Torreya jackii ( Taxaceae ) : A Special Species that is Genetically Admixed , Morphologically Distinct , and Geographically Sympatric with Parent Species

Torreya jackii Chun is an endangered species (Taxaceae) confined to a few localities in China. However, the species status of T. jackii within Torreya Arn. has not been clearly elucidated under a phylogenetic context. In this study, phylogenetic analyses based on the nuclear internal transcribed spacer (ITS) and amplified fragment length polymorphism (AFLP) indicated that T. jackii is closely related with a sympatric species T. grandis Fort. ex Lindl. that is present due to cultivation. However, analysis based on the concatenated sequences of seven chloroplast loci resolved T. jackii as the first branch within the genus. Given their overlapping distribution and synchronous blooming, we suggest that the plastid-nuclear incongruence was derived from the dilution of the nuclear genome of T. jackii by T. grandis via pollen-mediated introgression hybridization when the two species met due to cultivation. Introgressive hybridization is fairly common in plants but few cases have been recognized as independent species. Our study highlights the complexity of protecting endangered species and the need for caution to prevent the unreasonable expansion of economic crops into the distribution ranges of their wild relatives.


Introduction
Torreya Arn. is a genus in Taxaceae, which is distinguished by its completely enclosed seed within an aril and the production of two axillary female cones [1][2][3].Fossils of this genus are found widely across the northern hemisphere and the earliest has been dated to a Jurassic deposit in Europe [4].However, only six highly endangered members are extant and they have disjunct distributions in North America (NA) and eastern Asia (EA) [2].Two species, T. taxifolia Arn. and T. californica Torr., are endemic in NA, while four species are present in EA comprising T. nucifera (L.) Sieb.et Zucc., T. fargesii Franch., T. jackii Chun, and T. grandis Fort.ex Lindl.The first three species in EA are confined to a few localities in Japan or China, and the latter is cultivated due to its edible seed and oil in some southern provinces of China, where it overlaps the range with T. jackii (Figure 1) [1].
The morphological divisions within Torreya are not congruent with the geographical distribution of the species.This genus is divided into two sections in the only proposed infrageneric system [5,6].Sect.Ruminatae Hu is defined by deeply ruminated albumen and it contains two species from EA and one species from NA, whereas sect.Nuciferac Hu has only slightly ruminated albumen and it contains the remaining three species.It is unusual that species from the same continent are not morphologically approximated.Furthermore, despite their sympatry, T. jackii and T. grandis are morphologically distinct and ascribed to different sections due to differences in the albumen [5].Moreover, their leaves differ in length, where those of T. jackii are at least 10 cm whereas those of T. grandis are less than 6 cm [1].The morphological divisions within Torreya are not congruent with the geographical distribution of the species.This genus is divided into two sections in the only proposed infrageneric system [5,6].Sect.Ruminatae Hu is defined by deeply ruminated albumen and it contains two species from EA and one species from NA, whereas sect.Nuciferac Hu has only slightly ruminated albumen and it contains the remaining three species.It is unusual that species from the same continent are not morphologically approximated.Furthermore, despite their sympatry, T. jackii and T. grandis are morphologically distinct and ascribed to different sections due to differences in the albumen [5].Moreover, their leaves differ in length, where those of T. jackii are at least 10 cm whereas those of T. grandis are less than 6 cm [1].
The first comprehensive molecular phylogeny of Torreya reconstructed based on the nuclear internal transcribed spacer (ITS) contradicts the morphologically infrageneric division, but it is highly congruent with the geographical distribution [7] where the genera were resolved into two clades distributed in NA and EA.In addition, T. jackii and T. grandis were found to be closely related [7].However, a subsequent study based on chloroplast (cp) DNA obtained different phylogenies by resolving T. jackii as the most basal clade [8].Moreover, a second analysis based on ITS sequences that were mostly identical to those employed in the previous study found a different phylogeny [7,8].These discrepancies in terms of the geographical distribution, morphological division, genetic information from different genomes, and analyses based on different methods or samples suggest that the genus Torreya seems to have a complex history, especially the conflict of phylogenetic position of T. jackii in the genus.
In the present study, compared with previous studies [7,8] we increased the sample size and analyzed more genetic information from both the chloroplast and nuclear genomes in order to explore the phylogenetic relationships within Torreya, mainly to elucidate the underling mechanism responsible for the incongruent phylogenetic context of T. jackii according to different lines of evidence.

