuraco de VelezSP-CUV42° 26′ S, 73° 36′ W30 October 20022

To test for the occurrence of population substructuring, we applied a hierarchical sampling scheme. Each region (separated by at least 100 km) was divided into localities (separated by five to 60 km). In the Chiloe continental region, localities were separated by a minimum of 130 km. In most localities, both farmed and natural populations separated by less than 1 km were studied when possible. One to four quadrats, separated by 10 to 500 m were randomly sampled within each population. The number of localities per region, the type of populations (either farmed or natural), and the number of quadrats per populations are given in Table 1. In natural populations, the number of quadrats and the number of individuals sampled per quadrat were merely restricted by the density of thalli: whenever possible, a minimum of 20 individuals (haploids and diploids) per quadrat were sampled. In farmed populations characterized by higher densities, 25 thallus fragments separated by at least 1 m from each other to avoid sampling fragments from the same thallus could be sampled in each quadrat. Because of these sampling differences, the size of a quadrat was roughly 1 m2 in natural populations but reach 16 m2 (4 m × 4 m) in farmed populations. Because most analyses were performed on data from diploid plants, only quadrats including at least seven diploid individuals (i.e., 14 alleles) were considered.


The frequency of reproductive individuals and sex and ploidy (i.e., male and female haploid gametophytes and diploid tetrasporophytes) ratio were determined by observations under a binocular microscope of the reproductive organs in all sampled sites except P-DIC and F-TUB. This method was the only one used in F-MET, F-RMA, P-MOL, P-NIE, and P-RMA. In the other 23 sampling sites, ploidy level of vegetative individuals was also determined using flow cytometry whenever fresh material was available (i.e., 12 sites), otherwise when only dried samples were available, it was deduced from genotypic analyses (i.e., 11 sites).

The method of flow cytometry can be used only on fresh individuals. Flow cytometry analysis was applied to 472 individuals following a protocol modified after Marie et al. (2000), which consisted in extracting the nuclei from fresh tissue in 300-μl extraction buffer (15 mM MgCl2, 60 mM trisodium citrate, 60 mM sorbitol, 27.5 mM Hepes, 2.5 mM EDTA, and 2.5 mM sodium bisulfite). A suspension of 100 μl of nuclei was then stained with 7.5 μl of SYBR Green 1× in 200 μl of extraction buffer for 3 min. Samples were analyzed by a FACSort™ Cell Sorter Flow Cytometer System (Becton Dickinson, San Jose, CA). A subsample of 10 gametophytes and 10 tetrasporophytes previously identified by their reproductive organs was used to calibrate the method.

For the genotypic analysis, an individual was considered as diploid if it was heterozygous at any of the six loci that were genotyped. Therefore, the frequency of diploids was underestimated because some rare diploids may be homozygous for all the six microsatellite loci. This genotyping method was thus used only to analyze 466 individuals, for which the two other methods could not be used.


DNA extraction followed Cohen et al. (2004) and PCR amplification of the six microsatellite loci and allele size scoring were performed according to Guillemin et al. (2005). Products were run on 6.5% polyacrylamide denaturing gels in a LI-COR DNA sequencer model 4200™ (LI-COR, Lincoln, NE). When a sample failed to amplify (less than 0.6% of the PCR reactions), one to three new amplifications were tried and genotype agreement was checked. Because most r estimate values of null alleles (according to Brookfield 1996, eq. 4) were negative or close to zero, we considered that the frequencies of null alleles could be neglected.


The frequencies of reproductive individuals and of haploid and diploid individuals were compared (1) between farmed and natural populations and (2) between seasons, by two-way analysis of variance (ANOVA) fixed models on rank-transformed data, using MINITAB ver. 13.2 (MiniTab Inc., State College, PA) followed by Tukey comparisons of means (Hayter 1984).


Following Goudet et al. (1994), a cumulative pooling method was used to determine the relevant scale defining a panmictic unit. Unbiased estimators of Wright's F-statistics were calculated for each pooling scale (i.e., quadrats, locality, region, and all samples). We used the term Fi. rather than Fis for these estimators to emphasize that the relevant panmictic unit is a priori unknown. Because farmers are manipulating the mating system and exchanging fragments of thallus among farmed populations, this method was only applied to natural populations. Furthermore, as the distances between localities were not comparable in the region of Chiloe Continental (see sampling design and Fig. 2), samples from this region were not used for estimating the Fi. values at pooling scales larger than “locality” (i.e., not used for regional and global pooling scales). Significance was tested by running 5000 permutations of alleles among individuals within samples using FSTAT ver. 2.9 (Goudet 2001).


Based on the multilocus microsatellite genotype, putative genets were identified using GenClone 1.0 (Arnaud-Haond and Belkhir 2007). Hereafter, the multilocus genotypes will be referred to as MLGH for MultiLocus Genotype of Haploids and MLGD for MultiLocus Genotype of Diploids. At each site, genotypic diversity was measured in two ways. First, the frequency of different genotypes was calculated as D=MLG/N, where MLG (MLGH or MLGD) is the number of distinct multilocus genotypes and N is the total number of studied individuals (Ellstrand and Roose 1987). Second, because two identical multilocus genotypes can either be the result of sampling the same clone or sampling two different genets originated from two distinct sexual reproduction events but sharing exactly the same alleles at all loci (Arnaud-Haond et al. 2005), the frequency of genets was assessed from GD=GENET/N, where GENET is an estimate of the number of different genets and N is the total number of sampled ramets. To estimate GENET, we first calculated Psex, which is the probability for a given multilocus genotype to be observed in N samples as a consequence of different sexual reproduction events (Tibayrenc et al. 1990)


where Pgen is the probability of observing a given multilocus genotype (MLGH or MLGD) under the null hypothesis of free recombination (product of the observed frequency of each single-locus genotype) and m, the number of individuals in the sample sharing this particular genotype. For MLGH, the probability Pgen was calculated using the observed frequency in the haploid subsample. When Psex was smaller than 0.05, we considered duplicated multilocus genotypes (either MLGH or MLGD) as ramets (or clones) of the same genet. When Psex was greater than 0.05, duplicated genotypes were considered as different genets.

