Evolution,55(10),2001,pp.2011–2027 THE EVOLUTION OF AGRICULTURE IN BEETLES (CURCULIONIDAE: SCOLYTINAE AND PLATYPODINAE) BRIAN D. FARRELL,1,2 ANDREA S. SEQUEIRA,1 BRIAN C. O’MEARA,1 BENJAMIN B. NORMARK,1,3 JEFFREY H. CHUNG,1,4 AND BJARTE H. JORDAL1,5 1Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts 02138 2E-mail: [email protected] 3Department of Entomology, Fernald Hall, University of Massachusetts, Amherst, Massachuseets 01003 4Harvard Medical School, 25 Shattuck Street, Boston, Massachusetts 02115 5Department of Zoology, University of Bergen, Allegaten 41, N-5007 Bergen, Norway Abstract. Beetles in the weevil subfamilies Scolytinae and Platypodinae are unusual in that they burrow as adults inside trees for feeding and oviposition. Some of these beetles are known as ambrosia beetles for their obligate mutualisms with asexual fungi—known as ambrosia fungi—that are derived from plant pathogens in theascomycete group known as the ophiostomatoid fungi. Other beetles in these subfamilies are known as bark beetles and are associatedwithfree-living,pathogenicophiostomatoidfungithatfacilitatebeetleattackofphloemoftreeswithresin defenses. Using DNA sequences from six genes, including both copies of the nuclear gene encoding enolase, we performed a molecular phylogenetic study of bark and ambrosia beetles across thesetwosubfamiliestoestablishthe rate and direction of changes in life histories and their consequences for diversification. The ambrosia beetle habits have evolved repeatedly and are unreversed. The subfamily Platypodinae is derived from withintheScolytinae,near the tribe Scolytini. Comparison of the molecular branch lengths of ambrosia beetles and ambrosia fungi reveals a strong correlation, which a fungal molecular clock suggests spans 60 to 21 million years. Bark beetles have shifted from ancestral association with conifers to angiosperms and back again several times. Each shift to angiosperms is associated with elevated diversity, whereas the reverse shifts to conifers are associated with lowered diversity. The unusual habit of adult burrowing likely facilitated the diversification of these beetle-fungus associations, enabling them to use the biomass-rich resource that trees represent and set the stage for at least one origin of eusociality. Key words. Ambrosia, bark beetles, coevolution, haplodiploidy, insect-plant interactions, Platypodinae, Scolytinae. Received December 19, 2000. Accepted June 1, 2001. The interactions between insects and plants include the sive generations of ants (Chapela et al. 1994; Mueller et al. bulk of terrestrial multicellular species (Ehrlich and Raven 1998; Currie et al. 1999). 1964) and provide the resource basis for most of the rest. Whereas leaf-cutting attine ants strip trees of leaves, fun- The origins of herbivory and, in particular, feeding on an- gus-carryingbeetles(Fig.1)intheNeotropicsandelsewhere giosperms have resulted in enhanced rates of insect diver- bore into the center of tree trunks (Fig. 2) and branches or sification, producing nearly half of all insects (Mitter et al. carve intricate tunnel systems—often termed galleries—in 1988; Farrell 1998a), but insects often do not act alone in thephloem(Hubbard1897;Schedl1956;Browne1961;Stur- their depredations on plants. Many insect-plant interactions, geon and Mitton 1982; Wood 1986; Paine et al. 1997). The sometimes devastating attacks on forest trees by Scolytinae perhaps most, also involve microbial associates, including andPlatypodinae(approximately7500speciesaltogether)are plant pathogenic fungi (Paine et al.1997).Byexpandingthe usuallymediatedbyassociatedfungi,thearthropod-dispersed capacity of the insects to use plant resources, mutualistic ascomycetesclassifiedintheorderOphiostomatalesbutgen- interactions between herbivorous insects and fungi may erally referred to as the ‘‘ophiostomatoid fungi’’ (Fig. 3; cf. themselvespromotediversificationbyfungus-associatedlin- Wingfield et al. 1993). eages (Wilding et al. 1989; Holldobler and Wilson 1990; Ratherthanbuildfungalgardensundergroundliketheants, Wingfield et al. 1993; Chapela et al. 1994; Mueller et al. thesebeetlesbringthefungitotheirhosttrees,oftencarrying 1998). them via a striking array of glandular, invaginated cuticular In Neotropical forests, the most prodigious consumers of structures (mycangia) that serve to maintain fungal spores tree leaves are probably the attine ants and the symbiotic and mycelia in pure, often yeastlike,culturesforinoculation basidiomycete fungi they culture in colonies below ground. into the galleries ovipositing beetles dig into wood (Batra Because the ant fungi are unusually polyphagous—they are 1966; Francke-Grossman 1967; Beaver 1989; Malloch and able to utilize 50–77% of the species in a rainforest (Funk Blackwell1993).Onceintroduced,thesefungiformmycelia 1985;Cherrettetal.1989)—theseleaf-cuttingantsfeedtheir that help curtail tree defenses and/or serve as beetle food fungal gardens with leaves cut from a wide array of plant (Beaver1989).Theophiostomatoidfungicomprisethethree taxa,anapparentadvantageinhyperdiverseforests(Cherrett sexual, often pathogenic genera Ophiostoma (the largest ge- etal.1989;HolldoblerandWilson1990;Muelleretal.1998). nus, mostly associated with conifers), and Ceratocystis and Recent phylogenetic studies of the ants and their fungi has Ceratocystiopsis(bothmostlyassociatedwithvariousangio- revealed a single origin of gardening, approximately50mil- sperms), plus several dozen genera of asexual anamorphs. lion years ago, and repeated acquisitions of fungi that are Although most of these are known to be at least loosely apparently propagated as clones and passedbetweensucces- associatedwithbeetles,somehavebecome,liketheantfungi, 2011 (cid:113) 2001 The Societyforthe Study of Evolution.Allrightsreserved. 