Unraveling the Phylogeny Of

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 DOI 10.1600/036364411X553306 © Copyright2011 bythe American SocietyofPlantTaxonomists Systematic Botany positions of Dipsacales evolutionisobscured bytheuncertainsystematic difficult toanswer. Ourknowledgeandunderstandingof Dipsacales phylogeny, anumberofquestionshaveproven Abelia gests acloserelationship between 2007 ; Winkworth etal.2008 et al.2003 ). Morphology,; however, sug- Bell andDonoghue2005 ; ( Moore andDonoghue Pyck and Smets2000 ; Donoghue etal.2001 , 2003 ; Zhang Wall., relationship of lar investigationshaveprovided strong supportforasister tal inunderstandingDipsacalesevolution.Recentmolecu- Unraveling thephylogenyofthesekeygeneraisinstrumen- and or) sevenflowers, henceitscommonname,seven-son flower. nerved leaves and capitateinflorescences composedof(six China. Itshabitischaracterized bydarkgreen, opposite, tri- in 1916. has beenproblematic eversinceRehderdescribedthe genus whereas theLinnaeaclades.omits Dipelta Linnaeeae sensu Donoghue etal.2001 ) comprising Linnaea clade(Linnaeaceaesensu Backlund andPyck1998 ; clade s.l.corresponds tothetraditionalcircumscription ofthe Caprifoliaceae lineages( Fig. 1 ). Inwhatfollows,theLinnaea and wetherefore chosetoassigninformalnamesthemajor 2001 ). Unfortunately, thishasledtoconsiderable confusion phylogeny (e.g. Backlund andPyck1998 ; in response totheimproved understandingofDipsacales Donoghue et al. Zabelia ing withDipsacalesphylogenyhavesampledboth closely related toorcongenericwith Despite ourimproved knowledgeandunderstandingof The systematicpositionof Several newnamesandclassificationshavebeenproposed

4 National HerbariumoftheNetherlands,LeidenUniversityBranch,P.O. Box9514,NL-2300RA Leiden,TheNetherlands Leycesteria strongly indicateacloserelationship between with theMorinacladeisonlysupportedbyshared presence ofpsilatepollengrainswithanendocingulum.Fruit andseedmo clade s.l.Noneofourphylogeneticanalysessuggestacloserelationship between Heptacodium they are keygeneratohelpusunderstandevolutioninDipsacales.Ourmolecularanalysesstrongly supportasisterrelationshi ship withtheLinnaeaclades.l.(formertribeLinnaeeae). Linnaea s.l.cladesisessentialtosolvethismatter. that couldexplaintheuncertainsystematicpositionofgenus. A betterunderstandingoftheintergeneric relationships of aspects, itdoesnotaidinunravelingthesystematicpositionofgenus.Incase This studypresents molecularandmorphologicaldatainanattempttofurtherclarifythesystematicpositionsof R.Br. and Lonicera toassesstheirrelationship ( Pyck 2001 ). Abstract— Keywords— Maxim., Heptacodium at Jacobs, Bart (2011), 36(1):pp.231–252 Heptacodium L., (e.g. Rehder 1916 ). andtheLoniceraclade. Heptacodium Recent molecularinvestigationsplace Kolkwitzia Abelia 1 Zabelia Symphoricarpos Laboratory ofPlantSystematics,KatholiekeUniversiteitLeuven,Kasteelpark Arenberg 31, 2

Scientific InstituteofPublicHealth,JulietteWytsmanstraat 14,BE-1050Brussels, Belgium

5 isadeciduousshrub andendemic to Institute forSystematicBotanyandEcology, UlmUniversity, D-89081Ulm,Germany 3

Jodrell Laboratory, RoyalBotanicGardens, Kew, TW93DS,Richmond,Surrey, U.K. iscls, Lnaa ld s l oiea ld opooy, phylogeny. morphology , clade, Lonicera l., s. clade Linnaea Dipsacales, , 1 , 6 Rehderand (e.g. Zhang etal.2002 ), or Ke Geuten, Koen Urvln te hlgn o of Phylogeny the Unraveling Graebn., andthecladeholding Heptacodium miconioides (Caprifoliaceae): AnInterdisciplinaryApproach Duhamel,and Zabelia 6 Author forcorrespondence ( [email protected] ) Zabelia Zabelia Abelia Heptacodium Linnaea Zabelia ismostlyconsidered , however, seemstobecloserrelated totheMorinacladethananyothermemberofLinnaea 1 Nny Pyck, Nancy . Few studiesdeal- (Rehder)Makino. P.O. Box 2437,BE-3001Leuven,Belgium (Fg 1). (Fig. Abelia L.,and omnctn Eio: ak Simmons Mark Editor: Communicating Heptacodium , , Triosteum andeither Heptacodium Abelia Leycesteria Zabelia Rehder Lonicera Zabelia Abelia and 2 Sz Huysmans, Suzy hasalwaysbeenassumedtobecloselyrelated orevencongenericwith assistertotheLoniceraclade,whereas morphologysuggestsacloserelation- L. , and , ,

231 Zabelia later decidedto assignthespeciesof in s. l.isoneoftheleaststudied. logeny of proven difficult toresolve, whichmakesunravelling thephy- relationships oftheLoniceraandLinnaeas.l.clades have Zhang etal.2002 ; Winkworth etal.2008 Backlund 1996 ). Theintergeneric ; Pyck andSmets2000 with theuncertainsystematicpositionof ; Donoghue etal. 2001 ; Wall. Severalmorphologicalandmolecularstudieshavedealt leaf morphologybetween (1966) . Additionally, Rehder (1916) stressed thesimilarityin ther elaboratedby Troll andWeberling (1966) and Weberling of tioned thesimilaritybetweenflowersandinflorescences placed H.miconioides separate from though itisquestionablewhethershouldberegarded as (1952) described asecondspecies, and Smets2000 ; Zhang etal.2002 ; Jin andLi2007 ). with aconspicuous,persistent,purplecalyx( Airy Shaw Hara 1983 ; Pyck ovary developsintoadry, single-seeded,long,slenderachene with onefertileandtwosterilecarpels. After fertilization,the The actinomorphicflowershaveaninferior, tricarpellateovary the originaldescriptionof Zabelia ogy of calyx are shared bytheLinnaeaclades.l. Achene morphol- pels, itsdry, single-seededachenes,andaccrescent, persistent 2002 ). logical incongruences (e.g. Pyck and Smets 2000 ; Zhang etal. atic positionof Diervilleae, Caprifolieae,andLinnaeeae.Theunclearsystem- era Heptacodium Of allthemajorDipsacaleslineages,Linnaea clade Heptacodium Abelia Sambucus . Eventhoughwoodanatomyof Heptacodium , bothmembersoftheLinnaeaclades.l.However, in Heptacodium Heptacodium Abelia sect. Heptacodium 1 See Jansen, Steven and and asthesolememberofgenus. Rehder (1916) Heptacodium , and Heptacodium H. miconioides Zabelia Zabelia and Viburnum ’s tricarpellateovarywithtwosterilecar- isvirtuallyidenticaltothatof inCaprifoliaceae,comprisingthegen- Lonicera evenmore challenging. ( Rehder 1911 ), but Makino (1948) . Morphologically, , hybridization isaplausiblescenario Zabelia Heptacodium Heptacodium isduetomolecularandmorpho- inadditiontotheformertribes and 3 . Inthisstudy, weconsider , 5 Zabelia n rk Smets Erik and Zabelia Leycesteria Heptacodium H. jasminoides

Abelia wasinitially included and Zabelia isuniqueinseveral , Rhe (96 men- (1916) Rehder , the Loniceraand Leycesteria formosa Heptacodium sect. , whichwasfur- and ’ relationship ’s p between Zabelia rphology Ar Shaw, Airy Zabelia 1 , Abelia 4 Abelia

as (e.g. . toa and

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 systematic position of of moleculardata. However, the authorstressed theuncertain she wasthefirsttoconfirmVerlaque’s hypothesisbymeans a phylogeneticstudyincluding both phylogenetic frameworkismissing. In2001,Pyckconducted study tacklingthesystematic positionof and branches( Ogata 1991 ). Currently, athorough molecular and thepresence ofsixdistinct,longitudinalgrooves ontwigs in severalaspects,suchasthedevelopmentofaggregate rays ( Verlaque 1983 ). In Dipsacales, thewoodof close relationship withtheMorinacladeinsteadof 1991 , tinct in 1992 ). Thepollenmorphologyof Pollen morphologyandwoodanatomy, forexample,are dis- 1952 ; Ikuse andKurosawa 1954 tional evidencejustifyingMakino’sdecision(e.g. ; Erdtman Hara 1983 separate genus, ; Ogata 1991 ). (C) basedonnuclearandplastidsequencedata.Bootstrapbranchsupportisindicatedforvaluesexceeding70%. clade andLinninaare depictedin Fig. 1B and C , respectively. B-C.MaximumparsimonyhypothesisoftheLoniceraclad the ML bootstrap valueandthesecondnumbertoBIposteriorprobability. Maximumparsimonyhypotheses(andsupportvalues) bootstrap support,whereas thelastnumberdesignatesBIposteriorprobability. Ifonlytwosupportvaluesare giventhent (bootstrap branchsupport)or0.95(posteriorprobability) are indicated.Thefirstnumberrefers toMP bootstrapsupport,the [Volume 36 SYSTEMATIC BOTANY 232 Fig. 1 1. A. Maximumlikelihoodhypothesis(ln Zabelia (Edmn 92; Vrau 18 gt 1988, Ogata 1983; Verlaque 1952; (Erdtman Zabelia Zabelia . Subsequentstudiesprovided addi- . Abelia L Zabelia Zabelia =-24,336)based onnuclearandplastidsequencedata.Supportvaluesgreater thanorequalto70% and Zabelia suggestsa Zabelia ina broad isunique Abelia , and ,

seed morphologicaldata. logical aspectofthestudycombinespollen,wood,fruit, and sequence dataofthemajorDipsacaleslineages.Themorpho- our studycomprisesITS, Dipsacales evolutioningeneral.Themolecularaspect of the systematicpositionsof plinary studyaimedat(1)improving ourunderstandingof Eriksson andDonoghue (1997) . Sequencedataofthree additionalspecies Zhang etal.(2003) , Gould andDonoghue (2000) , Kim and(1999) Donoghue (2003 , and , 2005 ), Donoghue etal. (2001 , Donoghue (2007) 2004 ), , Hidalgo etal.(2004) , Simonovik etal.(2007) , Jacobs 2010 ), Slanc etal.(2006) Smith (2009) , , data of46Dipsacales(Appendix1)wereTheis etal.(2008) obtainedfrom Bell and Jacobs etal.(2009) , ; Bell (2004 , 2007 ), Moore and oeua Methods— Molecular This investigation presents theresults ofaninterdisci- aeil ad Methods and Materials T, ITS, trnK trnK Heptacodium , , matK , , matK , , atpB-rbcL second numberrefers toML , , e (B)andtheLinninaclade atpB-rbcL and he firstnumberrefers to , and Zabelia fortheLonicera trnL-F , and sequence (2)and trnL-F

