Blackwell Science, LtdOxford, UKBIJBiological Journal of the Linnean Society0024-4066The Linnean Society of London, 2004? 2004 834 441452 Original Article
MULTIPLE ORIGINS OF GALLING IN ERIOCOCCIDS L. G. COOK and P. J. GULLAN
Biological Journal of the Linnean Society, 2004, 83, 441–452. With 2 figures
The gall-inducing habit has evolved multiple times among the eriococcid scale insects (Sternorrhyncha: Coccoidea: Eriococcidae)
L. G. COOK* and P. J. GULLAN†
School of Botany and Zoology, The Australian National University, Canberra, ACT, 0200, Australia
Received 7 August 2003; accepted for publication 20 February 2004
The habit of inducing plant galls has evolved multiple times among insects but most species diversity occurs in only a few groups, such as gall midges and gall wasps. This phylogenetic clustering may reflect adaptive radiations in insect groups in which the trait has evolved. Alternatively, multiple independent origins of galling may suggest a selective advantage to the habit. We use DNA sequence data to examine the origins of galling among the most spe- ciose group of gall-inducing scale insects, the eriococcids. We determine that the galling habit has evolved multiple times, including four times in Australian taxa, suggesting that there has been a selective advantage to galling in Australia. Additionally, although most gall-inducing eriococcid species occur on Myrtaceae, we found that lineages feeding on Myrtaceae are no more likely to have evolved the galling habit than those feeding on other plant groups. However, most gall-inducing species-richness is clustered in only two clades (Apiomorpha and Lachnodius + Opisthoscelis), all of which occur exclusively on Eucalyptus s.s. The Eriococcidae and the large genus Eriococcus were determined to be non-monophyletic and each will require revision. © 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 83, 441–452.
ADDITIONAL KEYWORDS: adaptive radiation – Apiomorpha – Eriococcus – Eucalyptus – galling – host specificity – Myrtaceae.
INTRODUCTION Phylogenetic clustering of gall-inducing species is also apparent among the scale insects. Although gall The habit of inducing galls on plants has evolved mul- inducers are known from ten coccoid families, the tiple times among phytophagous insects, occurring in highest proportion (about 47%) of the more than 230 at least six orders. However, although there are more gallicolous species (Ferris, 1950; Beardsley, 1984; Gul- than 13 000 species of gall-inducing insects (Dreger- lan, 1984a; Ben-Dov, Miller & Gibson, 2003; Gullan, Jauffret & Shorthouse, 1992), most species diversity Miller & Cook, 2004) belongs to the Eriococcidae. This is clustered in only a few insect groups, such as proportion is even greater (about 57%) if species that the gall-midges (Diptera, Cecidomyiidae) (Dreger- produce only simple pits or depressions are excluded, Jauffret & Shorthouse, 1992) and the gall-wasps because almost all eriococcid gallers induce complex (Hymenoptera, Cynipidae) (Ronquist & Liljebblad, galls. The gall shapes are recognizably species-specific 2001). The phylogenetic clustering suggests that the and structural complexity is often considerable, with taxonomic distribution of species diversity reflects the gall induced by a female sometimes resembling a radiations in those groups that have evolved the habit fruit or other organ typically not found on that host rather than many independent origins of the ability to plant. The shape of the gall appears to be independent induce plant galls. of the host-plant species and the host-plant organ on which the gall grows and is determined by the species and sex of the scale insect (Gullan et al., 2004). Sexual dimorphism of gall morphology is apparent among *Corresponding author. E-mail: [email protected] †Current address: Department of Entomology, University of most eriococcid species in which males are known to California, Davis, CA 95616-8584, USA induce galls.
