Insights on a rare species, Geum atlanticum , new data to differentiate from Geum sylvaticum

. Geum atlanticum is a species described from North Africa and later found in Southern Iberian Peninsula. However, according to Flora iberica taxonomic treatment, it has been considered synonymous with G. sylvaticum , a species mainly distributed by Southeast of France, Iberian Peninsula, and North of Africa, sharing part of the distribution range and habitat. In this work, herbarium specimens of both G. atlanticum and G. sylvaticum were revised to evaluate morphological differences. Furthermore, the nuclear ribosomal ITS spacer was sequenced to study phylogenetic relationships among G. atlanticum and Iberian Geum species. The morphological study indicated clear differences with G. sylvaticum , and the phylogenetic analysis suggests that G. atlanticum was sister to G. sylvaticum . However, it shared several mutations with other Iberian Geum species. The number of populations is relatively low in the Iberian Peninsula, so its conservation status and its further inclusion in lists of threatened plants have been evaluated in this work.


Introduction
Geum is a world-wide distributed genus belonging to family Rosaceae. This genus includes approximately 56 species (Gajewski, 1959), although depending on the author, the number has changed over time (see explanation below). Of them, seven are distributed throughout the Iberian Peninsula (Laínz, 1998). Its taxonomy has been subject of debate since introgression and allopolyploidization seems to be extensive in the tribe Colurieae. In fact, several classifications traditionally have included species belonging to Geum in other genera or alternatively, several genera were synonymized as Geum (Scheutz, 1870;Greene, 1899;Bolle, 1933;Hara, 1935;Yuzepchuk, 1941). Its phylogeny suggests that introgression and polyploidization have been a major role in the evolution of Geum and other closely related genera, pointing out that the classic circumscription of Geum is not supported (Smedmark & Eriksson, 2002).
As commented before, seven Geum species are located in the Iberian Peninsula, belonging to three subgenera (Laínz, 1998); i) G. rivale L., G. sylvaticum Pourr. G. pyrenaicum Mill., G. urbanum L., and G. hispidum Fr. Included in subgenus Geum L., which shows its widest distribution in Europe, North America and Asia, ii) G. montanum L. included in subgenus Oreogeum Ser., typical from the mountains of Europe and iii) G. heterocarpum Boiss. included in subgenus Orthostylus (C.A. Mey.) F. Bolle, whose species are mainly located in the Mediterranean Region (Laínz, 1998). These species are well defined morphologically. Even several of them are considered as sister species in the phylogenetic study of Smedmark & Eriksson (2002).
However, as a result of our botanical explorations made in the Southeast Iberian Peninsula (mainly in Albacete province), we detected several populations of deviant G. sylvaticum, with longer habit and achenia, which inhabited near "typical" populations of the same species. These individuals showed several morphological traits that belong to G. atlanticum Desf., a species described from North Africa (Tlemcen, Algeria) (Desfontaines, 1798). After that, Font Quer (1928) proposed a new combination, G. sylvaticum var. atlanticum Font Quer & Pau. A few years later, Lindberg (1932) proposed that this variety was an endemic taxon from North, being followed by Quézel (1980) in the "Flore de l'Afrique du Nord", suggesting that G. sylvaticum var. atlanticum was exclusively located across the mountains of Algeria and Morocco, whereas G. sylvaticum s. str. would be distributed in the Iberian Peninsula and South of France. Interestingly, Quézel & Santa (1962) did not mention G. atlanticum in their work regarding Algerian flora, even as a synonym of G. sylvaticum. However, Cuatrecasas collected and reviewed accessions previously collected by Font Quer in Andalusia and Pau in Mongó mountain (Valencia) in the early twentieth century, which are the first identifications of G. atlanticum in the Iberian Peninsula, belonging to two herbarium accessions (MA55959, MA55990). Cuatrecasas proposed two new varieties for G. atlanticum; G. atlanticum var. cabrillense Cuatrec. nomen nudum (BC 20006), from Sierra de la Cabrilla (Jaén) and var. rosulatum Cuatrec. nomen nudum (BC91935), from Sierra del Pinar, in Grazalema (Cádiz).
More recently, several authors (Laínz, 1995(Laínz, , 1998Ibn Tattou, 1999;Romo, 2002) considered populations of G. atlanticum from Iberian Peninsula and Northern Africa as a synonym of G. sylvaticum without further explanation or suggesting that this issue is not completely solved. Laínz (1995Laínz ( , 1998 pointed out that some accessions cited as G. atlanticum showed morphological traits deviant from the typical G. sylvaticum. However, this author, arguing that cariology and chorology studies were inconclusive, suggested that this morphological variability alone was not enough evidence to separate these two taxa. In this work, we have performed a morphological and molecular analysis to elucidate the taxonomic status of G. atlanticum, focusing on the available material in order to i) explore their morphological and genetic differentiation from G. sylvaticum, and ii) to assess the conservation status of G. atlanticum according to IUCN criteria (IUCN, 2012).