To assess the connection between farmed populations and to examine whether the same genotype was used to seed different farms, an index of genotypic similarity (GS) between each pair of farms was calculated:


where MLGk is the number of copies of genotype k shared between farms i and j, and Ni and Nj are the total number of individuals sampled in each farm. To assess the pattern of thallus fragment exchange among farms of the same region (applicable to southern farms with the exception of F-CAS), the correlation between matrices of pairwise genotypic similarity (GS) and Ln (geographic distance) was tested with a Mantel test using the program ISOLDE (implemented in the software GENEPOP, Raymond and Rousset 1995).


Linkage disequilibrium (LD) was assessed using a single multilocus measurement of LD that is provided by the estimate inline image (Agapow and Burt 2001). This estimate is based on the association index (IA, Brown et al. 1980; Maynard-Smith et al. 1993) and was modified by Agapow and Burt (2001) to make it independent of the sample size. Multilocus LD was calculated for haploid and diploid subsamples in each population separately using the program MultiLocus1.2 (Agapow and Burt 2001). To test for departure from random associations between loci, the observed dataset was compared with 1000 simulated datasets in which sex and recombination were imposed by randomly reshuffling the alleles among individuals, independently for each locus (Agapow and Burt 2001). For the diploid dataset, the two alleles of the same locus were shuffled together to maintain associations between alleles within loci in the randomized dataset.


To test the effect of clonal propagation on genetic diversity and genetic structure, analyses were done (1) on all samples, disregarding if individuals belonged or not to the same genet, and (2) on the original dataset but including only a single copy of each genet. Genetic diversity in farmed and natural populations was assessed by computing the percent of polymorphic loci (P99%), the observed heterozygosity (HO), and the expected heterozygosity (HE) with the Genetix software (Belkhir et al. 2003). In addition, allelic richness (RS− 1) was estimated using the rarefaction procedure of Petit et al. (1998), which takes into account differences in sample size. This estimate was calculated in each population using the program Contrib (rarefaction size of N= 7, Petit et al. 1998). The Fis measurement of deviation from Hardy–Weinberg expectations was computed in diploid subsamples for farmed and natural populations with GENEPOP (Raymond and Rousset 1995) at the relevant sampling unit (defined by our previous hierarchical analyses). Fis was calculated for each locus and over all loci according to Weir and Cockerham (1984), and significance was tested by running 5000 permutations of alleles among individuals within samples using FSTAT ver. 2.9 (Goudet 2001).

Genetic differentiation among populations was analyzed by estimating FST (θ) (Weir and Cockerham 1984) at different spatial scales, and statistical significance of FST estimate values was assessed by running 5000 permutations of genotypes among samples, using FSTAT ver. 2.9 (Goudet 2001). For natural populations, isolation by distance was tested by a Mantel test on ln (geographical distance) and pairwise genetic distance (FST/(1 −FST)) using the program ISOLDE (GENEPOP, Raymond and Rousset 1995).


Because sexual reproduction regularly shuffles genes between the two phases of the life cycle, no differences in gene frequencies are expected between haploid and diploid subsamples of the same population. However, when sexual reproduction becomes sufficiently rare, the haploid and diploid phases turn into two isolated subpopulations that can diverge from each other under the effects of drift or selection. Differences in allele frequency between ploidy levels in each population were tested by exact tests on the distribution of allelic counts using GENEPOP (Raymond and Rousset 1995). Differences of gene diversity (HE) and allelic richness (RS− 1) between haploid and diploid subsamples were tested by the nonparametric Wilcoxon's sign rank test. In haploids, HE was corrected by a factor of (2n− 1)/(2n− 2) to avoid bias due to the different number of genes sampled between haploids and diploids (Engel et al. 2004). The allelic richness, RS− 1, was calculated for haploid subsamples using rarefaction for the same common sample size (N= 7) as for diploid subsamples (program Contrib, Petit et al. 1998).



Season has a significant effect on the number of reproductive individuals in farmed and natural populations, with more reproductive individuals observed in summer and autumn (Table 2A). On the average, the percentage of these individuals was higher in natural (71.8%± 27.4%) than in farmed populations (38.1%± 40.1%; Fig. 3A), but this difference was not significant (Table 2A). Indeed, considerable variation was observed among farmed populations, with percentages of reproductive individuals ranging from 0% (in F-CMA, F-LHE, F-PCA, F-PIA, and F-QUI), to 100% (in F-MET). Conversely, percentage of reproductive individuals never reached 100% in natural populations although it was always greater than 25% (Fig. 3A).

Table 2.  Two-way factorial ANOVA on frequency of reproductive individuals (A) and frequency of diploid individuals (B). Population type is defined as farmed or natural populations, and Season as winter/spring (low reproductive season) or summer/autumn (high reproductive season), the two factors are fixed.
 Population type 1 41.29 41.29 1.560.226
Season 1362.84362.8413.680.001
Population type 1 12.90 12.90 0.490.493
 × season 
Error20530.28 26.51
 Population type 1822.46822.4632.64<10−4
Season 1 11.55 11.55 0.460.506
Population type 1 43.21 43.21 1.710.204
 × season  
Error22554.39 25.20
Figure 3.

