Tendril-less regulates tendril formation in pea leaves

Tendrils are contact-sensitive, ﬁlamentous organs that permit climbing plants to tether to their taller neighbors. Tendrilled legume species are grown as ﬁeld crops, where the tendrils contribute to the physical support of the crop prior to harvest. The homeotic tendril-less ( tl ) mutation in garden pea ( Pisum sativum ), identiﬁed almost a century ago, transforms tendrils into leaﬂets. In this study, we used a systematic marker screen of fast neutron–generated tl deletion mutants to identify Tl as a Class I homeodomain leucine zipper (HDZIP) transcription factor. We conﬁrmed the tendril-less phenotype as loss of function by targeting induced local lesions in genomes (TILLING) in garden pea and by analysis of the tendril-less phenotype of the t mutant in sweet pea ( Lathyrus odoratus ). The conversion of tendrils into leaﬂets in both mutants demonstrates that the pea tendril is a modiﬁed leaﬂet, inhibited from completing laminar development by Tl . We provide evidence to show that lamina inhibition requires Unifoliata / LEAFY -mediated Tl expression in organs emerging in the distal region of the leaf primordium. Phylogenetic analyses show that Tl is an unusual Class I HDZIP protein and that tendrils evolved either once or twice in Papilionoid legumes. We suggest that tendrils arose in the Fabeae clade of Papilionoid legumes through acquisition of the Tl gene. pairs of in A of the pair


INTRODUCTION
Many climbing plants use specialized organs called tendrils for support. Some tendrils explore the physical environment with characteristic circling movements (Darwin, 1875) followed by contact-induced coiling (Jaffe and Galston, 1968), permitting the plant to obtain support by grasping onto and entwining its neighbors. Plant tendrils may be derived from a variety of structures, such as leaf parts, whole leaves, or stems (Bell, 1991); for example, the grapevine tendril is a gibberellin-inhibited inflorescence (Boss and Thomas, 2002). Such diverse derivations, and the fact that tendrilled taxa are widespread in flowering plants (Darwin, 1875), suggest that tendrils are an example of convergent evolution. These novel organs enable plants to reach the canopy, where they can spread and maximize opportunities for pollination, photosynthesis, and seed dispersal with minimal energy investment in expensive supporting structures. Indeed, the climbing habit is associated with species richness compared with nonclimbing sister taxa (Gianoli, 2004) (see Supplemental  Table 1 online), suggesting a selective advantage.
The three subfamilies of legumes, Caesalpinioideae, Mimosoideae, and Papilionoideae, together comprise >19,000 species, one of the largest flowering plant families (Lewis et al., 2005). Tendrils appear to have evolved independently at least once in each subfamily. In Bauhinia spp (Caesalpinioideae), tendrils arise at the base of the leaf, while tendrils form in the distal region of the leaf in Entada spp (Mimosoideae) and in peas, lentils, vetches, and chickpeas (Papilionoideae). All the economically important grain legume species are Papilionoids, which collectively provide approximately one-third of the total dietary protein needs of humans, as well as being used widely as animal feed. Under intensively planted field conditions, tendrils can form an interwoven network of support, conferring partial resistance to crop collapse or lodging. Therefore, a better understanding of tendril formation has the potential to aid agronomic performance and to provide insight on convergent morphological evolution.
Most legume leaves are compound (Lewis et al., 2005), with each leaf carrying one or more pairs of leaflets along the leaf axis. The leaf is further specialized in Papilionoid legume species belonging to the clades Cicereae (chickpeas) and Fabeae (peas, lentils, and vetches) where the organ formed at the terminal position of the leaf is a tendril, rather than a leaflet. Many species within the Fabeae are more extensively tendrilled, for example, garden pea (Pisum sativum) and sweet pea (Lathyrus odoratus) also produce pairs of tendrils in subterminal positions. A key regulator of the compound leaf trait in legumes is the meristem identity gene, Unifoliata (Uni), the ortholog of LEAFY (LFY) in Arabidopsis thaliana (Hofer et al., 1997). In garden pea, Uni maintains the meristematic potential of the compound leaf, enabling the sequential development of pairs of leaflets and tendrils in acropetal (first pair at the base and last pair at the tip) order. This role is shown by uni null mutants, which bear leaves composed of only a single leaflet (Hofer et al., 1997). A semidominant locus regulating tendril formation has long been known in garden pea (de Vilmorin, 1910;de Vilmorin and Bateson, 1911) and sweet pea (Punnett, 1923). Mutants in both species were originally called acacia (t) because of their tendril-less leaves; the locus was later renamed tendril-less (tl) in garden pea. It was not known if these were orthologous loci, and identification of the genes remained elusive. In this study, we employed an amplified fragment length polymorphism (AFLP) screening method to identify Tl as a Class I homeodomain leucine zipper (HDZIP) gene that confers the tendrilled trait on peas.