Morphological and chorological analysis
The morphological study was based on the material collected and deposited in BC, MNHN (P), MUB, MA, GDA, SALA, MPU, VAL and ANVL herbaria (Thiers, continuously updated). A portion of the accessions has been revised through digital images of preserved plants from some of the mentioned herbaria or digital platforms as GBIF, JSTOR or ANTHOS (GBIF, 2019;Anthos, 2020;MNHN, Chagnoux, 2020a;MNHN, Chagnoux, 2020b). Furthermore, morphological data of the newly located G. atlanticum populations from Southeastern Iberian Peninsula (mainly Albacete) were obtained from direct examination of living plants from Alcaraz and Segura range mountains. Herbarium specimens belonging to G. sylvaticum and G. atlanticum (including those described as G. sylvaticum var. atlanticum) are listed in Appendix 1 and quantitative and qualitative morphological traits studied are listed in Table 1. To evaluate differences in morphometric traits between the two taxa we conducted a T-test after checking the homogeneity of variance and normality of data with a Levene's test and a Shapiro-Wilk's test, respectively. All analyses were performed with SPSS v.24, IBM Corp., 2016.

Molecular analysis
For this work, the ITS (ITS1-5,8S-ITS2) region of nuclear ribosomal DNA from two individuals of G. atlanticum, G. sylvaticum and G. hyspidum was sequenced, which were complemented by previously published sequences (downloaded from GenBank) belonging to seven species of Geum and two sequences of Rubus ulmifolius Schott as outgroup taxon to root the tree. The taxa studied are listed in Table 2, with GenBank accession numbers and voucher details.

DNA extraction, PCR amplification and sequencing
Total DNA was extracted using the cetyltrimethyl ammonium bromide method (Doyle & Doyle, 1987) with slight modifications and stored frozen at -20ºC until amplification. The internal transcribed spacer region (ITS1-5.8S-ITS2) of nuclear ribosomal DNA was amplified with the primer pair 17S and 26S (Sun et al., 1994). The PCR reactions were carried out in an Eppendorf Mastercycler thermocycler using the following program: an initial cycle of 94°C for 5 min, 35 cycles of 94°C for 30 s, 52°C for 45 s, 72°C for 1 min, and a final extension step of 72°C for 8 min to complete the PCR. The reactions were conducted in a final volume of 50 μl using approximately 40 ng of DNA, 0.4 μM of direct and reverse primers, 5 μl of polymerase buffer (provided by the supplier of the enzyme), 0.2 mM of each of dNTPs, 2 mM of MgCl 2 and 2 U of Taq polymerase (Biotools). Finally, 2 μl of each amplification product were visualized on 1.5% agarose gels, and successful amplifications were cleaned using the GenElute PCR Clean-Up kit (SIGMA). Same primers were used in the sequencing reactions with the Big Dye sequencing kit and the products were separated in an automatic ABI 3700 sequencer following standard protocols. For each DNA sample, both strands were sequenced. Maximum likelihood (ML) and Bayesian inference (BI) analyses were performed with 17 nrITS sequences.
Sequences were checked for inaccurate base identification using Chromas Lite v2.01 (Technelysium Pty Ltd.). Consensus sequences of ITS were aligned using ClustalX (Thompson et al., 1997). BioEdit (Hall, 1999) was used to make minor alignment adjustments. A Bayesian and Maximum Likelihood (ML) analysis were performed. Both analyses were carried out under GTR + G + I model of evolution according to JModeltest 2 (Guindon & Gascuel, 2003;Darriba et al., 2012). MrBayes, v. 3.2 (Ronquist et al., 2012) was used to perform Bayesian analyses. Two simultaneous runs were initiated by starting from random trees. To ensure that the two runs converged onto a stationary distribution, analyses were run until the average standard deviation of the split frequencies was 0.01. Convergence was evaluated using the potential scale reduction factor (PSRF), and 1,000,000 generations were run by sampling every 100th generation at the following settings: Nst = 6, rates = invgamma. Burnin (the number of starting generations ruled out from further analyses) was set at 200,000 generations after visually inspecting the likelihood values in Excel. A 50% majority-rule consensus tree was constructed using the "sumt" command of MrBayes. ML phylogenetic trees were conducted with 1,000 bootstrap replicates by the rapid Bootstrap analysis in v.1.5b1 of RAxMLGUI (Silvestro & Michalak, 2012). The best-scoring ML trees were chosen as the final trees and bootstrap values were added to nodes. Trees were edited with Figtree, v.1.4 (Rambaut, 2012).