Percentage of reproductive individuals and ploidy ratio in farmed and natural populations. (A) Percentage of reproductive individuals in the 14 farmed populations (dark gray) and the 10 natural populations (white), as evidenced by the presence of reproductive structures on observed individuals. The means (horizontal lines) is surrounded by first and third quartiles (vertical bars) and the minimum and maximum values are indicated by diamonds. (B) Percentages of diploid individuals within populations determined by (a) observation under a binocular microscope, (b) flow cytometry, or (c) genotyping. The details of the methods used (i.e., a, b, and/or c) are given in parentheses for each population. The 15 studied farmed populations are noted in dark gray, the 11 natural populations in white and the two subpopulations escaped from farm in bright gray. In addition, the number of observed individuals appears in parentheses.

The percentage of diploid individuals was significantly higher in farmed (92%± 19%) than in natural populations (48%± 16%), regardless of the reproductive period (Table 2B). With the exception of F-LEN, all the farmed populations were composed mainly of diploid individuals (> 87%, Fig. 3B). In natural populations, the percentage of diploids was highly variable, ranging from 21 to 73% (Fig. 3B).

The relative abundance of reproductive individuals in the two subspontaneous populations was 16% in SP-BAI and 24% in SP-CUV, and the maximum percentage of diploids never reached more than 73% (Fig. 3B).


The cumulative pooling method showed an increase of heterozygote deficiency with the pooling units for natural populations (Fig. 4). Deficiency of heterozygotes, revealed by Fi., was low at quadrat and locality scales with values of 0.022 [−0.086 to 0.126, CI 95%; P= 0.979] and 0.041 [−0.066 to 0.137, CI 95%; P= 0.034], respectively (Fig. 4), but became highly significant when localities were pooled within regions (intra region level: Fi. value of 0.197 [0.066 − 0.347, CI 95%; P < 0.0002], Fig. 4). Consequently, localities can be considered as a relevant panmictic sampling unit for natural populations and all further analyses were performed at the locality scale for natural and farmed populations. Diploid subsamples at locality scale correspond to a mean sample size of 55.5 ± 14.7 and 37.3 ± 16.8 individuals for farmed and natural populations, respectively (Table 3).

Figure 4.

Multilocus unbiased estimates of Fi. calculated for each pooling scale in natural populations: quadrats, localities, regions, and all samples. For the regional pooling scale, individuals from the region of Chiloe Continental were not used because the sampling scheme between populations was not respected in this region. Bars indicate 95% CI calculated using bootstrapping over loci (1000 bootstraps), the P values were calculated using a permutation test (5000 permutations of alleles among individuals within samples, FSTAT version 2.9, Goudet 2001).

Table 3.  Estimation of the genetic diversity indices: percent of polymorphic loci (P99% criterion), allelic richness (RS− 1), expected heterozygosity (HE), and observed heterozygosity (HO). Two indices of genotypic diversity are given: the frequency of different genotypes (D) and the frequency of different genets (GD). N is the number of samples analyzed for each life-history phase (2n vs. n).
Farmed and/or natural population abbreviationsLife-history phaseNP99%RS− 1HEHODGD
  1. HE= Expected heterozygosity corrected by (2n− 1)/(2n− 2) for haploid subsamples, RS− 1 calculated using the program CONTRIB (Petit et al. 1998).

  F-LHE2n64 834.130.340.640.110.17
  F-LEN2n25 834.
n21  03.380.00 0.050.05
  F-CHA2n64 834.150.350.640.120.13
  F-MET2n28 674.140.340.650.180.18
  F-CAS2n56 333.720.
  F-RMA2n64 834.150.350.460.470.47
(Mean±SE for diploid subsample)55.47±14.7489±184.31±0.260.41± 0.110.61±0.180.28±0.200.31± 0.21
  P-LEN2n19 834.140.330.340.840.84
n 7 674.220.29 0.861.00
n31 834.230.36 0.611.00
n341004.650.54 0.850.88
  P-MOL2n49 673.970.270.310.310.47
n23 673.910.23 0.350.96
  P-NIE2n72 503.600.
n96 673.660.14 0.091.00
  P-MET2n41 674.110.330.400.660.83
n34 674.100.32 0.410.88
n381004.420.45 0.630.89
n14 834.420.42 0.861.00
  P-HOR2n34 674.
n69 674.100.32 0.301.00
  P-CTE2n12 673.770.160.190.330.33
n441003.720.16 0.230.27
  P-RMA2n47 673.990.280.240.430.57
n47 674.080.32 0.320.64
(Mean± SE for diploid subsample)37.27±16.7979±184.10±0.290.32±0.120.31± 0.110.64±0.280.79±0.23
(Mean±SE for haploid subsample)39.73±25.0078±154.14±0.300.32±0.12 0.50±0.270.86±0.22
  SP-BAI2n 7 834.410.400.570.570.57
n 7 834.350.35 0.571.00


A total of 35 alleles were observed over the 1266 diploid and 465 haploid genotyped individuals. Most loci showed moderate to low levels of polymorphism, with only three to eight alleles detected per locus. Allele frequencies for haploid and diploid subsamples from each locality are available on request. Of the 35 alleles detected, 18 occurred at a frequency lower than 0.01.

In all farmed populations, except for the diploid subsample of F-LEN, significant heterozygote excess was observed (all tests, P < 0.001) with significant negative Fis values ranging from −0.15 to −0.91 (Table 4). A deficiency of heterozygotes and a significant positive Fis value were detected in the only farm (F-LEN) with more haploid than diploid individuals (Tables 3 and 4). When clonal replicates were removed from the data, five Fis values were still significantly lower than zero, ranging from −0.09 to −0.52 (i.e., approximately half of those calculated when all samples were included in each farmed population; Table 4). Of the 11 natural populations, six showed significant Fis values with either excess or deficit of heterozygotes, ranging from −0.20 to 0.21 (Table 4). However, only three natural populations showed significant Fis values, all positive (between 0.18 and 0.22), after removing duplicate clones (Table 4).

