Genetic Analysis of Natural Variations in the Architecture of Arabidopsis thaliana Vegetative Leaves

Corresponding author: José Luis Micol, Campus de Elche, Edificio Vinalopó, Avenida del Ferrocarril s/n, 03202 Elche, Alicante, Spain., jlmicol{at}umh.es (E-mail)

Communicating editor: V. SUNDARESAN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

To ascertain whether intraspecific variability might be a source of information as regards the genetic controls underlying plant leaf morphogenesis, we analyzed variations in the architecture of vegetative leaves in a large sample of Arabidopsis thaliana natural races. A total of 188 accessions from the Arabidopsis Information Service collection were grown and qualitatively classified into 14 phenotypic classes, which were defined according to petiole length, marginal configuration, and overall lamina shape. Accessions displaying extreme and opposite variations in the above-mentioned leaf architectural traits were crossed and their F₂ progeny was found to be not classifiable into discrete phenotypic classes. Furthermore, the leaf trait-based classification was not correlated with estimates on the genetic distances between the accessions being crossed, calculated after determining variations in repeat number at 22 microsatellite loci. Since these results suggested that intraspecific variability in A. thaliana leaf morphology arises from an accumulation of mutations at quantitative trait loci (QTL), we studied a mapping population of recombinant inbred lines (RILs) derived from a Landsberg erecta-0 x Columbia-4 cross. A total of 100 RILs were grown and the third and seventh leaves of 15 individuals from each RIL were collected and morphometrically analyzed. We identified a total of 16 and 13 QTL harboring naturally occurring alleles that contribute to natural variations in the architecture of juvenile and adult leaves, respectively. Our QTL mapping results confirmed the multifactorial nature of the observed natural variations in leaf architecture.

MANY questions remain unanswered in the study of how the overall pattern of plant leaves is built, both at the level of leaf initiation and morphogenesis. Over the past decade, however, significant insights into several of the mechanisms operating in leaf ontogeny have been gained by studies of different plant species (reviewed in HAKE and SINHA 1991 ; SMITH and HAKE 1992 ; SINHA et al. 1993 ; TELFER and POETHIG 1994 ; TSUKAYA 1995 ; HALL and LANGDALE 1996 ; SYLVESTER et al. 1996 ; POETHIG 1997 ; BRUTNELL and LANGDALE 1998 ; TSIANTIS and LANGDALE 1998 ; VAN LIJSEBETTENS and CLARKE 1998 ; SCANLON 2000 ; BYRNE et al. 2001 ). Most of the published studies illuminating this issue have approached the causal analysis of leaf development by means of isolating mutants, whose characterization has allowed the identification of genes involved in leaf ontogeny. Nevertheless, variants in the architecture of leaves can be found not only by experimentally inducing mutations but also in the diversification of the species of interest in natural races. Both of these approaches might provide a panoramic view of the range of variation of the system under study and also specific variants to be studied to identify the genes controlling plant organogenesis.

The study of natural variation has proved useful for analyzing the genetic basis of some developmental processes in the model system Arabidopsis thaliana. Important contributions to their genetic dissection have been made by analyzing the progeny of intercrosses involving accessions (also named ecotypes) that differ in specific traits. Such an approach has allowed the identification of single genes controlling flowering time such as FLOWERING ALTERED (FLA; LEE et al. 1993 ), also named FRIGIDA (FRI; CLARKE and DEAN 1994 ; SANDA et al. 1997 ; JOHANSON et al. 2000 ); AERIAL ROSETTE (ART); and ENHANCER ROSETTE (EAR; GRBIC and BLEECKER 1996 ). In addition, quantitative trait loci (QTL) analysis has been shown to be useful in the identification of novel genes involved in some developmental processes such as those of EARLY DAY-LENGTH INSENSITIVE (EDI), FLOWERING F (FLF), FLG, and FLH (ALONSO-BLANCO et al. 1998 ), which affect flowering time; the ROSETTE LEAF NUMBER (RLN1–RLN5) loci (CLARKE et al. 1995 ), which affect vernalization responsiveness; the REDUCED TRICHOME NUMBER gene (RTN; LARKIN et al. 1996 ), affecting the number of trichomes on vegetative leaves; four QTL affecting seed oligosaccharide storability [BENTSINK et al. 2000; those of ESPRESSO (ESP), ANDANTE (AND), NON TROPO (NOT), and RALENTANDO (RAN)]; QTL affecting circadian rhythm (SWARUP et al. 1999 ); at least 11 loci affecting seed size (ALONSO-BLANCO et al. 1999 ); 11 QTL associated with several floral traits (JUENGER et al. 2000 ); 110 QTL for inflorescence developmental traits (UNGERER et al. 2002 ); and 12 QTL affecting light and hormone responses (BOREVITZ et al. 2002 ) among others.

Advances in molecular technologies, together with the information provided by genome projects, have made possible the Mendelization of QTL intervals, which are instrumental in positionally cloning the underlying genes. Recent examples are the fw2.2 gene of tomato (FRARY et al. 2000 ) and the EDI locus of A. thaliana, a QTL that turned out to be a novel allele of the CRYPTOCHROME2 (CRY2) gene (EL-DIN EL-ASSAL et al. 2001 ).

With a view to identifying genes involved in leaf morphogenesis, we have screened for new mutants with abnormal leaves (BERNA et al. 1999 ; ROBLES and MICOL 2001 ; PEREZ-PEREZ et al. 2002 ) and conducted genetic analyses of already obtained mutants (SERRANO-CARTAGENA et al. 1999 , SERRANO-CARTAGENA et al. 2000 ). The work presented here consists of a study of variations in the architecture of vegetative leaves in a large sample of A. thaliana accessions, with the purpose of ascertaining whether natural variability can act as a source of information for studying the genetic controls underlying leaf morphogenesis. We also performed a detailed morphometric analysis of leaf growth in one such accession, Enkheim-2 (En-2), which represents the genetic background for most of the mutant lines belonging to the large Arabidopsis Information Service (AIS) form mutants collection. Variations in the repeat number at 22 microsatellite loci were determined to measure genetic distances between the accessions under study, with the purpose of assessing correlation between their phylogeny and leaf morphology. We finally identified QTL contributing to natural variations in the architecture of the vegetative leaves of A. thaliana by analyzing a mapping population of recombinant inbred lines (RILs) obtained from a Landsberg erecta-0 (Ler-0) x Columbia-4 (Col-4) cross.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Plant materials, growth conditions, and crosses:
Seeds of A. thaliana (L.) Heyhn. accessions and RILs (LISTER and DEAN 1993 ) derived from a cross between Landsberg erecta-0 (Ler; NW20) and Columbia-4 (Col-4; N933) were supplied by the Nottingham Arabidopsis Stock Centre (NASC). The accessions used in this work are listed in Table 1, where they are noted by using the Arabidopsis Information Service and NASC (in parentheses) nomenclatures (ROBBELEN 1965 ; KRANZ 1978 ; KIRCHHEIM and KRANZ 1981 ; KRANZ and KIRCHHEIM 1987 ). The list of studied RILs included the following: N1901, N1903–N1971, N1973–N1999, N4664, and N4686.

