蘇鐵綺灰蝶的來源檢測與綺灰蝶屬食性演化之研究

全文

(1)Contents 致謝…………………..……........……………………………………………………..2 致謝中文摘要…………………..…………………………………………………………..3 中文摘要 Abstract……………………………...………………………………………………..5 Introduction………………….……...………………………………………………..7 Part I. Elucidating origins of the Cycad Blue…………………………..…………10 I-1. Introduction……..…..…...………………………………………..…………10 I-2. Materials and methods……..…………………………………..…………15 I-3. Results……………………………………………………………..…………20 I-4. Discussion……...………………………………………………..…………25 Part II. Molecular phylogenetics of Chilades butterflies…………………………33 II-1. Introduction………………………………………………………..………33 II-2. Materials and methods………………………………………..…………36 II-3. Results…………………………………………………………..…………41 II-4. Discussion……...………………………………………………..…………44 References……...………………………………………………………..…………47 Tables……...……………………………………………………………..…………59 Figures...………………………………………………………………..…………75. 1.

(2) 致謝本論文是集眾人幫助與鼓勵方有的成果，絕非憑一己之力就可完成。首先要特別感謝徐堉峰、顏聖紘以及 David C. Lees 博士，在撰寫論文初稿期間多次提供珍貴的建議及修改的方向。尤其是徐老師，以身作則的研究態度以及對蝴蝶新知識的執著，讓我體認到作為一名研究者應有的專注與態度。感謝口試委員張慧羽、王震哲及林思民博士，花費不少時間閱讀論文初稿，並指出論文錯誤及不適的地方，使本論文更趨近成熟。感謝實驗室眾多成員及實驗室之友的協助，讓我在研究的初期學會如何做實驗，並且在畢業的最後幾年可以全心全力的撰寫論文。實驗室的大師兄呂至堅博士，與你在環島採集時的談話，讓我受益良多。立豪，謝謝你在我忙不過來的時候，出手協助環島採集的實驗。羅桑、孟娟、宏軒、小油龍、Mayday 及阿南，謝謝你們的協助採集，讓我的論文更充實。感謝亭瑋在分生實驗上鼎力幫助。謝謝育綺在標本製作上迅速完美。有你們的參與，我才能完成論文。感謝眾多學者與同好提供珍貴的樣本作為研究，你們無私的協助讓我非常感激。Dr. Hideyuki Chiba、Dr. Masaya Yago 以及 Dr. Tsuyoshi Takeuchi 提供日本、韓國、及東南亞等地的重要樣本。廖培鈞博士、謝佳宏博士提供大陸地區的樣本； Dr. David Wright 提供美國南部的重要樣本； Dr. Steve Collins 、 Dr. Ichiro Nakamura 提供馬達加斯加島的樣本；Dr. Aubrey Moore 提供關島的重要樣本。我的好朋友羅益奎、吳明聰、陳賢明、施致易、陳文選、陳南瑛、朱晉溶、洪淑珍，你們與你們的家人，在研究的過程提供我無以回報的協助，感恩。最後感謝我的爸媽，提供生活上所有需要的幫助，並且默默的在背後支持兒子所選擇的人生道路。謝謝婷文在我攻讀博士期間全力的支持與協助，你是我心靈最大的支柱。謹在此將成果獻給家人、師長與朋友們。. 2.

(3) 中文摘要因食物或園藝造景之用而引入的外來植物，雖然不一定會直接危害原生物種的生存，但可能間接地影響到利用這些植物的昆蟲，進而改變其演化過程。以蘇鐵綺灰蝶(Chilades pandava)為例：自 1990 年以來，台灣地區引入大量的蘇鐵屬(Cycas)植物作為庭園造景之用，直接或間接地造成蘇鐵綺灰蝶的族群快速擴張，現今已分佈全台。本論文的第一部分以蘇鐵綺灰蝶的事件來討論外來園藝植物對原生昆蟲族群之遺傳結構的影響，並追溯大發生族群的來源。綺灰蝶屬的成員為體型最小的蝴蝶之一，分佈範圍可從非洲西部橫跨到亞洲東部，並且除了一般專食豆科及芸香科植物的蝶種外，亦包含有較罕見的專食蘇鐵科植物的成員。由於綺灰蝶屬的分佈範圍和食草的利用關係並不常見於其它屬的蝴蝶中，因此懷疑綺灰蝶屬是否為單系群(monophyly)。本論文將在第二部分討論綺灰蝶屬成員的親緣關係，以暸解此屬的食草利用格局。第一部份將世界各地共 810 隻蘇鐵綺灰蝶樣本，進行粒線體 COII 序列分析，得有 29 個基因型，其中台灣各地大發生的族群所擁有的主要基因型 C 僅分佈在台灣。此結果支持台灣各地大發生的族群來自島內族群擴張，並非人為引進蘇鐵時所夾帶的外來族群；在第二部分，為了建構綺灰蝶屬內的成員以及食性利用之間的關係，共採用. 3.

