A SYNOPSIS OF CYTOLOGICAL STUDIES IN GESNERIACEAE

Our knowledge of cytological data published on members of the family Gesneriaceae is summarized and critically evaluated in the light of current taxonomic treatments and phylogenetic hypotheses. There are about 1000 published chromosome counts, covering 56% of the genera but only 18% of the species. In particular the New World tribes Beslerieae and Napeantheae and the Old World tribe Didymocarpeae are underexplored at generic level. In Gesneriaceae chromosome data are a valuable source of taxonomic characters. From our current knowledge of the phylogenetic relationships in the family we know that basic chromosome numbers in the New World subfamily Gesnerioideae appear to be rather conserved, but that a more complex pattern of genome evolution seems to be present among the Old World tribes. Both polyploidy and dysploid changes have played a significant role in the evolution of the family. However, the number of species for which both cytological and molecular data are available is at present too low to reach firm conclusions on ancestral basic chromosome numbers, particularly for the Old World group. To facilitate wider access to cytological data on the Gesneriaceae , a website has been developed (http://www.rbge.org.uk/rbge/web/search/index.jsp), which is introduced in this paper.


INTRODUCTION
Members of the Gesneriaceae are predominantly tropical and subtropical herbs, currently grouped into three subfamilies (Wiehler, 1983;Burtt & Wiehler, 1995): Cyrtandroideae, all Old World (OW) except Rhynchoglossum azureum (Schltdl.) B.L. Burtt; the New World (NW) Gesnerioideae; and the southern hemisphere Coronantheroideae (sometimes treated as tribe Coronanthereae in subfamily Gesnerioideae). Recent molecular analyses do not indicate subfamily status for the Coronantheroideae as its members are placed closer to the NW tribes Beslerieae and Napeantheae and not as sister to the entire subfamily Gesnerioideae (Mayer et al., 2003;Wang et al., 2004). The monotypic genus Titanotrichum Soler. of the monogeneric tribe Titanotricheae is also more closely associated with NW tribes than previously thought and may eventually have to be transferred to subfamily Gesnerioideae (Wang et al., 2004). However, the purpose of the present paper is not to introduce a new Gesneriaceae systematic, but to give an overview of current cytological knowledge and to discuss chromosome evolution in this family, where possible, in the light of phylogenetic hypotheses. Thus, we use the taxonomic system of Burtt & Wiehler (1995).

HISTORY O F C H R O M O S O M E S T U D I E S I N GESNERIACEAE
The first reported chromosome counts in Gesneriaceae were published as early as 1923 by Oehlkers. During the following 30 years, a slowly increasing number of counts was recorded. In the 1950s and 1960s, due mainly to the work of Ratter, Milne & Prentice (1963-75) on OW taxa at the Royal Botanic Garden Edinburgh (RBGE) and Lee et al. (1950-67) on NW taxa, the number of counts increased significantly, reaching more than 200 and 115 respectively for the two research groups. Other major contributors were Rogers (1954-62), Eberle (1956-57), Fussell (1958, Morley et al. (1967-72), Davidse (1970-81), Wiehler (1972-76) and, more recently, Oliver & Skog (1981-85) for NW taxa. (For original citations see Ratter, 1975;Skog, 1984.) Kiehn and co-workers in Vienna added over 120 counts for OW taxa, mainly in subfamily Cyrtandroideae from Malaysia . Their papers also included discussion on the chromosome evolution of individual genera. Jong and co-workers (2000-01) continued the tradition of Ratter's work at RBGE with their contributions on Aeschynanthus Jack and Streptocarpus Lindl. Summaries of chromosome data available for Gesneriaceae were provided by Ratter (1975) and Skog (1984). To date about 1000 chromosome counts have been published for the family. These can be searched interactively in a database recently developed at RBGE (see section 9).
Continued cytological output in OW and NW taxa is evident from the overlapping years of publications by the major contributors (Table 1). It is TABLE 1 . List of selected authors of major cytological publications in Gesneriaceae, arranged by region of interest and date of publication. For original citations of literature before 1984 see Ratter (1975) and Skog (1984) New World Old World

