Amer. J. Bot. 76(12): 1720-1730. 1989.

CHLOROPLAST DNA SYSTEMATICS OF : RESOURCES, FEASIBILITY, AND AN EXAMPLE FROM THE ORCHIDACEAEI

MARK W. CHASE2 AND JEFFREY D. PALMER3 Department of , University of Michigan, Ann Arbor, Michigan 48109-1048

ABSTRACT Although chloroplast DNA (cpDNA) analysis has been widely and successfully applied to systematic and evolutionary problems in a wide variety of dicots, its use in monocots has thus far been limited to the . The cpDNAs ofgrasses are significantly altered in arrangement relative to the genomes of most vascular , and thus the available clone banks ofgrasses are not particularly useful in studying variation in the cpDNA ofother monocots. In this report, we present mapping studies demonstrating that cpDNAs offour lilioid monocots (Allium cepa, Alliaceae; sprengeri, ; Narcissus x hybridus, ; and On­ cidium excavatum, ), which, while varying in size over as much as 18 kilobase pairs, conform to the genome arrangement typical of most vascular plants. A nearly complete (99.2%) clone bank was constructed from restriction fragments of the chloroplast genome of excavatum; this bank should be useful in cpDNA analysis among the monocots and is available upon request. As an example of the utility of filter hybridization using this clone bank to detect systematically useful variation, we present a Wagner parsimony analysis of restriction site data from the controversial Trichocentrum and several sections of Oncid­ ium, popularly known as the "mule ear" and "rat tail ." Because of their vastly different floral morphology, the of Trichocentrum have never been placed in Oncidium, although several authors have recently suggested a close relationship to this vegetatively modified group. The analysis of cpDNA presented here supports this affinity; in fact, it places Tricho­ centrum as a derivative of the mule ear oncidiums.

AMONG THE meers, systematic studies ofchlo­ least three large inversions (Palmer, 1985; roplast DNA (cpDNA) variation are relatively Howe et aI., 1988). Clone banks constructed common and have been applied to a wide va­ from these highly rearranged genomes are of riety ofspecies at several taxonomic levels (see limited use in mapping studies of other vas­ review in Palmer et aI., 1988a). As is true of cular plants, and th us far the sources ofcpDNA systematic studies ofmonocots in general, the probes in filter hybridizations have been lim­ principal focus of molecular studies in the ited to those ofdicots, such as spinach, petunia, monocots has been the Poaceae (Tsunewaki and tobacco, that have the consensus arrange­ and Ogihara, 1983; Enomoto, Ogihara, and ment for flowering plants. The chloroplast ge­ Tsunewaki, 1985; Doebley, Renfroe, and Blan­ nomeofSpirodela oligorhiza (Araceae) has been ton, 1987; Lehvaslaiho, Saura, and Lokki, 1987; mapped (de Heij et aI., 1983) and was found Hilu, 1988). The chloroplast genome in the to be colinearwith theconsensus vascular grasses differs significantly from those found arrangement, although it was 30 kilo base pairs thus far in the dicots in the possession of at (kb) larger than that of Petunia. In this study we focused on four lilioid monocots: Allium cepa (Alliaceae), Asparagus I Received for publication 20 January 1989; revision sprengeri(Asparagaceae), Narcissus x hybridus accepted 25 May 1989. (Amaryllidaceae), and Oncidium excavatum We wish to acknowledge the support of a National Sci­ ence Foundation Fellowship in Environmental Biology, (Orchidaceae). Physical and gene mapping BSR-8600179, to MWC. We are also grateful for permis­ studies reveal that, unlike the grasses, the ge­ sion to remove leaves and from the orchid collec­ nome arrangement ofthese monocots is typical tions ofAlexis and Ted Linder (Great Lakes Orchids) and of that of most land plants. Ron Cnesinski (Taylor Orchids). Requests for the orchid cpDNA clone bank described here should be directed to Furthermore, we have constructed a clone MWC. We are grateful to William F. Thompson for pro­ bank of restriction fragments of one of these viding laboratory facilities for the portion of the work species, Oncidium excavatum and have used performed in 1982-1983. this bank in filter hybridizations to discover 2 Current address: Department of Biology, University of North Carolina, Chapel Hill, NC 27599-3280. systematically useful variation in the cpDNA 3 Current address: Department ofBiology, Indiana Uni­ of other orchids. As an example of the utility versity, Bloomington, IN 47405. of this approach in gathering useful new in-

1720 December 1989] CHASE AND PALMER-DNA SYSTEMATICS OF MONOCOTS 1721

TABLE I. Sources of DNA from species in the (vouchered by jlowers in FAA in the first author's private collection)

Species Origin- Voucher number I. Oncidium excavatum Cajamarca, Peru (GLO) 83427 2. Rossioglossum schlieperianum Chiriqui, Panama (GLO) 83449 3. Oncidium ampliatum Chiriqui, Panama (MWC) 84104 4. Psychopsis sanderae Peru (TO) 86126 5. Oncidium jonesianum Paraguay (TO) 86118 6. Oncidium ascendens Chiapas, Mexico (MWC) 82109 7. Oncidium flavovirens Colima, Mexico (MWC) 83109 8. Oncidium splendidum Guatemala (UMBG) 84552 9. Oncidium bicallosum Chiapas, Mexico (MWC) 83387 10. Trichocentrum panduratum Peru (TO) 86179 II. Trichocentrum tigrinurn Peru (RO) 83439 12. Oncidium pulvinatum Brazil (GLO) 82228 13. Oncidium hians Brazil (SBOE) 86134