Sterile (in 150-mm petri dishes containing 100 ml solid culture medium) and nonsterile (in pots containing a 2:2:1 mixture of perlite, vermiculite, and sphagnum moss) cultures were performed at 20° ± 1°, 60–70% relative humidity, and continuous illumination of 7000 lux as described in PONCE et al. 1998 . Crosses were performed with forceps under a binocular microscope, by transferring pollen from mature anthers to the stigmata of previously emasculated flowers (KRANZ and KIRCHHEIM 1987 ).

Photography and morphometric analysis:
Photographs of accession rosettes were taken with a Nikon F-601 AF camera equipped with an AF Micro Nikkor 105-mm f/2.8 macro lens 30 days after sowing. For the morphometric analysis of En-2 leaves, all the leaves from 10 plants were excised with forceps every 2 days, from the 10th to the 32nd day after sowing, immediately submerged in water to prevent dehydration, and mounted on slides before being photographed in a Leica MZ6 microscope. Morphometric analysis of these photographs was performed with the Sigmascan 2.0 program (Statistical Products & Service Solutions, Chicago).

For the morphometric analysis of RILs, seeds were sown on petri dishes. A total of 40 plants, corresponding to two RILs, were grown per petri dish. Two sowings were made per RIL. Leaves from the third and the seventh node from 15 plants of each RIL, chosen at random, were excised with forceps 25 days after sowing, immediately placed on the surface of agar medium to prevent dehydration, and covered by a transparent film. Photographs were taken with a Sony Cybershot FV-505 digital camera using a resolution of 1024 x 768 pixels. Images were digitally processed with the Adobe Photoshop 5.0 program (Adobe Systems, San Jose, CA) and morphometrically analyzed with the NIH Image program (developed at the U.S. National Institutes of Health and available at http://rsb.info.nih.gov/nih-image/).

Detection of microsatellite variation:
DNA isolation and PCR amplifications were performed following the high-throughput method described in PONCE et al. 1999 , which is based on multiplex PCR amplification of microsatellites, followed by fluorescent semiautomated detection of the amplification products in a Perkin-Elmer (Norwalk, CT) ABI PRISM 377 DNA sequencer. Genomic DNA samples of each accession were used as templates in five parallel multiplex PCR mixtures, each of which included three to six primer pairs. Each primer pair incorporates one oligonucleotide labeled with a fluorochrome [4, 7, 2', 4', 5', 7'-hexachloro-6-carboxyfluorescein (HEX), 6-carboxyfluorescein (6-FAM), or 4, 7, 2', 7'-tetrachloro-6-carboxyfluorescein (TET) phosphoramidites]. The microsatellites coamplified in each reaction mixture and the fluorochrome used to identify the corresponding amplification product were as follows: nga1126 (TET), nga162 (6-FAM), AthGAPAab (6-FAM), nga1111 (HEX), and MBK5 (HEX) in mixture 1; nga168 (6-FAM), nga12 (HEX), and AthCTR1 (TET) in mixture 2; AthZFPG (6-FAM), T27k12-Sp6 (6-FAM), AthGENEA (TET), nga1145 (HEX), AM4 (TET), and nga1139 (HEX) in mixture 3; AthACS (TET), nga111 (6-FAM), AthCHIB (HEX), nga6 (HEX), and nga139 (TET) in mixture 4; nga361 (TET), nga1107 (TET), and AthPHYC (6-FAM) in mixture 5. The sequences of the oligonucleotides used were as described in PONCE et al. 1999 .

Microsatellite lengths were determined using the GENESCAN 2.1 fragment analysis software (Applied Biosystems, Foster City, CA). The number of repeats of each allele was estimated by comparing the size of its amplification product with that of the Col-0 accession, which was determined by BELL and ECKER 1994 .

Gene diversity (or expected heterozygosity in a random mating population, H) was estimated following NEI 1973 and INNAN et al. 1997 : n(1 - p²)/(n - 1), where n is the number of samples and p is the frequency of an allele. The MICROSAT 1.5d program (E. MINCH, unpublished data; available at http://hpgl.stanford.edu/projects/microsat/) was used to obtain genetic distance measurements. This program allows the following parameters to be calculated for microsatellite data: the (µ)² genetic distance (GOLDSTEIN et al. 1995A ), the average square distance (D₁; GOLDSTEIN et al. 1995B ), the kinship coefficient (K_F; CAVALLI-SFORZA and BODMER 1971 ), the proportion of shared alleles (P_S; BOWCOCK et al. 1994 ), the fuzzy set similarity measure (F_S; DUBOIS and PRADE 1980 ), and the absolute difference parameter (D_AD; GOLDSTEIN et al. 1996 ). Distance matrices were obtained with the MICROSAT program and bootstrap resampling (n = 1000) was used to construct multiple phylogenetic trees with the NEIGHBOR program, which were derived into consensus trees with the CONSENSE program, both included in the PHYLIP 3.5c package (FELSENSTEIN 1993 ). Trees were plotted with the TreeView program (PAGE 1996 ).

Measurement of quantitative variation and statistical analysis:
Leaf morphology was studied in the third node (N3) and seventh node (N7) vegetative leaves 25 days after sowing. Measurements were taken for lamina area (LA), lamina perimeter (LP), lamina major chord or length (LL), lamina minor chord or width (LW), petiole length (PL), and petiole width (PW). In addition, the number of marginal serrations (NMS) was scored only in seventh leaves. Other measurements made 25 days after sowing were the total leaf number (TLN) and major (MRD) and minor (mrd) rosette diameters.