(4) 13 種綺灰蝶屬的成員以及 42 種近緣的蝴蝶作為分析(3437bp, 82 taxa)，發現目前歸類為綺灰蝶屬的蝴蝶並非單系群。但若將處理為同物異名的晶灰蝶屬(Freyeria)獨立出來，則綺灰蝶屬為單系群。儘管綺灰蝶屬仍包括利用裸子及被子植物的種類，但是利用蘇鐵為食的成員自成一群，顯示以蘇鐵為食的行為為單一的寄主轉移事件。此論文的結果架構出台灣地區蘇鐵綺灰蝶之擴散來源，同時指出外來園藝植物之引進使得蘇鐵綺灰蝶的角色轉變為園藝害蟲，並可能加害原生種台東蘇鐵的生存，建議長期監控園藝蘇鐵的貿易方能保護原生蘇鐵之生存。而根據分子親緣關係樹，我們將晶灰蝶屬自綺灰蝶屬中獨立出來，不僅指出綺灰蝶屬可能為非洲起源的類群，並且提供灰蝶親緣關係最複雜的類群之一 ― 藍灰蝶組 (Eliot’s Polyommatus section)重建其屬級分類的可能性。. 關鍵字關鍵字：族群擴張、分子親緣關係、園藝貿易、寄主轉移、裸子植物、藍灰蝶族、豆科、蘇鐵屬、綺灰蝶屬、晶灰蝶屬、藍灰蝶屬、藍灰蝶組、蘇鐵綺灰蝶。. 4.

(5) Abstract Foreign plants are usually introduced for food or aesthetic reasons. Most of these plants are non-invasive, but can alter the evolutionary trajectory of the associated native insects or inadvertently spread potential pests. A hitherto poorly documented example is the rapid expansion of Chilades pandava, a Cycas-feeding butterfly. Since about 1990, large numbers of the Sago Palm Cycas revoluta were introduced into Taiwan. Invading or introduced with this hostplant, Ch. pandava has rapidly spread to all major parts of Taiwan and to other places worldwide. In order to trace the source of outbreaks, I set this issue as the first part of this dissertation (Part I). In part II, the members in the genus Chilades include some of the smallest butterflies in the world. As for some other small butterflies classified in the same lycaenid tribe, Polyommatini, they are widely distributed, ranging from West Africa to East Asia. Larval hostplant associations are unusually wide for a single butterfly genus, including both gymnosperms and angiosperms, questioning the monophyly of Chilades butterflies and the hostplant associations among this genus butterflies. In part I, total 810 specimens were sampled covering 50 Taiwanese localities and other regions using mitochondrial COII sequences. Only 29 haplotypes were found, however, the haplotype C which dominates outbreak populations from western Taiwan was endemic to the island. This is consistent with the hypothesis of a local range expansion of Ch. pandava, rather than an introduction. In addition, the Taiwanese Central Mountain Ridge may constitute a primary biogeographic barrier 5.

(6) restricting gene flow between eastern and western populations. In part II, to reconstruct the relationships of the genus Chilades and map patterns of hostplant use onto this inferred tree, mitochondrial COI, COII and nuclear EF-1α sequences (3437 bp, total 82 taxa) were used. The topologies show that Chilades is polyphyletic containing two separated clades, and that gymnosperm specialists (feeding on cycads) are monophyletic and represent a single host shift from angiosperms (Fabaceae). The study of tracing the source of outbreaks populations not only flags an important new invasive insect that needs to be monitored and controlled within the horticultural trade and for in situ cycad conservation, but also provides a clearly documented case of the transformation of a native tropical butterfly into a pest via introduced horticultural plants. In the aspect of genus status, once butterflies belonging to the genus Freyeria (which is usually treated as synonym of Chilades, but is here resurrected, including Freyeria putli, F. minuscula, F. trochylus, and F. yunnanensis comb. nov.) are excluded, the genus Chilades becomes monophyletic, but still comprising both angiosperm and gymnosperm (cycad) feeders. In this part, I infer the correct phylogenetic placement of both Chilades and Freyeria, African origins for Chilades, and also an aspect to reconstructing the genera relationships among Eliot’s Polyommatus section. Keywords: Population outbreak; range expansion; molecular phylogeny; horticultural trade; host shift; gymnosperms; Polyommatini; Fabaceae; Cycadaceae; Cycas; Chilades; Freyeria; Polyommatus section; Chilades pandava. 6.

(7) Introduction Specialist cycad-feeding is rare among Lepidoptera. The Cycad Blue, Chilades pandava (Horsfield, [1829]), like some other congeneric species and American Eumaeus butterflies, has an obligate association with cycads (Rothschild et al., 1986; Schneider et al., 2002). Ch. pandava is a multivoltine lycaenid, taking only 20-30 days from egg to adult butterfly (Lee, 1989). Although the caterpillar of Ch. pandava can feed on the young leaves or soft tissue of Cycas plants (Chang, 1989; Corbet and Pendlebury, 1992; Bascombe et al., 1999; Parsons, 1999), the buds of a single Cycas plant are usually available only for a limited duration (Lan, 1999). Therefore, when the young leaves or soft tissue of Cycas plant become too tough for a caterpillar of Ch. pandava to utilize, the adult Cycad Blue tends to disperse to find new resources for the next generation, as in many other tropical butterflies which have ephemeral hostplants (Ehrlich, 1984). The natural distribution of Ch. pandava covers mainly tropical regions west of the Wallace Line, including Taiwan, southern mainland China, Southeast Asia, India, and Sri Lanka below 1000 meters (Hsu, 2002; Igarashi and Fukuda, 2000). There are four currently recognized subspecies, based upon geographic isolation and genitalic/androconial morphologies (Evans, 1932; Hsu, 1989; Shirôzu and Ueda, 1992). Taiwanese populations are currently considered to represent an endemic subspecies Ch. pandava peripatria, which is morphologically different from the populations in mainland China and Southeast Asia (Ch. p. pandava) and the Philippines (Ch. p. vapanda) (Hsu, 1989), and associated specifically with Cy. taitungensis in southeast 7.

