A story from the Miocene: Clock‐dated phylogeny of Sisymbrium L. (Sisymbrieae, Brassicaceae)

Abstract Morphological variability and imprecise generic boundaries have hindered systematic, taxonomical, and nomenclatural studies of Sisymbrium L. (Brassicaceae, Sisymbrieae DC.). The members of this almost exclusively Old‐World genus grow mostly on highly porous substrates across open steppe, semidesert, or ruderal habitats in the temperate zone of the Northern Hemisphere and African subtropics. The present study placed the biological history of Sisymbrium L. into time and space and rendered the tribus Sisymbrieae as monotypic. Five nuclear‐encoded and three chloroplast‐encoded loci of approximately 85% of all currently accepted species were investigated. Several accessions per species covering their whole distribution range allowed for a more representative assessment of intraspecific genetic diversity. In the light of fossil absence, the impact of different secondary calibration methods and taxon sets on time spans was tested, and we showed that such a combinatorial nested dating approach is beneficial. Multigene phylogeny accompanied with a time divergence estimation analysis placed the onset and development of this tribus into the western Irano‐Turanian floristic region during the Miocene. Continuous increase in continentality and decrease in temperatures promoted the diversity of the Sisymbrieae, which invaded the open grasslands habitats in Eurasia, Mediterranean, and South Africa throughout the Pliocene and Pleistocene. Our results support the assumption of the Irano‐Turanian region as a biodiversity reservoir for adjacent regions.

