Downloaded by guest on October 2, 2021 n nmlpliaos(–2.Atog h vlto fnovel of evolution the Although differ- to (9–12). adapt pollinators to animal populations allowed ent orientation, such features, and floral color other as and morphology the spur to in modifications petals of genus outgrowths the as legia fol- is evolved radiation spurs spurs a nectar nectar Floral of of example evolution textbook the lowing A (5). different morphologies with or plants spur pollinator between altogether—resulting isolation a pollinator reproductive of animal increased the in different body through a the by on specialization—either visitation pollen morphol- pollinator of spur placement to in differential lead changes can because rates ogy hypothesized are speciation Spurs increase (4–6). them to lack in than that speciose have and, lineages more are sister plants, spurs spurs their nectar flowering with Such lineages in cases, pollinators. all times pro- nearly animal many that for independently tissue, reward evolved are floral nectar spurs by Nectar a (4). formed duce innovation generally key a structures, of tubular example classic considered a are be spurs nectar to Floral (1–3). increasing clades particu- explain species-rich by to larly diversification used been of often have process opportunities, traits ecological the is, that promote innovations, to key of thought evolution The others. to relative T Aquilegia initial morphologies spur the genus. diverse the in of across evolution role subsequent the its in determining role tial part, a the of is evolution it which genetic the of of pathway elucidating further The including to nectaries. exploration, appears scientific of also development and of subsequent development identification the spur for of necessary I) be (phase the phase in early proliferation cell factor. regulating transcription in zinc-finger C2H2 a xrsinaaye,adfntoa sast dniyagn cru- gene development, a spur and nectar identify gene for to mapping, recent cial assays genetic functional relatively of and combination a analyses, a expression experienced use has We are radiation. that rapid genus spurs the lineage columbine Nectar about a the trait. known laceae), of this is feature of defining little basis a spurs, developmental nectar traditional and have the genetic of taxa none model As plant specialization. reproductive promote pollinator to multiple via ability in isolation their to rates due diversification of lineages evolution increased angiosperm independent with the prime associated with a innovation, spurs are key evolution- spurs a to nectar of interest floral example plants, of origin flowering been In long the biologists. has ary in innovations involved key lead these mechanisms the can of developmental understanding ways Therefore, and new rates. genetic in diversification environment increased allow their that 2020) to wings, 11, exploit April or review eyes to for as (received such organisms 2020 features, 30, novel July of approved evolution and The CA, Stanford, University, Stanford Bergmann, C. Dominique by Edited 02318 MA Cambridge, University, Harvard Biology, a key Ballerini a S. in Evangeline of spurs, development nectar the floral in innovation, role central a plays transcription factor, zinc-finger C2H2 a encoding POPOVICH, www.pnas.org/cgi/doi/10.1073/pnas.2006912117 clg,EouinadMrn ilg eatet nvriyo aiona at abr,C 30;and 93106; CA Barbara, Santa California, of University Department, Biology Marine and Evolution Ecology, ihsm iegsehbtn nrae ae fspeciation of rates increased exhibiting life, of lineages tree the some across with varies diversification species of pace he ancestor | ea development petal ilo o7mlinyasao() fe which after (8), ago years million 7 to million ∼5 POP Aquilegia pn pnmru vne o continued for avenues numerous up opens a,1,2 | etrsu,adeaiigispoten- its examining and spur, nectar etrspur nectar aMin Ya , POPOVICH Aquilegia | b e innovation key POP ol .Edwards B. Molly , (POP Aquilegia ly eta role central a plays ea uigthe during petal ,wihencodes which ), Aquilegia | mitosis (Ranuncu- Aqui- (7). b ln .Kramer M. Elena , h pretr hs I nwihmttcatvt essand ceases Cell activity (16). mitotic proceeds, anisotropically which development elongate petal in cells spur As II, differentiating 17). is phase the (16, spur enters the length spur until in the continue mm devel- they the 10 initially where to to are localized cup, that become spur divisions petal oping develop- cell the two I, throughout into phase dispersed down During broken phases. be mental can the which of nectary development, studies a its Previous with A). S1 adjacent Fig. outgrowth forms Appendix, tubular legia which (SI a tip spur, as the the point at and attachment petal, the the to of end distal The the features. key of legia identification the facilitating simple, will turn, in which, origin. innovation, its into key develop- insight a the provide of dissect organ, basis to genetic floral begin and to single mental opportunity a unique to a modifications provides by it formed is and recently itr.Gvnta the that evolutionary in Given is com- deep involve history. arose evolution traits that these their mechanisms of developmental many to plex genetic because part, led the in that contribut- difficult, discovering often as mechanisms (13–15), or recognized developmental bats, diversification is and and lineage fish, birds, cichlid to insects, of ing of jaws flight pharyngeal powered the the as such traits, doi:10.1073/pnas.2006912117/-/DCSupplemental at online information supporting contains article This 2 1 the under Published Submission.y Direct PNAS a is article This interest.y paper. y competing the no wrote declare E.M.K. authors and The E.S.B. and S.A.H. and data; M.B.E., analyzed Y.M., E.S.B. E.S.B., research; research; performed designed S.A.H. and E.S.B. contributions: Author rsn drs:Dprmn fBooia cecs aionaSaeUniversity, State California Sciences, Biological of 95819.y CA Department Sacramento, address: Present owo orsodnemyb drse.Eal alrn@sseuo hodges@ or [email protected] Email: addressed. be may lifesci.ucsb.edu. correspondence whom To nla eeomn,in lineage the development, of leaf function ancestral in an suggests taxa plant other columbine the in spurs, Here nectar genus floral interest. innovation, particular key a of of ment is gene, a innovations identify key we in of involved origin Under- changes of innovations. developmental the rates key and them increased genetic call the to we standing in lead lineage, environment a features their in such speciation with When interact ways. to novel them allow that tures fea- evolved have organisms history, evolutionary Throughout Significance opooecl iiini eas e ellrse nthe in step cellular key genus. the a in petals, spurs nectar in of division development cell promote to h eeomn ftesurdptlin petal spurred the of development The ea scmoe ftocmoet,telmnrbaeat blade laminar the components, two of composed is petal etrsu dnie e ellrpoessivle in involved processes cellular key identified spur nectar hl h ucinof function the While Aquilegia. y Aquilegia NSlicense.y PNAS b n ct .Hodges A. Scott and , hti rca otedevelop- the to crucial is that POPOVICH, Aquilegia b eateto raimcadEvolutionary and Organismic of Department Aquilegia, etrsu vle relatively evolved spur nectar . y POPOVICH https://www.pnas.org/lookup/suppl/ POPOVICH NSLts Articles Latest PNAS a,1 Aquilegia lofunctions also rhlg in orthologs srelatively is mm ∼7 | Aqui- Aqui- f9 of 1

PLANT BIOLOGY anisotropy is a major contributor to final spur length, and including plants whose petals bore a small pocket but no tubular variation in the degree of anisotropy largely explains the dif- spur or nectary, as well as plants with clearly developed tubular ferences in spur length between species (16). However, little spurs and nectaries (Fig. 1B). In order to evaluate the previ- is known about the genetic control of persistent mitosis in the ous observations of Prazmo (22), we measured the length of the spur cup during phase I of development, or the transition to petal pocket or spur from the attachment point to the apex or anisotropic cell expansion during phase II of development. An nectary, respectively, on the proximal side of the petals for 92 initial RNA sequencing (RNAseq) experiment identified genes F2 individuals. Plotting these length measurements produced a that are strongly differentially expressed (DE) between the bimodal histogram confirming that phenotypes could be differ- developing spur and blade tissue during phase I of petal develop- entiated into spurred and spurless individuals (SI Appendix, Fig. ment in Aquilegia coerulea ‘Origami,’ including the TEOSINTE S1C); 334 F2 individuals were phenotyped using this metric to BRANCHED/CYCLOIDEA/PCF gene AqTCP4, which acts to guide binning of individuals into spurred and spurless classes. restrain cell proliferation in the distal compartment of the spur, Phenotype counts were consistent with expectations for a reces- and the AqSTYLISH genes, a small family of transcription fac- sive allele at a single locus causing spur loss (SI Appendix, Fig. tors (TFs) that were subsequently found to be responsible for S1D; n = 334, χ2 = 0.44, degree of freedom = 1, P = 0.51). At nectary development (17, 18). While the set of DE genes from the same time, these data indicate that multiple genes contribute that study indicates that cell proliferation pathways are involved, to spur length variation downstream of this single locus respon- none of the candidates functionally explored so far have revealed sible for spur presence/absence. In this study, we have chosen to potential master regulators of spur development. focus first on the quantitative trait locus (QTL) for spur loss. One species of Aquilegia native to montane regions of central Using whole-genome skim sequencing, 286 individuals were China, Aquilegia ecalcarata, is the only known species of Aquile- genotyped, and 276 of those individuals were phenotyped and gia with petals that do not produce spurs or nectaries (19). While used to conduct QTL mapping of spur loss (phenotypes were once thought to have diverged from the lineage prior to the evo- not recorded for 10 sequenced individuals). As predicted by the lution of nectar spurs, phylogenetic analyses clearly indicate that phenotype counts, a single major locus associated with spur loss the spurless phenotype of A. ecalcarata represents a secondary was identified on chromosome 3 (Fig. 1C). An apparent sec- loss (8). Nonetheless, understanding spur loss in A. ecalcarata ond locus of moderate effect was detected on chromosome 2, has been suggested as key in helping to unravel the genetic and as well as several loci of small effect on chromosomes 5, 6, and 7. developmental basis of spur development (4). Although the A. Closer examination of individual genotypes at the loci on chro- ecalcarata petal no longer produces a nectar spur, other aspects mosomes 2 and 3 revealed that the association between genotype of its development remain intact, and the morphology of the A. and phenotype on chromosome 2 is likely caused by a deleterious ecalcarata petal is quite similar to the primitively spurless petals interaction between A. ecalcarata and A. sibirica alleles at these of the Aquilegia sister genus, Semiaquilegia (20). We previously loci (SI Appendix, Table S1). For example, individuals with an A. used A. ecalcarata in an effort to further narrow in on key compo- sibirica allele (homozygous SS or heterozygous ES) at the chro- nents of the genetic network involved in early spur development mosome 3 locus were never homozygous for the A. ecalcarata using comparative gene expression. A set of genes consistently allele (EE) at the chromosome 2 locus, and only one individ- DE between the developing petals of spurless A. ecalcarata ual homozygous for A. sibirica (SS) at the chromosome 3 locus and three phylogenetically divergent spurred taxa was identified had an A. ecalcarata allele (ES) at the chromosome 2 locus. (21). Cross-referencing these genes with those identified as DE Indeed, treating the genotype at the chromosome 3 QTL as a between blade and spur tissue in A. coerulea ‘Origami’ revealed covariate completely eliminates the association between geno- a list of only 35 genes consistently DE between petal samples type and phenotype at the chromosome 2 locus; however, when containing spur tissue and petal samples lacking spur tissue (21). treating the genotype on chromosome 2 as a covariate, the sig- Interestingly, a genetic cross conducted by W. Prazmo (22) in the nificant QTL on chromosome 3 remains (SI Appendix, Fig. S2). 1960s between A. ecalcarata and the spurred species A. vulgaris Therefore, in order to narrow in on the major genetic element showed that ∼25% of the F2 individuals did not produce a spur, controlling spur development in this cross, we focus on the locus while the ∼75% of the spurred F2 individuals showed continuous with the highest LOD score, the locus on chromosome 3, which variation in spur length. These findings suggest that spur loss in we refer to as POPOVICH (POP). A. ecalcarata is caused by a single locus that is epistatic to other Genotype at the POP QTL is highly predictive of phenotype— loci affecting morphological traits such as spur length and curva- all SS and SE individuals at the QTL produce spurs, whereas 71 ture (22). This genetic cross, combined with the developmental of 79 EE individuals lack nectar spurs (SI Appendix, Table S2). similarities between the spurless Semiaquilegia and A. ecalcarata Using recombination events that resulted in a shift in genotype petals, has been interpreted as an indication that there may be a between ES and EE that were associated with phenotype, we single genetic factor that plays a crucial role in early spur devel- narrowed our QTL region of interest to a <2-cM region in the opment (20). Identifying this element would provide a critical middle of chromosome 3. This region has a low recombination component of the genetic network involved in the development rate and encompasses ∼20 Mbp of sequence containing ∼1,100 of this key innovation, an important step in understanding how annotated genes. Although the spur loss phenotype could be a this novel feature evolved. Therefore, we made a similar genetic result of mutations affecting the coding sequence of one or more cross to that of Prazmo (22) and made use of tools available in proteins in the QTL, we first determined whether any of the 35 the modern genomics era to identify a gene encoding a C2H2 genes previously identified as consistently DE between Aquile- zinc-finger TF that is crucial to regulating mitosis during phase I gia phase I petal samples containing spur tissue (spur +) versus in the development of the Aquilegia nectar spur. those lacking spur tissue (spur –; ref. 21) are in the QTL for spur loss. Only one of the expression candidates, Aqcoe3G231100, Results occurs in the POP QTL. This candidate is expressed at much Quantitative Trait Locus Mapping and RNAseq Identify a Candidate higher levels in the spur tissue relative to the blade tissue Gene for Spur Development. In order to identify the genetic region of A. coerulea ‘Origami’ petals and is consistently expressed controlling spur loss in A. ecalcarata, we crossed a spurred at higher levels in the petals of the spurred taxa, A. sibirica, species, Aquilegia sibirica, to A. ecalcarata (Fig. 1A). A single Aquilegia formosa, and Aquilegia chrysantha, than in spurless A. individual from the F1 generation, whose petals developed nectar ecalcarata petals, where it is essentially only expressed at back- spurs (SI Appendix, Fig. S1B), was selfed to create an F2 genera- ground levels, during phase I of development (Fig. 1D; data from tion. Various petal phenotypes segregated in the F2 generation, refs. 17 and 21).

2 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.2006912117 Ballerini et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 alrn tal. et Ballerini cotyledons functional and leaves no in in ortholog is the expressed and is There (24), it (23). although mitosis AT4G17810, the of for involved maintenance information genes either or the differentiation repressing in cellular by for processes differentiation responsible developmental cellular genes plant and various growth in homologs mitotic between related transition and motifs, from repressor LxLxL-type contain devel- transcriptional genes floral These in EAR ). S3 (23), roles C2H2 Fig. important clade Appendix , of with (SI C1-1i subclade genes opment the of The number of (23). a part includes plants Aqcoe3G231100, of in containing one families TFs TF, TF zinc-finger Factor. C2H2 largest Transcription a the as Zinc-Finger annotated C2H2 is Aqcoe3G231100 a Encodes Aqcoe3G231100 in provided. levels are higher stages at these to expressed of corresponding is millimeters expression Aqcoe3G231100 in The (spur+; -). in cups (spur (D) development spur tissue S1). petal and spur Table of (spur–) lacking Appendix, I (n some (SI phase petals and spanning incompatibility entire +) stages for genomic from (spur several Data a tissue across samples. of spur (±SE), + result containing replicates spur a some biological including be multiple samples, to of petal reads appears various normalized 2 mean chromosome (C as on 4). plotted association to Aqcoe3G231100, (x 2 The 3 F (columns rate. chromosome morphologies the discovery spur on in false various QTL spurs exhibiting major plants of and a presence/absence 1) indicate (column the all with at genotype spurs produce of not do that plants including ( spurs curved i.1. Fig. CD AB Arabidopsis dniyn addt ou o prls in loss spur for locus candidate a Identifying .( Bottom). aebe on ofnto nrgltn the regulating in function to found been have o ahdt on) aafor Data point). data each for 3 = xmlso aito nptlmrhlg ergtn nteF the in segregating morphology petal in variation of Examples B) Arabidopsis .ecalcarata A. n eiaotruncatula , Medicago o ahdt on) eeomna tg srpeetdnmrcly1t si e.2,adapoiaesu lengths spur approximate and 21, ref. in as 5 to 1 numerically represented is stage Developmental point). data each for 4 = (spur–), rhlgo Aqcoe3G231100, of ortholog xs eoi akrbnpsto ycrmsm ncniogn) h ahdln ersnste5 LOD 5% the represents line dashed The centimorgans). in chromosome by position bin marker genomic axis, .sibirica A. PALMATE-LIKE lwr rmtecosprns spurless parents, cross the from Flowers (A) ecalcarata . A. .coerulea A. 2 (spur+), eeain oaih fteod LD crs(y scores (LOD) odds the of Logarithm generation. .formosa A. Oiai r rmrf 7weeRAwsioae rmptl isce noblades into dissected petals from isolated was RNA where 17 ref. from are ‘Origami’ su+,and (spur+), hs asn naioai ifrne(2Q swl sthree as well as (E26Q) difference acid amino an causing phism the the Q-repeat Comparing to this S4). in Fig. occurring iden- Appendix , variants (SI sequence acid region amino shared only 99% the than with tity) greater (both identical nearly the The between the S4). sequence carata Fig. near protein Appendix , predicted repeats (SI and (Q) protein nucleotide glutamine predicted of the region of a N-terminus the in across for conserved highly except is genus, protein predicted nine the overall, from that, and distributed cross phylogenetically the additional in segregating alleles Aqcoe3G231100 26). (25, development leaf compound PENTAFOLIATA1 nlzn h euneo the of sequence the Analyzing and .sibirica A. 2 .chrysantha A. .sibirica A. eeaino h rs between cross the of generation lee hr sasnl uloiepolymor- nucleotide single a is there allele, ,fntosi h euainof regulation the in functions (PALM1), lee ergtn nteF the in segregating alleles su+ r rmrf 1weeRAwsisolated was RNA where 21 ref. from are (spur+) xs o h rsneo bec fspurs of absence or presence the for axis) .ecalcarata A. .ecalcarata A. .ecalcarata A. Aquilegia NSLts Articles Latest PNAS eoewd association Genome-wide ) and (Top) .ecalcarata A. 2 and eeainare generation pce shows species .sibirica A. and .sibirica A. .sibirica, A. .ecal- A. | allele f9 of 3 with