CYTOLOGICAL C O V E R A G E
Representatives of all subfamilies have been counted. At tribal level, however, cytological data for one of the 11 tribes, the monogeneric NW Napeantheae, are still lacking. Across the family at least one species has been counted in 56% of the genera, with 65% and 56% respectively in the NW subfamilies Gesnerioideae and Coronantheroideae, and 51% in the OW subfamily Cyrtandroideae ( Table 2). The lower number for the latter is due mainly to the large OW tribe Didymocarpeae, with 70 genera the largest in the family, in which only 46% of genera have been counted.
The Gloxinieae is the most comprehensively investigated tribe in the NW (apart from the small tribe Gesnerieae), with 82% of the genera and 36% of the species counted. In comparison, in the Episcieae, with a similar number of genera, only about two-thirds of the genera and one-fifth of the species have been analysed. The tribe Beslerieae has hardly been investigated and only two species have been counted, despite it containing the large genus Besleria L., with more than 200 species. Amongst the Cyrtandroideae, tribe Epithemateae is most comprehensively covered, with counts for 26% of its taxa, representing 86% of the genera. However, this tribe includes only seven genera and around 84 species. In the largest OW tribe, Didymocarpeae, comprising more than 950 species, only a quarter (24%) of the species, covering less than half of the genera (46%), have been cytologically investigated. Tribe Cyrtandreae is the least investigated with only 6% of the species counted. This is due mainly to the small number of species investigated in the large genus Cyrtandra J. R. & G. Forst. (see below). Chromosome counts at species level across the family have been made for only about 18% (i.e. 590 out of 3347 species, see Table 2), mainly because of low coverage in some large genera (e.g. Cyrtandra, Besleria). On average, sampling of the NW and OW species has been at a similar level, 19% and 17%, respectively. Of the larger genera with more than 40 species, the NW Sinningia Nees (34%) and the OW Streptocarpus (42%) and Henckelia Spreng. (30%) are particularly well analysed (Table 3). On the other hand, the large genera Alloplectus Mart. (8%), Besleria (1%), Drymonia Mart. (7%) and Paradrymonia Hanst. (1%) of the NW, and Cyrtandra (6%) and Agalmyla Blume (2%) of the OW are particularly in need of investigation, as only 1-8% of their species have been counted. The low coverage in Cyrtandra is a result of the large number of species, around 600; of the c.60 accessions counted in 36 species the great majority reveal a diploid number of 2n=34; only three reports (Borgmann, 1964) give 2n=32.
Of the medium-sized NW genera with around 20 taxa, 92% of Achimenes Pers. species have been counted, followed by, in decreasing order: Kohleria Regel (77%), Codonanthe Hanst. (65%), Nematanthus Schrad. (40%), Gesneria s. 1. (Skog, 1976)    The number of genera without chromosome counts is higher in the OW (43 genera) than the NW (18 genera). This is not surprising as there are almost twice as many genera of Gesneriaceae in the OW and more small, monotypic genera: OW 32 genera (37%); NW eight genera (15%). The problem with monotypic genera is that plants usually occur in small populations in few locations and are often endangered, or already extinct as is apparently the case with Gyrogyne W.T. Wang from China (Y. Z. Wang, pers. comm.). Cytological data are thus available for only eight of the monotypic genera of the OW and for one genus in the NW. In the absence of morphological characters or character combinations that might indicate affinities with larger genera, obtaining cytological data for monotypic genera before they become extinct is of great importance. Unlike morphological or molecular data that can be retrieved even from very old herbarium specimens (Drabkova et al., 2002), cytological data can be reliably obtained only from living material. This underlines the importance of in situ protection for such groups as well as of well-curated ex situ collections.

