Cleistogamous flowering in barley arises from the suppression of microRNA-guided HvAP2 mRNA cleavage

Abstract

The cleistogamous flower sheds its pollen before opening, forcing plants with this habit to be almost entirely autogamous. Cleistogamy also provides a means of escape from cereal head blight infection and minimizes pollen-mediated gene flow. The lodicule in cleistogamous barley is atrophied. We have isolated cleistogamy 1 (Cly1) by positional cloning and show that it encodes a transcription factor containing two AP2 domains and a putative microRNA miR172 targeting site, which is an ortholog of Arabidopsis thaliana AP2. The expression of Cly1 was concentrated within the lodicule primordia. We established a perfect association between a synonymous nucleotide substitution at the miR172 targeting site and cleistogamy. Cleavage of mRNA directed by miR172 was detectable only in a noncleistogamous background. We conclude that the miR172-derived down-regulation of Cly1 promotes the development of the lodicules, thereby ensuring noncleistogamy, although the single nucleotide change at the miR172 targeting site results in the failure of the lodicules to develop properly, producing the cleistogamous phenotype.

Angiosperms appear in the fossil record at least 130 Mya, and have since diverged with respect to their mode of pollination (1, 2). Many species exploit specific vectors, commonly specific insect species, to transfer their pollen, although others rely passively on wind or water to ensure pollen dispersal. The effect of cross-pollination is to maintain the level of genetic variation in the population. Self-pollination is thought to represent an adaptation to unreliable pollen vectors and is most frequent in short-lived annuals and invasive species. Self-pollination is common in many domesticated species, where reproductive success and genetic uniformity are highly advantageous.

The hermaphroditic angiosperm flower consists of the carpel, surrounded by stamens (2). In the grasses, a small swollen structure (the lodicule) develops at the base of the carpel and stamens. When the floret opens, the palea and the lemma (the pair of bracts which encase the floret) are forced apart by the swelling of the lodicule. The anther filaments then elongate rapidly and exert the anthers, at which point the pollen is released. The style is also exerted, generally at a slightly later time. Wheat and barley are predominantly self-pollinated, even though much of their pollen is released only after the anthers have been exerted; this is because the stigmas become receptive before anther exertion and are able to capture sufficient self pollen not to require fertilization by wind-borne nonself pollen. The exertion of the anthers is so pronounced in some wild barleys that their rate of outcrossing is higher than that of cultivated barley (3).

Natural variants of barley have been described in which the palea and lemma remain tightly closed throughout the period of pollen release (4). Such closed flowering is known as cleistogamy (5). The size of the lodicule in the cleistogamous flower is typically smaller than that in the noncleistogamous type (6). The cleistogamous state in barley is recessive, under the control of a single gene at the cleistogamy 1 (cly1) locus, which maps to the long arm of chromosome 2H (7, 8). The objective of the present study was to isolate Cly1 to uncover the molecular mechanisms responsible for flower opening and to investigate the contribution of this gene in the domestication of cultivated barley.

Results

Lodicule Size and Cleistogamy.

Lodicule size differed markedly between cleistogamous and noncleistogamous cultivars (Fig. 1 A–E). The first notable difference in their size became apparent at the white anther stage, where cell division was much more active in the noncleistogamous type (Fig. 1 F and G). By the green anther stage, the noncleistogamous type lodicules were at least twice the size of the cleistogamous type ones. The most pronounced difference was in their depth (Fig. S1). There was no significant size difference (in length, width, or depth) between the lodicules of the noncleistogamous parental cultivar and its F₁ hybrid with a cleistogamous type (Fig. S1), confirming the observation that noncleistogamy is a fully dominant trait.

Fig. 1.

The lodicules of cleistogamous and noncleistogamous barley. (A) The lodicule [lo] located at the base of the stamen [st] open the spikelet by pushing apart the lemma [le] and palea [pa]. (B) A noncleistogamous and (C) a cleistogamous barley spike at anthesis. The lodicules of (D) a noncleistogamous and (E) a cleistogamous barley. A section of the spikelet in (F) a noncleistogamous and (G) a cleistogamous barley. The carpel [cp] is surrounded by the other floral organs. (Scale bars: D and E, 800 μm; F and G, 200 μm.)

Cly1 Encodes a Transcription Factor Belonging to the euAP2 Family.

