AOBPreview originally published online on November 16, 2004
Annals of Botany 2005 95(2):345-350; doi:10.1093/aob/mci031
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Annals of Botany 95/2 © Annals of Botany Company 2004; all rights reserved
Phenotypic Correlations Among Plant Parts in Iberian Papilionoideae (Fabaceae)
Departamento de Biología Vegetal y Ecología, Universidad de Sevilla, Apartado 1095, E-41080 Sevilla, Spain
* For correspondence. E-mail maliani{at}us.es
Received: 20 July 2004 Returned for revision: 6 September 2004 Accepted: 28 September 2004 Published electronically: 16 November 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
• Background and Aims Different plant organs may show varying degrees of form diversification or conservatism across phylogenetically related taxa. The present study uses data from a recent systematic study of Iberian Papilionoideae to investigate diversification and covariation in reproductive and vegetative plant parts. The appropriateness of imprecise (but comprehensive) taxonomic quantitative information is tested.
• Methods Organ size covariation and phenotypic correlations were studied among tribes, genera and species. Scale relationships were investigated by Reduced Major Axis regression. Variables used were the maximum dimensions of calyx, corolla, keel petal, fruit, seed, stipule, leaflet and petiole.
• Key Results As regards tribe averages, the length of the corolla and that of calyx correlated positively and significantly. In contrast, pod length was unrelated to corolla size and largely tribe-specific. Within genera, the sizes of calyx, corolla and fruit sometimes covaried linearly (e.g. Lathyrus species) and other times did not (Genista, Astragalus).
• Conclusions Information from taxonomic studies can be useful to establish major phenotypic correlations in plants. Results underscore the implications of tribal ownership in the Papilionoideae and illustrate the extensive morphological diversification of pods relative to flowers in this group.
Key words: Allometry, flowers, fruits, genera, isometry, Papilionoideae, petals, seeds, leaves, tribes
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Conservatism in the morphology of an organ within any plant lineage may result from taxon-specific developmental constraints (e.g. Niklas, 1997
, p. 46; Debat and David, 2001; Meiklejohn and Hartl, 2002), and/or functional restriction (Grant, 1949; Stebbins, 1974; Conner and Via, 1993; Conner and Sterling, 1995; Wolfe and Krstolic, 1999; see, however, Wilson, 1995; C.M. Herrera, 2001; Herrera, 2004). Furthermore, parts in which form has been intensely modified by diversifying forces will often vary independently from unaltered or relatively neutral ones, which can result in weak or non-existent phenotypic correlations between organs.
Several related concepts are covered by the term ‘morphology’ (architecture, size, geometry, shape; Niklas, 1994), but size is probably the simplest to use and quantify and also a major determinant of function in many instances. In the Genisteae (Papilionoideae), for example, the size of the banner petal varies independently from that of the floral pedicel, even though strong covariation exists between banner and keel (J. Herrera, 2001). It is likely that the diversifying forces that act on petals (as well as their functional and developmental constraints) are different from those that work on pedicels, and this has resulted in uncoupled size variability. By helping to identify sets of integrated characteristics, studying part covariation and phenotypic correlations in phylogenetically related taxa may increase our understanding of plant evolution. Nevertheless, such analyses are scarce both for animals (Riska, 1985; Lofsvold, 1986; Cheverud, 1989; Cowley and Atchley, 1990) and plants (Armbruster, 1988; Mazer and Hultgard, 1993; Armbruster et al., 1994; Torres, 2000).
Ideally, a convincing analysis of organ size covariation for any group of phylogenetically related plants should use accurate measurements for all of its species, although this may be impractical in large groups with many relatively rare taxa. The alternative of analysing in detail only a few widespread taxa bears a serious risk that the extant morphological diversity is insufficiently sampled. Resorting to part dimensions as indicated in systematic reviews may be a possible way to account for all or most species, although this information is usually imprecise and plagued by considerable error variance. The question thus arises of whether such low-power data can reveal any phenotypic correlations at all.