TYPES A N D S I G N I F I C A N C E O F C Y T O L O G I C A L D A T A
Cytotaxonomy is not necessarily merely concerned with collecting chromosome numbers for the sole purpose of establishing taxonomic groups. Theories on dysploid reduction or increase in chromosome number, caused, for example, by Robertsonian events, are often put forward to explain variation in closely related groups (e.g. Hair, 1963, for Podocarpaceae;Wang et al., 1998, for Epithemateae;, for Henckelia (Didymocarpeae); see section 8). In reality, however, explanations of observed differences in chromosome data are often more complex, such as comparisons of chromosome data with molecular phylogenies, e.g. the Epithemateae phylogeny by Mayer et al. (2003); see also section 8. Gesneriaceae, in general, possess relatively small chromosomes, frequently in the<1 to 2µm range, and as a result few karyotypes have been published so far, mainly in Chinese taxa (e. g. Wang et al., 1998;Wang & Gu, 1999;Zhou et al., 2004 [this issue]). Because of their relatively small size, it is necessary to exercise great care when exceptionally large nucleolar organizer regions (NORs) are present that may tend to detach; incorrect counts can be the result of not recognizing such 'satellites'. This has apparently happened, for example, in Lysimachia nemorum L. (Primulaceae), which has a published number of 2n=18 but a correct number of 2n=16 (P. M. Hollingsworth, pers. comm.). In Gesneriaceae such large NORs are thought to exist, for example in the Chinese genus Anna Pellegr., where 2n=34 was established with certainty for two species by Zhou et al. (2004). However, prometa-and metaphases frequently give the appearance of 2n=36, as the stalk linking the relatively large satellite to the not very much larger chromosome is often not clearly visible ( Fig. 1).
Detailed analyses of mitotic metaphase nuclei, beyond mere establishment of chromosome numbers, are essential where no difference in numbers between taxa can be established. These analyses basically involve detailed measurements of individual chromosome arms, establishment of centromere position and determination of the position and number of NORs. Interspecific comparisons of such karyotypes may well enable further species or even population distinctions, such as in Conandron ramondioides Sieb. & Zucc. where it was demonstrated that populations from Japan varied in NOR number in otherwise virtually indistinguishable chromosome complements (Kokubugata & Peng, 2002). Details of chromosomal rearrangements, such as inversions, duplications, deletions and translocations, can also be valuable characters for phylogenetic analyses (Borowik, 1995;Levin, 2002).
Additional cytological data can be gathered from characters in interphase and prophase nuclei. Much variation in interphase nuclei was demonstrated by Tischler in 1934, and has been applied in a number of different plant groups, such as Orchidaceae (Tanaka, 1971). Zhou et al. (2004) have summarized their own data and those published by others on interphase and prophase characters in OW Gesneriaceae. Although variation was found that spanned current tribal delimitations in subfamily Cyrtandroideae, the authors were careful not to draw far-reaching conclusions, as too few taxa have been analysed and preliminary molecular phylogenetic data suggest that the current tribes do not form natural groups (Wang & Li, 1998;Pfosser et al., unpublished).
Meiotic pachytene chromosomes are still relatively long and are thus ideal for morphological analyses, particularly as at this stage fewer chromosomal bodies are apparent, because of the pairing of homologues as bivalents. Eberle (1956) analysed pachytene chromosomes from a range of taxa across the family and noted fundamental differences between NW and OW taxa. Rhynchoglossum of tribe Epithemateae apparently possesses a chromosome type unique in the family, with all chromosome pairs being predominantly euchromatic and little heterochromatin with small chromomeres (Eberle, 1956: 310). Unfortunately Eberle did not study other members of the tribe. He also investigated one species in tribe Trichosporeae, Aeschynanthus tricolor Hook.f., and found one chromosome with a terminal NOR among the n=16 chromosomes. Ehrlich (1958), investigating Saintpaulia, used pachytene data to illustrate chromosome mutations, such as inversions, translocations and deletions. He also reported a single NOR chromosome with a small satellite.

