Origin and Timing of the Horizontal Transfer of a Pgic Gene from Poa to Festuca Ovina

Available online at www.sciencedirect.com

Molecular Phylogenetics and Evolution 46 (2008) 890–896 www.elsevier.com/locate/ympev

Origin and timing of the horizontal transfer of a PgiC gene from Poa to Festuca ovina

Pernilla Vallenback *, Maarit Jaarola, Lena Ghatnekar, Bengt O. Bengtsson

Department of Cell and Organism Biology, Lund University, Genetics Building, So¨lvegatan 29, SE-223 62 Lund, Sweden

Received 30 January 2007; revised 19 July 2007; accepted 26 November 2007 Available online 28 January 2008

Abstract

A segregating second locus, PgiC2, for the enzyme phosphoglucose isomerase (PGIC) is found in the grass sheep’s fescue, Festuca ovina. We have earlier reported that a phylogenetic analysis indicates that PgiC2 has been horizontally transferred from the reproductively separated grass genus Poa. Here we extend our analysis to include intron and exon information on 27 PgiC sequences from 18 species representing five genera, and confirm our earlier finding. The origin of PgiC2 can be traced to a group of closely interrelated, polyploid and partially asexual Poa species. The sequence most similar to PgiC2 is found in Poa palustris with a divergence, based on synonymous substitutions, of only 0.67%. This value suggests that the transfer took place less than 600,000 years ago (late Pleisto- cene), at a time when most extant Poa and Festuca species already existed. Ó 2007 Elsevier Inc. All rights reserved.

Keywords: Horizontal gene transfer; Festuca ovina; Poa palustris; Cytosolic phosphoglucose isomerase; PgiC2

1. Introduction transposable elements. For example, Mutator-like elements, MULEs, are widespread among plants (Turcotte Eukaryotic organisms normally reproduce sexually and et al., 2001; Yu et al., 2000) and are known to sometimes horizontal transfer of genes, i.e. movement of genes incorporate nuclear fragments (Jiang et al., 2004; Yu between sexually well-separated species, is rare (Andersson, et al., 2000). The ﬁrst unequivocal evidence of horizontal 2005; Keeling et al., 2005; Martin, 2005). During the last transfer of nuclear DNA in plants was recently reported years, some examples of horizontal gene transfer have, by Diao et al. (2006), who demonstrated that a MULE- however, been described between reproductively isolated sequence has been transferred between rice (Oryza)and plant species (Bergthorsson et al., 2003; Richardson and millet (Setaria) long after the separation of these lineages. Palmer, 2007). The majority of these studies concern genes However, as far as reported no active nuclear gene is asso- encoded by the mitochondria. Davis and Wurdack (2004), ciated with this transfer. for example, have described an instance of horizontal We have previously characterized an additional, second transfer of mitochondrial genes between two species in a locus for a functional nuclear gene speciﬁc for the grass host–parasite relationship, while Bergthorsson and col- sheep’s fescue, Festuca ovina (Ghatnekar et al., 2006). leagues (2004) have provided evidence for a massive trans- The gene, a cytosolic phosphoglucose isomerase denoted fer of mitochondrial genes into the basal angiosperm PgiC2, generates functional dimeric enzymes with the stan- Amborella trichopoda. dard locus of the species, PgiC1 (Ghatnekar, 1999). A phy- When horizontal transfer of nuclear genetic material has logenetic analysis based on exon sequences of 1.2 kb in been looked for in plants, most attention has been given to length indicated that PgiC2 has introgressed into F. ovina from the distant genus Poa (Ghatnekar et al., 2006). Two * Corresponding author. Fax: +46 46 147874. facts support the interpretation that this transgression is E-mail address: [email protected] (P. Vallenback). due to a recent horizontal transfer. First, the grass genera