(8) part of Taiwan (Hsu, 1987). The fourth subspecies, Ch. p. lanka, is restricted to the island of Sri Lanka. The genus Chilades Moore, [1881] is a member of the Eliot’s Polyommatus section classified in the tribe Polyommatini (Eliot, 1973). Bálint and Johnson (1997, 1999) revised this section and reclassified 33 genera into nine ones and excluded seven neotropical genera from this section. In a new concept of the Polyommatus section (which is referred as the Polyommatus genus-group), these authors consider some genera such as Edales, Luthrodes, Lachides, and Freyeria should be treated as synonyms of the genus Chilades. This means that the genus Chilades currently possesses about 20 species ranged from West Africa to East Asia (Table 1). Three Chilades species-groups were proposed according to their genitalic characters and distribution. The members of the cleotas-species group are all mainly distributed in Southeast Asia, and this group so far is known only to utilize Cycas plants as their larval hostplants (Table 2). The lycaenids belonging to the galba-species group are distributed mainly in Central Asia and their larvae feed on Fabaceae. However, the members of the lajus-species group are more widely distributed. Most of them occur in Africa, even C. minuscula is endemic to Madagascar, but three members (C. parrhasius, C. putli, C. trochylus) and the type species of the genus, C. lajus, are more widely distributed to SE Europe (Turkey), Central Asia, India, and Southeast Asia. In addition, species in the lajus-species group have more diverse hostplant use than the other two groups, feeding on Rutaceae, Fabaceae, and Boraginaceae. In this dissertation, two aspects of the cycad-feeding butterflies were discussed. Firstly, a hitherto poorly documented example is the rapid 8.

(9) expansion of Chilades pandava, a Cycas-feeding butterfly (Part I). Since about 1990, large numbers of the Sago Palm Cycas revoluta were introduced into Taiwan. Invading or introduced with this hostplant, Ch. pandava has rapidly spread to all major parts of Taiwan and to other places worldwide. Therefore, I tried to trace the source of outbreaks and to find out how the outbreaks happen (Part I). On the other hand (Part II), the members in the genus Chilades include some of the smallest butterflies in the world. As for some other small butterflies classified in the same lycaenid tribe, Polyommatini, they are widely distributed, ranging from West Africa to East Asia. Larval hostplant associations are unusually wide for a single butterfly genus, including both gymnosperms and angiosperms, questioning the monophyly of Chilades butterflies. I plan to reconstruct the relationships of Chilades butterflies and the associations with their hostplants.. 9.

(10) Part I. Elucidating origins of the Cycad Blue. I-1. Introduction The biodiversity of phytophagous insects may be inadvertently influenced by accidental, human-mediated activities (Chapin et al., 2000; Keane and Crawley, 2002; Tallamy, 2004; Strauss et al., 2006). The population size of some native insects may decrease when their native hostplants are replaced by aliens, but natives can also sometimes thrive and expand their ranges by exploiting these novel hosts (Stastny et al., 2006; Strauss et al., 2006; Carroll, 2007). An extreme example has been given for the native Californian butterfly fauna which has expanded on to alien plants in urban or suburban areas and some 40% of which now entirely depend on aliens (Shapiro, 2002). Invasive species can alter the evolutionary pathway of native species (Mooney and Cleland, 2001). However, when introduced as agricultural or ornamental plants, most alien plants are not naturally invasive, but in large quantity, they affect native biotas by direct replacement (Reichard and White, 2001). The unchecked spread of popular horticultural plants around the globe concomitant with modern globalization may be having a dramatic effect on their relatives. Oliver (2006) hypothesized that the native phytophagous insects which can utilize expanded alien plants should represent two clear genetic signatures: firstly, insects expanding their geographic range should exhibit lower genetic diversity in newer portions 10.