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 83, 441–452 441
442 L. G. COOK and P. J. GULLAN
The high number of galling eriococcids is not 2. multiple origins of galling in separate lineages, because of a numerical dominance by the Eriococcidae, suggesting that a common factor, such as a partic- which is only a distant fourth among the scale insect ular environmental condition, may have promoted families in terms of total number of species (Ben-Dov the evolution of the gall-inducing lifestyle. et al., 2003). Therefore, another explanation is needed The study includes a broad taxonomic and geograph- to explain the prevalence of gall-inducing taxa in this ical range of eriococcids, both galling and non-galling. group. The speciose genus Apiomorpha is represented for the Gall-inducing insects in general are prevalent in first time in a molecular analysis of the Eriococcidae. regions dominated by sclerophyllous vegetation (Price et al., 1998; Wright & Samways, 1998), leading to an argument that there is a selective advantage to galling METHODS in these habitats or intrinsic features of sclerophyll communities that may favour speciation among SPECIMENS AND DNA EXTRACTION gallers (e.g. Mendonça, 2001). Although only 27% of Exemplars were chosen to sample the diversity of the eriococcid species worldwide occur in Australia, more Eriococcidae and to increase representation of groups than 80% of eriococcid gall-inducers occur in Australia whose placements in an earlier molecular study (Cook (Miller & Gimpel, 2000; Gullan et al., 2004), predom- et al., 2002) were controversial. In particular, this inantly on sclerophyllous plants of the Myrtaceae. If study includes Eriococcus williamsi Danzig, the mor- the gall-inducing eriococcid taxa belong to only one or phological sister taxon of E. buxi, and both described a few lineages, then an extensive radiation in those species of Cylindrococcus as well as additional repre- groups may explain the prevalence of galling. How- sentatives of Beesoniidae and Stictococcidae. Pseudo- ever, if galling has evolved many times within Erio- coccidae, previously identified as a suitable outgroup coccidae another explanation, such as common (Hodgson, 2001; Cook et al., 2002), was used to polar- ecology, may be required. ize trees. The taxa used, their current taxonomic clas- Several recent studies have conflicted in their sug- sification, hosts, gall type (if applicable) and collection gested relationships among Australian gall-inducing localities are given in Table 1. Voucher specimens of eriococcids. A molecular phylogeny of the scale insects Australian taxa will be lodged in the Australian (Cook, Gullan & Trueman, 2002) suggests that galling National Insect Collection (ANIC), CSIRO, Canberra, has evolved at least twice among the eriococcids. Australia. Although some of the Australian eriococcid gallers Embryos and/or ovarian tissues were dissected from formed a monophyletic group, the Casuarinaceae- larger females. Small females and eggs were homoge- feeding galling genus Cylindrococcus appeared to be nized whole. DNA was extracted from fresh or etha- more closely related to the non-galling type species of nol-preserved specimens of embryos, eggs and/or Eriococcus (E. buxi Boyer de Fonscolombe), and to ovarian tissues using the salting-out method of Sun- Beesoniidae and Stictococcidae. However, only one nucks & Hales (1996). A single chloroform wash was representative of each of the two latter taxa was performed prior to precipitation if solids or excessive included in the study and Apiomorpha, the most spe- pigments were present. ciose gall-inducing scale insect genus, was not repre- sented at all. Hodgson’s (2001) phylogenetic study of adult male morphology also places Beesoniidae PCR AND SEQUENCING (which contains several gall-inducers) and Stictococ- The 5¢ region of the small subunit ribosomal RNA gene cidae among the eriococcids but generic placements (SSU rRNA) was amplified using the primers 2880 