Table 4.  Summary of statistical analyses for multilocus estimators of substructure. Fis and coefficient of multilocus linkage disequilibrium inline image calculated for each population, with and without the clonal duplicated genotypes.
Farm and/or natural population abbreviationsDiploid subsampleHaploid subsample
All individualsDuplicate clones removedAll individualsDuplicate clones removed
NFisinline imageNFisinline imageNinline imageNinline image
  1. ***P<0.001, **P<0.01, *P<0.05.

  F-CMA63−0.70*** 0.93*** 4 0.12 0.52**
  F-LHE64−0.88*** 0.15***11−0.52*** 0.05*
  F-PCH62−0.15** 0.36***25−0.07 0.14*
  F-LEN25 0.23** 0.09**19 0.05 0.0221 1 
  F-TUB64−0.17*** 0.06**44−0.09* 0.006
  F-PIA61−0.33*** 0.51***15−0.10 0.14**
  F-QUI60−0.51*** 0.40***18−0.21* 0.14**
  F-CHA64−0.82*** 0.36*** 9−0.50*** 0.39***
  F-MET28−0.91*** 0.36* 5−0.49* 0.06
  F-ANC29−0.34*** 0.42***14−0.19* 0.29***
  F-CUV64−0.65*** 0.41**10 0.13*** 0.16
  F-CAS56−0.70*** 0.60***14−0.41 0.29***
  F-QQL64−0.53*** 0.67*** 8−0.35** 0.43***
  F-TEN64−0.62*** 0.48***13−0.32** 0.17**
  F-RMA64−0.32*** 0.04*30−0.15*−0.006
  P-LEN19−0.02 0.0216−0.04 0.005 7 0.27* 7 0.27*
  P-DIC43 0.21*** 0.14***39 0.18* 0.14***31 0.0231 0.02
  P-TUB28 0.11−0.0228 0.11−0.0234−0.00930−0.02
  P-MOL49−0.13* 0.24***23 0.12 0.08**23−0.00922−0.03
  P-NIE72 0.00−0.0472 0.004−0.0496−0.0396−0.03
  P-MET41−0.20** 0.05*34−0.03−0.00934 0.00630−0.02
  P-MAU43−0.02−0.0243−0.02−0.0238 0.06**34 0.05*
  P-ANC22 0.18* 0.15***18 0.22* 0.12**14 0.0414 0.04
  P-HOR34 0.21** 0.0132 0.20* 0.00469−0.00369−0.003
  P-CTE12−0.14 0.36* 4 0.08 0.0244 0.29***12 0.03
  P-RMA47 0.13* 0.14***27 0.03 0.0347 0.08***30−0.01

All 15 farmed populations showed significant multilocus linkage disequilibria (LD) with values ranging from 0.04 to 0.93 (Table 4). Removing clonal replicates resulted in a decrease of inline image values (−0.006 <inline image < 0.52), but 10 of the farmed populations still retained significant multilocus LD (Table 4). In comparison, natural populations displayed relatively low values of inline image (−0.04 <inline image < 0.24), although six of 11 values were significant (Table 4). After removing clonal replicates, LD remained high and significant in only two natural populations. In addition, few inline image values were significant in haploid subsamples (four and two with and without clonal replicates respectively, Table 4).


None of the natural populations, except P-CTE (exact test, P < 0.0001), showed differences in allele frequencies between haploids and diploids (exact tests, P values ranging from 0.13 to 0.92). Moreover, expected heterozygosity (HE) and allelic richness (RS− 1) were not significantly different when comparing haploid and diploid subsamples from any of the 11 natural populations (correction for the different number of genes sampled between haploids and diploids was applied, bilateral Wilcoxon's signed rank test, n= 6, P > 0.07). On the contrary, allelic frequencies were significantly different between haploid and diploid subsamples of the farmed population F-LEN (exact tests, P < 0.0001), and values of HE and RS− 1 significantly higher in diploid subsamples than in haploids were observed (bilateral Wilcoxon's signed rank test, n= 6, P= 0.04 for the two tests).


Because of the low frequency of haploid individuals encountered in farmed populations, we used diploid subsamples to compare the genetic diversity between farmed and natural populations. Mean expected heterozygosity and allelic diversity were higher in farmed than in natural populations (Table 3) but differences were not significant (Mann–Whitney bilateral tests for RS− 1: P= 0.069) or only marginally significant (for HE: P= 0.053). In contrast, observed heterozygote frequencies were significantly higher in farmed (HO= 0.61 ± 0.18) than in natural populations (HO= 0.31 ± 0.11; Mann–Whitney bilateral tests P= 0.0004).

A total of 334 different MLGD and 92 different MLGH were detected for diploid and haploid individuals, respectively. Most multilocus genotypes were not shared among populations (private MLGD = 80% and private MLGH = 66%) and were unique (i.e., observed only once: 67% and 52%, respectively). MLGD genotypic diversity was more than two times higher in natural than in farmed populations (D= 0.6 ± 0.3 and 0.3 ± 0.2, respectively, Mann–Whitney bilateral tests z=−3.04, P= 0.001) ranging from 0.06 to 0.69 in farmed populations and from 0.17 to 1.00 in natural populations (Table 3). MLGH genotypic diversity varied from 0.09 to 0.86 in natural populations (D= 0.5 ± 0.3, Table 3). In F-LEN, the only farm in which a sufficient haploid subsample could be sampled, one private MLGH was fixed (D= 0.05, clone MLGH N°73, see online Supplementary material).