2012 BRIAN D. FARRELLETAL. FIG. 2. Gallery of Xyleborus celsus in hickory (from Hubbard FIG. 1. Dorsal view of adult female (left) and male (right) of 1897). Xyleboruscelsus(fromHubbard1897).Thisspeciesrangesinsize from 4.0 mm to 4.5 mm. (3–10 mm/day) and thus quick blocking of the resin canals (Paine et al. 1997). These associations with resin-bearing obligately associated, vertically transmitted, polyphagous conifersareancestralintheScolytinae(Sequeiraetal.2000) asexual domesticates and serve as the primary food of their and remain principal affiliations of temperate bark beetles. beetle farmers (Malloch and Blackwell 1993). These fungal Numerous authorshaveremarkedonasimilarassociationof cultivars are now classified in the genera Ambrosiella and tropicalbarkbeetleswithresin-orlatex-bearingangiosperms Raffaelea,butwerenamedambrosiafungifortheirenigmatic (e.g., in Mexico: Atkinson and Equihua 1986a,b; Noguera- identitytoearlynaturalistsperplexedoverthefoodofcertain MartinezandAtkinson1990;Africa:Schedl1956;Malaysia: tunnelingbeetles(ambrosiabeetles),thatseemednottocon- Browne1958,1961;alsoseeWoodandBright1992),whose sume wood (Hubbard 1897). sticky defenses they may likewise counter with fungal as- Ambrosia beetles total some 3400 described species in 10 sociates and tunneling strategies (Farrell et al. 1991; Dus- tribes. They lay their eggs in fungal gardens they culture in sourd and Denno 1994). Indeed, these beetles are some of galleries usually dug deep in the interior xylem. Extant am- themostubiquitousinsectscapturedinthefossilizedlegume brosia beetle genera in several tribes are represented in 30- and conifer resins that comprise the Dominican and Baltic million-year-oldamber,suggestingthatatleastsomeorigins ambers (Bright and Poinar 1994). of this association occurred even earlier in the Tertiary,per- hapsroughlycontemporaneouswiththeoriginofattineants. Whether the contemporary descendents of ambrosia beetles and fungi are themselves of comparable ages, however, has been an open question. Ant fungal gardeningaroseonceand is hypothesized to have arisen through intermediate stages of fungal consumption and then dispersal or the reverse (Mueller et al. 2001). Beetle fungal gardening differs both in numbers of origins and in the persistence of apparently earlier stages. In contrast to ants, there are several indepen- dent instances of ambrosia associations, in different beetle tribes, each of which also contains phloem-feeding beetles associated with Ophiostoma or other sexual free-living rel- atives of the ambrosia fungi (Whitney 1982). The nonam- brosia-feedingScolytinaetotalapproximately4100described speciesandarelargelyhost-specialisedphloem-feedersprop- erly known as bark beetles, plus approximately 400 species of typically more polyphagous beetles that feed on the pith oftwigsorinseeds(Schedl1956;Browne1961;Wood1982; Kirkendall 1993). Bark beetles that attack conifers are especially well stud- ied, and their associated Ophiostoma fungi havebeenshown to circumvent the conifers’ resinous defensesbyfastgrowth FIG. 3. Ambrosiafungus of Xyleboruscelsus(from Hubbard 1897). EVOLUTIONOF BEETLEAGRICULTURE 2013 Unlikethetypicallyspecialistbarkbeetles,individualam- andhostplants,weconductedamolecularphylogeneticstudy brosiabeetlespeciesmostoftenuseawidearrayofhosttaxa of 86 species of Scolytinae and Platypodinae, representing (Schedl 1956; Browne 1958; Beaver 1979, 1989; Noguera- the major tribes and ecological variation worldwide, plus Martinez and Atkinson 1990), made possible by the beetles’ outgroups.Weusedsixgenes,includingtwodivergentcopies direct feeding on fungi with broad host tolerances(Francke- of enolase that have not previously been used for phyloge- Grosmann 1967; Beaver 1989). Thus, domesticated fungi netic studies in insects. seem to enable both leaf-cutting ants and ambrosia beetles to adopt a generalist strategy apparently well suited to hy- MATERIALS AND METHODS perdiversetropicalforests(Beaver1979;Cherrettetal.1989; Beetleswerecollectedfromcolonizedhosts(seeAppendix Holldobler and Wilson 1990; Mueller et al. 1998). Whereas for collecting localities) and include genera from 20 of the both adults and larvae of phloem-feeders carve galleries as 25 tribes proposed by Wood and Bright (1992) for the Scol- they feed in isolation, ambrosia beetle larvae do little tun- ytinae and from two of the five tribes of Platypodinae (both neling; instead they feed gregariously in chambers on asex- classified as families in Wood and Bright 1992). Previous ually produced conidiospores induced in cultures kept pure comparative morphological studies of adults (Kuschel1966; by the parents (Hubbard 1897; Francke-Grosseman 1967; Thompson 1992) and cladistic analyses of larvae (Marvaldi Beaver 1989; Kirkendall et al. 1997), much as in attine ant 1997)haveproposedacloserelationshipbetweentheweevil