Sample Collections
Leaf samples were collected from at least two individuals of all six species of Torreya.T. jackii and T. grandis were the main focus of the present study, and we collected six individuals from four localities and three individuals from three localities for these two species, respectively.Amentotaxus argotaenia (Hance) Pilger and A. yunnanensis Li were selected as outgroup based on previous studies [9].Details of the samples, their voucher information, and GenBank accession numbers are listed in Table 1.The first comprehensive molecular phylogeny of Torreya reconstructed based on the nuclear internal transcribed spacer (ITS) contradicts the morphologically infrageneric division, but it is highly congruent with the geographical distribution [7] where the genera were resolved into two clades distributed in NA and EA.In addition, T. jackii and T. grandis were found to be closely related [7].However, a subsequent study based on chloroplast (cp) DNA obtained different phylogenies by resolving T. jackii as the most basal clade [8].Moreover, a second analysis based on ITS sequences that were mostly identical to those employed in the previous study found a different phylogeny [7,8].These discrepancies in terms of the geographical distribution, morphological division, genetic information from different genomes, and analyses based on different methods or samples suggest that the genus Torreya seems to have a complex history, especially the conflict of phylogenetic position of T. jackii in the genus.
In the present study, compared with previous studies [7,8] we increased the sample size and analyzed more genetic information from both the chloroplast and nuclear genomes in order to explore the phylogenetic relationships within Torreya, mainly to elucidate the underling mechanism responsible for the incongruent phylogenetic context of T. jackii according to different lines of evidence.

Sample Collections
Leaf samples were collected from at least two individuals of all six species of Torreya.T. jackii and T. grandis were the main focus of the present study, and we collected six individuals from four localities and three individuals from three localities for these two species, respectively.Amentotaxus argotaenia (Hance) Pilger and A. yunnanensis Li were selected as outgroup based on previous studies [9].Details of the samples, their voucher information, and GenBank accession numbers are listed in Table 1.

DNA Extraction, PCR Amplification and Sequencing
DNA was extracted from approximately 20 mg of leaves dried in silica gel using the cetyltrimethylammonium bromide (CTAB) method [10].Seven cp gene loci, rbcL, matK, trnL-trnF, psbB1-psbB2, rpl16, trnS-trnG, and trnD-trnT, and the nuclear internal transcribed spacer (ITS) were amplified with previously reported primers [11][12][13][14] according to the protocol described in our previous study [15].PCR products from the ITS was purified using an Agarose Gel DNA Purification kit (TakaRa Inc., Dalian, China) and then cloned with the pMD19-T vector (TakaRa Inc., Dalian, China) according to the recommended protocol, before transformation into competent Escherichia coli JM109.The transformed bacteria were screened overnight on solid Luria-Bertani medium containing 100 mg/mL ampicillin at 37 • C. Five or more positive clones were amplified and sequenced using the universal primers M13-47 and RV-M.For all the cp loci, the primers used for amplification were employed for sequencing, which was conducted at the Beijing Genomics Institute.New sequences have been deposited in GenBank under the accession numbers listed in Table 1.

Data Analyses
Three datasets were constructed where one comprised the nuclear ITS sequences, the second contained the concatenated sequences of the seven cpDNA regions, and the third comprised the AFLP bands.Multiple alignments of sequence data were obtained using ClustalX 1.83 [17] and refined manually with MEGA 5.0 [18].MEGA was also used to calculate the genetic distances according to the Kimura two-parameter model.The phylogenetic relationship was constructed using the maximum parsimony (MP) and maximum likelihood (ML) models implemented in PAUP 4.0 [19], and Bayesian inference (BI) in MrBayes 3.1.2[20].MP analyses were conducted by employing heuristic search to obtain the starting tree with stepwise, tree-bisection-reconnection (TBR) branch swapping, and the options of steepest descent, MulTrees, and Collapse.Bootstrap supports (BPs) were calculated by bootstrap analysis with 1000 replicates and the same settings as above.The optimal substitution models for the ML and BI analyses were found with ModelTest [21] based on Akaike's information criterion.The GTR+G+I model was selected for the seven combined cpDNA regions and the GTR+G model for the nuclear ITS gene.ML analyses were implemented via heuristic searches with 1000 replicates of random sequence addition, TBR branch swapping, and MulTrees.BPs were obtained using 1000 replicates by heuristic search with the same options.For BI, one cold and three hot Monte Carlo Markov chains were run twice for 2,000,000 generations, with sampling every 100 generations.Tracer v.1.5(http://tree.bio.ed.ac.uk/software/tracer/) was used to choose a suitable burn-in period.PAUP program was used to calculate a consensus tree and posterior probabilities from the sampled trees after the burn-in period.
The multi-locus profiles of amplified fragment length polymorphism (AFLP) were scored for the presence (1) or absence (0) of fragments measuring between 50 bp and 300 bp.A neighbor-joining tree was constructed based on the Nei and Li distance [22] from the presence or absence matrix using the TREECON 1.3b program [23].BPs obtained from 1000 pseudo-replicates were used as a measure of confidence in the reconstructed tree topology.