Genotypic diversity (D) varied greatly among regions, with the lowest values observed in the two natural populations from Valdivia (P-MOL and P-NIE, Table 3), and the highest values in Concepcion and Canal de Chacao (Table 3). On the basis of the MLGDs observed, it is likely that the individuals collected in the two subspontaneous populations SP-BAI and SP-CUV were seeded by individuals from the neighboring farmed populations. Indeed, only one to three recombination events are needed to achieve the MLGDs observed in SP-BAI and SP-CUV from the MLGDs observed within adjacent farmed populations (respectively F-CMA and F-CUV) (data not shown).

Among the 91 repeated MLGD and 35 repeated MLGH, 67 (74%) and nine (26%), respectively could be regarded as clones of the same genets. Considering the diploid subsamples, the majority of clonal individuals were encountered in farmed populations (67%). Moreover, the number of clonal individuals per genet tended to be higher in farmed than in natural populations: a maximum of 13 repeated clones of the same genetic individual were detected in natural populations (clone N°65 in P-RMA, see online Supplementary material) whereas more than 50 repeats of the same genet were observed in farmed populations (clone N°328 in F-CMA and F-LHE, see online Supplementary material).

In farmed populations, approximately 90% of the 426 genotypes were private (not shared among populations), but they represented only 40% of the sampled individuals (Fig. 5). To analyze the pattern of genotypic diversity among farmed populations, for each of them we calculated the frequency of private genotypes (whether they were unique or repeated clones of the same haploid or diploid genets) and the frequency of each of the 18 different genotypes shared among at least two farms (Fig. 5). Figure 5 shows clearly that two genets, N°328 and N°105, were very common across the farms. The genet N°328 was detected in 129 of the samples but only from the northern region (F-CMA, F-LHE, F-PCH, and the small subspontaneous population SP-BAI, Fig. 5). Conversely, the genet N°105 was recorded in 206 samples, all from southern farmed populations, with the exception of F-CAS (Fig. 5), and also in two of the southern natural populations (P-MET and SP-CUV, see online Supplementary material). These two widespread genets were heterozygous at four of the six study loci but are not closely related because they shared only six alleles. The only haploid subsample that could be studied from a farmed population (F-LEN) revealed the occurrence of a single genet shared over the 21 sampled individuals (genet N°73, Fig. 5). In a previous sampling of this farmed population 18 months earlier (January 2003), all the individuals genotyped (n= 18) were this same genet N°73 (data not shown). Mantel tests revealed a significant decrease in genotypic similarity with increasing geographical distance among southern farmed populations (Fig. 6, P= 0.028).

Figure 5.

Relative abundance of shared and private multilocus genotypes (MLGD and MLGH). The relative abundance is given over all the individuals genotyped in farmed populations. The numeration of the 18 genotypes shared among farmed populations indicated below the graph is the same as the one used in the text and in the online Supplementary material (in F-LEN only one private MLGH was encountered = MLGH No. 73). N, number of sampled individuals.

Figure 6.

Genotypic similarity (GS) index as a function of geographical distance in southern farmed populations.


An overall high and significant level of genetic differentiation among populations was found for both wild and cultivated stands analyzed separately (FST= 0.38 and 0.27, respectively, Table 5). Even within a region, the genetic differences among natural populations remained significant and very high (FST= 0.21 in Valdivia; FST= 0.32 in Concepcion; FST= 0.40 in Canal de Chacao; and FST= 0.45 in Chiloe Continental, all tests being significant at P < 0.0002). The genetic structure among farmed populations within a region varied from FST= 0.12 to FST= 0.22 in Coquimbo and Chiloe Island, respectively. The FST values estimated after removing the duplicate clones between the farmed populations and their nearby natural populations were low (FST < 0.06 in five of the seven pairs of populations, Table 5) and generally not significant. Raul Marin Balmaceda and Ancud were the only localities in which the genetic differentiation between farmed population and its associated natural population remained significant (Table 5).

Table 5. FST calculated for the various types of populations. Genetic differentiation between a farmed and its associated natural population was calculated for seven localities in which the two types of populations were present.
FSTAll individualsDuplicate clones removed
  1. NS, not significant.

  2. *P < 0.01, **P < 0.001, ***P < 0.0001).

  3. F-CMA, F-LHE, F-PCH.


All farmed populations0.27*** 0.25***
Northern farmed populations0.08*** 0.09***
Southern farmed populations§0.16*** 0.13***
All natural populations0.38*** 0.37***
F-TUB/P-TUB0.02** 0.004NS
F-LEN/P-LEN0.01NS 0.004NS
F-ANC/P-ANC0.11*** 0.06*
F-RMA/P-RMA0.09*** 0.04**

No significant pattern of isolation by distance was detected for natural populations (P= 0.48, data not shown). For farmed populations, a significant isolation by distance was detected when all farms were taken into account (P < 0.0001), a pattern that might be due to the genetic differences between northern and southern farms. The test was indeed not significant when performed within each group separately (P= 0.30 for southern farms and P= 0.34 for northern farms).


Our results demonstrated that farmed populations of G. chilensis are maintained mainly by asexual propagation, as indicated by the ploidy ratio and the population genetic structure. The predominance of diploid individuals in farmed populations suggests that cultivation significantly modified important life-history traits of the alga. In addition, we established that strong reduction in genetic diversity occurred in farmed populations and we propose that involuntary selection could operate during the ongoing first step of the domestication process. These consequences of farming practices and their evolutionary implications are discussed in turn below.