The present study attempts to establish major phenotypic correlations, and to differentiate diversified from conserved organs using data from a recent systematic account of Iberian Papilionoideae. Coupled variability (covariation) is investigated at different taxonomic levels (e.g. tribe, genus, species).
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Source of data
Data on dimensions were gathered from Flora Iberica, volume 7 (Talavera et al., 1999
, 2000), a comprehensive account of Iberian Papilionoideae that considers morphology, chromosome numbers and molecular studies (Talavera and Salgueiro, 1999). Even though the numbers of tribes and genera seem well established, the systematics of such a large and complex group is under permanent revision and is admittedly imperfect. It is therefore possible that further molecular research might indicate that some species be shifted from one genus to another (S. Talavera, pers. comm.). For the present study this represents an underlying source of error (i.e. the error variance will be larger than if all species were ‘correctly’ assigned to each genus) that can hinder detection of significant phenotypic correlations. Trait covariation will remain undetected in an unknown number cases and conclusions will most likely fall on the conservative side. Yet another source of error is uneven measurement accuracy among the 23 different systematists that participated in the work.
As is usual in systematic studies, the work of reference reported quantitative traits in the form of two numerals depicting the lower and upper size boundaries observed for each species within the considered geographical range (e.g. calyx 6–10 mm). The reliability of this kind of information was checked by confronting boundaries with data from a detailed morphometric study on Genisteae (Papilionoideae; J. Herrera, 2001) in which measurements had been gathered with a statistical concern (i.e. recording sample sizes, means, standard errors, etc). Size limits as given in the Flora correlated tightly with mean organ size in a subset of 14 species (Table 1). Correlations were near or above 0·9 in all cases (see Appendix for raw data) and, as a result, the coefficients of determination were so close to unity that if means had to be calculated from linear regression the estimation would be highly dependable. It can be concluded that, despite having a seemingly low resolution, boundary-like quantitative data can be used to test hypotheses regarding part size variation across taxa. Correlations with the mean were similarly tight for the upper and lower boundaries, but the upper limit has been used throughout the study because this was the only indication of organ size reported in a number of species. Furthermore, the use of the upper limit may help to purge the data set from abnormally small organs that sometimes appear under stressful conditions in plants.
|
Statistical analyses
A data matrix was built in which variables were the maximum observed dimension of the calyx, the corolla (which almost invariably meant length of the banner petal), keel petal, fruit, seed, petiole, leaflet and stipule. The criteria used to deal with organ heteromorphism were that if a species was reported to have both cleistogamous and chasmogamous flowers, only the latter's dimensions were included in the matrix; if leaves were heteromorphic, the dimensions of those appearing on fertile stems were chosen; if stipules were dimorphic, the length of the largest ones was used. Lastly, in species with compound leaves the length of the largest leaflet was considered. Dimensions were then transformed logarithmically and two coding variables (tribe and genus) added to the data set. The resulting matrix (available from the author upon request) had 499 cases (i.e. species) although, for a variety of reasons, data may often be missing. For example, leaf-related measurements were necessarily absent for many Papilionoideae which do not possess true leaves (e.g. many Genisteae). In a few cases the morphology of an organ was so peculiar (e.g. coiled pods in Medicago) that its dimensions, as reported in the systematic account, were not comparable to that of most other taxa and were discarded. But the main reason behind missing data was that systematists reviewing different genera often had different opinions regarding the usefulness of quantitative traits. Despite substantial editorial efforts to keep consistency across genera (S. Talavera, pers. Comm.), this often resulted into uneven data availability. Analyses below deal only with groups/traits in which there was enough quantitative information for a satisfactory statistical treatment.
