The Macaronesian endemic flora has traditionally been interpreted as a relict of a subtropical element that spanned across Europe in the Tertiary. This hypothesis is revisited in the moss subfamily Helicodontioideae based on molecular divergence estimates derived from two independent calibration techniques either employing fossil evidence or using an Monte Carlo Markov Chain (MCMC) to sample absolute rates of nucleotide substitution from a prior distribution encompassing a wide range of rates documented across land plants. Both analyses suggest that the monotypic Madeiran endemic genus Hedenasiastrum diverged of other Helicodontioideae about 40 million years, that is, well before Macaronesian archipelagos actually emerged, in agreement with the relict hypothesis. Hedenasiastrum is characterized by a plesiomorphic morphology, which is suggestive of a complete morphological stasis over 40 million years. Macaronesian endemic Rhynchostegiella species, whose polyphyletic origin involves multiple colonization events, evolved much more recently, and yet accumulated many more morphological novelties than H. percurrens. The Macaronesian moss flora thus appears as a complex mix of ancient relicts and more recently dispersed, fast-evolving taxa.

Macaronesia is a string of North Atlantic volcanic islands (the Azores, Madeira, Canaries, and Cape Verde) that emerged 0.4–20 million years ago and are characterized by high rates of endemism (see Juan et al. 2000 for review). Engler (1879), followed by many biogeographers (see Vanderpoorten et al. 2007 for review), proposed that Macaronesian endemics are the relics of biota that were widespread across Europe during the Tertiary and decimated on the continent during the glaciations. In contrast with the expectations of the refugium concept, however, several lines of evidence from analyses of moss species distributions (Vanderpoorten et al. 2007) and molecular evolution rates in angiosperms (Carine 2005) have recently questioned Engler's hypothesis. In mosses, the hypothesis of extinction of a Tertiary flora in all but Macaronesian areas is solely supported by extremely limited fossil evidence in the genus Echinodium (Frahm 2004). This interpretation is, however, weakened by the polyphyletic origin of the genus (Stech et al. 2008), thereby raising doubts about the actual sister relationship between fossil and extant Macaronesian Echinodium species. By contrast, the nested phylogenetic position of the Azorean endemic liverwort Leptoscyphus azoricus within a Neotropical clade (Devos and Vanderpoorten 2009); the close biogeographic affinities of several Macaronesian groups with the North and South American continents (see Vanderpoorten et al. [2007] for review); and the polyphyletic origin among Macaronesian populations of the moss Grimmia montana, which are genetically identical or most closely related to those of different continents (Vanderpoorten et al. 2008) all point to recurrent migrations between the latter and the Macaronesian archipelagos, potentially followed by in situ speciation.

The continental extinction and recent speciation hypotheses make different predictions regarding the rates of speciation and morphological evolution. In fact, as opposed to neoendemics, which originated from local speciation processes and often underwent spectacular adaptive radiations involving a sudden burst of morphological diversification (see Gillepsie and Roderick [2002] for review), paleoendemics, which survived continental extinctions on the islands, have most often retained a highly conserved morphology for millions, or tens of million years, to such an extent that extant taxa appear conspecific with fossil species (Sunding 1979).

In this article, we revisit hypotheses on the origin of Macaronesian endemism in the mosses from the Brachytheciaceae, subfamily Helicodontioideae. This group includes several Macaronesian endemics, namely the monotypic Madeiran Hedenasiastrum percurrens, and three Rhynchostegiella species: Rhynchostegiella bourgeana and R. trichophylla, which are restricted to the Canary Islands, and R. macilenta, whose distributions span Madeira and the Canaries. In addition, and unlike most bryophytes, a fairly well-documented fossil record is available (Miller 1984). We produced a molecular phylogeny of the Helicodontioideae to test the hypothesis of a radiation within Rhynchostegiella and date the origin of its Macaronesian endemic lineages, contrasting the results derived from a fossil calibration and an analysis employing absolute rates of molecular evolution. We then used the phylogeny to test the hypothesis of a long morphological stasis since the Tertiary period among Macaronesian endemics.