Data statistical analyses were performed by the SPSS version 7.5 software package (Statistical Products & Service Solutions) as described below. The normality of the studied traits was assayed by both chi-square (SNEDECOR and COCHRAN 1989 ) and Kolmogorov-Smirnov analyses (CHAKRAVARTI et al. 1967 ). For each trait, total phenotypic variance (V_P) was partitioned into sources attributable to genotype and error, using a random effects analysis of variance (ANOVA) as previously described (UNGERER et al. 2002 ). Components of variance were used to estimate the broad-sense heritability (H²) of the studied traits as V_G/(V_G + V_E), where V_G is the among-RIL variance and V_E is the residual (error) variance. The coefficient of genetic variation (CV_G) was calculated as (100 x )/, where is the mean value for the studied trait in the RIL population. Genetic correlations (r_G) among the studied traits in the RIL population were estimated as cov(x, y)/_xy (ROBERTSON 1959 ), and proximity matrices were constructed by using the SPSS version 7.5 software package.

QTL mapping:
We used a mapping population of 100 RILs, derived from a Ler-0 x Col-4 cross, whose residual heterozygosity is 0.42% (LISTER and DEAN 1993 ). Quantitative genetic analyses were conducted on the morphometric data obtained as described above from the pictures taken from excised leaves.

We constructed a linkage map using 173 molecular markers (42, 24, 29, 38, and 40 markers, respectively, for chromosomes 1, 2, 3, 4, and 5), all of which had already been genotyped by previous authors in at least 90 of the RILs studied here (information available at http://nasc.nott.ac.uk/RI_data/full_markers.text). These markers covered 519.5 cM, >85% of the Arabidopsis genome, and were spaced at intervals ranging from 0.5 to 8 cM, their average distance being 3 cM.

QTL analyses were performed by using the MAPQTL version 4.0 program (VAN OOIJEN and MALIEPAARD 1996 ). Marker interval candidates to possibly contain a QTL were identified by means of the consecutive use of interval mapping (IM; LANDER and BOTSTEIN 1989 ) and multiple-QTL model mapping (MQM; JANSEN 1993 ; JANSEN and STAM 1994 ) methods. The IM method was used first, and different combinations of the markers linked to the identified QTL were then tested as cofactors for MQM mapping. QTL intervals were gradually refined in this way, to finally select as cofactors those markers that maximized the variance explained by each QTL. Logarithm-of-odds (LOD) scores obtained in this way were used to construct QTL likelihood plots (LANDER and BOTSTEIN 1989 ), using the SigmaPlot 2000 version 6.0 program (Statistical Products & Service Solutions). A significance threshold of 2.7 was chosen for QTL identification after determination, separately performed for each trait, of experiment-wide significance thresholds by permutation analysis (CHURCHILL and DOERGE 1994 ; DOERGE and CHURCHILL 1996 ; UNGERER et al. 2002 ). One thousand permutations were performed for each trait, and the LOD score thresholds obtained ranged from 2.7 to 2.8 (P < 0.05) and from 3.3 to 3.5 (P < 0.01). For each identified QTL, 2-LOD support intervals were established as a 95% confidence level (VAN OOIJEN 1992 ) after fixing as cofactors the closest marker to all the detected QTL, with the exception of that cofactor linked to the QTL under study. Finally, the additive effect and the percentage of variance explained by each QTL, as well as the total variance explained by all of the QTL contributing to a single trait, were also obtained from the MapQTL 4.0 program using MQM mapping.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Phenotypic classification of accessions of the AIS collection:
To study leaf natural variations in a wide and representative sample of A. thaliana plants, wild-type lines were chosen from the large AIS collection. Such natural races of different geographical origins are named ecotypes or accessions (according to ALONSO-BLANCO and KOORNNEEF 2000 ). Seeds from 193 AIS accessions were obtained from the NASC (Table 1) and grown in a controlled environment under sterile culture conditions (see MATERIALS AND METHODS). A total of 24 seeds, corresponding to four accessions, were sown per 150-mm petri dish. The plants were photographed every 3 days, from the sixth day after sowing. Five accessions (Dijon-M, Enkheim-T, Hojda-Obi-Garm, Cen-0, and Chi-0) had to be discarded because of their inability to germinate, despite repeated attempts, in our working conditions. The remaining 188 accessions were qualitatively classified into 14 phenotypic groups, attending to the morphology of their rosette leaves. We took into account the overall shape of the lamina (spatulate, rounded, or lanceolate), its deviations from flatness or planarity (undulate, revolute, or involute leaves), and its marginal configuration (smooth or serrated margin), as well as the compactness of the basal rosette (either bushy or loose, which is mostly dependent upon the length of the petiole) as observed 30 days after sowing (Table 1).

Leaf initiation rate in accessions of the AIS collection:
In A. thaliana, as in many other plant species, leaves produced at different stages of development show morphological differences, which are known as heteroblasty (reviewed in LAWSON and POETHIG 1995 ; TSUKAYA et al. 2000 ). This is illustrated in Fig 1, which represents variations with time in the shape and size of a given leaf during its expansion, as well as differences between different leaves within a plant (taking as examples the second and the fifth rosette leaves of the En-2 accession). Early leaves are small with rounded laminae, whereas later leaves are larger, have lanceolate laminae, and expand more rapidly than the earlier ones.

View larger version (10K): In this window In a new window Download PPT slide   Figure 1. Variation with time of the size and shape of Enkheim-2 (En-2) vegetative leaves. (A) Second rosette leaf. (B) Fifth rosette leaf. Bar, 5 mm.

The above-mentioned differences between leaves in En-2 might be greater for other accessions, especially those displaying an increased number of leaves, which, in turn, would affect the whole aspect of the rosette. Hence, we thought it useful to determine the rates of leaf formation in the accessions under study. For that purpose, plants were photographed 6, 9, 12, 15, and 18 days after sowing, and the numbers of visible aerial organs (cotyledons, leaf primordia, and rosette leaves) were scored. The results are shown in Table 2, in which the growth rates reflect the moment when leaves and leaf primordia become visible under a dissecting microscope and not when they are produced by the shoot meristem.

As Table 2 shows, most accessions displayed the same vegetative developmental rates when cultured under the same conditions. A few exceptions were Bla-1, Bla-2, and Bla-3, which showed a delayed growth that merely seems to be a consequence of the belated appearance of their first pair of leaves. In other accessions with delayed growth, such as Condara and S96, all the leaves were produced at a lower rate. Only in 4 of the accessions studied (Do-0, Hl-0, Kä-0, and Kn-0) did vegetative leaves appear faster than in the remaining 184, with a difference with the remaining accessions of one leaf increase every 6 days. For these few accessions that showed atypical behavior, we tried to determine whether any correlation existed between their developmental profiles and the scarce information available on the environmental conditions of their habitat of origin (ANDERSON 1993 ); no obvious relationship was found.