Cly1 was isolated via positional cloning (Fig. 2A, Fig. S2, and Dataset S1). The gene was mapped in a 7 kbp interval between M191K21A11 and P101AP25’. This DNA stretch included most of the coding region of Cly1, with no other genes represented (Fig. S3). The sequence of the copy present in AZ (noncleistogamous, Cly1.a) differed from that in KNG (cleistogamous, cly1.b) at two nucleotide positions. One of these lay within the P101AP25’ sequence, which recombined with cly1, suggesting that the second (P22AP23’), which did cosegregate with cly1, corresponds to the functional site for Cly1.

Fig. 2.

Positional cloning of Cly1. (A) A genetic map of the critical region of the long arm of barley chromosome 2H constructed from a Azumamugi (AZ) x Kanto Nakate Gold (KNG) F₂ population. Barley EST sequence BF623536 is homologous to that of rice AP2. BACs M191K21 and M601E22 harbor Cly1, which encodes a transcription factor with two AP2 domains and a putative miR172 targeting site. The AZ and KNG Cly1 sequences differ from one another by a single nucleotide site within the target site (responsible for the Cly1.a and cly1.b alleles) and by a second nucleotide located upstream of the coding region (P101AP25’). (B) Sequence comparisons between alleles of the putative miR172 target site indicate that it is variation at the second and/or third variable positions (red stars), rather than at the first position (blue star), which is responsible for cleistogamy. (C) Lodicule size at the yellow anther stage and at anthesis in a 274 accession germplasm set. Noncleistogamous cultivars are marked by blue spots, and cleistogamous ones by red spots.

The full Cly1 gene extends 2691 bp from its start to its stop codon, has a GC content of 60.8% and is divided into 10 exons interspersed with 9 introns (Fig. 2A). The length of the coding sequence is 1464 bp, and it is flanked by a 5′ UTR (480 bp) and 3′ UTR (346–368 bp). Cly1 encodes a predicted 487 residue polypeptide with two AP2 domains (9, 10), the first sited between residues 112 and 178 and the second between 203 and 270 (Fig. S4). Exon 10 includes a putative miR172 target sequence (Fig. 2A), which is also present in a number of AP2 genes (11, 12). Two auxin-response elements (13) were present, one approximately 3 kbp, and the other approximately 2 kbp upstream of the start codon. The sequence resembles those of Arabidopsis thaliana AP2 (AT4G36920.1) and TOE3 (AT5G67180.1), and belongs to the euAP2 clade (Fig. S5). The transcript was highly homologous to predicted transcripts of the rice AP2-like gene Os04g0649100. The single nucleotide polymorphism at P22AP23’, which lies within a putative miR172 target sequence in Exon 10, was synonymous (Fig. S4).

The miR172 Site Sequence Polymorphism Is Associated with Lodicule Size and Cleistogamy.

Sequence variation among the set of 274 barley accessions was observed at three base positions within the miR172 target sequence (Fig. 2B and Dataset S2). None of these variants implied any variation in peptide sequence, as all are sited in the third base position of their respective codon. The first variant (…C(A/C)G…) was uncorrelated with cleistogamy, but both the second and third (…(A/G)TCATC(A/C)…) were correlated. The observation that all of the noncleistogamous types were of haplotype ATCATCA at the latter site indicated that this haplotype is necessary for the determination of noncleistogamy (Fig. 2B). The haplotype was also present in the wild barley line. The haplotype of the cleistogamous types was either GTCATCA or ATCATCC (Fig. 2B). The initial G was present in seven cleistogamous accessions in addition to KNG, indicating that this single nucleotide change was sufficient for the change from noncleistogamy to cleistogamy. This notion was supported by the genetic delimitation of the cly1 locus and the miR172-guided mRNA cleavage analysis (Fig. 2A). A second cly1 allele carries the alternative haplotype ATCATCC (where a C replaces an A in comparison with the noncleistogamous type). This haplotype was present in four cleistogamous accessions (Fig. 2B).

With respect to lodicule size, the set of 274 barley accessions fell into two distinct classes (Fig. 2C). The lodicule of the noncleistogamous types was at its largest by swelling at anthesis, at which point it was 2.3 fold deeper and 1.4 fold wider than at the yellow anther stage, whereas that of the cleistogamous types remained small. The swelling of the lodicule is primarily responsible for pushing open the lemma and thereby opening the floret. Entry SV235 was the single exception to the otherwise perfect relationship between lodicule size and cleistogamy. This line appeared to be cleistogamous in the field but its lodicule size was intermediate and was responsive to 2,4-D treatment to swell to be noncleistogamous. The SV235 coding sequence was identical to that of the cleistogamous cultivar KNG, suggesting that in SV235 either Cly1 expression is reduced or a different pathway is involved in the control of lodicule development.

miR172 Guided Cleavage of the Cly1 Transcript.