Material and Methods


The 17 genera of Helicodontioideae as circumscribed by Aigoin et al. (2009) were sampled (Table 1). Each genus was represented by one to three species, with a special emphasis on Rhynchostegiella, for which all eight species were sampled. Three other species, namely R. papuensis, R. leptoneura, and R. muriculata, clearly do not belong to the genus, and the appropriate taxonomic changes will be presented elsewhere. Aerobryidium filamentosum, a species of the sister family Meteroriaceae, as well as representative taxa of each of the three subfamilies of Brachytheciaceae was sampled as outgroups (Table 1). Forty-three morphological characters that are variable across the Helicodontioideae, including 34 gametophytic and nine sporophytic ones, were scored for each of the sampled taxa (Appendices S1 and S2) (see Huttunen and Ignatov [2004] for a thorough account on character significance and description).

Table 1.  Taxon sampling, voucher information (for sequences produced for the present study), and GenBank accession numbers.
SpeciestrnL/trnFatpB/rbcLpsbT/psbHpsbA/trnHVoucher specimen for sequences
Aerobryidium filamentosum (Hook.) M. Fleisch.AF397789AF417347 
Aerolindigia capillacea (Hornsch.) M. MenzelFJ262414FJ262441FJ262474FJ262499Ecuador, Toapanta & Caranqui 1437 (MO)
Brachytheciastrum collinum (Schleich. ex Müll. Hal.) Ignatov and HuttunenAY184776AY663296AY184757 
Brachythecium salebrosum (Hoffm. ex F. Weber and D. Mohr) Schimp.AF397857AY663309AF417448AY312896 
Bryhnia novae-angliae (Sull. and Lesq.) GroutAF161122AJ288397AF417405 
Bryoandersonia illecebra (Hedw.) H. Rob.AF397819 AF417365FJ262501USA, Bowers 22214 (MA)
Cirriphyllum crassinervium (Taylor) Loeske and M. Fleisch.FJ262415FJ262443FJ262476FJ262502France, Vanderpoorten 413 (LG)
Cirriphyllum koponenii (Ignatov) Ignatov and HuttunenFJ262416FJ262444AF417446FJ262503Papua New Guinea, Koponen 32122 (H)
Cirriphyllum piliferum (Hedw.) GroutFJ262417FJ262445AF417403FJ262504Finland, Koponen & Huttunen 1324 (H)
Clasmatodon parvulus (Hampe) Sull.FJ262418FJ262446FJ262477FJ262505USA, Vanderpoorten 4748 (LG)
Donrichardsia macroneuron (Grout) H.A. Crum and L.E. AndersonAY009848AF322323 
Donrichardsia patulifolia (Cardot and Thér.) Ignatov and HuttunenFJ262419FJ262447FJ262478FJ262506China, Koponen et al. 53920 (H)
Eurhynchiastrum pulchellum (Hedw.) Ignatov and HuttunenFJ262420FJ262448FJ262479FJ262507France, Sotiaux & Sotiaux 14670 (herb. Sotiaux)
Eurhynchiella zeyheri (Spreng. ex Müll. Hal.) M. Fleisch.FJ262421FJ262449FJ262480FJ262530South Africa, Hylander B78518 (S)
Hedenasiastrum percurrens (Hedenäs) Ignatov and VanderpoortenFJ262442FJ262475FJ262500Madeira, Hedenäs B42883 (S)
Helicodontium capillare (Hedw.) A. JaegerFJ262422FJ262450FJ262481FJ262512Venezuela, Vanderpoorten V180 (LG)
Homalotheciella subcapillata (Hedw.) Broth.FJ262423FJ262451F417462FJ262513U.S.A., Allen & Redfearn sn (NY)