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 25 specimens,comprising 24speciesin12genera, wascollectedfrom 5 and6,respectively. characters andcharacterstates, matricescanbefoundinappendices (DELTRAN) transformationhypotheseswere evaluated. A descriptionof nuclear andchloroplast data.Bothaccelerated(ACCTRAN) anddelayed mony optimizationanalysesmade useofourML phylogeny basedon mented inMacCladev4.04( Maddison andMaddison2002 ). Theparsi- seed characterswasstudiedusingparsimony optimization(PO)asimple- forming aclade. (2) theLinnaeaclades.l.beingmonophyletic;and(3) 233 JACOBSET AL.: HEPTACODIUM AND ZABELIA White etal.(1990 ; ITS), Young etal.(1999 ; al. 2009 , 2010 ). Amplification andsequencingprimerswere adoptedfrom was isolatedusingamodifiedCTAB protocol ( Tel-Zur etal.1999 ; Jacobs et (Appendix 1)were addedbyusingthefollowingmethods.Total DNA 2011] bined dataset:(1) analyses aswellthree constrained ML hypotheses basedonourcom- for thesetestsincludedtheresulting MP, BI,andML hypothesesfrom our can result intype-1errors ( Shimodaira 2002 ). Thealternativehypotheses reduces thebiasofSH-test, althoughunderspecialcircumstances it in reducing type-1errors. The AU-test, ontheotherhand,significantly tion aspointedoutbyseveralstudies(e.g. Shimodaira 2002 ), itdoeswell ferences. Although theSH-testisheavily biased intermsoftree selec- ( Shimodaira andHasegawa2001 ) toassessthe significanceofthesedif- approximately unbiased(AU; Shimodaira 2002 ) testsusingConselv.0.1j the Shimodaira-Hasegawa(SH; Shimodaira andHasegawa 1999 ) andthe between andwithinthemajorCaprifoliaceaelineages.We madeuseof lyzing binarydata,indeldatawere excludedfrom the ML analyses. model. SinceRAxML-VI-HPCv.2.2 doesnotimplement amodelforana- bootstrapping wasdoneon2,000replicates withthesamesubstitution tion model,andeachmarkerplacedinaseparatepartition.Nonparametric random initialstartingtree foreachinference, GTRMIX setasthesubstitu- Each analysiscomprised1,000inferences ontheoriginalalignmentwitha quate. We usedRAxML-VI-HPCv.2.2 ( Stamatakis 2006 ) forML analysis. independent BayesianMCMCruns andshowedaburn-inof2,500wasade- ( Rambaut andDrummond 2007 ) wasusedtodetermineconvergence ofthe with twosimultaneousruns andasamplefrequency of100.Tracer v1.5 to variable.Bayesianinference analyseswere run for2,000,000 generations tion andanalyzedusinganF81-basedmodelwithascertainmentbiasset tition withpartitionsunlinked.Indeldatawere assignedtoaseparateparti- + IGmodelforthenucleardata.Eachlocuswasplacedinaseparatepar- selection. A GTR+Gmodelwasselectedforeachplastidmarkeranda implemented inMrModeltestv.2.2 ( Nylander 2004 ), were usedformodel criterion ( Akaike 1974 ) andthelikelihoodratiotest( Felsenstein 1988 and Huelsenbeck2003 ), as ) wasusedforBIanalysis.The Akaike information settings identicaltothoseoftheoriginalanalyses.MrBayesv.3.1 ( Ronquist addition replicates were carriedoutforeachbootstrap pseudoreplicate with were performedtoassessbranchsupport.Heuristicsearches of100random in effect. Bootstrap(BS; Felsenstein 1985 ) analyseson100pseudoreplicates trees heldateachstep,andtree-bisection-reconnection branchswapping tion replicates withallcharactersunordered andequallyweighted,five as theoutgroup. We conductedheuristicsearches of1,000randomaddi- Jacobs etal.2010 ), sevenmembersof Adoxaceae (Appendix1)were used 10 ( Swofford 2002 ). Basedonprevious studies(e.g. Winkworth etal.2008 ; plastid data;and(3)nucleardata. Felsenstein 1973 ) were usedtoanalyzethree datasets:(1)nucleardata;(2) ence (BI; Yang andRannala1997 ), andmaximumlikelihood (ML; number S10255). our analyses.Matricesandtrees were submittedtoTreeBASE (study and oftenopentodebate.We therefore chosetoomitITSindeldatafrom the ITSregion, homologyassessmentforindelsinthisregion iscomplex Ochoterena (2000) . As aresult ofthefrequency withwhichindelsoccurin characters followingsimpleindelcodingasdescribedby Simmons and terion ( Simmons 2004 ). Indelsofplastidregions were codedasbinary in GeneiousPro v.4.7 ( Drummond etal.2009 ) usingthesimilaritycri- and amaximumof25iterations.Thisalignmentwasmanuallyoptimized was usedforinitialsequencealignmentwithdefaultparametersapplied ing facilities(Macrogen, Seoul,SouthKorea). MUSCLEv4.0( Edgar 2004 ) 72°C. PurifiedPCRproducts were sequencedbytheMacrogen sequenc- sec, and72°Cfor30–90sec.Reactionsendedwitha3minincubationat lowed by30–35cyclesconsistingof95°Cfor30–90sec,50–52°C (PCR; 25µl)commencedwithaninitialdenaturationat95°C(180sec),fol- atpB-rbcL Mrhlgcl Methods— Morphological Reconstruction— State Ancestral Tree topologiesbasedonthevariousdatasetsimplieddifferences Maximum parsimonyanalyseswere conductedusingPAUP* v.4.0 beta Analyses— Phylogenetic ), and Taberlet etal.(1991 ; Heptacodium Maximum parsimony(MP),Bayesianinfer- For ourpalynologicalstudy, materialof andtheLinninacladebeingmonophyletic; Evolution ofpollen,wood,fruit, and trnL-F trnK ). Polymerasechainreactions , , matK ), Manen etal.(1994 ; Abelia and Zabelia

JSM-6360 scanningelectron microscope. sputtered withgold.Observationandphotographingwasdonea After anultrasonictreatment, thematerialwascriticalpointdriedand SEM, thesameprotocols were appliedwiththefollowing modifications. a PL-B622CFPixeLINKdigitalcamera(PixeLINK,Ottawa,Canada).For and photographedwithaLeitzDialux20lightmicroscope equippedwith tions (5µm). After dyingwithtoluidineblue(0.01%),slideswere observed (Microm, Walldorf, Germany)wasusedtocutlongitudinalandcross sec- Histo-Technik, Wehrheim, Germany). An HM360rotary microtome dehydrated inanethanolseries,andembeddedTechnovit 7100(Kulzer glutaraldehyde (2.5%)buffered withsodiumcacodylate buffer (0.05M), or collectedinthefield(Appendix4).ForLM,materialwasfixed through collaborationwithseedbanks,botanicalgardens, andherbaria, tomical datafor features forhardwood identification( IAWA Committee1989 ). Wood ana- digital camera.Wood terminologyfollowstheIAWA listofmicroscopic done withaLeitzDialux20lightmicroscope equippedwithaDP50-CU mounted inEuparal( Jansen etal.1998 ). Observationandphotographywas blue (35/65),dehydratedinanethanolseries,cleared withParasolve,and washed indeionizedwater, stained withamixture ofsafraninandalcian transverse andlongitudinalsections(20µm).Sectionswere bleached, eter). We usedasledgemicrotome (Reichert,Vienna, Austria) tomake wood sampleswere collectedfrom fairlythinbranches(<2cmindiam- were collectedfrom herbariaandlivingcollections(Appendix3).Most available ( Blackmore andCannon1983 ; Verlaque 1983 ). The Morinacladewasnotsampledsincepalynologicaldataisreadily Punt etal.(2007 ; of tenpollengrains.Pollenterminologyfollowstheonlineedition http://www3.bio.uu.nl/palaeo/glossary/index.htm ). to measure thepolaraxis(P)andequatorialdiameter(E)ofaminimum A LeitzDialux20lightmicroscope (Leitz,Wetzlar, Germany)wasused Germany). We observedtheinnerloculewallfororbiculepresence. Germany) equippedwithaDP50-CUdigitalcamera(Olympus,Hamburg, were observedwithaLeitzDialux20lightmicroscope (Leitz,Wetzlar, (Jeol, Tokyo, Japan).Forlightmicroscopy (LM),glycerinejellyslides photography wasdonewithaJSM-5800-LV scanningelectron microscope with gold(Spi-Supplies,West Chester, Pennsylvania).Observationand washed withethanol(70%),mountedonstubs,air-dried, andsputtered in aceticacid(100%).Forscanningelectron microscopy (SEM),pollenwas sis, flowerswere rehydrated inan Agepon solution(1:200)anddehydrated lyzed followingReitsma’s(1969)wettingagentmethod.Priortoacetoly- herbaria andlivingcollections(Appendix2). All specimenswere aceto- showed our plastiddata setrejected theBItopology (nuclear),which not sistertotheValeriana clade.Incontrast, testsbasedon data setrejected alternativehypothesesinwhich rejected either. The AU- andSH-tests basedonournuclear Heptacodium clade s.l.wasnotrejected byourtests.Thepossibility of Abelia the AU- andSH-testsclearlyrejected thehypothesisinwhich exclusion of by (1)incongruences betweenITSandplastiddata;(2)in-or SH-tests ( Table 1 ) showedthatincongruences were caused Diervilla, Lonicera,andLinninaclades. latter analysisplaces the exceptionofMP analysisbasedonnucleardata.The Heptacodium monophyly oftheLinnaeaclades.ismoderatetostrong. any othermemberoftheLinnaeaclades.l.Supportfor more closelyrelated totheMorinaorValeriana cladesthanto Linnaea clades.l.isparaphyleticinallanalyseswith lower fortheMP topology basedonnucleardata).The strong supportinouranalyses (bootstrapvaluesare slightly clade s.l.,allmajorcaprifolioidlineagesreceive moderateto Fruits and/orseedsof39species,comprising26genera,were acquired Wood samplesof28specimens,comprising26speciesin12genera, Visual inspectionoftheresulting topologiesand AU- and Analyses— Phylogenetic and Zabelia Zabelia formingacladewiththeLinnina cladewasnot issistertotheLoniceracladeinallanalyseswith Zabelia Linnaea borealis assistertotheValeriana clade.Thisillustrates formedaclade,monophyleticLinnaea ; and(3)the optimalitycriterion. Although Heptacodium wasobtainedfrom Greguss (1959) . Results With theexceptionofLinnaea inapolytomywiththe Zabelia Zabelia was