a GLO = Great Lakes Orchids; MWC = Mark W. Chase, field collected; TO = Taylor Orchids; UMBG = Matthaei Botanical Garden, University of Michigan; RO = R. J. Rands Orchids; SBOE = Santa Barbara Orchid Estate. formation that differentiates opposing hypoth­ Sources of material ofeach orchid species are eses about phylogenetic relationships, we pre­ listed in Table 1. All DNAs were further pu­ sent an analysis ofcpDNA variation among a rified by means of cesium chloride-ethidium controversial group oforchids. Some members bromide gradients. ofthe oncidioid orchids (subtribe Oncidiinae) Restriction endonuclease digests, electro­ display a series of vegetative modifications phoresis, transfer to nylon filters, labelling of (succulence) that have earned for them the recombinant plasmids, filter hybridizations, common names of "mule ear" and "rat tail and autoradiography were carried out accord­ oncidiums." Members ofthe genus Trichocen­ ing to Palmer (1986). Washing ofthe filters was trum also exhibit the "mule ear" condition but done under the following conditions: two have a floral morphology that is radically dif­ washes of 15 minutes at room temperature, ferent from any species on Oncidium. Dressler followed by two washes for 30-60 min at 65 and Williams (1982) suggested that Trichocen­ C, with a 2 x SSC, 0.5% SDS wash buffer. After trum is closely related to these vegetatively a round of hybridization, the probes were modified oncidiums. Others (Wirth, 1964; stripped from the filters by 3-4 washes ofboi1­ Chase, 1986) viewed the exceptionally dis­ ing 0.1 x SSe. These filters (Zetabind, AMF tinctive floral morphology of Trichocentrum Cuno) went through over 30 rounds of hy­ as support for a view that the similar vegetative bridizations and were still in usable condition. features are due to parallelism. A series ofother A restriction site map for 6 enzymes was members of Oncidium have less extreme veg­ constructed for O. excavatum (Fig. 1) using the etative features that also suggest a relationship mapping strategy ofPalmer(1986). The cpDNA to the mule eargroup. We examined restriction probes for this mapping were the clone bank site variation in cpDNA, as detected by filter of Lactuca sativa (Jansen and Palmer, 1987). hybridization, among these species in to For the region in which L. sativa has a 22 kb test the conflicting ideas of their phylogenetic inversion, we substituted the three uninverted relationships. Petunia x hybrida clones (Palmer et aI., 1983). Once this restriction fragment map was com­ MATERIALS AND METHODS-CpDNAs ofAl­ pleted, gene probes from tobacco, spinach, and lium cepa, Asparagus sprengeri, Narcissus pea were hybridized to the same filters to iden­ x hybridus and Oncidium excavatum were iso­ tify their locations in this genome (for probe lated by the sucrose gradient technique (Palm­ descriptions, see Jansen and Palmer, 1987). er, 1986). For all other taxa, total cellular DNAs The representatives of three other lilioid were isolated by the modified CTAB method monocots, Allium cepa, Asparagus sprengeri, ofDoyle and Doyle (1987). Most ofthe mem­ and Narcissus x hybridus, were mapped for two bers of the mule ear oncidium group are suc­ restriction enzymes, Pvull and Sad, using the culent, and the standard method of isolation cloned genome of Vigna radiata (mung bean; of total cellular DNA did not work well or at Palmer and Thompson, 1981) as the source of all. This problem was remedied by using a 3 x cpDNA probes. (At the time this mapping was CTAB buffer rather than the standard 2 x . performed, more appropriate clone banks [i.e., --.J 4 4 N ..-- N ...", ...... -< : UU- .. N-NN '" r;; ...... '" ... ~ ~~:. o 0 e A. A. -- 0."1...... '0 '0 . - ~ ... ; - ~ .. '" '" ~ eo eo ...... eo '"-...'" -...'" '" - "foe- ...... e- eo S PP S PX X Xb P X X X - NN X s X XH 5 ~3.83.7 I 5.8 I 5.0 3.0 ~M.316.7 12.8 3.9 I See left side of Inverted repeat IV 4.9 1 5.8 I 4.6 I 4.9 12.21 5.6 14.0 14.5 3.0 ~811.8 Xho I 6.2 0~13.01 14.5 I 9.2 1 10.4 I 4.9 I I 13.3 I 17.8 Pst I 12.8 15.0 I 47 25.0 1-r°.7 18.4 Kpn 1 0•71" 31 I 4.8 1 6.7 I 5.1 1 I.ll-° 14.4 1 4.5 11.~16.2 Sma I ~.914.5 1 14.4 I .l!J0 25 1 6.1 25 1·.9J.I,1 4.2 1 13.4 Sac I 1 4.2 1J1I:'0 23 1 5.9 1 7.0 12.51 92 1f.6 Sal I I 0~6 36 I 15.3 1 ~ tTl I I I I I I I I I ~ 90 100 110 120 130 140 143/0 10 20 Q z ...... 8 +-- _4 :;.:l Q U .. -e ..J -- -- ... '" ~'!e~~~;;~ z -- . '",e..!-.. ~ ;l> - ii~ -:-: D..~ . ... -=...... -= ... - ...... ;,e-.!!f'e-e-e- = .. r- H PPP PX P K X C XXPX o I .. 1\1 I I" " II I II .tL- I >r1 14.0 27 i 19.4 45 12.2 1.,j;;6 23 (92) I I II I I 30 40 50 60 70 80

Fig. 1. Restriction fragment and gene map of cpDNA of Oncidium excavatum. The clone bank fragments are indicated in the uppermost, shaded strip, and the endonuclease ...... , restriction sites are C, ClaI; H, HindIII; N, NsiI; P, Pstl; S, Sacl; Xb, XbaI; X, Xhol. Crosshatched fragments are uncloned; fragment numbers (Table 2) begin at the left edge <: of the inverted repeat. Darkened bars indicate the position of the inverted repeat. ~ -.J 0\ December 1989] CHASE AND PALMER-DNA SYSTEMATICS OF MONOCOTS 1723

Allium cepa 4.1 4.1 0.9 18.7 21 18.5 14.6 18.5 22 PvuII I 14.8 r I H I I JL I LL I L I OJ ILL I I 144.9 kb 16.2 18.8 9.7 12.8 17.2 7.5 13.3 9.8 7.8 5.6 7.0 11.1 7.5 13.3 10.6 0.9 0.6 0.6 0.9 1.9 2.5 4.3 3.4 1.1 6.2 kb

Asparagus sprengeri

4.1 4.1 0.9 PvuII 119.6JL 19.4 32 25 15.4 L1L I I J1L II 150.2 kb 16.2 18.8 12.8 17.2 7.0 11.1 13.3 10.6 0.9 0.6 4.2 3.3 1.1 Sad 148.6 kb