Homalothecium fulgescens (Mitt. ex Müll. Hal.) A. JaegerAF397877EF530978AF417466  
Meteoridium remotifolium (Müll. Hal.) Müll. Hal.AY306783AF417418FJ262514Suriname, Newton 4399 (BM)
Microeurhynchium pumilum (Wilson) Ignatov and VanderpoortenFJ262434FJ262466FJ262493FJ262509France, Vanderpoorten 4415 (LG)
Okamuraea brachydictyon (Cardot) Nog.FJ262424FJ262452AY184771FJ262515China, Koponen et al. 48969 (H)
Oxyrrhynchium hians (Hedw.) LoeskeFJ262425FJ262453FJ262482FJ262508France, Vanderpoorten 401 (LG)
Oxyrrhynchium savatieri (Schimp. ex Besch.) Broth.AF397859FJ262454AF397859FJ262510China, Koponen et al. 51775 (H)
Oxyrrhynchium vagans (A. Jaeger) Ignatov and HuttunenAF397862FJ262455AF417450FJ262511China, Koponen et al. 49717 (H)
Platyhypnidium austrinum (Hook. f. and Wilson) M. Fleisch.AY184791FJ262483FJ262516Australia, Streimann 49544 (H)
Platyhypnidium riparioides (Hedw.) DixonAY857573AY857595   
Pseudorhynchostegiella duriaei (Mont.) Ignatov and VanderpoortenFJ262428FJ262459FJ262486FJ262519Madeira, Hedenäs B9050 (S)
Pseudoscleropodium purum (Hedw.) M. Fleisch.AF397797AF233567AF417470  
Remyella brachypodia (M. Fleisch.) Ignatov and HuttunenAF397854FJ262456AF417423Papua New Guinea, Koponen 33007 (H)
Rhynchostegiella bourgeana (Mitt.) Broth.FJ262426FJ262457FJ262484FJ262517El Hierro, Dirkse 1882 (L)
Rhynchostegiella curviseta (Brid.) Limpr.FJ262427FJ262458FJ262485FJ262518France, Vanderpoorten 421 (LG)
Rhynchostegiella holstii (Broth.) Broth.FJ262429FJ262460FJ262487FJ262520South Africa, Hylander B78534 (S)
Rhynchostegiella leptoneura CardotFJ262431FJ262462FJ262489FJ262522China, He 36074 (MO)
Rhynchostegiella litorea (De Not.) Limpr.FJ262432FJ262463FJ262490FJ262523Madeira, Hedenäs B9057 (S)
Rhynchostegiella macilenta (Renauld and Cardot) CardotFJ262433FJ262464FJ262491FJ262524Madeira, Hedenäs B4503 (S)
Rhynchostegiella muriculata (Hook. f. and Wilson) Broth.DQ208214FJ262465FJ262492FJ262525Australia, Streimann 49628 (H)
Rhynchostegiella papuensis E.B. BartramAF417439 
Rhynchostegiella tenella (Dicks.) Limpr.FJ262436FJ262468FJ262495FJ262527Russia, Ignatov & Ignatova 05-6143 (MW)
Rhynchostegiella teneriffae (Mont.) Dirkse and Bouman #1FJ262430FJ262461FJ262488FJ262521Madeira, Hedenäs B9096 (S)
Rhynchostegiella teneriffae (Mont.) Dirkse and Bouman #2FJ262437FJ262469FJ262496FJ262528France, Vanderpoorten 362 (LG)
Rhynchostegiella teneriffae (Mont.) Dirkse and Bouman #3FJ262435FJ262467FJ262494FJ262526Switzerland, Hedenäs B11895 (S)
Rhynchostegiella trichophylla Dirkse and BoumanFJ262438FJ262470FJ262497FJ262529Gran Canaria, Dirkse 13843 (L)
Rhynchostegium psilopodium Ignatov and HuttunenFJ262439FJ262471FJ262498FJ262531China, Koponen et al. 51803
Scorpiurium circinatum (Brid.) M. Fleisch. and LoeskeAF397834FJ262472AF417410 France, Vanderpoorten M16 (LG)
Squamidium brasiliense (Hornsch.) Broth.FJ262440FJ262473AF417393FJ262532Tanzania, Pócs et al. 88161 (R)
Zelometeorium patulum (Hedw.) ManuelAF397787 AF417362  