PLANT BIOLOGY additional Q residues in the A. ecalcarata allele. It is unlikely that expected, expression of Aqcoe3G231100 was lowest in A. ecal- this amino acid variation would result in the spur loss seen in A. carata homozygotes, highest in A. sibirica homozygotes, and ecalcarata, as the A. ecalcarata allele has a predicted amino acid intermediate in the heterozygotes (Fig. 2A). Aside from the cor- sequence identical to that found in Aquilegia japonica, which has relation between genotype at the POP QTL and spur presence nectar spurs (SI Appendix, Fig. S4). (i.e., EE are spurless and ES and SS are spurred), there does not appear to be a strong correlation between Aqcoe3G231100 Cis -Regulatory Changes Contribute to Differential Expression of expression level and spur length in the spurred F2 individuals that Aqcoe3G231100 in A. ecalcarata . Given that there are no predicted were sampled (R2 = 0.21, P = 0.058, n = 14; SI Appendix, Fig. amino acid differences that would cause functional effects in the S5), consistent with the observation that spur length is controlled protein and that this gene was first identified in part by tran- by multiple loci. scriptional differences between spur+ and spur– petal tissue, Focusing on the 10 heterozygotes, the small number of coding we predicted that the functional difference in the A. ecalcarata sequence nucleotide variants between the A. sibirica and A. ecal- allele would be due to cis-regulatory differences in gene expres- carata alleles were used to assign reads spanning these positions sion (while possible, it is less likely that a trans-regulator would as having been transcribed off of each parental allele. This analy- be encoded in the same QTL). To test this, we explored allele- sis showed a strong pattern in which, for nearly every F2 heterozy- specific expression patterns of our gene using RNAseq on F2 gote assessed, a greater number of reads was produced by the A. individuals with different genotypes at the POP QTL. Gene sibirica allele (directional Wilcoxon signed rank test, P = 0.009), expression was assayed in developing petals of 20 F2 hybrids: 5 further suggesting that cis-regulatory differences between these homozygous A. sibirica (SS), 5 homozygous A. ecalcarata (EE), alleles significantly contribute to the difference in expression and 10 heterozygous (ES) at the POP QTL. All F2 plants assayed seen both between the parental species and in the F2s with dif- had spur presence/absence phenotypes consistent with their POP ferent POP QTL genotypes and otherwise variable genetic back- QTL genotypes—the EE individuals were spurless, and the ES grounds (Fig. 2B). Comparing the upstream sequence between and SS individuals had spurs. Normalized expression patterns of A. ecalcarata, A. sibirica, and several other spurred species indi- Aqcoe3G231100 in the F2s significantly differed by their QTL cates that that there are few variants in the ∼1.5 kb directly genotype (ANOVA, F = 31.41, P = 1.96e-6; Dunnett modi- upstream of the transcription start site; however, there is a region fied Tukey–Kramer test results in SI Appendix, Table S3). As containing a number of single nucleotide and indel variants

AB C

DE F G H

I

Fig. 2. Expression of Aqcoe3G231100 in F2 hybrids and in A. coerulea ‘Origami.’ (A) Expression of Aqcoe3G231100 in petals (developmental stage ∼3 to 4) of F2 individuals that are either homozygous A. ecalcarata (EE, n = 5), homozygous A. sibirica (SS, n = 5), or heterozygous (ES, n = 10) at the spur loss QTL. Each point represents normalized reads as assessed by RNAseq of a single F2 individual, with the black bar indicating mean expression of all individuals for each genotype. (B) The number of Aqcoe3G231100 assignable reads transcribed from the A. ecalcarata allele (E reads) vs. the A. sibirica allele (S reads) for each of the 10 heterozygous F2s in A. The dashed line represents the 1:1 ratio of E and S reads. (C) Relative expression of Aqcoe3G231100 in spur tissue of A. coerulea ‘Origami’ across the transition from mitotic growth (phase I, purple) to cell differentiation (phase II, gray) as assayed at a variety of developmental stages in three different plants. Developmental staging of petals is presented as the sample spur length relative to the spur length of a fully developed petal for each plant (i.e., proportion of full spur length). For various proportions, the range of spur lengths across the three plants assessed is also provided, in millimeters. (D–H) Spatial expression of Aqcoe3G231100 in petals at various developmental stages using in situ hybridization. (D) In a stage 7 floral meristem, Aqcoe3G231100 is expressed throughout early differentiating petal (p), stamen (st), staminode (sd), and carplel (car) primordia (floral stages based on ref. 27; see SI Appendix, Table S4 for stage comparison with RNAseq stage). (E) Aqcoe3G231100 is broadly expressed in the early petal of a stage 9 flower. (F) As development progresses, Aqcoe3G231100 expression contracts to the tip of the developing spur in an early stage 10 flower. (G) Late in stage 10, Aqcoe3G231100 is restricted to the adaxial layer of cells at the spur tip. (H) Magnification of the spur tip in G. (I) The sense probe showed no signal. (Scale bars: D, H, and I, 100 µm; E, F, and G, 200 µm.)

4 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.2006912117 Ballerini et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 etitdt h prcp(i.2F (Fig. cup spur and becomes the gradually but to 2E) restricted (Fig. petals in broadly expressed tially nAqcoe3G231100-AqANS An (D) c2). in and flower the (c1 from indicated dissected petals KD two the Aqcoe3G231100-AqANS 3. Fig. most development petal in of 2 stages later development (Fig. in maintained through is it wanes organs, floral Aqcoe3G231100 of sion S7 (SI in carpels Fig. and and Appendix, meristem staminodes, floral stamens, the petals, in differentiating expressed early broadly qRT-PCR is and see gene 27; the RNAseq data), ref. with in comparisons as stage staging developmental (floral of 7 stages to developmental 3 early across coerulea hybridization A. situ in used when occurring was as transition 16 phase are petals ref. (this ‘Origami’ phase in mitotic elongation from identified cell transitioned previously the petals phase as to through dropped growth increased dramatically Expression but pattern. I each similar spurs In nectar a in 2C). showed Aqcoe3G231100 Fig. of II; expression the (phase differentia- assessed, elongation cell plant development cell into anisotropic petal I) and as (phase tion growth spurs mitotic in from Aqcoe3G231100 development transitioned of of I profile phase sion during 1D petals on (Fig. only conducted hor- were the in patterns expression variety assess qRT-PCR ticultural to used hybridization we situ Spur in Aqcoe3G231100, and during of patterns Activity expression Mitotic tial with Consistent Development. in Broadly Aqcoe3G231100 Is of ‘Origami’ Expression Spatial and Temporal expression. allele in into differences research to future for focus of (Chr03:27,452,100 site start 27,452,400; transcription to the of upstream kb ∼2 alrn tal. et Ballerini Aqcoe3G231100- AqANS and (R (red) length WT spur of and number number cell cell between against correlation plotted and positive length strong Spur a (G) indicating three. fit, other best the to relative arrows) by Aqcoe3G231100- an of Example (E expressed. B CDEF D AC oass pta xrsinpten fAceG310 we Aqcoe3G231100, of patterns expression spatial assess To mm.) 1 F, .Twr h n fpaeI xrsini vnulylim- eventually is expression I, phase of end the Toward E). hntpso ISo qo3210 in Aqcoe3G231100 of VIGS of Phenotypes E–H n e.2) eue R-C otakteexpres- the track to qRT-PCR used we 21), ref. and Tflwrwt utpespl eoe osedvlpn ea etrsusta aentbgnpouigatoynnpget.( F pigments. anthocyanin producing begun not have that spurs nectar petal developing see to removed sepals multiple with flower WT A ) Oiai oes uigflrldvlpetstages development floral During flowers. ‘Origami’ and nodrt ute xmn h eprladspa- and temporal the examine further to order In .Ti einmyb point a be may region This S6). Fig. Appendix, SI .coerulea A. i.S7 Fig. Appendix, SI A mt 0m nlnt;Fg 2C). Fig. length; in mm 10 to mm ∼8 and IStetdpatwt w Dptl hwn ral eue etr pr.( spurs. nectars reduced greatly showing petals KD two with plant treated VIGS D B C ihtefu Dptl niae d od) (E d4). to (d1 indicated petals KD four the with qN VIGS AqANS .Atog h expres- the Although 2D). Fig. and ’ Oiai’A ro Nsqassays RNAseq prior As ‘Origami.’ rgltr lmnsrelated elements cis-regulatory and rae ln sdt eiygn D hwn togrdcini prdvlpeti w eas(indicated petals two in development spur in reduction strong a showing KD, gene verify to used plant treated al S4 Table Appendix, SI .Telcsi ini- is locus The A–C). i.S7 Fig. Appendix, SI .coerulea A. .coerulea A. .coerulea A. T (B WT. (A) anthesis. at Flowers (A–D) ‘Origami.’ D for C ’ IStetdpatwt orK easwt aibyrdcdsus (D spurs. reduced variably with petals KD four with plant treated VIGS and ape sdt eiyadass h ereo eeK were KD stages gene developmental of 3 earlier degree at (Fig. the petal morphology assess on produced, based and is identified verify pigment to used nectaries before samples produce largely phenotypes still development, severe which petal less of 3 while some (Fig. nec- spurs, nectary, the shorter the of included strongest loss including near-complete The spur, the VIGS tar targeted. in of resulted phe- was evidence phenotypes spur showed spur Aqcoe3G231100 of that experiencing when range petals tissue A in response tissue. of present (WT) were marker wild-type notypes exact, versus not KD target although pigments floral general, red a the produce designed to in were required is constructs knockdown gene, which reporter VIGS, gene (AqANS), a with target target also variable of to highly degree is temporal (KD) and spatial, tive, in expression Aqcoe3G231100 the Number. down of knock Cell transiently to in (VIGS) Silencing Reduction Gene a Development Spur by Altered Caused Dramatically in Results Knockdown Gene with associated generally tissues. detected active is mitotically was expression 2I been signal (Fig. has Aqcoe3G231100 No probe (this tially, 17). sense ref. assayed a in was using leaves expression young which for in confirmed phase the ing i.S7 Fig. , Appendix (SI leaves in young detected D, and also expres- ovules, to was placenta, addition expression In the Aqcoe3G231100 markers. petals, division in cell in sion I, seen phase not in late is Aqcoe3G231100 which of localization adaxial the by of tion indicated as 2 cells (Fig. active HISTONE4 mitotically cup of spur contraction petal the mirrors the of surface H adaxial the to ited .Ti atr fporsiesailrsrcint h prtip spur the to restriction spatial progressive of pattern This ). .coerulea A. and F, F lwr tadvlpetlsaewe qo3210 snormally is Aqcoe3G231100 when stage developmental a at Flowers ) 2 E 0.95, = A–F ,alo hc r rsmbymttclyatv dur- active mitotically presumably are which of all G), and xrsin(6 7,wt h oal excep- notable the with 17), (16, expression (HIS4) .A qo3210 sepesdol al in early only expressed is Aqcoe3G231100 As ). P Oiai easadptl n a eue as used be can and petals and sepals ‘Origami’ F 3.0e-9, = .I ardsmls(Tv.K easfrom petals KD vs. (WT samples paired In ). n .coerulea A. C 4.(cl bars: (Scale 14). = 0 easdsetdfo h oe in flower the from dissected Petals ) G .Tu,bt eprlyadspa- and temporally both Thus, ). ) NHCAII SYNTHASE ANTHOCYANIDIN AqANS Oiai’A h quantita- the As ’Origami.’ D(ry easadteln of line the and petals (gray) KD iecdVG oto.(C control. VIGS silenced eue Virus-Induced used We A, NSLts Articles Latest PNAS B, C, C 0 , and D, D 0 0 cm; 1 , | G Petals ) C f9 of 5 An ) with and A, E )

PLANT BIOLOGY the same flower), KD petals had, on average, 36.1% of the the leaf phenotypes in Medicago and Aquilegia differ consid- expression of Aqcoe3G231100 in comparison with WT petals erably, the fact that the perturbation of normal PALM1/POP (n = 13, SE 14.4%; SI Appendix, Fig. S8A). activity results in mutant leaf phenotypes in both genera may Developmental studies of several Aquilegia species have shown reflect an ancestral function of the gene lineage in leaf develop- that interspecific variation in spur length can largely be attributed ment. Further study will be required to determine when the petal to differences in anisotropic cell elongation, rather than differ- expression domain evolved for POP orthologs in the Ranuncu- ences in overall cell number (16). In order to assess whether the laceae and whether this coincides with the evolution of spurs in differences in spur length seen in the VIGS KD petals are due Aquilegia. to differences in cell elongation, such as those seen with inter- More generally, it is interesting to note that several closely specific variation, or in cell number, which would suggest that related C2H2 homologs control crucial aspects of plant organ Aqcoe3G231100 plays a role in regulating mitosis, we counted development by repressing genes regulating cell morphogenetic and measured spur cells and assessed spur length from both activity. For instance, in Arabidopsis, the closely related homolog WT and KD petals. These comparisons show that variation in RABBIT EARS (RBE)(SI Appendix, Fig. S4) also influences petal spur length is not correlated with cell length (R2 = 0.02, P = shape by controlling the timing of the developmental shift from 0.60, n = 14; SI Appendix, Fig. S8B), but is highly correlated with mitotic growth to differentiation (29). The ultimate role of RBE cell number (R2 = 0.95, P = 3.0e-9, n = 14; Fig. 3G), indicat- appears to be more similar to POP in that both loci function ing that the reduction of spur length in KD spurs is primarily to maintain mitosis in petals, which RBE achieves by directly due to the production of fewer cells. Despite the fact that in situ repressing the cell differentiation factor TCP4 (29, 30). Studies hybridization indicates expression of Aqcoe3G231100 in other of AqTCP4 have confirmed a parallel function in Aquilegia petal floral organs, no other obvious phenotypes were noted in flo- spurs, where the gene is responsible for shutting down cell pro- ral tissue; however, many plants exhibiting KD in floral tissue liferation in what we have termed the distal compartment of the also exhibited a leaf phenotype (SI Appendix, Fig. S9A). In these developing spur (17). Although PALM1 and RBE appear to play leaves, the aspect ratio of leaflet shape was altered from rounded relatively narrow roles in leaf and petal development, respec- in control leaves to elongated and lanceolate in presumably tively, another more distantly related C2H2, JAGGED (JAG), silenced leaves (SI Appendix, Fig. S9 B and C). In more-extreme functions more broadly in the regulation of cell proliferation by cases, dissections between leaflet lobes were exaggerated such directly repressing several cell cycle inhibitors (KRPs; ref. 31). that the overall leaf architecture shifted from ternately compound The Aquilegia ortholog, AqJAG, similarly promotes cell prolifer- to biternately compound leaves (SI Appendix, Fig. S9C). ation in most lateral organs, including petals. Although not spur specific, the role of AqJAG in promoting cell proliferation and Discussion laminar expansion is necessary for spur development, as KD of Here we present multiple lines of evidence that Aqcoe3G231100 AqJAG also results in the loss of spur and nectary development is critical to the development of nectar spurs in Aquilegia. As (32). What all of these C2H2 TFs have in common is the presence prior studies have shown, the gene is expressed in the petals of a repressive EAR domain (23). The previous study of petal of spurred Aquilegia taxa but not in the petals of A. ecal- transcriptomics that compared A. ecalcarata to three spurred carata, which has secondarily lost nectar spurs (21). The gene species found that there were generally 50 to 100% more genes is located under the POP QTL, which is strongly associated with up-regulated in A. ecalcarata compared to spurred taxa at equiv- the presence or absence of spurs. Spatial and temporal expres- alent developmental stages (21). That study hypothesized that sion patterns reveal highly localized expression in developing this was due to a heterochronic shift toward earlier differentia- spurs during phase I, the mitotic phase of spur development. tion in A. ecalcarata petals. Loss of expression of a transcriptional Moreover, although we have not pinpointed the genetic vari- repressor (POP) could be the basis for this shift from prolonged ants responsible for differential expression of Aqcoe3G231100 cell division in spurred species to rapid differentiation in the between the species, allele-specific expression patterns in het- spurless A. ecalcarata. erozygotes indicate that cis-regulatory differences contribute to It is notable that POP expression is spatially more restricted the lack of Aqcoe3G231100 expression in A. ecalcarata. Finally, than what we have previously observed for cell division mark- knocking down expression of the gene in the long spurred variety ers in the developing petal spur (16, 17). In particular, in petal A. coerulea ‘Origami’ causes a dramatic reduction in the number stages that still show broad AqHIS4 expression, POP expression of spur cells produced, resulting in shortened spurs and, in some is localized more narrowly toward the tip of the spur, and, at later cases, the near-complete loss of the spur, including the nectary— developmental stages, it becomes restricted to the adaxial surface traits consistent with the A. ecalcarata phenotype. Given these inside the developing organ. If POP is actively promoting this multiple lines of evidence, we are designating Aqcoe3G231100 comparatively broader domain of cell division, its function must as POPOVICH (POP). have a non-cell-autonomous component. In terms of the adax- Our results suggest that POP plays a key morphogenetic role ial localization of its expression, there is a parallel observed in in spur development by promoting cell division in the Aquile- the Medicago ortholog PALM1, which is negatively regulated by gia spur cup. Based solely on what little is known about the an ortholog of the abaxial factor AUXIN RESPONSE FACTOR function of POP orthologs in other taxa, this locus would seem 3 (ARF3) (33). Conserved negative regulation by the abaxial to be an unlikely candidate gene for controlling spur develop- identity program in Aquilegia petals would explain the adaxial ment. The M. truncatula ortholog of POP, PALM1, regulates localization of POP, but still leaves open the question of how the complexity of compound leaf development by suppress- POP promotes cell divisions more broadly in the organ. ing mitotic activity in developing leaves (25). This is achieved Future studies to dissect the genetic interactions among POP, through the direct repression of SINGLE LEAFLET1 (SGL1), AqTCP4, and AqJAG, as well as identifying additional targets of a LEAFY/FLORICAULA ortholog that functions to maintain POP, will provide new insight into the regulatory architecture indeterminacy during the development of compound leaves in M. of Aquilegia spur development. Although POP clearly plays a truncatula (25, 28). Nothing is known about the function of the pivotal role in the earliest phase of spur development, this Arabidopsis POP/PALM1 ortholog, although it is also expressed function appears to be epistatic to continuous morphological in young developing leaves (24), as we similarly saw with POP. variation in traits such as spur length and curvature that is segre- The expression in developing leaves of POP and its orthologs in gating in the spurred individuals of the F2 generation, suggesting M. truncatula and Arabidopsis suggests that early leaf expression that other loci controlling spur morphology vary in this cross. is a common feature of POP orthologs in eudicots, and, although The identification of such loci will allow us to determine whether