(11) of the insects’ range than older ones. Secondly, levels of gene flow should increase among isolated populations when these have been connected, due to anthropogenically mediated range expansions of their hostplants. The invasive insects will thus exhibit relatively low population genetic differentiation, evidenced by reduced local adaptation and hybridization among formerly isolated lineages (Pannell and Dorken, 2006; Vellend et al., 2007). Cycads are one of the most popular cultivated gymnosperm groups, with ancient origins (Hu et al., 1999; Treutlein and Wink, 2002). These plants have a worldwide scattered distribution and occur in many kinds of habitats, such as in the understory of tropical rain forest and seasonally dry forest, grasslands, and at high elevation habitats in eastern Africa (Norstog and Nicholls, 1997; Whitelock, 2002). Cycads comprise 11 genera and about 305 species (Hill et al., 2004). One of the largest genera, Cycas (about 90 species), is mainly distributed in mainland China, Southeast Asia, India, Australia, and South Africa (Jones, 1993). Cycas plants have been used as herb medicines, food, and ornaments for thousands of years (Whitelock, 2002). However, cycads are now endangered as a consequence of extensive collections from the wild and of decreasing natural habitats through urbanization (Donaldson, 2003). Consequently, all the extant cycads are listed under Appendix I or II of Convention on International Trade of Endangered Species of Wild Fauna and Flora (CITES). Two Cycas species are now commonly found in Taiwan: the endemic Cy. taitungensis and the introduced “Sago Palm” Cy. revoluta (Liu et al., 1994; Shen et al., 1994). The distribution of the native cycad is restricted to southeastern Taiwan where it is considered a 11.

(12) glacial relict (Huang et al., 2001). On the contrary, Sago Palm is planted ubiquitously, especially in urban and suburban parts of western Taiwan where native Cycas species do not occur. Despite the fact that the Cycad Blue has only recently been reported from Taiwan, the nature of its origin is controversial. The first report of the Cycad Blue dates back to 1976 (Hsu, 1987). In the following years, this butterfly was only occasionally found in urban or suburban areas (e. g. Konishi, 1987; Chang, 1989; Lee, 1989). From that time, this butterfly was rarely reported up to 1990, but the population irrupted suddenly throughout the whole island by 2000 (Lan, 1999; Wu et al., unpublished data). Some have considered this butterfly as an introduced species in Taiwan, because no record was registered in previous exhaustive surveys (e.g. Shirôzu, 1960; Hamano, 1987). However, this butterfly was later considered to be native in Taiwan after large numbers of individuals were found to coexist with Cy. taitungensis in a cycad nature reserve (Hsu, 1987; Hsu, 1989). Moreover, three previously undocumented specimens of Ch. pandava were found in the collection of Jinhaku Sonan, a pioneer of butterfly studies in Taiwan, bearing labels that indicates they were collected in Taitung, southeast part of Taiwan, in 1937 (Wu et al., unpublished data). These specimens lend support to the idea that Ch. pandava should be treated as a species native to Taiwan, rather than of exotic origin. In other regions, the Cycad Blue has become an invasive pest after Sago Palm or other Cycas species were introduced as horticultural plants. Even the recent occurrences of the Cycad Blue on Guam and Madagascar are cases of severe defoliation events on native cycads that were never 12.

(13) observed in the past (Moore, 2008; DCL per. com.). Establishing the origins of these introduced populations is urgent to help put in place control measures that could prevent invasions in parts of the world with native cycad populations, and where ornamentals are being increasing introduced. Population outbreaks of the Cycad Blue in Taiwan could originate from alien or native sources. In this study, two alternative hypotheses for how population outbreaks developed in Taiwan are proposed (Figure 1). Firstly, if the Cycad Blues that feed on Sago Palms are derived from native populations, outbreaks may be caused by range expansion of the native populations induced through widespread horticultural planting of Sago Palm (as Hypothesis 1 in Figure 1). In this case, as in the typical genetic signature of range expansion proposed by Oliver (2006), I would expect genetic diversities of native Taiwanese Cycad Blue populations (No. 9 in Figure 2) to be higher than those of the local outbreak populations (No. 1-8). Additionally, Taiwanese populations should be more closely related to each other than to those of other regions or be monophyletic. In other words, significant genetic structure or high genetic differentiation would be found between Taiwanese and non-Taiwanese populations. Alternatively, individual larvae and/or adults of Cycad Blue may have been directly transported with the Sago Palms to locations where they established outbreak populations (as Hypothesis 2 in Figure 1). Under this scenario, Taiwanese outbreak populations could also exhibit low genetic variation due to founder effects or alternatively high genetic variation due to multiple colonization events (e.g. the case study in Darling et al., 2008), but the haplotype composition of outbreak 13.

(14) populations would be different than that of Taiwanese native populations. In addition, the introduced outbreak populations should be more closely related to source regions than to the native population in Taiwan. That is, recent colonization would either obscure the genetic structuring between Taiwanese and other regions, or the level of genetic differentiation would be low. These two scenarios are not exclusive. Outbreak populations could result from range expansion of both native and introduced populations, with the Cycad Blue exhibiting a genetic signature combining features of both alternative hypotheses (M1 & 2 in Figure 1). In this study, the genetic signatures of Taiwanese populations of the Cycad Blue were examined to shed light on how a rare butterfly species can be rapidly transformed into a pest, both locally and in other parts of the world. Mitochondrial COII sequences were employed in this study using standard phylogeographic methodologies to test the above hypotheses and also to address the following specific questions: (1) Does the genetic structure among the three principal subspecies regions support the current subspecies status within Ch. pandava? (2) Do the outbreak populations of the Cycad Blue in Taiwan derive from native or recently introduced populations? What are the possible geographic sources of introduced populations of the Cycad Blue in various parts of the world? (3) Is the pattern of haplotype distribution influenced by the population dynamics of this tropical butterfly? (4) Have geographic barriers such as straits and mountains affected the genetic structure of this lycaenid?. 14.

