Ann. Bot. Fennici 35: 123–130 ISSN 0003-3847 Helsinki 30 June 1998 © Finnish Zoological and Botanical Publishing Board 1998

Allozyme diversity and phylogenetic relationships among diploid annual bromes (, )

Tatjana Oja & Vello Jaaska

Oja, T. & Jaaska, V., Institute of Zoology and Botany, University of Tartu, Riia 181, Tartu, EE-2400, Estonia Received 14 October 1997, accepted 11 March 1998

Phylogenetic relationships and genetic differentiation among eleven diploid annual brome species were evaluated by cladistic and phenetic analysis of the allozyme diversity of eight enzymes detected by PAGE. All species lacked heterozygous allozyme pheno- types, indicating prevalent autogamy and self-fertilization. Bromus japonicus Thunb. and B. squarrosus L. had the same allozymes, supporting their close genetic affinity. The placement of B. pumilio (Trin.) P. M. Smith in its own section Boissiera (Hochst. ex. Steudel) P. M. Smith was supported by its basal position in a separate clade. Mor- phologically uniform B. intermedius Guss. was the most polymorphic species, reveal- ing six isoenzyme lineages. The diploids of the section Genea Dum. were distinguished in a separate cluster on both cladistic and phenetic allozyme trees. Key words: Bromus, isoenzymes, phylogenetic relationships, Poaceae,

INTRODUCTION dependent taxonomic unit. She noted that there is a continuous range of variation between section Bromus L. (Poaceae) is a taxonomically complex Bromus and section Genea via the B. pectinatus genus with about 130 species of annual and per- Thunb. complex of section Bromus. This com- ennial, diploid and polyploid brome grasses of plex was suggested in order to link the Genea spe- wide geographic distribution in Eurasia, North and cies with the section Bromus through the diploid South America, and Australia. There is still B. japonicus. Recently, Pillay and Hilu (1995) and no consensus on the taxonomic treatment of Bro- Pillay (1996) reported the use of advanced con- mus. The genus has been divided into six sections temporary molecular methods with cpDNA and (Smith 1972), seven subgenera (Stebbins 1981) rDNA markers in the investigations of the genetic or even five different genera (Tsvelev 1976). The relationships in the genus Bromus. Pillay and Hilu number and rank of the divisions depends on the (1995), on the basis of the cpDNA data, suggested characters used as a basis for their delimitation that B. sterilis L. and B. tectorum L. are recently (serological, cytological and morphological, re- derived species that share a common maternal spectively), and phylogenetic relationships among ancestry. They also found that subgenera Steno- them are still inconsistent. Sales (1993) doubted bromus (Griseb.) Hackel and Bromus are not dis- the reality of the section Genea Dum. as an in- tinct entities and probably originated from simi- 124 Oja & Jaaska • ANN. BOT. FENNICI 35 (1998) lar ancestors related to the subgenus Festucaria 1. Bromus pumilio (Trin.) P. Smith: 1 accession, from Den- (Coss. & Dur.) Hackel. Pillay (1996) concluded mark. it unlikely that the subgenera vs. sections of Bro- 2. Bromus danthoniae Trin.: 2 Turkish accessions. 3. Bromus alopecuros Poir.: 3 Turkish accessions. mus had independent origins. Despite this new 4. Bromus scoparius L.: 3 accessions, from Kirghizia (1), molecular information, phylogenetic relationships (1) and Central (1). among the subdivisions of Bromus are still far 5. Bromus japonicus Thunb.: 21 accessions, from from clear. (8), Central (7), (1), Kirghizia (1), Ger- In addition to inconsistent intrageneric classi- many (1), (2) and the (1). fication, there is a lot of confusion about species 6. Bromus squarrosus L.: 16 accessions, from Turkey (10), Hungary (2), (1), Greece (1) and (2). delimitation within the sections, due to their great 7. Bromus intermedius Guss.: 17 accessions, from Turkey. and frequently continuous morphological variabil- 8. Bromus pseudosecalinus P. Smith: 1 accession, from ity, and less well known genetic relationships England. among them. Among the diploids, some species 9. Bromus sterilis L.: 27 accessions, from Turkey (7) and are easily distinguished by unique morphological from central and southern (20). characters, like Bromus pumilio (Trin.) P. M. 10. Bromus tectorum L.: 34 accessions, from central and southern Europe (17) and from Turkey (17), including Smith and B. danthoniae Trin., but other diploids six Turkish accessions of B. tectorum L. ssp. lucidus of the type section are less clearly differentiated Sales (= B. sericeus Drob.) and difficult to identify since their diagnostic char- 11. Hordeum bogdanii Wilensky: 2 accessions, from Paki- acters are often overlapping and difficult to de- stan and China. scribe exactly (Smith 1973). Smith and Sales Vouchers are deposited in TAA, Herbarium of the In- (1993) tried to clarify the taxonomy of some dif- stitute of Zoology and Botany (Tartu, Estonia). Some of ficult taxa of the section Bromus and recognized them were grown from the seed accessions in order to check four affinity groups among them. Ainouche and their botanical identifications. Bayer (1996) confirmed the weak divergence of the diploid species within the section Bromus ac- cording to the ITS sequences of the nrDNA. Isoenzyme analysis and nomenclature The goal of the present study was to analyze the differentiation among diploid brome grasses Enzyme extracts were prepared from the shoots (the pri- mary leaf with the coleoptile) of 5–10-day-old, etiolated between and within sections on the basis of elec- seedlings subjected to electrophoresis in vertical polyacry- trophoretic allozyme characters and to compare lamide gel slabs and stained as described in Jaaska and Jaas- allozyme diversity between taxa in their native ka (1986, 1990). Four individual seedlings of each acces- East Mediterranean and other regions (Europe and sion were routinely analyzed for all isoenzymes. For some Central Russia). Cladistic and phenetic analysis enzymes and accessions, up to ten individuals were analyzed of the allozyme characters was used to understand in order to check the constancy or polymorphism of the isoenzyme patterns. Isoenzymes encoded by separate loci the phylogenetic relationships and genetic affini- of a diploid genome are designated by capital letters followed ties among diploid species as ancestral to poly- by a number denoting the allozymic variants (in short, ploids. This was necessary in order to clarify the morphs) in the order of decreasing mobility. Heterozygous relationships within the whole genus. We also ana- phenotypes are designated by codominant allozymes sepa- lyzed how divergence by isoenzyme phenotypes rated by a diagonal line, i.e. as a fraction. The term “fixed was correlated with the morphological species de- heterozygosity” is used when heterozygous phenotypes of isoenzymes were observed for all analyzed individuals with- limitation. out segregation into homozygous phenotypes.