Poa and Festuca belong to different subtribes within Poeae some numbers, and the exons and introns sequenced. The (Poaceae) and are completely reproductively separated; in species were chosen based on our earlier results on fact, no sexual hybrid between them has ever been recorded sequence similarities and differences with PgiC2 (Ghatne- (Hegi, 1996; Knobloch, 1968). Second, PgiC2 is not fixed in kar et al., 2006). All plants were collected at sites in south- F. ovina but segregates in the species, with no trace of an ern Sweden. inactive gene copy in plants not expressing PgiC2. One sequence was obtained from each diploid species, Here we present a more detailed phylogenetic analysis of while two different gene copies were sequenced from each PgiC2. We include additional Poa species and base our plant of the four polyploid Poa species used in the analysis most complete study on a 3.7 kb long sequence containing (the very variable species P. pratensis was represented by both exons and introns. With this new information we do two plants). Thus, sequences from diploid plants are desig- not only corroborate our earlier findings, but can show nated without any numerals, while sequences from poly- from where within the speciose Poa genus the PgiC2 gene ploid species are designated first with the plant’s number has come. In addition, we estimate that the horizontal and then with the sequence’s number, see Table 1. Parts transfer took place in late Pleistocene, at a time when most of some of the sequences listed in Table 1 have earlier been extant Poa and Festuca species already existed. reported by Ghatnekar et al. (2006). Five of the analyzed species belong to the fine-leaved 2. Materials and methods festucoids, Aira praecox, F. tracophylla, F. rubra, F. polesica and F. ovina, while four belong to the broad-leaved 2.1. Plant material festucoids, F. altissima, F. pratensis, F. arundinacea and Lolium perenne (Catalan et al., 2004). Of the eight Poa spe- A total of 27 sequences of the gene for cytosolic phos- cies analyzed, P. annua, P. chaixii, P. trivialis, P. supina, P. phoglucose isomerase, PgiC, were investigated from plants nemoralis, P. palustris, P. pratensis and P. angustifolia, the belonging to five genera within subfamilies Poeae and Bro- last four belong to a well-characterized group of closely meae of the grass family Poaceae. Table 1 summarizes the interrelated, partially apomictic polyploids (Patterson species and subspecies analyzed, their recorded chromo- et al., 2005).

Table 1 Data on the 27 sequences used in the analyses Species, subspecies Plant/sequence Chromosome number Sequenced parts Aira praecox 1–1 14 Exons 5–10, 16–21 Bromus sterilis 1–1 14 Exons 5–10, 16–21 Festuca altissima 1–1 14 Exons 5–10, 16–21 Festuca arundinacea 1–1 14 Exons 5–10, 16–21 Festuca ovina PgiC1 a 14 Exons and introns 5–21 Festuca ovina PgiC2 b 14 Exons and introns 5–21 Festuca polesica 1–1 14 Exons 5–10, 16–21 Festuca pratensis 1–1 14 Exons 5–10, 16–21 Festuca communtata 1–1 14 Exons 5–10, 16–21 Festuca rubra litoralis 1–1 14 Exons 5–10, 16–21 Festuca rubra rubra 1–1 14 Exons 5–10, 16–21 Festuca tracophylla 1–1 14 Exons 5–10, 16–21 Lolium perenne 1–1 14 Exons 5–10, 16–21 Poa angustifolia 1–1 46–63, 72 Exons and introns 5–21 Poa angustifolia 1–2 46–63, 72 Exons and introns 5–21 Poa annua 1–1 28 Exons 5–10, 16–21 Poa chaixii 1–1 14 Exons and introns 5–21 Poa nemoralis 1–1 28, 35, 42, 49–70 Exons and introns 5–21 Poa nemoralis 1–2 28, 35, 42, 49–70 Exons and introns 5–21 Poa palustris 1–1 21, 28–42 Exons and introns 5–21 Poa palustris 1–2 21, 28–42 Exons and introns 5–21 Poa pratensis 1–1 28–124 Exons and introns 5–21 Poa pratensis 1–2 28–124 Exons and introns 5–21 Poa pratensis 2–1 28–124 Exons and introns 5–21 Poa pratensis 2–2 28–124 Exons and introns 5–21 Poa supina 1–1 14 Exons 5–10, 16–21 Poa trivialis 1–1 14 Exons and introns 5–21 Chromosome number according to Lid (1979). a The earlier described sequence d from locus PgiC1 (see Ghatnekar et al., 2006). b The earlier described sequence c from locus PgiC2 (see Ghatnekar et al., 2006). 892 P. Vallenback et al. / Molecular Phylogenetics and Evolution 46 (2008) 890–896