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 tionship between BS) andposteriorprobability (1.00PP)support.Thecloserela- ine ld .l oohltc1 .804 .508 .50.98 0.65 0 0.88 0.15 7 0.42 0.28 13 Linnaea clades.l.monophyletic mative. Missing dataaccountsfor28%ofthe nucleardataset characters ofwhich 367are variableand 290parsimonyinfor- ing bothlineagesastheDipsacus clades.l. support ( Fig. 1A ), andwe therefore refer tothecladecompris- lifera Fig. 1A ). Ouranalysesconsistentlyrecover sister totheremainder ofCaprifoliaceae(100%BS;1.00 PP; (unresolved), allanalysesrecover theDiervillacladeas 1A-B ). Apart from theMP analysisbasedonnucleardata ity criterionandtheinclusionorexclusionofindeldata( Figs. positions of ered byourcombinedanalyses( Fig. 1A ), whilethesystematic Leycesteria clade, followedby recovers ing lineageintheLoniceraclade. Analysis of plastiddata although BIandML showthat to resolve theintergeneric relationships intheLoniceraclade, Valeriana officinalis form aclade,whichissisterto Heptacodium Abelia core valerians, followed by Juss. isrecovered asthesistertoremainder oftheclade, relationship islow, though.IntheValeriana clade, 88 Scabiosacolumbaria M. Cannonand 0.93 M. Cannonisrecovered assisterto 88 [Volumeslightly betweenanalyses. 36 0.85 in theMorina,Dipsacus,andValeriana cladesdiffer only tionships inthemajorcaprifolioidlineages.Relationships 0.70 Zabelia caprifolioid lineagesandtheirrelation to 0 0.31 tid data. once more theincongruences between thenuclearandplas- 6 nuclear SYSTEMATIC BOTANY nuclear ML BI MP are indicatedinbold. 234 between based onnucleardata,however, resolves asisterrelationship Moench and Ncer aa Set— Data Nuclear Before wefocusontherelationships betweenthemajor Table Wall. exDC.assistertotheDipsacuscladewithstrong and , wegiveabriefoverviewoftheintergeneric rela- . 1. Comparing topologiesusinglikelihoodsandapproximately unbiased(AU)andShimodaira-Hasegawa(SH)tests.Significantt Lonicera Zabelia S. columbaria and n inn oohltc2 .701 401 .91 .00.66 0.10 17 0.69 0.14 14 0.19 0.07 20 andLinninamonophyletic Lonicera Nardostachys grandiflora Succisella inflexa monophyletic obnd1 .502 .2 0.92 0.62 0.47 0 0.62 3 0.62 0.26 0 0.75 0 0.15 0.26 0.06 0.26 21 16 0.15 7 0.15 16 0.27 21 16 combined 21 0.22 plastid combined noindels combined 15 plastid noindels plastid combined noindels nuclear obnd1 .003 0 30 0.30 0.20 15 34 33 combined plastid noindels plastid Symphoricarpos Fedia cornucopiae Morina longifolia topology asthesistertoremainder oftheLonicera Leycesteria L.( Fig. 1A ). Nucleardataaloneare unable L.appearstobesister Triosteum and and The nucleardatasetconsistsof795 Triosteum S. pratensis Acanthocalyx albus and . Thesister relationship between Beck( Fig. 1A ). TheMP analysis receives highbootstrap(>85% Centranthus ruber Gaertn.and Wallich exDC.( Figs. 1A , C ). Triosteum Symphoricarpos Cryptothladia chinensis changewiththeoptimal- DC.( Fig. 1A ). Insidethe . Supportforthelatter D 23 43 lnL isthefirstdiverg- Triplostegia glandu- Heptacodium Valerianella Succisa pratensis (Hand.-Mazz.) isalsorecov- ML nuclear (L.)DC.and < 0.05 < 0.05 < 0.05 < 0.05 < 0.05 <0.05 <0.05 AU 01 .609 .70.89 0.07 0.89 0.89 4 0.07 0.07 0.97 4 4 0.76 0.97 0.97 0 0.77 0.77 0.11 0 0 0.11 0.11 00 .608 4 14 0.82 0.16 3 24 0.05 0.05

Patrinia Mill. (Pai) and <0.05 SH 0.13

(1.00 PP).IntheLoniceraclade, Diervilla cladeisthesistertoremainder ofCaprifoliaceae resolved inboththeLinnina andtheLoniceraclade.The (< 60%BS).TheBItopologybasedonnucleardataismore clade (100%BS); (74% BS); polytomy withallmajorlineagesispresent: Linnaeaclades. Lonicera in theLoniceracladeare unresolved andthemonophylyof and (4)theLinninaclade(91%BS).Intergeneric relationships BS); (2)Loniceraclade(69%(3) polytomy offourcladesisrecovered: (1)Diervillaclade(100% all trees withatree length a tree lengthof1,070,we baseourstrictconsensustree on 1989 ] =0.75).SincetheMP analysisalsorecovers 49trees with and Farris1969 ] =0.56,ensembleretention index[RI; Farris a tree lengthof1,069(ensemble consistencyindex[CI; Kluge v.4.0 beta10).TheMP analysisgeneratesoneshortesttree with (including gaps,whichare treated asmissingdatabyPAUP* to apolytomyof BS) toacladecomprising clade s.(unsupported)andthisentire cladeissister(98% the MorinaandDipsacuss.l.cladesare sistertotheLinnaea ogy. However, insteadof thetrichotomyinLinninaclade, clades (0.93PP).TheML topologyissimilartothe BItopol- Valeriana clade(0.74PP);and(3)MorinaDipsacuss.l. comprises: (1)Linnaeaclades.(0.91PP);(2) Linnina (<0.95PP).Intheclade,abasaltrichotomy sister totheLoniceraclade(0.89PP)andbothwere sisterto and theremaining Caprifoliaceae, theDiervillacladeissisterto theremainder of est trees with alengthof2,358(CI= 0.76,RI=0.86).In v.4.0 beta 10).Maximumparsimonyresults infiveshort- (including gaps,whichare treated asmissingdatabyPAUP* mative characters,whereas missingdataaccountsfor 30.2% informative. Indeldataaccounts for158parsimonyinfor- characters ofwhich1,473were variableand959parsimony H.miconioides no support(<50%BS).Supportforasisterrelationship of sister relationship of Patd aa Set— Data Plastid D 293 lnL 16 and Zabelia Triosteum ML plastid andtheLoniceracladeislow(<60%BS). <0.05 <0.05 < 0.05 < 0.05 < 0.05 < 0.05 AU (100%BS);Morinaclade(90%Dipsacus Triplostegia glandulifera Leycesteria

02 800 0.34 0.05 28 0.26 03 7 37 0.33 Lonicera isunsupported.IntheLinninaclade,a Zabelia The plastiddatasetconsistsof3,908 <0.05 <0.05 < 0.05 < 0.05 < 0.05 < 0.05 < 0.05 Zabelia SH seis species. £ , , 1,070.InCaprifoliaceae,abasal Symphoricarpos andtheValeriana cladegains

008 0.95 0.95 0.84 0.95 0.84 0 0.84 0 0 8 80 82 andtheValeriana clade.The Triosteum D Heptacodium miconioides 314 114 lnL Heptacodium miconioides ; andValeriana clade issister(0.95PP) , , ML combined Lonicera dioica <0.05 < 0.05 <0.05 < 0.05 < 0.05 < 0.05 AU Zabelia

0.62 0.24 est results <0.05 < 0.05 < 0.05 <0.05 and SH L., L., is is ;

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 lower, thoughnotsignificantly. alter thetopologyinanyway. Support, however, isslightly PP; <60%BS; Fig. 1A ). Linnaea clades.s.,and data results inlowersupport forthesisterrelationship ofthe 1A ). Itisnot surprising, however, thattheinclusionofnuclear cal totheBIandML topologiesbasedonplastiddata( Fig. BS; Figs. 1A , C Dipsacus clades.l.remains sistertotheValeriana clade(93% ). BIand ML topologies,however, are identi- a cladeholding sister totherest oftheLinnina clade(100%BS),followedby plastid datainoneaspect,thatis,theLinnaeaclades.is sus topology( Fig. 1C ) differs from theMP topologybasedon 0.70, RI=0.82). As regards theLinninaclade, theMP consen- ysis results intwoshortest trees withalengthof3,436(CI= treated asmissingdataby PAUP* v.4.0 beta10).OurMP anal- 30.0% ofthecombineddataset(includinggaps,whichare informative characters,whereas missingdataaccountsfor simony informative.Indeldataaccountsfor158parsimony 4,703 charactersofwhich1,840are variableand1,249par- Morina clade,and(3)theDipsacuss.l.Valeriana clades. omy comprising(1)theLinnaeaclades.s.,(2) omy intheLinninacladecollapsesandturnsintoatrichot- consensus topology. BasedonMP, however, thebasaldichot- sion ofindeldatafrom ourBIanalysisdoesnotalterthe Zabelia for thesisterrelationships betweentheLinnaeaclades.s.,and are identicaltothoserecovered byMP. Supportishighexcept Relationships betweenthemajorlineagesinLinninaclade and theLoniceracladeiswell-supported(1.00PP;91%BS). BI andML topologies,thesisterrelationship of the Dipsacuscladereceives strong support(95%BS). Inthe Orbicules are absentontheinnerloculewall. in sexinepatterntowards thepores orthepoles( Figs. 2A , C ). that are about2µmlong( Fig. 2C ). There isnodifferentiation pattern isminutelyperforateandechinate,withrobust spines granular nexineisobservedaround theectopori.Thesexine lum. Eventhoughadistinctendoaperture islacking,azoneof the polararea isprotruding, thisisnotcausedbyavestibu- nants ofthepore membraneatthemargins ( Fig. 2C ). Although 3-zonoporate ( Fig. 2B ), withthepores beingsmallandrem- in polarviewismore orlesstriangular( Fig. 2B ). Pollenis spherical toellipticalinequatorialview( Fig. 2A monads, are large (averageP 71µm,averageE79µm)and ). Shape 01 JCB TA. ETCDU N AEI 235 JACOBSET AL.: HEPTACODIUM AND ZABELIA clade s.issister(69%BS)to maximum support(100%BS).Inthelatterclade,Linnaea Valeriana clade,andtherest oftheLinninacladereceives split betweenacladewiththeDipsacuss.l.and moderate support(88%BS).IntheLinninaclade,abasal relationship oftheLoniceracladeand miconioides the family(100%BS)andLoniceracladeplus 2011] imum support.Thesisterrelationship of (95% BS).Themonophylyofallmajorlineagesreceives max- The relatively shortectocolpihavestraight, distinctedges at thecorners( Fig. 2D ). Pollenis 3-zonocolporate ( Figs. 2D-E ). view. Shapeinpolarviewistriangularwith apertures located (average P 79µm,averageE70µm)andprolate inequatorial Palynology— Omitting indeldatafrom theMP andBIanalysesdoesnot Set— Data Combined Although posteriorprobabilities are slightlylower, exclu- Z ABELIA andtheMorinaclade(<60%BS;<0.95PP). — issistertotheLinninaclade(64%BS).The Pollen grains,dispersed asmonads,are large H Zabelia EPTACODIUM andtheMorinaclade(68%BS).The Zabelia The combineddatasetconsistsof — Pollen grains, dispersedas plustheMorinaclade(<0.95 Zabelia andtheMorinaclade H. miconioides T. glandulifera Zabelia H. miconioides Heptacodium and the gains and occidentalis the tectum.Few species( terized byacombinationofspines andsmallperforationsin 2M), ( Fig. Several taxa( ing ontheinnerloculewall. out supratectalprocesses ( Fig. 2F ). Orbiculesare entirely lack- the ecto-andendoaperture. Thesexinepatternispsilatewith- bulges outslightly( Fig. 2E ). A fastigiumislackingbetween On SEMimages,theendocingulumarea attheequatoroften a lolongateorrectangular openingmaybevisible( Fig. 2E ). tion oftheendocingulumandectocolpi(atequator), lum attheinnernexinewalllayer( Fig. 2E ). At theintersec- finely granular. All speciesinvestigatedhaveanendocingu- and mostlyacuteends( Figs. 2D-E ). Thecolpusmembraneis 1.14 ( remaining taxarangesfrom 0.86( ing anoblateshapeinequatorialview. TheP/Evaluesofthe made aspollenwasonlyvisibleinpolarview(LM),indicat- and tematic significanceofpollensizeistherefore limited. variation insizeisobservedin,forexample, Buchw.) Makino(averageP 88.5µm).However, interspecific Wall. (averageE92.5µm)and largest pollengrainsare observedin Weigela decora phoricarpos orbiculatus overview ofthepalynologicaldataoursampling. cies ofallrelevant lineages.We refer to Table 2 pollen variationinDipsacales,wesampledanumberofspe- foradetailed (cf. Figure 2B ). Thisisnottrue for angular andiscausedbyprotruding apertures atthecorners torial view. The amborshapeinpolarviewisoftensubtri- Dipeltafloribunda orbiculatus are observed( tures. Insomespeciesoccasional grainswithfourapertures Linnaeaborealis study. A fastigium/vestibulumispresent in underlines thatsoundobservationsrequire athorough LM region oftheendoaperture ( Punt etal.2007 ; Figs. 2G , N).This of theinnerpartexinefrom thedomedsexinein It istherefore thesamestructure appearingasaseparation whereas inpororate grainsitisreferred to asvestibulum. pollen grains,thecavityisdesignatedbytermfastigium, ence ofacavityunderneaththeectoaperture. Incolporate A feature thatdeservesspecialattentionisthe possiblepres- shaped apertures, and ectopori, thatis, erture isgenerallyacolpus ( Figs. 2I-K ). Onlytwotaxahave Diervilla Bureau &Franch.and tigated havesupratectalspines exceptfor in Dipsacalesisechinate( Figs. 2A-C , G,I-O). All speciesinves- (e.g. it isusuallyashortcolpusperpendiculartotheectocolpus is oftendifficult toassess.Whenanendoaperture ispresent, ( Fig. 2N ), and To situatethepollenof Shape— Size— Apertures— Sexine— Dipelta floribunda Lonicera Zabelia umbellata , , The smallestpollengrainsare observedin Leycesteria , , Hook.,and Dipelta For severaltaxa(e.g. The mostcommonsexineornamentationpattern Abelia xgrandiflora and (Nakai)Nakai(averageP 40.0µm),whereas the Leycesteria formosa Weigela . Lonicera syringantha All speciesinvestigatedhavethree aper- ( Fig. 2H ), and ( Fig. 2O ), and Leycesteria Diervilla , , Lonicera Maxim.),P measurements couldnotbe (+/−).Thepresence ofanendoaperture ), indicatingasphericalshapeinequa- Moench(averageE42.0µm)and Zabelia Triosteum himalayanum Weigela Lonicera syringantha Heptacodium Mill.( Fig. 2L ) withshortcolpus- ) oraporus asin (+/−), ( Figs. 2D-F ) whichare psilate. (Rovelliex André) Rehder, Wall., Thunb.withgenuinepori. Zabelia umbellata Linnaea borealis Weigela coraeensis Linnaea borealis Abelia Symphoricarpos Maxim., Symphoricarpos inthelarger contextof Triosteum himalayanum Triosteum himalayanum (Fg 2 , 2G), (Fig. Lonicera thibetica Zabelia Diervilla , , Abelia ) havearough Symphoricarpos Symphoricarpos ). Theectoap- ) are charac- (Graebn.& Dipelta , , Thunb.)to , , (Fg 2G), (Fig. Triosteum . Thesys- Triosteum . Sym- , and , ,