taxonomic studies on Sisymbrium followed including Payson (1922) and Schulz (1924Schulz ( , 1928). Schulz's circumscription of the genus was based on nonapomorphic single characters only, including linear and terete fruits (siliques) and seeds with incumbent cotyledons.
Thus, it was criticized in subsequent studies on Sisymbrium by Rollins (1943) and Romanczuk (1981Romanczuk ( , 1982. It was Rollins (1982) who first regarded Sisymbrium as an exclusively Old World taxon.
Its distribution area was believed to be limited to the dry habitats of the western USA, but it was also recently discovered in northwestern China (Chen et al., 2019). While being highly controversial in the past, the generic limits of Sisymbrium are nowadays wellunderstood and generally accepted (following Al-Shehbaz, 2012, 2015Rollins, 1982) with a notable exception of Sisymbrium aculeolatum Boiss. and Sisymbrium afghanicum Gilli (Warwick & Al-Shehbaz, 2003). While these two taxa have been formally moved from Sisymbrium to Neotorularia Hedge & Léonard (Léonard, 1986), there was only little justification for such a transfer (Appel & Al-Shehbaz, 2002). Furthermore, the whole plastome phylogeny placed S. aculeolatum next to Sisymbrieae (Walden et al., 2020).
Hitherto it remains unclear, whether the two aforementioned species belong to the genus Sisymbrium or not.
Together with the monotypic Ochthodium (Nikolov et al., 2019;Walden et al., 2020) Sisymbrium belongs to the tribus Sisymbrieae, which shows a close relationship with the other tribes in the Lineage II of the Brassicaceae phylogeny-Brassiceae, Isatideae, and Thelypodieae (BrassiBase: Kiefer et al., 2014;Koch et al., 2012Koch et al., , 2018. Currently, there is no consent either which of these tribes are sister group to the Sisymbrieae, or which of the other clades build a sister group relationship within the Lineage II (BrassiBase: https://brass ibase.cos.uni-heide lberg.de). A number of single, multi-locus, and whole-genome studies recovered contradicting sister group relationships with Sisymbrieae Beilstein et al., 2010;Couvreur et al., 2010;Franzke et al., 2011;German et al., 2009;Huang et al., 2016Huang et al., , 2020Liu et al., 2020;Nikolov et al., 2019;Walden et al., 2020;Warwick et al., 2010). Warwick et al. (2002) published the last and the most comprehensive phylogeny of Sisymbrium L. based on nuclear genes, which also served as an anchor point for our study. However, over the past two decades, new species of Sisymbrium have been discovered and validly described (Al-Shehbaz, 2015;Blanca et al., 2015;Mutlu & Karakuş, 2015), several taxonomic revisions took place (Al-Shehbaz, 2004a, 2004b, 2006a, 2006bGerman & Al-Shehbaz, 2018;Warwick & Al-Shehbaz, 2003), and the shortcomings of the use of single potentially paralogue markers have been uncovered (Koch et al., 2007;Pirie et al., 2007). No comprehensive chloroplast-based Sisymbrium phylogeny has been published until now. Furthermore, only scarce genetic data of individual chloroplast regions (Arias & Pires, 2012;Hall et al., 2002;Koch et al., 2007;Warwick et al., 2004Warwick et al., , 2006 or whole plastomes (Hohmann et al., 2015;Nikolov et al., 2019;Walden et al., 2020) are available, preventing a comprehensive comparative study. Out of 514 different Sisymbrium names that can be found in the literature, currently only 50 of them represent accepted species and subspecies, leaving out 466 synonyms and one name with an unresolved status (Kiefer et al., 2014;Koch et al., 2018).
The southern East European Plain is part of a nowadays continuous Eurasian steppe belt, which is the vastest grassland region in the world with 8,000 km in length and up to 1,000 km in width. This ecoregion was under strong influence of glacial and interglacial cycles during the past five MYR, which continuously caused contractions, expansions, and latitudinal range shifts of the Eurasian steppe belt . In contrast, the western and central regional subcenters of the Irano-Turanian region (Léonard, 1988) were not as strongly affected by the Quaternary glaciations as the Eurasian steppe belt. Only intermountanious valley glaciations were recorded from the Pleistocene cycles, leaving ancestral flora at least partially intact (Agakhanjanz & Breckle, 1995). This is corroborated also by high levels of endemic species in Iranian mountain ranges and current distribution patterns of several representative Irano-Turanian plant taxa (Djamali, Baumel, et al., 2012;Noroozi et al., 2008). The western and central regional subcenters of the Irano-Turanian region may have played an important role by partially "supplying" adjacent regions-northern and eastern Irano-Turanian subcenters, as well as the Mediterranean Basin and Saharo-Arabian floristic region-with floral elements (Magyari et al., 2008;Manafzadeh et al., 2014Manafzadeh et al., , 2017Takhtajan et al., 1986;Žerdoner Čalasan et al., 2019).
The overall aim of our studies is to place the biological history of Sisymbrium L. into time and space. To achieve that we carried out a multigene phylogeny using biparental fast-evolving ITS and ETS regions, as well as low copy nuclear markers Bra13, Bra246, Bra813, Bra1402, and maternal chloroplast markers trnQ-rps16, psbA-trnH, and ycfb1 of a representative taxon set, including approximately 85% of all currently recognized species. Due to fossil absence, we further tested the impact of different tree priors, calibration methods (substitution rates vs. secondary calibration points), and different taxon sets on the topology as well as time spans of the Sisymbrieae clade. In addition, aiming for taxonomic stability, we also formally transferred Ochthodium aegyptiacum DC. to Sisymbrium and investigated the phylogenetic position of Sisymbirum aculeolatum and S. afghanicum. To achieve the latter and test the robustness of our time estimation, we extended the phylogenetic analysis and time divergence estimation to the whole Lineage II.

| Phylogenetic analyses of Lineage II and Sisymbrieae
The tribus Sisymbrieae is embedded into the Lineage II of the Brassicaceae family (BrassiBase: Kiefer et al., 2014;Koch et al., 2012Koch et al., , 2018 While the Lineage II phylogenies were based solely on ITS sequence information, the phylogeny of Sisymbrieae was based on two rDNA loci ITS and ETS, four low copy Bra loci and three cpDNA encoded loci (see first 55 lines in Appendix S1 for taxon sample).
Single-locus phylogenies were carried out, and the topologies were assessed using IQ-TREE (Nguyen et al., 2015). We concatenated the six nuclear DNA loci and three cpDNA loci, respectively, as no major topological incongruences were observed within the single loci.
However, as concatenation might lead to artificial topologies and bias statistical branch support (Kubatko & Degnan, 2007), we additionally tested both, nuclear DNA-based and cpDNA-based Sisymbrieae datasets using ASTRAL (Zhang et al., 2017), showing bootstrapping (BS). It was proven that this algorithm is statistically consistent under the multi-species coalescent model (Mirarab et al., 2014). Both IQ-TREE analysis as well as ASTRAL analysis were run under default settings, with increased 1,000 bootstrapped gene alignments and 1,000 bootstrapping trees, respectively. The topologies retrieved from Bayesian Inference and Maximum Likelihood were then compared with topologies retrieved from ASTRAL.