gardens (Mueller et al. 1998; Currie et al. 1999). The only subfamily Cossoninae and the Scolytinae. We were able to other major agricultural insect group, the eusocialmacroter- obtain two genera from the Cossoninae belonging to two mitine termites, are largely detritivores, feeding their basid- differenttribes(CossoniniandAraucarini)toincludeasout- iomycetefungi(genusTermitomyces)withleaf-litterandoth- groups. er plant remains found near the surface of the soil in their AfricanandIndomalayanhabitats(WoodandThomas1989). Recent progress in molecular systematics of the ambrosia Amplification and Sequencing fungi Ambrosiella and Raffaelea and other ophiostomatoid Polymerase chain reaction (PCR) and cycle sequencing fungi (Cassar and Blackwell 1996; Jones and Blackwell were used to obtain partial sequences of six genes: EF-1(cid:97); 1998) provide opportunities for comparisons with bark and COI; two copies of enolase, here termed enolase 1ni (no ambrosia beetle phylogenesis. On the fungal side, the phy- intron) and enolase 2I (intron-containing; Fig. 4); 18S; and logenyofferssupportfortheacquisition-clonalityhypothesis 28S (D2 and D3 expansion segments). DNA was extracted developedforants(Chapelaetal.1994),inthattheseasexual from individual beetles preserved in ethanol following Sun- ambrosia fungal genera are each polyphyletic and multiply nucksandHales(1996)withthemodificationsintroducedby derived from the sexual genera Ceratocystis, Ceratocystiop- Normark (1999). sis, and Ophiostoma (Cassar and Blackwell 1996; Jones and PCR reactions(50(cid:109)l)typicallycontained0.2mMofeach Blackwell1998).Therehavenotbeencomparablemolecular primer, 0.8 mM dNTPs, Qiagen (Valencia, CA) PCR buffer surveys across the many groups of ambrosia beetles from withadditionalMgCl toafinalconcentrationof2mM(18S 2 whichthesefungihavebeenstudied.Wehavethereforesam- and 28S) or 2.5 mM (EF-1(cid:97), COI, and enolase), and 1.25 pledallofthemajorlineagesofScolytinaeandPlatypodinae units Qiagen Taq DNA polymerase. For 18S and COI, the to identify the number, placement, and apparent ages of or- temperature profile was 40 cycles of 95(cid:56)C for 30 sec, 47(cid:56)C igins of ambrosia beetles and to evaluate whether this habit for 60 sec, and 72(cid:56)C for 60 sec. The cycling profile for 28S is ever reversed. Identification of ambrosia beetle sister wasthesameexceptthattheannealingtemperaturewas50(cid:56)C. groups will also make it possible to test whether polyphagy For EF-1(cid:97), a touchdown profile of 42 cycles was used, with tendstopromotediversification,bypreventingextinctionand annealing temperature decreasing from 58(cid:56)C to 42(cid:56)C by 2(cid:56)C increasing geographic range, or depress diversification, by every third cycle and the final 18 cycles at 42(cid:56)C. After am- increasing gene flow among distant populations (Kelley and plification, double-stranded PCR products were purified us- Farrell 1998; Dobler and Farrell 1999; Kelley et al. 1999). ing the Qiagen PCR purification kit to remove primers and Thestriking variation in resourceuseinScolytinaeispar- unincorporateddNTPspriortosequencing.Cyclesequencing alleled by variationinreproductivestrategies,andnumerous reactionswereperformedwiththeABIprismDyeTerminator hypotheses have been developed on the interplay between Cycle Sequencing Kit (Perkin-Elmer, Norwalk, CT) on an thesetwodimensionsoftheirlifehistories(Kirkendall1993; ABI 370 automated sequencer. Primer sequences are given Kirkendall et al. 1997). For example, most bark beetles are for enolase in Table 1, for EF-1(cid:97) and COI in Normark etal. monogynous,presumablybecausethetunnelsfacilitatemate- (1999); and for 18S and 28S in Sequeira et al. (2000). guarding; and gregarious feeding in ambrosia chambers or inpithorseedshasbeenthoughttoleadtothemanyinstances Enolase ofinbreeding(Kirkendall1993).Moreover,ambrosiabeetles showparentalcare,atleastoneoriginofhaplodiploidy(Nor- Enolase, also known as 2-phospho-D-glyceratehydrolase, mark et al. 1999) and eusociality (Kent and Simpson 1992). is a glycolytic enzyme responsible for converting two PG Thus, it seems that all three agricultural insect groups (am- molecules into PEP through a dehydration reaction during brosiabeetles,attineants,andmacrotermitinetermites)show glycolysis. The enolase gene is generally a single-copy (in some degree of sociality. invertebrates)nucleargeneofapproximately1302bplength. To provide a phylogenetic framework for tests of hypoth- Although cited as a promising candidate for molecular sys- eses on the evolution of interactions between beetles, fungi, tematics by Friedlander et al. (1992), it had never been de- 2014 BRIAN D. FARRELLETAL. FIG. 4. Diagram of the intron/exon structure of enolase 2I in members of four different tribesincluded in this study.Numbersreferto the position of the intron with respect to the Drosophila melanogaster sequence (Genbank accession no. X17034). veloped for insect systematics. Initial primersforthesingle- deletions. Ribosomal sequences did have insertion-deletion copy nuclear protein coding gene were developed using differences, and thus were aligned using Clustal X (Aladdin GenBank sequences for Drosophila, the decapod crustacean Systems Inc., Heidelberg, Germany) with the default gap Penaeus monodon (Boonchuoy et al. 1999), and