Four chloroplast regions (trnL-trnF, atpB-rbcL, psbT-psbH, and psbA-trnH) were selected for exhibiting the appropriate level of variation at the genus level in the Brachytheciaceae (Huttunen and Ignatov 2004). DNA extraction, PCR and sequencing protocols, sequence editing, alignment, indel scoring, and selection of models for DNA substitutions and indel evolution follow Aigoin et al. (2009). Phylogenetic reconstruction was conducted with MrBayes 3.1.2. Four independent Monte Carlo Markov Chains (MCMCs) of 2,000,000 iterations each were run and trees and model parameters were sampled every 10,000 generations. The convergence of the MCMCs was verified visually from the likelihood values, and trees of the “burn-in” were discarded.

A significant departure of alternative topologies involving a monophyletic origin of the three Macaronesian endemic Rhynchostegiella species, namely R. bourgeana, R. macilenta, and R. trichophylla, was tested by constrained analyses. The MCMC analysis described above was rerun under the constraint that only trees fitting with a monophyletic origin of the Macaronesian endemic Rhynchostegiella species were sampled. We then determined whether the constraint induced a significant loss of likelihood by means of the Bayes factors, as assessed by twice the difference in the log marginal likelihood between the two runs.


Times of divergence were calculated to determine the origin of the most recent common ancestor (hereafter, MRCA) of H. percurrens and of the three Macaronesian endemics of Rhynchostegiella. Examination of rate variation among branches of the phylogeny suggested a strong departure of the sequence data from the molecular clock (see below). Divergence times were therefore estimated using a Bayesian MCMC method under a relaxed-clock model employing an uncorrelated lognormal model of rate variation among branches in the tree as implemented by BEAST 1.4.8. Nonequivocal Miocene fossils of Cirriphyllum piliferum and Oxyrrhynchium hians, and Pliocene fossils of Clasmatodon parvulus (Miller 1984) were used for calibration at three internal nodes. In bryophytes, wherein the probability of fossilization is very low owing to the absence of lignified tissues and much reduced cuticle, the presence of a species is likely to be recorded in the sediment only well after its actual phylogenetic origin. We therefore used asymmetric distribution priors with a mode at the beginning of the period considered and asymptotically skewed toward the end of it (Fig. 1).

Figure 1.

Fifty percent majority-rule consensus tree from the Bayesian analysis of four chloroplastic genes in Helicodontiodeae, with branch lengths averaged over the 284 trees sampled from the posterior probability distribution. Posterior probabilities are indicated under each branch. Fossil calibration points are labeled by stars and the range of the prior distribution encompassing the geological era during which the fossil was found is illustrated in boxes. The dates indicated at five nodes (black circles) correspond to the median (and 95% confidence interval) of the age of selected nodes, as inferred from the molecular dating analysis when the fossil calibration (F) or absolute substitution rates of chloroplastic DNA (R) were used. The four Macaronesian endemic species of Helicodontioideae appear in bold.

In a second set of analyses, absolute nucleotide substitution rates documented for a large number of plant lineages (including bryophytes) were used. Accordingly, a normal distribution, with a mean and standard deviation of 5.0 × 10−10 and 10−10 substitutions/site/year, respectively, which corresponds to the average absolute substitution rate of cpDNA across land plants and largely encompasses their variation range (see Schnabel and Wendel [1998] for review), was used as a prior on the absolute rates of evolution for the four cpDNA regions combined.

Although BEAST allows for topologies, branch lengths, and rates to be coestimated, we chose to perform the dating analyses onto topologies that were derived independently from the kind of calibration used. Therefore, we constrained the topology to match the 50% majority-rule consensus of the trees sampled from the posterior probability distribution generated by MrBayes, but used the MCMC implemented by BEAST to sample branch lengths and substitution rates depending on their posterior probabilities. A Yule prior on branching rates was employed and two independent MCMC analyses were each run for 10,000,000 steps. Parameter values were sampled every 1000 cycle over the 10,000,000 MCMC steps. Convergence and acceptable mixing of the sampled parameters was checked using Tracer 1.4. After discarding the burn-in steps, the two runs were combined to obtain an estimate of the posterior probability distribution of the divergence dates of the ancestral nodes.


Ancestral character state reconstructions were performed to contrast the hypotheses that Macaronesian endemics are characterized by plesiomorphic or derived character states. All reconstructions were performed after pruning the outgroups from the trees. The probabilities of a change in a branch were calculated by estimating the instantaneous forward and backward rates among the two states by implementing the Markov model of “Multistate” in BayesTraits 1.0.

To contrast alternative hypotheses regarding the morphology of the MRCA of the Macaronesian endemics, we used the “local” approach, wherein the significance of the reconstruction is explicitly tested at each node of interest (Pagel 1999). For that purpose, we fixed each of the MRCAs of H. percurrens and Rhynchostegiella at one of the two states it can take. Then, an MCMC was used to visit the sample of trees generated by the MrBayes analysis and the space of rate parameter values. In the absence of information on rates, uniform distributions ranging between 0 and 100 were used as priors. The likelihood of the new combination of a rate and a tree was calculated and this new state of the chain was accepted or rejected following evaluation by the Metropolis–Hastings term. The rate, at which parameters were changed (“ratedev”), was set at the beginning of each run so that the acceptance rate of the proposed change globally ranges between 20% and 40%. The chain was run for 10,000,000 generations and was sampled for rate parameters every 10,000 generations. A second, independent chain was run to sample rate parameters and derive overall likelihoods of the reconstructions when the node of interest was fixed at its alternative state. Bayes factors were then used to determine the support for alternative state at each node of interest.