| Time estimation analyses of Lineage II and Sisymbrieae
To obtain the appropriate crown age of Sisymbrieae, we first dated the whole Lineage II, based on the same ITS-based taxon set as in the phylogenetic studies, one including the four tribes of the There are no reliable Brassicaceae fossils that could potentially serve as time constraints (Franzke et al., 2011), with a possible exception of East European Bunias fossil from the Pliocene (Mai, 1995).
Furthermore, in the light of general fossil absence, it is hard to establish a stable evolutionary time frame. Thus, we tested three different time constraint approaches, firstly using secondary calibration points retrieved from a whole plastome analysis  and as an alternative using the internal transcribed spacer substitution rate for herbaceous annual/perennial angiosperms of 4.13 × 10 -9 sub/ site/yr (Kay et al., 2006). The third approach included secondary calibration points as well as the published ITS substitution rate. While Huang et al. (2020) also published the crown age of Sisymbrieae, we did not incorporate this age into our dating analysis of the Lineage II, but used it as an independent checkpoint to estimate the accuracy of the ITS-based substitution rate calibration as a time constraint.
Furthermore, we wanted to test how different datasets might influ- All sequence evolution models used in this study were assessed using the Akaike information criterion (AIC) implemented in the jModelTest2 v2.1.6 (Darriba et al., 2012). All the analyses were carried out at the CIPRES Science Gateway computing facility (Miller et al., 2010). The aligned matrices are available as *.nex files upon request. Recent studies have shown that the impact of the tree priors in Bayesian phylogenetics is in general not as strong as previously thought (Ritchie et al., 2017;Sarver et al., 2019). Nevertheless, one should set the priors accordingly, as they might have an impact on the accuracy of the analysis, especially of those based on mixed inter-and intraspecies datasets (Ritchie et al., 2017). Thus, as all of our taxon samples-the Lineage II with and without the putative Eutremeae and Thlaspideae outgroups as well as Sisymbrieae dataset-were not limited to one accession per species only (which would automatically exclude the Coalescence tree prior) and to furthermore allow for potential extinction events (embedded into the Birth-Death tree prior, but not into Yule tree prior), all analyses were done in parallel using the Yule, Coalescence, and Birth-Death tree prior.

| Ancestral range reconstruction of Sisymbrieae
The time divergence tree from BEAST analysis performed under coalescent tree prior was used for the ancestral range reconstruction.
Distribution maps and contemporary range estimations followed the same methodology described in detail in Section 2.1. The analysis was carried out using RASP4 v4.0 ancestral state reconstruction tool (Yu et al., 2015(Yu et al., , 2020. DEC, DIVALIKE, and BAYAREALIKE biogeographic models with and without corresponding jumping parameter "+j" were tested using BioGeoBEARS v1.1.1 algorithm implemented in RASP through R (Matzke, 2013a(Matzke, , 2013b(Matzke, , 2014. coexist was set to three, following the widely distributed S. irio that can be found in maximally three out of eight areas specified above.

| RE SULTS
Details on individual alignments of Lineage II and Sisymbrieae can be inferred from Appendix S4. Based on the ITS sequence information, three investigated accessions (see first column in Appendix S1) were placed into morphologically similar genus Erysimum and were excluded from further analyses. After assessment of the genetic diversity of Sisymbrium based on ITS locus, a reduced representative taxon sample set of Sisymbrieae was used for all the subsequent analyses that covered the whole known genetic diversity within this tribus (see Materials & Methods section, for taxon sample refer to the first 55 lines in Appendix S1). The 10-15 sequenced clones per locus per accession consistently retrieved two major individual genetic copies that were included into all subsequent analyses (see Appendix S1, diverging clone copies depicted in *).  (Tables 1 and 2).