(15) I-2. Materials and methods Specimens and sampling locations 810 individuals of Chilades pandava representing 50 locations from Taiwan and its neighbors were used in the study (Figure 2; Table 3). Nine locations were sampled in Taiwan, including six populations west of the Central Mountain Ridge (CMR) (No. 1-6, Taipei, Xinzhu, Taizhong, Jiayi, Kaohsiung and Yilan) and three populations east of the CMR (No. 7-9, Hualian, Guanshan, and Luye). The CMR is reported as a biogeographic boundary for other Taiwanese biota (e. g. Cheng et al., 2005; Chen et al., 2006; Tzeng et al., 2006). Among these sites, Luye (No. 9) is the first and only locality where Ch. pandava was found feeding on the native cycad, Cy. taitungensis (Hsu, 1989). Because tropical organisms often have high dispersal ability as needed to search for temporally scattered resources and related large fluctuations in abundance (Ehrlich, 1984), they should exhibit a mixed genetic composition, especially when local assembly takes place from different sources over a protracted time period. For the purposes of observing the effect of temporal variation on their genetic pattern, butterfly specimens were sampled each month from February, 2000 to October, 2002 (Table 4). Two subspecies neighboring Taiwan were also sampled to investigate their relationship with the endemic Ch. p. peripatria. Samples collected from regions west of Taiwan Strait belong to Ch. p. pandava. This subspecies has the largest distributional range, extending from mainland China to Southeast Asia and India (Igarashi and Fukuda, 2000; Hsu, 2002). The specimens were collected mainly in mainland China and 15.

(16) Southeast Asia (Figure 2, locations No.16-45), regions expected to be outbreak sources for the Cycad Blue because of their large export trade in Cycas plants. Another subspecies sampled in this study is Ch. p. vapanda, which. is. endemic. to. the. Philippines. archipelago. (No.46-47).. Additionally, two evidently recently introduced populations, namely Guam (No. 50) and Madagascar (No. 48), and seven recently colonized populations of Ch. pandava, namely Korea (No. 49; Takeuchi, 2006), Japan (No. 13-15) and the islands near Taiwan (No. 10-12) were sampled to broaden the search for the origins of these recent outbreak populations. Cycad Blue occurrences from these islands near Taiwan are firstly reported in this paper, which have no native cycads but do have some introduced Cy. revoluta. In addition, three congeneric species, Ch. cleotas (Malekula Island, Vanuatu), Ch. laius (Kaohsiung, Taiwan), and Ch. mindora (Dinagat Island, Philippines) were used as the outgroups for phylogenetic analyses. All the specimens were stored at –80℃, except for a small number of museum specimens.. DNA extraction and sequencing Genomic DNA was obtained from the thoracic muscle tissue or legs using the Purgene DNA Isolation kit (Gentra Systems, Minnesota, USA), following the extraction protocol of manufacturer. Precipitated DNAs were resuspended in 100µl of dH2O and we used the primer pairs: Pierre (5’-AGAGC CTCTC CTTTA ATAGA ACA-3’) and Eva (5’-GAGAC CATTA CTTGC TTTCA GTCAT CT-3’) to amplify the partial mitochondrial cytochrome oxidase II gene (COII) by polymerase chain 16.

(17) reaction (PCR) (Caterino and Sperling, 1999). Each PCR reaction was carried out in a final volume of 25µL with 0.8µL of 10µM dNTP, 1.5µL of 25mM MgCl2, 0.5µL of each 10µM primers, 2.5µL of 10X Taq buffer, 0.1µL of Amersham Taq (Amersham Biosciences, Buckinghamshire, UK), and finally I added dH20 up to 25µL. PCR was carried out as the following three steps: an initial denaturation step of 94℃ (2min), followed by 35 cycles consisting of denaturation at 94℃ (30 sec), annealing at 55℃ (30 sec), extension at 72℃ (1min), and a final extension step of 72℃ (7min). Different annealing temperatures (50-58℃) were used to improve PCR quality when the above PCR conditions failed. Some museum specimens resistant to amplification via the above primer pairs. could. be. amplified. through. using. the. internal. primers. Cppcox-J-3300 (5’-ATAWG AATCA AATTC AATRT TT-3’) and Cppcox-N-3400 (5’-TTATT GCWTT ACCTT CWTTA CG-3’). Finally, the products were run on 1.0% agarose gels in 1X TBE buffer to ensure that the lengths of PCR fragments were correctly amplified. Blank controls were also run each time to check that no contamination occurred during the PCR process. PCR products were cleaned using Gel/PCR DNA Fragments Extraction kit (Geneaid, Taipei, Taiwan) when only a single DNA band was visible in a gel. DNA sequence reactions were conducted using a 96 well Gel/PCR Clean Up kit (Geneaid) on an ABI3730 DNA Analyzer (Applied Biosystems). Both directions were sequenced. Finally, the sequences were checked and assembled into contiguous arrays using Sequencher 4.5 (GeneCode, Boston, USA).. Analyses Neutrality of the mtDNA COII gene sequence data was tested in case that some selection was hidden (Otto, 2000). A combination of Tajima’s 17.

