Author for correspondence: Clara de Vega Tel: +34 954557052 Fax: +34 954557059 Email: email@example.com
• Speciation via race formation is an important evolutionary process in parasites, producing changes that favour their development on particular host species. Here, the holoparasitic plant Cytinus, which has diverse host species in the family Cistaceae, has been used to study the occurrence of such races.
• Amplified fragment length polymorphism (AFLP) analyses were performed on 174 individuals of 22 populations parasitizing 10 Cistaceae species in the Western Mediterranean basin.
• Neighbour-joining, multivariate ordination analyses, and individual-based Bayesian analyses, clustered Cytinus populations into five well-characterized genetic races that, overall, agreed with the taxonomic sections of their hosts. In the AMOVA, among-races differences accounted for almost 50% of the genetic variation. The isolation-by-distance model was not supported by a Mantel test among Cytinus populations (r = 0.012; P = 0.456). All races showed low within-population genetic diversity, probably as a result of restricted pollen flow aggravated by flowering asynchrony, restricted seed dispersion, or stochastic processes.
• The genetic differentiation among the five races of Cytinus is congruent with the view that these races are well-characterized lineages that have evolved independently as a result of selective pressures imposed by their hosts. This pattern, with genetically distinctive groups associated with the infrageneric sections of the host species, has not been reported previously for parasitic angiosperms.
Speciation in parasitic plants occurs somewhat differently than in other angiosperms. In addition to systems of reproduction and dispersion, genetic drift and mutations, interactions of parasitic plants with their hosts are a significant selective force in speciation (Zuber & Widmer, 2000; Jerome & Ford, 2002; Linhart etal., 2003). Host–parasite associations are generally not random and can be quite stable, indicating a reasonably long evolutionary history (Hoberg, 1997; Olivier etal., 1998). When a parasite interacts with several host species over a long period of time, high genetic similarity may appear between populations of the parasite on the same host (Linhart etal., 2003) because of genetic changes that favour its development on that particular host species (Norton & Carpenter, 1998). In this way, parasites may evolve different races on different host species that could be considered incipient species (Glazner etal., 1988). Speciation via adaptation to different hosts is therefore an important process in parasite evolution (e.g. Olivier etal., 1998; Jerome & Ford, 2002; Diegisser etal., 2006).
The genus Cytinus (Cytinaceae) is an example of a parasitic plant closely linked to its hosts by virtue of being a holoparasitic endophyte. This genus was long considered to belong to the family Rafflesiaceae, but differences in the morphology of flowers, ovaries and seeds, together with recent data from phylogenetic studies, support recognition of Cytinus as belonging to a separate family, Cytinaceae, order Malvales (Bouman & Meijer, 1994; Nickrent et al., 2004). Cytinus comprises some eight species (Mabberley, 1997) that occur as holoparasites on the roots of diverse plant families. All Cytinus species show a marked reduction of morphological characters, with scale-like leaves and lack of external roots. The vegetative body is reduced to an endophytic system, often compared to that of a fungal plectenchyme (de Vega et al., 2007); only in the blooming period are the plants visible, with inflorescences bursting through the host root tissues.
Cytinus has one centre of diversification in the Mediterranean and Macaronesia, where the species are monoecious and parasitize roots of Cistaceae (Maire, 1961; Webb, 1964), and another centre in South Africa and Madagascar, where plants are dioecious and parasitize roots of Asteraceae, Hamamelidaceae, Rhamnaceae, Rosaceae and Rutaceae (Visser, 1981; Goldblatt & Manning, 2000). Recent taxonomic treatments have recognized either one or two species of Cytinus in the Mediterranean area. Greuter et al. (1989) adopted the former approach, recognizing the species Cytinus hypocistis (L.) L. with four subspecies: subsp. hypocistis, subsp. macranthus Wettst., subsp. lutescens, and subsp. clusii Nyman. The differences among these subspecies are very subtle because of their reduced morphological features, and lie mainly in flower colour and in the identity of the host parasitized: subsp. hypocistis (with small, ochre to light-yellow flowers) parasitizes white-flowered Cistus species; subsp. clusii Nyman (with ivory-white to pale pinkish flowers) parasitizes pink-flowered Cistus; subsp. macranthus Wettst. (with intensely yellow, largish flowers) is found on Halimium species; and subsp. lutescens (Batt.) Maire (with small, bright yellow flowers) is found on diverse species of Fumana, Helianthemun and Halimium. The authors of recent floristic studies in the Iberian Peninsula (Villar, 1997; Villar & López-Sáez, 2002) have followed Webb (1964) in recognizing two species, Cytinus hypocistis (L.) L. and Cytinus ruber (Fourr.) Fritsch (=Cytinus hypocistis subsp. clusii). In this view, C. ruber has ivory-white to pale pinkish flowers, and parasitizes pink-flowered Cistus species, whilst C. hypocistis has two subspecies, subsp. hypocistis, with small, yellow flowers, and subsp. macranthus, with larger, bright yellow flowers; both of these subspecies are reported to parasitize white-flowered Cistus, and Halimium and Helianthemum species. However, in Cytinus populations both the flower size and the shades of yellow of the flowers are markedly variable features (C. de Vega, unpublished data), and thus identification has depended mainly on the identity of hosts.