MATERIAL AND METHODS Data analysis

Plant material Pee-Wee program vers. 2.5.1 (Goloboff 1993) which finds trees with the best total fit of characters and calculates de- Most of the seeds were collected in Turkey and Central cay indexes (DI) of branch support (Bremer 1994) was used Russia by the first author. The accessions of other geographi- for the cladistic analysis of the allozyme data. Phenetic ana- cal origins, also collected in the wild from known locations, lysis was perfomed with the TYTAN program provided by were received from different botanical gardens. Dr. A. Batko (Warsaw University, ) by applying the ANN. BOT. FENNICI 35 (1998) • Allozyme diversity and phylogenetic relationships 125

Manhattan distance method combined with the UPGMA meth- The genetic interpretation of zymograms followed od. Both analyses were conducted on the multilocus allo- the results of our previous comparative studies of zyme lineages identified in the species studied. The isoen- isoenzymes in diploid and polyploid bromes (Oja zyme data matrix was constructed using allozymes as binary absence/presence characters. Rare morphs were excluded & Jaaska 1996, Oja 1998). from the matrix. The autogamous diploid perennial Hordeum A characteristic feature of the diploid annual bogdanii Wilensky of the sister tribe Triticeae Dum. was bromes examined was the total absence of hetero- chosen as an outgroup for comparison in cladistic analysis. zygous allozyme phenotypes of most heterozymes except PGD, despite the existence of intraspecific allozyme polymorphism of several heterozymes. RESULTS AND DISCUSSION This result is in accordance with the data on the autogamy and self-fertilization of annual bromes Data on the isoenzyme variation among the ten (Smith 1972) and shows prevalent inbreeding in brome-grass species is summarized in Table 1. the natural populations of all diploids examined.