2.2. DNA isolation, PCR amplification, cloning and DNA structed from multiple equally parsimonious trees. The ML sequencing analyses were carried out using the heuristic search approach and ‘‘as is” as well as ‘‘random” addition repli- Leaf material was ground to powder in liquid nitrogen cate. The hierarchical likelihood ratio test (hLRT) and and total genomic DNA was isolated using the Qiagen the Akaike information criterion (AIC), implemented in DNeasy Plant Mini Kit. PCR was performed using stan- the computer program MODELTEST version 3.7 (Posada dard techniques. Taq DNA polymerase (Roche) was used and Crandall, 1998), were used to establish the most appro- in all nested PCR amplifications and Expand Long Tem- priate model of DNA substitution for the NJ and ML anal- plate PCR System (Roche) was used for cloning purposes. yses. The MODELTEST analyses were performed with and Primers were constructed using Oligo version 6.0 without outgroup. We also used a simpler substitution (MedProbe). Information on primers can be obtained from model, the Kimura 2-parameter method (K2P; Kimura, the authors upon request. 1980), for comparison. Relative stability of MP, ML and Due to selfing or lack of population variation, PgiC NJ phylogenetic trees was assessed with bootstrap analyses sequences could be obtained from most plants by direct using 10,000 replicates, with exception of the ML analysis sequencing of PCR products. However, cloning was on the first data set which was based on 500 replicates. required to obtain sequences from the polyploid Poa spe- Estimates of the number of substitutions per synony- cies. About 5.0 kb of PgiC genes were amplified in a single mous and nonsynonymous site, KS and KA, were calculated PCR. Standard procedures were used for cloning, utilizing using DnaSP version 3 (Rozas et al., 2003). the pGEMR-T Easy Vector System (Promega), high effi- We tested the PgiC exon sequences for a molecular clock ciency competent cells (Epicurian colir XL blue, AHdiag- by comparing the log likelihood scores of ML trees con- nostics), and agarose plates with ampicillin for positive structed with and without a molecular clock constraint selection. Five colonies were picked from each cloned plant (Felsenstein, 1988). and the DNA of each colony was amplified with the vector specific primers SP6 and T7. Two different alleles were 3. Results sequenced from each plant (see Table 1). PCR products were purified using the QIAquick PCR All sequences analyzed shared the exon–intron structure Purification Kit (Qiagen). The Big Dye Terminator cycle of PgiC1 in F. ovina, as described by Ghatnekar et al. sequencing kit (Applied Biosystems) was used for sequenc- (2006). The sequence data have been deposited in GenBank ing and resulting products were run on an ABI 3100 capil- (Accession Nos. EU379314-EU379327). lary automated DNA sequencer. To determine the origin of PgiC2, two phylogenetic studies were performed. The first study was based on a data 2.3. Sequence and phylogenetic analyses set consisting of 27 sequences of 912 nucleotides from exons 5 to 10 plus exons 16 to 21 (see Table 1). NJ, MP Sequences were aligned, inspected and edited using and ML analyses were performed using Bromus sterilis as Sequencher version 4.5 (Genes Codes Corp.). Ambiguous outgroup. The three phylogenetic methods employed gen- bases were resolved by eye. Exon/intron borders (always erated highly similar tree topologies, and the NJ and ML GT ...AG) were determined using information from previ- algorithms recovered the same basic topology independent ously studied F. ovina alleles (Ghatnekar et al., 2006). Con- of substitution model used. The ML tree is shown in Fig. 1. catenated sequences were constructed from contigs for It is based on the model that gave the highest log likelihood exons 5–10 (498 bp) and exons 16–21 (414 bp; exon 16 was score: GTR with a gamma shape parameter (G) of 0.7215 only sequenced in its distal part and exon 21 in its proximal and a frequency of invariable sites (I) estimated at 0.4257. part). Thus, our first data set was based on exon data only The separation of the festucoid sequences (Catalan et al., and contained 27 sequences of altogether 912 bp. 2004) from the Poa sequences is strongly supported in all A second data set with information from both exons and of the NJ, MP and ML analyses. The PgiC2 sequence falls introns was constructed from 12 Poa sequences plus F. ovi- well nested within the Poa lineage, whereas the PgiC1 na PgiC1 and PgiC2. Data spanned from exon 5 to the sequence clusters, as expected, with the fine-leaved fescues. proximal part of exon 21. After manual exclusion of indels For a more detailed study of the origin of PgiC2, a sec- in introns, a total aligned data matrix of 3698 bp was ond data set was created. The aligned data matrix included obtained. Data on nucleotide substitutions and amino acid 14 sequences of 3698 consecutive nucleotides, now includ- replacements were assembled using MacClade version 4.06 ing introns, ranging from exon 5 to exon 21 (Table 1). Data (Maddison and Maddison, 2000). were taken from PgiC1, PgiC2 and the twelve Poa Phylogenetic relationships among PgiC sequences were sequences that clustered closely with PgiC2 in the tree pre- inferred using neighbor-joining (NJ), maximum parsimony sented in Fig. 1. This data set was analyzed with and with- (MP) and maximum likelihood (ML) algorithms imple- out F. ovina PgiC1 as outgroup. Rooted and unrooted trees mented in PAUP version 4.0b10 (Swofford, 1998). The recovered the same topology, and the NJ, MP and ML MP analyses were carried out using the branch-and-bound analyses gave highly similar results irrespective of substitu- option. Strict and 50% majority consensus trees were con- tion model used. Fig. 2 shows the rooted ML tree based on P. Vallenback et al. / Molecular Phylogenetics and Evolution 46 (2008) 890–896 893