In situ RNA hybridization revealed that the Cly1 transcript was well represented in the lodicule up to the stamen primordium stage (Fig. 3 A and B), with signal detectable in both the noncleistogamous and cleistogamous types (Fig. 3 C and D). Although quantitative PCR analysis showed that carriers of both the Cly1.a (AZ, noncleistogamous) and the cly1.b (KNG, cleistogamous) allele shared essentially the same pattern of gene expression, from the awn primordium stage to the yellow anther stage, transcript abundance was somewhat higher in KNG than in AZ (Fig. 3E).

Fig. 3.

Expression of Cly1. (A) The 3′ end of Cly1, including Exon 10 which carries the putative miR172 target site (red bar). The two regions targeted by RT-PCR shown by arrows. The zigzag line indicates the 5′ RACE RNA oligo-adapter ligated to cleaved mRNA. The probe used for in situ hybridization is also shown. (B) Cly1 expression in an immature spike at the awn primordium stage of the noncleistogamous cultivar AZ, as detected by in situ hybridization with an anti-sense (Left) and sense (Right) Cly1 3′ UTR transcript. (C) A higher magnification of the presentation shown in (B). [lo] lodicule, [st] stamen. (D) Cly1 expression in the cleistogamous cultivar KNG. (E) Cly1 expression in an immature spike during development. The constitutively expressed actin gene was used as a control. Dark gray bars represent AZ and light gray bars KNG. The numbers below each bar refer to the developmental stage assayed (1, lemma primordium stage; 2, stamen primordium stage; 3, awn primordium stage; 4, white anther stage; 5, green anther stage; 6. yellow anther stage; 7, spike at anthesis). (F) Modified 5′ RACE ligations from (A) were amplified by nested PCR. (G) Cleavage at the miR172 site within Cly1. The 5′ termini of miR172-guided mRNA cleavage were determined by cloning and sequencing of the amplicon generated in (F). A rice miR172 sequence was aligned with the Cly1 mRNA sequences. Vertical arrows indicate the inferred 5′ termini of miR172-guided cleavage, and the number above each arrow indicates the proportion of clones showing these sites. (Scale bars: B, 500 μm; C and D, 250 μm.).

A modified 5′ RACE experiment was carried out to investigate miR172-guided mRNA cleavage of Cly1. The 5′ ends of the cleaved transcripts could be ligated to the GeneRacer RNA oligomer. The nested PCR targeting the miR172 site produced an approximately 400 bp amplicon from Cly1.a carrier but not from cly1.b carrier (Fig. 3F). This was taken to indicate that miR172-guided cleavage is occurring in the former plants but not in the latter (Fig. 3F). The majority (32/48) of sequences cloned from the Cly1.a RACE amplicon revealed a cleavage site within the miR172 target site, with the bulk of the cleavage taking place between the A and U nucleotides (Fig. 3G). However, most of the cly1.b amplicon clones represented ribosomal RNA or mRNA of genes unrelated to Cly1 with just two out of 48 clones containing sequence of Cly1 (Fig. 3G). Thus most of the cly1.b RACE products appear to represent random mRNA breakage. Overall, at the mRNA level, it is clear that lodicule development and hence cleistogamy arises due to a difference in miR172 guided cleavage of Cly1 transcripts.

The Origin of Cleistogamous Barley.

The phylogenetic analysis suggested two distinct lineages of cleistogamy (Fig. 4). The KNG haplotype (cly1.b) differed from the wild-type AZ sequence by only two nucleotide substitutions, one at position 3084 within the miR172 target sequence and the other at position 626 in the first exon (Fig. S6). The likelihood is therefore that the KNG haplotype was a direct descendant of the wild-type AZ haplotype (Fig. 4). The AZ haplotype may be descended from the SV012 haplotype because the G present at position 2522 is shared with both noncleistogamous cultivars and wild barley (Fig. S6). The origin of the second cleistogamous haplotype (cly1.c) is uncertain. Clearly, however, the two cleistogamous haplotypes appear to represent independent mutational events (Fig. 4). The 3kb Cly1 coding sequence was invariant across both cleistogamous haplotyes, which indicates that the origin of cleistogamous barley was probably relatively recent.