| Phylogenetics and divergence time estimation of Sisymbrieae
The separately concatenated nuclear DNA loci and cpDNA loci, respectively, showed only partially congruent results. The most prominent difference was a strong bias in the resolution    Table 3). Four geographically well-defined clades were retrieved from the analyses. The mean stem age of the disjunct forest Eurasian clade was dated to the Pliocene/Pleistocene, while its mean crown age was dated to early Pleistocene. The mean stem age of the South African clade was dated to the Pliocene, while its mean crown age was dated to the Calabrian in the Pleistocene (Figure 4). Furthermore, the mean stem age of the Caucasian clade was dated earlier to the late Miocene, while its mean crown age was dated as well to the Calabrian in the Pleistocene, and the mean stem as well as crown age of the Mediterranean clade was dated to the late Miocene ( Figure 4).

| Ancestral range reconstruction of Sisymbrieae
The most optimal model according to the AIC algorithm was DEC + j (Appendix S7). Ancestral range reconstruction placed the origin of the tribus Sisymbrieae into the western Irano-Turanian floristic region and Mediterranean (Figure 5a and 5b). In the latter, the highest number of speciation events i. e. 23 was inferred, followed by Euro-Siberian

TA B L E 3 Statistics of the crown and stem ages of geographically defined clades depending on the calibration method based on the dataset of Sisymbrieae
This study focuses on Sisymbrium L.-yellow-flowering herbaceous Brassicaceae, which grows predominantly in dry steppe, semidesert, and rural habitats of Northern hemisphere (Figure 1).

| Taxon sampling strategy and differences in the topology
Following the contemporary big scale phylogenies, genetic as well as cytological evidence, we excluded taxa Orychophragmus (Al-Shehbaz, 1985;Couvreur et al., 2010;German et al., 2009;Gomez-Campo, 1980;Liu et al., 2012;Lysak et al., 2007) and Sinalliaria (Kiefer et al., 2014;Koch et al., 2012Koch et al., , 2018  . The bold numbers next to nodes represent median ages in millions of years. Codes next to species names refer to the isolation codes from Appendix S1 Sisymbrieae in Walden et al. (2020)  to the topological incongruences, the nuclear-and chloroplast-based matrices in Sisymbrieae dataset were not concatenated and are presented as separate phylogenetic reconstructions in Figure 3 (ML and BI analyses) and Appendices S5 and S6 (ASTRAL analyses).
While the clades A (European S. strictissimum and East Asian S. luteum and S. yunnanense), B (South African clade), and C (Caucasian clade) were recognized already by Warwick et al., 2002, an additional geographically well-defined Mediterranean clade D was retrieved in this study (Figures 3 and 4, Appendices S5 and S6). The minor topological change in the placement of S. loeselii and closer relationship to S. erucastrifolium in the chloroplast-based phylogenies fits well with the morphological characteristics shared between these two species. Contrarily, major topological difference in the placement of Sisymbrium volgense might be attributed to its hybrid nature, where the progenitor not only genetically (this study) but also morphologically highly resembles the putative maternal lineage Sisymbrium polymorphum s.l. (Dorofeyev, 1997). Surprisingly, according to the nuclear signals, genetically distantly related Sisymbrium damascenum is placed in close proximity to the S. polymorphum s.l. and S. volgense in the chloroplast-based tree. This relationship cannot be explained neither by morphological similarities nor by geographical proximity and further genome-wide research would be necessary to irrefutably elucidate this peculiar topological incongruence.
The initial study of Mutlu and Karakuş (2015) associated