(18) D (Tajima, 1989), Fu & Li’s D* (Fu and Li, 1993) and Fay and Wu’s H (Fay and Wu, 2000) was used, and statistical significance for these neutral tests was assessed by coalescent simulations with 10,000 replicates performed by using DNASP 4.10 (Rozas et al., 2003). General population genetics, such as the number of unique haplotypes, variable nucleotide positions and measured genetic diversities including nucleotide diversity (π) and haplotype diversity (h) were also described by using DNASP 4.10. If Taiwanese outbreaks were derived from populations outside the island, genetic differentiation between subspecies would be obscured. Therefore, an exact test was employed to check the null hypothesis of random distribution of haplotypes, using the software Arlequin 3.1 (Excoffier et al., 2005). To evaluate evidence for genetic structure among different geographic regions, the pairwise FST between populations were evaluated as implemented in the program DNASP 4.10. The analysis of molecular variance (AMOVA) was also performed to partition total variance into variance components attributable to inter-individual, and/or inter-population differences (Excoffier et al., 1992). Three different levels of hierarchical components were included in this analysis (called Φ-statistics): ΦCT, the degree of differentiation among all regions; ΦSC, the degree of differentiation among local populations within regions, and ΦST, the degree of differentiation among all local populations. In this analysis, datasets A-C were grouped as below. In dataset A (the taxonomic subspecies dataset), the whole sampled populations were grouped into three groups according to their subspecies status, including Ch. p. peripatria (Nos. 1-9), Ch. p. pandava (Nos. 16-45), and Ch. p. 18.

(19) vapanda (Nos. 46-47). In dataset B (the geographical “disjunct regions” dataset), distinct regions were defined as Taiwan (Nos. 1-9), eastern mainland China (Nos. 16-22), Hainan (Nos. 23-24), central mainland China (Nos. 25, and 30), western mainland China (Nos. 26-29), Central Indochina (Nos. 31-33), north of Indochina Peninsula (Nos. 34-35), western Malaysia (Nos. 36-39), Borneo (Nos. 40-41), Java (Nos. 42-45) and Philippine region (Nos. 46-47) based on the geographical separation by mountains, straits or long distance. In dataset C (the geographical “Taiwanese populations” dataset), the grouping was similar to dataset B, except that the Taiwan region was divided into the eastern (Nos. 1-5) and western populations (Nos. 6-9) based on the barrier of the Taiwanese CMR. These three datasets were compared to each other to detect changes in the above three differentiation parameters. The significance of the components was computed using a nonparametric permutation test (10,000 permutations) as performed by Arlequin 3.1. Some new outbreak populations of the Cycad Blue (No. 10-15, 48-50) were excluded from the AMOVA because these locations were not included in the native distribution of Ch. pandava and it was not possible to define their group membership. In order to infer the most basal haplotype of Ch. pandava and the phylogenetic origins of this species in Taiwan, all haplotypes (29 ingroups; 3 outgroup; 621bp) were aligned and analyzed by maximum parsimony (MP), maximum likelihood (ML), and Bayesian analysis. Although the phylogenies based on MP, ML and Bayesian methods all showed high branch support recovering Ch. pandava in the same monophyletic group (MP: tree length=99, CI=0.869, RI=0.797, bootstrap 19.

(20) value=100; ML: HKY85 model, -ln=-1357.68, bootstrap value=85; Bayesian. inference:. GTR+I+G. model,. -ln=-1407.50,. posterior. probabilities=0.85), the relationships among the haplotypes or three subspecies of Ch. pandava were unresolved (tree not shown). Because conspecific populations often have lower divergences than at interspecific level, I also performed haplotype network joining for studying closed relationships (Posada and Crandall, 2001). Haplotype networks were constructed with the software TCS 1.21 based on the principle of parsimony (Clement et al., 2000). Each branch in the network was supported with a 0.95 probability (over 0.95). This setting provided plausibility for the uncertainty of the exact cladogram when only a part of dataset was used (Templeton et al., 1992).. I-3. Results Sequence information A total of 810 specimens of Ch. pandava and 3 congeneric species were sequenced for 621 base pairs (bp) of the partial COII gene (GenBank Accession nunbers FJ941955-FJ942767). All sequences could be translated into amino acids. No stop codon was found but one specimen from Mindoro (Figure 2; No. 43) showed a 3-bp indel. In total 99 polymorphic sites were detected, most due to outgroup sequences, but 38 among populations of Ch. pandava. The neutral tests, Tajima’s D test (average value of D= -0.065; P value=0.83, NS.), Fu and Li’s test (D* test: average value of D=-0.025; P value=0.79, NS.), and Fay and Wu’s H test 20.