The chloroplast matrix contains 19% variable sites. The 50% majority-rule consensus of the 284 trees sampled from the posterior probability distribution is presented in Figure 1. Macaronesian endemics within the Helicondontioideae appear at the two extremities of the phylogeny. The Madeiran endemic Hedenasiastrum percurrens is sister to all other genera of the Helicodontioideae, and this relationship is supported with a posterior probability (hereafter, p.p.) of 1.00. By contrast, Rhynchostegiella is one of the most recently diverging groups of species of the Helicodontioideae. Within Rhynchostegiella, the sub-Saharan African R. holstii is part of a polytomy with two other clades. The first clade includes the Canarian endemics R. bourgeana and R. trichophylla with a p.p. of 0.98. Within the second clade, the Macaronesian endemic R. macilenta is resolved as sister to a Madeiran accession of R. teneriffae with a p.p. of 0.96. The hypothesis of a monophyletic origin of the three Macaronesian endemic species can be significantly rejected. In fact, constraining all three Macaronesian endemic species to monophyly resulted in a significantly lower log marginal likelihood by comparison with the unconstrained analysis (log marginal likelihood of −6884.07 and −6861.42 for the constrained and unconstrained analyses, respectively).


The coefficient of variation of rates among branches of the phylogeny was 0.63 (95% highest posterior density 0.44–0.83) for the first analysis (fossil calibration) and 0.60 (95% highest posterior density 0.43–0.77) for the second one (rate sampling from a prior distribution). This suggests a strong departure of the data from a molecular clock, a condition in which the coefficient of variation equals zero. The two kinds of calibration used for molecular dating by a relaxed-clock model provided congruent results regarding both the medians of the posterior probability distributions and their 95% confidence intervals (Fig. 1). Hedenasiastrum percurrens emerged 38 [20;74] million years with the fossil calibration and 42 [25;65] when absolute substitution rates were employed. The divergence and diversification of Rhynchostegiella, respectively, date back to 20 [9;35] and 11 [4;20] million years with the fossil calibration and 22 [13;36] and 12 [6;20] with absolute rates of molecular evolution. The MRCA of the Macaronesian endemics R. bourgeana and R. trichophylla was dated back to 7 [2;14] million years with the fossil calibration and 8 [3;15] when a distribution of absolute substitution rates was used as a prior, whereas R. macilenta is estimated to have diverged more recently, 1 [0;4] million year according to the fossil calibration, and 1 [0;4] when absolute rates were used (Fig. 1).


The marginal log-likelihoods of the reconstruction of ancestral morphological character states at the MRCA of the Macaronesian endemics are presented in Table 2. When the MRCA of the Helicodontioideae was constrained to be morphologically identical to the Madeiran endemic H. percurrens, the marginal log-likelihoods of the reconstruction were significantly higher for 20 of the 32 investigated gametophytic characters. For the remaining 12 characters, the marginal log-likelihood obtained after fixing the root of the Helicodontioideae at one or the alternative state were not significantly different (i.e., the Bayes factors were <2), so that the state at the root was ambiguous for those characters. Altogether, the two Canarian endemics R. bourgeana and R. trichophylla differ from the MRCA of the genus by six character states (#12, 13, 15, 20, 23, 26). Finally, R. macilenta, which is the phylogenetically most recently diverging Macaronesian endemic within the Helicodontioideae, differs from the MRCA of Rhynchostegiella by nine character states (#5, 10, 11, 12, 13, 18, 20, 26, and 35).