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 sexine ismicroperforate withsupratectalspinesofvaryingsize. D. ing porivisible.B.Pollengraininpolar view, notedifferentiated nexineornamentation inaperturalregion (vestibulumabsen [Volume 36 SYSTEMATIC BOTANY 236 E-F. of cavitybelowectoaperture, i.e.fastigium. O. Equatorial viewon acharacteristicprotruding aperture. M,N. tation. G. at equatorinpolarview, pollengrainwithfourapertures. I. amabilis Fig. Zabelia triflora 2 2. Palynology (A,C-F, I-M,O,SEM;B,G-H,N. A-C, LM). . Equatorialviewonmesocolpiumwithevenlydistributed spinesontectum.K. Abelia sp.Polarview, notethree extruding apertures withacavitybetweenthewalllayers,i.e.fastigium.H. . E.Detailofectocolpus.Noteslightlyelevatedarea atequatorindicatingpresence ofendocingulum. F. Detailofpsilatese Dipelta ventricosa Dipelta ventricosa Triosteum himalayanum . Detailofechinate sexineornamentation. Zabelia taihyonii Heptacodium miconioides . Equatorialviewonectocolpus,echinate sexineornamentation.J. . Pollen graininpolarviewwiththree slightlyprotruding colpivisible. . M. Short protruding ectocolpuswithfastigium beneath.N.Detail Lonicera involucrata . A. Pollengrainin equatorialviewwithtwoslightlyprotrud- . Detailofectocolpus. L. Dipelta floribunda t). C.Detailofectoporus, Diervilla lonicera . Opticalsection xine ornamen- Kolkwitzia . Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 01 JCB TA. ETCDU N AEI 237 JACOBSET AL.: HEPTACODIUM AND ZABELIA 2011]

Table 2. Summary of pollen characters of the sampled species. P: polar axis, E: equatorial diameter, EC: echinate, PF: perforate, G: gemmate, GR: granulate, R: rugulate, PS: psilate. Missing data are represented by a dash (not for “fastigium/vestibulum”).

Taxon/Character P (µm) E (µm) P/E number of apertures ectoaperture endoaperture fastigium or vestibulum sexine supratectal elements Abelia spathulata 59-(65)-72 58-(64)-71 1.00-(1.03)-1.09 3 colpus ? + EC spines Abelia x grandiflora 68-(71)-75 68-(72)-77 0.93-(0.99)-1.04 3-(4) colpus colpus + EC spines Diervilla lonicera 46-(49)-55 49-(52)-57 0.91-(0.95)-0.98 3 colpus-porus porus + EC, G, GR gemmae, spines, granules Diervilla sessilifolia 45-(48)-51 45-(49)-53 0.92-(0.97)-1.04 3 colpus-porus porus + EC, G, GR gemmae, spines, granules Dipelta floribunda - 63-(72)-78 - 3-(4) colpus - - EC, (PF)(micro)spines Dipelta ventricosa 51-(57)-61 55-(58)-63 0.93-(0.99)-1.05 3 colpus - - EC, PF(micro)spines Heptacodium miconioides 63-(70)-79 68-(76)-84 0.86-(0.92)-1.00 3 porus - - EC, PF spines Heptacodium miconioides - 77-(80)-83 - 3 porus granular zone around porus - EC, PF spines Kolkwitzia amabilis 60-(63)-68 54-(57)-62 1.05-(1.10)-1.14 3 colpus - - EC (micro)spines Leycesteria formosa 57-(59)-61 59-(60)-62 0.95-(0.98)-1.03 3 colpus colpus + EC, PFspines Linnaea borealis 42-(44)-49 43-(46)-50 0.86-(0.96)-1.06 3-(4) colpus - - EC, PF(micro)spines Linnaea borealis 50-(52)-55 49-(52)-55 0.95-(1.00)-1.04 3 colpus - - EC, PF(micro)spines Lonicera involucrata 56-(57)-58 59-(59)-60 0.93-(0.95)-0.97 3 colpus colpus - EC (micro) spines Lonicera syringantha 56-(57)-60 56-(57)-63 0.95-(1.00)-1.07 3-(4) colpus colpus - EC, Rspines Lonicera thibetica 44-(47)-49 45-(46)-48 0.98-(1.01)-1.04 3 colpus short colpus - PS - Symphoricarpos occidentalis - 52-(55)-58 - 3 colpus - + EC, R (micro)spines Symphoricarpos orbiculatus - 40-(42)-45 - 3-(4) colpus - + EC (micro)spines Triosteum himalayanum - 85-(92)-97 - 3 colpus - + EC, PF, R spines Triosteum perfoliatum - ±80 - 3 colpus colpus + ? spines Weigela coraeensis 46-(47)-50 50-(54)-57 0.81-(0.86)-0.93 3 porus - + EC, GR (micro)spines, granules Weigela decora 35-(40)-45 39-(44)-47 0.90-(0.92)-0.96 3 porus - + EC, G, GR gemmae, spines, granules Weigela maximowiczii - ±60 - 3 porus - - EC, G, GR gemmae, spines, granules Zabelia taihyonii 73-(81)-87 65-(71)-79 1.05-(1.14)-1.21 3 colpus endocingulum - PS - Zabelia triflora 57-(65)-72 55-(60)-65 1.02-(1.08)-1.17 3 colpus endocingulum - PS - Zabelia umbellata 78-(88)-95 72-(77)-84 1.07-(1.14)-1.19 3 colpus endocingulum - PS - Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 of uprightand square marginal cells. Perforatedraycells of theraycells are procumbent orsquare withseveralrows (average 934µm).Raysare sometimesaxiallyfused.The body are 2–5-seriate( Fig. 3F ) withheightbetween400and2,200µm the bodyoffibre-tracheids. Axial parenchyma isdiffuse. Rays (average 890µm).Helicalthickenings are present throughout are thick-walledwithfiberlengthbetween650and1,320 µm have distinctlybordered pits(5–7µmindiameter).Fibers with narrow, fibriform-likevesselelements.Fiber-tracheids lar tracheidsare sometimespresent. Ifpresent, theyintergrade between 340and960µm(average532µm).Vasicentric/vascu- (average 113 vesselspermm [Volume 36 SYSTEMATIC BOTANY tapetum. trace ofsporopollenin polymerisationontheremnants ofthe ence oforbiculesintheanthers.Nonethemshowany with bothsupratectalspinesandgemmaeand/orgranules. Diervilla tectum surfacebetweenthespines,designatedasrugulate. 238 ranging from 140–240vesselspermm to 50µm(average28indiameter).Vessel densityishigh, ment tails.Thetangentialvesseldiameterrangesfrom 17µm helical thickeningsare mainlyfoundinthenarrow vessel ele- pitting isbordered orwithslightlyreduced borders. Slender ing from 6µmto15(average9indiameter).Vessel-ray xylem. Intervesselpitsare alternateandnonvestured, rang- perforation platesare occasionallypresent nearthe primary perforation platesare simple( Fig. 3B ), althoughscalariform solitary andsomewhatangularinoutline( Fig. 3A sity andfiberthickness.Thewoodisdiffuse-porous, vessels ). Vessel are distincttoindistinct,markedbydifferences invesselden- Vessel densityishigh,rangingfrom 76–172vesselspermm diameter rangesfrom 11–50 µm(average24indiameter). ent innarrow vesselelements( Fig. 3E ). Thetangentialvessel reduced borders. Helicalthickeningsare abundantlypres- in diameter).Vessel-ray pittingisbordered orwithslightly nate andnonvestured, rangingfrom 5–9µm(average7 foration platesare simple( Fig. 3E ). Intervesselpitsare alter- wood isring-porous withsolitaryvessels( Fig. 3D ). Vessel per- by differences invesseldensityanddiameter( Fig. 3D ). The and 840µmin ate withheightbetween250and1,900µm(average689,807, parenchyma issparselydiffuse ( Fig. 3A ). Raysare 3–4-seri- present throughout thebodyoffiber-tracheids ( Fig. 3C ). Axial in 1981–0475–2 ( between 410and620µm(average511 µm)in 700 µm(average590µm)in per mm and 1,300µm(average840µm)in Fibers are thin-tothick-walled withfiberlengthbetween700 have distinctlybordered pits(5–7µmindiameter; Fig. 3C ). 439 µm)in crystals percell( Fig. 3C ). 17 rayspermm).Someraycellscontainonelarge prismatic density ishighandrangesfrom 14–21rayspermm(average ( Fig. 3B ). Perforatedraycellsare simplein square withseveralrows ofuprightandsquare marginal cells axially fused.Thebodyoftheraycellsare procumbent or 4041, Coombes 1981–0475 od Anatomy— Wood Orbicules— Z H. miconioides ABELIA and ( Fig. 2L ) and 2 ). Meanvesselelementlengthisbetween400and — Pyck 92–0130–16 Growth ringboundariesare distinct,marked H. miconioides ), andbetween700980µm(average870µm) H. miconioides All specimenswere investigatedforthepres- ( Pyck 92–0130–16 ), andbetween300600µm(average H EPTACODIUM Weigela ( , respectively). Raysare sometimes Pyck 92–0130–16 2 ( H. miconioides ). Meanvesselelementlengthis Coombes 1981–0475 share adistinctexinepattern Got rn boundaries ring — Growth ). Helicalthickeningsare H. miconioides 2 (average195vessels H. miconioides ( ) Fiber-tracheids ). Pyck 1987–4041 H. miconioides , , Pyck 1987– ( Coombes . Ray . ), 2