severalver- opening:gapextensioncosts(15:6)andthensubjectedtoeye tebrates. A series of PCR, sequencing, and primer modifi- inspection, where the minimum regions of ambiguous align- cationwasperformeduntilsequencescouldberetrievedfrom ment were selected forexclusion,includingfewgap-bearing a representative set of taxa. For PCR either a touchdown regions in the analysis. profile as above or a standard 40-cycle program with 47(cid:56)C to 42(cid:56)C annealing temperature were used (depending on the Gene Divergences taxon). Therelativesequencedivergencesbygenewerecompared by plotting the uncorrected pairwise distance between two Sequence Alignment taxa for a gene of interest against the HKY (cid:49) (cid:71) corrected AllsequenceswerecompiledusingSequencher3.0(Gene- pairwise distance for COI for that pair. The likelihood cor- codes Corp., Ann Arbor, MI). There were no insertions or rection was used to moreclosely approximatetimeonthex- deletionsinCOI.IntronswereremovedfromEF-1(cid:97)andfrom axis, although the overall result is similar with uncorrected theintron-bearingcopyofenolasepriortophylogeneticanal- EF-1(cid:97) or COI divergence.Forribosomal genes,thedistance ysis,yieldingprotein-codingsequenceswithnoinsertionsor for all positions was used, but third codon positions were excluded for protein-coding genes because saturation ob- scures the more slowly evolving first and second positions TABLE1. Primersforenolase.Namesrefertothegene(en,enolase), (Fig. 5a). To visualize the relative divergences per codon the direction (s, sense; a, antisense), and the position of the 3(cid:57) end positiontheproportionsofoverallgenedivergencebycodon with respect to the coding portion of the Drosophila melanogaster sequence(Genbankaccessionno.X17034).Usereferstoamplification position was graphed and the ratio of total first and second (a) and sequencing (s). N, Y, R, S, and X are base ambiguities. codon position:third position changes calculated (Fig. 5b). Name Use Sequence Phylogenetic Analysis ens65c a,s GACTCCCGTGGNAACCCCACNGTGGAGGT ens65b a,s GACTCCCGTGGNAACCCCACNGTNGAGGT Phylogenetic analysis was performed by maximum-par- ens65a a,s GACTCCCGYGGNAAYCCCACNGTNGAGGT simony using PAUP (ver. 4.0b4a, Swofford 2000). Amino ens287 a,s GARATYGAYGARTTYATGATYAA acidsequencesofCOI,themostvariableproteincodinggene ens312a a GACGGCACCGAGAACAAGAGC (Fig. 5a), were combined in one matrix with nucleotide se- ens312b a TGGACGGCACCGAGAACAA ens509 a,s TGGCSATGCAGGARTTCATGAT quences for all other genes (EF-1(cid:97), 18S, 28S, enolase 1ni, ens552n a,s TTYACCGARGCXATGAARATG and enolase2I).All substitutions wereweightedequallyand ens552i s TTYACCGARGCXATGMGXATG the few included gaps were treated as missing data. ens791 a,s TACGATTTGGACTTCAAGA Heuristic searches used 100 random-addition-sequence ena385 s GCCARATCNGCAATGTGTYTGTAAAGTGG startingtreesandstartingfromrandomtreeswithnomaxtrees ena493 s ATCATRAAYTCYTGCAT ena474 s CATGAATTCCTGCATGGCCAGCTTGTTGCC limit, with TBR branch swapping on best trees only. These ena544 a,s CGTCCATRCCAATYTCAAT same parameters were also used with an implementation of ena776 a,s TTYGGRTTCTTGAARTCCA the parsimony ratchet procedure (courtesy of P. Lewis and ena764 a,s TCTTGAAGTCCAAATCGTA D. Sikes, Univ. Connecticut; based on Nixon 1999), with ena886 a,s CCAGTCRTCYTGRTCRAAXGG ena1014 a ATCTGGTTGACCTTCAGSAGSAGGCA 1000 replicates and 15% weighting. For bootstrap analyses, ena1170 a,s GGAGCACCGGTCTTGATCTGACC 1000 pseudoreplicatesweregeneratedwith10randomtaxon ena1223 a,s CTGGTTGTACTTGGCCAGACGCTC additions. For incongruence testing among the six datasets ena1228 a,s CTCCTCCTCAATGCGCARGATCTG the ILD test (Farris et al. 1995) was used as implementedin EVOLUTIONOF BEETLEAGRICULTURE 2015 FIG. 5. (a) Divergence by gene. Uncorrected nucleotide divergence (p, uncorrected pairwise divergence) between two taxa for each gene region plotted against the HKY (cid:49) (cid:71) corrected pairwise distance for COI for that taxon pair. (b) Proportions of overall gene divergence by codon position. Numbers on the y-axis indicate the proportion of a given gene’s changes at a particular codon position. Uncorrected total number of changes, summed over all taxa, are reported above the bars. The striped bar in each bar group depicts the relative proportion of changes in first and second codon positions combined with respect to those occurring in third codon positions. PAUP,using100replicatesand50randomadditionsequenc- For comparisons of ambrosia-fungus phylogeny with bee- es in each replicate (TBR, maxtrees (cid:53) 2000, excluding un- tlephylogeny,weobtainedthe18Ssequencesfortheophios- informative characters). tomatoid fungi from TreeBase (matrix accession number To create the constraint trees for the nodes from the com- M712; Berbee and Taylor 2001) and additional sequences bined maximum-parsimony tree and to calculate decay in- for Ophiostoma and all available Ceratocystis and Raffaelea dexes (Bremer 1994), we used Autodecay 4.0 (Eriksson sequencesfromGenbank(CeratocystisvirescensU32419,C. 1998), again with 100 random additions for the heuristic fimbriata U43777, Ambrosiella brunnea U40023, A. ferru- PAUP runs (TBR limited to 106rearrangementsperaddition ginea U40016, A. gnathotrichi U40015, A. hartigii U40017, sequence replicate). A. ips U40018, A. macrospora U40019, A. sulcati U40020, Totestwhetherthedistributionofeachofthreelife-history