Our objective in this study was to look for genetic race formation in Cytinus with respect to different host species. To do so we used amplified fragment length polymorphism (AFLP) markers to investigate the genetic structure of Cytinus populations on its most common Cistaceae hosts. Sampling was focused on both sides of the Strait of Gibraltar (i.e. SW Spain and NW Morocco) because all the recognized Cytinus taxa occur in this area, and it is also the major centre of diversity of Cistaceae.
Materials and Methods
According to the occurrence of Cytinus on Halimium, Helianthemum and pink- and white-flowered Cistus species, our sample included the four subspecies recognized by Greuter et al. (1989), and the two species and subspecies recognized by Villar (1997) and Villar & López-Sáez (2002) from their taxonomic viewpoint. For convenience, the material used in this study will be referred to as Cytinus.
Heinricher (1917) reported that Cytinus requires 3 yr in the host before producing the first inflorescence. The longevity of Cytinus is unknown, although our observations indicate that its inflorescences burst through the same host root for at least 5 yr (de Vega et al., 2007). In the spring, Cytinus produces a simple short spike with female and male flowers at the base and top of the inflorescence, respectively. Hand-pollinations have shown that plants are self-compatible (C. de Vega, unpublished). Experimental studies have also shown that ants are the main pollinators, and these vectors generally move all over the same inflorescence or visit neighbouring inflorescences seeking nectar (C. de Vega, unpublished); thus, high levels of endogamy are likely to occur in natural populations. Populations of Cytinus are isolated and small (usually no more than 5–15 flowering individuals in the same host population), and the flowering period of the populations of the parasite on different host species is usually asynchronous (C. de Vega, unpublished).
The hosts of Cytinus are species of Cistaceae that are important components in many fire-prone, Mediterranean shrubland ecosystems (e.g. ‘maquis’ and ‘garrigue’). Cistaceae shrubs behave as obligate seeders adapted to disturbances operating in Mediterranean ecosystems, particularly fire (Trabaud, 1995). Most Cistaceae species have a relatively short life span (10–20 yr), although Cistus ladanifer L. is more long-lived, with a life span of 30–50 yr (Arianotsou-Faraggitaki & Margaris, 1982; Roy & Sonie, 1992; Patón et al., 1998). Many species such as C. ladanifer, Cistus populifolius L., Halimium halimifolium (L.) Willk. and Halimium multiflorum (Dunal) Maire prefer acid soils, but some prefer basic soils (e.g. Cistus albidus L.), and others are substrate indifferent (Cistus salviifolius L. and Cistus monspeliensis L.). Commonly, members of the Cistaceae, even congeneric species, co-occur at the same site (Herrera, 1987).
The Western Mediterranean is the major centre of Cistaceae diversity, particularly on both sides of the Strait of Gibraltar, and in this region we sampled Cytinus populations on 10 of their most common host species: C. albidus, C. ladanifer, C. monspeliensis, C. salviifolius, C. populifolius, Halimium atlanticum Humbert & Maire, H. halimifolium, H. multiflorum, Halimium lasiocalycinum (Boiss. & Reut.) Gross ex Engl., and Helianthemum glaucum Pers. (=Helianthemum croceum (Desf.) Pers.). Most recent treatments of the genus Cistus for the Iberian flora (Demoly & Montserrat, 1993; López González, 1993; Nogueira et al., 1993) have essentially followed the classifications of the genus by Dunal (1824), Willkomm (1856) and Dansereau (1939) (see Guzman & Vargas, 2005) which recognize the pink-flowered species (including C. albidus) in a separate subgenus or section, but also split the white-flowered species, with C. ladanifer in a separate section from C. monspeliensis, C. salviifolius and C. populifolius. A comprehensive molecular phylogenetic study of the family is lacking, but Guzman & Vargas (2005) studied the molecular phylogeny of the genus Cistus and they distinguished two main lineages: one comprising pink-flowered species and the other white-flowered species. However, the markers (internal transcribed spacer (ITS), trnl intron, 3’exon and trnl-F spacer (trnL-trnF) and megakaryocyte associated tyrosine kinase gene (matK) used by those authors did not produce a resolution of the group of white-flowered Cistus species, but instead showed a large, unresolved polytomy. In a more recent study which combined ribulose-biphosphate carboxylase gene (rbcL), matK and trnS-G sequences for 11 white-flowered species of Cistus, they recognized C. ladanifer as a monophyletic group, while the phylogenetic relationships among the other species remained unclear (B. Guzman & P. Vargas, pers. comm.).