Morphometric analysis of the expansion of En-2 vegetative leaves:
One accession, En-2, was chosen to make quantitative the above-mentioned qualitative observations. The En-2 accession is of particular interest since it represents the genetic background of >100 mutant lines displaying altered leaf morphology, isolated by either G. Röbbelen or A. R. Kranz, which make up the so-called AIS form mutants collection (BURGER 1971 ; KRANZ 1978 ). A genetic analysis of these mutants has been published (SERRANO-CARTAGENA et al. 1999 ).

We morphometrically analyzed the variation with time of the shape and size of En-2 vegetative leaves, which were collected, photographed, and studied to obtain the results shown in Fig 2. In our morphometric analysis of the wild-type leaf, we chose the parameters of length, width, area, and perimeter of leaves (Table 3). We found a significant correlation between the leaf order and these parameters: At full expansion the first two leaves are smaller than the later ones. A similar correlation was observed between the length/width ratio and time, since the first two leaves are rounded, while later ones are elongated. Both the growth rate and the final length of the first pair of leaves were smaller than those of the second pair, and those of the latter, in turn, were smaller than those of the following ones. As regards the variation with time of leaf shape and size, lamina growth was seen to be much faster in the earlier stages of leaf expansion in all the studied leaves.

View larger version (31K): In this window In a new window Download PPT slide   Figure 2. Morphometric analysis of the expansion of En-2 vegetative leaves. Variations with time are represented for (A) area, (B) length, (C) width, (D) length/width ratio, and (E) perimeter of all rosette leaves.

Study of the inheritance patterns of leaf form variants in accessions of the AIS collection:
Although it has traditionally been assumed that traits displaying continuous variation are polygenic, several individual genes with large effects responsible for apparently quantitative variations have been identified thanks to the study of A. thaliana accessions (see the Introduction). To determine whether the morphological traits under study were monogenic or polygenic, crosses were performed involving pairs of accessions, each of them displaying a given trait in one of two extreme and opposite ways. Transmission of the following three traits was studied: petiole length (long or short), leaf marginal configuration (entire or serrated), and overall leaf shape (lanceolate or rounded). Fig 3 includes photographs of the studied accessions, some of which showed in an extreme manner more than one of the above-mentioned traits.

View larger version (154K): In this window In a new window Download PPT slide   Figure 3. Representative individuals of accessions belonging to the phenotypic classes chosen to study the inheritance patterns of natural variations in leaf morphology: (A) Ga-0, (B) Kä-0, and (C) Be-1, with serrated leaf margin; (D) Aa-0, (E) Cvi-0, and (F) Bd-0, with entire leaf margin; (G) La-1 and (H) Bla-14, with rounded leaves; (H) Bla-14, (I) Fi-0, and (J) Hl-2, with short petiole; (K) Bla-1, (L) Gr-3, and (M) Li-5-3, with long petiole; (N) Ep-0, (O) Hl-0, and (P) Jl-5, with lanceolate leaves. Photographs were taken 30 days after sowing. Bars, 10 mm.

Reciprocal intercrosses were attempted between the accessions indicated in Table 4, and when their F₁ progeny was studied a low degree of phenotypic variation was found in all cases between individuals presumed to be of identical genotype, as is to be expected for outcrosses of inbred lines. The results obtained suggested inheritance patterns deviating from those expected for traits depending upon single biallelic genes. On the one hand, different results were obtained for the mode of inheritance of a given trait, when comparing all the crosses involving a given accession, the only exception being the Li-5-3 accession (with long petiole; see Table 4), whose F₁ progeny showed long petiole. On the other hand, at least in two cases, those of Ga-0 x Gr-3 (differing in their petiole length) and Hl-0 (with lanceolate leaves) x La-1 (with rounded leaves), differences between reciprocal crosses were observed.

To determine if a single gene was responsible for the variation in the morphological traits under study, we tried to analyze their F₂ phenotypic segregations. About 50 F₂ individuals derived from selfed F₁ plants were studied from all the successful crosses between accessions differing in marginal configuration, from three crosses between accessions differing in petiole length [Ga-0 x Bla-1 (F₁ progeny with long petiole), Ga-0 x Gr-3 (F₁ with short petiole), Bla-14 x Bla-1 (F₁ progeny with intermediate petiole length)], and from three crosses between accessions differing in overall leaf shape [Hl-0 x Bla-14 (F₁ with lanceolate leaves), La-1 x Hl-0 (F₁ with rounded leaves), and Jl-5 x Bla-14 (F₁ with intermediate leaves)]. Since the F₂ population constituted a phenotypic continuum in all cases, we concluded that the natural variability of the studied traits was likely to be of multigenic nature.

Microsatellite repeat number variation among accessions:
To determine genetic distances between the accessions under study, PCR amplification products were obtained and sized by means of fragment analysis at 22 microsatellites in those accessions displaying the traits in an extreme manner (see Table 1), only homozygous individuals being found (Table 5). The only exception was Gr-3, for which two alleles were detected in a single locus. We estimated the level of microsatellite polymorphism on the basis of the number of alleles and gene diversity, the latter being found to vary from 0.309 (AthPHYC, 2 alleles) to 0.984 (nga6, 15 alleles) with an average of 0.827 over the 22 microsatellites.

Several distance matrices were obtained with the MICROSAT program, calculated on the basis of different genetic distance measurements (see MATERIALS AND METHODS). Consensus phylogenetic trees were constructed from resampled data using either the neighbor-joining or the UPGMA methods. According to the phylogenetic trees obtained, two examples of which are presented in Fig 4, the studied accessions group into clusters that, in general, do not correlate with those made according to petiole length, marginal configuration, and overall lamina shape of vegetative leaves. This is in agreement with our results in the bootstrap analysis, given that most of the nodes are not well supported, suggesting star-like phylogenies for the accessions, as previously proposed (SHARBEL et al. 2000 ; NORDBORG et al. 2002 ). Only in a few cases were accessions with similar leaf phenotypes clustered together (Jl-5, Ler, and Aa-0, with long petiole, entire margin, and lanceolate leaves).

View larger version (33K): In this window In a new window Download PPT slide   Figure 4. Neighbor-joining trees of the Arabidopsis thaliana accessions studied in this work. Phylogram trees were constructed from a distance matrix calculated using the distance measurements D1 (A) and DKF (B). Sp and Lp, short and long petiole; Em and Sm, entire and serrated margin; Ll and Rl, lanceolate and rounded leaves. The numbers at the branchpoints indicate the number of times that the accessions to the right of the branch point grouped together out of 1000 bootstrap replicates.