| Differences in the dating and retrieval of multiple copies
There have been some substantial differences in the employed calibration methods, datasets, and matrices. Huang et al. (2020) used a generally slowly evolving whole plastome datasets of a representative taxon sample of rosids, where reliable fossils are available, to constrain early nodes. Subsequently, they applied these now secondary constraints to the ITS-based tribe levels under the assumption that plastome and genome-based time estimations produce converging results (Hohmann et al., 2015;Huang et al., 2016). Contrarily, we applied the average ITS-based substitution rate for herbaceous annual/ perennial plant species on a reduced taxon sample, covering not all Brassicaceae tribes, but only Lineage II with and without the two putative outgroups (Eutremeae and Thlaspideae). There were some differences in the time spans of crown ages of Brassiceae, Isatideae, and Thelypodieae between Huang et al. (2020) and our analysis (Tables 1   and 2). These time frame shifts can be easily explained by different time constraint strategies with different backgrounds as well as by differences in the taxon sets. Interestingly, when only secondary calibration points from Huang et al. (2020) are used in our study, the crown age of Sisymbrieae does not correspond with the crown age inferred from Huang et al. (2020). This illustrates how much of an influence different taxon samples has on the retrieved time spans. Nevertheless, the crown age of Sisymbrieae remained relatively stable in the present study, regardless of the taxon sample and used constraints, indicating that at least within Sisymbrieae, secondary calibration based on plastome datasets indeed recovers highly congruent results with the ITS-based substitution rate calibration .
The whole Brassicaceae family is characterized by an α-WGD, which is only a recent one in the series of two additional older known WGDs that predated the split of Brassicaceae. These were followed by an extensive diploidization and shaped the evolutionary history of whole angiosperms and Brassicales, respectively (Barker et al., 2009;Fawcett et al., 2009;Franzke et al., 2011;Huang et al., 2020;Mandáková et al., 2017;Schranz et al., 2012). Recent studies have shown that approximately 40% of all Brassicaceae underwent an additional neopolyploidization event that has not yet been followed by diploidization (Hohmann et al., 2015;Kagale et al., 2014). Despite the fact that there is no evidence that Sisymbrium underwent a recent polyploidization event, several nuclear-encoded multi-copy loci (marked with * in the Appendix S1) in four species possessed two different copies, however, in all of the cases the copies clustered monophyletically. This indicates that the copies are probably remnants of polyploidization events and are not of hybrid nature. While there were always only two copies that were retrieved from the majority of the clones, individual SNPs, and chimeric copies were also detected (however, only in 1-2 clones out of 15 analyzed). These can be attributed to sequencing errors, paralogous nature of loci (as none of the investigated nuclear DNA loci were single copy), or cloning artifacts.

| Biogeography of Sisymbrium in South Africa
We hypothesize that Sisymbrium migrated to southern Africa from the Irano-Turanian floristic region through the East African Riff Mountains sometime in the Pliocene (stem ages dated to approximately 3.5-4.0 MYA: Figure 4 and Figure 5). During that time, this region was under strong influence of Pliocene uplift, which arguably had the most prominent effect on its present day topography (Partridge & Maud, 1987). During the Late Pliocene, an asymmetrical uplift of the whole continent and the consequent formation of the Post-African II erosion surface took place (Partridge & Maud, 1987).
These resulted in a development of a temperature as well as precipitation gradient. Middle and Late Pleistocene climate fluctuations have caused, together with the eustatic sea-level changes, a reduction and shift in the precipitation profile. During glacial cycles, the regions that nowadays experience summer rainfall were under winter rainfall regime (Hoag & Svenning, 2017;van Zinderen Bakker, 1978).
Cold air masses that were produced more often due to the development of the Antarctic polar cap were penetrating deeply into the African subcontinent (van Zinderen Bakker, 1978). During glacial periods, open dry grassland and semidesert environments were favored (Knight & Grab, 2016;Partridge & Maud, 1987) in which dry-adapted Sisymbrium species could thrive. This is also supported by our time estimation analysis, which placed the crown age of South African Sisymbrium species into the same period (Figure 4). Plant disjunctions between the Mediterranean, Irano-Turanian floristic region, and southern Africa are a common phenomenon tightly connected to the xeric conditions that connect all three areas (Goldblatt, 1978).
While the most common explanation for such a disjunction is the spread from southern Africa northwards (Désamoré et al., 2011;Durvasula et al., 2017;Klak et al., 2017;Moore et al., 2010;Valente et al., 2011), our dataset indicates that Sisymbrium occupied southern Africa from the Irano-Turanian floristic region, a more seldom pattern observed in only some other plant species (Carlson et al., 2012;Pyankov et al., 2002).

