Hybridisation and introgression in British Helosciadium (Apiaceae)

Reticulation between Helosciadium repens (Jacq.) W.D.J. Koch and H. nodiflorum (L.) W.D.J. Koch (Apiaceae) has been the source of much speculation, but until now supporting evidence has remained largely anecdotal. In the current study interspecific hybridisation and introgression between the two species was confirmed using DNA barcoding. The parentage of three putative hybrids collected from Port Meadow, Oxfordshire (UK) was determined using a maternally-inherited chloroplast marker (rps16-trnK) and two biparentally-inherited nuclear markers (LEAFYi2, ITS). Two of the individuals are early-generation hybrids between H. repens and H. nodiflorum, F1 or otherwise, while the third is most likely a backcross to H. repens. These individuals are the first confirmed hybrids/hybrid derivatives between the two parental species, and represent a new addition to the British flora. The hybrids closely resemble H. nodiflorum var. longipedunculatum F.W. Schultz, and in our view should be treated as H. × longipedunculatum (F.W. Schultz) Desjardins.

Helosciadium repens is a small plant that grows in open, wet places; it has a creeping habit and roots freely at every node (Stace, 2019). In horticultural conditions it behaves as a long-lived clonal perennial, but in the wild it usually behaves as an annual, or at best, a short-lived perennial that is killed off annually by prolonged winter flooding, summer droughts and/or heavy grazing by stock. These extreme seasonal conditions being an essential requirement for the maintenance of the open bare ground that is required by H. repens. The taxon is predominantly distributed in Western Europe, but is also found in parts of Central and Southern Europe, and there are isolated populations in North Africa (McDonald & Lambrick, 2006). It is in decline all over its European range (Lansdown, 2011) and the remaining wild populations are reported to have a narrow genetic base (Herden et al., 2019). The taxon was formerly scattered around the British Isles, being found in parts of Eastern England and the Scottish Lowlands (Stace, 2019), but it is now restricted to just two sites, one at Port Meadow, Oxfordshire (v.c.23) and another at Walthamstow Marshes, Essex (v.c.18). An introduced population is also present at North Hinksey Meadow, Oxfordshire (v.c.23;JNCC, 2013;Stroh et al., 2016). Helosciadium nodiflorum is much more common than H. repens. It is widely distributed in Europe, often being found wherever there is suitable habitat (Tutin, 1980). It is also much more variable, displaying high levels of phenotypic plasticity and presenting in a wide variety of forms, depending on the environmental conditions. The typical upright form, var. vulgare F.W. Schultz, is easy to distinguish, but smaller creeping forms, such as var. pseudo-repens H.C. Watson and f. simulans Ridd., can resemble H. repens and be mistaken for it in the field (Riddelsdell & Baker, 1906;Riddelsdell, 1914b). This overlap between the two species has led to ongoing identification problems for field botanists, particularly in the absence of flowering material, and has led some to question whether H. repens is even resident to the British Isles (Tutin, 1962). Furthermore, a number of common garden experiments have suggested that British plants appearing to be H. repens in the wild, tend to revert to a more typical H. nodiflorum state in cultivation (Druce, 1927;Grassly et al., 1996). However, a genetic study using random amplified polymorphic DNA (RAPD) markers was able to confirm the presence of pure H. repens at Port Meadow. The genetic profiles of British H. repens clustered with reference populations from central Europe and, when cultivated, retained the distinctiveness of their field morphology (Grassly et al., 1996). Helosciadium repens and H. nodiflorum therefore appear to be closely-related, but distinct species and can be separated using a combination of characters (H. nodiflorum vs. H. repens): (i) leaflets longer than wide vs. as long as wide; (ii) leaflets apically acute vs. asymmetrically bifid (i.e. unequally lobed); (iii) peduncles shorter than rays vs. longer than rays; (iv) bracts 0-2 vs. 3-7; and (v) fruits longer than wide vs. wider than long (Riddelsdell & Baker, 1906;Tutin, 1980;Stace, 2019;O'Mahony in Stace et al., 2015).
However, discrimination of the two species is further complicated by accounts of putative interspecific hybrids. Riddelsdell (1917b) Druce (1928) made the combination Apium × riddelsdellii Druce nomen nudum. However, when Tutin (1975) examined these specimens he regarded them as only variants of H. nodiflorum. Walters (1980) presented a collection from Chippenham Fen, Cambridgeshire (v.c.29), which resembled a small, creeping H. nodiflorum, but with poor pollen and without ripe fruits. It was initially thought to be a possible example of H. repens × H. nodiflorum, but was eventually revealed to be × Beruladium procurrens -the intergeneric hybrid between B. erecta and H. nodiflorum (Desjardins et al., 2015). Crackles (1976) reported a number of putative H. repens × H. nodiflorum specimens from Hornsea Mere, Yorkshire (v.c.61); similar to H. repens, but with less than three bracts at the base of most umbels, and apparently sterile fruits.
Previous accounts of H. repens × H. nodiflorum, while tantalising, are based on morphological determination alone, and have not been verified by alternative methods (e.g. secondary chemistry, cytology, DNA barcoding etc.), neither have they been compared with artificially resynthesised specimens (Tutin, 1975). The current study therefore aimed to unequivocally determine the parentage of three putative hybrid specimens collected from Port Meadow, where the prospective parental species grow together. Sequence data from three gene regions were used: rps16-trnK, the ITS and LEAFYi2. rps16-trnK, an intergenic spacer, is a maternallyinherited chloroplast marker and was used to identify the female parent. The ITS, ribosomal DNA (rDNA), and LEAFYi2, an intron of a low-copy nuclear gene (LCNG), are biparentally-inherited nuclear markers and were used to identify the male and female parents.
The putative hybrids were identified in the field and collected as follows. In July 2001, A.G.S collected RXN1 from SP 500 078, which was displaying the leaflet and petiole characters of H. nodiflorum with the bract and peduncle characters of H. repens (O'Mahony in Stace, 2015). In August 2014 J.A.W collected RXN2 from amongst a population of H. repens at SP 500 075 with H. nodiflorum growing nearby. This specimen resembled H. repens in the field, but was generally bigger with larger-lobed leaflets (Fig. 1). In October 2017, J.A.W collected RXN5 from SP 499 078, observing that it was clearly different from the surrounding H. repens.