Morphological results
According to the morphological examination of herbarium accessions and fresh material from Southeastern Iberian Peninsula (mainly Albacete populations), we have observed clear distinctive patterns among G. sylvaticum and the fresh material plus several herbarium accessions from Spain and North Africa. These features are listed in Table  1. Most of the morphometric variables studied showed no significant differences between G. sylvaticum and G. atlanticum. Only total length of achene (seminiferous part plus beak), length of beak, and stem's leaves showed significant differences (Table 1). G. atlanticum achenia are longer than that of G. sylvaticum (up to 17 mm vs. 12 mm respectively) and with long patent hairs on the base of the achene. Moreover, the number of achenia per flower is higher in G. atlanticum (>40 per flower) than in G. sylvaticum. Moreover, the habit from the G. atlanticum morphological type usually shows longer stems with longer leaves than G. sylvaticum. Another remarkable difference is the absence of carpophore in G. atlanticum (Figure 2).

Molecular Results
The two ITS1-5.8S-ITS2 sequences generated here for G. atlanticum from fresh material of two Albacete populations were identical and different from those corresponding to the rest of Geum species. We have observed that G. sylvaticum sequences (from Albacete) show several mutations compared to G. atlanticum, and these mutations are shared between G. atlanticum and G. hispidum (from Serranía de Cuenca range mountain). Pairwise comparisons of percentage sequence divergence for the ITS region among all species used in the molecular analyses are summarized in Table 3. Sequence divergence between the species of Rubus and Geum ranged between 10.91% (R. ulmifolius vs. G. atlanticum) and 13.51% (R. ulmifolius vs. G. speciosum Albov). Sequence divergence among the Geum species ranged between 0.15% (G. atlanticum vs. G. hispidum) and 6.25% (G. speciosum vs. G. heterocarpum). Sequence divergence among G. atlanticum and G. sylvaticum is 1.04%. The final sequence alignment of the 15 ingroups and two outgroup taxa was 719 bp long. Of these, 605 sites were constant, 25 were variable but parsimony uninformative, and 89 were parsimoniously informative. Both the ML and Bayesian inference searches resulted in trees with a similar topology. Therefore, the Bayesian posterior probability (PP) and ML bootstrap values (Bs) are provided in the same tree for all the analyses ( Figure  3). The phylogenetic tree shows G. heterocarpum as basal species and G. speciosum as sister species (Bs = 100%, PP = 100%). This species is related to a clade with low bootstrap support (Bs = 72%) but high Bayesian support (PP = 100%), in which G. montanum and G. rivale are located a sister species, related to G. aleppicum Jacq. and G. geniculatum Michx. in a subclade with low Bs and PP support, whereas in other subclade are located G. hispidum and G. urbanum, as sister species of a subclade which includes G. atlanticum and G. sylvaticum. These subclades also show low support, suggesting the difficulty of elucidating the phylogenetic relationships in Geum.

Distribution and habitat of G. atlanticum
According to the available data from studied herbarium accessions as well as bibliographic references with locations of G. atlanticum (Muñoz Medina, 1951;Barbero et al., 1981;González Bueno, 1988;Valdés et al., 2006;Mateos & Valdés, 2009;Romo, 2009;Chambouleyron, 2012) (Figure 1), this taxon should be considered an Ibero-African species, present in North Africa and the South and Southeast of the Iberian Peninsula. In North Africa, it is distributed for humid mountains from Morocco (Rif, Middle Atlas and rarer in High Atlas) and Algeria (Tell Atlas, from Tlemcen, Blida, Djurdjura, to the mountains of Babor and Constantine, and Saharan Atlas in Bellezma and Aurès mountains). In Northern Africa, it inhabits mesophylous forests of conifers as Cedrus atlantica (Endl.) Manetti ex Carrrière, Abies spp. and Quercus spp. at 1000-2400 m asl, in several types of soils. In the Iberian Peninsula, including the newly located populations, it is distributed along the Baetic mountains; Grazalema, Sierra Nevada, Baza mountains and the Subbaetic ranges of Cazorla, Segura and Alcaraz. Moreover, there is a narrow and isolated population in the South of Valencia province (Montgó mountain) and at least another in Sierra Morena (Despeñaperros), where it is very rare. In the Iberian Peninsula, G. atlanticum, as well as in Northern Africa, occurs in mesophylous forests. However, it accompanies other species, for example, Pinus pinaster Aiton, P. nigra subsp. salzmannii (Dunal) Franco, Quercus rotundifolia Lam., Q. faginea Lam. and Taxus baccata L., always within the meso-supramediterranean (oromediterranean) belt, in subhumid or humid ombrotype, on limestone or siliceous soils. It flowers from May to July and fruit formation takes place from June to August. Bootstrap values (Bs) followed by Bayesian posterior probabilities (PP) above 50% are given below the branches. The tree was rooted with Rubus ulmifolius as outgroup.