The Papilionoideae share a common phylogeny, so species cannot be considered independent data points. Congeneric species, for example, are more likely to resemble each other morphologically than taxa in different genera or tribes (i.e. there is phylogenetic constraint). Therefore, combining all Papilionoideae in a single statistical sample would probably result in an inflated number of degrees of freedom. The problem was partially accounted for in the present study by performing analyses at several levels within the systematic hierarchy (tribe, genus and species). For tribe-based analysis the variates were organ dimensions averaged over all the species in each tribe (13 tribes were represented, but the five that had fewer than four species and one that lacked usable data were discarded). In genus-based correlations, variates were dimensions averaged over all species in each genus (albeit partially, quantitative data were available for 49 out of 53 genera covered in the reference work). Lastly, species-based analyses were restricted to eight genera which are particularly diverse and rich in quantitative data (Genista, Cytisus, Ulex, Astragalus, Vicia, Lathyrus, Ononis and Hyppocrepis). These account altogether for 245 species (i.e. 45 % of Iberian Papilionoideae; see Table 3 for the numbers of species in each genus).
|
Scale relationships among organs were investigated by Reduced Major Axis (RMA) regression, a method preferred over least-squares regression whenever both variables are, as here, subject to natural variation and measurement error (Niklas, 1994
). In a RMA regression the confidence ellipse that encircles the data points is computed and the slope of the ellipse's major axis amounts to the coefficient of regression (
). Numerically, this coefficient is simply the standard deviation of one variable divided by the standard deviation of the other, so that if the ratio (i.e. the major axis' slope) is 1 there exists isometry (meaning that organ size variations are proportional and one part does not accelerate or decelerate relative to the other). The shape of the confidence ellipse around data points is also visually informative: if it is very narrow, most of the variance is being accounted for by the major axis (i.e. the variables are tightly correlated), but if it approaches a circle the variance about the major axis is relatively scant (and the correlation is weak). |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Tribe-based patterns
Averages for the maximum dimension of the corolla, calyx and fruit are plotted in Fig. 1. Across tribes, corolla length and calyx length correlated significantly (r = 0·882, Bonferroni adjusted P = 0·02; n = 7 tribes including 36 genera). In contrast, fruit and corolla varied independently (r = 0·37, ns). Although the tribe Ononideae might be viewed as an outlier (on average they have fruits unusually short relative to the calyx; Fig. 1), excluding this tribe from the analysis did not alter the general pattern of lack of correlation between fruit and corolla.
|
The major axis of the confidence ellipse in Fig. 1 is perceptibly parallel to the diagonal and, accordingly, the RMA regression of calyx length on corolla length yielded a scaling exponent or ‘slope’ not significantly different from unity (
RMA = 1·2, with lower and upper confidence limits L1 = 0·55 and L2 = 1·85). The scale relationship was thus isometric, implying that flowers (but not fruits) were geometrically similar among tribes.
Genus-based patterns
Correlations among eight quantitative traits (averaged over species within genera) are shown in Table 2. Reproductive parts such as petals, fruit and seed were often involved in significant correlations, whereas traits relating to leaves rarely did. Moreover, the two subsets of traits (reproductive or vegetative) appeared largely unconnected to each other. The only exception in this regard was the significant correlation between the length of the seeds and that of stipules.
|
Most tribes contributed few (1–8; mean 4) genera to the sample, although the Genisteae were relatively overrepresented with 16. Reanalysing the data after withdrawal of the Genisteae only caused the weak correlation among fruit and seed length to vanish, whereas the overall pattern (i.e. strong reproductive correlations, uncorrelated vegetative parts) remained largely unchanged.
Species-based patterns
Table 3 summarizes results of Reduced Major Axis regressions of calyx and fruit dimensions over corolla length for eight species-rich genera. Invariably, calyx and corolla correlated positively and significantly in length. Furthermore, the scaling exponent (i.e. slope) for the regressions never departed significantly from unity, pointing to isometry among species for these organs. On the other hand, whether corolla and pod size were correlated or not depended on the genus. Provided that a significant correlation among these organs existed, the scaling exponents could be indistinguishable from unity (as in Vicia or Ulex); or significantly larger than 1 (i.e. much variation in fruit size was not related to corolla size, as in Ononis and Genista).