Although our results showed that on average 40% of farmed individuals of G. chilensis produce reproductive structures, the genetic analyses clearly showed that recruitment of individuals resulting from sexual reproduction (i.e., settlement of spores) is very infrequent in farmed populations. This result is congruent with the fact that cultivation takes place in sandy bays or muddy estuaries, characterized by soft and unstable substrata that prevent normal spore settlement and germination, as compared with the rocky bottoms where natural populations normally occur (as reported for different Gracilaria sp.: Causey et al. 1946; Stokke 1957; Simonetti et al. 1970). These findings strongly suggest that the life cycle is not completed under farming conditions. A rapid loss of fertility has been reported in clonally propagated vascular plants, especially when selection for sexual reproduction is relaxed and when crops are cultivated for their vegetative parts (Zohary 2004). For example, potato (hermaphroditic) plants that are exclusively cultivated for their tubers are notorious for sterility (Ross 1986). In his study on the flowering behavior and male/female fertility in 676 accessions of cultivated potatoes, Gopal (1993) found that 20% were entirely sterile and 25% were sterile for the male or female function. In the case of G. chilensis, accurate comparisons of spore quantity and quality between individuals from natural and farmed populations should be obtained to test for the occurrence of reproductive vestigialization. In this species, it has been hypothesized that selection for vegetative growth favors sterility as maturation of thalli diverts resources from growth (Guimaraes et al. 1999). However, our results do not show a significant difference in the average frequency of reproductive individuals between farmed and wild populations, and more data are needed to test the effect of clonal propagation by cuttings on the investment in sexual reproduction in farmed populations. As this species has only been cultivated since 1985 (Buschmann et al. 2001), we can also hypothesize that farmed individuals are currently evolving toward a functional sterility, and that the complete loss of mature reproductive organs could be a subsequent step of the domestication process not yet achieved.

Detection of asexual propagation can be difficult to discern as infrequent events of recombination can remove signatures of clonality (Tibayrenc et al. 1990; Balloux et al. 2003; Bengtsson 2003; de Meeus and Balloux 2004; Halkett et al. 2005b). An interesting theoretical prediction of the models is that clonal reproduction generates massive heterozygote excess (thus largely negative Fis values) through the accumulation of mutations in the clonal lineages, because the two alleles at each locus irreversibly diverge within individuals (Balloux et al. 2003; Bengtsson 2003; Halkett et al. 2005b). In the case of Gracilaria, clonal propagation has occurred in farmed populations only within a few decades, but has resulted in massive replication of clonal individuals that can cover hectares of high-density biomass. This is equivalent of a very high number of generations of clonal propagation for each single genotype, which could thereby have allowed accumulation of numerous mutations in relatively few years of farming (see the second part of the discussion below). Moreover, the combination of rare sex events and strong selection against inbred seedlings has been recently proposed as an additional explanation by Pujol et al. (2005a) for the accumulation of heterozygotes in mainly asexually propagated crops. Pujol et al. (2005a), analyzing the heterozygote excess in cassava farming, found a significant positive correlation between the level of heterozygosity at neutral markers and the size of the cultivated plants. In cassava farming, a regular but low rate of sexual reproduction was observed and expected to produce a large proportion of homozygous genotypes due to strong inbreeding. However, because homozygotes individuals were smaller than heterozygotes, only new heterozygote genotypes were retained in the farmed populations. Whatever the mechanism involved, relatively few studies have validated this prediction of increased heterozygosity with clonal propagation by a strict comparison of sexual and asexual populations within the same species (but see in aphids: Simon et al. 1999; Delmotte et al. 2002; Halkett et al. 2005a; and in plants van der Hulst et al. 2000; Dorken and Eckert 2001; Eckert et al. 2003). In aphids, studies have further shown that heterozygote excess was inversely proportional to the frequency of sexual reproduction (Simon et al. 1999; Delmotte et al. 2002; Vorburger et al. 2003; Halkett et al. 2005a). In this context, our study represents a significant contribution because it broadens the range of organisms to which the predictions apply. In G. chilensis farmed populations, high linkage disequilibria, the occurrence of numerous repeated clones and severe heterozygote excesses were detected with significant negative Fis values. Interestingly, the clonal propagation occurring in the haplo–diploid species G. chilensis also led to an uncoupling of haploid and diploid phases in the only farm where this analysis could be done. None of the natural populations showed such a pattern: the two phases were indeed regularly connected by sexual reproduction. Differences in allele frequency between haploid gametophytes and diploid tetrasporophytes have also been observed in another red alga, Gelidium arbuscula reproducing mainly asexually (Sosa et al. 1998). On the other hand, the occurrence of sex, even occasionally, can be invoked to explain the lack of significant differences between haploid and diploid subpopulations found in Gracilaria gracilis (Wattier et al. 1997; Engel et al. 2004), Macrocytis pyrifera (Coyer et al. 1994), Cladophoropsis membranacea (van der Strate et al. 2002) and Gelidium canariensis (Sosa and Garcia-Reina 1993).