Generation of New tl Mutant Alleles and Identification of Tl
We used fast neutron (FN) mutagenesis to generate new tl deletion alleles in a conveniently dwarf, early flowering garden pea genotype. This allowed us to screen DNA samples for markers that distinguished wild-type plants from mutants. Compared with the wild-type leaf ( Figure 1A), narrow, subterminal leaflets were found in place of tendrils in heterozygous FN mutants ( Figure 1B), as expected for this semidominant mutation (de Vilmorin and Bateson, 1911;Marx, 1973), while the homozygous FN mutants displayed a classic homeotic transformation of tendrils into leaflets ( Figure 1C). Tendril-less F1 progeny were obtained from tendril-less FN mutants crossed to lines carrying the tl-w type allele, confirming that the new FN mutants all carried allelic mutations. Notably, the wild-type, heterozygous (tl/Tl) and homozygous (tl/tl) garden pea phenotypes resembled phenotypes in sweet pea that were correspondingly wild-type ( Figure  1D), heterozygous ( Figure 1E), or homozygous ( Figure 1F) at the t locus (Punnett, 1923), which suggested that tl and t might be orthologous loci.
Genomic DNA from five new FN tl lines was pooled and screened for AFLP markers (Vos et al., 1995) that differentiated the mutant pool from the wild type. A 218-bp marker, absent from all five FN alleles and present in the wild type, was sequenced, and primers derived from this were used to screen a ninefold redundant P. sativum cv Cameor HindIII BAC library. BAC genomic DNA sequencing revealed that the marker spanned the second intron-exon junction of a gene encoding a 237-amino acid Class I HDZIP protein (Figure 2A). The entire coding sequence was deleted from all tl FN mutants used in the AFLP screen, indicating that deletions larger than 1 kb occur in this mutagenized population and that a loss of function causes the tendril-less phenotype. A reverse genetics screen for this gene in an ethane methanesulfonate (EMS)-mutagenized targetinginduced local lesions in genomes (TILLING) population (Dalmais et al., 2008) confirmed the identity of Tl. An M2 plant carrying a W117STOP TILLING lesion, designated tl-18 ( Figure 2A; see Supplemental Table 2 online), subsequently yielded tendril-less M3 progeny homozygous for the mutation ( Figure 1G).