Narcissus Xhybridus

4.1 4.1 0.9 2.7 4.04.0 18.0 3 17.8 17.8 48 PvuII I JL 1 JiJ I 1 Hf-:,---,:,--'--:-II 1.8 II LlhLJ II I 157.0 kb 16.2 9.7 12.8 17.2 7.5 13.3 9.8 7.8 5.6 7.0 11.1 7.5 13.3 10.6 0.6 1.0 1.5 3.4 1.1 Sad 157.9 kb

Fig. 2. Restriction fragment maps of Allium, Asparagus, and Narcissus for the endonucleases Pvull and Sad, with the relative position of the mung bean clones used in the filter hybridizations. Lines from the mung bean clones to fragments in the other species indicate fragments hybridizing to the mung bean sequences. The mung bean map is drawn with the position of a 50 kb, legume-specific inversion reversed, in order to approximate the arrangement typical of most vascular plants (i.e., the 7.5 kb and 13.3 kb fragments, which contain the ends of the inversion, are shown twice). Position of the inverted repeat is indicated by the dark bars above each map (end points have not been accurately determined and are only approximations). ones that did not contain inversions] were un­ the restriction enzymes Pstl and XhoI, which available.) In Fig. 2, we have corrected for the produced 23 fragments from 0.2-14.0 kb. Ini­ 50 kb inversion found in Vigna radiata, while tial cloning was into pIC20H (Marsh, Ertle, a second, larger inversion that affects the whole and Wilkes, 1984) by the shotgun method. Lat­ large single copy region (Palmer, Osorio, and er cloning was into Bluescript sk+ (Stratagene, Thompson, 1988b) does not require compen­ Inc.) by gel-isolating fragments from digests sation. separated in low melt agarose. Fragments long­ The chloroplast genome of Oncidium ex­ er than 8.5 kb were subcloned by digesting the cavatum was cloned using a combination of larger fragments with an appropriate set ofen- 1724 AMERICAN JOURNAL OF BOTANY [Vol. 76

TABLE 2. Clones and subclones ofOncidium excavatum

Fragment number Size Enzymes Vector I 3.8 kb PstI-XhoI pIC20H 2 9.5 kb XhoI-PstI pIC20H a. 3.7 kb XhoI-XbaI BS sk+ b. 5.8 kb XbaI-PstI BS sk+ 3 5.0 kb PstI-XhoI pIC20H 4 3.0 kb XhoI-PstI pIC20H 5 (uncloned) 0.8 kb XhoI-XhoI 6 9.1 kb XhoI-PstI pIC20H a. 2.3 kb XhoI-SacI BS sk+ b. 6.7 kb SacI-PstI pIC20H 7 2.8 kb PstI-PstI BS sk+ 8t 6.3 kb PstI-XhoI pIC20H a. 3.9 kb PstI-SacI pIC20H 9t 9.4 kb XhoI-XhoI BS sk+ a.* 0.35 kb NsiI-NsiI BS sk+ b.* 4.9 kb NsiI-PstI BS sk+ 10 10.2 kb XhoI-XhoI BS sk+ a. 5.7 kb XhoI-SacI BS sk+ b. 4.5 kb SacI-XhoI BS sk+ II 4.9 kb XhoI-XhoI pIC20H 12 12.3 kb XhoI-PstI BS sk+ a. 2.2 kb XhoI-HindIII BS sk+ b. 5.6 kb HindIII-HindIII BS sk+ c. 4.5 kb HindIII-PstI BS sk+ 13 1.0 kb PstI-PstI pIC20H 14 0.8 kb PstI-PstI pIC20H 15 4.8 kb PstI-PstI pIC20H 16 4.0 kb PstI-XhoI pIC20H 17 7.4 kb XhoI-PstI pIC20H 18 8.9 kb PstI-XhoI pIC20H a. 6.0 kb PstI-KpnI BS sk+ b. 2.9 kb KpnI-XhoI pIC20H 19 14.0 kb XhoI-XhoI BS sk+ a. 8.3 kb XhoI-ClaI BS sk+ b. 5.7 kb ClaI-XhoI BS sk+ 20 2.7 kb XhoI-XhoI BS sk+ 21 1.5 kb XhoI-PstI pIC20H 22 2.0 kb PstI-XhoI pIC20H 23 (uncloned) 0.2 kb XhoI-PstI t Only the single copy portions of these cloned fragments have been subcloned. * Hybrid cloning sites. zymes, separating them in low melt agarose, gus, and Narcissus are shown in Fig. 2. Large cutting the fragments out, and ligating them inversions, such as have been found in the with the appropriately cut vector. In this way, grasses, are absent from these monocot ge­ a clone bankwith a total of30 cloned fragments nomes. At this level of analysis, they are co­ was produced (Table 2). The O. excavatum linear with that of the typical, unrearranged clone bank was then used to map restriction dicot genome, such as tobacco or spinach, and sites for ten enzymes that cut a typical chlo­ with that ofOncidium (Fig. I). The major dif­ roplast genome 35-80 times each. Phyloge­ ference among these genomes is size: Oncidium netic analysis was performed using the Wagner is 143 kb long, Allium, 145 kb; Asparagus, 149 parsimony programs, PAUP (D. Swofford; in­ kb; and Narcissus. 157 kb. The majorstructural cluding the branch and bound and strict con­ feature of these genomes is a large inverted sensus options) and PHYLIP (Felsenstein, repeat, which we estimate in Oncidium to have 1985; including the bootstrap option). Se­ a maximum and minimum sizes of 25.6 and quence divergence was calculated for the most 24.4 kb, respectively. The smaller size of this closely and distantly related species pairs using genome appears to be the product of major equations 9 and 10 of Nei and Li (1979). reduction in a few regions rather than uniform reduction throughout the genome. For exam­ REsuLTs-Genome size and structure-The ple, an 18.8 kb SacI fragment from lettuce (con­ linearized cpDNA maps for Allium. Aspara- taining the entire small single copy region) hy- December 1989] CHASE AND PALMER-DNA SYSTEMATICS OF MONOCOTS 1725