Furthermore, our results support the idea that farming practices could have involuntarily selected for diploid strains. As this diploid dominance exists in the industrial and artisanal farmed populations and over the whole geographical range of Gracilaria cultivation, it seems to be directly linked to human-assisted embedding. This may be due either to the fact that one phase (the diploid one in our case) is better at fragmenting than the other, or to the fact that diploids were preferentially propagated by farmers because of a heterosis effect (advantage in growth over haploids for example). We do not have any argument to support the first hypothesis, although in vegetatively propagated crops, “cultivars” are, as a rule, highly heterozygous, suggesting the occurrence of heterosis (Zohary 2004). Although we did not experimentally test for differences in fitness between haploids and diploids, several lines of evidence support the hypothesis that diploidy could have been indirectly selected because of heterosis. First, we established that diploid rather than haploid thalli have been retained in all but one farmed population. Second, most of the diploid genotypes retained in farmed populations are heterozygous, a situation validated by the generalized heterozygosity excess detected in all farmed populations except F-LEN and which persisted in 12 of these farmed populations even when clonal replicates were removed. Finally, two predominant multilocus genotypes shared among farmed populations, one in southern Chile (native area) and the other in northern Chile (introduced area), were found to be heterozygous for four of the six loci studied. We hypothesized that the dominance of heterozygous diploids in the farmed populations is due to, on the one hand, the accumulation of mutation in clonal lineages (see below) and on the other hand to heterosis. These findings add to the debate about conscious or unconscious selection by farmers (Gepts 2004). Current techniques for producing new cultivars are sophisticated, generally based on a strong knowledge of the genetic determinism of the traits of interest. The conscious selection hypothesis suggests that early farmers in the Neolithic were also aware of the biology and of the life cycle of plants before cultivating them, whereas the unconscious selection hypothesis advocates that the first farmers could not have set out to specifically select for these considerable changes. Hillman and Davies (1990) have suggested that a combination of unconscious and conscious selection may have operated in succession, with the former taking place during the early stages of domestication, when changes were too small to be noticeable. In this context, it seems that the recent domestication of G. chilensis might be the result of unconscious selection resulting from a single-step process of controlling vegetative propagation (Zohary 2004), rather than benefiting from the available knowledge on the biology of the species and from the theoretical bases of the genetics of quantitative traits.

The excess of diploids associated with farming practices may have important consequences on the evolution of the haploid–diploid life cycle. Indeed, evolutionary theory predicts that such complex life cycles can persist only when ecological differences between alternate phases exist (for review see Valero et al. 1992; Mable and Otto 1998). A synthesis of the functional properties of the isomorphic biphasic algal life cycle has been recently published by Thornber (2006). It follows that different phases of the same species can vary in abundance, in demographic parameters such as mortality and fecundity, in their physiology, and in their resistance to herbivory. Demographic models predict that in a dioecious obligately sexually reproducing species with the two phases being ecologically equivalent (equal per capita demographic rates), a diploid frequency of 0.41 is expected at equilibrium (Destombe et al. 1989; equivalent to the square root haploid:diploid ratio computed by Thornber and Gaines 2004). However, this frequency may shift if there are differences in “fitness” among haploid and diploid stages. For example, in G. gracilis, demographic surveys established that the frequency of diploids was significantly higher than 0.41, suggesting a twofold advantage in viability of diploids (Destombe et al. 1989). In this study, the frequency of diploids in wild populations was highly variable (0.21 to 0.73), although a clear dominance of either haploid or diploid individuals was not observed. Such fluctuations in ploidy ratios among populations have been frequently documented but specific experimental tests of adaptive differences between isomorphic generations are required to unravel the causes underlying the patterns of phase dominance (for a review see Fierst et al. 2005). On the other hand, our study revealed a sharp contrast among farmed and wild populations regarding the ploidy ratio, with quasi fixation of diploid individuals within farmed populations (> 0.87 in all farms but F-LEN). This diploid dominance highlights the absence of coupling between the two phases, a likely consequence of clonality as the predominant mechanism for stock propagation in farms.


Because domestication typically induces genetic erosion in cultivated crops (Brush et al. 1995; Buckler et al. 2001), we expected cultivated populations of G. chilensis to be less genetically variable than the natural stands. Indeed, diversity in terms of the number of genotypes was significantly lower in cultivated populations, which retained only one-third of the genotypic diversity recorded in wild populations. This result is in agreement with other research that has compared wild and cultivated populations of seaweeds. These studies were carried out on Porphyra yezoensis, one of the most economically important red algae known as “nori” (Miura et al. 1979; Niwa and Aruga 2006), Undaria pinnatifida or “wakame,” the worldwide cultivated and invasive brown alga (Huh and Huh 2002; Voisin et al. 2005), and the green sea lettuce Ulva prolifera, used as a side dish in Korea (Huh et al. 2004). All these studies have concluded that domestication led to an erosion of the genetic diversity in farmed populations.

However, because of the heterozygote excess in farmed populations (see above), gene diversity was not different between wild and cultivated populations. This result highlights the importance of analyzing multilocus information in addition to single locus genetic parameters. Despite the typically low level of genotypic diversity in farmed populations, a substantial variation does exist, with some farms much more diverse than others. Some of the sampled farmed populations (F-TUB in Concepcion and F-RMA in Raul Marin) showed a surprisingly high level of genotypic diversity and were characterized by a large number of unique and private genotypes (50% in F-TUB and 30% in F-RMA). This result suggests that these farmed populations were seeded by large numbers of different genotypes and that the “domestication bottleneck” was less effective. The alternative explanation that new recombinant genotypes came from sexual reproduction is untenable because haploid individuals were not observed in those farmed populations.

Whether farmed or wild, G. chilensis exhibits a relatively low level of genetic diversity (e.g., mean number of alleles lower than 5 and HE lower than 0.4). To what extent is this result associated with the collapse of wild stands in the 80s due to overexploitation (Santelices and Ugarte 1987; Vasquez and Westermeier 1993; Norambuena 1996)? A comparison with the New Zealand wild populations of the same species (Cohen et al. 2004) that have never experienced such a bottleneck (Nelson 1987) is a promising perspective to measure the effect of overexploitation on the pattern of genetic diversity in this species. Nearly half of the total number of alleles (16 out of 35) were observed at very low frequencies and restricted to a single population, indicating that they probably appeared by mutations within that population. Interestingly, most of these rare putative mutation events (11 out of 16) were observed in farmed populations. In each farmed population, the multilocus genotype that carried such a private rare allele differed from the most frequent multilocus genotype by only this allele. It is likely that this pattern reflects the accumulation of somatic mutations, a factor that allows an occasional increase of intravarietal diversity in clonally propagated crops (e.g., in cassava, Elias et al. 2000). On the other hand, the mutation load accumulated in clonal lineages might become a problem for the maintenance of selected strains of agronomic value (Meneses and Santelices 1999).