exceeds 1,100µm. between 600–1,000µm.In cially commonin septate andthin-tothick-walled. Thin-walledfibersare espe- reduced tosimplepitsin in bothradialandtangential walls.Libriformfibershave ranging from 4–8µmindiameter. Fiber pitsare common intergrade withnarrow, fibriform-likevesselelements. vascular tracheidsare sometimespresent. Ifpresent, they (390 µm)and vessel elementlengthislessthan400µmin than 100vessels/mm nigra age 33µm)were measured in average valuesbetween20and30µm.Wider vessels(aver- tangential diameterofvesselluminaisusuallysmallwith in mosa L.crocothyrsos is mostlybetween400and700µm,in of absent. Thelongitudinalgrooves thatcharacterizethebranches per mm(average14raysmm).Mineralinclusionsare are simple.Raydensityishighandrangesfrom 11–22 rays encountered in generally diffuse-porous ( Fig. 3G , I,M),ring-porous woodis distinctly present in ( Fig. 3G ), and They are absentin preceding paragraphs. descriptions astheirwoodanatomyhasbeendescribedinthe Heptacodium the woodanatomyofDipsacales,werefer to Table 3 as theyare (mostly)herbaceous.Foradetailedaccountof and Valeriana cladesaswell omy ofthemajorDipsacaleslineages.TheMorina,Dipsacus, . as smallgapsinthegrowth ringsofitsring-porous wood. Diervilla Symphoricarpos scalariform in mon (e.g. Fig. 3I , M).Vessel perforationplatesare exclusively index 1.5).Inseveralspecies,anangularvesseloutlineiscom- in shortradialmultiples two vesselsoccurin sively solitary( Fig. 3G , H,I,M)Occasionalradialmultiplesof semiring porous. crata Leycesteria Dipeltaventricosa thickenings are apparent invesselsandtracheidsof ray pittingwasobservedin distinct borders orslightly reduced borders. Simplevessel- largest exceed10µm.Pitsare nonvestured. Vessel-ray pitshave lariform. Thesmallestintervesselpitsare 4µm,whereas the ally alternate( Fig. 3K ), theyare sometimesoppositetosca- Symphoricarpos simple withabout5%scalariformperforationsin are exclusivelysimplein ally more than20barsper perforation plate.Vessel perforations Fibers— Got Rings— Growth In thefollowingparagraphs,wedescribewoodanat- Vessels— Zabelia L. syringantha . Thewoodof , , (average45µm).Vessel densityishigh(usuallymore L. crocothyrsos , , correspond tofurrows inthexylem,clearlyvisible Sambucus nigra , , Fiber pittingisusuallydistinctlybordered ( Fig. 3O ), Viburnum opulus and Vessels are exclusivelysolitarytoalmostexclu- Symphoricarpos orbiculatus , and ( Fig. 3O ). Helicalthickeningsare absentin ( Fig. 3K ). Although intervesselpitsare gener- Weigela maximowiczii Abelia Symphoricarpos ), and , , Zabelia Kolkwitzia amabilis Diervilla Weigela coraeensis Lonicera thibetica , n and , Growth ringsare distincttoindistinct. 2 Viburnum opulus Diervilla sessilifolia , , Symphoricarpos ). Although meanvesselelementlength Viburnum opulus Diervilla willnotbeincludedinthefollowing , , Weigela coraeensis Viburnum opulus and Sambucus nigra , and Sambucus nigra Abelia grandiflora Sambucus nigra Sambucus nigra Viburnum opulus ( Fig. 3H ), and , , Weigela Triosteum Weigela Dipelta , itisabout1,000µm.Mean and , , (S.Moore) Rehder, and (Fg 3 , 3H), (Fig. Leycesteria . Vessels are solitaryor (325µm).Vasicentric/ . Meanfiberlengthis L. Although wood is , , Buckley, ( Fig. 3J ) withgener- Sambucus nigra were notincluded Kolkwitzia amabilis orpredominantly Leycesteria formosa , meanfiberlength , and (Fg 3 . Helical 3N). (Fig. . Fibersare non- (vesselgrouping Lonicera thibetica , , Lonicera involu- and Leycesteria for- , , Lonicera Lonicera Weigela Lonicera Leycesteria Sambucus Abelia . The . (not L.is , and , and , , ,

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 1–4-seriate rays,fiberswithdistinctly bordered pits.G. H. 239 JACOBSET AL.: HEPTACODIUM AND ZABELIA 2011] longitudinal groove inthetwig.E-F. prismatic crystals (PC)andhelicalthickeningsassociated withfiber-tracheids (arrows). diffuse to diffuse-in-aggregates axialparenchyma. N. wide earlywood vessel (right).L. coraeensis tracheid (centre), prismaticcrystalinraycell(rightarrow). D. simple perforationplates.C.Tangential longitudinalsection(TLS),fiber-tracheid pitswithadistinctpitborder (leftarrow Fig. Symphoricarpos albus 3 3. Wood anatomy(SEM). A-C. . RLS, scalariformperforationplateswith>40barsperplate. K. . TS,ring-porous woodwithsolitaryvessels. I. Diervilla sessilifolia Zabelia taihyonii Heptacodium miconioides . TLS,1–2-seriaterays.M. Sambucus nigra . E.RLS,narrow vesselselementswithsimpleperforationplatesand helicalthickenings.F. TLS, Leycesteria crocothyrsos . A. Transverse section(TS),solitary, narrow vessels.B.Radiallongitudinalsection(RLS), Zabelia umbellata . RLS,simplevessel-ray pitting.O. Viburnum opulus . TS,diffuse-porous woodwithsolitary, narrow vessels, growth ringsabsent. Dipelta ventricosa . TS,firstgrowth ringnearpithwitha“gap”(arrows) formedbya Symphoricarpos albus . TS,indistinctgrowth rings,solitary, narrow vessels. J. . TS,solitary, narrow vesselswith angularvesseloutline, Symphoricarpos orbiculatus . RLS, simpleperforationplates,vesselpittingin ), helicalthickeningsinfiber- . TLS,raycellswith two Weigela Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 4 SSEAI OAY [Volume 36 SYSTEMATIC BOTANY 240

Table 3. Summary of wood characters of the sampled species. V-V pits = intervessel pits; V-R pits = vessel-ray pits; FP = fiber pits.

Taxon/Character growth porosity vessel groupings angular perforation plates V-V pits V-R pitting helical thickenings helical thickenings vessel vessels/mm2 vessel element rings vessels in vessels in tracheids/ fibers diameter length Abelia chinensis +/− diffuse solitary - scalariform alternate bordered - + 15-(25)-30 50-(60)-70 450-(644)-800 Abelia graebneriana +/− diffuse solitary - scalariform scalariform/ bordered + + 14-(21)-30 180-(219)-264 450-(651)-800 opposite Abelia x grandiflora +/− diffuse solitary - scalariform alternate bordered - + 15-(21)-25 33-(66)-82 600-(790)-1,000 Abelia spathulata +/− diffuse solitary - scalariform alternate/ bordered +/− + 10-(19)-25 120-(167)-204 400-(554)-870 scalariform Diervilla lonicera +/− diffuse solitary + scalariform scalariform/ bordered - - 20-(27)-35 136-(163)-184 310-(497)-640 opposite Diervilla sessilifolia - diffuse solitary + scalariform scalariform bordered - - 17-(26)-32 41-(57)-64 700-(900)-1,100 Dipelta floribunda +/− diffuse solitary +/− scalariform alternate bordered - - 15-(22)-35 144-(162)-176 500-(670)-850 Dipelta ventricosa + diffuse solitary - scalariform alternate bordered +/− - 20-(25)-33 100-(140)-168 470-(717)-900 Heptacodium miconioides + diffuse solitary - simple alternate bordered + + 25-(32)-50 160-(208)-236 410-(511)-620 Heptacodium miconioides +/− diffuse solitary - simple alternate bordered + + 140-(197)-240 400-(590)-700 Heptacodium miconioides + diffuse solitary +/− simple (1% scalariform) alternate bordered + + 20-(24)-35 160-(180)-200 300-(439)-600 Kolkwitzia amabilis +/− diffuse solitary - scalariform alternate bordered - +/− 18-(24)-30 53-(64)-70 420-(562)-700 Leycesteria crocothyrsos - diffuse solitary +/− scalariform alternate bordered + + 20-(27)-40 50-(62)-72 780-(986)-1,200 Leycesteria formosa ? diffuse solitary + scalariform alternate bordered + + 15-(22)-30 232-(249)-284 700-(1,002)-1,500 Lonicera involucrata + ring solitary + simple (scalariform 5%) alternate bordered + + 17-(26)-32 124-(177)-208 300-(458)-560 Lonicera syringantha +/− diffuse solitary +/− simple (scalariform 1%) alternate bordered +++ + 15-(19)-23 112-(161)-196 300-(456)-600 Lonicera thibetica + semi-ring solitary simple alternate bordered +++ + 10-(19)-25 140-(188)-236 330-(390)-500 Sambucus nigra +/− semi-ring solitary/radial +/− simple alternate simple +/− + 30-(45)-75 100-(134)-152 300-(474)-660 multiples (1.5) Symphoricarpos albus + ring solitary - simple (scalariform 5%) alternate bordered + + 20-(31)-60 100-(120)-140 270-(497)-800 Symphoricarpos occidentalis + ring solitary + simple (scalariform 5%) alternate bordered + + 15-(24)-35 84-(96)-110 400-(542)-720 Symphoricarpos occidentalis + ring solitary +/− simple (scalariform 10%) alternate bordered + + 20-(27)-37 144-(170)-198 220-(630)-1,150 Symphoricarpos orbiculatus + ring solitary simple alternate bordered - + 10-(24)-35 120-(139)-156 200-(326)-450 Viburnum opulus + diffuse solitary (80–90%)+ scalariform alternate bordered - - 22-(33)-42 120-(138)-158 470-(741)-900 Weigela coraeensis +/− diffuse solitary + scalariform alternate bordered - - 20-(30)-60 68-(98)-160 800-(972)-1,200 Weigela maximowiczii - diffuse solitary + scalariform opposite/ bordered - - 13-(22)-30 84-(131)-160 350-(595)-840 alternate Zabelia taihyoni + ring solitary - simple alternate bordered +++ + 18-(27)-50 88-(110)-172 350-(585)-960 Zabelia triflora + ring solitary - simple alternate bordered + + 11-(16)-24 76-(104)-120 360-(550)-700 Zabelia umbellata + ring solitary - simple alternate bordered + + 15-(29)-46 120-(125)-134 340-(463)-700 Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 01 JCB TA. ETCDU N AEI 241 JACOBSET AL.: HEPTACODIUM AND ZABELIA 2011]

Table 3. (Continued). Summary of wood characters of the sampled species.