DIAGNOSTIC P O T E N T I A L O F C H R O M O S O M E D A T A
In Gesneriaceae the small size of the chromosomes frequently allows only the reporting of chromosome numbers for the taxa investigated. However, even the establishment of ploidy levels or approximate numbers can be of taxonomic importance, and this has been used to support taxonomic decisions. For instance Didymocarpus, in its traditional circumscription, included over 250 species, but Weber & Burtt (1998) remodelled the genus and split it into three genera, Didymocarpus s. str., Hovanella A. Weber & B.L. Burtt and Henckelia. In addition to morphological and biogeographical data they used differences in chromosome number to support their decision. Henckelia has rod-shaped chromosomes and a basic number of predominantly n=9; Didymocarpus s. str. has more globular chromosomes and a range of chromosome numbers, excluding n=9, while the Madagascan endemic Hovanella has n=14  or n=15 (Möller, unpublished).
Another example is the genus Streptocarpus. It is divided, mainly on the basis of vegetative morphology, into two subgenera, Streptocarpus Fritsch and Streptocarpella Fritsch. The former includes mainly unifoliate and rosulate taxa, while the latter comprises mainly caulescent species (Hilliard & Burtt, 1971 Mangenot & Mangenot, 1957; see footnote in section 5(c)) and a group of woody caulescent species from Madagascar with n=16. However, the chromosomes of the latter have a morphology different from those in subgenus Streptocarpus and can be distinguished easily (Jong & Möller, 2000).

DISTRIBUTION O F H A P L O I D C H R O M O S O M E N U M B E R S I N D I F F E R E N T T A X O N O M I C G R O U P S (a) Gesnerioideae
Within the tribes of Gesnerioideae the Gloxinieae are relatively diverse in haploid numbers (n=10, 11, 12, 13), but n=13 prevails (Table 2). Tribe Episcieae is very uniform: only two out of 21 genera have n=8, the rest n=9. The three genera in Gesnerieae all have n=14. The Beslerieae are seriously undersampled, only two out of more than 200 taxa having been counted, with n=16. Although the subfamily appears cytologically relatively uniform within tribes, it differs greatly between tribes.

(b) Coronanthereae
Plotting the frequencies of haploid numbers by subfamily reveals that the Gesnerioideae and Coronanthereae cluster in distinctly different groups (Fig. 2). While taxa in the Gesnerioideae have haploid numbers ranging from n=8 to n=16 (with a distinct peak at n=9), the Coronanthereae have numbers between n=37 and n=45. This might be an indication that the two groups belong to distinctly different phylogenetic lineages (but see sections 7 and 8).
Only five out of the nine genera in the Coronanthereae have been counted so far, but all have high haploid numbers. The exact somatic number of all taxa examined has not been exactly determined, although for Rhabdothamnus solandri Cunn. an exact meiotic count of n=37 has been recorded (Hair & Beuzenberg, 1960). FIG

(c) Cyrtandroideae
While the NW subfamily Gesnerioideae is characterized by a narrow range of low chromosome numbers, taxa in the OW subfamily Cyrtandroideae possess a wide range of haploid numbers from n=4 to n=64, with a peak at n=16 (Fig. 2). The majority have n=9, 15, 16 or 17. The Didymocarpeae are the most diverse with at least 11 different haploid numbers (Table 2). Apart from the n=20 for Titanotricheae, the Didymocarpeae includes all numbers recorded for the Cyrtandroideae. Thus, although very variable intragenerically, the tribe Epithemateae with n=9-27 fits well into the range of the rest of this subfamily.