Domestication affects not only the pattern of genetic diversity but also the level of differentiation among farmed populations. In G. chilensis, FST values among farmed populations were markedly lower than those among natural populations. It is likely that farming practices have enhanced the homogeneity among farms, mainly as a result of inoculate exchange. Such activities can efficiently homogenize geographically separated crops and, at the same time, maintain genetic polymorphism within farmed populations (Elias et al. 2000). In fact, two groups of farmed populations were clearly distinguished: a northern group that comprises the three northern industrial farms and a large group with all the southern farms located in the region of Chiloe. No genotypes were shared between these two groups of farms and only a few dominant and widespread genotypes were observed within each group. We were not able to find natural populations of G. chilensis in the region of Coquimbo, despite the fact that some studies reported the occurrence of natural stands extensively collected between 1965 and 1985 in the area (Santelices and Ugarte 1987). Moreover, G. chilensis is absent between Coquimbo bay (F-LHE, F-PCH) and Concepcion bay (P-DIC), whereas it is present in every bay and estuary south of Concepcion. These findings suggest that the highly differentiated genotypic composition of the northern farms may reflect the existence of ancient and isolated natural populations in the northern region that now seem to have totally disappeared in favor of the current farmed populations probably established using local genetic material.

The existence of few widespread genotypes in southern farmed populations is likely due to exchanges accompanied by homogeneous selection within farms, rather than seeding from a single source. The existence of widespread genotypes due to transport and homogeneous selection in different sites has been documented in other clonally propagated plants (e.g., transplanted populations of the aquatic plant Zostera marina: Reusch 2001; in potato, Brush et al. 1995 and cassava Elias et al. 2000). This artisanal farming practice ensured the maintenance of substantial levels of genotypic diversity within southern farmed populations and allowed the spread of selected genotypes at local scale. In contrast, industrial farming practices, more developed in northern Chile, led to a drastic loss of genotypic diversity and selection of a single dominant genotype (No. 328).

Another possible cause of the higher genotypic diversity observed in southern artisanal farmed populations could be the result of gene exchanges with adjacent natural stands. Indeed, when farmed and natural populations were sampled at the same locality, the two types of populations shared the same pool of alleles exhibiting relatively low levels of genetic differentiation, but with farmed populations maintaining only a small part of the diversity present in the wild adjacent population. Even though the question of the direction of gene flow between wild and farmed populations remains open, the phenomenon of gene introgression between farmed and adjacent wild populations can be critical for the sustainable management of genetic resources in cultivated species. For example, the spread of cultivated strains within natural populations of U. pinnatifida has been suggested to disturb the native genetic diversity (Uwai et al. 2006). On the other hand, traditional farming management that combines both sexual and asexual reproduction by the regular intentional incorporation of plants originating from volunteer seedlings into the pool of clonally multiplied genotypes is crucial in preserving the genetic diversity in clonally propagated crops (potatoes: Quiros et al. 1992; cassava: Elias et al. 2001; Pujol et al. 2005a; and yams: Scarcelli et al. 2006). In G. chilensis however, incorporation of plants resulting from sexual reproduction seems not to be achieved even in artisanal farmed populations. Because of this, and considering the very narrow genetic basis upon which domestication was achieved, the cultivation process may have led to a drastic reduction of genetic and genotypic variability in G. chilensis farmed populations, and consequently to a higher susceptibility of these homogeneous clonal populations to parasite attacks.

Indeed, Leonardi et al. (2006) reported that G. chilensis farmed populations became rapidly affected by highly prevalent epiphytes and endophytes, reducing the yield and quality of the production. This raises the question of the role of farming practices on the increase of parasitism. Because most farms are dominated by a single or a few genotypes, a pathogen that has overcome host reaction defenses on a single thallus is able to infect most of the farmed population. In addition, the exclusive use of vegetative propagation during human-assisted embedding probably favors vertical transmission of epiphytes. These epidemiological aspects were ignored in the design of farming practices and thus can be considered as an unexpected consequence of domestication of G. chilensis. Future prospects to improve the agronomic value of the domesticated strains should consider epiphyte resistance as relevant characters.

In conclusion, this study suggests that strong selection could have occurred at the beginning of domestication of the red alga G. chilensis, leading to dramatic changes in the cultivated stands. We also suggest that the cultivation process has already resulted in functional sterility, as revealed by the absence of coupling between haploid and diploid phases in farmed populations, although mature reproductive organs are still observed in farms. Finally, and considering the narrow genetic basis upon which domestication is conducted in this alga, G. chilensis farmed populations could be threatened in the near future by a drastic reduction of their genotypic variability.

Associate Editor: J. Shykoff


We are grateful to A. Pizarro, O. Alcalde, and R. Westermeier for giving access to their farms and culture facilities, to E. Martínez and J. Morales for help in field sampling, to D. Marie for his help in flow cytometry, and to C. Engel, D. Roze, I. Olivieri, A. Mann, and S. Krueger for constructive comments and for improving the English. We would like to thank D. McKey and an anonymous reviewer for their detailed suggestions and comments improving the manuscript. Thanks to the associate editor J. Shykoff for her diligence in the review process. This study was financed by the European Commission INCO-DEV Programme (INCO-EPIFIGHT ICA4-CT-2001–10021), the Fondation BETTENCOURT SCHUELLER, “COUP d'ELAN à la Recherche 2001,” the French Embassy in Chile and research grant FONDAP-FONDECYT 1501 0001-program 7. MLG was supported by a CONICYT grant during her stay in Chile.