In the present study we used a concept of population based on parasitized host: if Cytinus parasitized two different host species at a given site, then we assumed that these were two different populations of the parasite. This procedure was important because Cistaceae species very often occur sympatrically (e.g. C. ladanifer, C. salviifolius and H. halimifolium at Hinojos, Spain; C. albidus and C. populifolius at Constantina, Spain; and C. salviifolius and H. multiflorum at Asilah, Morocco; Table 1). Overall, 174 individuals from 22 populations of Cytinus were collected, parasitizing 10 Cistaceae species at different localities in SW Spain and NW Morocco (Table 1, Fig. 1). In each population we collected young flowers of different Cytinus individuals chosen at random, and dried them in silica gel. Given that different floral clusters on the same host are in effect different individuals (C. de Vega, unpublished), we collected only one individual per host plant infected. Voucher specimens from all populations were deposited in the herbarium of the University of Seville (SEV).
Table 1. Sampled populations of Cytinus grouped by host species
Locality and herbarium voucher
Population size (NIHP)
Flower colour: Y, yellow; W, white.
NIHP, population size of Cytinus based on the number of infected host plants in each population.SEV, University of Seville.
On Cistus albidus (sect. Cistus)
Spain, Constantina (Sevilla). 722 m; 37°56′N, 5°37′W (SEV 215856)
Spain, Zahara de la Sierra (Cádiz). 841 m; 36°48′N, 5°22′W (SEV 215855)
Spain, Sierra de las Nieves (Málaga). 1200 m; 36°39′N, 5°3′W (SEV 215857)
On Cistus ladanifer (sect. Ladanium)
Spain, Constantina (Sevilla). 881 m; 37°55′N, 5°34′W (SEV 215846)
Spain, Hinojos-Cuadrejón (Huelva). 90 m; 37°17′N, 6°24′W (SEV 215845)
On Cistus monspeliensis (sect. Ledonia)
Morocco, Jbel Tassaot (Tetouan). 700 m; 35°15′N, 5°10′W (SEV 215859)
Morocco, Jbel Talsemtam (Tetouan). 998 m; 35°4′N, 5°9′W (SEV 215860)
On Cistus populifolius (sect. Ledonia)
Spain, Constantina (Sevilla). 722 m; 37°56′N, 5°37′W (SEV 215861)
Spain, Constantina (Sevilla). 713 m; 37°57′N, 5°36′W (SEV 215862)
On Cistus salviifolius (sect. Ledonia)
Spain, Jabugo (Huelva). 684 m; 37°54′N, 6°43′W (SEV 215851)
Spain, Aracena (Huelva). 725 m; 37°52′N, 6°38′W (SEV 215852)
Spain, Hinojos-Membrillo (Huelva). 103 m; 37°17′N, 6°25′W (SEV 215849)
Spain, Hinojos-Las Palomas2 (Huelva). 106 m; 37°18′N, 6°25′W (SEV 215850)
Morocco, Asilah (Tanger). 200 m; 35°17′N, 6°3′W (SEV 215847)
On Halimium atlanticum (sect. ?)
Morocco, Jbel Tazzeka (Taza). 1845 m; 34°4′N, 4°10′W (SEV 215867)
On Halimium halimifolium (sect. Chrysorhodion)
Spain, Hinojos-Las Palomas1 (Huelva). 102 m; 37°18′N, 6°25′W (SEV 215866)
Spain, Hinojos-Las Palomas2 (Huelva). 105 m; 37°18′N, 6°25′W (SEV 215865)
Morocco, Essaoira (Safi). 100 m; 31°30′N, 9°41′W (SEV 215863)
On Halimium lasiocalycinum (sect. Chrysorhodion)
Morocco, Ketama (Tetouan). 1166 m; 34°59′N, 4°52′W (SEV 215869)
Morocco, Asilah (Tanger). 200 m; 35°17′N, 6°3′W (SEV 215864)
On Helianthemum croceum (sect. Helianthemum)
Morocco, Azrou (Meknes). 1947 m; 33°14′N, 5°12′W (SEV 215870)
DNA isolation and AFLP fingerprinting
Total genomic DNA was extracted from dried tissue following the 2 × cetyltrimethylammonium bromide (CTAB) protocol (Doyle & Doyle, 1987) with the following slight modifications: after precipitation with isopropanol and subsequent centrifugation, the DNA pellet was washed with 70% ethanol, dried in a vacuum centrifuge, and re-suspended in 20 mM Tris-HCl pH 8.0, 0.1 mM EDTA (TE buffer); DNA extracts were then treated with RNAse at 37°C for 30 min. The quality of the isolated DNA was checked on 1% Tris-acetate-EDTA (TAE) agarose gel, and the quantity of DNA was estimated photometrically (UV-160A; Shimadzu, Kyoto, Japan).