differ from those in Cook et al. (2002). Hodgson’s [for] (5¢-CTGGTTGATCCTGCCAGTAG) (Tautz et al., study suggested that the Australian gall-inducing 1988) and B- [rev] (5¢-CCGCGGCTGCTGGCACCAGA) genera Apiomorpha, Cystococcus, Lachnodius and (von Dohlen & Moran, 1995). Two new internal Opisthoscelis were related to each other. However, primers (E10for: GGAAGGAACGCTCTTATTAG, Hodgson’s results should be treated with caution and E10rev: CGGTTTTGATCTAATAAGAGC) were because male morphology may be subject to conver- designed for this region for Apiomorpha. An internal gent selection in galling taxa. region of Elongation Factor 1-alpha (EF-1a) was In this study, we used DNA sequence data from amplified using primers M44-1 and rcM52.6 (Cho three gene regions (two nuclear and one mitochon- et al., 1995). These primers did not amplify EF-1a drial) to investigate whether the observed species in all scale insects and new primers (EF-1a[a]for: richness of gall-inducing eriococcids in Australia may GATGCTCCGGGACAYAGAG and EF-1a[c]rev: best be explained by: GGTTTCAAGACACCAGTTTC) were designed inter- 1. an evolutionary radiation (possibly adaptive) nal to the original primers. An internal region of the within a single lineage, or mitochondrial COII gene was amplified using the
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 83, 441–452
MULTIPLE ORIGINS OF GALLING IN ERIOCOCCIDS 443 primers C2-J-3400 (Liu & Beckenbach, 1992) and C2- mented in PAUP* 4.0b10 (Swofford, 2002) for each full N-3661 (Simon et al., 1994). data set and separately for third positions in COII Each reaction (25 mL) contained 5 pmol of each and EF-1a. Substitutional saturation was assessed primer, 10 mM Tris/HCl (pH 8.3), 3 mM MgCl, 50 mM by plotting the transition–transversion ratio (ti/tv) KCl, 0.2 mM of each dNTP, 1 unit of Taq-polymerase against genetic distance and checking whether the (Fisher Taq F1, Biotech) and 2 mL of template. Ampli- curve had levelled. fication was carried out in a Corbett Research FTS-960 A variety of different approaches to phylogenetic thermal sequencer using 45 s denaturation at 94 ∞C, reconstruction were used, primarily to guard against 30 s extension at 72 ∞C and using a 5 ∞C stepdown pro- undetected violations of underlying assumptions in gramme for annealing, with the first cycle at 65 ∞C and any one model. It was assumed that, when different annealing in later cycles reduced by 5 ∞C after every models and weighting schemes resulted in the same second cycle until reaching 55 ∞C (40 ∞C for COII). An topology, the resultant phylogeny was robust in additional 35 cycles were run with annealing set at respect of inference method. Robustness does not 55 ∞C (40 ∞C for COII). PCR products were precipitated imply that the phylogeny is accurate. Congruence and purified using ammonium acetate/ethanol (1 : 10) among data sets provides increased confidence of rela- precipitation followed by a 70% ethanol wash. PCR tionships (Miyamoto & Fitch, 1995). products that contained minor bands were individu- Bootstrap tests (Felsenstein, 1985) were conducted ally excised from a 1% agarose gel following electro- for maximum parsimony (MP) and distance-based phoresis and purified using a BRESAclean DNA analyses using 1000 replicates and the same settings purification kit (GeneWorks cat. no. BT-3000) follow- as the original analyses. The inferred amino acid ing the manufacturer’s instructions. A 10-mL sequenc- sequences were not used for phylogenetic analyses ing reaction was carried out using ABI (Perkin-Elmer) because the nucleotide compositional bias, and/or BigDye v.2 terminator chemistry following the manu- codon bias, may affect the inferred amino acid facturer’s instructions. Sequencing was performed sequence (Foster, Jermiin & Hickey, 1997; Singer & using an ABI Prism 377 automated DNA sequencer. Hickey, 2000) and therefore exacerbate, rather than All DNA fragments were sequenced in both the for- correct for, compositional bias. ward and the reverse directions.