6 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.2006912117 Ballerini et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 xml,sm ftevraini F in variation the of some through example, function they not or alrn tal. et Ballerini spurs nectar of evolution the regarding in questions many While Conclusion development. leaf because that haps note in in changes to loss by interesting spur caused is of it component case, major either the In (35). scenarios these muta- in before genetics morphology at spur other occurred of these eventually tions aspects in first smaller occurring in affecting mutations happened loci have with could accumulated fashion, loss secondarily stepwise spur a have hand, other could the they On mutations. point which at phenotype moot, loss spur the POP that indicates cross in this scenario, in one loss spur to leading narios development. nectary and length spur function not increasing the is between of itself study Further nectary growth. the of spur that promote suggest to results required These length. development loci spur development spur nectary essential stunt at the and nectaries of not KD the does stud- and of (34), Previous stages removal initiated. developmental mechanical that be that expression early can observed of nectary have duration the ies or is before there number required that cell such is of indirect, more threshold be some could Alterna- interaction development. the nectary promote tively, to insufficient be would lower be scenario, could this Under POP it program. of developmental First, several nectary level the are observation. higher There this a formation. that for nectary explanations proper possible and length spur F studied the both previously of POP phenotypes of Interestingly, up-stream (18). development indirectly, This least nectaries. AqSTY lack at phenotypes it, KD places VIGS stronger and F alleles both as development, number. than due rather appears length, spur small cell a in that changes spur, such to a elongation cell generate affecting to by or able be causing may by either variation length spur affecting circumstances, certain of most Although 7. F and the 5 chromosomes on eight QTL one small-effect least the the at explain have all help individuals also These may loci F such spurred of in com- identification primary differences the length The be spur to interspecific shown of been ponent has which dif- anisotropy, length cell spur of the F the Alternatively, among cross. ferences this in variation length com- analysis POP limited the of results paring the number, cell modulating variable that in show of studies variants levels KD potential VIGS the and though these Even regulators. of combination a through and the by number trolled cell to related rma vltoayprpcie ubro uainlsce- mutational of number a perspective, evolutionary an From that indicate also studies Our .ecalcarata A. POP Aquilegia AqSTY2 slne oessgetta hr sarltosi between relationship a is there that suggest flowers -silenced ln,rneigadtoa oiseict prmorphology spur to specific loci additional rendering alone, 2 xrsini nieyt eamjrdtriato spur of determinant major a be to unlikely is expression xrsinwudpooeol iie prgot and growth spur limited only promote would expression ihsc nallccmiainlc pr 3/8,under (30/38), spurs lack combination allelic an such with s POP nsu eeomn ilhl lcdt h relationship the elucidate help will development spur in ooos hc r seta oncay u o spur not but nectary, to essential are which homologs, POP 2 htaehmzgu for homozygous are that s .ecalcarata A. xrsinwt prlnt nte1 F 14 the in length spur with expression tl ean h dnicto fti e regulator key this of identification the remain, still opeeyeiiae etre ihu impacting without nectaries eliminates completely rgltr ifrne tthe at differences cis-regulatory POP xrsincndaaial le prlnt by length spur alter dramatically can expression ol aebe agl civdb uain at mutations by achieved largely been have could POP a litoi fet nohrtat uhas such traits other on effects pleiotropic has POP 2 xrsint asaoeatrsodlevel threshold a above pass to expression a ecue ycagsi h degree the in changes by caused be may s .sibirica A. POP n t lss eaie a epclarify help may relatives closest its and 2 lnshmzgu for homozygous plants xrsin ahrta ucin per- function, than rather expression, tde ouigo population on focusing Studies POP. POP .ecalcarata A. xrsini eurdt activate to required is expression POP POP xrsinlvl,ete con- either levels, expression lee tteeo te loci other or these at alleles oiflec prlnt.For length. spur influence to 2 .ecalcarata A. prlnt a edirectly be may length spur .ecalcarata A. .sibirica A. sncsayfrnectary for necessary is a ecniee.In considered. be can POP leea ohof both at allele Aquilegia lee at alleles 2 per obe to appears ou tefor itself locus ugs that suggest s .ecalcarata A. AqSTY1 trans- 2 POP. (16). and diinllnso eunig(0 ppie n)wr sdt sequence to used Three were individual. was end) per it paired coverage bp data, (100 sequence in sequencing the variation of processing 100-bp high lanes In as additional was lane. sequenced there single and that a pooled apparent in were reads individuals end library 96 paired of that each assumed for initially libraries F was were the it amounts 1/20th of input prep, using DNA coverage library and prep that the library Given indices during (36). for standardized Nextera Inc.) allowing (Illumina protocol Illumina volumes modified (ThermoFisher reaction using fluorometer a 2.0 prepared and the Qubit reagents were for the using as Libraries quantified protocol as Scientific). same DNA of the ng using 2 frozen flash sibirica were A. 100-bp that as leaves sequenced young and to Inc.) reads (Illumina end kit paired Nextera Illumina the of volume the the for for Inc.). library available (Qiagen sequencing beads tissue outlined MagAttract no variations and was with reagents There DNeasy 38, Qiagen ref. using in tissue described the that For to below. similar method ing (SI trait Genotyping. binary presence/absence a as scored S1D). and Fig. was Appendix, scans absence greater or the presence with from Spur those “spurred.” assessed while designated were “spurless,” Appendix, spurs designated 8-mm (SI than were spurs pockets short 7-mm with than petals petals spurred These petals. from S1 spurred Fig. shape to relative in length differed in also difference clear pocket a F petal with some the ets that of mea- such bimodal extent were clearly distal individuals was the 92 (ImageJ; to phenotyping, spur point on or attachment V370 pass the Perfection first from were Epson an Sheets sured a paper. on clear As resolution white inch scanner. of of per piece Photo side pixels a 720 sticky to were at affixed scanned Petals the then then anthesis. onto was at that placed tape plant and packing the longitudinally, from folded removed removed, were mature to flowers into Phenotyping. put plants 100 of batch Each by flowering. staggered promote was 4 vernalization to at 100, vernalized of were pro- batches plants After leaves, in pots. moved vegetative gallon 12 were in least they up at matured, ending ducing plants eventually vegetative pots, As larger mo. successively 3 into of germi- conditions span seedlings light a 500 natural approximately over total, nated under In germinate months. to spring facil- the allowed greenhouse throughout were Barbara Santa seeds (UC), and California ities, of University the to individual moved in 4 soil at Agriculture) stratified Gro and (Sun trays #4 plug Sunshine in sown were seeds 1,000 F F sibirica An Nurseries. Suncrest from acquired Cross. Genetic Methods and Materials controlling locus gene identified. candidate have critical pure not which would a innovation, approaches key of com- a identity of combining development the of the gene discover power of to the analyses aspects and highlights function notable mapping, it most genetic that transcriptomics, the is parative of study one this Perhaps of genus. the deter- across to in ogy variation interest morphological of subtle be in more also variation will whether it mine broadly, More ecalcarata. A. tn h pcfi eei ai o osof loss for under- basis to genetic seek we specific as the speciation, of stand basis genetic the concerning nectar to of how leading identification lineage regarding the in expression questions petal development. acquired address spur in in can evolved role spurs we its that addition, understand pathway better In developmental to and focus regulates genetic of it the point elucidate of critical to identification a the work With with research. us future provides for development spur of 2 eeainwsgnrtdb efigasnl F single a selfing by generated was generation C stemtra aetand parent maternal the as and aet euniglbaisfrteF the for libraries Sequencing parent. h F The n laslce etre.Bsdo hs easwt less with petals this, on Based nectaries. lacked always and D) o eryeeyF every nearly For otclua ie of lines Horticultural .sibirica A. i.S1 C Fig. Appendix, SI 2 2 ol erltvl osset n qa oue of volumes equal and consistent, relatively be would s eegntpduigawoegnm kmsequenc- skim whole-genome a using genotyped were s POP Aquilegia ∼50× ilfrhralwu oepoequestions explore to us allow further will aet N a xrce rmdscae leaf desiccated from extracted was DNA parent, .sibirica A. oeae o h F the For coverage. ◦ o k fe taicto,pu ry were trays plug stratification, After wk. 4 for C o ntotal, In mo. ∼1 2 ydtriighwadwhen and how determining by POP ln htflwrd h rtadsecond and first the flowered, that plant .ecalcarata A. 2 aetwscntutduigahalf- a using constructed was parent omdwal endptlpock- petal defined weakly formed s .sibirica A. 1 rhlg otiue oother to contributes orthologs .ecalcarata A. .Tedsrbto fteelengths these of distribution The ). eeainwsgnrtdwith generated was generation Aquilegia 2 NSLts Articles Latest PNAS ,DAwsetatdfrom extracted was DNA s, 0 lnsflowered. plants ∼400 steptra aet The parent. paternal the as and 2 eepeae using prepared were s 1 POP aeto h F the or parent ln.Approximately plant. ecnnow can we POP, .ecalcarata A. prmorphol- spur The Aquilegia. xrsinin expression ◦ o wk 8 for C | 1 POP f9 of 7 The . were A.