AFLP profiles were generated following established procedures (Vos et al., 1995) and according to the PE Applied Biosystems (Foster City, CA, USA; 1996) protocol with minor modifications (see Tremetsberger et al., 2003). An amount of 0.5 µg of genomic DNA was used in the restriction and ligation reactions with EcoRI and MseI restriction enzymes, conducted in a thermal cycler for 2 h at 37°C. Preselective amplifications (5 µl from restriction and ligation reactions) were performed using a single selective nucleotide primer pair (EcoRI-A and MseI-C). For selective amplification we conducted a previous screening of 18 primer combinations and chose and fluorescently labelled the three that revealed both the highest levels of intra- and inter-population variability and the clearest banding patterns. These three primer combinations were: MseI-CTC/EcoRI-ACA (FAM), MseI-CAT/EcoRI-ACG (HEX), and MseI-CAG/EcoRI-AGC (NED). Products from selective amplifications were separated on a 5% polyacrylamide gel with an internal size standard (GeneScan 500 (ROX); PE Applied Biosystems) on an automated DNA sequencer (ABI PRISM 377, Applied Biosystems). The peak profiles were first checked and aligned using the ABI PRISM genescan 2.1 analysis software (PE Applied Biosystems) and then viewed and scored as binary characters in genographer v. 1.6 (Montana State University; available at http://hordeum.oscs.montana.edu/genographer). AFLP profiles were scored for the presence (1) or absence (0) of fragments in the 50–500-bp range. To check AFLP reproducibility and reliability, 15 randomly selected samples were duplicated, and the profile of 1 s and 0 s yielded an average of 95% scoring repeatability.
Allele frequencies were estimated using the Bayesian approach proposed by Zhivotovsky (1999), based on nonuniform prior distribution of null-allele frequencies. This procedure produces less biased estimates of allele frequencies with dominant markers than some other methods (for a detailed comparison of methods, see Krauss, 2000). However, previous knowledge of the level of inbreeding (FIS) is required. To deal with this problem, we estimated allele frequencies under two contrasting assumptions: (1) FIS = 0, which assumes that populations are under Hardy–Weinberg equilibrium, and (2) FIS = 1, which assumes complete inbreeding, and equates allele frequencies to presence/absence data (Lynch & Milligan, 1994). The whole set of analyses, performed under the two assumptions, yielded similar results. We show results only for the FIS = 1 assumption, justified by the results of a 5-yr study on the mating system of Cytinus which indicate high levels of autogamy in the species (C. de Vega, unpublished).
Reynolds’ pairwise genetic distances among populations were calculated using the aflp-surv 1.0 software (Vekemans, 2002), and were subjected to neighbour-joining (NJ) analysis using the programs neighbor and consense of the phylip 3.57 package (Felsenstein, 2004). Branch support was assessed by bootstrap analysis (1000 replicates). In order to examine the relative position of Cytinus populations in a multidimensional genetic space, a pairwise FST matrix was analysed through nonmetric multidimensional scaling (NMDS) ordination analysis, and the first two components extracted were plotted graphically using statistica 6 (StatSoft, 2001).
As an alternative approach to detect possible hidden spatial population structure in the Cytinus populations across the sampled geographic area, we applied two individual-based Bayesian approaches to our AFLP data set. Firstly, we used the software structure version 2.2 (Pritchard et al., 2000; Falush et al., 2007) to estimate the most probable number of genetically differentiated Cytinus groups (K). structure is a Markov chain Monte Carlo-based approach that clusters individuals to minimize Hardy–Weinberg and gametic phase disequilibria within groups (Pritchard et al., 2000). We assumed a model with population admixture and that the allele frequencies were correlated within populations. Runs for each K value (from 1 to 22) were independently replicated 10 times with a burn-in of 105 followed by 105 iterations of Monte Carlo Markov Chain (MCMC). The estimate of the posterior probability of the data (logeP(D)) was then standardized using the ad hoc ΔK statistic, based on the rate of change between successive K values, to infer the uppermost level of structure in the data set (Evanno et al., 2005).
Secondly, the same data set was analysed using baps (Corander et al., 2004) to assess whether these results were robust to different analytical methodologies. baps uses a stochastic optimization algorithm approach finding the posterior mode of the genetic structure and has a faster execution speed compared with MCMC methods. The optimal number of genetic groups (K) was estimated using the ‘clustering of individuals’ option. Values of K from 1 to 22 were entered 10 times each. The output of this initial analysis was then used as the basis for an admixture analysis using the ‘admixture based on mixture clustering’ option. For this, the minimum size of a population was set at three individuals and runs of 105 iterations.