(21) (average value of H=0.015; P value=0.60, NS.), all showed that the accumulated mutations of the dataset were not seriously affected by positive selection, and the dataset was suitable for population analyses without bias from positive selection.. Genetic diversities Gene diversities of each location were listed on Table 5. Overall haplotype and nucleotide diversities of Ch. pandava were 0.791 and 0.00446, respectively. Although the haplotype diversities of Taiwanese populations (h=0.1-0.5) appear lower than those from non-Taiwanese regions (h=0.2-0.8), the statistical analysis between these two is not significant (t=0.989, P=0.329, df =37), even though nucleotide diversities were not different between Taiwan and other locations (π=0-0.003; t=0.78, P=0.440, df =37). However, populations sampled from the eastern part of Taiwan (h=0.1-0.5) had higher haplotype diversity than those from the western part (h=0-0.1; t=2.869, P=0.024, df =7). The populations from western part of CMR (No. 1-6) all showed low gene diversities, even the locations of Taizhong and Yilan became fixed over three years of observation (Table 4).. Haplotype distribution Twenty-nine COII haplotypes of Ch. pandava were obtained from 810 specimens (Table 5). This number of haplotypes is less than in some other widely distributed species, such as Lampides boeticus (29 haplotypes among only 57 specimens: Lohman et al., 2008) and Zizina maha (27 haplotypes among 121 specimens, Yago et al., 2008), but most of these 21.

(22) haplotypes of Ch. pandava were localized. Haplotype A-D and F were only found in Taiwan. Other major haplotypes were detected in mainland China or Southeast Asia, except haplotype M and N, which were only found in the Philippine archipelago. Most locations were comprised one or two haplotypes, but some locations such as Luye, Guanshan, Nanning, and Fuzhou, possessed at least four haplotypes. In Taiwan, the haplotype distribution was asymmetrical. Haplotype A and B were mainly found in the eastern part of CMR, but haplotype C was dominant in the western part. Moreover, other rare haplotypes such as F and D were only detected at local locations, Jiayi and Luye respectively. Only haplotype E showed a disjunct distribution. This haplotype was found both in Taiwan (No. 1, 2 & 8) and Malaysia (No. 34). Most localities which Ch. pandava recently colonized possessed single haplotype. The specimens from Ludo Is. (Figure 2, No. 10), and Yonaguni Is. (No. 13) both possessed haplotype A, previously detected in eastern Taiwan (No.7-9). Specimens from Jeju Is. (No. 49) and Orchid Is. (No. 11) possessed haplotype C, detected both in eastern and western Taiwan. Specimens from Guam (No. 50) possessed only haplotype M, which also occurs in Luzon (No. 46, Quezon Province) in the Philippine archipelago. The specimens from Madagascar (No. 48) possessed haplotype O, found mainly in southern part mainland China or Southeast Asia. Japanese regions had single haplotypes in each region: haplotype A was detected on Yonaguni Is. (No. 13), whilst haplotype H was found on Honshū Is. (Osaka, No. 14 and Takarazuka, No. 15). Among all recently colonized populations, only Pengjia Islet (No. 12) had two haplotypes, haplotype O and H, both detected dominantly in mainland China and 22.

(23) Southeast Asia.. Haplotype network The 29 haplotypes of Ch. pandava used to construct a haplotype network (Figure 3). A star-like network indicates rapid population expansion. Most unique haplotypes (colored in Figure 3) were represented on tip clades, meanwhile common haplotypes were represented in interiorly nested clades (e. g. haplotype O and H). Haplotypes belonging to the subspecies Ch. p. pandava and Ch. p. peripatria were connected together by a single step mutation, while the haplotype of Ch. p. vapanda exhibits a long branch connection with Ch. p. pandava (9 steps). In Taiwan, the haplotypes were connected together, and the first Taiwanese haplotype connected to outside haplotypes was haplotype B, only found in eastern part of Taiwan. Curiously, the dominant haplotype of the western part (Haplotype C) was a tip clade connected to haplotype A instead of to other haplotypes from mainland China and Southeast Asia or the Philippines. This relationship among haplotypes supports the hypothesis that western populations may derive from eastern populations (Hypothesis 1 of Figure 1).. Population differentiation and genetic structure of Ch. pandava The exact test performed in Arlequin 3.1 software showed that haplotypes of Ch. pandava were not randomly distributed (P<0.0001 with 10,000 steps in Markov chain). Therefore, population differentiation of the Cycad Blue should exist. High genetic differentiation of Cycad Blue populations was detected through performing a pairwise FST (Table 6). 23.

(24) For example, in the populations of Ch. p. pandava, the proportion of FST values higher than 0.8 was 46.5%, in contrast to the 15.8% of FST values that were lower than 0.3, reflecting low gene flow in this widely distributed subspecies (Table 6A). High FST values were also found in Taiwan when western part populations (No. 1-6) were compared to the eastern part (Table 6B, No. 7-9). In other words, FST values calculated through comparing eastern and western populations were higher than the pairwise FST values obtained within eastern populations alone (t=2.07, P=0.05, df =19) or within western populations alone (t=28.716, P<0.0001, df =28). Although the FST values of Taiwan native populations (Table 6B, No. 9) compared with Taiwanese outbreak populations (No. 1-8) were not significantly different compared to non-Taiwanese regions (G1-10; t=1.726, P=0.10, df =16, Table 6B), the pairwise FST values obtained from comparing within Taiwanese populations (No.1-9) were significant lower than the pairwise FST values obtained from comparing Taiwanese and non-Taiwanese populations (No.1-9 vs. G1-10 in Table 6B; t=9.179, P<0.0001, df =124). This significant difference indicates that Taiwanese outbreak populations were not derived from non-Taiwanese populations, and that the native Taiwanese population is the only possible source of Taiwanese outbreak populations. In general, significant genetic structure of Ch. pandava was also observed at various hierarchical levels by AMOVA (Table 7). Those significant differences also support hypothesis 1 that Taiwanese populations exhibit genetic structure in comparison with non-Taiwanese populations. In the three subspecies dataset, the variation among groups accounted for most of the variance (70.29%). When the sampled 24.