| Biogeography of Sisymbrium in the Mediterranean
The Mediterranean climate regime we know today was established around 3.2 MYA (Suc, 1984). Changes in the precipitation reduction and development of seasonality resulted in a compositional and structural change of the Mediterranean forests that lost subtropical elements and started to resemble contemporary Mediterranean forests with drought-adapted species (Kondraskov et al., 2015;Thompson, 2005). Similarly to the Eurasian steppe belt, the Mediterranean basin was also under strong influence of Pleistocene climatic oscillations causing recurrent range shifts that have shaped current species' distributions (Hewitt, 2011;Weiss & Ferrand, 2007).
Prior to the Sisymbrium migration into the Mediterranean, this basin was already under strong influence of gradual global cooling and aridification, which have been initiated as early as in the middle Miocene (van Dam, 2006;Zachos et al., 2008).  González-Sampériz et al., 2010). With its complex topography and relatively stable climate during the Late Quaternary, the Iberian Peninsula served as a Pleistocene refugial center for numerous plant species (Dubreuil et al., 2008;Heredia et al., 2007;Médail & Diadema, 2009;Pérez-Collazos et al., 2009) and could have potentially endorsed diversification of Sisymbrium species. High number of Sisymbrium species (up to 12, depending on the taxonomic treatments) on the Iberian Peninsula (Ball, 1964;Pujadas Salvá, 1993)

| Biogeography of Sisymbrium in the Irano-Turanian region & Caucasus
The western part of the Irano-Turanian floristic region is one of the global biodiversity hotspots of the Old World (Mittermeier et al., 2000) and a putative origin of several important crop wild relatives, including the family of Brassicaceae (Franzke et al., 2011;Hedge, 1976;Karl & Koch, 2013;Noroozi et al., 2019). Furthermore, due to its position connecting the eastern and western Eurasian floras, this region has also been proposed to be the main source of plant taxa for adjacent floristic regions, especially the Mediterranean (Herrera, 2010;Jabbour & Renner, 2012;Manafzadeh et al., 2014;Quezel, 1985;Thompson, 2005, this study).
This region has experienced a predominantly dry and stable climate since the early Eocene (Manafzadeh et al., 2017). Mountain ranges of Alborz, Zagros, Kopet Dagh, and Pamir had formed from the middle Miocene through the early Pliocene creating considerable habitat heterogeneity (Bobek, 1937(Bobek, , 1953Manafzadeh et al., 2017). Furthermore, climate stability of this region might also sustain genetic diversity, as it deters from climate-dependent extinction (Cowling et al., 2015;Galley et al., 2009;Schwery et al., 2015).
Thus, it is not surprising that western Irano-Turanian floristic region harbors approximately 27,000 species, and up to 40% of representatives that constitute this flora are endemic (Sales & Hedge, 2013;Takhtajan et al., 1986). This pattern can also be applied to Sisymbrium.
Fourteen Sisymbrieae species (approximately 30% of the whole tribe) occur in the western part of the Irano-Turanian floristic region, and half of them (sensu BrassiBase) are confined to this floristic region. Furthermore, ten individual speciation events within this area were inferred (third highest after Mediterranean) and the ancestral range reconstruction put the origin of this tribus somewhere into the Mediterranean and Irano-Turanian floristic region ( Figure 5), again supporting the hypothesis that this tribe (as well as the whole family) originated in the Irano-Turanian floristic region.
Adjacent to the Irano-Turanian floristic region is another biodiversity hotspot of the Old World-the Caucasus. Currently, there are around 7,500 plant taxa described from the Caucasus region, with 35% having a status of an endemic species (Gagnidze et al., 2002;Schatz et al., 2009), including four Sisymbrium taxa.
Palaeontological, palaeoclimatical, and genetic data point out that the Caucasus region was one of the main glacial refugial centers for fauna and dendroflora during the Quaternary together with Iberian, Italian, and Balkan Peninsula (Hewitt, 2000(Hewitt, , 2011Stewart et al., 2010). Furthermore, intermontane basins and high mountain plateaus of the greater Caucasus region were also recognized as a cryptic southern refugium for dry-and cold-adapted species during the interglacial periods (Fjellheim et al., 2006;Skrede et al., 2006). The exact extent of Pliocene and Pleistocene glaciations of the Caucasus is, however, unclear (Milanovsky, 2008;Mitchell & Westaway, 1999). Nevertheless, recent isotope dating studies and an overall high level of endemism in this region (Gagnidze et al., 2002;Nakhutsrisvili et al., 2017) suggest that the Caucasian glaciations were not as vast as previously thought.
Pliocene to Pleistocene transition had a tremendous effect on the vegetation composition of the Greater Caucasus area. With the decrease in temperature and humidity, the thermophilic forests flora was gradually replaced by a more robust temperate dendroflora without a modern analogue (Klopotovskaya, 1973;Naidina & Richards, 2016). Gradual intensification of the continental climate and reduced humidity caused an expansion of forest steppes and steppes through the Greater Caucasus area. Palaeovegetational analyses dated this expansion to be not older than 1.8 MYA (Connor & Kvavadze, 2009;Joannin et al., 2010;Messager et al., 2013;Naidina & Richards, 2016;Tagieva et al., 2013). This coincides with Sisymbrium clade endemic to Caucasus region, which started to diverge around the same time and its most recent common ancestor that invaded this area through East European Plain in the Pliocene (Figures 4 and 5).