A.sulfureU40021, A.xyleboriU40022,Leucostomapersonii traits (host used: conifers vs. angiosperms; feedingsubstrate M83259, Ophiostoma bicolor AB007666, O. europhioides used:phoem,xylem,ambrosia,pith,orseeds;andinbreeding AB007667,O.penicillatumAB007668,O.piceaeAB007663, vs. outcrossing; Wood 1982; Kirkendall 1993) on the topol- O. piliferum U20377, Raffaeleaalbimanens U44474, R.arxii ogy differs significantly from a randomly distributed char- U44475, R. canadensis U44480, R. santoroi U44477, R. sul- acter, we compared the observed number of changes sepa- cati U44481, R. tritirachium U44478). The very similar 18S rately for each of these three characters on the tree with a sequences for 28 fungal taxa werealigned usingSequencher randomizeddistributionoftherespectivestatesofeachchar- 3.1 and yield a matrix of 1960 characters (174 informative) acter produced using the PTP utility in PAUP and holding thatwasusedinmaximum-parsimonyanalyseswithheuristic their relative frequencies and the beetle topology constant searchparametersandbootstrapanalysesasabove.WithMo- (Kelley and Farrell 1998). deltest 3.0 (Posada 1998) the following parameters werese- 2016 BRIAN D. FARRELLETAL. TABLE 2. Sister group comparison of diversity of angiosperm feeding and conifer feeding. Comparisons 1–4 are from thepresentstudy,5– 10 are published in Farrell (1998a). Comparisons 1, 5–10 represent shifts from conifers to angiosperms, and 2–4 represent shifts from angiosperms to conifers. Angiospermfeeding Gymnospermfeeding 1. Scolytinae ((cid:50) Ipini) 5200 Hylastini(cid:49) Tomicini 180 2. Acanthotomicus(cid:49) 95 Orthotomicus 11 Premnobius 24 3. Xyleborini/Dryocoetini 1500 Ipini ((cid:50) Acanthotomicus(cid:49) Premnobius) 195 4. Corthylina 458 Pityophthorus 200 5. Apion 1500 Antliarhininae 12 6. Belinae 150 Allocoryninae(cid:49) Oxycorinae 30 7. Higher Curculionidae 44,002 Nemonychidae 85 or Anthribidae 3000 8. Higher Cerambycidae 25,000 Aseminae(cid:49) Spondylinae 78 9. Megalopodinae 400 Palophaginae 3 10. Higher Chrysomelidae 33,400 Aulacoscelinae(cid:49) Orsodacninae 26 lectedforthe18Sfungaldataset(TrN(cid:49)I(cid:49)G,generaltime themostlikelymodel(maximum-likelihoodparameters),and reversiblemodel,submodelabaaea,estimatingtheproportion then the selected model was used to select among the max- ofinvariablesitesandtheshapeofthe(cid:71)parameter,withthe imum-parsimony trees (see Results) obtained from the com- empirical base frequencies, and estimating the ts/tv ratio). bined analyses. Branch lengths were optimized on the se- This model was used to choose between the maximum-par- lected tree using the corresponding parameters. Linear cor- simonytreesobtainedfromthe100randomaditionsequences relation was performed for the 18S maximum-likelihood search. Maximum-likelihood branch-length optimizations branchlengths optimized on the fungal 18S topology andthe wereperformedonthetopologyofthemostlikelytreeusing 18S beetle branchlengths optimized on the combined beetle the Describe Tree feature in PAUP. These were then used topology from the root to each origin of ambrosia beetles with theBerbeeandTaylor(1995,2001)calibrationforfun- and beetle associated fungi. gal 18S. Ages for nodes in Figure 7 were inferred using the divergencetimebetweenNeurosporaandHypomycesinBer- RESULTS bee and Taylor (2001) and the maximum-likelihood-opti- Ourmatrixof159previouslypublishedsequencesand106 mized branch lengths under a molecular clock. Standard er- new sequences, across six genes, contains 6104 characters rors correspond to branch length/age ranges given by the with 1613 parsimony-informative sites (see Appendix for maximum-likelihood optimization on the maximum-parsi- Genbank Accession nos.). Partition homogeneity tests indi- mony fungal topologies. cate no significant incongruence among the six datasets (P (cid:53)0.0698).Fromthe1936bpofthe18Salignment,272bases Diversity Tests were excluded because they could not be unambiguously Each of the life-history traits of interest were mapped on aligned,producingafinalmatrixof1664characters.The28S the most parsimonious tree, and sister groups identified for alignment produced a matrix of 993 bp, which displays two diversity contrasts (Tables 2, 3). Tallying of relative diver- 400-bp regions of great apparent homology plus one much sitiesofthetribesandgenerarequiresassumptionsofmono- morevariableregionwithambiguousalignment.Thevariable phyly of taxa not included in our analysis. Inasmuch as this regionwasexcludedfromtheanalysis,leaving892positions is true of both sides of each sister group comparison, this includedintheanalysiswithafewgapsanalyzedasmissing approach should not bias the results. data. For EF-1(cid:97), evidence of two loci that differ in intron/ exon structure was found in somemembersofthesubfamily Relative Age of Ambrosia Associations Scolytinae(Normarketal.1999).Forthisstudyweusedonly thecopyhavingoneintronbetweencodingpositions753and Toevaluatewhetherthereisacorrespondencebetweenthe 754.Twocopiesofenolasewerediscovered(enolase1niand ages of origin of ambrosia fungi and the origin of ambrosia enolase2I)andwereanalyzedseparately(seeFig.4forintron associated beetles, the following analysis was carried out: structure diagrams). Although considerable variation was Modeltest 3.0 was used on the 18S beetle dataset to