Table 1. Electrophoretic variants of alcohol (ADH), malate (MDH) and shikimate (SKD) dehydrogenase, aspartate aminotransferase (AAT), superoxide dismutase (SOD), peroxidase (PRX), 6-phosphogluconate dehydrogenase (PGD) and phosphoglucoisomerase (PGI) heterozymes in Bromus species: Allozyme lineages within the species are num- bered, with their geographic origin designed with letter T (Turkey), CR (Central Russia), Eu (Europe) and CA (Central Asia). No. = the number of accessions analyzed, s = a slightly slower variant of the numbered morph, and f = a slightly faster variant of the numbered morph, r = a rare morph. The unique (autapomorphic) allozymes are in bold. ————————————————————————————————————————————————— Species and No. ADH MDH SKD AAT AAT SOD SOD PRX PGD PGI allozyme lineages A A A B C A B F A|B A ————————————————————————————————————————————————— Hordeum bogdanii 232f 3 622 19 3 5 Section Boissiera B. pumilio Eu 1 3 2 2 6 32 2 1 2|3 7 Section Bromus B. danthoniae T 23325 32 2 6 2|35 B. alopecuros T1232253422s 24 B. alopecuros T2 1 3 2 2 5 3 4 12 23 B. scoparius CA1 2 3 3 3s 436 24 2 4 B. scoparius CA2 1 7 324 36 24 2 4 B. japonicus CR 8332534262|34 B. jap-sq T1 9332434262|34 B. jap-sq T2 2332434262|35 B. squarrosus T1233253426f 2|3 4 B. squarrosus T25332434261|3 4 B. squarrosus CR1332434252|34 B. intermedius T11322534252|34 B. intermedius T21323434252|34 B. intermedius T323235342 5,3 2|3 4 B. intermedius T44323434252|33 B. intermedius T57323534252|33 B. intermedius T61323534242|33 B. arvensis T1 1322534252|33 B. arvensis T2 1323534242|35 B. arvensis Eu 123224 34 2 4 2|33 B. pseudosecalinus Eu 1 3 3 3 3 34 2 5 2|4 4 Section Genea B. sterilis T1 443253428 2 7 B. sterilis T2 243253428 2 6 B. sterilis T3 143253428 2 5 B. sterilis Eu 20 4 3 2 5 3 4 2 8,6r 2 6 B. tectorum T1743251 425 2 6 B. tectorum Eu 24 4 3,2r 2 5 1 425 2 6 ————————————————————————————————————————————————— 126 Oja & Jaaska • ANN. BOT. FENNICI 35 (1998)