Genus-specific scale relationships that link calyx, corolla and fruit size can be assessed graphically in Fig. 2. In Lathyrus, fruit and calyx covaried linearly with the corolla (i.e. their lengths might change across species, but organs were just scaled-up versions of the same model). In contrast, in the genus Genista large-flowered species had disproportionately long fruits. Lastly, and for different reasons, corolla and pod lengths were unrelated in Astragalus and Hippocrepis. In the latter, fruit size had so little variance that it could actually be considered a constant. In the former, however, there was extensive fruit size variation among species, but one which was totally uncoupled from floral variations.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Although means might instinctively seem to be the best statistic to establish whether organs correlate in size across a large plant group, upper size limits may also serve provided that they correlate tightly with the mean. This condition was met in the present study, and should apply whenever dimensions distribute normally and continuously within the range of a species. Those taxa that include subspecies are exceptional in this regard, since their quantitative traits are more likely to distribute in multimodal, non-normal ways. Relatively few Iberian Papilionoideae (8 %) include subspecies, but these nonetheless represent a source of error variance that should be added to those mentioned in the Introduction (i.e. incorrect assignment of species to genera and varying accuracies of systematists).
Plant life-form in Iberian Papilionoideae spans various categories, from small annuals (e.g. many Trifolium), to large shrubs (e.g. Cytisus). As a result, plant sizes range over several orders of magnitude and a number of ‘part–whole correlations’ among organs could be expected ensuing from a common dependence on plant size (Sokal and Rohlf, 1981; Jackson and Somers, 1991). In other words, genera in which individual plants are large will often have relatively large organs (flowers, fruits, leaves, stipules, etc.). This would explain the otherwise unexpected coupled variation of stipule and seed lengths detected in the generic comparisons, and also the one that existed among petioles and leaflets.
Regardless of whether species, tribes or genera were considered, a simple (isometric) scale relationship linked the largest dimensions of calyx and corolla. Extensive variation regarding organ shape can be assumed for the studied group as a whole, but the basic geometry of flowers was remarkably constant, probably due to a general morphogenetic constraint operating on all Papilionoid flowers (for a differentiation between shape and geometry see Niklas, 1994, p. 11). Nevertheless, the calyx-to-corolla ratio, which is remarkably steady in the Papilionoideae, represents only an average tendency: exceptions to the norm exist and can be illustrated by the Genisteae, in which calyx and corolla do not correlate at all in size (e.g. the calyx may be long relative to the corolla, as in Ulex or Stauracanthus; or rather short, as in Cytisus; J. Herrera, 2001). Within the Papilionoideae, deviant groups such as the Genisteae indicate that even primary floral morphogenetic restrictions can sometimes be disrupted by evolution.
The second most frequent phenotypic correlation detected was between corollas and fruits, which correlated significantly in size whenever congeneric species or genera were considered. This is hardly surprising since, at least among closely related taxa, those with larger flowers are likely to also have longer ovaries, which in turn will develop into longer fruits. Exceptional in this regard was Astragalus (Fig. 2), a species-rich genus in which speciation has been suggested to relate to ecological factors or biochemical features, rather than to morphological innovation (Sanderson and Wojciechowski, 1996).
In contrast to genera and species, pod and corolla dimensions were uncorrelated when tribe averages were considered. The size relationship for these organs was actually tribe-specific and indicative of significant past evolutionary modifications of fruits that long ago disrupted the primary pattern (i.e. corolla–fruit covariation).
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
I thank M. Arista, P. E. Gibbs, A. M. Sánchez-Lafuente, S. Talavera and P. Wilson for helpful comments on the manuscript. Special acknowledgement is made to Professor Salvador Talavera, who shared valuable expertise on the biology and systematics of Papilionoideae. This study was funded by Plan Andaluz de Investigación (Junta de Andalucía) and by grant BOS2000-0328 of the Spanish Dirección General de Enseñanza Superior e Investigación Científica.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
-
Armbruster WS. 1988. Multilevel comparative analysis of the morphology, function, and evolution of Dalechampia blossoms. Ecology 69: 1746–1761.[CrossRef]
Armbruster WS, Edwards ME, Debevec EM. 1994. Floral character displacement generates assemblage structure of western Australian Triggerplants (Stylidium). Ecology 75: 315–329.[CrossRef][ISI]
Cheverud JM. 1989. A comparative analysis of morphological variation patterns in the papionins. Evolution 43: 1737–1747.[CrossRef]
Conner J, Via S. 1993. Patterns of phenotypic and genetic correlations among morphological and life-history traits in Wild Radish, Raphanus raphanistrum. Evolution 47: 704–711.[CrossRef][ISI]
Conner JK, Sterling A. 1995. Testing hypotheses of functional relationships: a comparative survey of correlation patterns among floral traits in five insect-pollinated plants. American Journal of Botany 82: 1399–1406.[CrossRef]
Cowley DE, Atchley WR. 1990. Development and quantitative genetics of correlation structure among body parts of Drosophila melanogaster. American Naturalist 135: 242–268.[CrossRef]
Debat V, David P. 2001. Mapping phenotypes: canalization, plasticity and developmental stability. Trends in Ecology and Evolution 16: 555–561.[CrossRef]
Grant V. 1949. Pollination systems as isolating mechanisms in angiosperms. Evolution 3: 82–97.