choose foundinintronlengths,thealignmentofthecodingsequenc- es was unequivocal. All introns were removed prior to phy- TABLE3. Thediversityofambrosia-feedinggroupsiscontrastedwith logenetic analysis. the diversity of their sister groups of other habits. Compared Gene Divergences Ambrosia beetlegroup Diversity Sistergroup Diversity The uncorrected pairwise distance against the HKY (cid:49) (cid:71) Xyleborini 1300 Ozopemon 25 corrected pairwise distance for COI for each pair plotted in Corthylina 458 Pityophthorus 385 Figure 5a indicates a trend in decreasing rate of evolution Platypodini 1500 Scolytini 192 from 28S, COI, the enolases, EF-1(cid:97), and 18S. When com- Premnobius 24 Acanthotomicus 95 paring the proportions of overall gene divergence by codon EVOLUTIONOF BEETLEAGRICULTURE 2017 position in the four protein-coding regions (Fig. 5b), COI to be consistently more diverse (Table 2, P (cid:44) 0.001, sign displays the highest ratio of first and second codon position test), consistent with the general pattern in beetles (Farrell substitutions to third codon position substitutions (0.339 vs. 1998a). The seven origins of ambrosia feeding arelessclear 0.178 for EF-1(cid:97)), suggesting that a greater proportion of (Table 3). Compared to their respective sister groups, Xy- changescouldbeobscuredduetomultiplehitsatthirdcodon leborini, Platypodini, and Corthylina are more diverse, but positions.ToreducetheeffectofsaturationweusedtheCOI Premnobius is less diverse. The three remaining origins of aminoacid sequences in the combined dataset. ambrosia beetles represented in our samples are much less confidently placed. Of these the Xyloterini is more diverse Combined Phylogenetic Analyses than the species of Hylesinus, Scolytoplatypodini is either more or less diverse than these respective genera of Micra- The parsimony searches (100 random addition sequences) cini, depending on which represents the actual sister group, starting with random trees resulted in 33 maximum-parsi- and Sueus is less diverse than the Ipini (cid:49) Xyleborini. mony trees of length 11,319 steps (the parsimony ratchet found trees four steps longer). Tree support varied among nodes at different levels of relationship, with strongest sup- Fungal Phylogeny and Age of Ambrosia Associations portfornodesatsomeofthedeepestandmostshallowgroup- We found 18 maximum-parsimony trees in our reanalysis ings (Fig. 6). of 18S sequences for Ambrosiella, Raffaelea plus the other The phylogeny estimate is in broad agreement with the ophiostomatoidgenera,andoutgroups,andthesedonotcon- outline provided by Wood (1982) and Nobuchi (1969), with flictwiththepreviouslypublishedseparateanalysesofthese some notable exceptions. Of the tribes recognized by Wood genera (Cassar and Blackwell 1996; Jones and Blackwell and represented by multiple genera in our sample, several 1998). As reported by M. Blackwell and associates, we find are monophyletic in our results (Hylastini,Hypoborini,Cor- threeseparateoriginsofAmbrosiellacultivars,withnearrel- thylini, Crypturgini, Xyloterini, Platypodini); a few are par- atives from Ophiostoma and Ceratocystis (Fig. 7). The Am- aphyletic with respect to a few close relatives(Phloeosinini, brosiella associated with Corthylini shares an ancestor with Ipini, Dryocoetini), and one is paraphyletic with respect to Raffaelea associates of Platypodinae. The best fitting maxi- all other bark beetles (Tomicini). Three tribes are strikingly mum-likelihood model is the general time reversible (GTR) polyphyletic in our results: Xyleborini, Cryphalini, and Hy- model estimating the proportion of invariable sites and es- lesinini. Xyleborini is polyphyletic due to the distant place- timating the shape of the (cid:71) parameter, with the empirical ment of Premnobius, whose membership in Xyleborini had base frequencies, and estimating the ts/tv ratio. Using this long been controversial (Normark et al. 1999). Polyphyly of model to calculate branchlengths for the best fitting maxi- Cryphalini and Hylesinini is less well supported, the hy- mum-parsimony tree, and the Berbee and Taylor (2001)cal- pothesis of monophyly for the Hylesinini has been rejected ibration, the age estimates for ambrosia fungi are 60 ((cid:54)7.9) in previous studies where topologies constrainingthemono- million yearsfortheAmbrosiellaand Raffaeleaassociatesof phylyofthegroupwerefoundtobesignificantlylongerthan the maximum-parsimony tree (P (cid:53) 0.001, Sequeira et al. Corthylini and Platypodinae, 35 ((cid:54)4.3)million yearsforthe Ambrosiella cultivars associated with Xyleborini, and 21 2000). This can be due to the inclusion of several members ((cid:54)2.7) million years for the presumably facultative Ambro- of the extremely species-rich and morphologically diverse siella associates of Ips. Significant correlation was found of Hylesinopsis (R. Beaver, pers. comm.). Of the subfamilies the maximum-likelihood-optimized 18S branch lengths of recognizedbyWood,onlyPlatypodinaeismonophyletic;the ambrosiabeetleswiththoseoftheirassociatedambrosiafun- tribes of Hylesininae and Scolytinae are interdigitated to gi (R2 (cid:53) 0.8872, P (cid:53) 0.019; Fig. 8). someextent.StrikingsimilaritiestoWood’stopologyinclude the basal position of Tomicini, Hylastini, and most Hylesi- nini, and the close relationship of Platypodinae to Scolytini, DISCUSSION