No case of heterozygous isozyme phenotype was rum ssp. lucidus by general morphology (synapto- recorded in the accessions Bromus intermedius spermous, but still with lower glume 1-veined) and B. squarrosus, in contrast to the recent find- had an isozyme phenotype absolutely identical to ings of Ainouche et al. (1995). that of B. tectorum subsp. tectorum. This evidence PGD was exceptional among the enzymes ana- is consistent with the conclusion of Sales (1991) lyzed by showing fixed heterozygosity with per- who reduced B. sericeus to B. tectorum subsp. manent three-banded phenotypes in several dip- lucidus on the basis of the morphological and re- loids. Of the ten species studied, only Bromus alo- productive characters, and favouring even the lower pecuros and B. scoparius from the type section taxonomic rank of mere morphological variety. and B. sterilis, B. tectorum from the Genea sec- The accessions attributed to the Bromus ja- tion, had the homozygous variant A2. The most ponicus-squarrosus complex by morphology dis- plausible explanation for the permanent occur- played homologous variation with shared morphs rence of a symmetrical three-banded, heterozy- at several heterozymes. The accessions from Tur- gous phenotype of dimeric 6-PGD in the autoga- key demonstrated relatively greater interpopula- mous diploids is that it is encoded by a duplicated tional variability than the accessions from other locus with divergent alleles. It is likely that the parts of the species range. The Turkish accessions permanent three-banded phenotype reflects het- of B. japonicus-squarrosus showed polymor- erozygosity due to gene duplication with concur- phism of SKD-A, PGD-A and PGI-A with two rent modification and the appearance of a modified, allozymes and of PRX-F with three allozymes. paralogous locus encoding heterozyme PGD-B. The accessions from Europe and Central Russia Intraspecific variation with two allozymes of attributed by morphology to typical B. japonicus ADH-A, SKD-A, PGD-A, AAT-B, SOD-B, PRX-F and B. squarrosus, respectively, were differentiated and PGI-A was observed in all seven diploids for from each other by alternative allozymes of AAT- which sufficient numbers of accessions were ana- B and PRX-F. In general, taxonomists have rec- lysed: Bromus alopecuros, B. scoparius, B. japo- ognized that B. japonicus is a highly variable taxon nicus, B. squarrosus, B. intermedius, B. arvensis, with at least two subspecies: typical subsp. japon- and B. sterilis. An important consequence of the icus and subsp. anatolicus, the latter almost ex- prevalent self-fertilizing breeding system in the actly fitting the description of B. squarrosus diploid bromes examined is their intraspecific dif- (Smith 1980, 1985a). Our experience with grow- ferentiation into distinct multilocus allozyme ge- ing voucher reproductions has shown the great notypes maintained as multilocus allozyme (iso- influence of environmental conditions on brome- zyme) lineages. Thus, isoenzyme analyses revealed grass phenotype. For example, an original acces- two isozyme lineages in B. alopecuros with differ- sion of B. japonicus collected in Turkey had a ent allozymes of 3 heterozymes (SOD-B, PRX-F lower lemma 10 × 6 mm and uppermost and PGI-A) and in B. scoparius with differences 20 mm, while its reproduction in Estonia had a in two allozymes of ADH-A and SKD-A. Bromus lower lemma 13 × 8 mm and uppermost spikelet arvensis from Turkey also showed two isozyme 35 mm, thus resembling B. squarrosus more. phenotypes (differentiated by SKD-A and PGI-A) There seems to be a continuous range of variation which both differed from the European accessions from B. japonicus to B. squarrosus. A set of eleven in AAT-B allozymes. Turkish accessions was morphologically interme- Accessions of Bromus sterilis, from Turkey, diate between typical B. japonicus and B. squar- displayed three isozyme lineages, which differed rosus (designed as jap-sq in Table 1). They were only in PGI-A allozymes. The European acces- differentiated into two isozyme lineages. The data sions were identical with the B. sterilis lineage in Table 1 show that genetic differentiation in the T2 from Turkey, except some having PRX-F6 in- complex by allozymes is independent of differen- stead of common PRX-F8. All accessions of tiation by morphology. B. tectorum from Turkey were monomorphic, The allozyme data suggests that Bromus ja- whereas some accessions from Europe had only ponicus and B. squarrosus are closely related not one additional allozyme MDH-A2. Six Turkish only morphologically, but also genetically. The accessions which could be attributed to B. tecto- complex needs more detailed investigations in ANN. BOT. FENNICI 35 (1998) • Allozyme diversity and phylogenetic relationships 127 order to determine the correct taxonomic rank of a length of 87 steps and a total fit of 178.9. The B. japonicus and B. squarrosus, but our allozyme consensus tree of the 18 fittest trees (Fig. 1) shows data favours their intraspecific status. that Bromus pumilio is linked to Hordeum bog- Isoenzyme evidence on three unique allo- danii as the outgroup species and is basally para- zymes in its electrophoretic phenotype clearly phyletic with respect to all other diploids studied. shows that Bromus pumilio stands apart from the This indicates its outstanding position in the ge- other diploids of the section Bromus, where it has nus and supports its recognition in a separate sec- been placed in some treatments (Smith 1970). This tion by Smith (1985b). Another distinctly differen- is in accordance with its divergence by morpho- tiated species among annual diploids seems to be logical and dispersal mechanism characteristics B. danthoniae which is basally paraphyletic to the (Smith 1985b). The suggestion to put B. pumilio other species of the section Bromus. Decay in- in its own section Boissiera (Smith 1985b) is thus dexes of the B. danthoniae clade from the sister supported by our allozyme data. Contrary to a clades of the B. japonicus-squarrosus complex report about the tetraploid chromosome number and of B. pumilio are 3.4 and 4.3 respectively. for B. pumilio (Smith 1972), our isoenzyme re- Smith (1985b) also supposed that B. pumilio is sults showed that B. pumilio should be a diploid most closely related to B. danthoniae and B. alo- species. This was confirmed by showing that our pecuros in the genus. The clear allozyme differen- accession of B. pumilio had 2n = 14 (H. Laarmann tiation of B. danthoniae is congruent with its diver- unpubl.). The diploid chromosome number was gence by a unique morphological character (this also given by Avdulov (1931). species has a three-awned lemmas) in the section Bromus intermedius was the most polymor- Bromus. phic species among the diploids studied, reveal- The Bromus japonicus-squarrosus complex ing six isozyme lineages. Despite remarkable allo- has a central position on the allozyme cladogram, zyme variability, all the examined accessions of but was not clearly resolved into two different spe- B. intermedius were morphologically quite uni- cies. The morphologically intermediate allozyme form with slender, tangled panicle branches and lineage japonicus-squarrosus T2 together with the hairy with recurved awns, as described typical B. japonicus from Central Russia are ba- by Smith (1985a). It is remarkable that the pheno- sal in the consensus cladogram and are only weak- type of specimens of Turkish accessions of B. in- ly (DI only 0.4) differentiated from the interme- termedius reproduced in Estonia was less affected diate lineage T1 and the typical B. squarrosus lin- by environmental growing conditions than in the eages T1 and T2. case of B. japonicus-squarrosus. Bromus alopecuros and B. scoparius of the The ten heterozymes of the eight enzymes ex- section Bromus are linked on the cladogram as amined revealed 36 shared and 14 unique morphs sister species and formed one subclade with spe- (putative allozymes) occurring with different fre- cies belonging to the section Genea. Bromus alo- quencies and patterns within and among the spe- pecuros seems to be less diverged from the B. ja- cies. Heterozymes differed remarkably in the ex- ponicus-squarrosus complex, than B. scoparius, tent of genetic variability, AAT-C and SOD-B with the Bremer’s supports (DI values) 1.0 and were the most conservative heterozymes in this 3.8 respectively. section, with one morph in common to most spe- Section Genea is recognised on the cladogram cies. PRX-F, PGI-A and PGD-A were the most as a monophyletic group which is well diverged variable heterozymes with nine, five and four from the central Bromus japonicus-squarrosus morphs, respectively. complex (DI = 4.1) and is linked to the B. alo- Cladistic analysis of the allozymic variation pecuros–B. scoparius clade. The cladogram shows pattern among the species and their isozyme line- the B. tectorum and B. sterilis T2 lineage as sister ages was performed on the basis of the presence/ species, supporting their close phylogenetic af- absence (0/1) data matrix of 36 phylogenetically finity. The cpDNA data also suggests that B. steri- informative shared allozyme characters compiled lis and B. tectorum are recently derived species from the data in Table 1. Analysis of the data ma- that share a common maternal ancestry (Pillay & trix with Pee-Wee gave 18 trees of best fit, having Hilu 1995). 128 Oja & Jaaska • ANN. BOT. FENNICI 35 (1998)