POLYPLOIDY
The figures given in Table 2 and in the following discussion refer to haploid numbers and multiples thereof found within genera. Potential palaeopolyploidy, which might be assumed for species exhibiting n=x=17, is not taken into account, as this seems too speculative on the basis of available data.
The level of polyploidy inferred from the prevalent basic numbers shows an interesting trend congruent with the diversity of haploid numbers at subfamily level (Table 2). Polyploidy in the Gesnerioideae is mostly limited to tetraploidy and is greatest in tribe Episcieae with 11%. By contrast, the entire subfamily Coronantheroideae exhibits high haploid numbers, between 37 and 45. Speculation on basic numbers and ploidy levels is however hampered by two factors. First, most counts in this group are approximations, with exact numbers being unknown. Secondly, the affinities of this group with other NW groups are ambiguous (see section 8).
In the Cyrtandroideae, on the other hand, a high degree of polyploidy is apparent, with up to 15% of the counts for Trichosporeae being possibly polyploids. Levels of polyploidy may also be high, e.g. 10x in Henckelia, or 8x in Streptocarpus.
The highest degree of inferred polyploidy is 18% of counts observed for the Epithemateae.
The geographical distribution of polyploids is intriguing in Streptocarpus, where the only polyploids observed so far come from Madagascar and adjacent islands with none on the African mainland. Furthermore, polyploids are present in both subgenera, the predominantly caulescent Streptocarpella and the predominantly unifoliate or rosulate Streptocarpus, even though they evolved before the genus split into African and Madagascan lineages (Möller & Cronk, 2001). This shows that these polyploidizations are parallel evolutionary events. The situation in Streptocarpus is also remarkable with regard to the general trend of chromosomal stasis observed in taxa speciating on islands (Stuessy & Crawford, 1998) and might indicate another example of rapid speciation accompanied by cytological changes in island plants such as the Hawaiian silversword alliance, Asteraceae-Madiinae (Carr & Kyhos, 1986) or the genus Psychotria L. (Rubiaceae) (Kiehn & Lorence, 1996).

GENETIC V E R S U S C Y T O L O G I C A L D I V E R S I T Y
Chromosome evolution may involve structural rearrangements (see Streptocarpus above) and also dysploid changes in basic chromosome number. It is presumed to be a function of diversification time (Lagercrantz, 1998;Levin, 2002). It is also assumed that chloroplast DNA (cpDNA), being inherited uniparentally, evolves clock-like, steadily accumulating mutations over time, although there are exceptions (Clegg et al., 1994;Muse & Gaut, 1997;Mayer et al., 2003). By correlating divergence in cpDNA (here given as pairwise trnL-F intron/spacer sequence differences) with diversity in haploid numbers, it is possible to show whether chromosome evolution is likely to be linked to divergence time. In Gesneriaceae there seems to be close correlation at subfamily level ( Fig. 3). At tribal level, however, the data indicate that the extant taxa in the Didymocarpeae represent an assemblage more diverse in chromosome numbers, while those in the Epithemateae are more variable in cpDNA. This supports the view that the Epithemateae is the oldest tribe, representing an assemblage of relict taxa.
The Coronantheroideae, on the other hand, with their often monotypic genera and a distribution suggestive of a relict group, show, from six out of nine genera investigated, genetic divergence similar to other NW tribes, dismissing the suggestion that they are a palaeopolyploid assemblage of taxa. An explanation for their present day distribution needs further investigation.