Overall genetic differentiation among populations (FST; Weir & Cockerham, 1984) was estimated using aflp-surv 1.0 (Vekemans, 2002), and its significance tested by random permutation (1000 permutations) of individuals among populations. Additionally, we obtained an independent estimate of genetic differentiation (θB) following the Bayesian approach developed by Holsinger et al. (2002) as implemented in hickory 1.0 (Holsinger & Lewis, 2003). This approach has the major advantage that it does not require any previous knowledge of levels of within-population inbreeding, and it has been shown to provide accurate and reliable estimates of genetic differentiation with dominant markers (e.g. Holsinger & Wallace, 2004). We ran the four models available in the program for dominant markers (full model, f = 0 model, θ = 0 model, and the f free model), and the deviance information criterion (DIC; Spiegelhalter et al., 2002) was used to determine the model showing the best fit to the data.
We conducted three analyses of the molecular variance (AMOVA; Excoffier et al., 1992) for the hierarchical partitioning of the total genetic diversity: (1) among populations and among individuals within populations; (2) among host species, among populations within host species and among individuals within populations; and (3) among groups (five groups representing the taxonomic sections of the host species; see the Results), among populations within groups, and among individuals within populations. AMOVA analyses were performed in arlequin 3.0 (Excoffier et al., 2005).
Finally, we checked whether genetic structure among Cytinus populations was correlated with geographic distance, according to a pattern of isolation by distance, using a Mantel test with the program passage version 1.0 (Rosenberg, 2001). Pairwise genetic distances were computed as FST/1 − FST (Rousset, 1997) in arlequin 3.0 (Excoffier et al., 2005).
For each population, we report the proportion of polymorphic loci at the 95% level (P) and Nei's gene diversity (HJ) as implemented in aflp-surv 1.0 (Vekemans, 2002), and the mean proportion of pairwise band differences between individuals in each population (D) computed with arlequin 3.0 (Excoffier et al., 2005). For each of the five groups representing the taxonomic sections of the host species, we report the number of exclusive and diagnostic fragments, the total genetic diversity (HT), the average gene diversity within populations (HS) and the coefficient of genetic differentiation (GST) using aflp-surv 1.0 (Vekemans, 2002). As differences in sampling intensity between populations could bias our comparisons of genetic diversity, we computed the band richness (Br) standardized to the smallest sample size by means of a rarefaction method with the aflp-div 1.0 software (Coart et al., 2005).
The three primer combinations yielded 361 unambiguously scored loci, of which 318 (88.18%) were polymorphic. Each of the 174 samples displayed a unique banding pattern.
Using the ΔK estimator derived from the rate of change of the posterior probability for each K, the Bayesian analysis of structure clustered the 174 Cytinus individuals into five genetically distinct groups: the first group includes the populations of Cytinus parasitizing C. ladanifer; the second includes those parasitic on C. monspeliensis, C. salviifolius and C. populifolius; the third comprises exclusively Cytinus populations on C. albidus; the fourth comprises populations growing on H. halimifolium, H. multiflorum and H. lasiocalycinum; and finally, the fifth includes populations of Cytinus on H. atlanticum and H. croceum (Fig. 2). The host species of the first four groups fall into the following sections of the genera Cistus and Halimium: C. ladanifer (sect. Ladanium (Spach) Gren. & Godr.); C. monspeliensis, C. salviifolius, and C. populifolius (sect. Ledonia Dunal); C. albidus (sect. Cistus); and H. halimifolium, H. multiflorum, and H. lasiocalycinum (sect. Chrysorhodion Spach); sectional classifications according to Demoly & Montserrat (1993), López González (1993), and Nogueira et al. (1993). We use hereafter the names of these sections to designate these four genetic groups of Cytinus. The hosts of the fifth group are species that belong to two different genera: H. atlanticum (section unknown) and H. croceum (sect. Helianthemum), and we will refer to the populations of Cytinus on these two species as the Helianthemum/Halimium group. Analysis with baps supported these five genetic groups in Cytinus according to the genera and taxonomic sections of their host species (Fig. 2). Individuals from the same population were always assigned to the same cluster.
The NJ tree grouped Cytinus populations into the five main genetic groups which were supported with high bootstrap values (89–100%; Fig. 3). The tree also revealed weak evidence for host specificity in the genetic group of Cytinus for the white-flowered species of Cistus included in section Ledonia. The number of exclusive fragments (i.e. occurring only in some individuals of a genetic group) was 40 in Cytinus populations on Halimium sect. Chrysorhodion, 25 in Cytinus on Cistus sect. Ledonia, 16 in Cytinus on Cistus sect. Ladanium, and five in Cytinus on Cistus sect. Cistus. In these four genetic groups, diagnostic fragments (i.e. appearing in all individuals of a genetic group and not in another) were not detected. However, populations Ha15 and Hc22, from Cytinus parasitizing the Halimium/Helianthemum group (two Moroccan species that were located 130 km apart), shared 25 exclusive fragments and six diagnostic fragments, indicating, as does the 100% bootstrap value, that these two populations were closely related.