Fig. 1. The consensus tree of the 18 fittest trees for the diploid annual Bromus species from the Pee-Wee analyses of the data matrix of 36 shared allozymes of 10 heterozymes. Decay indexes are given at branches.

The six morphologically cryptic allozyme lin- Bromus pumilio and B. danthoniae are two ba- eages within Bromus intermedius are less differen- sally ancestral brome species on the phylogram tiated from the B. japonicus-squarrosus complex that revealed the duplication of the PGD locus (DI = 0.8), being divided into two closely related with PGD-AB2/3, whereas the outgroup species, groups and mixed with B. arvensis. The recent perennial Hordeum bogdanii, showed no duplica- study by Ainouche and Bayer (1997) on the basis tion. The same duplication was found in most of of rDNA ITS sequences also showed that the dip- the diploids studied. This indicates that duplica- loid bromes Bromus japonicus, B. squarrosus, tion with the emergence of paralogous PGD-A2 B. arvensis and B. intermedius belong to a weakly has occurred in B. pumilio and is maintained in differentiated clade within the section Bromus, most diploid bromes. Only B. alopecuros, B. sco- which agrees well with our results. Both the ITS parius and diploids of the section Genea (B. ste- and allozyme cladograms reveal the same sister rilis and B. tectorum), which fell within the same species couples, e.g. B. japonicus–B. squarrosus derived subclade on the cladogram again, lacked and B. intermedius–B. arvensis. the duplication and had only PGD-A2. This sug- The only remarkable discrepancy between the gests inactivation of the duplicated PGD-B3 in ITS and allozyme data is in the phylogenetic po- this derived subclade. Further divergence of du- sition of the section Genea. Thus, the ITS clado- plicated PGD-AB with the appearance of PGD-A1 gram placed Bromus tectorum of the section and PGD-B4 was detected for B. squarrosus T2 Genea as a sister species to B. anomalus Rupr. ex and B. pseudosecalinus, respectively. Fourn. of the section Pnigma Dum. (Ainouche & Phenetic analysis was performed on the basis Bayer 1997), whereas our allozyme cladogram of the presence/absence data matrix of the 36 linked the Genea diploids to the B. alopecuros– shared, plus 14 unique, allozymes compiled from B. scoparius subclade of the type section. The Table 1, by applying Manhattan distance with the possible reason for this disagreement is that no UPGMA method of clustering. diploid Pnigma species was available in our study, The allozyme phenogram (Fig. 2) is rather sim- and therefore the cladistic program found the most ilar to the cladogram in the general topology in closely related species of those studied. recognizing similar clusters of related species, but ANN. BOT. FENNICI 35 (1998) • Allozyme diversity and phylogenetic relationships 129