Characterization of Previously Described tl Mutant Alleles
Five tl mutants described in the literature had been observed or generated in a variety of pea genotypes (see Supplemental  Table 2 online), so we determined first whether each mutant and its reported progenitor were related using a fingerprinting technique (Ellis et al., 1998). Sequence-specific amplification polymorphism displays confirmed that alleles tl-7, tl-13, tl-16, and tl-17 were maintained in stocks near-isogenic to their specified progenitor wild-type lines. The progenitor of the original spontaneous tl mutant allele, tl-w, is unknown (de Vilmorin and Bateson, 1911); however, as a result of backcrossing (Marx, 1973), line JI 1197 carrying tl-w is near-isogenic to wild-type line JI 1194, and this was also confirmed by fingerprinting.
Sequencing revealed that tl-w contains a 3-bp insertion encoding an additional Ile residue at position 149, within the leucine zipper domain (Figure 2A). We predict that this transcribed allele ( Figure 2B) produces a nonfunctional protein, altered in its capacity for homo-or heterodimerization via its leucine zipper. We investigated three independent tl deletion alleles that are not transcribed ( Figure 2B) and phenotypically resemble the tl-w type line. Alleles tl-13 and tl-16 (Vassileva, 1979) are radiationinduced complete gene deletions, and tl-12 (Figure 2A), a spontaneous allele, carries a 1908-bp deletion, including the promoter and the first six amino acids of coding sequence (see Supplemental Table 2 online). Allele tl-17 ( Figure 2A) carries an EMSinduced C/T transition, resulting in a Q226STOP mutation, which would truncate the C terminus by 12 amino acids. This allele is transcribed ( Figure 2B), yet the phenotype of the mutant is similar to the deletion alleles, suggesting that the C-terminal domain may have an important function in stabilizing the protein.
The spontaneous allele tl-pet (Lamm, 1957) differs from other tl alleles in that distal leaflets are borne on elongated stalks (see Supplemental Figure 1 online). We were unable to identify the progenitor of tl-pet, but compared with all other alleles, it carries a unique G/C transversion in the promoter (Figure 2A). The altered nucleotide corresponds to the second position of a putative GGTCCAT auxin-responsive cis-regulatory element (Lescot et al., 2002), 115 bp from the ATG corresponding to the start codon and 33 bp upstream of a predicted TATA box (Bucher, 1990). Transcription of this allele is severely impaired ( Figure 2B), suggesting that Tl may be an auxin-regulated gene. Auxin inhibitor studies in pea (DeMason and Chawla, 2004) and patterns of auxin transport during primordium formation suggest that auxin plays a role in regulating primordium type and polarity (Heisler et al., 2005). Studies in compound-leaved Cardamine hirsuta showed that pinformed1 mutants, which fail to accumulate auxin in their leaf rachis, fail to separate leaflet from rachis correctly (Barkoulas et al., 2008). An inability to respond appropriately to auxin may account for the stalked leaflet phenotype of tl-pet mutant leaves.

Tl-Related Genes in Other Species
Class I HDZIP genes that play a role in diverse developmental processes have been described (Ariel et al., 2007;Komatsuda et al., 2007). Therefore, the relationship between Tl and other Class I HDZIP sequences in nontendrilled model plant species is of interest. A maximum likelihood tree based on aligned Class I HDZIP domains ( Figure 3) shows that Tl and its immediate relatives (see below) are most closely related to At2g36610 and At5g03790 from Arabidopsis and AC139525_29 from Medicago truncatula. At2g36610 encodes an unusual plant Class I HDZIP protein that contains an eight-amino acid insertion between helix 1 and helix 2 of the homeodomain (see Supplemental Figure 2 online), followed by an exceptionally short C terminus. These features mean that it is less likely to resemble a Tl progenitor sequence than At5g03790 and AC139525_29. At5g03790, also known as LATE MERISTEM IDENTITY1 (LMI1), is a gene identified as a direct promoter binding target of LFY (William et al., 2004) that acts together with LFY to promote floral meristem identity in Arabidopsis (Saddic et al., 2006).
A clear LMI1 ortholog with conserved functions in leaves and flowers has not been identified in pea or closely related legumes, but one candidate is AC139525_29, which we derived from Medicago BAC sequence data after manual editing to remove sequence corresponding to a 1249-nucleotide second intron. Spliced full-length transcripts were confirmed by sequencing products obtained from PCR on reverse transcribed cDNA. Although AC139525_29 is a potential LMI1 ortholog, it is not well supported as a Tl ortholog because it shares only 62% open reading frame nucleotide sequence identity and maps (http:// www.medicago.org/genome/cvit_blast.php) to a nonsyntenic position on Medicago chromosome 1 (highest TBLASTN similarity score of 1e-40).
To identify a possible molecular basis for differences in Tl gene function compared with other Class I HDZIPs, we compared their HDZIP regions ( Figure 4). The legume Tl sequences are distinguished by their shorter leucine zippers, which contain only four Leu residues, whereas most other Class I HDZIPs have zippers comprised of five or six hydrophobic residues. Residues in DNA binding helix 3 of the homeodomain are identical or  conservatively substituted in all sequences compared, except for a Thr replacement for Ala at position 123 of Tl (Figure 4).