PLANT BIOLOGY additional F2 libraries, including those from the first lane that had low cov- jgi.doe.gov/) and processed to generate read counts per transcript as erage, however these libraries were quantified using qPCR prior to pooling described in ref. 38. Normalized counts per transcript were estimated using into groups of 96 individuals. In total, 287 F2s were sequenced aiming for the trimmed mean of M-values method (via the calcNormFactors command) coverage depth of 1x, with those having greater than 0.1x coverage usable using the R package edgeR (45, 46). Differences in mean normalized expres- for genotyping (n = 286). All sequencing was conducted on a HiSeq3000 at sion of Aqcoe3G231100 were assessed using ANOVA as implemented in R, the Vincent J. Coates Genomics Sequencing Laboratory (UC Berkeley). and post hoc testing was done using the Dunnett modified Tukey–Kramer In order to identify informative sites to determine ancestry in the F2s, test (for unequal sample sizes) implemented using the DTK package in R the F2 raw sequence reads from a single lane (96 individuals) were merged (https://CRAN.R-project.org/package=DTK). Spur length for the ES and SS to reconstitute a mock F1. Reads for the A. sibirica parent, the mock individuals was measured from the attachment point to the nectary on F1, and the F2s were aligned to the A. coerulea ‘Goldsmith’ v3.1 refer- the proximal edge of the spur using ImageJ, and these measurements were ence genome (https://phytozome.jgi.doe.gov) using the Burrows–Wheeler used to test for a correlation between Aqcoe3G231100 expression and spur aligner (37) (details in ref. 38). Informative sites for determining ances- length. We did not have petal scans at anthesis for one ES individual, so try in the F2 generation were those where the mock F1 is heterozygous the sample size for this test was n = 14. For the 10 heterozygous individu- and the A. sibirica parent is homozygous. These were identified using als, reads spanning variant sites between the A. sibirica and A. ecalcarata SAMtools 0.1.19 (38, 39). Using SAMtools 0.1.19 and custom scripts (available alleles of Aqcoe3G231100 were counted using the Integrative Genomics at https://github.com/anjiballerini/POP), the number of reads at informa- Viewer (47), and a one-tailed Wilcoxon signed rank test was used to test tive sites indicative of A. sibirica or A. ecalcarata ancestry were counted for overrepresentation of A. sibirica derived reads. for each F2. These read counts were then processed in R (40) to deter- mine the frequency of reads from each parent across nonoverlapping Developmental qRT-PCR. Four flowers at various developmental stages were genomic windows, with read frequencies >0.9 or <0.1 indicating homozy- collected from each of three A. coerulea ‘Origami’ plants during a 1-h gosity for either A. sibirica or A. ecalcarata and frequencies between 0.4 period. As there was some variability in spur length from plant to plant, and 0.6 indicating heterozygosity. Windows were either 0.5 Mb or 1 Mb a fully developed flower at anthesis was used to get a measure for final in size depending on recombination rate estimates from previous crosses, spur length for each plant. For each flower collected, spurs were measured, with larger windows used in low-recombination regions of the genome. and the ratio of spur length to final spur length was used for developmental Several regions were identified as inconsistent between the A. coerulea staging. Spur tissue was isolated from each petal at the attachment point, ‘Goldsmith’ v3.1 physical map and the genetic map of a previous cross and tissue was flash frozen and kept at –80 ◦C prior to RNA isolation. RNA between A. formosa and A. pubescens (38). In these regions, custom bin isolation and genomic DNA removal was done using the RNeasy Plus Mini sizes were created to allow these marker bins to map appropriately in the Kit (Qiagen). Complementary DNA (cDNA) was synthesized using the Verso genetic map. cDNA Synthesis Kit and oligo dT primers (Thermo Fisher). A fragment of Aqcoe3G231100 was amplified and quantified using iTaq Universal SYBR 0 QTL Mapping. The F2 genotypes of different marker bins were used to Green Supermix (Bio-Rad) and primers 5 -GACCAATAACCTGATGGCATCCT- estimate a genetic map using the R package R/qtl v1.35-3 (41). The 30 and 50-CGGGGTGGTCTTGATGATCC-30. Expression of AqIPP2 (ISOPENTYL genetic map was then combined with the spur presence/absence phe- PYROPHOSPHATE:DIMETHYLALLYL PHYROPHOSPHATE ISOMERASE2) was notypes and F2 genotypes for 276 individuals to conduct QTL mapping used to normalize Aqcoe3G231100 expression (27). The relative expression (Dataset S1; gen.w.phen.csv, R/qtl v1.35-3; ref. 41). The spur phenotype of Aqcoe3G231100 was calculated using the ∆ ∆ Ct method with primer was scored as a binary trait (presence/absence), and the R/qtl command efficiency correction. scanone was used to calculate logarithm of the odds (LOD) scores for each marker using the expectation-maximization (EM) algorithm. Ten In Situ Hybridization. A 250-bp fragment of Aqcoe3G231100, including part thousand permutations were used to determine the 5% LOD cutoff for of the 50 untranslated region and the beginning of the open reading significant QTL. frame (which is a nonconserved region), was PCR amplified using primers 50-GTATTCGGAGCGAGGTTCACT-30 and 50-AACCCTACCAGGCAAAAC-30, and Gene Tree. The Aqcoe3G231100 predicted protein was queried against cloned into pCR4-TOPO vector (Thermo Fisher). In situ hybridization steps the proteomes of Arabidopsis thaliana (AT), M. truncatula (Medtr), and followed the protocol described by ref. 48. Slides were stained in 1% cal- Vitis vinifera (GSVIVG) using the BLAST algorithm in Phytozome 12 coflour white for 5 min before visualization. Images were taken using the (https://phytozome.jgi.doe.gov). Predicted proteins were aligned using the Zeiss AxioImager microscope at the Arnold Arboretum of Harvard University. ClustalW algorithm (42) in Geneious (v9.1.16, https://www.geneious.com), adjusted by eye, and trimmed to alignable regions. Initial relationships VIGS and Phenotypic Assessment. VIGS targeting Aqcoe3G231100 was con- were estimated using the Neighbor-Joining method (43) implemented in ducted following previously published protocols (49, 50). As Aqcoe3G231100 Geneious (v9.1.16, https://www.geneious.com), and the sequence list was consists of a single exon, a 223-bp sequence of the 30 end was trimmed to include sequences in several clades closely related to the amplified directly from A. coerulea ‘Origami’ DNA using primers 50- one containing Aqcoe3G231100 as well as those related to A. thaliana CGGAATTCCATCCTGCCCAACCCTCAAT-30 and 50-CGGCTCTAGACGGGGTGG- JAG, which was used as an outgroup. After trimming the sequence list, TCTTGATGATCC-30, which include sequences to build in EcoR1 and XbaI phylogenetic relationships were estimated using random accelerated max- restriction enzyme sites, respectively. This region does not have similarity imum likelihood (RAxML, ref. 44) with the Jones–Taylor–Thornton model with other C2H2 homologs. The fragment was then cloned into a TRV2 con- of amino acid replacement as implemented in the CIPRES web portal struct containing sequence from the Aquilegia ANS gene (49). A. coerulea (http://www.phylo.org/). We used previously published sequences from ‘Origami’ (Goldsmith, Syngenta) plants were grown until they had approxi- nine different species of Aquilegia (A. aurea - SRR405095, A. barnebyi - mately five to seven