Herrera CM. 2001. Deconstructing a floral phenotype: do pollinators select for corolla integration in Lavandula latifolia. Journal of Evolutionary Biology 14: 574–584.[CrossRef]
Herrera J. 2001. The variability of organs differentially involved in pollination, and correlations of traits in Genisteae (Leguminosae: Papilionoideae). Annals of Botany 88: 1027–1037.
Herrera J. 2004. Lifetime fecundity and floral variation in Tuberaria guttata (Cistaceae), a Mediterranean annual. Plant Ecology 172: 219–225.[CrossRef]
Jackson DA, Somers KM. 1991. The spectre of spurious correlations. Oecologia 86: 147–151.[CrossRef]
Lofsvold D. 1986. Quantitative genetics of morphological differentiation in Peromyscus. I. Tests of the homogeneity of genetic covariance structure among species and subspecies. Evolution 40: 559–573.[CrossRef]
Mazer SJ, Hultgard UM. 1993. Variation and covariation among floral traits within and among four species of northern European Primula (Primulaceae). American Journal of Botany 80: 474–485.[CrossRef]
Meiklejohn CD, Hartl DL. 2002. A single mode of canalization. Trends in Ecology and Evolution 17: 468–473.[CrossRef]
Niklas KJ. 1994. Plant allometry. Chicago: University of Chicago Press.
Niklas KJ. 1997. The evolutionary biology of plants. Chicago: University of Chicago Press.
Riska B. 1985. Group size factors and geographic variation of morphometric correlations. Evolution 39: 792–803.[CrossRef]
Sanderson MJ, Wojciechowski MF. 1996. Diversification rates in a temperate legume clade: are there ‘so many species’ of Astragalus (Fabaceae). American Journal of Botany 83: 1488–1502[CrossRef]
Sokal RR, Rohlf FJ. 1981. Biometry. 2nd Ed. San Francisco: Freeman.
Stebbins GL. 1974. Flowering plants: evolution above the species level. Cambridge, MA: Belknap.
Talavera S, Aedo C, Castroviejo S, Romero Zarco C, Sáez L, Salgueiro FJ, Velayos M. 1999. Flora Iberica. Vol VII (I). Leguminosae (partim). Madrid: Real Jardín Botánico, CSIC.
Talavera S, Aedo C, Castroviejo S, Romero Zarco C, Sáez L, Salgueiro FJ, Velayos M. 2000. Flora Iberica. Vol VII (II). Leguminosae (partim). Madrid: Real Jardín Botánico, CSIC.
Talavera S, Salgueiro FJ. 1999. Sobre el tratamiento de la familia Leguminosae en Flora Iberica. Lagascalia 21: 155–222.
Torres C. 2000. Pollen size evolution: correlation between pollen volume and pistil length in Asteraceae. Sexual Plant Reproduction 12: 365–370.[CrossRef]
Wilson P. 1995. Selection for pollination success and the mechanical fit of Impatiens flowers around bumblebee bodies. Biological Journal of the Linnean Society 55: 355–383.[CrossRef]
Wolfe LM, Krstolic JL. 1999. Floral symmetry and its influence on variance in flower size. American Naturalist 154: 484–488.[CrossRef][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||