BAYESIAN INFERENCE PHYLOGENETIC ANALYSIS Bayesian analyses were performed using MrBayes Sequences were edited using Sequencher 3.02 (Gibbs (Huelsenbeck & Ronquist, 2001) assuming a GTR+G & Cockerill, 1995) and the protein encoding regions model. Each run comprised four chains starting from (COII and EF-1a) were checked for the presence of random trees, and 500 000 generations with trees stop codons in putative exons. A similarity search was saved at each ten. Five runs were performed for each performed on all sequences using the ‘basic blast’ data set to check for convergence among runs. The option (Altschul et al., 1997) at the NCBI website majority rule consensus tree was calculated after the (http://www.ncbi.nlm.nih.gov/BLAST/). The SSU removal of trees saved during the ‘burnin’ period. rDNA data were aligned by eye with reference to the predicted secondary structure model for eukaryote nuclear SSU rRNA (Van de Peer et al., 1998). Some MAXIMUM PARSIMONY highly variable regions, particularly E10 helices of V2, Heuristic searches comprising 100 random addition did not have the same predicted secondary structures sequence starting trees, TBR branch swapping and no and were not able to be unambiguously aligned across maxtrees restrictions were performed using PAUP* all taxa. Alignable groups of sequences were offset 4.0b10 (Swofford, 2002). The SSU rDNA alignment with respect to other such groups. The introns of EF- was analysed under two weighting schemes: equal 1a also were not able to be aligned across taxa and weights, and the V2 region weighted half that of the were excluded from further analyses. Sequence data more conserved regions. Two weighting schemes for from each gene region were analysed separately codon position (1st : 2nd : 3rd = 2 : 3 : 1 or 4 : 6 : 1) because not all taxa could be amplified for COII or EF- were used for the protein coding regions. Analyses 1a. Additionally, the three data sets represent two were also run for COII with third codon positions genomes under different modes of inheritance. excluded. Most methods of phylogeny reconstruction assume that the base composition is at equilibrium (station- ary), with the base composition of homologous DISTANCE sequences the same for all taxa under study. Station- The LogDet model (Lockhart et al., 1994) was used arity was tested using a chi-squares test as imple- because base composition of third codon positions of
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 83, 441–452
444 L. G. COOK and P. J. GULLAN ) a EF-1 Cactaceae, GenBank Accession ( No. = Sapindaceae, = ) COII GenBank Accession ( No. Myrtaceae, Myrtaceae, Sn Casuarinaceae, Ct Casuarinaceae, = = GenBank Accession (18S) No. Cornaceae, Cs Cornaceae, Lamiaceae, Lamiaceae, M = = N/A AY795513 AY791955 AY795493 stem swelling AY795525 – AY795488 enclosing AY795522 – – enclosing AY795524 – AY794591 enclosing AY795535 – AY795504 N/A AY795517 – – enclosing AY795523 AY791951 – enclosingenclosingenclosing AY795554/5 AY795551 AY791963 – AY795499 AY791962 AY795501 – AY795500 enclosing AY795552/3 AY791964 – Gall type N/AN/A AY795506 – AY795508 AY791948 – – M
Cs rosette AY795515 AY791956 AY795492 Cs rosette AY795516 AY791957 –
Icacinaceae, Icacinaceae, L W
M M M M M Cu N/A AY795505 – AY795483 =
M Cr N/A AY795547 – – M M
I Cu N/A AY795507 – AY795484 Ct N/A U20402 – –
Fg N/A AY795511 – –
† Ct N/A AY795538 – AY795502 M
? M N/A AY795518 – –
? D rosette AY795512 – – B
Sn N/A AY795539 – AY795498
elwitschia Acer Buxus Allocasuarina Leptospermum Griselinia Melaleuca Allocasuarina Corymbia Apodytes Baeckia Kunzea Corymbia Eucalyptus Eucalyptus Eucalyptus Eucalyptus Opuntia Opuntia Shorea Quercus Host genus Callitris Eucalyptus Calocedrus W Fagaceae, Fagaceae, I = Fabaceae, Fabaceae, Fg = Gullan Australia
Ericaceae, Ericaceae, Fb = Maskell Australia
Maskell Australia (Tepper) Australia
Beardsley * N/A U20429 – – Fuller Australia