Fig. 4.

A phylogenetic analysis of the Cly1 sequence suggests that the KNG (cleistogamous) allele cly1.b evolved from the AZ (noncleistogamous) Cly1.a allele. The immediate ancestor of SV230 type cleistogamous cultivars remains unknown. Hordeum vulgare ssp. spontaneum accession OUH602 was used as the outgroup to construct this phylogeny.

Discussion

Cly1 Is an Ortholog of A. thaliana AP2

It has been suggested that the monocotyledonous lodicule is structurally homologous with the dicotyledonous petal (2). The lodicule is small compared to the other floral organs, and in cleistogamous barley, its size is even more reduced. The internal structure of the noncleistogamous type lodicule is better organized than that of the cleistogamous type (Fig. 1 E and F). In particular, the former contains the vascular tissue which permits the rapid water uptake required for their swelling at anthesis. The morphological difference in the lodicules of the two flowering types reflects altered organ development. The ABCE model maintains that gene networks play a major role in establishing the identity of the floral organs. Specifically, the class A genes AP1 and AP2 overlap in function with those of the class B genes AP3 and PISTILLATA (PI) to determine the identity of petals and stamens, although an antagonistic interaction takes place between the class A and the class C gene AGAMOUS (AG) (14). It has been argued that AP2 function is restricted to dicotyledonous plants on the grounds that the monocotyledonous flower lacks sepals and petals (14). However, Cly1 shares the characteristics of the euAP2 lineage with its two AP2 domains and an miR172 target sequence downstream of the coding sequence (15). A number of “AP2-like” genes have been identified in rice and maize, including SUPERNUMERARY BRACT (SNB) (16), indeterminate spikelet1 (IDS1), and sister of indeterminate spikelet1 (sid1) (17–19). Barley HvAP2L1 (20) is also a member of this family (Fig. S5). These AP2-like genes are phylogenetically related to one another but are distinct from A. thaliana AP2 and barley Cly1. They may therefore represent duplicated paralogues, the function of which is not equivalent to that of AP2 orthologs. Phylogenetically, Cly1 is closely related to rice Os04g0649100 (Fig. S5) and to date remains the sole rice AP2-like gene which is closely related to AP2. The combination of genetic, phylogenetic, and gene expression data strongly supports the notion that Cly1 (HvAP2) is an ortholog of A. thaliana AP2.

Cleistogamy Is Determined by the Action of MicroRNA.

We have demonstrated that cleistogamy in barley can be determined by a single nucleotide substitution at the miR172 target sequence in Cly1, which suppressed the miRNA-guided cleavage of cly1.b mRNA. It has been proposed that AP2 genes containing miR172 target sequences are primarily regulated by translational repression (11, 12) or by a combination of miR172-directed mRNA cleavage and translational repression (21). This represents an ancient regulatory system which must have been elaborated before the monocot/dicot split (22). The miR172-directed mRNA cleavage system has evidently been retained not just in barley but probably in the grasses more generally, because the putative miR172 target site is present also in a number of ESTs, extracted from a range of grass species, homologous with the Cly1 transcript sequence (Fig. S7). The cleavage of the Cly1.a transcript at this site is analogous to that which affects the transcripts of A. thaliana AP2 and maize gl15 (23, 24). The cleavage site lies 3 bp downstream of the variable base that distinguishes the AZ and KNG sequences (Fig. 3G). Base mismatch is responsible for the loss of recognition of target sequences, such as in the Ts6 allele of maize ids1, where a single base mismatch at the 3′ end of miR172 produces indeterminacy in the spikelet meristem (18). The position 2–12th nucleotide of miR172 target site are specifically sensitive to mismatches, because the region is responsible for pairing with the 5′ portion of the miRNA (21, 25). The cly1.c allele reflects a point mutation at the eighth nucleotide of the miRNA172 sequence (Fig. 2B), therefore cly1.c probably affect cleavage site recognition in the same way.