Results
The aligned lengths and variations in all the studied loci are listed in Table 2.The combined dataset of seven cp regions comprised 8241 bp, where 487 were parsimony-informative.The average genetic distances between species are listed in Table 3 (upper right).The overall genetic distance was 0.28% while those within T. grandis and T. jackii were both 0. One to two different ITS sequences were found in each species, with a total of 33 sequences and a combined length of 1152 bp, where 212 were parsimony-informative.The average genetic distances between species are listed in Table 3 (lower left).The overall genetic distance was 1.17% while those within T. grandis and T. jackii were 0.06% and 0.32%, respectively.The AFLP dataset comprised 813 bands and 623 were variable, where 507 were parsimony-informative.The MP strict consensus tree reconstructed with cpDNA datasets was largely consistent with those obtained by ML and BI in terms of the topology, as shown in Figure 2, where the support values from all three methods are indicated above the branches.The species T. jackii from EA branched off first with strong support, followed by the two NA species, T. taxifolia and T. californica, and the remaining Torreya species were resolved into two clades, which one contained T. yunnanensis and T. fargesii with strong support, and the other contained T. grandis and T. nucifera with moderate support.
The MP strict consensus tree (Figure 3) based on the nuclear ITS dataset recovered the same topology as the ML and BI trees.Compared with the tree obtained using the cp gene loci, the major difference was the position of T. jackii, which appeared in the base position according to the cp phylogeny, whereas it was closely related to T. grandis in the tree based on the ITS phylogeny.The AFLP data recovered an almost identical topology (Figure 4) to that using the ITS sequences, except T. grandis was resolved into polyphyly.The MP strict consensus tree (Figure 3) based on the nuclear ITS dataset recovered the same topology as the ML and BI trees.Compared with the tree obtained using the cp gene loci, the major difference was the position of T. jackii, which appeared in the base position according to the cp phylogeny, whereas it was closely related to T. grandis in the tree based on the ITS phylogeny.The AFLP data recovered an almost identical topology (Figure 4) to that using the ITS sequences, except T. grandis was resolved into polyphyly.

Discussion
Our phylogenetic analyses based on seven cp loci showed that T. jackii is the most basal branch of Torreya.The remaining species were resolved into two clades, where one comprised two species from NA and the other contained three Asian species, T. grandis, T. fargesii, and T. nucifera (Figure 2).Both ITS and AFLP, which are mostly representative of the nuclear genome, recovered a largely similar phylogeny (Figure 3), except that T. jackii was resolved as closely related to T. grandis.Among the six T. jackii sequences, two were identical to that in T. grandis whereas the other four differed by a distance of only 0.32%, which was much smaller than the overall mean distance (1.17%) for ITS sequences (Table 3).Incongruent phylogenies were also reported in a previous study based on different samples and analyses [7,8], thereby suggesting that technical issues such as insufficient taxon sampling, long branch attraction, or sequencing errors might not explain the incongruence [24].