Fig. 1. Phylogenetic relationship among 27 PgiC sequences. Maximum likelihood tree based on 912 bp exon data (for details see text). Bootstrap values are for neighbor-joining, maximum parsimony and maximum likelihood methods, respectively. Sequences from diploid species are specified by the name of the species. From each polyploid plant two different sequences were derived; they are specified by the plant’s and the sequence’s number (see also Table 1). The sequences from the standard gene in F. ovina, PgiC1, and the horizontally transferred gene, PgiC2, are indicated with arrows. the model that generated the highest log likelihood score: alis—in Fig. 2 all three analytical methods give 100% boot- TVM+G (G = 0.5479). In accordance with earlier studies strap support to the branch with PgiC2, P. palustris 1–2 (Patterson et al., 2005), the sequences from the polyploid and P. nemoralis 1–1. However, not even the extensive sec- species P. pratensis, P. nemoralis, P. angustifolia and P. ond data set incorporating information also from introns is palustris show intricate genealogical relationships that rich enough to unequivocally determine the pattern of reflect their complex genomic relationships. branching within this clade since here the three methods As to the origin of PgiC2, the two studies represented by of analyses give different results. In the NJ analysis P. Figs. 1 and 2 differ in their details but agree with respect to palustris 1–2 is the sequence most closely related to PgiC2 their main results: (i) The PgiC2 sequence groups with with a bootstrap support of 100%, while in the MP analysis sequences from the polyploid Poa species; the sequences P. nemoralis 1–1 falls closest to PgiC2.InFig. 2, based on from the diploid species P. chaixii and P. trivialis plus the ML analysis no clear-cut pattern of relationship two of the sequences from P. pratensis are basal to this between these sequences is seen. This discrepancy is pre- branch (ii) The studies also agree in that PgiC2 clusters sumably due to these sequences being so closely related most closely with sequences from P. palustris and P. nemor- to each other that they carry signs of being affected by 894 P. Vallenback et al. / Molecular Phylogenetics and Evolution 46 (2008) 890–896

Fig. 2. Phylogenetic relationship among twelve Poa PgiC sequences plus F. ovina PgiC1 and PgiC2. Maximum likelihood tree based on 3698 bp exon and intron data. Other details as in Fig. 1. recombination. A detailed comparison shows that most of the sequence—just as expected for sequences undergoing divergence between PgiC2 and P. nemoralis 1–1 exists in recombination. Limiting our attention to base pair differ- the first 300 bp of the sequence with 29 bp differences, to ences over the whole analyzed region, sequence 1–2 from be compared with only 32 differences over the rest of the P. palustris is unquestionably the sequence most similar 3300 bp. Between PgiC2 and the P. palustris 1–2 sequence to PgiC2 in our analyzed sample of sequences. there are 35 differences evenly distributed along the entire Estimates of divergence, in terms of KS and KA, between sequence. Different branching patterns would therefore be some of the sequences are presented in Table 2. According deduced if the analysis were restricted do different parts to KS, PgiC2 is about thirty times more similar to P. palus- P. Vallenback et al. / Molecular Phylogenetics and Evolution 46 (2008) 890–896 895