Despite the fact that populations of different genetic groups of Cytinus often occurred sympatrically (Table 1), they never grouped together either in individual or in population-based analyses (Figs 2, 3). Moreover, the Mantel test of genetic vs geographical distance did not support isolation-by-distance as a model of differentiation among Cytinus populations (r = 0.011; P = 0.456). For example, all populations on Halimium sect. Chrysorhodion (Hh16–Hh18, Hm21 and Hl19–Hl20) clustered together (Fig. 3), despite the fact that some of them were located approx. 700 km apart (e.g. Hh18 at Essaoira, Morocco was approx. 700 km from Hh16 and Hh17, both located in Huelva, Spain). Results of the nonmetric multidimensional scaling analysis revealed a similar genetic clustering pattern (Fig. 4). When populations of Cytinus were plotted over the first two extracted dimensions, the five main clades of the Bayesian analyses and that of the NJ tree could be distinguished.
Assuming a situation of complete inbreeding within populations (FIS = 1), the estimate of genetic differentiation (FST) among populations was 0.552, while the Bayesian estimator displayed a slightly lower value, with θB = 0.486. When individuals were pooled into the five genetic groups, the two descriptors yielded high, similar values of genetic differentiation (0.503 and 0.473, respectively).
AMOVA performed at the population level showed that most of the genetic variation was found among populations (60.34%; Table 2). This result agreed with the previous estimates of genetic differentiation. When populations were pooled at the host-species level, AMOVA showed that occurrence on host species accounted for the highest proportion of variance (51.45%), whilst differences among populations on the same host were markedly lower (11.66%). When the five genetic groups defined by host occurrence were analysed, the AMOVA revealed that 44.92% of the total variance was explained by differences among Cytinus genetic groups, with a lower proportion (20.18%) found among populations within genetic groups.
Table 2. Results of three analyses of molecular variance (AMOVA) for 22 populations of Cytinus
% of total
Populations were grouped according to their host occurrence or Cytinus genetic groups. Total variation was partitioned (a) among populations, (b) among host species and among populations nested within host species, and (c) among genetic groups and among populations nested within subspecies.
SS, sum of squares; VC, variance components.
(a) Population level
(b) Grouped by host species
Among host species
Among populations within host species
(c) Grouped by Cytinus genetic groups
Among Cytinus genetic groups
Among populations within Cytinus groups
Total diversity for genetic groups of Cytinus was relatively low (HT = 0.180, Table 3). Populations were well differentiated, as shown by the high coefficient of genetic differentiation among populations (GST = 0.555), whilst those of Cytinus on Cistus sect. Ladanium had the lowest total genetic diversity (HT = 0.077), and they also showed the lowest values of within-population genetic diversity (Table 3). By contrast, populations of Cytinus on Halimium sect. Chrysorhodion had the highest value of total genetic diversity (HT = 0.150) and also the highest within-population genetic diversity (HS = 0.095).
Table 3. Genetic diversity statistics for genetic groups in Cytinus
HT, total genetic diversity; HS, within-population genetic diversity; GST, coefficient of genetic differentiation.
On Cistus sect. Cistus
On Cistus sect. Ladanium
On Cistus sect. Ledonia
On Halimium sect. Chrysorhodion
On Halimium/Helianthemum Group
Populations of Cytinus on Cistus sect. Ladanium showed low values of genetic variability, P ranging from 15 to 18.6% and HJ from 0.055 to 0.069 (Table 4), as did those of Cytinus on Cistus sect. Ledonia, although in the latter these values were very variable (P ranged from 7.8 to 38.5% and HJ from 0.038 to 0.130). Moreover, most populations of Cytinus on Cistus sect. Ledonia had low band richness (Br), showing that the lowest genetic variability was not biased by the number of sampled individuals (Table 4). Populations of Cytinus on Halimium sect. Chrysorhodion showed the highest levels of genetic variability (P ranged from 23.8 to 38.2%, HJ from 0.075 to 0.119 and Br from 1.100 to 1.157).
Table 4. Genetic variability of 22 populations of Cytinus
n, sample size; P, percentage of polymorphic loci at 95% criterion; Br, band richness; HJ± 1 SE, Nei's gene diversity; D± 1 SE, mean number of pairwise differences between individuals within populations. For population abbreviations see legend to Fig. 1.