...------Oncidiumexcavatum ...------Oncidium excavatum rr:------...:~:....RoSsiog/ossumschlieperianum .------....::....Rossiog/ossum schlieperianum ...------....:....Oncidium ampliatum ...------"-Oncidium ampliatum .------...:8:....Psychopsis sanderae ...----....:..Psychopsis sanderae 4 9:""Oncidium 4 r---...: jonesianum .----~ Oncidiumjonesianum 98% '----...:5:....0ncidium ascendens 6 '---....::....Oncidium ascendens ...---,,4c..Oncidiumjlavovirens Oncidiumflavovirens 97% 5c..O r--..: ncidium splendidum Oncidium splendidum 2 5 Oncidium bicallosum 5 Oncidium bicallosum 55% 6 Trichocentrumpanduratum 7 6 Trichocentrum panduratum 8 Trichocentrum tigrinum 8 Trichocentrum tigrinum .-- 1'- Oncidium pulvinatum 15 .-- 1:.... Oncidiumpu/vinatum 3 '-----...:3:....0ncidium hians 4 '--__...:3:.... Oncidium hians r------Oncidium excavatum 2n=56 ...------Rossiog/ossum sch/ieperianum 2n=44 .------Oncidium ampliatum 2n=44 r-----Psychopsis sanderae 2n=38* Oncidiumjonesianum 2n=30] rat-tail '----Oncidium ascendens 2n=36* oncidiums r---Oncidiumjlavovirens 2n=36J '--1----0ncidium sp/endidum 2n=36 ::C\~:;s Oncidium bicallosum 2n= 28 Trichocentrum panduratum * Trichocentrum tigrinum 2n=24 Oncidium pulvinatum 2n=42J Brazilian 5 '-----Oncidium hians 2n=42* :;:.~~-;:,a;s Fig. 3-5. Wagner parsimony of the vegetatively modified members of Oncidium, Psychopsis, and Trichocen­ trum, rooted with Oncidium excavatum and Rossioglossum. The numbers at each node and are the number of site mutations that define each lineage. 3. One of two equally parsimonious trees found by both PAUP and PHYLIP. These two trees each have 79 steps and consistency indices of 0.92. The bootstrap option of PHYLIP was used to establish a measure of statistical significance for each monophyletic lineage, and this is expressed as a percentage. 4. The alternate, equally most parsimonious tree found by PAUP and PHYLIP. The only difference between this tree and that in Fig. 3 is the relationship of O. splendidum and O. f/avovirens to each other and the to which they are sister taxa. 5. Strict consensus tree (from PAUP), to which chromosome numbers (Tanaka and Kamemoto, 1984) have been added. Taxa marked with an asterisk have not been counted; in cases in which a closely related species has been counted, those chromosome numbers have been presented. bridizes to only a 13.4 Sad fragment in the monocots, particularly the orchids, led us to orchid. Fine structure mapping experiments clone an orchid chloroplast genome. All ofthe suggest that these Sad sites are homologous, 21 original orchid clones that exceeded 8.5 kb which means that in the small single copy re­ in length were subcloned so that the clone bank gion alone the orchid is 5.4 kb shorter than now consists of 30 fragments. These clones lettuce. Many ofthe Sad and PvuII sites in all represent 98.7% of the O. excavatum genome four genomes may in fact be conserved; ifthis (99.2% of the sequence diversity). Fragments is indeed true then it appears that some regions that extend over the boundaries ofthe inverted of these genomes are expanding while others repeat, clones 6, 8, and 9, were subcloned using are shrinking and not always in the same di­ sites near the edge of the inverted repeat in rection as the genome is changing as a whole. order to generate more specific probes that will For example, if the PvuII sites that produce produce simpler and more readily interpret­ the 4.1 kb and 0.9 kb fragments flanking a large able hybridization patterns. fragment comprising most of the inverted re­ peat are homologous (Fig. 2), then Narcissus, Sequence divergence-Sequence divergence with the largest genome, is the most condensed values (Nei and Li, 1979) were calculated for for this fragment. Further studies are needed the most distantly and most closely related to determine whether such size differences in species pairs among these 13 orchid species the small single copy region are due primarily (Fig. 3). The most closely related pair, O. hians to shorter intergenic spacers, fewer and shorter and O. pulvinatum, and the most distantly re­ introns, or fewer genes. lated, O. excavatum and T. tigrinum, diverged by 0.14% and 2.77%, respectively, which are Clone bank composition-Our desire to un­ typical ofvalues reported for intrageneric stud­ dertake further experiments in the lilioid ies (Palmer and Zamir, 1982; Palmer et al., 1726 AMERICAN JOURNAL OF BOTANY [Vol. 76