Table 2.  Ancestral character state reconstructions at the most recent common ancestors of the Macaronesian endemics within the Helicodontioideae. See Appendix S1 for character identification and description. Bayes factors (BF) measure twice the difference between the log of the harmonic means (HM) returned by the model when the MRCA is successively set at its two possible states, with a difference of 2–5 considered as positive evidence (*), 5–10 strong evidence (**), and >10 very strong evidence (***). The state returning the highest log marginal likelihood is underlined.
 Character state ofHMBFCharacter states ofHMBF
H. percurrensConstrained MRCAR. bourgeanaR. trichophyllaR. macilentaConstrained Rhynchostegiella MRCA
Character 100/1−11.35/−15.518.3**0000/1−12.21/−19.6114.8***
Character 200/1−21.61/−21.650.11100/1−22.06/−21.331.5
Character 310/1−14.65/−14.910.5000/1−14.14/−20.3912.5***
Character 400/1−13.34/−16.756.8**0000/1−13.24/−20.5314.6***
Character 500/1−14.93/−18.076.3**0010/1−14.47/−21.9515.0***
Character 600/1−11.46/−15.417.9**0000/1−11.73/−19.8216.2***
Character 700/1−9.71/−14.659.9**0000/1−9.97/−19.2818.6***
Character 800/1−22.97/−23.20.5 000/1−22.92/−23.270.7
Character 900/1−22.42/−23.612.4*0000/1−22.19/−24.795.2**
Character 1000/1−18.79/−20.844.1*0010/1−18.68/−22.076.8**
Character 1100/1−21.94/−22.811.70010/1−21.55/−23.594.1*
Character 1200/1−18.46/−20.413.9*0100/1−20.39/−18.423.9*
Character 1300/1−21.92/−22.541.20100/1−23.39/−21.713.4*
Character 1410/1−16.66/−16.710.11110/1−21.02/−15.7910.5***
Character 1500/1−9.95/−15.3310.8***0100/1−10.82/−16.1710.7***
Character 1600/1−19.42/−21.464.1*0000/1−19.31/−24.259.9**
Character 1700/1−13.66/−16.936.5**0000/1−13.42/−21.1415.4***
Character 1800/1−23.05/−23.821.50010/1−22.90/−24.022.2*
Character 1910/1−20.39/−20.490.20110/1−20.74/−20.340.8
Character 2000/1−17.95/−19.893.9*0110/1−15.98/−21.1310.3***
Character 2100/1−14.77/−17.45.3**0000/1−14.88/−21.0312.3***
Character 2200/1−21.27/−22.231.9 000/1−21.16/−22.562.8*
Character 2300/1−22.55/−23.010.90100/1−22.13/−24.344.4*
Character 2400/1−14.02/−17.256.5**0000/1−14.13/−20.7513.2***
Character 2500/1−11.39/−14.766.7**000/1−11.23/−18.4814.5***
Character 2610/1−23.61/−22.711.81000/1−24.38/−22.194.4*
Character 2710/1−11.49/−10.521.910/1−13.75/−10.506.5**
Character 2800/1−13.5/−14.211.41  0/1−13.50/−14.401.8
Character 2900/1−15.98/−18.394.8*0000/1−15.79/−22.0112.4***
Character 3000/1−18.83/−20.964.3*0000/1−18.78/−23.9210.3***
Character 3100/1−17.11/−19.023.8*0000/1−16.87/−22.1710.6***
Character 3200/1−10.21/−14.699.0**0000/1−10.63/−19.3717.5***
Character 330/10000/1−12.19/−17.8411.3***
Character 34 0/1  0000/1−10.22/−17.8915.3***
Character 350/10010/1−18.08/−20.174.2*
Character 36 0/1  0000/1−11.73/−17.1510.8***
Character 370/10000/1−14.79/−18.417.2**
Character 38 0/1  0000/1−15.89/−16.932.1*
Character 390/10000/1−12.97/−16.737.5**
Character 40 0/1  0000/1−13.20/−16.787.2**
Character 410/10000/1−11.62/−15.768.3**
Character 42 0/1  0000/1−11.85/−16.449.2**
Character 430/10000/1−10.71/−17.6313.8***