also suggested in Kuschel et al. (2000). OurestimateofScolytinaephylogenyilluminatessomeof Primary associations with conifers versus angiosperms the consequences of shifts in resource use and mating sys- showsignificantdeviationfromarandomdistributionacross tems. As in other groups of beetles and other insects, much thephylogeny(PTPtest,P(cid:53)0.001;KelleyandFarrell1998). of thediversificationinlife-historytraitsandlineagesseems Conifer associations are basal in the group, and followedby occasioned by use of angiosperm hosts (Farrell 1998a). several shifts between angiosperms and conifers (Fig. 6). Feeding substrate (phloem, xylem, pith, seeds, and am- Ambrosia brosia) also shows significant deviation from a random dis- tributionwithrespecttothephylogeny(PTPtest,P(cid:44)0.001). Fungusgardeninghasevolvedatleastseventimesinthese Phloem feeding is basal, followed by seven separate origins beetles (Fig. 6). It is not yet clear whether adoption of the of ambrosia feeding, all unreversed (Fig. 6). fungus-gardening habit generally enhances diversification Distribution of breeding systems(inbreedingvs.outcross- rate(Table3),buttheseobligateassociationswithfungispan ing) also departs significantly from a random distribution a range of ages that underscorestheirapparentstability.The with respect to the phylogeny (PTP test, P (cid:44) 0.001). three largest radiations, Platypodinae, Xyleborini, and Cor- thylini, are primarily tropical and comprise98% ofthe3400 Sister Group Diversity Tests described ambrosia beetle species, whereas the three small Thethreesistergroupcontrastsbetweenangiosperm-feed- groups, Xyloterini, Scolytoplatypodini, and Hyorhynchini, ingandconifer-feedinglineagesshowtheangiospermfeeders haveahigherproportionoftemperatespecies.Thetemperate 2018 BRIAN D. FARRELLETAL. FIG. 6. Strict consensus of 33 most parsimonious trees from the combined analysis of COI amino acids, 18S and 28S (hypervariable regionsexcluded),EF-1(cid:97)(intronexcluded),andbothcopiesofenolase(enolase1niandenolase2I;intronsexcluded).Length(cid:53)11,319 steps; CI (informative characters only) (cid:53) 0.2995; RI (cid:53) 0.4573. Above each internal branch are the bootstrap values, and the decay indices are below, both for the combined analysis. Branch color for the nonambrosia-feeding Scolytinae and Platypodinaeindicatesthe hostplantgroupfortherespectivegenera(black,conifers;gray,angiosperms).Ambrosiafeedingisindicatedbystripedbranches,whereas direct feeding on other host tissues is indicated by hatch marks with letters (F, phloem; X, xylem; P, pith; S, seeds; letters are marked EVOLUTIONOF BEETLEAGRICULTURE 2019 FIG.7. Oneofthe18mostparsimonioustreesfromtheanalysisofthe18Sdatasetforophiostomatoidfungiandalliedgenera(Length (cid:53)623steps)displayingthemostlikelyscoreunderthemodelchosenforthisdatasetwithModeltest3.0(Posada1998).Branchlengths correspond to themaximum-likelihoodoptimizationofthe18Sdataoverthechosentopologyenforcingamolecularclock.Numberson the time scale correspond to the ages of bark beetle associated clades using the calibration for fungal 18S introduced by Berbee and Taylor (2001) and standard errors correspond to branch length/age ranges given by the maximum-likelihood optimization on the 18 maximum-parsimony fungal topologies. zone is generally drier than the tropics and less conducive ago, soon after the onset of diversification of conifers, the to fungal growth, which may explain the greater abundance principalhostsofOphiostoma(Harrington1993;Kile1993). and diversity of ambrosia beetles in the tropics (Beaver Thus,thesepathogenspredatetheScolytinaeandwereprob- 1989). However, further investigation oftherelationshipsof ably vectored by other insects, perhaps including the weevil temperate and tropical ambrosia beetles would be required antecedentsofScolytinae.Theoldestoriginofambrosiafun- todeterminewhetherdiversificationratesareactuallyhigher gi is the early Tertiary, some 60 million years ago, in the in the tropics. platypodine-associated Raffaelea and the related corthyline- The two genera of ambrosia fungi, Raffaelea and Ambro- associated Ambrosiella (Cassar and Blackwell 1996; Jones siella,arebothpolyphyletic,andeacharoseatleastfivetimes and Blackwell 1998; Fig. 7). The genus Ambrosiella, asso- (CassarandBlackwell1996;JonesandBlackwell1998)from ciated with the scolytine ambrosia beetle tribes Xyloterini, within the ophiostomatoid clade that includes Ophiostoma Xyleborini, and Corthylini (and apparently, Scolytoplaty- and Ceratocystis (Berbee and Taylor 1995, 2001). The podini: Kinuura and Hijii 1991), comprises two primary ophiostomatoids apparently arose some 200 million years clades of which one is associated with Xyleborini and Xy- ‹ with a (cid:49) if the group is known to contain species that feed on different tissues). The presence of inbreeding in a group is indicatedby anasterisk.Barsbesidetaxonnamesindicatetribe,subfamily,andfamilyclassificationafterWood(1986)andWoodandBright(1993). 