GENOME E V O L U T I O N : C U R R E N T K N O W L E D G E
In recent years molecular data have been at the centre of numerous phylogenetic investigations. However, the continuing value of cytological data for taxonomic and evolutionary investigations cannot be emphasized enough. Knowledge of chromosome evolution (historic and recent) and inheritance in a plant group is directly relevant to a proper interpretation of any molecular analysis. After all, the nuclear genes from which sequences for phylogenetic research are often obtained reside on these chromosomes and frequently represent only a very small fraction of the entire genome (e.g. one out of 25,000 genes in Arabidopsis L.).
Chromosome evolution may involve single base pair changes or larger chromosomal rearrangements such as translocations, insertions, deletions, duplications or inversions, but may also involve the fusion or fission of whole chromosomes (Robertsonian events, leading to dysploidy), or even of whole chromosome complements (polyploidy). The latter can be divided into autopolyploidy (unreduced gametes resulting in genome duplication -tetraploidy) and allopolyploidy (hybridization followed by chromosome duplication -amphidiploidy). Mapping available chromosome data onto molecular phylogenies provides evidence that all the mechanisms indicated above are involved in the evolution of the Gesneriaceae genome. F IG . 3. Diversity of chromosome numbers plotted against maximum trnL-F intron/ spacer diversity, across tribes and subfamilies of Gesneriaceae. sf. Coro.=subfamily Coronantheroideae, sf.Cyrt.=subfamily Cyrtandroideae, sf.Gesn.=subfamily Gesnerioideae, tr.Bes.= tribe Beslerieae, tr.Cyrt.=tribe Cyrtandreae, tr.Didym.=tribe Didymocarpeae, tr.Ep.= tribe Episcieae, tr.Epith.=tribe Epithemateae, tr.Glox.=tribe Gloxinieae, tr.Trich.=tribe Trichosporeae. Hypotheses have been put forward suggesting that increasing or decreasing dysploid series can accompany plant evolution (for summaries see Sato, 1962;Levin, 2002). However, such assumptions are often based only on the existence of a range of basic numbers that can be arranged in, more or less complete, ascending or descending series. Examples in Gesneriaceae are given by Wang et al. (1998), , and . If such assumed series are not backed up by other independent characters (from morphology, phytogeography, molecular phylogeny or most importantly by comparisons of karyotype morphology and pairing relationships in hybrids), they do not necessarily reflect actual events or real evolutionary patterns. Exceptions include the single discordant counts as discussed for Henckelia by . Chromosome evolution is best discussed in the light of independent evolutionary hypotheses.
There is as yet no comprehensive, reliable molecular phylogeny available for the whole family Gesneriaceae. Recently a phylogeny of NW taxa has been published (Zimmer et al., 2002). Subsequent attempts to map available cytological data onto this phylogeny immediately revealed a recurring problem: many taxa that have been analysed cytologically were not included in the phylogeny and vice versa. Thus, the overlap between cytological and molecular data is surprisingly low: of the 57 ingroup taxa studied by Zimmer et al. (2002), 23 (40%) have no chromosome count. However, NW taxa of subfamily Gesnerioideae are relatively constant in basic chromosome number, and the phylogeny presented indicates that certain groups (e.g. Gesneriaeae with n=14) are very stable, others include a single dysploid reduction (e.g. n=9 to n=8 in Episcieae), while taxa with n=13 do not form a monophyletic clade because of a split between Gloxinieae and a Sinningia clade. It is interesting to note that n=11 has apparently evolved several times independently, always from n=13. Although the overall pattern of chromosome evolution in the NW taxa appears more complex than expected from the conserved chromosome complements within genera, there is no obvious case of an increasing or decreasing dysploid series. Zimmer et al. (2002) propose a basic number for subfamily Gesnerioideae of n=13 (excluding Beslerieae and Napeantheae). Although they excluded subfamily Coronantheroideae from their analyses, this may not have affected their results, even though it was earlier suggested (Smith et al., 1997) and recently proved (Wang et al., 2004) that subfamily Gesnerioideae is not monophyletic. Beslerieae and Napeantheae, both of which have been little investigated cytologically, appear basal to the rest of Gesnerioideae plus Coronanthereae. Based on the available phylogenetic and cytological evidence, the Coronanthereae may represent a distinct monophyletic lineage of highly polyploid species, as part of the NW subfamily Gesnerioideae. For constructive discussion on the basic number for all the NW taxa, it is clearly important to obtain a stable phylogeny, and to investigate comprehensively the cytology of the Beslerieae and Napeantheae.
The OW taxa, in contrast, have much more variation in basic chromosome number (Table 2). Mapping the available data onto a preliminary molecular phylogeny of OW taxa (Pfosser et al., unpublished) reveals interesting patterns, although the overlap of chromosomal and molecular data is not very extensive. This cpDNA-based phylogeny reflects shared chromosome number rather than the current tribal arrangement, with groups belonging to different tribes but possessing a common chromosome number clustering together. Not surprisingly, Chirita with its range of basic numbers appears polyphyletic across the tree. There are however too many gaps in the cytological data for a full discussion of the chromosome evolution of this phylogenetic hypothesis.
Nevertheless, useful insights into genome evolution can be gained when focusing on phylogenies of individual genera with a higher cytological coverage.  observed that the evolutionary relationships between Whytockia W.W. Sm. and Monophyllaea and within Monophyllaea suggested by morphological advancements were paralleled by an increasing dysploid series with an ancestral x=9 (unchanged in Whytockia), to x=10/11 or 12 in Monophyllaea. Mapping the available chromosome numbers for this relationship onto the molecular phylogeny of Mayer et al. (2003) gives an impression of increasing dysploidy, though not of a continuous series. Optimizing the available chromosome numbers onto this molecular tree confirms that the basic number for Epithemateae is x=9, and that there are four undisputed increases: to x=10 (Rhynchoglossum notonianum (Wall.) B.L. Burtt), to x=10/11 (Monophyllaea), to x=11 (Loxonia Jack) and to x=12 (Epithema membranacea (King) R. Kiew), with a further increase from x=11 to x=12 (Monophyllaea glauca C.B. Clarke), and a possible reduction from x=9 to x=8 in Epithema saxatile Blume, n=8 and 9 (Fig. 4). To demonstrate a continuous series of dysploid increase, however, would require molecular analysis of the polymorphic races of Monophyllaea.
On the other hand, in the large genus Cyrtandra a high chromosome number stability of 2n=34 is indicated, particularly since the available counts cover taxa representing the whole area of its distribution (Kiehn, unpublished). Differences may be present, however, at structural level even between closely related taxa (Storey, 1966;Kokubugata & Madulid, 2000). There are at present insufficient data to speculate on the age of the genus, but it may be relatively young and therefore uniform in its karyotype. The species richness of the genus may be explained by comparatively recent bird-dispersal of the fleshy fruits, in particular to islands in the eastern range of the genus where further speciation by isolation occurred.
In Streptocarpus, the basic chromosome number is highly conserved, having changed just once, coinciding with the origin of the subgenera (Fig. 5). Since then, speciation has occurred with no further change in basic chromosome number, although the ploidy level has increased several times independently in Madagascan taxa (see section 6). Aeschynanthus, by contrast, exhibits a very different chromosome evolution (Fig. 6). The resulting phylogeny has two main clades as in Streptocarpus. The ancestral number in Aeschynanthus appears to be n=16, and this has not been affected by the appearance of these clades. However, within the clades there is much variation in basic number, indicating that dysploid reductions in particular often coincided with, or even predated, speciation events, as seen in A. FIG. 4. cpDNA phylogeny of tribe Epithemateae from Mayer et al. (2003) with basic chromosome numbers optimized using MacClade 3.07 (Maddison & Maddison, 1997). For taxa lacking a square no cytological data are available.
radicans Jack with two cytotypes, n=15 and 16. This suggests that chromosome fusion occurs relatively easily, though this needs further testing. This chromosome behaviour is fundamentally different from that of Streptocarpus. Although chromosome number appears to be conserved in Streptocarpus, there are however still significant differences in chromosome structure and karyotype of taxa with identical chromosome numbers (Fig. 7).