Taxon/Character FP distinctly septate fiber thickness mean fiber length axial ray ray height sheath perforated rays/mm prismatic crystals bordered fibers parenchyma width cells ray cells Abelia chinensis + - (very) thick 900-(1,073)-1,300 diffuse 3–4 600-(1,362)-2,400 - scalariform 14-(18)-21 absent Abelia graebneriana + - thin-thick 800-(990)-1,250 sparsely diffuse 2–3 400-(844)-1,500 - scalariform 9-(11)-14 absent Abelia x grandiflora + - thin 900-(1,112)-1,280 diffuse 2–3 500-(1,138)-2,300 - - 11-(16)-18 absent Abelia spathulata + - thin-thick 450-(748)-900 diffuse 3–4 300-(1,125)-2,500 - scalariform 15-(16)-22 absent Diervilla lonicera + - thin 350-(657)-900 sparsely diffuse 2-(3) 150-(352)-650 - scalariform 20-(21)-24 absent Diervilla sessilifolia + - thin 1,000-(1,080)-1,220 sparsely diffuse 1-(2) 1,500 - scalariform 21-(23)-27 absent Dipelta floribunda + - thick 600-(871)-1,150 diffuse 2–3 400-(850)-1,200 - scalariform/ reticulate 10-(13)-16 absent Dipelta ventricosa + - thin-thick 760-(993)-1,220 diffuse 2–3 400-(850)-1,900 - scalariform 12-(14)-15 absent Heptacodium miconioides + - thin-thick 700-(841)-1,000 sparsely diffuse 3–4 400-(807)-1,900 - simple 14-(16)-18 in ray cells Heptacodium miconioides + - thin-thick 730-(840)-1,300 sparsely diffuse 3-(4) 250-(689)-1,150 - simple 16-(17)-21 in ray cells Heptacodium miconioides + - thick 700-(870)-980 sparsely diffuse 3 300-(840)-1,350 - - 16-(17)-18 in ray cells Kolkwitzia amabilis + - thick 650-(796)-1,000 diffuse 2–3 ? - scalariform 13-(16)-19 in ray cells Leycesteria crocothyrsos + - thin-thick 900-(1,120)-1,400 sparsely diffuse 3–4 > 2,400 - - 11-(12)-14 absent Leycesteria formosa + - thin-thick 700-(1,133)-2,100 absent 1-(3) > 2,000 - scalariform 14-(19)-22 druses in bark Lonicera involucrata + - thin-thick 550-(638)-750 absent 2–3 250-(790)-1,300 - simple 21-(23)-24 absent Lonicera syringantha + - thin-thick 700-(814)-1,100 absent 2 250-(546)-1,000 - simple 14-(17)-20 absent Lonicera thibetica + - thin-thick 500-(627)-700 sparsely diffuse 2 250-(622)-1,900 - - 18-(20)-23 druses in bark Sambucus nigra - - thin-thick 500-(700)-900 absent 2–3 200-(519)-1,000 - 4-(7)-9 in ray cells Symphoricarpos albus + - thin-thick 460-(765)-960 diffuse 3 800-(1,875)-2,400 - simple 13-(18)-21 absent Symphoricarpos occidentalis + - thin-thick 550-(753)-950 sparsely diffuse 2-(3) 300-(793)-1,600 - simple 13-(18)-21 absent Symphoricarpos occidentalis + - thin-thick 600-(811)-1,150 sparsely diffuse 2 300-(813)-1,300 - simple/ scalariform 13-(16)-18 absent Symphoricarpos orbiculatus + - thin-thick 400-(561)-700 diffuse 2–3-4 300-(660)-1,200 - simple 17-(19)-22 in ray cells Viburnum opulus + - thin-thick 700-(943)-1,100 absent 1 (2)250-(447)-750 - - 10-(11)-13 absent Weigela coraeensis + - thin 950-(1,227)-1,500 sparsely diffuse 1 > 2,000 - - 18-(19)-22 absent Weigela maximowiczii + - thin 700-(660)-950 sparsely diffuse 1–2 > 2,000 - - 18-(22)-31 absent Zabelia taihyonii + - thick 810-(1,013)-1,320 diffuse 3–4 500-(1,118)-2,200 - - 11-(14)-16 absent Zabelia triflora + - thick 650-(825)-1,000 diffuse 2–3 600-(900)-2,100 - simple 15-(17)-22 absent Zabelia umbellata + - thick 670-(833)-1,000 diffuse 3–4-5 400-(784)-1,300 - simple 11-(12)-13 absent Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 Lonicera ray heightislessthan550µmin higher thanonemm. Axially fusedraysare common.Mean cells, insomecasesexceeding10rows. Raysare generally lular withseveralrows ofuprightand/orsquare marginal Weigela coraeensis 2-seriate in albus up tofourorfivecellswide( rays are mostlyonetothree cellswide,sometimes theyare ally composedofuptoeightcellsperstrand.Eventhough S.orbiculatus Abelia parenchyma (sometimesdiffuse-in-aggregates) ispresent in observed intheraycellsof sel elementsof similar totheperforationplatetypesencountered intheves- counted more than20rays permm.Perforatedraycellsare µm), betica nigra observed more than12rays permm,exceptfor and and observed in Symphoricarposorbiculatus pericarp withitssclerifiedendocarp ( Fig. 4C , E ). The firstmechanicallayerprotecting theseedisthin, dry mature seed.Severalfree supernumerarybractsare present. opment. Remainsofthemesotestalcrystalsare present inthe mesotesta (andendotesta)are compressed duringseeddevel- the outlineofcompressed seedcoatcellsisvisible.The smooth seedsurface( Fig. 4F ). A subtlesculpture causedby of compressed parenchymatous cells( Fig. 4D ) resulting ina ( Figs. 4D-F ). Themature seedcoatiscomposed ofathinlayer date aminuteembryoincopious,thin-walledendosperm seeds are long,slender, andspindle-shaped,accommo- ized bythered topurple,persistent,accrescent calyx.Mature 4E ). At maturity, thelong,cylindrical achenesare character- carp, andasingle-layered, thin-walled,fibrous endocarp( Fig. fruit iscomposedofathinexocarp,dry, compressed meso- mesocarp, andvascularbundles.Thepericarpofthemature results inacompression ofthesterilecarpels,most an earlystageofseeddevelopment( Fig. 2C-D ). Seedgrowth oxalate crystal.Theendotestaappearstobecompressed at raphal bundle.Manymesotestalcellscontainalarge calcium layered mesotesta( Fig. 4C ) comprisingalarge, amphicribral oping seedcoatconsistsofaunilayered exotestaandamulti- and anunsclerified,unilayered endocarp( Fig. 4C ). Thedevel- mesocarp withuptotenwell-developedvascularbundles, carp iscomposedofaunilayered exocarp,amultilayered sterile ovules( Fig. 4B ). Duringfruit development,theperi- a singleovule( Fig. 4A ) andtwosterilecarpelswithseveral The tricarpellateovaryiscomposedofonefertilecarpelwith [Volume 36 SYSTEMATIC BOTANY formosa 242 mesocarp, and aunilayered endocarp.The mature endocarp pericarp iscomposed ofaunilayered exocarp, amultilayered elliptic, fertileloculewithone fertileovule( Figs. 4G-H ). The sterile carpelswithseveral ovulesandoneabaxial, Ry Cells— Ray Fruit andSeedMorphologyAnatomy— Z Parenchyma— ABELIA Sambucus nigra Viburnum opulus , and and , , Lonicera syringantha , , Symphoricarpos occidentalis , , Dipelta , , Lonicera Symphoricarpos — Viburnum opulus S. orbiculatus The ovaryiscomposedoftwocircular adaxial Abelia graebneriana ), and Diervilla sessilifolia (Fg 3 , 3M), (Fig. Abelia Aggregate raysare absent.Raysare 1-or (present in , and Axial parenchyma isabsentin Zabelia . Sparselydiffuse axialparenchyma was , , (447µm).Sheathcellsare absent.We Diervilla ). W. maximowiczii , and (546µm), . Axial parenchyma strandsare usu- (Fg 3O). (Fig. K. amabilis . In K. amabilis Abelia spathulata Zabelia L. thibetica , , Diervilla Rehder, (Fg 3 , 3L), (Fig. Dipelta , and Diervilla lonicera Sambucus nigra . Prismaticcrystalswere , , , , Symphoricarpos and , , Weigela Kolkwitzia Diervilla . Raysare heterocel- Sambucus nigra ) ), Viburnum opulus Viburnum opulus Weigela , , Symphoricarpos H . Diffuse axial , , EPTACODIUM , , Lonicera thi- Leycesteria Mill.(352 (519µm), Leycesteria , weeven Sambucus (notin , and , — , , , Patrinia pels are sterileintheValeriana clade,inseveralgenera(e.g. one seed,ovarymorphologyisunknown. Although twocar- ing oneovule. Although Tetradoxa omeiensis (no carpelabortion),eachcontainingoneovule,whereas Sambucus late ovarieswithonefertilecarpelcontainingovule. in pels varies,however(onesterilecarpelin both havetetracarpellateovaries.Thenumberofsterilecar- eral ovulesineachcarpel. Lonicera late ovarywitheachcarpelholdingmanyovules,whereas as regards ovarymorphology. taining numerous ovules. TheLoniceracladeisfairlydiverse is knownforitsbicarpellatecapsuleswitheachcarpelcon- one fertileovule,andsterilecarpel.TheDiervillaclade clade havebicarpellateovarieswithonefertilecarpelthathas and twofertilecarpels( Fig. 4M ). MembersoftheDipsacus Dipelta and onefertilecarpelwithovule( Figs. 4J-L , R). 4G , J-L,R)withtwosterilecarpels,containingseveralovules, clade ischaracterizedbyinferior, tricarpellateovaries( Figs. are phylogeneticallyimportant inDipsacales.TheLinnina characterized byfleshyfruits (except only cladewithfleshyfruits. Adoxaceae, however, are also the Diervillaclade.InCaprifoliaceae,Loniceracladeis of theLinninaclade(e.g. Fig. 4T ). Fruits are onlydehiscentin two-seeded achenes. respectively. Triosteum Lonicera shaped seeds,whichare mostlywingedin of theDiervillacladeholdmany, small elliptictokidney- encountered in sal bywindandwater. pels are persistentinthemature achenesandpromote disper- miconioides (cf. seed. Fruits are somewhatcompressed andslightlycurved indehiscent acheneholdingoneslender, spindle-shaped pressed duringseeddevelopment.Themature fruit isadry, development. Theparenchymatous seedcoatbecomescom- tile ovules,theadjacentendocarpdoessclerifyduringfruit opment. Eventhoughthesterileloculesdonotcontainfer- is builtupoffibersandstartssclerifyingduringfruit devel- ries in in (1) capsulesintheDiervillaclade;(2)drupes in Adoxaceae (not Table 4 seed morphologyandanatomyofDipsacales,werefer to . InDipsacales,fourmainfruit typesare encountered: fruit wallwithitssclerifiedendocarp. copious. As in to mesocarpcompression. Theendospermisthin-walledand chymatous and moderatelydevelopedin S.corydalifolia in allother Adoxaceae rangesfrom fourto five(apartfrom other Caprifoliaceaehavesingle-seed fruits. Seednumber Oay Morphology— Ovary Se Number— Seed Fut Morphology— Fruit Se Ca Anatomy— Coat Seed Sinadoxa corydalifolia Symphoricarpos Heptacodium miconioides Leycesteria , however, hasatetracarpellateovarywithtwosterile , , hastwoorthree carpels(sometimesfive)withsev- holdbetweensixand20seeds,while thedrupes of Fedia and and (Fg 4 , 4A), (Fig. Dipelta withitssingle-seededfruits). Gaertn.,and Symphoricarpos Heptacodium Diervilla Adoxa moschatellina and ). In Adoxaceae, Fruits withnumerous seeds(>20)are only hasatetracarpellateovary, eachcontain- isunique intheLinninacladewithits Sinadoxa corydalifolia Lonicera Sinadoxa corydalifolia ) ), , , Triosteum Weigela For adetailedaccountoffruit and In Adoxaceae, theseedcoat isparen- Sinadoxa corydalifolia Carpel numberandcarpelabortion , Fig. 4E ) Thepericarpisthindue , thefirstmechanicallayeris Valerianella containthree andtwopyrenes, ; and(4)achenesin Triosteum Leycesteria , and , and haveuptofivecarpels Viburnum Leycesteria Sinadoxa corydalifolia Symphoricarpos Mill.)thesterilecar- and Adoxa hasapentacarpel- , , , andallmembers Triosteum Weigela Viburnum hasfruits with hastricarpel- Symphoricarpos and . Thecapsules Heptacodium . Berriesof Sambucus andtwo , andall ; (3)ber- ). ,