Morphometric analysis of leaf architecture in the recombinant inbred lines:
We morphometrically analyzed several leaf architectural traits (see MATERIALS AND METHODS) in a mapping population of 100 RILs. We chose the third node leaf, assumed to be representative of juvenile leaves in the Ler-0 and Col-4 accessions, and the seventh node leaf, assumed to be a typical adult leaf. Both types of leaves were collected 25 days after sowing, from plants grown in strictly controlled environmental conditions (see MATERIALS AND METHODS). Although third leaves of Ler-0 and Col-4 were fully expanded, seventh leaves were not. With this approach we sought to identify not only the QTL involved in leaf morphology and expansion but also others responsible for the heteroblastic differences between juvenile and adult leaves.

We found variation both among RILs and between their parental accessions (Fig 5 and Table 6). Normal distributions of the phenotypic traits under study were obtained (Fig 6) for all the parameters analyzed, as was to be expected for quantitative traits controlled by multiple loci. In addition, in some RILs we found phenotypic values that were higher or lower than those shown by their ancestor accessions, indicating that Ler-0 and Col-4 may contain both positive and negatively balanced alleles for leaf architectural traits. The largest of these transgressions was observed for petiole length in seventh leaves (PLN7), with the N1909 line displaying a 78.8% increase above the Col-4 parental values and a 55.9% decrease of the N1977 line compared with Ler-0 values. The smallest transgression was shown for lamina length in third leaves (LLN3), with a 12.2% increase of N1909 with regard to Ler-0 and a 12.7% decrease of N1930 with regard to Col-4. In addition, broad-sense heritabilities of the traits under study were calculated (see MATERIALS AND METHODS), which ranged from 84.3% for LWN7 to 98.6% for LPN3. These high values of heritability are very likely due to the strictly controlled environmental conditions used in this study.

View larger version (76K): In this window In a new window Download PPT slide   Figure 5. Third (A–H) and seventh (I–P) leaves of selected RILs and their parental accessions (A and I, Ler-0; B and J, Col-4). Notable phenotypic variations are shown for leaves of some RILs displaying long petioles and large lamina (C and K, N1909; G, N1971), lanceolate (D, N1931) or rounded lamina (E, N1940), short petiole and small lamina (F, N1950; H, N1977; M, N1957; O, N1983), rounded lamina and smooth margin (L, N1935), lanceolate lamina and thick petiole (N, N1977), and highly serrated margin (P, N1989). Bar, 5 mm.

View larger version (52K): In this window In a new window Download PPT slide   Figure 6. Frequency distribution for the morphological traits studied in the RIL population and parental lines. N3, third node leaf; N7, seventh node leaf; TLN, total leaf number. Lamina perimeter (LP), lamina length (LL), petiole length (PL), lamina width (LW), petiole width (PW), major rosette diameter (MRD), and minor rosette diameter (mrd) are indicated in millimeters and lamina area (LA) is in square millimeters. Normal curves that fit the observed data are shown.

We determined the statistical correlations between the above-mentioned variables (Table 7), as estimative of genetic correlations between the characters under study, and found that all the pairs of traits of a given type of leaf displayed positive and highly significant ( = 0.01) correlations, the only exception being those involving petiole width. We found that the area, perimeter, length, and width of the lamina were highly correlated, suggesting the existence of a common genetic control. On the contrary, petiole and lamina lengths were slightly correlated (Table 7), suggesting that several loci differentially participate in the elongation processes of these two leaf subdomains. Similar positive correlations among morphological traits have also been found for floral organs in A. thaliana (JUENGER et al. 2000 ). We found lower correlations between the same traits in leaves from different nodes than between the different traits from the same leaf node, perhaps reflecting differences in leaf developmental profiles between juvenile and adult leaves. As an example, LLN3 showed strong genetic correlations with LAN3 (r_G = 0.943), LWN3 (r_G = 0.822), or PLN3 (r_G = 0.641) and a lower degree of genetic correlation with LLN7 (r_G = 0.494).

We analyzed the NMS in the studied lines as an estimate of the number of hydathodes, which is known to differ between juvenile and adult leaves (TSUKAYA and UCHIMIYA 1997 ; CANDELA et al. 1999 ). Only leaves from the seventh node showed any variation, which was not sufficient to carry out a quantitative analysis, the mean values obtained ranging from RILs with none up to 10 marginal serrations (data not shown).

In addition, we analyzed the TLN in the RIL population 25 days after sowing, where differences could be due to differences in both the flowering time and the time profile of leaf production. Significant variation was found with differences of up to four leaves between extreme values. Major and minor rosette diameters were also measured but not used for QTL mapping because differences between these parameters are dependent not only on leaf length but also on the angle of the petiole to the stem (not measured).

QTL mapping:
The above-mentioned results on the morphometric analysis of two representative leaves, one juvenile and one adult, from 15 plants from each of 100 RILs were used for QTL mapping. We first determined the number and location of the QTL contributing to the traits under study in juvenile and adult leaves: lamina area and perimeter, lamina width and length, petiole length and width, number of leaf marginal serrations, total leaf number, and rosette diameters (Table 8). A LOD score significance threshold of 2.7 was calculated for QTL identification (see MATERIALS AND METHODS). We have taken into account, however, four putative QTL (see ju-PLE2 and ju-LaWI1 in Table 9 and ad-LaSI2 and ad-PLE1 in Table 10), whose LOD scores are below the 2.7 threshold but still make a significant contribution to the explained variance. The QTL identified (from 2 to 11 QTL for PWN7 and LPN3 or LWN3, respectively) explained a higher percentage of the phenotypic variance found for juvenile leaves (average of 69.3%) than for the adult ones (56.9%). In all cases, a large percentage of this variance was explained by one or two QTL with a strong effect, and the remaining variation was apparently due to a large number of weaker (with smaller effects) QTL, as previously proposed (ROBERTSON 1985 ). The proportion of the total phenotypic variance explained for each leaf trait by these QTL ranged from 48.0% in LWN7 to 72.3% in PLN3.