On Cistus sect. Cistus
0.117 ± 0.011
37.000 ± 19.514
0.084 ± 0.008
22.166 ± 12.466
0.073 ± 0.008
17.666 ± 10.005
On Cistus sect. Ladanium
0.069 ± 0.007
20.178 ± 10.002
0.055 ± 0.006
17.029 ± 7.784
On Cistus sect. Ledonia
0.077 ± 0.008
21.800 ± 11.244
0.100 ± 0.009
28.400 ± 15.059
0.059 ± 0.007
14.200 ± 7.700
0.060 ± 0.007
11.333 ± 7.122
0.058 ± 0.006
16.511 ± 8.045
0.053 ± 0.006
15.022 ± 7.349
0.059 ± 0.006
18.419 ± 8.658
0.038 ± 0.005
10.780 ± 5.863
0.130 ± 0.010
43.311 ± 20.567
On Halimium/Helianthemum group
0.068 ± 0.007
18.200 ± 9.773
0.080 ± 0.008
22.600 ± 12.054
On Halimium sect. Chrysorhodion
0.075 ± 0.008
26.648 ± 13.883
0.084 ± 0.008
26.989 ± 14.136
0.085 ± 0.009
24.000 ± 14.696
0.119 ± 0.010
33.333 ± 20.275
0.106 ± 0.009
30.166 ± 16.839
0.117 ± 0.009
37.888 ± 18.034
In most cases, not only do the Cytinus samples from different populations on the same host species group together in the NJ analysis, but the majority of the parasite populations also show a high degree of relatedness at the infrageneric sectional level of their host species. Thus, populations of Cytinus parasitizing host species belonging to the same taxonomic section were genetically very similar, and, overall, the parasite plants could be separated clearly into four genetic groups that infest, respectively, species of Cistus sect. Ladanium, Cistus sect. Ledonia, Cistus sect. Cistus, and Halimium sect. Chrysorhodion. Our results are also partially in accord with the phylogenetic studies on Cistus by Guzman & Vargas (2005). Those authors, using ITS, trnL-trnF and matK markers, distinguished two phyletic lineages comprising pink-flowered vs white-flowered Cistus species, but with the latter taxa emerging as an unresolved polytomy. In an attempt to obtain better resolution using rbcL, matK and trnS-G sequences for 11 white-flowered species of Cistus, B. Guzman & P. Vargas (pers. comm.) recognized C. ladanifer as a monophyletic group, while they did not find evidence of monophyly of sect. Ledonia. Our data indicate a genetic group of Cytinus on Cistus sect. Cistus (with pink-flowered species), and two distinct genetic groups/races that infest the white-flowered species, one on C. ladanifer and the other common to C. monspeliensis, C. salviifolius and C. populifolius, the former species included in sect. Ladanium, and the latter three in sect. Ledonia, in traditional infrageneric classifications.
The only exception to this remarkable specialization at the sectional level occurred for the Cytinus population sampled on H. atlanticum (Ha15 from Jbel Tazzeka), which showed a closer affinity to the population of the parasite sampled on H. croceum (Hc22 from Azrou) than to the parasite populations sampled on the other three Halimium host species from Spain and Morocco (Hh16, Hh17, Hm21, Hl19 and Hl20). This group of Cytinus on H. croceum has been described as parasitizing other species of Helianthemum, and the genera Fumana and Halimium (Maire, 1961). Thus, this is the group with the highest diversity of susceptible host species.
This pattern, as found in Cytinus, with genetically distinctive groups associated with the infrageneric sections of the host species, has not been reported previously for parasitic angiosperms. In the few cases where molecular studies were employed, two contrasting patterns were described: specificity at the host species level (e.g. Jerome & Ford, 2002), or little specificity, with plants capable of parasitizing different genera and families (Nickrent & Stell, 1990; Zuber & Widmer, 2000; Linhart et al., 2003). A high genetic similarity between parasite populations on the same host species suggests an ancient origin of the host–parasite relationship (Olivier et al., 1998). The high number of fragments that are exclusive to the genetically defined groups of Cytinus, and in some cases also of diagnostic fragments, suggests that these groups are well-characterized lineages that have evolved independently. The AMOVA results, revealing strong genetic differentiation among the different groups of Cytinus, are congruent with the view that these groups are independent lineages that may be regarded as genetic races or cryptic species.
If several host species of the same section have parasites that are closely related vs parasites of species of other sections, this would imply that the ancestral species of those sections was already infected with a distinctive race of Cytinus before speciation occurred to create the related group of species that now make up the section. Thus, the already differentiated parasite race would have been carried passively through the adaptive radiation of each group of Cistus/Halimium/Helianthemum species. Differentiation of the genera of Cistaceae took place during the Oligocene to Miocene (B. Guzmán & P. Vargas, pers. comm.), and the differentiation of the Cistus/Halimium complex occurred in the late Miocene or Pliocene (Guzman & Vargas, 2005), followed by the evolution of Cistus species into two lineages with purple and white flowers, respectively. Consequently, the divergence between races of Cytinus may predate the Oligocene to Miocene period. An ancient Cytinus–Cistaceae association such as this implies a long period of evolutionary divergence, during which adaptation of Cytinus to different host species could have occurred.