'GESNERIACEAE WEBCYTE', A C Y T O L O G I C A L D A T A B A S E
Chromosome data on Gesneriaceae are scattered in the literature, and they have been compiled and discussed only twice in the past (Ratter, 1975;Skog, 1984). The indexes to plant chromosome numbers (last issue: Goldblatt & Johnson, 2000) merely summarize chromosome number reports, and normally refer only to literature covering the two years before their publication. Currently, there are 55 species of Gesneriaceae listed. No doubt there are also many unpublished cytological data on Gesneriaceae. The plan to create a platform holding all chromosome data on Gesneriaceae, with the possibility of contributing previously unpublished data, was supported by all participants in the Gesneriaceae workshop held at RBGE in 2002(Möller et al., 2002. The ability to access up-to-date knowledge on cytology quickly would increase effective use of existing data and encourage the closing of gaps in our knowledge. In order to fill this need, the cytological database 'Gesneriaceae WebCyte' (http://www.rbge.org.uk/rbge/web/search/index.jsp) has been developed at RBGE in collaboration with M.K. (second author of this paper) and Larry Skog of the Smithsonian Institution, USA (Möller et al., 2002 ongoing). It is regularly updated, includes all cytological data available for the Gesneriaceae and allows their instant display, including statistics and images of chromosome preparations if available. Thus all these data are now immediately accessible to those in the research community who may be interested in Gesneriaceae cytology (Fig. 8).