Discussion
As commented before, G. atlanticum was described by Desfontaines from the mountains near Tlemcen, in Algeria (Desfontaines, 1798). However, several years later, Pau and Font Quer synonymized populations from North Africa as a variety of G. sylvaticum (G. sylvaticum var. atlanticum), with individuals from Morocco sharing an exsiccatae of this taxon with several herbaria. Individuals with such morphological features were located in the Iberian Peninsula, being named as the former variety, although Lindberg (1932) and Quézel (1980) suggested that this taxon was endemic from North Africa. Later, Laínz (1995) attempted to clarify the taxonomic status for these individuals with the search of fresh material in the previously cited populations to obtain chromosome numbers that shed light. Unfortunately, he could not obtain enough material to perform a satisfactory analysis and suggested to maintain all the Iberian accessions as G. sylvaticum, although pointing out that more studies were needed to clarify the taxonomic status of G. atlanticum (Laínz, 1998).
In this work, we have sampled new individuals and populations with the morphological traits typical of G. atlanticum, as the number of achenia per flower, the size and indumentum of achenia, and the absence of carpophore, which are unique morphological traits that are fixed and consistent in every individual analyzed, and are absent in G. sylvaticum. Moreover, the phylogenetic study shows that G. atlanticum is closely related to G. sylvaticum, as expected, but it shares several mutations with G. hispidum and G. urbanum. Indeed, in the ML tree, it is located in an intermediate position between G. sylvaticum and the pair G. hispidum-G. urbanum. These circumstances point out that G. atlanticum should not be considered as a synonym of G. sylvaticum, but a different species, as suggested by Desfontaines. Its distribution area would cover the North of Africa and the South and Southeast of the Iberian Peninsula, while G. sylvaticum would be restricted to the Iberian Peninsula and South of France. G. atlanticum distribution range seems to coincide with other mesophylous and orophylous Iberoafrican species, which probably migrated between Africa and Europe through Baetic-Rifean belt (Blanca, 1990;Lavergne et al., 2013;Sánchez-Robles et al., 2014). Its ecology is quite similar in every population. However, in North Africa, it inhabits at a higher altitude, probably due to the lower latitude of these populations.
Although no morphological differences have been observed between African and Iberian individuals studied, some studies seem necessary to clarify the taxonomic status and phylogenetic position of this taxon. Taking into account the importance of introgression and polyploydization in the evolution of Geum genus (Gajewski 1958;Smedmark & Eriksson 2002), new chromosome counts, cytogenetic studies and new phylogenetic studies using more nuclear and plastid regions in which more populations were sampled (including several from North Africa), are needed. Also, a phylogeographic study in which the highest possible number of populations from the Iberian Peninsula and North Africa were involved could help to establish hypotheses about the origin and evolution of populations of this taxon.

Implication in Conservation
According to available data, G. atlanticum is widely distributed in the most humid mountains of North Africa. Although overgrazing and environmental degradation in the most suitable habitats (Thomas, 2013;Benabid et al., 2015) might have had a great impact on the survival of populations in recent times. However, a significant number of populations are located in National Parks in both Morocco and Algeria, which suggests that these populations would be free of anthropogenic threats. On the contrary, taking into account that recently located populations are narrow and their number of individuals are scarce (less than one hundred individuals per population), G. atlanticum seems to be a rare species in the South of the Iberian Peninsula, although it has probably gone undetected. However, most of the populations are located in protected territories under European level (Natura 2000 Network), or Regional Parks of Andalusia, Castilla-La Mancha, and Valencia.
The number of known populations suggests that G. atlanticum is not threatened at a global level. Nevertheless, according to IUCN criteria (IUCN, 2012), G. atlanticum should be considered as Data Deficient (DD). An extensive sampling of historical localities might help to elucidate its conservation status. Due to the scarcity of populations in the Iberian Peninsula is feasible that G. atlanticum was protected in several Autonomous Communities from Spain where it would probably be threatened at the regional level. Further population studies are needed to address its threat level and to perform adequate conservation policies.