true leaves and then vernalized at 4 ◦C for 4 wk. A total SRR7965809, A. chrysantha - SRR408559, A. formosa - SRR408554, A. japon- of 331 plants were treated in multiple rounds. Of treated plants, 21.25% ica - SRR413499, A. longissima - SRR7965810, A. oxysepala var. oxysepala - exhibited an AqANS phenotype in sepals, 19.34% exhibited an AqANS phe- SRR413921, A. pubescens - SRR7943924, and A. vulgaris - SRR404349; ref. 38), notype in petals, and 17.22% exhibited a spur phenotype (89.1% of those in addition to the A. sibirica parent (SRR11508011) and an individual A. ecal- spurs showing an AqANS phenotype). KD of Aqcoe3G231100 was verified carata plant (SRR892115) from the same grower as the parent used in the in a subset of petal samples by comparing expression of the gene in KD cross to compare sequence variation in Aqcoe3G231100 across the Aquilegia and WT petals from the same flower using qRT-PCR using the same protocol phylogeny. as in Developmental qRT-PCR. Because Aqcoe3G231100 is only expressed early in petal development, largely before petals begin producing antho- RNAseq. Petal tissue from 20 F2 plants, 5 A. sibirica homozygotes, 5 A. cyanin pigments, in order to test targeted KD of Aqcoe3G231100, petals ecalcarata homozygotes, and 10 heterozygotes at the spur loss QTL was with KD phenotype were identified early in development based primarily on collected at the developmental stage equivalent to stage 3 in ref. 21 and shape rather than the lack of pigment production. As we do not know what flash frozen in liquid nitrogen. Total RNA was extracted using the Qiagen the final spur length would be and there is variation in KD effect across a RNeasy Micro kit (Qiagen). For each sample, RNA quantity and quality were single flower, it cannot be ruled out that some petals classified as WT also assessed prior to library construction as in ref. 21. Libraries were quanti- experienced some KD relative to true WT petals. fied using qPCR, pooled aiming for equal representation across samples, and Cell count and length measurements were made on fully developed WT sequenced in a single lane on the HiSeq4000 (Illumina Inc.) as 50-bp single (n = 4) and KD (n = 10) petals. WT and KD petals were harvested at anthe- end reads, at the UC Davis Genome Center. Raw reads were aligned to the sis and fixed overnight in FAA (3.7% formaldehyde, 5% glacial acetic acid, A. coerulea ‘Goldsmith’ v3.1 reference transcriptome (https://phytozome. 50% ethanol) at 4 ◦C. The tissue was dehydrated to 95% ethanol and then

8 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.2006912117 Ballerini et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 2 .B htal .A ogs olntrsit rv nraigyln etrsusin spurs nectar long increasingly drive shifts Pollinator Hodges, A. S. Whittall, B. J. between isolation 12. Floral Hodges, A. S. Fulton, plant M. sphingophilous and 11. ornithophilous between isolation floral of Origin Grant, V. 10. 6 .G,J hn .Chen, R. Chen, J. Ge, L. 26. the in innovations key for evidence Fossil Mayhew, J. P. Ross, J. A. Nicholson, B. D. 13. 4 .V lpkv,A .Ksao,E .Grsmv .D oahv,A .Pnn high Chen A J. Penin, A. 25. A. Logacheva, D. M. Gerasimov, S. E. Kasianov, S. A. Klepikova, V. A. 24. genus the on studies Genetic Prazmo, W. petal 22. early of transcriptomics Comparative Hodges, A. S. Kramer, M. E. Ballerini, S. E. of 21. ontogeny Floral Hodges, sympatric A. three S. of Tucker, C. isolation S. and traits 20. Floral Huang, Q. S. Sun, F. J. three- Yu, Q. for Tang, L. basis L. the Molecular of 19. Homologs Kramer, Kramer, M. E. M. Bunn, E. I. J. Levy, Min, Y. C. 18. Puzey, Evolution J. Mahadevan, Collani, L. S. Kramer, Yant, M. E. L. Hodges, 17. A. S. Gerbode, J. S. fishes: Puzey, cichlid and R. centrarchid evolutionary J. in and mechanics 16. biting ecological Pharyngeal of Drucker, synthesis G. E. A Galis, F. Mendes, P. 15. Braga, P. H. P. Peixoto, P. F. 14. 3 .C nlrct .Sho,S B S. Schoof, H. Englbrecht, C. C. 23. alrn tal. et Ballerini and article the the in included the and are in data https://github.com/anjiballerini/POP is study Otherwise, mapping at S1). for found used map- be file processing genotype-phenotype can National for the Scripts data PRJNA623619. at ping BioProject: (SRA) (https://www.ncbi.nlm.nih.gov/sra) following Archive Information the Read under Biotechnology Sequence for the in Center available is RNAseq Availability. Data con- 52). the (51, LSM700 using FIJI Zeiss measured in tool lengths a selection their on line and straight counted mounted were 512 Cells AxioCam microscope. 20× focal Zeiss at a from nectary continuously using imaged to illumination a was point underneath spur overnight attachment Each dry clips. the to the binder left and with by up, secured 18006) facing coverslip no. spur Sciences microscope the of Microscopy on side mounted (Electron proximal were 60 Petals Cytoseal fuchsin. using acid slides 1% in overnight stained .V rn,Ioainadhbiiainbetween hybridization and Isolation Grant, V. 9. Fior S. 8. Schluter, D. 7. key Kay a M. spurs nectar K. floral ecological Are 6. diversification: plant their Spurring Arnold, L. M. and Hodges, A. diversification. S. innovations and 5. spurs evolutionary nectar Floral Key Hodges, A. Hauser, S. L. 4. D. Heard, hier- B. S. different at 3. process and pattern novel evolutionary of origin “The Cracraft, J. 2. Simpson, G. 1. oubn flowers. columbine pubescens. species. Bot. Evol. Syst. J. A Aliso regions. cpDNA evolving (2013). rapidly 579–592 of sequencing next-generation Oxford, Press, 2000). University Kingdom, United (Oxford Eds. Barrett, H. 311–325. C. pp. S. 2006), Harder, Kingdom, United L. Flowers, of lution Sci. Biol. B Soc. Biol. innovation?Proc. (1997). mechanisms. 21–46. pp. 1990), IL, Chicago, Press, in levels” archical 1953). NY, ne rncito atresnilfrcmon efmrhgnssin morphogenesis leaf compound for truncatula. essential factor transcription finger (2010). 10754–10759 107, legume model the in factor scription diversity. insect of evolution N-e profiling. RNA-seq the of map resolution the in (2004). proteins finger zinc C2H2 vulgaris development. spur nectar of with species associated diverse four across development (Ranunculaceae). (2007). Aquilegia in development (2015). the of elaboration dimensional in anisotropy. diversity spur-length of innovation. evolutionary key a into Insights scales. spatial across diversity bat of determinants ptoeprlrcntuto fthe of reconstruction Spatiotemporal al., et oto fdsetdla opooyb y()i()zn ne tran- finger zinc Cys(2)His(2) a by morphology leaf dissected of Control al., et rc al cd c.U.S.A. Sci. Acad. Natl. Proc. .and L. pce nteQnigMutis China. Mountains, Qinling the in species Foa hrcesadseisdvricto”in diversification” species and characters “Floral al., et rc il o.BBo.Sci. Biol. B Soc. Biol. Proc. ln inl Behav. Signal. Plant rc il Sci. Biol. Proc. h ao etrso Evolution of Features Major The h clg fAatv Radiation Adaptive of Ecology The it Biol. Hist. qiei ecalcarata Aquilegia Aquilegia. a eunedt sdfrgntcmpigadteF the and mapping genetic for used data sequence Raw n.J ln Sci. Plant J. Int. .H iei d Uiest fChicago of (University Ed. Niteki, H. M. Innovations, Evolutionary ln J. Plant Nature rbdpi thaliana Arabidopsis 5–7 (1995). 151–173 10, amt-iepentafoliata1 Palmate-like 4–6 (1952). 341–360 2, 6014 (2012). 1640–1645 279, 0817 (2016). 1058–1070 88, 0–0 (2007). 706–709 447, e Phytol. New Aquilegia rc il o.BBo.Sci. Biol. B Soc. Biol. Proc. 1413 (2010). 1134–1137 5, h,Cnevto,dvricto n xaso of expansion and diversification Conservation, ohm, ¨ Aquilegia 5–7 (2005). 