| Biogeography of Sisymbrium in Eurasia
A peculiar relationship is observed between European S. strictissimum and its sister clade, consisting of East Asian S. luteum and S. yunnanense. All three species exhibit similar morphology (Vassilczenko, 1939;Zhou et al., 2001) and distinctive habitat preference. While most of the other Sisymbrium species are either widely distributed weeds or show a tendency toward open dry habitats, these three species can be found in more humid and predominantly shaded habitats in forests, thickets, ravines, or mountain slopes near water bodies (Brach & Song, 2006;Brandes, 1991). Assuming that the ecology of the last common ancestor of these species did not vary greatly, the expansion of the clade through the late Miocene/early Pliocene forests of central Asia is highly probable. During this era, most of Europe was covered in warm-temperate evergreen broadleaved and mixed forest, Western and Central Asia (between 45°N and 53°N) in temperate deciduous broadleaf forests, while the coastline of Eastern Asia was dominated by similar forest types as present in Europe, which transitioned into temperate deciduous broadleaved savanna biome in the inland (Bezrukova et al., 2003;Haywood et al., 2002;Ivanov et al., 2011;Pound et al., 2012). According to our study, the diversification into two distinct contemporary eastern (S. luteum and S. yunnanense) and western lineage (S. strictissimum) took place around early Pleistocene (Figures 4 and 5). This coincides with the era when the continuous forest belt started to fall apart due to the increased aridity and lower temperatures (Demske et al., 2002;Frenzel, 1968;Velichko, 2005).
Despite their wide distribution range, the fast-evolving ITS locus in S. loeselii and S. officinale remained monomorphic across the species' whole distribution area. While the latter showed no ambiguous signals in either of the investigated loci, all Sisymbrium loeselii accessions had to be cloned to retrieve two unambiguous ITS copies. Because the multi-copy ITS region is under strong influence of concerted evolution , the persisting two ITS copies might be an indicator of a recent hybridization event through which S. loeselii emerged. The young age is also supported by our time estimation analyses, placing the split between the two coexisting ITS copies into the western Euro-Siberian steppe of the middle Pleistocene (Figures 4 and 5). Both species are nowadays distributed worldwide and can be commonly found near human settlements, along roadsides and hedges, or growing on heavily nitrogen contaminated dry and loamy soils ( Malyschev & Peschkova, 2004). Exhibiting a weed-like ecology, these two species are excellent competitors that can successfully occupy wide ranges of disturbed habitats in a short period of time (Holzner & Numata, 2013). Therefore, quick and extensive repeated migrations through human migratory pathways might be accounted for the lack of geographically correlated genetic footprint even in the fast-evolving ITS region. Nevertheless, sound conclusions on this issue can be drawn based only on extensive nextgeneration sequencing datasets and proper sampling efforts.
In contrast to S. loeselii and S. officinale, some Sisymbrium species exhibit geographically correlated diverging ITS copies. Genetic splits found in widely distributed species were mostly of younger age indicating a potential influence of middle and late Pleistocene events. This era is characterized by an intensified development of permafrost and continental climate, major trans-as well as regressions of water bodies and major continental glaciations that got gradually reduced toward the end of the Pleistocene Velichko, 2005). These parame-  (Plenk et al., 2017), and Schivereckia (Friesen et al., 2020). This could likewise be the case in a two-way split in Sisymbrium altissimum, S. irio, S. volgense, or three-way split in predominantly Mediterranean S. erysimioides, all dated to the late or the end of middle Pleistocene. A slightly older split can be inferred from Sisymbrium heteromallum dating to the middle Pleistocene in the continental Asia, and an even older split between S. polymorphum accessions, dated to late Miocene/early Pliocene into Middle and Central Asia (Figures 4 and 5). The latter one is of great interest for two reasons. Firstly, Sisymbrium polymorphum is, as the epithet already suggests, morphologically as well as genetically an extremely variable species. This is also illustrated in our case, where S. linifolium is retrieved as a sister group to the eastern group of S. polymorphum s.l. accessions, rendering the S. polymorphum paraphyletic. This paraphyly also found in Chen et al., 2019, coupled with immense morphological variability of this species group, indicates a complex problem, which is out of the scope of this study. Secondly, Sisymbrium polymorphum is a widely distributed species found solely across the entire Eurasian steppe belt.
Assuming that its ecology has not changed extensively since the past makes the species a suitable proxy to infer past florogenetic patterns of the Eurasian steppe belt. Molecular signals in typical steppe plant species reflect the climate-landscape history of the steppe and thus allow for a finer resolution of the history of the steppe belt in comparison with floristic and fossil-based methods.
Overall, all these splits cannot be rigorously assigned to certain geological events due to a limited taxon sampling. Nevertheless, our study might serve as the starting point for other Sisymbriumspecific studies that investigate different aspects of the onset and development of different Eurasian geographical entities.