TABLE 3. Chloroplast DNA restriction site mutations for TABLE 3. Continued the nine enzymes used in phylogenetic analyses of13 species in the Oncidiinae (Table 1). Only the mutations Enzyme- Region Mutation Taxa that group two or more species (synapomorphies) are 12.2 ~ 7.6 + 4.6 12,13 listed. Map coordinates are cited to the nearest 0.5 kb 57. H 21.5 58. H 38.0 2.1 + 9.0 ~ II.I 4-13 (see Fig. 2) for each mutation. which are listed with II.I ~ 10.5 + 0.5 the ancestral state (see Methods) first and the derived 59. H 38.5 5, 6 11.1 ~ 9.8 + 1.3 state second. Taxa sharing a given mutation are num­ 60. H 40.0 12, 13 11.1 ~ 5.6 + 5.5 9-11 bered as in Table 1 61. H 44.5 62. 1 5.5 1.0 + 0.4 ~ 2.4* 5-11 63. I 16.0 2.7 + 1.4 ~ 4.1 7-11 Enzymes- Region Mutation Taxa 64. I 10.5 4.1 + 1.0 ~ 5.1* 5, 6 bl. A 96.0 6.6 ~ 5.2 + 1.3 4, 7, 8 65. I 19.0 2.2 + 2.2 ~ 4.4 4-13 -z. A 98.5 0.9 + 1.2 ~ 2.1 5, 6 66. I 41.0 7.4 ~ 5.9 + 1.5* 10, II 3. A 10.5 0.3 + 0.7 ~ 1.0* 4,12,13 a Enzyme abbreviations: A, BanI!; B, BamHI; C, XbaI; 4. A 50.5 2.2 ~ 1.6 + 0.6* 2-13 D, HindIII; E, ClaI; F, BanI; G, BglII; H, EcoRI; I, EcoRV. 5. A 53.0 1.6 ~ 1.1 + 0.5* 5,6 NsiI was mapped but no mutated sites were discovered. 6. A 70.5 2.4 + 3.4 ~ 5.8 2-13 b Homoplasious site mutation relative to most parsi­ 7. A 80.0 9.6 ~ 6.3 + 2.3 2-13 monious . 8. A 78.5 9.4 ~ 8.3 + 1.1 2-13 * Exact position uncertain because two or more frag­ 9. B 97.0 18.0 + 2.0 ~ 20.0 8-11 ments are unordered. 10. B 33.5 10.9 ~ 6.9 + 4.0 4-13 II. B 75.0 4.4 + 2.6 ~ 7.0 5-11 12. C 95.0 15.2 ~ 11.4 + 3.8 12,13 13. C 78.0 4.3 + 1.0 ~ 5.3* 10, II 1983; Sytsma and Schaal, 1985; Sytsma and 14. C 2.0 4.7 + 0.8 ~ 5.5 5-13 Gottlieb, 1986; Doebley et al., 1987). 15. C 11.5 3.8 + 9.2 ~ 13.0 5-11 16. C 26.0 4.0 ~ 2.3 + 1.7 5, 6 Phylogenetic analyses-Three major regions 17. C 23.5 4.0 + 1.1 ~ 5.1 12,13 of size variation were uncovered among the 18. C 22.5 1.1 + 5.1 ~ 6.2* 12,13 19. C 23.5 5.1 ~ 4.0 + 1.1 7-11 orchid genomes studied here. Point mutations 20. C 44.5 16.0 + 3.1 ~ 19.1 12,13 in these regions were not utilized in this anal­ b21. C 47.5 3.1 + 9.5 ~ 12.6 8-10 ysis unless the fragments in which point mu­ 22. C 52.5 9.5 ~ 4.3 + 5.2 12,13 tations were hypothesized appeared to be free 23. C 68.5 4.2 + 2.6 ~ 6.8* 10, II ofsize variation. These three regions were the 24. C 65.0 4.3 ~ 3.6 + 0.7* 9-11 13.4 kb small single copy region, which varied 25. C 71.5 4.2 ~ 2.6 + 1.6 2-13 26. C 76.0 2.7 + 1.0 ~ 3.7* 10, II by as much as 1 kb between species, and the b27. D 94.0 11.8 + 6.1 ~ 17.9 4, 7, 8 regions between coordinates 24-32 and 48-52 28. D 0.5 11.8 ~ 9.8 + 2.0 10, II (Fig. 1), which were less variable in size. Length b29. D 49.0 5.6 + 10.9 ~ 16.5 8-10, 12, 13 mutations responsible for size variation were 30. E 93.0 1.6 ~ 1.2 + 0.4 3-13 not used in phylogenetic analyses because of 31. E 92.5 1.2 ~ 0.8 + 0.6 7,8 the nearly continuous size variation observed, b32. E 96.5 1.5 ~ 1.1 + 0.4* 4,7,8 33. E 10.5 12.3 ~ 8.5 + 3.8 5-11 which made assessments of identity and po­ 34. E 16.0 2.4 ~ 1.4 + 1.0* 5-13 larity impractical. Several otherregions ofsmall 35. E 40.5 7.0 + 3.5 ~ 10.5 12,13 size variation were also detected, but these were 36. E 47.0 8.7 ~ 6.6 + 2.1 4-11 not of significant size or distribution to affect 37. E 45.5 5.2 ~ 3.6 + 1.6 12,13 this analysis. 38. E 78.5 5.2 + 2.0 ~ 7.2 4-13 Approximately 370 restriction sites were 39. F 92.0 1.6 + 1.1 ~ 2.7 2-13 40. F 100.0 3.5 ~ 3.1 + 0.4* 8-11 mapped for each species, and 125 site muta­ 41. F 5.5 2.6 + 4.2 ~ 6.8 3-11 tions were found, of which 66 were shared by 42. F 64.0 14.0 ~ 3.5 + 10.5 2-13 two or more species (Table 3). Polarity ofthese 43. G 83.0 5.1 ~ 1.5 + 3.6 12,13 mutations was determined by outgroup com­ 44. G 91.5 2.4 ~ 1.9 + 0.5* 2-13 parison to O. excavatum and Rossioglossum 45. G 36.5 2.9 ~ 1.1 + 1.8* 10, II schlieperianum (see the second paragraph of 46. G 52.0 1.6 ~ 1.1 + 0.5* 5,6 47. G 58.0 7.4 ~ 3.1 + 4.3 10, II the discussion for the reasons these two taxa 48. G 59.5 7.4 ~ 6.1 + 1.3 12,13 were chosen). These two outgroups agreed with 49. G 36.5 1.1 + 1.8 ~ 2.9 10, II respect to polarity for all but the 7 or 8 mu­ 50. G 46.0 1.1 + 1.2 ~ 2.3 7-11 tations that occurred between these two species 51. H 4.0 2.0 ~ 1.5 + 0.5* 5-11 and that could not be polarized. These data 52. H 11.0 12.2 + 2.0 ~ 14.2 7-11 were then used to construct Wagner parsimony 53. H 14.0 12.2 ~ 10.2 + 2.0 3-11 54. H 12.5 10.2 ~ 7.8 + 2.4 12,13 trees with both PAUP and PHYLIP. b55. H 19.0 12.2 ~ 9.8 + 2.4 3-6,12,13 Two, equally parsimonious trees (Fig. 3, 4), 56. H 23.0 12.2 + 3.1 ~ 15.3 7-11 which differ only in the relationship of two species, O.jlavovirens and O. splendidum, were December 1989] CHASE AND PALMER-DNA SYSTEMATICS OF MONOCOTS 1727 found. In Fig. 3, O. flavovirens is the sister While restriction site analysis is unlikely to species to a group of species that includes O. provide the phylogenetically useful informa­ splendidum. while in Fig. 4 these two species