Table 2 from P. palustris and P. nemoralis (Fig. 2); according to Divergence estimates between selected PgiC sequences the KS values, PgiC2 is most similar to P. palustris 1–2 Sequences No. of KS KA (Table 2). However, due to the complex relationship bp between the genomes in these polyploid taxa and their F. ovina PgiC2–P. palustris 1257 0.0067 0.0021 occasional sexuality (Patterson et al., 2005), it seems unsafe 1–2 to extend the search for the origin of PgiC2 down to the F. ovina PgiC2–P. nemoralis 1257 0.0450 0.0010 species level—that we have been able to determine the 1–1 F. ovina PgiC2–PgiC1 1257 0.2113 0.0137 exact Poa lineage from which PgiC2 has come is sufficient P. palustris 1–1–1–2 1257 0.0067 0.0010 for our purpose. P. nemoralis 1–1–1–2 1257 0.0853 0.0010 An interesting sign of the close relationship between P. pratensis 1–1–1–2 1257 0.0917 0.0121 these taxa and their genomes, is the indication of recombi- P. pratensis 2–1–2–2 1257 0.1082 0.0084 nation found in the analyzed sequence 1–1 from P. nemor- P. annua–P. supina 912 0.0047 0.0014 0 Five alleles from PgiC1a 1158 0.0297–0.0645 0.0011–0.0045 alis, where the 5 -end of the sequence is clearly less similar F. polesica–F. ovina 912 0.0331–0.0677 0.0014–0.0043 to PgiC2 than the 30-end. This difference is unique to P. PgiC1allelesa nemoralis 1–1, since in general the PgiC2 sequence is evenly Values not corrected for possible multiple substitutions. diverged from the other sequences over the whole analyzed a Alleles PgiC1 a1, a2, b, c and d (see Ghatnekar et al., 2006). region (data not shown). There are two likely routes for the transfer of PgiC2 into Festuca: either it was moved by an external vector e.g. an tris 1–2 than to PgiC1 from F. ovina (all alleles from PgiC1 insect or a fungus, which transmitted the sequence perhaps give similar estimates). The estimates of K are small and A bracketed by a virus/transposing element, or genetic mate- uncertain since they are often based on very few nucleotide rial from Poa became introduced into the Festuca back- substitutions, but their general pattern correlates well with ground via a highly abnormal fertilization event, in which the estimates of K . S at least a chromosomal fragment from a Poa pollen partic- The uncorrected divergence between PgiC2 and P. ipated. If the second scenario turns out to be true—which palustris 1–2 in intron sequence is 35/3480 = 1.0%. This ongoing sequencing outside the active gene will indi- value is consistent with the estimated divergence at synon- cate—our present results tell that the chromosomal mate- ymous sites in exons, K = 0.67%, given that the K value is S S rial would have come from a Poa plant where the control based on only two substitutions among about 300 synony- of the fertilization process is relaxed, as indicated by the mous positions. broad range of chromosome numbers and the incomplete For the later discussion of the timing of the transfer of apomixis found in this group of polyploid species (Asker PgiC2, it is valuable to notice that the evolution in the and Jerling, 1992). ML tree fits a molecular clock (with outgroup: The time of transfer of PgiC2 into F. ovina can be given an v2 = 25.62, df = 25, p > 0.05; without outgroup: v2 = upper estimate by the degree of divergence between this 24.78, df = 24, p > 0.05). sequence and the Poa sequence most similar to it. The transfer must have happened after the coalescence of these two 4. Discussion sequences. Using KS we estimate this divergence at 0.67%, a comparatively small value (see Table 2; KS is the estimator The results presented here confirm that PgiC2 has, by for which we have information from the largest number of some as yet unknown process of horizontal transfer, moved sequence comparisons). This value is close to the lowest esti- from Poa into F. ovina where it has become integrated into mate of the divergence between sequences from polyploids the nuclear genome. In the phylogenetic analyses the PgiC2 within Poa. Only the divergence between two hybridizing gene clusters well inside a highly supported Poa lineage Poa species, P. annua and P. supina, is smaller (0. 47%), while with both diploid and polyploid species more basal (Figs. the divergence between electrophoretically distinct alleles 1 and 2). In addition, PgiC2 is only slightly diverged from from PgiC1 is five to ten times larger. So is also the diver- the most similar sequences from P. palustris and P. nemor- gence between these alleles and the sequence retrieved from alis (Table 2). Together, these results exclude that PgiC2 F. polesica, a species morphologically similar to F. ovina.In originates from a genome of a third unidentified genus, biological terms, the transfer from Poa to Festuca must or that it is the result of gene duplication in the lineage therefore have occurred well after the formation of the leading to F. ovina. Thus, given the complete reproductive major lineages within these genera as well as most extant separation between Festuca and Poa, this event represents species. A numerical estimate of the upper limit for the time an unquestionable example of a horizontal transfer of a of transfer can be made assuming that synonymous sites in functional nuclear gene between two plant species. plant nuclear sequences evolve at a rate of 5.8–8.1 109 Our detailed analysis shows that PgiC2 has its origin per year (Muse, 2000; based on data from Gaut, 1998, and among the polyploid Poa species that contain, among oth- Wolfe et al., 1987). The approximate constancy of the diver- ers, P. palustris, P. nemoralis, P. angustifolia and P. praten- gence rate in the PgiC tree is assured by the fit to a molecular sis. PgiC2 appears to be particularly similar to sequences clock. According to this logic, the transfer occurred less than 896 P. Vallenback et al. / Molecular Phylogenetics and Evolution 46 (2008) 890–896