| Taxonomic update on Sisymbrium
In view of the above-mentioned paraphyly of Sisymbrium that embeds the single species of Ochthodium, the latter genus is synonymized here with the prior one. Before molecular phylogenetic studies, proximity of the two genera has never been assumed due to considerable differences in the fruit and seed characters (manyseeded dehiscent linear-cylindric or conical siliques with smooth papery valves and seeds with incumbent cotyledons in Sisymbrium versus two-seeded indehiscent ellipsoid to subglobose silicles with verrucose corky valves and seeds with accumbent cotyledons in Ochthodium) that were traditionally the key features in the systematics of the family. It is therefore understandable that Warwick Koch versus Draba L. of Arabideae DC. (Al-Shehbaz & Koch, 2003), several segregate genera versus Heliophila L. of Heliophileae DC. (Mummenhoff et al., 2005) or Tchihatchewia Boiss. versus Hesperis L.

ACK N OWLED G M ENTS
The authors thank all the curators of the following herbaria, who kindly provided the material for sequencing: B, MSB, GDA, GDAC, HBG, HEID, HUJ, K, L, P, MW, PRE, and WAG. Furthermore, we would also like to thank the "Plantarium" web-community members, who allowed us to use their photographs for Figure 1

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

E TH I C A L A PPROVA L
This article does not contain any studies with animals carried out by any of the authors.

SA M PLI N G A N D FI E LD S TU D I E S
The study was performed in compliance with the Convention on Biological Diversity (CBD).