tion about major groups of lilioid monocots, together form a to the same set of the approach promises to be useful within more species that included O. splendidum in Fig. 3. closely related taxa, and it is in this frame of These two trees have 79 steps and consistency reference that the study of the vegetatively indices of 0.92 (8% homoplasy) for all muta­ modified members ofthe orchid subtribe On­ tions and 0.84 (16%) for informative ones (syn­ cidiinae is presented. The make-up and cir­ apomorphies only). Four additional trees were cumscription of the group of species (Table 1) found at 80 steps and six more at 81 steps, and treated here as an example of the application these ten plus the two most parsimonious trees ofthis clone bank were determined by an over­ were used to construct the strict consensus tree all analysis ofvegetative features (Chase, 1986) (PAUP; Fig. 5). We did not use a Dollo par­ and micromorphological details of the com­ simony analysis because the level of homo­ plex pollinaria (Chase, 1987) in subtribe On­ plasy detected was so low. Relative to the cidiinae. Most species in subtribe Oncidiinae cladograms in Fig. 3 and 4, the homoplasious have been shown to be 2n = 56 (Tanaka and mutations (from Table 3) are of the following Kamemoto, 1984), while the members of this types: two parallel site gains (1,27); one parallel vegetatively modified group have a range of site loss (32); two gain/losses (3, 55); and two lower numbers, 2n = 24-44 (Fig. 5), which loss/gains (21, 29). have been demonstrated to be derived by aneu­ The bootstrap option of PHYLIP was used ploid changes (probably reduction; Chase and to place confidence intervals (in percentage of Olmstead, 1988). Additionally, a survey ofseed samples in which the same placement oc­ morphology ofthe subtribe (Chase and Pippen, curred) on the monophyletic groups identified 1988) found a unique sculpturing of the cell in the majority rule consensus tree (Fig. 3). One walls to be restricted to these same species. A hundred samples were run, and only two nodes preliminary analysis of cpDNA variation in were found to have less than 92% confidence the subtribe (Chase and Palmer, 1988) was also intervals. Not surprisingly, the weakest node used to support the naturalness of this clade was the one (54%) at which the alternate equal­ relative to the rest of the subtribe. While O. ly parsimonious arrangement occurs. In the excavatum is representative ofthe largest clade strict consensus tree (Fig. 5), these two weakest of Oncidium (2n = 56 and the conserved nodes collapse into trichotomies. These anal­ section; Dressler and Williams, 1982; Brum­ yses demonstrate strong support for the mon­ mit, 1985), Rossioglossum (2n = 44) appears ophyly of the vegetatively modified members to be a more closely related taxon that would ofOncidium relative to the two outgroup taxa perhaps permit a more accurate assessment of and place genus Trichocentrum as a derivative polarity of the mutations found in this group. ofthe mule ear oncidiums, as exemplified here Indeed, a number of authors (Wirth, 1964; by O. bicallosum. Dressler and Williams, 1982; Dressler, 1981) have suggested it as a close outlier to the veg­ DISCUSSION - While the four lilioid mono­ etatively modified group. When considered cots lack the cpDNA inversions typical of from the perspective ofGaray and Stacy's sec­ grasses and were colinear with other flowering tional revision ofOncidium (1974), this group plants lacking such inversions, they vary con­ ofspecies would appear to be almost randomly siderably in size. Allium. Asparagus. and Nar­ chosen (Table 4), but the abundant evidence cissus all belong to the lato, in cited above suggested that these nine species the order , while the Orchidaceae is (exclusive of Trichocentrum) were a mono­ placed traditionally near the Liliales in its own phyletic unit. The exact relationship ofTricho­ order, . In size, the cpDNAs ofAl­ centrum to this vegetatively modified clade was lium and Oncidium are more similar than that the point of contention and the focus of this of Allium is to the other members of the tra­ example. ditional Liliaceae, but in specific regions the The cpDNAdata clear! y support neither pre­ orchid is quite different. We assert that it would viously suggested hypothesis: Trichocentrum be imprudent to try to infer anything about the is not an associated, monophyletic, outlier rel­ systematic relationships of these four genera ative to the others or a sister group to the mule by comparisons of their restriction site maps. earclade, but rather a closely related derivative The obvious and widespread size variation, as of the larger vegetatively modified lineage. In well as the small number of restriction sites the preliminary molecular analysis of Chase mapped, would undoubtedly be a major source and Palmer (1988), no other members of the of error in any such study. Oncidiinae fell within this clade, although sev- 1728 AMERICAN JOURNAL OF BOTANY [Vol. 76