Tl Is Expressed in Tendril Primordia
Surgical experiments have shown that terminal tendril fate in pea leaves is not determined at leaf initiation (plastochron 1, Figure  5A) but is acquired later, between plastochrons 3 and 5 and subsequent to the specification of all other lateral organs (Gould et al., 1994). In situ hybridization experiments revealed that Tl mRNA accumulates in terminal tendril primordial cells at plastochron 3 ( Figure 5B), the earliest stage at which tendril fate is determined, and continues to be expressed there until at least plastochron 4 ( Figure 5D). Tl mRNA was detected in plastochron 4 tendril primordia but not in adjacent leaflet primordia ( Figure  5E). Tl transcripts did not accumulate in other vegetative organs, the shoot apex, or developing flowers.
The transcription start site of Tl was mapped by 59 rapid amplification of cDNA ends (RACE) PCR to a CA dinucleotide within a CCANTG LFY binding motif (William et al., 2004), 49 nucleotides upstream of the ATG corresponding to the start codon. Uni is the garden pea ortholog of the Arabidopsis gene LFY (Weigel et al., 1992;Hofer et al., 1997), and it plays a role in maintaining the meristematic potential of the compound leaf, enabling pairs of leaflets, followed by pairs of tendrils, to develop. Uni can fully complement Arabidopsis lfy mutants; therefore, it must share LFY activities (Maizel et al., 2005;Wang et al., 2008), including promoter binding. We postulated that if the predicted LFY binding motif represented an actual Uni binding motif, then Tl transcription would be dependent on Uni. We tested this hypothesis genetically by generating uni tl double mutants and found that unifoliolate uni single mutant and uni tl double mutant phenotypes were indistinguishable in a segregating population (t test, P = 0.63; see Supplemental Table 3 online). This shows that uni is epistatic to tl and that these two genes most likely act in the same developmental pathway. Next, we examined the expression patterns of the two genes in situ, predicting that if Uni regulates transcription initiation by binding to the CCANTG motif, then the two genes would have overlapping domains of expression. Adjacent tissue sections hybridized with a Uni probe ( Figure  5C), showed accumulation of Uni mRNA at leaf initiation and also later, in the distal region of plastochron 3 primordia, where Tl expression was observed ( Figure 5B). This confirmed results from earlier work showing that Uni expression can be detected as late as plastochron 4 (Gourlay et al., 2000). The expression of Uni and Tl therefore overlaps both temporally, during plastochrons 3 and 4, and spatially, in the distal region of leaf primordia, where tendril initiation occurs.  Finally, we tested the dependence of Tl expression on Uni by quantitative PCR. Uni expression is known to be upregulated in the afila (af) genotype (Gourlay et al., 2000), a prolifically tendrilled mutant of pea used widely in agriculture ( Figure 5G). Our expectation that Tl mRNA would also accumulate to higher levels in this genotype due to the increased number of tendril primordia was confirmed by in situ hybridization ( Figure 5F) and quantitative PCR ( Figure 5H). The af genotype was used as a sensitive reporter of Tl expression in further quantitative PCR analyses. Tl mRNA levels were shown to be reduced fivefold in a uni,af genotype compared with a Uni,af genotype ( Figure 5H), indicating that Tl transcription is positively regulated by Uni. Similar results were obtained in a uni mutant genotype compared with its corresponding wild type (JI 2822), where Tl expression was reduced 2.5-fold in the mutant. Notably, Tl expression was not abolished completely in uni af mutants or uni mutants, suggesting that a basal level of Tl transcription can occur, even in the absence of Uni.