557–574 166, rbdpi thaliana Arabidopsis 7973 (1993). 7729–7733 90, Maxim. eiaotruncatula. Medicago 2725 (1999). 2247–2252 266, 4–4 (1995). 343–348 262, easi civdsll hog cell-shape through solely achieved is petals 0010 (2019). 1090–1100 221, M Genom. BMC Aquilegia ea spur. petal .Eo.Biol. Evol. J. caSc o.Pol. Bot. Soc. Acta Aquilegia eeomna rncitm ae on based transcriptome developmental Aquilegia, Clmi nvriyPes e York, New Press, University (Columbia anfiainudrbrightfield under magnification qiei formosa Aquilegia STYLISH M Ecol. BMC Aquilegia ln ytmt Evol. Systemat. Plant Ofr nvriyPes Oxford, Press, University (Oxford qiei formosa Aquilegia noe oe y()i()zinc Cys(2)His(2) novel a encodes .I rse between Crosses I. L. 0483(2014). 20141823 281, n.J ln Sci. Plant J. Int. eelfwgnsconsistently genes few reveal rc il Sci. Biol. Proc. genome. 6 (2019). 668 20, and Semiaquilegia, 4–7 (1996). 641–670 9, eefml oto nectary control family gene rc al cd c.U.S.A. Sci. Acad. Natl. Proc. IAppendix SI 8(2018). 18 18, ai aito through radiation rapid 2–4 (1960). 423–442 29, IAppendix SI M Genom. BMC and e Phytol. New clg n Evo- and Ecology 20142778 282, and .pubescens A. S81–S88 158, 121–128 267, Medicago (Dataset Aquilegia Aquilegia Enemion . 39 5, 198, 2 . 2 .A cnie,W .Rsad .W lcii I mg oIae:2 er fimage of years 25 ImageJ: to Image NIH Eliceiri, W. K. Rasband, S. W. Schneider, A. C. 52. differential for method normalization scaling A Oshlack, A. for package Robinson, bioconductor D. A M. edgeR: 46. Smyth, K. G. McCarthy, J. D. RAxML Robinson, the for D. M. algorithm bootstrap 45. rapid A Rougemont, reconstructing J. for Hoover, method P. Stamatakis, new A. A method: 44. neighbor-joining The Nei, M. Saitou, sensitivity N. the 43. Improving W: CLUSTAL Gibson, J. T. Higgins, G. D. Thompson, D. J. 42. Wu, H. Broman, W. K. 41. Team, Core R 40. Li H. 39. Baym M. 36. 0 .Sam,E .Kae,Vrsidcdgn iecn nterpdccigcolumbine Schindelin cycling rapid J. the in 51. silencing gene Virus-induced Kramer, M. E. Sharma, B. analyses 50. functional for tool a as silencing gene Virus-induced Kramer, M. E. Gould, B. 49. structures: reproductive plant of evolution the studying for “Methods Kramer, M. E. 48. Thorvaldsd H. Robinson, T. J. 47. Burrows-Wheeler with alignment read short Filiault accurate L. and D. Fast 38. Li, H. Durbin, R. 37. plants,” in change Huang shape L. of mechanisms 35. and evolution microRNA “Plant Puzey, R. J. 34. Madue F. Berbel, A. Peng, J. 33. The Kramer, M. E. Min, Y. 32. Mui M. J. Schiessl, of K. targeting miR319a 31. Jack, T. King, S. Nag, A. 30. Li J. 29. 7 .S alrn,E .Kae,Evrnetladmlclraayi ftefloral the of analysis molecular and Wang Environmental H. Kramer, 28. M. E. Ballerini, S. E. 27. ainGat hs udn oishdn oei h eino the of the design writing the in or in data, role of no Instrumen- interpretation had OD018174 and bodies manuscript. S10 analysis, funding NIH collection, by These study, Sequencing at supported Instrumentation Grant. Genomics Cores Berkeley, Shared Coates tation UC Analysis NIH J. at Vincent by Expression the Laboratory supported and was and Center, sequencing Technologies Grant 1S10OD010786-01, Genome The Grant NSF DNA Davis E.M.K. E.S.B., the to UC to 1456217 the by award Grant out Discovery NSF and carried Karp S.A.H., Harvey by to Barbara funded 1456317 Santa was Kirschstein L. Research UC Ruth (Grant (F32GM103154). the the Award NSF under NIH Service The the facilitated Research leader. by that National supported inspirational greenhouses was E.S.B. the an growth. construct plant being to Hannah- funds for F C. provided Popovich the OIA-0963547) and G. maintaining Taber thank in D. We invaluable project. provided Shahandeh were the M. Bick and throughout Mem- laboratories qPCR. feedback Finkelstein conducting and constructive and S.A.H. the DNA, of extracting bers planting, to contributing for ACKNOWLEDGMENTS. analysis. Methods data. RNA-seq of analysis expression data. expression gene digital (2010). 139–140 of analysis expression differential servers. web trees. phylogenetic choice. matrix weight position- (1994). and weighting, penalties sequence gap through alignment specific sequence multiple progressive of crosses. 2017). Austria, Vienna, Computing, Statistical for Foundation (2009). 2078–2079 history. unique a (2018). with e36426 chromosome 7, polymorphic extraordinarily an reveals and genomes. (2018). legia in development qiei coerulea Aquilegia (2007). 6 3, plant model emerging the in 617–635. pp. 2005), Data chemical in techniques” expression gene Comparative Viewer. Genomics Integrative the with transform. (2012). MA Cambridge, University, Harvard thesis, PhD repressing directly in by expression patterning leaf compound stems. and leaves both in types cell adaxial inhibitors. cycle cell Kip-related repressing U.S.A. by Sci. growth tissue to ing in (2008). ortholog eudicot lower the in transition Arabidopsis. tal., et h euneAinetMpfra n SAMtools. and format Alignment/Map Sequence The al., et pce eeldb 0ncergn fragments. gene nuclear 10 by revealed species Bioinformatics a.Methods Nat. eei iest n vltoayhsoyo orcoeyrelated closely four of history evolutionary and diversity Genetic al., et 7–8 (2012). 676–682 9, IGELEAFLET1 SINGLE ABTEARS RABBIT lSOne PloS 8023 (2014). 2830–2835 111, oto fcmon efdvlpetby development leaf compound of Control al., et nxesv utpee irr rprto o megabase-sized for preparation library multiplexed Inexpensive al., et Bioinformatics y.Bio. Sys. The al., et tal. et eiaotruncatula. Medicago .A imr .H olo,Es Esve cdmc a ig,CA, Diego, San Academic, (Elsevier Eds. Roalson, H. E. Zimmer, A. E. , ... (R v.3.6.3 Computing, Statistical for Environment and Language A R: .Ep Bot. Exp. J. Arabidopsis. ii noe-orepafr o ilgcliaeanalysis. biological-image for platform open-source An Fiji: , ehd o.Biol. Mol. ‘Origami,’Methods o.Bo.Evol. Biol. Mol. 0206(2015). e0128036 10, o .Sbosi rbdpi AGDlnsflrlognpattern- organ floral links JAGGED Arabidopsis Sablowski, R. no, ˜ 5–7 (2008). 758–771 57, 8–9 (2003). 889–890 19, 7–7 (2012). 671–675 9, euae h rncito of transcription the regulates .Sn .A hrhl,Rql T apn nexperimental in mapping QTL R/qtl: Churchill, A. G. Sen, S. ´ etakJ ehe,D idee n .Pretorius J. and Birdseye, D. Wechter, J. thank We Aquilegia 7416 (2009). 1754–1760 25, 4368 (2016). 6473–6480 67, ti,A .Wne,A ei,J .Msrv ain review Variant Mesirov, P. J. Zehir, A. Wenger, M. A. ottir, qiei JAGGED Aquilegia in ´ rc al cd c.U.S.A. Sci. Acad. Natl. Proc. Aquilegia o .Ce,AXNRSOS ATR regulates FACTOR3 RESPONSE AUXIN Chen, R. no, Aquilegia. ˜ eiaotruncatula. Medicago eoepoie nih noaatv radiation adaptive into insight provides genome 0–2 (1987). 406–425 4, rn.PatSci. Plant Front. eoeBiol. Genome ac Res. Canc. clmie Ranunculaceae). (columbine, EvoDevo e Phytol. New oeua vlto:PouigteBio- the Producing Evolution: Molecular TCP4 ooo rmtspoieainof proliferation promotes homolog 2 18 (2013). 71–81 975, 3–3 (2017). e31–e34 77, ouainadVG plants. VIGS and population uli cd Res. Acids Nucleic 2 (2010). R25 11, (2011). 4 2, 2–3 (2017). 927–932 8, NSLts Articles Latest PNAS AMT-IEPENTAFOLIATA1 PALMATE-LIKE TCP4 sciia o ea rwhand growth petal for critical is ln Physiol. Plant .Ssea.Evol. Systemat. J. 3–4 (2017). 536–548 216, 23–23 (2009). 22534–22539 106, uigptldevelopment petal during FLORICAULA/LEAFY Bioinformatics Bioinformatics rc al Acad. Natl. Proc. 1759–1772 146, ln Methods Plant 4673–4680 22, 129–138 56, | f9 of 9 Aqui- eLife Nat. 26, 25,

PLANT BIOLOGY