miR172-directed mRNA cleavage only occurred in the noncleistogamous type, and the significant decrease in transcript abundance detected indicated that a portion of the transcript had been cleaved in the noncleistogamous type (Fig. 3 E and G). The question that arises is why the Cly1.a allele (encoding noncleistogamy) is dominant over the cly1.b allele (cleistogamy), because the expectation is that the level of noncleaved HvAP2 mRNA present in the Cly1.acly1.b heterozygote should be sufficient to generate a sufficient level of HvAP2 protein to suppress lodicule development. In practice, however, the heterozygote is noncleistogamous (Fig. S1). One possible model to account for this unexpected dominance relationship is based on the miRNA-guided formation of transacting small RNAs (ta-siRNAs) derived from a single transcript (26). Following miR172-guided cleavage of HvAP2 mRNA (Fig. 3G), the fragmented mRNA becomes available as a precursor for the formation of double stranded siRNA, which in turn triggers the degradation of intact HvAP2 mRNA molecules and results in the negative regulation of Cly1.a. It may be of some significance that the siRNA sequences derived from AP2 transcription in A. thaliana, rice and barley share certain repetitive elements (Fig. S8). If the siRNAs generated from the Cly1.a allele are able to interact with both Cly1.a and cly1.b transcripts, then the Cly1.acly1.b heterozygote will become noncleistogamous. An alternative model could be built on the basis that there may be a threshold quantity (or proportion) of the cly1.b translation product above which lodicule development is suppressed. If a proportion of the transcript is cleaved (Fig. 3E), the amount (or proportion) of HvAP2 mRNA, and in consequence HvAP2 product, falls below this threshold level in the Cly1.acly1.b heterozygote, resulting in the expression of noncleistogamy. Note, however, that lodicule size is a quantitative trait (Fig. 2C), so it is not clear how a quantitative trait could be compatible with such a threshold. Testing of these models will require the quantification of HvAP2 translation products in the lodicule primordia.

In A. thaliana, the miR172-mediated repression of AP2 acts through an AGAMOUS (AG)-dependent pathway to promote floral determinacy and to define the inner boundary of the B gene expression domain (27). If the same processes apply in barley, the miR172-mediated down-regulation of HvAP2 would be predicted to reduce the expression of class B genes in noncleistogamous barley, leading to the development of normal lodicules (Fig. 1F). The de-repression of HvAP2 function in cleistogamous barley caused by the mutation at the miR172-target site would then permit the recovery of class B gene expression to allow the formation of rudimentary lodicules (Fig. 1G). In rice, loss-of-function mutants of class B genes (such as the PI homologs OsMADS2 and OsMADS4, and the AP3 homolog SUPERWOMAN1) produce a homeotic transformation of the lodicule into a bract-like organ (28–30). The abolition in maize of Zmm16 (PI) and SILKY1 (AP3) activity has a similar phenotypic effect (31, 32). Regarding A class genes, the snb mutant, a rice AP2-like gene, allow for lodicule formation and its boundary specification (16). Rice class C genes (OsMADS3 and OsMADS58) in rice share a similar function with AtAG (33). Thus, the possibility is that the formation of lodicules in the cereals is homologous to the formation of petals in A. thaliana. We hypothesize that the de-repression of HvAP2 in cleistogamous lines up-regulates B-class genes, presumably via the suppression of HvAG. This activity would deliver an imperfect homeotic transformation, leading to the suppression of cell division and the loss of conductive tissue, which together prevent the swelling of lodicules characteristic of the cleistogamous trait.

Evolution of Cleistogamy.

Noncleistogamy is an ancestral character in barley because wild barleys all show this phenotype. The cleistogamous type arises from a point mutation, which has apparently originated at least twice, given that two independent mutations produce the phenotype. An analogous history has been deduced regarding the evolution of the six-rowed barley spike, which appears to have arisen independently three times (34). One of the two cleistogamous types probably originated in Northern Europe, because it is common among British barley cultivars, whereas the second probably originated in southern Europe or in the Western Mediterranean region (Dataset S2). Both mutations are likely to have been relatively recent events, because there is no further sequence polymorphism or recombination within either of the cleistogamous alleles (Fig. S6), and the geographic distribution of the cleistogamous barleys is well separated from the site of the crop's domestication (35). In contrast, sequence variation has been observed within the noncoding region of the six-rowed spike gene (vrs1.a), which first arose as a mutation from the wild-type two-rowed type some 8,000 years B.P. in the Fertile Crescent (34). Cleistogamy is common among the grasses, but quite rare among the Triticeae (5). In polyploid species—in particular bread and durum wheat—a recessive mutation affecting one of the Cly1 homeologues is unlikely to deliver a cleistogamous phenotype because it will be masked by the dominant allele present at the remaining homeologue(s). The particular advantage of barley is that it is diploid, thus allowing for the ready expression of a recessive mutant phenotype.