Tl Is Present in Other Tendrilled Legumes
The phenotypic similarity of the tendril-less garden pea and sweet pea mutants (Figure 1) suggested that tl and t might be orthologous loci. To test for cosegregation of t and sweet pea Tl, an F2 population of 185 individuals was generated from a cross between sweet pea cultivar America and a tendril-less, homozygous t/t mutant (Punnett, 1923). Garden pea Tl gene primers that flanked the HDZIP region were used to amplify a PCR marker suitable for scoring the F2 population. A presence-absence polymorphism distinguished the wild-type and t/t parents, which suggested that the mutant carried a gene deletion. Absence of the marker cosegregated with the tendril-less phenotype of all 45 homozygous t/t F2 segregants, indicating that it is likely that tl and t are orthologous loci and that these closely related legumes share the same genetic mechanism for the regulation of tendril formation.
Tendrils are characteristic of the Papilionoid clade Fabeae, to which peas (Pisum spp and Lathyrus spp), vetches (Vicia spp), and lentils (Lens spp) belong ( Figure 6). Various wild species of chickpea (Cicer spp) are also tendrilled but are contained within a separate clade, the Cicereae ( Figure 6). Since all other Papilionoid legumes are untendrilled, this phylogeny suggests that tendrils arose either once (marked with an arrowhead), with at least two independent losses, or, twice (marked with arrows) independently. Garden pea Tl gene primers were able to amplify HDZIP PCR products from tendrilled sister taxa representatives, common vetch (Vicia sativa), narbon bean (Vicia narbonensis), and lentil (Lens culinaris). The sequences of these products comprise a single distinct Tl-like clade (Figure 3). Alignment of these sequences (see Supplemental Figure 2 online) identified the striking common feature of a Thr substitution for Ala in the presumed DNA binding domain (Figure 4), suggesting that the mechanism for tendril formation may be shared more widely within the Fabeae clade. Efforts to amplify a Tl ortholog from wild tendrilled chickpea species, Cicer anatolicum and Cicer canariensis, were unsuccessful. We were also unsuccessful in attempts to amplify Tl orthologs from nontendrilled taxa that are closely related, but outside the Fabeae clade, such as Medicago and clover (Trifolieae clade), and no clear ortholog is present in the Medicago genome sequence. We concluded that either these species do not have Tl orthologs, or the sequence is too diverged to detect. Our evidence supports the possibility that the evolution of tendrils in the Fabeae was dependent on acquisition of Tl, but it remains an open question whether the same acquisition event, an independent acquisition of Tl, or an independent event altogether, led to tendrils in the Cicereae (Figure 6).