The two calibration techniques used in the present study, that is the use of fossils and the sampling of absolute substitution rates from a prior distribution encompassing a wide range of rates documented across land plants, rely on completely different assumptions. Yet, they returned highly congruent results, which strongly reinforces the confidence in the molecular dating inferences. The MRCA of Rhynchostegiella dates back to about 10 million years. This timing is compatible with both a paleo- and a neo-endemic origin of its Macaronesian endemic species. The polyphyletic origin of the latter, however, does not lend support to Engler’ refugium hypothesis. Like in some angiosperms (see Carine et al. [2004] for review), the colonization of Macaronesia by Rhynchostegiella species seems to have involved at least two dispersal events. Rhynchostegiella was one of the few bryophyte genera, wherein several Macaronesian endemics had been described, and therefore was one of the few candidates for adaptive radiations. The polyphyletic origin of its Macaronesian endemics invalidates that hypothesis, although the sister relationship between R. bourgeana and R. trichophylla suggests that local speciation may, to a minimal extent, have happened. Furthermore, and although this needs to be confirmed by additional sampling, the Macaronesian endemic R. macilenta is resolved as a sister to a sympatric accession of R. teneriffae, rendering the latter species paraphyletic. Such a situation has been recently increasingly described in other bryophyte species and has been interpreted in terms of recent speciation and incomplete lineage sorting (see Sotiaux et al. [2009] for review), thereby reinforcing the idea of a neoendemic origin of R. macilenta.

Although the polyphyletic origin and recent evolution of the Macaronesian endemic Rhynchostegiella species is thus suggestive of multiple colonization events followed by rather recent divergence, the Madeiran endemic moss H. percurrens diverged about 40 million years, that is well before Madeira actually emerged, 5.2 million years. This time frame, together with the fairly long branch leading to this unique species, is suggestive of numerous extinctions and a relictual origin. Hedenasiastrum percurrens actually becomes the only clear-cut documented case of relict in the Macaronesian bryophyte flora. In the context of a “dispersalist counter-revolution” fueled by the outcome of recent molecular phylogenies and molecular dating studies (de Queiroz 2005), and of a reassessment of the recent origin of emblematic island endemics (e.g., Goldberg et al. 2008; Grandcolas et al. 2008), including Macaronesia (e.g., Jordal and Hewitt 2004; Emerson and Oromí 2005; Dimitrov et al. 2008), the present study argues for a pluralistic view of Macaronesian biogeography. Hedenasiastrum percurrens thus falls within the increasingly limited category of confirmed Tertiary relicts. The acceptance of a relictual origin of H. percurrens, together with its morphological identity with the MRCA of the Helicodontioideae, dated at 40 million years, lends support to the traditional view of bryophytes as “unmoving sphinxes of the past” (see Devos and Vanderpoorten [2009] for review). In fact, the morphological similarity between extant and fossil species suggests that some groups of mosses have persisted with little morphological change for at least 80 million years (see Devos and Vanderpoorten [2009] for review). The question of whether this similarity is indeed due to the direct sister relationship or homoplasy remains, however, open. For example, the oldest liverwort fossil, the Devonian Pallaviciinites devonicus, exhibits a striking resemblance with the unrelated Pallavicinia and Symphyogyna (Heinrichs et al. 2007), suggesting that the similarity between fossil and extant taxa likely results from homoplasy.

Thus, although evidence for cryptic speciation in bryophytes is mounting (see Devos and Vanderpoorten [2009] for review), morphological stasis over periods of time of a range of tens of million years actually appears as an extremely rare phenomenon. In contrast, fast rates of morphological evolution or diversification have been repeatedly reported in bryophytes (Shaw et al. 2003; Wall 2005; Devos and Vanderpoorten 2009; Sotiaux et al. 2009). In the case of H. percurrens, however, although character evolution might be constrained, so that all the space of morphologies cannot be necessarily explored (Beldade et al. 2002), convergence or parallelism occurring to develop what is the same taxon two or more times is, arguably, less likely than plesiomorphy (Zander 2008). The retention of so many plesiomorphic traits during a 40-million-year period is remarkable and can be compared to other spectacular examples of living fossils (see Lee et al. [2006] for review). Surprisingly, and despite their much more recent origin, all three Macaronesian endemic Rhynchostegiella species accumulated many more morphological novelties than H. percurrens. Although the remarkable stasis of H. percurrens might originate from stabilizing selection, the reasons for such striking differences in evolutionary rates within the same group of mosses remain completely unexplained.

Associate Editor: J. Vamosi


DA, ND, and AV acknowledge financial support from the Belgian Funds for Scientific Research (F.R.S.–FNRS) and the Léopold III Funds, and MI the RFBR 07-04-00013. Many thanks are due to M. Carine, P.-H. Fabre, R. G. Gillepsie, F. Kjellberg, and three anonymous reviewers for their constructive comments on an earlier draft of this article.