illiams Australia akagi Malaysia (Froggatt) Australia (Maskell) Australia
roggatt Australia De Lotto Australia
T W
(Cockerell) *
F Brain South Africa
(Kuwana) Japan
(Maskell) New Zealand
(Maskell) Australia (Ehrhorn) USA (Froggatt) Australia
Schrader Australia
(Signoret) Hungary
(Boyer de Fonscolombe) France
sp. Australia
sp. Africa
Dipterocarpaceae, Dipterocarpaceae, E elwitschiaceae) = W
aracoccus = Eriococcus aceris Eriococcus buxi Eremococcus turbinata ‘callo2’ Eriococcidae sp. Australia Cylindrococcus spiniferus spinosus Eriochiton Cylindrococcus casuarinae Cystococcus echiniformis Calycicoccus merwei Calycicoccus Callococcus pulchellus Callococcus acaciae Apiomorpha spinifer Apiomorpha strombylosa Ascelis praemollis Apiomorpha munita tereticornuta Apiomorpha minor Dactylopius confusus Gallacoccus heckrothi Dactylopius austrinus Pseudococcus callitris Pseudococcus Beesonia napiformis Dysmicoccus neobrevipes Ehrhornia cupressi P Coccoidea species used in this study. Family classification of host genus is shown after name (Cr Family Coccoidea species used in this study. Cupressaceae, Cupressaceae, D Sapotaceae, Sapotaceae, W = = able 1. lassification Species Source *von Dohlen & Moran (1995). not native to collection locality. †Introduced species, Eriococcidae Dactylopiidae St Current family c Beesoniidae Pseudococcidae T Cu
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 83, 441–452 MULTIPLE ORIGINS OF GALLING IN ERIOCOCCIDS 445 ) a EF-1 GenBank Accession ( No. ) COII GenBank Accession ( No. GenBank Accession (18S) No. Gall type rosette AY795527 AY791950 AY795487 blister AY795519 – – N/A AY795541 – – rosette AY795526 – – N/A AY795548 – AY795496 N/A AY795544 – – N/A AY795549 AY791961 AY795495 N/AN/A AY795545 AY795546 – – – – pitenclosingN/A AY795520 AY795528 AY791954 AY795550 AY795489 – – AY795485 – N/A AY795531 – – N/AN/A AY795534 AY795529 – – – – N/Apit/enclosing AY795521 AY791953 AY795514 AY795490 – – N/A AY795530 AY791949 – N/A AY795543 – – N/A AY795542 AY791960 – N/A AY795540 AY791959 AY795497 St N/A AY795509 – AY795494 M
Cs N/A AY795532 – – Ct N/A AY795536 – AY795503
L
E P N N M M M M M M M M M
Cs N/A AY795533 AY791954 AY795486
M M
U Fb leaf fold AY795537 AY791958 –
Fb N/A AY795510 – – P
B b
odocarpus Host genus Chrysophyllum Melaleuca Acacia Ulmus Eucalyptus Hakea Olneya Pseudopanax Melaleuca Eucalyptus Eucalyptus Eucalyptus Leptospermum Casuarina Eucalyptus Nothofagus Eucalyptus Leucopogon Nothofagus P Eucalyptus Prostanthera Eucalyptus Buxus Mammillaria Eucalyptus Allocasuarina roggatt Australia roggatt Australia F Gullan & Strong Australia
F
Green Australia
Hoy New Zealand Maskell Australia roggatt Australia
Maskell Australia
Green Australia (Maskell) Australia
F
Cockerell Nigeria Danzig Georgia Cockerell USA Maskell Australia Maskell sp.2 Australia Maskell sp.1 Australia e (Maskell) New Zealand
roggatt Australia Hoy New Zealand Richard Kenya (Modeer) USA
Maskell Australia F
socialis pustulans ferrugineus ’ ’ ’ sp. ‘Nc’sp. Australia
sp. Australia sp. Australia
sp. ‘Hakea’sp. Australia sp. ‘CA1’sp. USA
Sphaerococcus Sphaerococcus Sphaerococcus Stictococcus sjostedti Hockiana insolitus Hockiana ‘ Eriococcus spurius ‘ Eriococcus ‘ Eriococcus Scutare lanuginosa Eriococcus irregularis Eriococcus leptospermi Eriococcus serratilobis Opisthoscelis mammularis Ourococcus Phacelococcus subcorticalis Eriococcus fossilis Madarococcus viridulus Madarococcus Lachnodius Madarococcus totara Eriococcus eucalypti Eriococcus williamsi Eriococcus coriaceus Eriococcus eucalypti Eriococcus tepperi Eriococcus coccineus Eriococcus casuarinae lassification Species Source Current family c *von Dohlen & Moran (1995). not native to collection locality. †Introduced species, Stictococcidae