DISCUSSION
All Papilionoid legumes, apart from taxa in the Fabeae and Cicereae clades, lack tendrils. We propose that the wild-type Tl allele (such as that carried by present-day tendrilled peas) arose as a semidominant mutation that suppressed leaflet blade development in a tendril-less progenitor legume, either in the Fabeae clade ;18 million years ago, or in the Fabeae-Cicereae clade ;33 million years ago (Lewis et al., 2005). We propose that this distinct allele survived the constraints of selection by providing a novel phenotype, permitting adaptation to a new and advantageous climbing growth habit.
Tendrils have arisen many times in flowering plants (Darwin, 1875;Bell, 1991). In grapevine, for example, the tendril is a gibberellin-inhibited inflorescence, as shown by a dwarf, tendrilless GIBBERELLIN-INSENSITIVE1 mutant (Boss and Thomas, 2002). In legumes, they are recently acquired specializations of the leaf, and here we have shown that pea tendril formation likely involves interaction between the Class I HDZIP gene Tl and the meristem identity gene Uni/ LFY.
The closest related gene to Tl in Arabidopsis, LMI1, was reported to have additional LFY-independent roles in leaf development, including promotion of leaf margin serrations and suppression of blade outgrowth from the petiole (Saddic et al., 2006). Thus, certain aspects of Tl and LMI1 function appear to be similar, such as blade suppression and transcriptional regulation by LFY. It is not clear, however, that regulation involves transcription initiation in both cases, and the suppression of leaf blade outgrowth probably involves different mechanisms in pea and Arabidopsis, since this process is largely Uni dependent in the former, but LFY independent in the latter. Other aspects of Tl, such as its semidominant inheritance and its tightly delimited expression domain in the leaf, are not characteristic of LMI1.
Tl encodes an unusual Class I HDZIP protein, and in all the tendrilled species, we have examined this protein has a Thr replacement for Ala at position 123 of the DNA binding helix 3 of the homeodomain. The closest relative identified in Medicago does not seem to be a good candidate for a Tl ortholog. This suggests one of three possibilities: (1) within the lineage leading to the Fabeae, after its divergence from the Trifoleae, a gene duplication occurred generating LMI1 paralogs, one of which we see as Tl but the other has not yet been found or is lost. (2) The Tl gene and the Medicago gene AC139525_29 are both LMI1 orthologs, but the Tl-like genes in tendrilled taxa relatives have Tree adapted from http://www.tolweb.org/IRLC_%28Inverted_Repeat-lacking_clade%29/60358. Arrows point to possible independent origins of tendrils in Fabeae and Cicereae clades. Arrowhead points to possible single origin of tendrils in Papilionoid legumes. evolved a new, presumably additional, function and structure. (3) The common ancestor of the tendrilled legumes and Medicago had duplicate LMI1 homologs. One paralog retained LMI1 structure and function while the other diverged as Tl. In the Medicago lineage, the Tl-like paralog was lost. This ancestral duplication may be preserved in some Cicer species. In all three of these possible histories, Tl has a novel sequence and function.
Comparison of the HDZIP regions of Tl-like proteins (see Supplemental Figure 2 online) showed that the legume Tl sequences have shorter leucine zippers that may optimize Tl for homodimerization or specific heterodimerization partners. The Thr replacement for Ala at position 123 of Tl in the DNA binding helix 3 of the homeodomain (Figure 4) is in the equivalent position to Met-54 of Antennapedia, a well-studied example of a DNAbound eukaryotic homeodomain protein (Fraenkel and Pabo, 1998). The hydrophobic side chain of Met-54 points into the major groove and makes contact with the DNA recognition motif. This suggests that Tl is either impaired significantly in DNA binding or is optimized for a different target from other Class I HDZIPs. Plant Class IV HDZIP proteins have a Thr-containing, rather than an Ala-containing, Helix 3 DNA binding motif, and Outer Cell Layer1 (Ingram et al., 1999) from maize (Zea mays) in particular has conserved adjacent Arg residues like Tl (see GenBank accession Y17898). This suggests that binding targets may exist for the Tl homeodomain in peas. If binding targets existed in a nontendrilled progenitor, then the origin of Tl may have established a new network of interactions that promoted tendril formation.
A Tl-instigated developmental pathway transforms lateral organ primordia from their default leaflet fate to a tendril fate where vascular bundles surround a central pith and adaxial polarity is suppressed (Tattersall et al., 2005). This transformation could occur if Tl interfered, either positively, to provide alternative binding sites to a heterodimeric partner protein as suggested above or in a dominant-negative manner to prevent DNA binding of a partner protein. Both models can explain the semidominant effect of Tl because partner proteins would either acquire abaxial polarizing activity or be prevented from exerting their adaxial polarizing activity due to competitive dimerization with Tl in lateral primordia. The resulting tendril would be interpreted as an abaxialized leaflet. Putative partners might include Class IV HDZIP proteins or pea orthologs of the Class III HDZIP proteins PHABULOSA, PHAVOLUTA, and REVOLUTA, which are known to play a role in the establishment of lateral organ adaxial identity in Arabidopsis (McConnell et al., 2001;Emery et al., 2003). Identification of Tl partners is essential to test these models and to gain further understanding of the tendrilled trait in crop legume species.