2020 BRIAN D. FARRELLETAL. genera, which are consistent with rapid radiation (Schedl 1956;Browne1961;Nobuchi1969;Beaver1989;Woodand Bright 1992). Whereasthespeciesofambrosiabeetlesareextremelyuni- form, the diversity of their mycangia is striking (Francke- Grossman 1967; Beaver 1989). Within the Platypodinaeand Xyleborini,themycangiaofverycloserelatives(i.e.,sharing an ambrosia-associated ancestor) often occur in different body regions (e.g., the mouthparts, basal leg segments, tho- rax,orelytra; Francke-Grossman1967;Beaver1989)ordif- fer between the sexes. These mycangia have thus evolved very rapidly. Tests of the hypothesis that these fungi and their herbivores are undergoing rapid turnover could draw onpatternsinmolecularevolutionandanalysesofdispersion on the phylogeny (Moran 1996; Kelley and Farrell 1998). There are a few species that apparently lackmycangia(Bea- FIG. 8. Maximum-likelihood (ML) optimized branch lengthscal- culatedontheMPselectedfungal18Stopologyandtheoptimized ver 1989), but most species have not been subjected to the branchlengthsfor18Sonthebeetletopologyforeachofthebeetle- detailed histological studies required to detect and demon- ambrosia associations show significant linear correlation (R2 (cid:53) strate mycangial presence. The widespread distribution of 0.8872, P (cid:53) 0.019). mycangia (across Xyleborini, Ipini, Drycoetini, Xyloterini, Corthylini, Tomicini, Scolytoplatypodini, Cryphalini, Both- loterini and the other is associated with Corthylini. Because rosternini, Hylesinini, and the Platypodinae) suggests the Xyleborini and Xyloterini represent independent origins of presenceofmycangiaclosetotheancestoroftheScolytinae, theambrosiabeetlehabit,lateraltransmissionorindependent which was evidently feeding on phloem of conifers in as- acquistion of ambrosiafungi has apparentlyoccurred,ashas sociation with ophiostomatoid fungi. However, the origins been also been demonstrated for the fungi associated with of ambrosia feeding all followed shifts to angiosperms (al- attine ants (Mueller et al. 1998). though some temperate ambrosia beetle species are able to ThemostrecentoriginofAmbrosiellacomprisesfungithat use conifer hosts; Wood 1982). are associated with the largely conifer phloem-feeding Ips and Hylurgops, also associated with the largely conifer-at- Shifts between Conifers and Angiosperms tackingOphiostoma,andmaythusrepresenttheinitialstages of obligatefungalmutualisms(LutzoniandPagel1997).Al- The estimate of Scolytinae phylogeny supports the hy- though strictly asexual lineages are susceptible to accumu- pothesis thatshiftsfromconiferfeedingtoangiospermfeed- lationofdeleteriousmutations(Moran1996)andarethought ing have tended to enhance rates of species diversification togoextinctrapidlyinmostcases(BartonandCharlesworth in beetles (Fig. 5, Table 2) and other herbivores (Farrell 1998),ambrosiafungiappeartohavepersistedsincetheearly 1998a, 1999). While the shifts from coniferstoangiosperms to mid Tertiary, like a few other asexual clades (Judson and in the hylesinine tribes and within Ipini further support the Normark 1996; Mark Welch and Meselson 2000). hypothesis that use of angiosperms fosters insect diversity, The correlated branchlengths (Fig. 8) and inferredagesof ourresultsforscolytinesstrengthenthispatternbyproviding ambrosia beetles and fungi (Fig. 7) apparently reflects their evidence that shifts back to conifers following angiosperm co-descent from less tightly associated ancestors and cor- colonization areassociatedwithlowereddiversity(Table2). roboratesourviewthat30millionyearoldDominicanamber Both early Mesozoic ((cid:59)200 million years ago;Farrell1998) is too ancient to have captured the Xyleborini (Jordal et al. and late Tertiary (10–20 million years ago; this study) col- 2000).Thus,Xyleborini,Corthylini,andPlatypodinaeareall onizations of conifers are associated with lower diversity, abundantambrosiabeetlesonHispaniolatoday,andallattack suggesting that the number of conifer host species is the resinous legumes (as well as many other plants) related to limitingqualityofthisplantgroup,ratherthantheparticular the source of Dominican amber (Wood and Bright 1992; time period in which shifts occur. Bright and Poinar 1994). Although the corthyline and pla- Nevertheless, the three independent conifer-associations typodinefossilsinDominicanamberarecommon,Xyleborini (Hylastini (cid:49) Tomicini, Ipini, and Pityophthorus) each com- is conspicuously absent from these amber deposits. The or- priseapproximately200species,andthusareamongthemost igin of 1300 xyleborine species in 20 million years is more species-rich conifer associations known. Many other small than twice the rate for the Platypodinae (1500 species in 60 groups of conifer-associated beetles occur in otherwise an- millionyears)andmayreflectthecombinationofpolyphagy giosperm-affiliatedcladesintheScolytinaeandothergroups. enabled by ambrosia-feeding together with inbreeding and Thesebark beetlescollectivelyusenearlyallofthe300spe- haplodiploidy in this group (Jordal et al. 2000). Moreover, ciesofconifers(e.g.,someDendroctonususeupto30species the colonizing ability of xyleborines afforded by the com- of Pinus; Kelley and Farrell 1998), and most conifers have bined inbreeding and haplodiploidy may also explain their multiplebeetleassociates(SturgeonandMitton1982;Bright disproportionate representation on islands (Jordal et al. andStock1982;WoodandBright1992).Therelativelygreat 2001). Indeed, both Xyleborus and the outcrossing Platypus diversity of scolytines associated with conifers may reflect arefamouslyspecies-richandmorphologicallyhomogeneous their very small body size range (1–11 mm). Thus, different