A possible basis for the deliberate selection of cleistogamous barley is its association with spike density, given that the spikes of cly1 carriers are of the erectum form (36). The control of rachis internode length by cly1 may be via linkage between cly1 and a QTL for spike internode length (36) or a pleiotrophic effect by the AP2 transcription factor on the elongation of rachis internode. The cultivar Golden Melon forms erectum type spikes and was a selection from the nutant cv. Chevallier made in England during the 1800s. A second erectum type (cv. Goldthorpe) was selected from a field of cv. Chevallier in England in 1889 (37). It is unclear whether either or both these erectum types were de novo mutations of cv. Chevallier or whether instead they arose from preexisting seed admixture, but in either case, the presence of an erectum spike in nutant spike barley was novel enough to have been remarked upon at the time.

Materials and Methods

Assessment of Flowering Phenotype.

A set of 274 barley cultivars, chosen to represent the global distribution of barley, was obtained from the Research Institute for Bioresources, Okayama University, Kurashiki, Japan (Dataset S2). Additional cultivars, along with some noncultivated lines are maintained in our own laboratory. Plants were sown in the field at Tsukuba, and a sample of at least three spikes, including the flag leaf and peduncle, was taken from each accession at the yellow anther stage immediately before anthesis. These were exposed to 100 mg/l 2,4-D at room temperature for two days (6), after which spikelets in the central portion of each spike were excised, the lemmas removed, and the lodicule photographed. The digital images were used to measure the width and height and depth of the lodicule, using the image analysis program Makijaku v1.1, provided by Dr. H. Iwata.

Genetic Mapping.

F₂ populations were bred from the crosses Kanto Nakate Gold (KNG, cly1.b) x Azumamugi (AZ, Cly1.a) and OUH602 (Cly1.a) x KNG (Fig. S2), and phenotyped for flowering habit in the F₃ generation. Lodicule size was maximized by applying the spike culture method (6). De novo markers for fine genetic mapping were obtained by exploiting synteny between rice and barley (Dataset S1 and SI Methods) and developed from barley BAC sequence (see below).

BAC Contigs, Sequencing, and Annotation.

A BAC library constructed from cv. Morex (38) was screened by PCR following (39). BAC contig assembly was achieved by HindIII fingerprinting. Individual BACs were shotgun sequenced using standard methodology (40). Once assembled, the contigs were correctly ordered and oriented by aligning vector cloning sites and bridge subclone end-sequences. Intercontig gaps were filled by full sequencing of the bridge subclones, using either primer walking or GPS-1 transposon sequencing (New England Biolabs). Method for sequence annotation was described in SI Methods.

In Situ RNA Hybridization.

A 280 bp fragment from the 3′ UTR of Cly1 was PCR amplified with primers BF623536U514U794 and BF623536U514M060H23L1053 (SI Methods). In vitro transcription and in situ RNA hybridization followed methods described previously (34).

Quantitative PCR.

The level of Cly1 expression was assessed by real-time PCR, based on a TaqMan (Applied Biosystems) assay. The reaction mix contained 900nM of each gene-specific primer along with 250nM of Taqman probe. Quantification was performed using the comparative Ct (ΔΔCt) method, and relevant data are presented as the fold difference in gene expression normalized against the expression of the constitutively expressed actin gene, and related to the expression of a calibrator sample (KNG at the lemma primordium stage).

Mapping the MiRNA-Guided Cleavage Site.

A RNA-ligase mediated 5′ RACE (23), using the GeneRacer Kit (Invitrogen) was applied to total RNA extracted from whole spikes at the awn primordium stage. The dephosphorylation and decapping steps were omitted, so that only the 5′ends of the truncated transcripts could be ligated to the GeneRacer RNA oligomer. A nested PCR employed a primer targeting the GeneRacer RNA oligomer, initially in combination with a reverse gene specific primer (BF623536U514M060H23L1053), and subsequently with an inner gene specific primer (BF623536U514M060H23L1031) (SI Methods). PCR products were separated by agarose gel electrophoresis and cloned into the TA vector (Invitrogen). Randomly selected clones without any prior size selection were chosen for DNA sequencing.

Phylogenetic Analysis.

Amino acid alignments were performed using ClustalW2 (http://www.ebi.ac.uk/Tools/clustalw2/). A phylogenetic tree was constructed by the neighbor-joining method using PAUP4.0b10 (41).