CONCLUSIONS A N D O U T L O O K
Although tremendous efforts have been made to accumulate chromosomal data in Gesneriaceae, with about 1000 published counts, there are still significant gaps in the coverage of the family, particularly in certain NW and OW tribes and small monotypic genera. The NW tribes Beslerieae and Napeantheae and the large, heterogeneous OW tribe Didymocarpeae are in particular need of investigation. There is also a frustrating lack of correspondence between the availability of molecular and of cytological data, a lack which should be rectified. Of particular interest would be a more detailed investigation of the genera in tribe Epithemateae with their variable FIG. 6. Simplified Aeschynanthus Jack ITS phylogeny based on Denduangboripant et al. (2001). For A. humilis Hemsl. no  basic numbers. Although investigated previously, species in subfamily Coronantheroideae need re-evaluation to establish their exact chromosome numbers. This should allow establishment of their basic number and, together with improved molecular data, provide better information on their phylogenetic position. At present, this group of species appears to be a highly polyploid lineage within subfamily Gesnerioideae.
Knowledge of chromosome morphology becomes more important for closely related taxa in cases where they possess the same basic chromosome number, especially in the light of molecular phylogenies. Therefore, besides the need to count chromosome numbers for additional species, more studies of chromosome structure and gene locations are desirable. Although some recent publications include images of sufficient quality to show morphological features, older publications relied on drawings which may or may not reflect reality. For comparative studies it seems prudent to produce new preparations. This is of particular importance in cases where different pretreatment or staining techniques can affect chromosome morphology (Zhou et al., 2004). Karyotyping was hampered in the past by the generally small size of Gesneriaceae chromosomes. This may now be easier with the introduction of new, highly sensitive 'chromosome painting' techniques such as: • Fluorescent in situ hybridization (FISH) for the localization of conserved multicopy loci. • Genomic in situ hybridization (GISH) for detection of hybrids and quantification of introgression. • Primed in situ hybridization (PRINS), for localization of single genes or determination of loci number of low-copy genes.
These techniques may even allow the acquisition of new data at nucleotide level. Living research collections are an important source of cytological material, but are often not fully exploited at present. Resources should be channelled into this area to recover and secure as much information as possible. Gesneriaceae are often short-lived and may not persist for more than a few months or years in cultivation. Also they should be analysed while physiologically in peak condition to maximize cytological success. This involves a battle for resources and time.
How can we tackle the task of closing existing gaps in our data and knowledge? Extended, targeted field work, preferably including the collection of living plants or seeds, and collaborative work will be the main tools to achieve maximum impact with limited institutional resources. This includes the establishment and maintenance of well-curated, decentralized ex situ living collections, crucial for cytological investigations, and the organization of regular workshops bringing together experts on the family and providing a platform for discussion, such as the workshop at RBGE in 2002(Möller et al., 2002. The new cytological database at RBGE, 'Gesneriaceae WebCyte', is intended to provide an optimized and continually updated overview of chromosome data in Gesneriaceae. It should serve as a depository for cytological data submitted by all researchers. On the basis of the data gathered there, collaborative project proposals to close existing gaps in our knowledge should be easier to draw up, thus stimulating a greater interest in the cytology of this exciting plant family.

A C K N O W L E D G E M E N T S
We would like to thank Kwiton Jong and Jimmy Ratter most warmly for helpful discussions and comments. The research by M.K. was carried out within the framework of Austrian Science Foundation (FWF) Research Proj. B-13107-Bio.