DNA extraction, amplification and sequencing
Total genomic DNA (gDNA) was isolated from dried leaf material (20 mg) using the DNeasy Plant Mini Kit (Qiagen). rps16-trnK and the ITS were amplified by PCR, purified and sequenced as per Desjardins et al. (2015). LEAFYi2 was amplified with the primers LFsxl1-2 (5' CAC CCA CGA CCI TTY ATI GTI ACI GAR CCI GGI GA 3') and LFtxr (5' (Frohlich & Meyerowitz, 1997). Samples that gave mixed signals were also sequenced from clones. Cloning was conducted using the pGEM ® -T Easy Vector System (Promega) and α-Select Competent Cells taken from E. coli (Bioline).
Recombinant plasmids were selected for by blue-white screening and the size of the insert determined by colony PCR with M13 primers. Plasmid DNA was isolated from cell cultures using the E.Z.N.A. ® Plasmid Mini Kit (Omega Bio-tek). A minimum of five colonies were sequenced per accession. Sanger sequencing reactions were outsourced to GATC Biotech (Konstanz, Germany).

Phylogenetic analysis
Generated sequence reads were viewed, trimmed and blasted with Geneious R7 (created by Biomatters; available from http://www.geneious.com/). Additional sequences were downloaded from the GenBank database (Supplementary information 1). Sequences were aligned using the Clustal W algorithm, and adjusted by eye. Length-mutational events (indels) were incorporated for the analysis of LEAFYi2 using a simple gap coding method (Simmons & Ochoterena, 2000). Copies acquired from the putative hybrid specimens were investigated by direct sequence comparison with reference taxa and by phylogenetic analysis. Maximum parsimony (MP) analysis was conducted on sequence data using PAUP* 4.0 (Swofford, 2002).
Hedera helix L. (Araliaceae) was used as the OUTGROUP for MP analysis of rps16-trnK and the ITS, but no sensible alignment could be made with ingroup taxa for LEAFYi2, so Sium latifolium L. and B. erecta were used instead. Topology searches were carried out using a branch and bound algorithm with the addition method FURTHEST. Bootstrapping = 1000 replicates.