Fig. 2. UPGMA phenogram of Manhattan distances for the diploid annual bromes, based on a presence/absence data matrix of 36 shared and 14 unique allozymes of 10 heterozymes compiled from Table 1. Scale relative, 0 = minimal distance. supplementing it by evaluating the extent of ge- netic trees in a separate cluster which is linked to netic divergence between the species and allozyme B. alopecuros of the section Bromus. lineages. The most significant difference between the Thus, Bromus pumilio is distinguished in a sep- phenogram and cladogram is in the position of arate phenetic branch, having also appeared in a Bromus scoparius which, according to the pheno- separate clade on the cladogram. gram, is the most divergent species among the The phenogram shows that Bromus japonicus annual diploids studied, supported by unique allo- CR and B. squarrosus T1 are the least divergent zymes ADH-A7, SKD-A3 and SOD-A6, followed couple linked with other allozyme lineages of the by B. pumilio with unique allozymes AAT-B6, B. japonicus-squarrosus complex in the same PRX-F1 and PGI-A7 (Table 1). cluster. Bromus danthoniae is linked with the The results presented above should be consid- B. japonicus-squarrosus cluster at a higher diver- ered only as a preliminary attempt to understand gence distance. As in the cladogram, the B. inter- phylogenetic relationships in the genus Bromus medius–B. arvensis couple is joined in a separate by applying phenetic and cladistic analysis of the clade which is linked with the B. japonicus-squar- variability pattern of molecular allozyme char- rosus cluster at a higher divergence distance. Dip- acters among the ancestral diploid species. Fu- loids B. sterilis and B. tectorum, belonging to the ture studies involving perennial diploids of the section Genea, appear on both cladistic and phe- section Pnigma, which were not available for the 130 Oja & Jaaska • ANN. BOT. FENNICI 35 (1998) present study, as well as more accessions and spe- zen 181: 301–320. cies of annual diploids with the use of a larger Jaaska, V. & Jaaska, V. 1990: Isoenzyme variation in Asian number of molecular markers will hopefully give beans. — Bot. Acta 103: 281–290. Oja, T. 1998: Isoenzyme diversity and phylogenetic affini- a more substantiated picture of phylogenetic rela- ties in the section Bromus of the grass genus Bromus tionships in the brome genus at the ancestral dip- (Poaceae). — Biochem. Syst. Ecol. [In print.] loid level. Oja, T. & Jaaska, V. 1996: Isoenzyme data on the genetic divergence and allopolyploidy in the section Genea of Acknowledgements: The research was supported partly by the grass genus Bromus (Poaceae). — Hereditas 125: the grant Nr. 2010 from the Estonian Science Foundation 249–255. to Dr. V. Jaaska. We thank Prof. Erast Parmasto (Depart- Pillay, M. 1996: Genomic organization of ribosomal RNA ment of Mycology, Institute of Zoology and Botany, Univer- genes in Bromus (Poaceae). — Genome 39: 198–205. sity of Tartu) for his assistance with the computer programs. Pillay, M. & Hilu, K. W. 1995: Chloroplast-DNA restric- tion site analysis in the Genus Bromus (Poaceae). — Am. J. Bot. 82: 239–249. REFERENCES Sales, F. 1991: A re-assessment of Bromus tectorum: a com- puter analysis, synaptospermy and chorispermy. — Flora et Vegetatio Mundi IX: 29–41. Ainouche, M. & Bayer, R. J. 1996: On the origins of two Mediterranean allotetraploid Bromus species: Bromus Sales, F. 1993: Taxonomy and nomenclature of Bromus hordeaceus L. and B. lanceolatus Roth. (Poaceae).— sect. Genea. — Edinburgh J. Bot. 50: 1–31. Am. J. Bot. 83: 135. Smith, P. M. 1970: Taxonomy and nomenclature of the Ainouche, M. & Bayer, R. J. 1997: On the origin of the brome-grasses (Bromus L. s.l.). — Notes RBG Edin- tetraploid Bromus species (section Bromus, Poaceae) burgh 30: 361–375. – insights from internal transcribed spacer sequences Smith, P. M. 1972: Serology and species relationships in of nuclear ribosomal DNA. — Genome 40: 730–743. annual bromes (Bromus L. sect. Bromus). — Ann. Bot. Ainouche, M., Misset, M. T. & Huon, A. 1995: Genetic 36: 1–30. diversity in Mediterranean diploid and tetraploid Bro- Smith, P. M. 1973: Observations on some critical Brome- mus L. (section Bromus Sm.) populations.— Genome grasses. — Watsonia 9: 319–332. 38: 879–888. Smith, P. M. 1980: Bromus L. — In: Tutin, T. G., Heywood, Avdulov, N. P. [Avdulov N. P.] 1931: Karyo-systematische V. H., Burges, N. A., Moore, D. M., Valentine, D. H., Untersuchungen der Familie Gramineen. — Trudy Walters, C. M. & Webb, D. A. (eds.), Flora Europea: 5: Prikl. Bot. Priloz. 44: 1–342. 182–189. Cambridge Univ. Press, Cambridge. Bremer, K. 1994: Branch support and tree stability. — Smith, P. M. 1985a: Bromus L. — In: Davis, P. H. (ed.), Cladistics 10: 295–304. Flora of Turkey and the East Aegean Islands: 9: 272– Crawford, D. J. 1989: Enzyme electrophoresis and 301. Univ. Press, Edinburgh. systematics. — In: Soltis, D. & Soltis, P. (eds.), Iso- Smith, P. M. 1985b: Observations on Turkish brome- zymes in plant biology: 146–164. Portland, Oregon. grasses. I. Some new taxa, new combinations and notes Goloboff, P. A. 1993: Estimating character weights during on typification. — Notes Royal Bot. Garden Edinburgh tree search. — Cladistics 9: 83–91. 42: 491–501. Gottlieb, L. D. 1977: Electrophoretic evidence and plant Smith, P. M. & Sales F. 1993: Bromus L. sect. Bromus: systematics. — Ann. Missouri Bot. Garden 64: 161–180. Taxonomy and relationships of some species with small Jaaska, V. & Jaaska, V. 1984: Isoenzymes of aromatic alco- spikelets. — Edinburgh J. Bot. 50: 149–171. hol dehydrogenase in rye and triticale. — Biochem. Stebbins, G. L. 1981: Chromosomes and evolution in the Physiol. Pflanzen 179: 21–30. genus Bromus (Gramineae). — Bot. Jahrb. Syst. 102: Jaaska, V. & Jaaska, V. 1986: Isoenzyme variation in the 359–379. barley genus Hordeum L. 1. Alcohol dehydrogenase Tsvelev, N. N. [Cvelëv N. N.] 1976: [Poaceae USSR]. — and superoxide dismutase. — Biochem. Physiol. Pflan- Nauka, Leningrad. 778 pp. [In Russian.]