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 abaxial fertilelocule.H.Fertilelocule withdevelopingseedandsterileloculetwo,small,ovules. I-K. endosperm. F. Mature seedwithsmoothsurface.G-H. mature seedwiththin,amorphous seedcoat.E.Transverse sectionofmature fruit withslender, spindle-shapedseedwithminute matous mesocarp,sclerifiedendocarp,unilayered exotesta,multilayered mesotesta,compressed endotesta,andadjacentendosperm on developingseedinfertilecarpel.B.Immature fruit withviewonsterileovulesincarpels.C.Cross sectionofimm 243 JACOBSET AL.: HEPTACODIUM AND ZABELIA 2011] endosperm. P. fertile locule.K.Cylindricalovarywith fertileloculeexposed.Fertilecontainsoneanatropous ovule.L. with bilobedcalyxandfree supernumerarybractssubtending theflower. J.Cross sectionofimmature fruit withtwoadaxialste pair ofsupernumerary bracts,covered with glandulartrichomes,enclosingthe fruit. sterile loculescontaining severalsterileovules.S. pratensis sclerified exotesta.O. of ovarywithtwo,lateral,sterilelocules andtwofertile(adaxialabaxial)locules.N. fruit withoneseedinfertilelocule,twopersistent, sterilelocules,andtwoinflated,sclerifiedsupernumerarybracts.M. Fig. 4 4. Fruit andseedmorphologyanatomy(A-B,F, G-N,Q-T, SEM;C-E,O-P, LM). A-F. . Cross section ofmature diaspore. R. Knautia macedonica Diervilla sessilifolia . Transverse sectionofmature fruit withepicalyx,thin,parenchymatous seedcoat,andadjacentendosperm.Q. . Cross sectionofseedwithsclerifiedexotesta, compressed mesotesta, weaklysclerifiedendotesta,andadjacent Patrinia scabiosifolia Triosteum perfoliatum Zabelia triflora . Cross sectionofmature fruit withoneabaxial fertileloculeandtwoadaxial,persistent . Cross sectionofpyrene. T. . G. Cross sectionofimmature fruit withtwoadaxialsterileloculesand one Lonicera dioica Heptacodium miconioides Linnaea borealis . Cross sectionofmature seed withwelldeveloped, Kolkwitzia amabilis . Lateralview of mature fruit withone Abelia parvifolia Dipelta yunnanensis . A. Immature fruit withview rile loculesandoneabaxial ature fruit withparenchy- . Cross sectionofmature . I.Developing flower embryoandcopious . D.Cross sectionof . Cross section Succisa Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 Fedia Dipsacus clade,whereas thecore valerians( with copiousendospermand embryosreminiscent of the clade, thebasallineages(e.g. cross section(lessdorsoventrallyflattened).IntheValeriana ( Fig. 4Q ), althoughtheembryosare elliptictocircular in oventrally flattened.Thesameistrue fortheDipsacusclade as theseed.InMorinaclade,embryosare longanddors- embryo islongandcylindrical,aboutthesamelength Diervilla, Lonicera,andLinnaeas.l.clades.In omy are closely associatedwithseedprotection andthefirst toward seedswithlarge embryosandlackingendosperm. ter ofseedlength)in and Caprifoliaceae.Embryosare small(lessthanonequar- endosperm isphylogeneticallyinformativeinboth Adoxaceae usually compressed duringseeddevelopment( Fig. 4D , P). all otherCaprifoliaceae,theseedcoatisparenchymatous and The mesotestaiscompressed duringseeddevelopment.In exotesta andathin,weaklysclerifiedendotesta( Figs. 4N-O ). clades, theseedcoatiscomposedofasclerified,unilayered or welldevelopedin Zabeliatriflora Weigela subsessilis Weigela florida Viburnum opulus Valerianella rimosa Valeriana sambucifolia Valeriana officinalis Triosteum perfoliatum Triosteum hirsutum [Volume 36 Symphoricarposoccidentalis Symphoricarposalbus Succisapratensis Sambucusebulus Plectritisanomala Patriniascabiosifoli Patriniagibbosa Morinalongifolia Loniceramaximowiczii Loniceramaackii Loniceraalpigena Linnaeaborealis Leycesteriaformosa Leycesteriacrocothyrsos Kolkwitziaamabilis Knautiamacedonica SYSTEMATIC BOTANY Knautiaarvensis Heptacodiummiconioides Fediacornucopiae Dipsacussylvestris Dipsacusfullonum Dipeltayunnanensis Dipeltafloribunda Diervillasessilifolia Diervillarivularis Cephalariaalpina Centranthusruber Adoxamoschatellina Abeliaparvifolia Abeliachinensis 244 Frt ehncl Layer— Mechanical First Endosperm— and Embryo Table , , Plectritis . 4. Summary offruit andseedcharactersofthesampledspecies. Taxon

cpue202+--se ot-+sal+ small + seed - - coat - 20+ 0 2 capsule ahn rcs--sal- small - - bracts - + 1 2 3 achene ahn eiap--lre- large - - pericarp - + 1 2 3 achene

ahn eiap--sal- small - - pericarp - + 1 2 3 achene bry23()062 edca ml - small + seed - + coat - 6–20 0 2–3-(5) berry ahn pclx+-lre- large - + epicalyx - + 1 1 2 achene ahn pclx+-lre- large - + epicalyx - + 1 1 2 achene due45045++edcr ag - large - - endocarp + + 4–5 0 4–5 drupe ahn eiap+-lre- large - + pericarp - - + large 1 - 2 + epicalyx 3 achene - + 1 1 2 achene bry23()062 edca ml - small + seed - + coat - 6–20 0 2–3-(5) berry due321++edcr ml - small - - endocarp + + 1 2 3 drupe ahn eiap--entire - - seed - pericarp - + 1 2 3 achene ahn eiap--entire - - seed - pericarp - + 1 2 3 achene cpue202+--se ot-+sal- small + seed - - coat - 20+ 0 2 capsule ahn rcs--sal- small - - bracts - + 2 2 4 achene ahn eiap--entire - - seed - pericarp - + 1 2 3 achene cpue202+--se ot-+sal+ small + seed - - coat - 20+ 0 2 capsule Suksd., ahn pclx+-lre- large - + epicalyx - + 1 1 2 achene ahn eiap--entire - - seed - pericarp - + 1 2 3 achene ahn pclx+-lre- large - + epicalyx - + 1 1 2 achene bry502+-+se ot-+sal- small + seed - + coat - 20+ 0 5 berry ahn rcs--sal- small - - bracts - + 1 2 3 achene ahn eiap--lre- large - - pericarp - + 1 2 3 achene cpue202+--se ot-+sal- small + seed - - coat - 20+ 0 2 capsule ahn eiap--entire - - seed - pericarp - + 1 2 3 achene due413++edcr ml - small + - endocarp + + 3 1 4 drupe ahn pclx+-lre- large - + epicalyx - + 1 1 2 achene due45045++edcr ml - small - - endocarp + + 4–5 0 4–5 drupe ahn rcs--sal- small - - bracts - + 2 2 4 achene due422++edcr ml - small + - endocarp + + 2 2 4 drupe due413++edcr ml - small + - endocarp + + 3 1 4 drupe ahn eiap--entire - - seed - pericarp - + 1 2 3 achene bry23()062 edca ml - small + - seed + coat - 6–20 0 2–3-(5) berry bry502+-+se ot-+sal- small + - seed coat + - 20+ 0 5 berry ahn eiap--sal- small - - pericarp - + 1 2 3 achene Viburnum due422++edcr ml - small + - endocarp + + 2 2 4 drupe Viburnum Valeriana achene achene type fruit Embryo sizeandtheamountof Fruit typeandseedcoatanat- . IntheDiervillaandLonicera Patrinia , , L., number of Adoxa moschatellina 3 3 carpels Valerianella , Fig. 4R ) haveseeds sterile carpels number of Centranthus 2 2 ) haveevolved Sambucus , andthe number of seeds 1 1 DC., , the , endocarp part of diaspore function compared totheepicalyx( Fig. 4Q ). wall isreduced toathinpaperylayer with littleprotective specialized andmorphologically diverse. As aresult, thefruit of thediaspore. TheepicalyxoftheDipsacuscladeis more fused supernumerarybracts, is welldeveloped,butnotpart ily sclerifiedintheMorinaclade. Theepicalyx,theresult of bracts isinflatedandweaklysclerified. Theendocarpisheav- bracts partially( Linnaeaborealis by wind,whereas, in develops intolarge, paperystructures promoting dispersal for theembryo.In pletely ( clade ( Fig. 4Q ). In Valeriana ( Fig. 4R ) clades,andthe epicalyxintheDipsacus is thepericarpinLinnaea( Figs. 4G , J,K),Morina,and corydalifolia 4S), ( Fig. function isfulfilledbythesclerifiedendocarpin cal layeristhesclerifiedseedcoat( Figs. 4N-O ), whereas this Diervilla outermost layerofthediaspore andprotects theembryo.In mechanical layer. Thefirstmechanicallayerisusuallythe + + mesocarp Dipelta , , fleshy Symphoricarpos Leycesteria - - ). IntheLinninaclade,firstmechanical layer ) enclosetheseedandprovide someprotection ( Fig. 4T ), however, thefree supernumerary L. borealis first mechanical pericarp pericarp Dipelta , , Dipelta layer Lonicera K. amabilis , and Adoxaceae (nodatafor , Fg 4 4T ; Fig. , , onepairofsupernumerarybracts , , Kolkwitzia amabilis , and epicalyx present - - , onepairofsupernumerary K. amabilis Weigela seed coat sclerified - - thefirstmechani- , Fig. 4L ) orcom- small small (Fg 4 , and 4L), (Fig. embryo size Triosteum Sinadoxa winged seeds - -