Results
Chloroplast DNA marker None of the rps16-trnK sequences from the three putative hybrids was identical (99.02 -99.76 % identity), indicating three distinct haplotypes. RXN1 clustered with H. repens in a weakly-supported clade (60% bootstrap support, BS; Fig. 2  Internal transcribed spacer The ITS ribotypes of H. repens and H. nodiflorum differ at five polymorphic sites. A direct sequence comparison of the ITS ribotype of putative hybrid RXN1 matched H. repens at all five of these sites, and the phylogenetic analysis placed it in a weaklysupported H. repens clade (61% BS; Fig. 3). When sequenced directly (5'  3') the ITS sequences of putative hybrids RXN2 and RXN5 were visibly heterozygous at all five of these polymorphic sites, and subsequent gene cloning detected the presence of two distinct copies in both individuals, which were designated copy 1 and 2. Copy 1 copies possessed the H. nodiflorum character state at all five polymorphic sites, and the phylogenetic analysis placed them in a strongly-supported H. nodiflorum clade (96% BS). Copy 2 copies possessed the H. repens character state at all five polymorphic sites, and the phylogenetic analysis placed them in the H. repens clade.
Putative hybrids RXN2 and RXN5 therefore possess two divergent forms of the ITS, one matching H. nodiflorum and another matching H. repens, while putative hybrid RXN1 is apparently homozygous and possesses only a single form of the ITS, which matches H. repens.  Fig. 4). A direct sequence comparison of copy 2 copies matched H. nodiflorum, and the phylogenetic analysis placed them in a strongly-supported H. nodiflorum-clade (96% BS). The putative hybrids therefore all appear to possess two divergent copies of the LCNG LEAFYi2, one originating from H. repens and another from H. nodiflorum.

Molecular Confirmation
In the case of RXN2 and RXN5 both biparentally-inherited nuclear markers (LEAFYi2 and the ITS) revealed contributions from H. nodiflorum and H. repens, with the maternally-inherited chloroplast marker (rps16-trnK) identifying H. nodiflorum as the female parent. RXN2 and RXN5 therefore appear to be early-generation hybrids between H. nodiflorum and H. repens. Interestingly, while the direction of hybridisation is the same, RXN2 and RXN5 possess distinct chloroplast haplotypes and appear to have arisen independently.
In the case of RXN1 the ITS and rps16-trnK sequences matched H. repens, with no apparent contribution from H. nodiflorum, but its LEAFYi2 sequence revealed contributions from both species. RXN1 is therefore most likely a backcross to H. repens, for while it retains both parental copies of the LCNG, LEAFYi2, it has seemingly lost the H. nodiflorum copy of the ITS. Discrepancy between nuclear datasets, e.g. rDNA and LCNGs, is not uncommon and can be due to differential patterns of inheritance (Small et al., 2004). In first-generation hybrids nuclear markers are typically inherited in a predictable, biparental and additive fashion but, in the case of repetitive DNA, such as the ITS, additional copies can be lost from generation to generation. This is because repeat-units tend to evolve in unison via mechanisms of concerted evolution, e.g. gene conversion, unequal crossovers etc. (Baldwin et al., 1995). In the wake of a reticulate event, two divergent copies, initially present in the amalgamated genome, can be lost if the sequence becomes homogenised to a single parental type (Álvarez & Wendel, 2003). Homogenisation can be bidirectional (Wendel et al., 1995), but in backcrosses it typically occurs in the direction of the recurrent parent (Aguilar et al., 1999). LCNGs, like LEAFYi2, are thought to be less likely to undergo concerted evolution than rDNA, and can maintain the signal of reticulation even when it has been lost from rDNA markers (Small et al., 2004). This appears to be the case here and in RXN1 the signal of hybridisation, while apparently lost from the ITS sequence, is preserved in LEAFYi2. This trace would have been entirely overlooked if only the ITS had been relied upon, as is so often the case in phylogenetic and hybridisation studies (Álvarez & Wendel, 2003), and highlights the importance of analysing additional nuclear datasets.