TABLE 4. Subgeneric and sectional assignments in Oncidium (Garay and Stacy, 1974) ofthe species from Table 1. (Taxa marked with an asterisk have not been treated as members ofgenus Oncidium)

Species and author Subgenus Section Oncidium excavatum Lindley Oncidium Excavata *Rossioglossum schlieperianum Reichenb. f. Oncidium ampliatum Lindley Oncidium Oblongata Psychopsis sanderae (Rolfe) Liickel and Braem Oncidium Glanduligera Oncidium jonesianum Reichenb. f. Oncidium Cebolletae Oncidium ascendens Lindley Oncidium Cebolletae Oncidiumflavovirens L. O. Williams Cyrtochilum Cyrtochilum Oncidium splendidum A. Rich. Oncidium Pluri tuberculata Oncidium bicallosum Lindley Oncidium Plurituberculata *Trichocentrum panduratum Schweinfurth *Trichocentrum tigrinum Linden and Reichenb. f. Oncidium pulvinatum Lindley Oncidium Pulvinata Oncidium hians Lindley Oncidium Paucituberculata

eral othergenera (Cuitlauzina, Lemboglossum, vegetative and floral modification, not an in­ Osmoglossum, Palumbina, and Ticoglossum) termediate as one would expect in a hybrid. appear to be as closely related to this group as Additionally, the chromosome numbers found Rossioglossum. An analysis of the whole sub­ in this clade (discussed in more detail below) tribe is underway and will address the rela­ exhibit a definite trend, and Trichocentrum oc­ tionships more fully. The most that can be cupies its lowest point. concluded on the basis of this analysis is that Aneuploid change in chromosome number including the vegetatively modified taxa in On­ is a major feature ofseveral groups within the cidium does not result in monophyletic taxa Oncidiinae (Chase, 1986; Chase andOlmstead, unless Psychopsis, Rossioglossum, and Tricho­ 1988). Chase (1986) hypothesized that the centrum are not also included. An already pro­ morphological evidence was compatible with posed alternative would be the removal ofthis aneuploid reduction: progressively lower num­ clade from Oncidium as genus Lophiaris bers were associated with increasingly modi­ (Dressler, 1981, prudently did not make any fied vegetative features. Diploid chromosome formal transfers). Such taxonomic decisions numbers have been added to Fig. 5, and the should not be contemplated in isolation, but result is clearly compatible with an hypothesis rather be a part of a comprehensive generic that aneuploid decreases are responsible for the revision of the Oncidiinae (approximately variation in chromosome number in this clade. 1,500 species in 75 poorly delineated genera). Most orchid subtribes are relatively uniform The focus ofthis smallerstudywas the position in nonfloral aspects. The Oncidiinae, in con­ of Trichocentrum, and an hypothesis of close, trast, is one of the most diverse in the derivative relationship has been supported. from cytological, ecological, and morpholog­ The Orchidaceae, and the Oncidiinae in par­ ical (both floral and vegetative) standpoints ticular, are known from the horticultural lit­ (Dressler, 1981). For example, the range of erature as a group in which few barriers to chromosome number in this subtribe, 2n = hybridization exist, and this introduces a ques­ 10-60, exceeds that of the rest of the family. tion as to whether Trichocentrum might not In their , gross floral morphology has have originated as a hybrid between a mule been emphasized, and often, in cases in which ear type and another member of subtribe On­ the floral and vegetative features lead to vastly cidiinae. Ifthe female parent had been a mem­ different generic limits, the ones based on the ber ofthe former type, then the hybrid would latter have been disregarded. The resulting ge­ exhibit a chloroplast genome typical of this neric treatments ofall but the most recent au­ group. We offer several points to counter such thors have been totally artificial. Garay and an explanation. Documented hybrids from na­ Stacy's (1974) arrangement of the sections of ture are rare in the Oncidiinae (MWC, personal Oncidium was by their own admissions arti­ observation); this is due, we believe, to the fact ficial, i.e., closely related species were often that relationships with specific pollinators rare­ placed in different sections. The eleven species ly break down (natural barriers to hybridiza­ ofthe ingroup in this study were placed in seven tion are mechanical, not physiological or ge­ sections, in each of the two subgenera of On­ netic). More importantly and specific to this cidium (Table 4), as well as in Psychopsis and case, Trichocentrum is an extreme example of Trichocentrum. This disassociation of taxa so December 1989] CHASE AND PALMER-DNA SYSTEMATICS OF MONOCOTS 1729 well circumscribed by their distinctive vege­ been convinced that vegetative features offered tative features, as well as their much lower a more reliable basis for generic circumscrip­ chromosome numbers, is a result of the su­ tion. The evidence from analysis of cpDNA premacy of gross floral features in taxonomic has dramatically presented us with an hypoth­ matters. esis ofa previously unsuspected degree ofpias­ We suspect that the problem illustrated by ticity in floral morphology and casts suspicion this study is not an isolated case in the Or­ on evolutionary hypotheses that do not utilize chidaceae (and in many other all features of these plants. families, as well). Apparently rapid, drastic al­ teration of floral morphology has long been LITERATURE CITED suggested as a major factor in the evolution of the family (Dodson, 1962). If the overall ge­ BRUMMIT, R. K. 1985. Report of the committee for the netic basis is as equally minoras the differences Spermatophyta, 679. Taxon 34: 661. in cpDNA, then the floral differences that char­ CHASE, M. W. 1986. A reappraisal of the oncidioid or­ chids. Syst. Bot. II: 477-491. acterize Trichocentrum and O. bicallosum may --. 1987. Systematic implications of pollinarium result from relatively few alterations in the morphology in Oncidium Sw., Odontoglossum Kunth, gene(s) that control gross floral features. Ifthis and allied genera (Orchidaceae). Lindleyana 2: 8-28. is a model for orchids in general (and we intend --, AND R. G. OLMSTEAD. 1988. Isozyme number to test this hypothesis by genetic studies ex­ in subtribe Oncidiinae (Orchidaceae): an evaluation amining the transmission offloral types), then of polyploidy. Amer. J. Bot. 75: 1080-1085. --,AND J. D. PALMER. 1988. Chloroplast DNA vari­ gross floral morphology cannot be trusted to ation, geographical distribution, and morphological lead us to accurate phylogenetic hypotheses. parallelism in subtribe Oncidiinae (Orchidaceae). This has an ominous portent for a family in Amer. J. Bot. 75: 163-164. which floral morphology has been so heavily --, AND J. S. PIPPEN. 1988. Seed morphology in the emphasized. Oncidiinae and related subtribes (Orchidaceae). Syst. The application of molecular data to phy­ Bot. 13: 313-323. DE HEIJ, H. T., H. LUSTIG, D. M. MOESKOPS, W. A. Bo­ logenetic programs in the Orchidaceaeand oth­ VENBERG, C. BISANZ, AND G. S. P. GROOT. 1983. er lilioid monocot families has been lagging Chloroplast DNAs of Spinacia, Petunia, and Spiro­ behind that of similar studies in the dicots. dela have a similar gene organization. Curro Genet. The orchid clone bank described in this paper 7: 1-6. is a major step in fostering the practicality of DODSON, C. H. 1962. The importance of in the evolution ofthe orchids oftropical America. Amer. studies similar to the one reported here that Orchid. Soc. Bull. 31: 525-534, 641-649, 731-735. have a fairly narrow(interspecific, intergeneric) DOEBLEY, J., W. RENFROE, AND A. BLANTON. 1987. Re­ focus. One ofthe major phylogenetic problems striction site variation in the Zea chloroplast genome. facing systematists is that of interfamilial re­ Genetics 117: 139-147. lationships, which in general cannot be dealt DOYLE, J. J., AND J. L. DOYLE. 1987. A rapid DNA with by comparative restriction site mapping isolation procedure for small quantities of fresh leaf (Palmer et al., 1988a). The chloroplast ge­ tissue. Phytochem. Bull. 19: 11-15. DRESSLER, R. L. 1981. The orchids: natural history and nomes of the three different groups tradition­ classification. Harvard University Press, Cambridge. ally recognized as members of the Liliaceae --, AND N. H. WILLIAMS. 1982. Proposal for the sensu lato, while apparently identical in ge­ conservation of the generic name 1779 Oncidium nome arrangement, are probably too different Swartz (Orchidaceae) with a conserved type species, in size for this approach to be used. Sequence Oncidium altissimum Sw. Taxon 31: 752-754. data from conserved genes, chloroplast or oth­ ENOMOTO, S. Y., Y. OGIHARA, AND K. TSUNEWAKl. 1985. Studies on the origins of crop species by restriction erwise, will be needed in order to construct endonuclease analysis of organellar DNA. I. Phylo­ phylogenetic hypotheses for taxa that are di­ genetic relationships among ten cereals revealed by vergent in cpDNA size and sequence. None­ the restriction fragment patterns ofchloroplast DNA. theless, comparative restriction site mapping Jap. J. Genet. 60: 411-424. has added a new dimension to the study of FELSENSTEIN, J. 1985. Confidence limits on phylogenies: infrafamilial relationships in the Orchidaceae. an approach using the bootstrap. Evolution 39: 783­ 791. The evidence from several recent studies fo­ GARAY. L. A., AND J. E. STACY. 1974. Synopsis of the cused on the Oncidiinae supported the position genus Oncidium. Bradea I: 393-428. of Dressler and Williams (1982) that Tricho­ HILU,K. W. 1988. Chloroplast DNA in the systematics centrum is closely related to the vegetatively and evolution of the Poaceae. In T. R. Soderstrom, modified members of Oncidium. The grossly K. W. Hilu, C. S. Campbell, and M. E. Barkworth different floral morphology of [eds.), Grass systematics and evolution, 65-72. Smith­ Trichocentrum sonian Press, Washington, DC. suggests no derivative relationship from that HOWE, C. J., R. F. BARKER, C. N. BOWMAN, AND T. A. ofOncidium, and its modified habit only places DYER. 1988. Common features of three inversions it in affinity with this group. Few workers had in wheat chloroplast DNA. Curro Genet. 13: 343-349. 1730 AMERICAN JOURNAL OF BOTANY [Vol. 76