Discussion
Our phylogenetic analyses based on seven cp loci showed that T. jackii is the most basal branch of Torreya.The remaining species were resolved into two clades, where one comprised two species from NA and the other contained three Asian species, T. grandis, T. fargesii, and T. nucifera (Figure 2).Both ITS and AFLP, which are mostly representative of the nuclear genome, recovered a largely similar phylogeny (Figure 3), except that T. jackii was resolved as closely related to T. grandis.Among the six T. jackii sequences, two were identical to that in T. grandis whereas the other four differed by a distance of only 0.32%, which was much smaller than the overall mean distance (1.17%) for ITS sequences (Table 3).Incongruent phylogenies were also reported in a previous study based on different samples and analyses [7,8], thereby suggesting that technical issues such as insufficient taxon sampling, long branch attraction, or sequencing errors might not explain the incongruence [24].A number of biological factors could explain the incongruence between the phylogenies based on nuclear and cp DNA, including convergent evolution, lineage sorting, and introgressive hybridization [25].Given our wide range of samples and the fact that we analyzed the cp and nuclear genomes, the first two explanations are unlikely to apply to T. jackii, but the last is difficult to reject [24].Introgressive hybridization is a process where genes are transferred via the formation of an initial F 1 hybrid that subsequently crosses with individuals from one or both of the parental species [26,27].The occurrence of this process requires that the two species meet some requirements including contact in space, overlapping blossoming periods, the same number of chromosomes, and the maternal inherence of cp DNA [26].We found that T. jackii and T. grandis may satisfy all these requirements.First, T. jackii inhabits southern China where T. grandis is cultivated (Figure 1).In most localities, the two species are well within the distance of pollen dispersal, which is mediated by the wind in both species.Second, the male plants from both species blossom from late March until middle April, and the females from late March until early April [28][29][30].Third, both species possess 11 chromosomes including one sex chromosome [31].Finally, maternal inheritance is documented in Torreya, although the genome is usually inherited by the paternal donor in gymnosperms [32].
Both parents could be ancestors of the backcross, but a pattern that is often seen in instances of introgression is the asymmetric transfer of genetic material [33].Therefore, one of the hybridizing lineages acts mainly as a donor and the other taxon as a recipient.It has been suggested that the quantity of pollen may be important and that pollen might be preferentially dispersed from a species with more pollen to one with less [34,35].T. grandis produce a much higher quantity of pollen than T. jackii due to cultivation.Therefore, the female individuals of T. jackii (actually, its ancestor) might have been fertilized by the pollen from T. grandis to yield the F 1 hybrid, before multiple backcrosses with T. grandis.Finally, the genetic material from the mother donor was diluted over the generations to yield T. jackii.
Introgressive hybridization usually occurs between the ranges of two species and it leads either to speciation by adapting to a habitat different from those of the parents or extinction via genetic assimilation [26,27].In each case, the offspring of hybridization are only a transient phase and they are seldom recognized as independent species.However, T. jackii is widely recognized as a species with little controversy, possibly due to its distinct morphology [1,2,36].We suggest that this species is actually quite special and different from an ordinary species because it is not genetically or geographically isolated from one of the parents, and its cp donor or maternal ancestor might have become extinct, possibly due to the cultivation of T. grandis within the range of T. jackii.Therefore, we advise caution to prevent the unreasonable expansion of economic crops into the distribution ranges of their wild relatives, thereby highlighting the complex issues involved with the protection of endangered species.

Conclusions
In the present study, we explored the phylogenetic relationships of Torreya species based on the chloroplast and nuclear genome data, found the plastid-nuclear incongruent phylogenetic position of endangered species T. jackii, and elucidated the introgression hybridization responsible for the incongruence due to the cultivation of sympatric species T. grandis.This finding implicates the complexity of protecting endangered species and the need for caution to prevent the unreasonable expansion of economic crops into the distribution ranges of their wild relatives.

Forests 2018, 9 , 11 Figure 1 .
Figure 1.Geographical distribution of the extant Torreya species.A, worldwide distribution (orange regions); B, distribution in Eastern Asia; C, distribution in North America.

Figure 2 .
Figure 2. Strict consensus of the most parsimonious tree obtained for Torreya based on the seven combined chloroplast regions.Numbers above the branches denote the bootstrap values based on 1000 replicates from the maximum parsimony (MP) and maximum likelihood (ML) analyses, and the posterior probabilities from Bayesian analyses.EA, Eastern Asia; NA, North America.

Figure 2 .
Figure 2. Strict consensus of the most parsimonious tree obtained for Torreya based on the seven combined chloroplast regions.Numbers above the branches denote the bootstrap values based on 1000 replicates from the maximum parsimony (MP) and maximum likelihood (ML) analyses, and the posterior probabilities from Bayesian analyses.EA, Eastern Asia; NA, North America.

Figure 3 .
Figure 3. Strict consensus of the most parsimonious tree obtained for Torreya based on the nuclear internal transcribed spacer (ITS).The other details are the same as those in Figure 2.

Figure 3 .
Figure 3. Strict consensus of the most parsimonious tree obtained for Torreya based on the nuclear internal transcribed spacer (ITS).The other details are the same as those in Figure 2.

Figure 4 .
Figure 4. Neighbor-joining tree based on the amplified fragment length polymorphism (AFLP) data.Numbers above the branches denote the bootstrap values based on 1000 replicates from MP.

Figure 4 .
Figure 4. Neighbor-joining tree based on the amplified fragment length polymorphism (AFLP) data.Numbers above the branches denote the bootstrap values based on 1000 replicates from MP.

Table 2 .
Statistic for the sequences used in this study.PI: parsimony-informative sites; CI: consistency index; RI: retention index.