The high level of genetic differentiation among the various races of Cytinus shown by our data, and the results of the AMOVA, indicate a low level of historical gene flow. Various aspects of the biology of Cytinus can explain this situation: even when populations of two or more races of Cytinus coexist in the same community (such as the cork-oak (Quercus suber L.) forests of southern Spain), they have partially or totally asynchronous flowering (C. de Vega, unpublished). This feature alone would limit gene flow via pollen, and help maintain the genetic differentiation of the populations; but, additionally, Cytinus seeds are dispersed by ants and mice (C. de Vega, unpublished), and because these vectors move in small areas, seed-mediated gene flow is also likely to be restricted. Poor pollen flow and limited seed dispersal may explain the high population differentiation found in this study.
The values of genetic diversity in populations of Cytinus are much lower than mean values in populations of angiosperms in general, and they are also lower than those found for species with high rates of endogamy (Nybom, 2004). They are also lower than those reported for some other parasitic plants (e.g. Bharathalakshmi et al., 1990; Jerome & Ford, 2002; Mutikainen & Koskela, 2002; Linhart et al., 2003), although such comparisons with other parasites require caution as some of the studies were performed using allozymes. There are various nonexclusive biological and ecological factors that could contribute to the low intra-populational genetic diversity in Cytinus: one obvious factor is that the parasite is a self-compatible taxon with high rates of endogamy, but another might be the small effective size of its populations. Although cistaceous host species form very large populations with thousands of individuals, the number of plants infected by Cytinus, at least as indicated by the production of inflorescences, is surprisingly low, with usually no more than 5–15 infected plants per population. A small population size entails bottlenecks in which genetic drift may cause reduction in the gene pool (Ellstrand & Elam, 1993). Another factor contributing to the low intra-populational diversity of Cytinus might be the occurrence of a founder effect when new host populations are infested, and this, together with a probable low rate of recombination, could lead to a rapid loss of variability during the initial stages of establishment of Cytinus populations.
The striking association between the Cytinus groups defined in the Bayesian analyses, the NMDS and the NJ tree and the taxonomic sections of the host genera points to a clear host specificity. Closely related parasites tend to live on closely related hosts, which can lead to the parasites and hosts having congruent phylogenies (Norton & Carpenter, 1998). This tendency, known as the Farenholz rule, has led to the view that the phylogenetic trees of the parasites and of their hosts may be topologically identical (Hafner & Nadler, 1988; Hafner & Page, 1995; Downie & Gullan, 2005). This rule has recently received considerable attention from parasitologists, because, if true, the phylogenies of the parasites can be predicted from the phylogenies of the hosts, and vice versa (Norton & Carpenter, 1998). A test case of this principle for Mediterranean Cytinus will be provided if, in future studies on the molecular phylogeny of the family Cistaceae, white-flowered Cistus species of section Ledonia are recognized as a monophyletic group and H. atlanticum (a poorly known species) is in fact recognized as being close to the genus Helianthemum.
As to taxonomic implications, the results from our limited sampling (SW Spain and NW Morocco) of populations of Cytinus support the recognition of at least five main genetic groups. One group comprises ivory-white to pink-flowered Cytinus on pink-flowered Cistus, which would correspond to C. hypocistis subsp. clusii or C. ruber. The other four are yellow-flowered forms. We observed two distinct groups of Cytinus on white-flowered species of Cistus: a group of parasites infesting white-flowered C. ladanifer (sect. Ladanium), and another infesting white-flowered species of sect Ledonia (C. monspeliensis, C. populifolius and C. salviifolius). These two groups would correspond to C. hypocistis subsp. hypocistis but, as noted above, our results have depicted them as two independent lineages. The fourth group of Cytinus grows on Halimium sect. Chrysorhodion, and it would correspond to subsp. macranthus. Finally, Cytinus on the Halimium/Helianthemum group would accord with C. hypocistis subsp. lutescens, which is the subspecies with the broadest range of host species including the genera Helianthemum, Halimium and Fumana (e.g. Maire, 1961). Cytinus hypocistis sens. lat. is a circum-Mediterranean complex, and our sampling in SW Spain and Morocco does not allow assessment of the taxonomy of Cytinus in the Mediterranean area. More detailed sampling will be necessary before their taxonomic boundaries can be resolved satisfactorily.
We thank R. Vega-Durán and E. Durán for help with the field sampling, and G. Kadlec for technical assistance in the laboratory. Special thanks to R. G. Albaladejo, P. E. Gibbs and three anonymous reviewers for helpful comments. This work was supported by a PhD grant from the Spanish MEC to the first author and by two projects from the Spanish CICYT (REN2002-04354-C02-02 and CGL 2005-01951) and one project from the Junta de Andalucía (EXC/2005/RNM-204).