Taxonomy
The hybrid H. repens × H. nodiflorum appears to be a permanent fixture of the British flora; having arisen independently in a number different locations (Crackles, 1976;Stace et al., 2015). A validly published hybrid binomial is therefore warranted, and is in line with other hybrid combinations in this genus (e.g. Riddelsdell, 1914a;Desjardins et al., 2015;Rita et al., 2016). Druce (1928) proposed the name Apium × riddelsdellii, but this is invalid as it was published without description. A number of described varieties/forms of H. nodiflorum are also said to possess intermediate features between H. repens and H. nodiflorum (e.g. var. ochreatum (DC.) DC., var. pseudo-repens, var. longipedunculatum F.W. Schultz).
The best candidate among these is H. nodiflorum var. longipedunculatum, which has been linked with the hybrid by a number of eminent botanists (e.g. Rothmaler, 1963;Stace, 1997). This variety was originally recognised by Schultz (1854), and later described in further detail by Riddelsdell & Baker (1906) who examined a number of specimens from Duddingston Loch, Edinburgh (v.c.83) and Gullane Links, East Lothian (v.c.82). While Riddelsdell (1917a) never regarded this variety as the cross between H. repens and H. nodiflorum, preferring his own candidates from Binsey Common (Riddelsdell, 1917b), it appears to be a strong contender for a number of reasons. (i) Its leaflets are somewhat intermediate between H. repens and H. nodiflorum, being ovate to broadly ovate, coarsely serrate and occasionally lobed; (ii) its peduncles resemble those of H. repens, being typically longer than the rays of the umbel; and (iii) an involucre of 1-3 bracts is always present, a rare thing in pure H. nodiflorum (Riddelsdell & Baker, 1906 BM001187310, BM001154374). In both cases the leaflets are ovate to broadly ovate, with or without lobes, peduncles are always present and are equal to or longer than rays; bracts are also present at the base of most umbels. Ancient DNA analysis of the type specimens would be definitive but, given the age and historical value of the specimens, destructive sampling was not considered. Nevertheless, morphological comparison alone has convinced us that var. longipedunculatum does most likely represent the hybrid between H. repens and H. nodiflorum, in line with Rothmaler (1963) and Stace (1997).
There is a valid hybrid binomial under Apium (see Rothmaler, 1963) (1854), did not designate a holotype, neither did he give a precise locality nor cite the collector of the specimens he procured. However, he writes that the material was originally collected from the Edinburgh area and that he received it by way of London, presumably on loan from the British Museum. He further notes that the consignment consisted of two specimens, affixed to a single sheet, and that they were catalogued under the name H. repens Koch. It was from these two specimens that Schultz (1854)  revealed the specimens in question -BM001144095 (Fig. 5.). Two small examples, adjacent to one another, collected by G. Lawson from Gullane Links in July 1845, and archived under the name H. repens Koch, which are almost certainly the original syntypes. We here designate the larger of the two (left) as the lectotype. Note that in the current mounting these two specimens occupy only the top right hand corner of a sheet, the rest being taken up by another larger specimen (BM001144094). This third specimen was also collected from Gullane Links, by G. Don (date unknown), and Riddelsdell & Baker (1906) cited it as a further example of var.
longipedunculatum. However, the label reads 30. Sium repens Jacq., and Schultz (1854) makes no mention of such a specimen. The conclusion being that Schultz did not have this additional specimen to hand when making his original description of var. longipedunculatum, and that the two collections were combined at a later date.
For ease of viewing this additional specimen has been cropped out of