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 01 JCB TA. ETCDU N AEI 245 and Even thoughourstatisticaltestsreject thehypothesisof 1A ). TheLinnaeaclades.ismonophyletic,though( Fig. 1A ). Zabelia et al.2003 ; Winkworth etal.2008 ). Theunexpectedpositionof JACOBSET AL.: HEPTACODIUMin which AND ZABELIA in topologiesidenticaltopreviously publishedphylogenies our sampling.Excluding tigations, whichisattributabletotheinclusionof clade, however, differs significantlyfrom previous inves- relative oftheLoniceraclade.ThephylogenyLinnina Winkworth etal.2008 ) byresolving Donoghue etal.2001 , 2003 ; Moore andDonoghue2007 ; 2011] and et al.(2008) byconfirmingthesisterrelationship of Lonicera clade,ourresults agree withthefindingsof Theis to shedmore lightonthe intergeneric relationships ofthe al. 2008 ; Winkworth etal.2008 ). Eventhoughweare unable clade (e.g. Backlund 1996 ; Donoghue etal.2001 , 2003 fidently assesstheintergeneric relationships ofthe Lonicera ; Theis et trating oncemore clade s.l.beingmonophyleticisnotrejected ( Table 1 ), illus- Lonicera supported sisterrelationship between followed by ous with clade. ThesituationintheLinnaeaclades.islessambigu- (> 12permm)of heterocellular rays. Although weoccasionally bordered pits,diffuse axialparenchyma, andahighdensity distinct toslightlyreduced pitborders, ðbres with distinctly with vessels,tracheids,and fibers, vessel-raypittingwith tary, relatively narrow vessels,helicalthickeningsassociated shares severalfeatures withotherDipsacales,suchassoli- signal ofourwoodanatomical data( Fig. 6 ). nificant degree ofhomoplasyobscuringthephylogenetic tematic positionofthegenus.Thisislargely duetothesig- 1952 ), itdoesnotsignificantly contributetoassessingthesys- omy supports shape ofthepollengrains( Fig. 5 ). Eventhoughwoodanat- vestibulum underneaththeectoaperture determinesthe whereas intheDiervillaandLoniceracladesafastigiumor pollen’s distinctshapeistheresult ofprotruding apertures, Diervilla andLoniceraclades( Figs. 1 , 5 ciated withthesimilarlyshapedpollengrainsfoundin ), in subtriangular pollengrainsof and Linnaeas.clades( Backlund 1996 ). Eventhoughthe forations are foundin 1981 ; Chia-Chi andChao-Xing1988 ), pollenwithmicroper- in Caprifoliaceae( Fig. 5 ; e.g. Böhnke-Gütlein andWeberling with ectopori,suggestingthatectoporievolved(atleast)twice eages. While tinct features, someofwhich are shared withothermajorlin- synapomorphies. Thepollenof caprifolioid taxa,eventhoughweare unabletodefineclear data suggestthat ical evidenceforthisrelationship isscarce. Ourpalynological ship between amabilis Our analysesagree withprevious investigations(e.g. So far, phylogeneticinvestigationshavebeenunabletocon- Heptacodium— Symphoricarpos Zabelia causestheLinnaeaclades.l.tobeparaphyletic( Fig. . or Abelia Zabelia Triosteum formingaclade,thepossibilityofLinnaea Linnaea borealis Heptacodium Heptacodium Heptacodium beingthefirstdiverging lineageoftheclade, isabsent(e.g. Zhang etal.2002 ; Donoghue Heptacodium Despite thewell-supportedsisterrelation- Zabelia . Ourresults furtherindicatethateither isthebasalmostlineageofLonicera Heptacodium Discussion ’s equivocalsystematicposition. andtheDiervillacladehavepollen andtheLoniceraclade,morpholog- ’s inclusioninDipsacales( Metcalfe Zabelia . Ourresults alsorecover awell- fallswithinthelowergradeof Heptacodium Heptacodium from ouranalysesresults aswellintheLonicera Heptacodium Dipelta seemtobeasso- haveseveraldis- Heptacodium and astheclosest Heptacodium Leycesteria Kolkwitzia Zabelia Abelia , the , in

wood anatomy doesnothelpusclarify thesystematic and (2)within (1) alongthelineleadingto two shiftsfrom echinatetopsilatepollenhavetaken place: species (e.g. ( Fig. 5 ), theoccurrence ofpsilate polleninseveral acter stateforCaprifoliaceae ( Fig. 5 ). Basedonouranalyses ( Verlaque 1983 ). Echinatepollenistheplesiomorphicchar- an endocingulumisrestricted to found inaselectnumberof len withanendocingulum. Although psilatepollen isalso this relationship. Bothlineageshavepsilate,tricolporatepol- tion, palynologicaldataprovide morphologicalsupportfor Morina cladeiswell-supportedinmostanalyses.Inaddi- also foundintheseeminglyunrelated genus and seedmorphology, ing asmallembryoandcopiousendosperm.Intermsoffruit tent calyxandalong,slender, spindle-shapedseedcontain- have slender, cylindricalacheneswithanaccrescent, persis- and seedsof in anyway. Despitethestrikingsimilarities betweenfruits and statisticaltests( Table 1 ) donotsupportthisrelationship Makino raised ship between hypothesis canbediscarded. Eventhoughacloserelation- (with lowsupport)andweare therefore confidentthatthis tionship isonlyresolved by ouranalysisbasedonnucleardata (2) theMorinaclade;or(3)Valeriana clade.Thelatterrela- sider three hypothesesas regards the conclusionthat (1996) analyzedalarge morphologicaldatasetandcameto Kim etal.(1999) onthebasisof Linnaea clades.l.,anhypothesisthatwasalsoformulatedby Pyck (2001) . Thisrelationship impliestheparaphylyof Morina cladewasalready suggestedby Verlaque (1983) and 7). (Fig. clade leading to tricarpellate ovaryinCaprifoliaceae;(1)alongthebranch clade impliestheoccurrence oftwoindependentshiftstoa ble sisterrelationship between Heptacodium sules asfruit types,speakagainstacloserelationship between coats, multipleseedsperfruit, anddrupes, berries,andcap- Linnaea clades.l.Notabledifferences, suchassclerifiedseed endosperm. However, thesefeatures are alsopresent inthe and Loniceraclades,thatis,smallembryoscopious with thetwobasalmostcladesofCaprifoliaceae,Diervilla orbiculatus observed prismaticcrystalsintheraycellsof Abelia ( Fig. 7 ). Similaritiesinfruit andseedmorphologybetween seeds withathin,inconspicuous,parenchymatous seedcoat ile carpels,adry, single-seeded,indehiscentachene,and Both share atrilocularovarywithonefertileandtwoster- ( Hara 1983 ; Donoghue etal.2001 close relationship between , 2003 ; tomical data,fruit andseedmorphologystrongly supporta Zhang etal.2002 ). to theambiguoussignalofourpalynologicalandwoodana- crystals seemtobemostcommonin Finally, thecloserelationship between With the current andpastinvestigationsinmind,wecon- Zabelia— , , Heptacodium , , Heptacodium Kolkwitzia amabilis andtheLonicera(orDiervilla)clade.Thepossi- The sisterrelationship between L. thibetica Abelia Abelia Lonicera Zabelia , and and and Abelia togenericlevel,ouranalyses( Fig. 1 ) ( Fig. 5 ). As isthecasefor ) istheresult ofconvergent evolution; ; and(2)attheoriginofLinnina Zabelia Heptacodium Zabelia Zabelia and Heptacodium Lonicera , and Zabelia Zabelia hasbeensuggestedeversince matK Heptacodium , virtuallyidenticalfruits are are striking. All three genera Zabelia Zabelia Sambucus nigra sequencedata. Backlund species,thepresence of bares littleresemblace Heptacodium are sistergenera. andtheMorinaclade; andtheLinninaclade and theMorinaclade : sisterto(1) andtheLonicera Zabelia Heptacodium Zabelia Symphoricarpos Heptacodium . Incontrast , prismatic , andthe and the Lonicera Abelia . , ;

Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 The epicalyx is theresult ofthe fusionofoneormore several pairsof free supernumerarybracts (noepicalyx). thin, unilayered, sclerifiedendocarp,asmallembryo, and oped epicalyx,whereas fruits of a large, dorsoventrallyflattenedembryo,andawell-devel- Fruits oftheMorinacladehaveathick, sclerified endocarp, between morphology supportingthehypothesized sisterrelationship and Qian2003 ; Zhou etal.2004 ). Similaritiesinfruit andseed position of DELTRAN optimizationsoccur, the ACCTRAN hypothesisisingray. A dashedlineconnectstherespective ACCTRAN andDELTRAN hypot the character, whereas thebottomnumberrefers tothecharacterstate.Nocharactesare homoplasious.When conflicts betweent [Volume 36 SYSTEMATIC BOTANY 246 have considered ( Fig. 6 ). Based onfruit andseedmorphology, severalauthors has evolvedindependentlyin ity inDipsacales,ourresults indicatethatring-porous wood Fig. 5 5. Parsimony optimizationanalysisoffourpollencharacters(bottomleft)basedonourcombinedML phylogeny. Theupper Zabelia Zabelia andtheMorinacladeare seeminglylacking. Zabelia . Despitetherare occurrence ofring-poros- congenericwith Symphoricarpos Zabelia (n (and Abelia Abelia and (e.g. Zhou ) ae a have ) Zabelia dorsoventrally compressed embryosof theMorinaclade took place( Fig. 7 ), althoughitmustbenotedthatthe merary bracts.InCaprifoliaceae, twoshiftstolarge embryos as theresult ofthefusionanadditionalpairsupernu- oped asecondepicalyx(subtending the8-veinedepicalyx) phylogeny, whichdidnotinclude (2003 , Fig. 4A this hypothesishasalsobeen suggestedby ), theybasedtheirhypothesison adifferent Donoghue etal. fusion oftwopairsbracts(8-veinedepicalyx). Although once attheoriginofDipsacusclades.l.asresult ofthe fusion ofthree pairsofbracts(12-veinedepicalyx);and(2) (1) onceattheoriginofMorinacladeasresult ofthe an epicalyxoccurred (atleast)twiceinCaprifoliaceae( Fig. 7 ): supernumerary bracts.Ourresults indicatethattheshiftto Zabelia . . Triplostegia he ACCTRAN and he ACCTRAN number indicates heses. devel- Copyright (c) American Society for Plant Taxonomists. All rights reserved. Delivered by Ingenta to IP: 192.168.39.151 on: Wed, 29 Sep 2021 22:51:14 2002 ). s. l.clades,can beexplainedbyhybridization ( Zhang et al. ber between 2002 ). Differences inkaryotypeandbasechromosome num- clade an