Plant Material
All garden pea (Pisum sativum), lentil (Lens culinaris), vetch (Vicia sativa), and chickpea (Cicer arietinum) lines were obtained from the John Innes Pisum Germplasm collection, apart from garden pea lines MC1a/1 from J.W. and M3 4092-1 from C.L. Sweet pea (Lathyrus odoratus) cv America was obtained from Chiltern Seeds (http://www.chilternseeds.co.uk/), and the t/t line was commercially available as Snoopea. Plants were grown in 16 h daylength in John Innes No. 1 compost with 30% extra grit. DNA was prepared from leaves according to Ellis et al. (1998).

Fingerprinting
Wild-type progenitor and tl mutant pairs were analyzed as described (Ellis et al., 1998). TaqI-digested genomic DNA ligated to a TaqI adaptor was used as a template for PCR with a 33 P-labeled primer matching the polypurine tract of the Pea Dispersed Repeat1 retrotransposon and a TaqI primer with two selective bases (AA). Reactions were loaded side by side and separated by gel electrophoresis on a 4.5% denaturing polyacrylamide gel. Dried gels were displayed using a Typhoon 8600 PhosphorImager (Molecular Dynamics, GE Healthcare UK).

Marker Screening
AFLP marker screening was performed as described (Vos et al., 1995), except that enzyme PstI was substituted for EcoRI. Genomic DNA (0.5 mg) was digested with PstI and MseI, and adapters were ligated. PstI adapter 1 and 2 together with MseI adapter 1 and 2 sequences are given in Supplemental Table 4 online. The ligation reaction was diluted 10-fold, and 2 mL was used in a 20-mL preamplification reaction using PstI and MseI primers (see Supplemental Table 4 online) with one selective base. Twenty cycles of preamplification PCR in 20-mL reactions, containing 15 ng of each primer, were performed according to Vos et al. (1995). Reactions were diluted 10-fold, and 1 mL was used as template in 10-mL AFLP reactions containing 15 ng 6-carboxyfluorescein-labeled PstI primer with two bases of selection and 15 ng MseI primer with three bases of selection. After 35 cycles of PCR, fragments were separated by capillary electrophoresis and displayed using GeneMapper v3.7 software (Applied Biosystems).

Cloning
An 892-bp cDNA clone, c21, was amplified from JI 2822 cDNA using primers TLHDcDNA59 and 34cDNA39 (see Supplemental Table 4 online) and cloned into a TopoTA vector (Invitrogen). This sequence has been deposited under accession number 1119567 in GenBank. A 2371-bp clone, 34g40, was obtained by PCR amplification from JI 2822 genomic DNA using primers TLHD59nest and TLHD39nest2 (see Supplemental  Table 4 online) using Phusion Hot Start High-Fidelity DNA Polymerase (Finnzymes). The product was A-tailed with Taq polymerase before insertion into a TopoTA vector (Invitrogen). This sequence has been deposited under accession number 1119577 in GenBank.