600,000 years ago (1/2 [0.0067/5.8] 109 = 577,586) in late Asker, S.E., Jerling, L., 1992. Apomixis in Plants. CRC Press Inc, Boca Pleistocene. If the calculation instead is based on the degree Raton. of divergence in introns (1.0%) the estimate becomes about Bergthorsson, U., Adams, K.L., Thomason, B., Palmer, J.D., 2003. Widespread horizontal transfer of mitochondrial genes in flowering 867,000 years. The original evolution of the transgene must, plants. Nature 424, 197–201. thus, have taken place in a world that from a biogeograph- Bergthorsson, U., Richardson, A.O., Young, G.J., Goertzen, L.R., ical point of view did not differ drastically from the world we Palmer, J.D., 2004. Massive horizontal transfer of mitochondrial presently live in. genes from diverse land plant donors to the basal angiosperm In principle, also a lower bound for the transfer event Amborella. Proc. Natl. Acad. Sci. USA 101, 17747–17752. Catalan, P., Torrecilla, P., Rodriguez, J.A.L., Olmstead, R.G., 2004. could be calculated from sequence data. PgiC2 as it exists Phylogeny of the festucoid grasses of subtribe Loliinae and allies today in F. ovina is a complex locus with usually one active (Poeae, Pooideae) inferred from ITS and trnL-F sequences. Mol. gene and at least one pseudogene (Ghatnekar et al., 2006). Phylogenet. Evol. 31, 517–541. The divergence between these two genes, as well as their Davis, C.C., Wurdack, K.J., 2004. Host-to-parasite gene transfer in respective allelic divergences, must all have taken place flowering plants: phylogenetic evidence from malpighiales. Science 305, 676–678. after the transfer into F. ovina. These divergences are, how- Diao, X., Freeling, M., Lisch, D.R., 2006. Horizontal transfer of a plant ever, so small that no meaningful numerical estimates can transposon. PLoS. Biol. 4, 119–128. be based on them (Ghatnekar et al., 2006). This kind of Felsenstein, J., 1988. Phylogenies and quantitative characters. Annu. Rev. analysis will therefore have to wait until a more fine- Ecol. Syst. 19, 445–471. grained survey has been made of the allelic differentiation Gaut, B.S., 1998. Molecular clocks and nucleotide substitution rates in higher plants. Evol. Biol. 30, 93–120. within PgiC2 and of its geographic distribution. Ghatnekar, L., 1999. A polymorphic duplicated locus for cytosolic PGI The data presented in Table 2 may be used for a com- segregating in sheep’s fescue (Festuca ovina L.). Heredity 83, 451– parison of the divergence between different sequences with 459. respect to their nonsynonymous and synonymous sites. Ghatnekar, L., Jaarola, M., Bengtsson, B.O., 2006. The introgression of a The former estimate is clearly smaller than the second functional nuclear gene from Poa to Festuca ovina. Proc. R. Soc. Ser. B Biol. 273, 395–399. one in all comparisons. For example, in the comparison Hegi, G., 1996. Illustrierte Flora von Mittel-Europa. Parey, Berlin. between PgiC2 and P. palustris 1–2 the KA/KS ratio is Jiang, N., Bao, Z., Zhang, X., Eddy, S.R., Wessler, S.R., 2004. Pack- about 0.3 (based on two synonymous and two nonsynony- MULE transposable elements mediate gene evolution in plants. mous substitutions). It thus appears as if stabilizing selec- Nature 431, 569–573. tion for conserved enzyme function represents the