JANSEN, R. K., AND J. D. PALMER. 1987. Chloroplast 1983. Chloroplast DNA evolution and the origin of DNA from lettuce and (): struc­ amphidiploid Brassica. Theor. Appl, Genet. 65: 181­ ture, gene localization and characterization of a large 189. inversion. Curr. Genet. II: 553-564. ---, AND W. F. THOMPSON. 1981. Clone banks of ---, AND ---. 1988. Phylogenetic implications of mung bean, pea, and spinach chloroplast DNA. Gene chloroplast DNA restriction site variation in the Mu­ 15: 21-26. tisieae (Asteraceae). Amer. J. Bot. 75: 753-766. --, AND D. ZAMIR. 1982. Chloroplast DNA evolu­ LEHVASLAIHO, H., A. SAURA, ANDJ. LOKKI. 1987. Chlo­ tion and phylogenetic relationships in Lycopersicon. roplast DNA variation in the grass tribe Festuceae. Proc. Nat!. Acad. Sci. USA 79: 5006-5010. Theor. Appl, Genet. 74: 298-302. SYTSMA, K. J., AND L. D. GOTTLIEB. 1986. Chloroplast MARSH. J. L., M. ERFLE, AND E. J. WILKES. 1984. The DNA evolution and phylogenetic relationships in pIC plasmid and phage vectors with versatile cloning Clarkia sect. Peripetasma (Onagraceae). Evolution 40: sites for recombinant selection by insertional inacti­ 1248-1262. vation. Gene 32: 481-485. --, AND B. A. SCHAAL. 1985. of the NEI, M., AND W. LI. 1979. Mathematical models for Lisianthus skinneri (Gentianaceae) studying genetic variation in terms of restriction en­ in Panama using DNA restriction fragment analysis. donucleases. Proc. Nat!. Acad. Sci. USA 76: 5269­ Evolution 39: 594-608. 5273. TANAKA, R., AND K. KAMEMOTO. 1984. Chromosomes PALMER, J. D. 1985. Comparative organization ofchlo­ in orchids: counting and numbers. In J. Arditti [ed.], roplast genomes. Annual Rev. Genet. 19: 325-354. Orchid biology: reviews and perspectives, III, 321­ ---. 1986. Isolation and structural analysis of chlo­ 410. Cornell University Press, Ithaca, NY. roplast DNA. Meth. Enzymo!. 118: 167-186. TSUNEWAKI, K., AND Y. OGIHARA. 1983. The molecular ---, R. K. JANSEN, H. J. MICHAELS, M. W. CHASE, AND basis of genetic diversity among cytoplasms of Trit­ J. R. MANHART. 1988a. Chloroplast DNA variation icum and Aegilops species. II. On the origin of poly­ and plant phylogeny. Ann. Missouri Bot. Gard. 75: ploid wheat cytoplasms as suggested by chloroplast 1180-1206. DNA restriction fragment patterns. Genetics 104: IS 5­ ---, B. OSARIO, AND W. F. THOMPSON. 1988b. Evo­ 171. lutionary significance of inversions in legume chlo­ WIRTH, M. 1964. Supraspecific variation and classifi­ roplast DNAs. Curr. Genet. 14: 65-74. cation in the Oncidiinae (Orchidaceae). Ph.D. disser­ ---, C. R. SHIELDS, D. B. COHEN, AND T. J. ORTON. tation, Washington University, St. Louis, MO.