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 83, 441–452 446 L. G. COOK and P. J. GULLAN both COII and EF-1a varied significantly among taxa taxa share the trait of interest. Host plants were (i.e. non-stationary). Characters were weighted by assigned to one of two categories: Myrtaceae or ‘other codon position (2 : 3 : 1) and a minimum evolution family’. Myrtaceae was treated separately because it search criterion was used. is the dominant host family in terms of eriococcid spe- cies richness. Myrtaceae was further divided into either Eucalyptus or ‘other genus’. Species of eriococ- TRAIT MAPPING cids were identified as being either galling or non-gall- The gall-inducing habit was mapped onto the SSU ing. Categories in which both galling and non-galling rRNA tree because it was generated using sequences taxa occur were scored in counts for both. If one host- from all taxa and was not strongly contradicted by plant taxon is more favoured during the evolution of either of the other two data sets. The putative origins the gall-induction habit it should support more lin- of galling were assessed by MP optimization using eages of gall-inducers than do other host-plant taxa. ACCTRAN (accelerated transition) and DELTRAN However, if a host-plant taxon is favoured by eriococ- (delayed transition) but other data, such as adult cids irrespective of whether or not they induce galls female morphology, also were considered. then this should to be taken into account in assessing Data on host use and gall-inducing habit (Tables 1 host-use bias. and 2) were obtained from Miller & Gimpel (2000), Gullan et al. (2004), ScaleNet (Ben-Dov et al., 2003) RESULTS and our unpublished records to determine whether there may be a host-preference bias among Australian There was no well-supported conflict among DNA gall-inducing eriococcids. Three hierarchical catego- sequence data sets (Figs 1, 2), although the resolution ries of taxon assemblages were used: species, genera among some of the terminal taxa differed. For exam- and lineages, with lineages being clades in which all ple, Apiomorpha was not recovered as a monophyletic group using EF-1a (Fig. 2A) despite strong support in the other two data sets. There was low support for Table 2. The host affiliations of all described species of most nodes in the COII data set (Fig. 2B), probably Australian Eriococcidae. Numbers in parentheses refer to owing to substitutional saturation in all codon posi- the number of taxa (galling/non-galling). Lineages are tions (>30% divergence). Similarly, third codon posi- those identified from phylogenies derived in this study and tions were saturated in EF-1a and there were only 25 do not include all described genera. Totals are not neces- potentially informative sites among first and second sarily the sum of observations because taxa with hosts codon positions. in more than one category are scored for each category. The gall-inducing eriococcids did not form a mono- Host genus is shown only for taxa feeding on species of phyletic group in any of the analyses. Three main Myrtaceae clades of eriococcids, each comprising both gall-induc- ing and non-galling taxa, were recovered from all Other plant three data sets. The ‘BSE’ clade comprises represen- Host family Myrtaceae families Total tatives of three previously recognized families: Beeso- species 125 (82/43) 32 (3/29) 154 (85/69) niidae, Stictococcidae and Eriococcus s.s. genera 17 (11/8) 5 (2/3) 20 (13/9) (Eriococcidae), thus rendering the Eriococcidae non- lineages 11 (7/6) 6 (3/4) 12 (10/9) monophyletic. The ‘A’ (acanthococcid) clade contains a range of taxa including Dactylopius (Dactylopiidae), the gall-inducing Apiomorpha and Eriococcus aceris Host genus Eucalyptus s.s. Other genera (Signoret), the type species of Acanthococcus if Erio- species 93 (66/27) 32 (16/16) coccus and Acanthococcus are recognized as separate genera 12 (7/6) 6 (5/3) genera (Melville, 1982). The ‘G’ (Gondwanan) clade lineages 6 (3/4) 6 (4/2) comprises species from Australia and New Zealand and contains the subclade ‘MF’ (Myrtaceae-feeding) in