Since the genetic correlations between the traits under study were highly significant and similar heritabilities were found, we assumed it would be possible to detect the same QTL in different analyses, for which reason we grouped all the QTL that displayed neighboring map positions identified from the analyses of different leaf parameters (Table 9 and Table 10). As a result, we were able to distinguish between the QTL affecting all the studied parameters of the whole organ (leaf size QTL or LSI), the lamina (lamina size QTL or LaSI), or the petiole (petiole size QTL or PSI) from those affecting either the organ's length (lamina length QTL or LaLE and petiole length QTL or PLE) or its width (lamina width QTL or LaWI and petiole width QTL or PWI). For juvenile leaves, a total of 16 QTL were found in this way (Table 9 and Fig 7). Five QTL were shown to affect whole leaf size (ju-LSI1 to ju-LSI5), 2 of which, those linked to er (ju-LSI2 in chromosome 2) and to g4028 (ju-LSI5 in chromosome 5), would explain >40% of the observed variance. One QTL was specific for petiole size (ju-PSI1), and 3 were responsible for 25% of the variance in lamina size (ju-LaSI1 to ju-LaSI3). Several minor-effect QTL for length (1 QTL for leaf length and 3 for petiole length) or width (3 QTL for lamina width) were also identified. For these juvenile leaf QTL, LOD values ranged from 2.47 (ju-LaWI2, identified in the LWN3 analysis) to 14.45 (ju-LSI2, identified in the PLN3 analysis) and the 2-LOD intervals ranged from 44.4 cM (ju-PLE2, identified in the PLN3 analysis) to 4.4 cM (ju-LSI5, identified in the LAN3 analysis) with an average of 13.6 cM.

View larger version (33K): In this window In a new window Download PPT slide   Figure 7. Likelihood plots for QTL affecting juvenile leaf architectural traits. LOD scores are indicated along the ordinate and map positions along the abscissa in centimorgans. The dotted line at LOD 2.7 indicates the significance threshold for QTL identification.

In the case of adult leaves we found 13 QTL (Table 10 and Fig 8), 4 of which affected leaf size (the one linked to the g4028 marker was also found in juvenile leaves), 5 lamina size QTL (that of ad-LaSI1 on chromosome 2 was responsible for 10% of the variance), and 4 QTL affecting only petiole size (2 of which were responsible for nearly 50% of the variance). Surprisingly, 1 of these 2 QTL was that closest to the er marker, which also affected juvenile leaf size but was not detected in the adult leaf lamina analyses. For the adult leaves, LOD values ranged from 1.47 (ad-LSI2 detected with the PWN7 analysis) to 18.30 (ad-PSI1 detected with the PWN7 analysis) and the 2-LOD intervals ranged from 42.3 cM (ad-PLE2, identified in the PLN7 analysis) to 4.9 cM (ad-PLE3, identified in the PLN7 analysis) with an average of 17.1 cM. Due to the high phenotypic correlations between juvenile and adult leaves, we expected to find similar QTL affecting the same trait in different leaves. However, only 8 QTL were found to be apparently common to juvenile and adult leaves.

View larger version (31K): In this window In a new window Download PPT slide   Figure 8. Likelihood plots for QTL affecting adult leaf architectural traits. See Fig 7 legend.

Altogether, 21 QTL affecting leaf morphology were found, 8 of which were shared by juvenile and adult leaves, in the genome of A. thaliana (4, 2, 3, 5, and 7 QTL were found in chromosomes 1, 2, 3, 4, and 5, respectively). The proportion of the phenotypic variance explained by individual QTL ranged from 3.4% (ju-PSI1, identified in the PLN3 analysis) to 26.7% (ju-LSI2, identified in the PLN3 analysis) in juvenile leaves and 2.3% (ad-LSI2, identified in the PWN7 analysis) to 44.1% (ad-PSI2, identified in the PWN7 analysis) in the adult ones.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Substantial variability exists in leaf architectural traits among A. thaliana accessions:
The AIS collection of accessions is one of the largest sets of wild-type races of A. thaliana. Maintained for years by A. R. Kranz, it was created by F. Laibach and enlarged later with the addition of new accessions by other researchers such as G. Röbbelen, D. Ratcliffe, and C. Gómez-Campo (KRANZ 1978 ). This collection of accessions constitutes an excellent sample of the variability existing in leaf shape and size as well as in rosette structure among A. thaliana natural races. Some of this variability has probably arisen because of flowering time diversity, which causes differences in the number of rosette leaves, which are continuously generated until bolting. Differences in the levels of activity of genes controlling leaf morphogenesis could also contribute to natural variations in vegetative leaf shape and size, traits that are traditionally assumed to be quantitative, due to the joint action of several genes whose expression is probably influenced by the environment. Controlled conditions of substrate, light, temperature, and humidity should reduce such environmental phenotypic variation, allowing the genetic component of the traits under study to be unraveled.

Very few studies have been published on the inter- and intraecotypic variability of life history traits, such as the timing of leaf primordia initiation and the number of leaves in A. thaliana plants grown under controlled culture conditions (DOBROVOLNA 1967 ; ZWAN et al. 2000 ). Developmental traits were shown to be the most variable, not only between different accessions but also within a given accession. Similar results were obtained by KARBE and ROBBELEN 1968 , who studied rosette height and leaf number, shape, marginal configuration, and color. These results indicate apparent heterozygosity, at least for the studied traits (DOBROVOLNA 1967 ), in contrast with the high level of autogamy that A. thaliana is known to present (ROBBELEN 1971 ).

We have studied the variability of a large group of accessions from the AIS collection under controlled culture conditions, classifying them into 14 phenotypic groups according to overall leaf shape, leaf marginal configuration, and rosette structure. Only minor intraecotypic variability was found for the studied traits, all the individuals of a given accession being unambiguously assigned to the same phenotypic group, the only exceptions being Jl-5 and Li-5-3. Given that these two accessions included two clearly distinguishable subpopulations, we chose plants belonging to the subpopulation displaying the trait of interest for further analysis. We also studied leaf initiation rates and found that only a few accessions deviated from the predominant pattern. Therefore, substantial variability exists in the shape and size of A. thaliana vegetative leaves as well as in rosette structure, which is unlikely to be related with differences between vegetative growth rates.

Leaf variants among accessions represented variations less extreme than those found in mutant searches. Many of the leaf mutant lines included in the AIS form mutants collection display phenotypes that are more extreme than those of any of the natural variants studied here (BURGER 1971 ; KRANZ 1978 ; SERRANO-CARTAGENA et al. 1999 , SERRANO-CARTAGENA et al. 2000 ). Likewise, most of the lines that we have isolated in a large-scale screen for EMS-induced leaf mutants (BERNA et al. 1999 ) presented phenotypes that are more extreme than those of AIS ecotypes. Genetic analysis of two such collections of mutants has shown that almost all the studied traits were monogenic and that mutations affecting at least 57 different genes in the former (SERRANO-CARTAGENA et al. 1999 ) and 94 in the latter group of lines (BERNA et al. 1999 ) were responsible for the leaf phenotypes.