RT-PCR
Total RNA was extracted from shoot apices dissected from 3-week-old seedlings using a Qiagen RNeasy Plant Mini kit (Qiagen Sciences). DNA was removed from 80 mg of RNA samples by digestion with 30 Kunitz units DNaseI (GE Healthcare) in 100 mL 13 One-Phor-All buffer. Two micrograms of RNA was reverse transcribed with Moloney Murine Leukemia Virus reverse transcriptase (Invitrogen) from an oligo(T) primer. One microliter of first-strand cDNA was used in 20-mL PCR reactions containing 0.25 mM primers TLHDF1 and 34F1 for 40 cycles with an annealing temperature of 568C. Primers flanking introns 19, 20, and 21 of a pea ARGONAUTE1 cDNA clone were used in control reactions. Primers PsAGO1 and PsAGO2, flanking introns 19, 20, and 21 of a pea ARGO-NAUTE1 cDNA clone were used in control reactions, see Supplemental Table 4 online for primer sequences.
Quantitative PCR cDNA was synthesized from total RNA prepared as above, except that DNaseI (Ambion) treatment was performed according to the manufacturer's instructions. RNA was quantified by spectrophotometry (Nano-Drop Technologies), and 5 mg of RNA was reverse transcribed using SuperScript RNaseH reverse transcriptase (Invitrogen). Twenty nanograms of cDNA was used as template in 10-mL PCR reactions containing 13 SYBR Green JumpStart Taq ReadyMix (Sigma-Aldrich) and 0.2 mM forward and reverse primers for 40 cycles with an annealing temperature of 608C. Samples were amplified on a Chromo4 Real Time PCR machine (Bio-Rad) and analyzed with MJ Opticon monitor software V3.1. A 94-bp Tl amplicon spanning intron 2 was amplified with primers PsTlF and PsTlspanR. A 104-bp control amplicon corresponding to pea actin was amplified with primers PsActF and PsActR59 with primer PsActF sequence modified from GenBank accession number U81047 (see Supplemental Table 4 online for primer sequences).

RACE-PCR
RNA ligase-mediated 59 RACE was performed on 10 mg of shoot apex total RNA. A FirstChoice RLM-RACE kit (Ambion) was used according to the manufacturer's protocol. Nested PCR was performed using 59 RACE primers supplied and Tl gene-specific primers 34F6 and 34F6adj. A single amplified product was sequenced directly. RNA was reverse transcribed, and two rounds of 39 RACE PCR were performed according to the manufacturer's protocol using primers supplied and Tl gene-specific primers 34cDNA59 and 34PstextR1 (see Supplemental Table 4 online for primer sequences). Heterogeneous products were cloned into a TopoTA vector and sequenced.

RNA in Situ Hybridization
A 400-bp Tl cDNA 39 fragment was amplified using primers 34cDNA59 and 34cDNA39 (see Supplemental Table 4 online) and cloned into a TopoTA vector (Invitrogen) to generate clone 34/19 lacking the HDZIP region for use as an in situ hybridization probe. Digoxygenin-labeled antisense probes were generated from NotI-digested clone 34/19 transcribed with T3 RNA polymerase and an EcoRI-digested Uni cDNA clone transcribed with T7 RNA polymerase. Control sense probes were generated from PmeI-digested clone 34/19 transcribed with T7 RNA polymerase and a XhoI-digested Uni cDNA clone transcribed with T3 RNA polymerase. Sectioning, hybridization, and microscopy were performed as described previously (Hofer et al., 1997).

Phylogenetic Analysis
The deduced amino acid sequences of 35 Arabidopsis thaliana, Medicago truncatula, and other legume Class I HDZIP genes were aligned using ClustalW version 2.0.5 (see Supplemental Figure 2 online for alignment). Residues 132 to 237 of the HDZIP region were selected to estimate maximum likelihood trees using PROML in PHYLIP version 3.67 (http://evolution.genetics.washington.edu/phylip.html) with the Jones-Taylor-Thornton probability model of change between amino acids and the Class II HDZIP At4g37790 defined as an outgroup. Bootstrap support was obtained from 100 replicates for majority-rule consensus tree branches.

Accession Numbers
Sequence data from this article can be found in the GenBank/EMBL data libraries under the following accession numbers: EU938524 (P. sativum Tl mRNA, complete coding sequence), EU938525 (P. sativum Tl gene, complete coding sequence), EU938526 (L. odoratus Tl gene), EU938527 (L. culinaris Tl gene), EU938528 (Vicia narbonensis Tl gene), and EU938529 (V. sativa Tl gene). The pea ARGONAUTE1 sequence is available as accession number EF108450. The pea actin sequence corresponds to accession number U81047.

Supplemental Data
The following materials are available in the online version of this article.