Keeling, P.J., Burger, G., Durnford, D.G., Lang, B.F., Lee, R.W., Pearlman, R.E., Roger, A.J., Gray, M.W., 2005. The tree of dominant evolutionary force acting on the PgiC gene and eukaryotes. Trends Ecol. Evol. 20, 670–676. that this also applies to PgiC2. Kimura, M., 1980. A simple method for estimating evolutionary rates of The horizontal transfer of PgiC2 that we now have con- base substitutions through comparative studies of nucleotide firmed, is not only the first known example of a transfer of sequences. J. Mol. Evol. 16, 111–120. a functional nuclear gene in plants but also highly interesting Knobloch, I.W., 1968. A Check List of Crosses in the Graminae. Michigan State University, East Lansing. due to its recent occurrence. Most described horizontal gene Lid, J., 1979. Norsk og svensk flora. Det Norske Samlaget, Oslo. transfers involving plants happened a long time ago (Rich- Maddison, D.R., Maddison, W.P., 2000. MacClade 4: Analysis of ardson and Palmer, 2007), while in this case the event must Phylogeny and Character Evolution, Version 4.0. have taken place between evolutionary units that still are Martin, W., 2005. Molecular evolution—lateral gene transfer and other extant. Thus, there is a good chance that both the molecular possibilities. Heredity 94, 565–566. Muse, S.V., 2000. Examining rates and patterns of nucleotide substitution mechanisms by which the transfer/incorporation occurred, in plants. Plant. Mol. Biol. 42, 25–43. and the selective forces causing the spread of the new con- Patterson, J.T., Larson, S.R., Johnson, P.G., 2005. Genome relationships struct, will be amenable to direct analysis—all based on the in polyploid Poa pratensis and other Poa species inferred from results on origin and timing of the event that we here present. phylogenetic analysis of nuclear and chloroplast DNA sequences. Genome 48, 76–87. Posada, D., Crandall, K.A., 1998. MODELTEST: testing the model of Acknowledgments DNA substitution. Bioinformatics 14, 817–818. Richardson, A.O., Palmer, J.D., 2007. Horizontal gene transfer in plants. We thank B. Jacobsson for help with cultivating the J. Exp. Biol. 58, 1–9. plants and A. Ceplitis for valuable comments on a previous Rozas, J., Sanchez-DelBarrio, J.C., Messeguer, X., Rozas, R., 2003. version of the manuscript. Helpful suggestions from a con- DnaSP, DNA polymorphism analyses by the coalescent and other methods. Bioinformatics 19, 2496–2497. structive reviewer are also gratefully acknowledged. The Swofford, D.L., 1998. PAUP* Phylogenetic Analysis using Parsimony (* research was supported by the Jo¨rgen Lindstro¨m Fund, and Other Methods). the Erik Philip-So¨rensen Fund, the Nilsson-Ehle Fund Turcotte, K., Srinivasan, S., Bureau, T., 2001. Survey of transposable and the Swedish Research Council. elements from rice genomic sequences. Plant J. 25, 169–179. Wolfe, K.H., Li, W.H., Sharp, P.M., 1987. Rates of nucleotide substitution vary greatly among plant mitochondrial, chloroplast, References and nuclear DNAs. Proc. Natl. Acad. Sci. USA 84, 9054–9058. Yu, Z., Wright, S.I., Bureau, T.E., 2000. Mutator-like elements in Andersson, J.O., 2005. Lateral gene transfer in eukaryotes. Cell. Mol. Life Arabidopsis thaliana: structure, diversity and evolution. Genetics 156, Sci. 62, 1182–1197. 2019–2031.