Quantitative analysis of the expansion of En-2 rosette leaves:
To analyze mutants displaying morphological aberrations, criteria are required to determine the nature of their differences with wild-type individuals. A considerable amount of information on wild-type leaf growth has been obtained in different plant species (reviewed in CUSSET 1986 ; DALE 1988 ), some of them dicotyledonous, such as Phaseolus vulgaris (DALE 1964 ), Vicia faba (DENNET et al. 1978 ), Glycine max (BARTHOU and BUIS 1988 ), Lycopersicon esculentum (DENGLER 1984 ), Nicotiana tabacum (POETHIG and SUSSEX 1985 ), Vitis riparia (LACROIX and POSLUSZNY 1990 ), and Cucurbita argyrosperma (JONES 1993 ), and others monocotyledonous, such as Festuca arundinacea (SKINNER and NELSON 1994 ). Studies on the expansion of A. thaliana cotyledons (TSUKAYA et al. 1994 ), the first vegetative leaf (PYKE et al. 1991 ), and all rosette leaves (TSUGE et al. 1996 ) have been published. Morphometric analysis has been performed in only some of the above-mentioned cases.

We performed a morphometric analysis of the expansion of the rosette leaves of the Enkheim-2 accession, which was chosen because it represents the genetic background for most of the mutant lines belonging to the large AIS form mutants collection. Our quantitative results agree with the qualitative observations that previous authors made on other accessions and provide a framework for the phenotypic characterization of AIS leaf form mutants, making possible their precise comparison with the wild-type pattern of leaf organogenesis. Our data on the growth of all the vegetative leaves of En-2 confirm and extend previous studies on the first leaf of Ler (PYKE et al. 1991 ) and the cotyledon of Col-0 (TSUKAYA et al. 1994 ). The results obtained on the area, perimeter, length, and width of all the rosette leaves, collected from the 10th to the 32nd day after sowing, provide information on the morphological differences between successive leaves as well as on the variation with time of the shape and size of each leaf and will facilitate quantitative comparisons between the En-2 accession and leaf mutants with an En-2 background, such as those of the AIS collection. The leaf parameters studied in this work in En-2 allowed us to reach conclusions similar to those made by previous authors studying different genetic backgrounds on the heteroblastic differences between A. thaliana vegetative leaves (TSUKAYA et al. 2000 ) and the proportionality of leaf expansion in the dimensions of width and length (TSUGE et al. 1996 ).

Natural variation in leaf architecture is multifactorial in A. thaliana:
After the classification of wild-type strains according to leaf architectural traits, we tried to determine the genetic basis of the observed variations. For that purpose, we chose accessions displaying extremes of the most representative trait of each phenotypic group and performed crosses involving accessions displaying a given trait in two extreme and opposite ways, their F₁ and F₂ progenies being analyzed. Of note was the high number of unsuccessful crosses, suggesting that accession diversification has led to some degree of incompatibility.

We expected to obtain, at least in some cases, a discrete number of phenotypic groups among the F₂ progeny of intercrosses, making it possible to estimate the number of genes underlying some of the traits under study. However, despite the substantial number of crosses performed, no obvious phenotypic classes were found in their F₂ progeny, so we could not directly estimate the number of genes controlling the traits under study. Our failure to find natural monogenic variants in petiole length, marginal configuration, and overall lamina shape contrasts with our own results with regard to venation pattern, another leaf architectural trait (CANDELA et al. 1999 ). In fact, a search for natural variations in venation patterning in the first vegetative leaves of 266 accessions resulted in finding 1, Ei-5, which shows unequivocally different patterning from that of the rest and which is inherited as a monogenic recessive trait.

Polymorphic microsatellites, often referred to as simple sequence repeats (SSR) or simple sequence length polymorphisms (SSLP), have been found in A. thaliana (BELL and ECKER 1994 ), as well as in other eukaryotes (TAUTZ 1989 ; WEBER and MAY 1989 ; HEARNE et al. 1992 ). Microsatellites have been proven useful for estimating genetic distances between closely related species, as well as between subpopulations of a single species (BOWCOCK et al. 1994 ), such as in A. thaliana, where they vary greatly among accessions (INNAN et al. 1997 ; ZWAN et al. 2000 ).

The morphological traits studied allowed us to define clusters of accessions that clearly do not correlate with genetic distances measured according to microsatellite polymorphisms, a type of molecular marker that we used because of their high level of polymorphism and ease of genotyping. INNAN et al. 1997 determined the level of polymorphism at 20 microsatellite loci in a worldwide sample of 42 A. thaliana accessions, while we analyzed 22 microsatellites in 16 accessions. The values of gene diversity found in all the loci studied, 0.794 (INNAN et al. 1997 ) and 0.827 (this work), are remarkably close since only 2 accessions (Aa-0 and La-1) and five markers (nga111, nga168, AthCHIB, nga162, and AthCTR1) were used in both studies. As in previous studies (INNAN et al. 1997 ; ZWAN et al. 2000 ), we found no clear correlations between genetic distances and the geographic origins of A. thaliana accessions.

Furthermore, we found that microsatellite-specific distance measurements did not correlate with morphological grouping of the accessions studied. Neighbor-joining phylogenetic trees of three different topologies were obtained from distance matrices calculated using (a) the D₁ and (µ)² distance measurements; (b) the D_KF, D_PS, and D_FS distance measurements; and (c) the D_AD distance measurement, none of which showed clustering of the accessions studied in relation to the morphological traits under study (Fig 4). The fact that the trait-based leaf morphological clustering and microsatellite-based phylogeny did not correlate in the studied accessions reinforced the hypothesis that intraspecific variability in leaf morphology arises from the accumulation of mutations at quantitative trait loci in A. thaliana. This is further shown by the QTL analyses performed.

Although the most likely explanation for our results is that the studied leaf phenotypes are controlled by QTL, the experimental approach required to test such a hypothesis in the F₂ individuals obtained was considered to be beyond the scope of this work, since this would require not only a detailed morphometric and statistical analysis of the F₂ progeny plants obtained, but also their individual genotyping for at least 100 molecular markers, together with that of their parental accessions.

QTL affecting leaf morphology:
It has been known for a long time that rosette leaves produced throughout the vegetative development of A. thaliana can be distinguished from one another by their size and shape (TSUKAYA et al. 2000 ). To undertake a quantitative genetic dissection of leaf morphogenesis, we studied the third and seventh vegetative leaves from plants belonging to a mapping population of 100 RILs derived from a Ler-0 x Col-4 cross (LISTER and DEAN 1993 ). Leaves were collected 25 days after sowing, when third leaves are fully expanded but seventh leaves are not. We used 173 markers covering 519.5 cM of the Arabidopsis genome, which, if randomly chosen, could help to detect a QTL by marker linkage within 3 cM and with a probability of 85%.

We identified 16 QTL affecting highly correlated leaf morphological traits in juvenile leaves: 5 affecting the overall form of the leaf organ, 6 specific for the lamina, 4 affecting only the petiole, and 1 modifying the length but not the width of the whole leaf. In addition, a total of 13 QTL were identified in adult leaves: 4 leaf size QTL, 5 specific for the lamina, and 4 specific for the petiole. In both analyses we found three pairs of linked QTL with opposite effects (ju-LSI3 and ju-LSI4, ju-LaSI1 and ju-LaSI2, and ad-LSI2 and ad-LSI3), which could be discriminated thanks to the high density of markers employed. In juvenile leaves, at least 50% of the variance in leaf size could be explained by two large-effect QTL: ju-LSI2, which is linked to ER, and ju-LSI5.

Among the QTL identified for juvenile leaves, those represented by ju-LSI3, ju-LSI5, ju-LaSI2, ju-LLE1, and ju-LaWI2 had alleles that increased the phenotypic values of the Ler-0 parental, whereas ju-LSI1, ju-LSI2, ju-LSI4, ju-LaSI1, and ju-LaSI3 had alleles showing a positive effect on the Col-4 parental. For the adult leaves, the Col-4 alleles of ad-LSI1, ad-LSI2, ad-LaSI1, ad-LaSI2, ad-LaSI5, ad-PLE1, and ad-PLE3 had a positive effect over the variance, whereas the Ler-0 alleles of ad-LSI3, ad-LSI4, ad-LaSI3, and ad-LaSI4 increased the variance.

As regards the other parameters analyzed, the total leaf number was scored to detect the QTL responsible for the time profile of production of vegetative leaves, although some flowering-time QTL were expected to be identified in our analyses. We found two major QTL (TLN1, linked to the JGB9 marker in chromosome 4, and TLN2, linked to ve018 in chromosome 2), which were responsible for 25% of the observed variance and whose map positions make them candidates to be alleles of FWA and TOC2, respectively.

Candidate genes:
Although it is assumed that the understanding of the genetic architecture of quantitative traits, which begins by mapping QTL to broad genomic regions, should end with the molecular definition of QTL alleles (MACKAY 2001 ), the resolution of QTL analysis is not sufficient for the positional cloning of candidate genes (DARVASI et al. 1993 ; BOEHNKE 1994 ). Consequently, the confirmation that a particular gene is, in fact, a candidate requires narrowing a QTL interval by Mendelization using near-isogenic lines (NILs) to finally achieve the molecular identity of the natural alleles involved in the trait under study. The isolation of near-isogenic lines, which is under progress for some of the major QTL, will help us to narrow the genome interval harboring the QTL identified in this work.

From our results concerning the map positions and phenotypic effects of the identified QTL affecting leaf morphological traits, we looked for candidate genes by using the available genetic and molecular data on leaf mutants. A large leaf mutant collection has been obtained in our laboratory (BERNA et al. 1999 ; ROBLES and MICOL 2001 ), some of whose phenotypic classes correspond to alterations in the size or proportions of leaves, such as those named Rotunda (Ron) and Orbiculata (Orb), with rounded leaves; Exigua (Exi), with small leaves; and Elongata (Elo), with lanceolate leaves and long petioles. Several of the QTL detected in this work had map positions neighboring those of some mutants of the above-mentioned phenotypic classes (ROBLES and MICOL 2001 ). The ju-LLE1 QTL maps near the lower telomere of chromosome 1, close to the EXI6 gene. Another QTL affecting leaf size, in both juvenile and adult leaves, that mapped near the g4028 marker (ju-LSI5 and ad-LSI4) in chromosome 5, had a strong effect on the phenotypic variance and mapped close to the EXI2 and ELO3 genes, whose mutant alleles cause altered proportions of the lamina and petiole size. Another QTL, ju-LaSI3, maps close to the ORB1 gene, in chromosome 5, and both ju-LaWI2 and ad-LSI3 are located in the neighborhood of the ELO2 gene, in chromosome 3. The EXI1 gene maps in chromosome 4, close to a QTL involved in lamina size in both juvenile and adult leaves (ju-LaSI2 and ad-LaSI4).

Mutant alleles of the ER gene, which encodes a leucine-rich repeat (LRR) receptor protein kinase, display a compact inflorescence, blunted fruits, and short petioles (TORII et al. 1996 ). Several QTL mapping close to ER have been identified in A. thaliana (ALONSO-BLANCO et al. 1999 ; SWARUP et al. 1999 ; JUENGER et al. 2000 ; BOREVITZ et al. 2002 ). It is likely that ju-LSI2 and ad-PSI2 are probably caused by the erecta mutation carried by the Ler-0 accession. The effect of ER, or that of its linked QTL, was detected both for the laminae and the petioles of juvenile leaves, but only for the petioles of adult leaves. These results are congruent with the described pattern of expression of the ER gene, whose transcripts are more abundant in juvenile than in adult leaves (TORII et al. 1996 ; YOKOYAMA et al. 1998 ).

	ACKNOWLEDGMENTS

We are grateful to J. M. Barrero, H. Candela, S. García, S. Jover, M. R. Ponce, P. Robles, V. Quesada, J. Rozas, and two anonymous referees for comments on the manuscript; to the NASC for providing seeds of accessions; and to S. Gerber and J. M. Serrano for their expert technical assistance. We are especially indebted to C. Alonso-Blanco and an anonymous reviewer for their useful suggestions. This research was supported by PB91-0749, APC95-019, PB95-0685, and PB98-1389 grants from the Ministerio de Educación y Cultura of Spain. J. M. Pérez-Pérez and J. Serrano-Cartagena were fellows of the Conselleria de Cultura, Educació i Ciència of the Generalitat Valenciana.

Manuscript received April 5, 2002; Accepted for publication July 12, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ALONSO-BLANCO, C. and M. KOORNNEEF, 2000  Naturally occurring variation in Arabidopsis: an underexploited resource for plant genetics. Trends Plant Sci. 5:22-29.[Medline]

ALONSO-BLANCO, C., S. E. EL-ASSAL, G. COUPLAND, and M. KOORN