(25) populations were grouped according to their geographical distribution, the percentage of the variation among groups increased to 72.47%. Moreover, when Taiwanese populations were divided into eastern and western populations by the CMR, the percentage of the variation among groups increased to 78.11%.. I-4. Discussion Molecular systematics and phylogeography of Cycad Blue Mitochondrial DNA sequences often represent little variation in widely distributed species. However, in two other recently investigated cases of polyommatine lycaenids feeding on “weedy” species, mtDNA provides good resolution on genetic structure and systematic validity. The species Lampides boeticus, one of the most widely distributed Old World butterflies, has never been considered to geographical differentiation (Lohman 2008). However, phylogeographic analyses show this butterfly forms three distinct groups based on mitochondrial COI and cytB genes. Yago et al. (2008) address taxonomic problems in the genus Zizina, small butterflies often difficult to identify by wing pattern. The mitochondrial ND5 gene in combination with male genitalic morphology allows reliable identification of Zizina taxa. As in the above widely distributed Asian butterflies, most sampled populations of Ch. pandava also showed little genetic variation (only 29 haplotypes were found in 810 specimens). However, in this case genetic data supports the validity of existing subspecies. Ch. p. pandava is the most widely distributed subspecies. 25.

(26) This subspecies could in fact represent the source population where it feeds on several native Cycas populations that are distributed in south of mainland China and Indochina (Jones, 1993). Ch. p. pandava exhibits one to four step mutations to Ch. p. peripatria endemic in Taiwan, and nine step mutations to Ch. p. vapanda endemic in the Philippines. Based on the mitochondrial COII gene, Ch. p. pandava also possesses 21 haplotypes, considerably more than Ch. p. peripatria (five haplotypes) and Ch. p. vapanda (two haplotypes). Although Ch. p. lanka was not surveyed in this study, this subspecies may be more closely related to Ch. p. pandava than to Ch. p. peripatria or to Ch. p. vapanda, based on its geographic proximity. These molecular data in combination with significant morphological differences (Hsu, 1989) both supports the subspecies validity of Ch. p. peripatria. Low genetic variation and the star-like haplotype network of the Cycad Blue (Figure 3) is consistent with a population bottleneck after rapid range expansion, as also in the example of the highly invasive Horse Chestnut leaf miner moth Cameraria ohridella in Europe (Valade et al., 2009). At the same time, Cycad Blue hostplants have suffered severe reduction through habitat destruction and collecting for the horticultural trade and for subsequent planting in urban and suburban areas (Donaldson, 2003). On the one hand, the gene flow of Cycad Blue may nevertheless have been limited due to CITES restrictions forbidding transport of wild cycads without permits, in accord with COII gene data that show a localized, highly endemic distribution of haplotypes (Figure 2; Table 5). On the other hand, at present, the planted range of wild cycads is much vaster than native range in Asia, representing many opportunities 26.

(27) for the Cycad Blue to increase its population size and range. A strongly analogous situation is found in another cycad feeding lycaenid that was formerly of conservation concern, the local race of the Atala Hairstreak butterfly (Eumaeus atala) in southeastern USA. This butterfly, represent in the Caribbean, became extinct between 1937 and 1959 in Florida (Landolt, 1984). However, after initial reestablishment in greenhouses, the Atala Hairstreak is now commonly found in southeast Florida feeding on the genera Zamia and Cycas wherever they are planted horticulturally (Hall and Butler, 1995).. The origin of Taiwanese populations The results support the hypothesis that Taiwanese outbreak populations, especially in the western part of CMR, were mostly caused by range expansion of the native population. The hypothesis that outbreak populations were delivered by means of direct introduction along with alien cycads was rejected because the major haplotypes (haplotypes A, B, and C) represented the dominant populations which were only found in Taiwan. Western populations (haplotype C) showed a different dominant haplotype from eastern populations (haplotype A, B), supporting a hypothesis of longer coexistence of alien and native populations in Taiwan. However, the haplotype network suggests that haplotype C was derived from haplotype A, occurring only in eastern part of Taiwan (Figure 3). In this case, the outbreak populations of western populations in Taiwan would have been maintained entirely by horticultural cycads. The Cycad Blue has thus expanded its range through a single rather than through multiple colonization events, to become widespread around the 27.

(28) whole island. There is no evidence that the striking biogeographic difference in haplotypes divided by the CMR (C compared with A and B) is related to a now extinct population of Cycas (such as Cy. taitungensis) native to western Taiwan, i.e. that the pattern is explained by divergence in allopatry. Taiwanese herbarium records also show that few horticultural cycads were planted in Taipei and southwestern Taiwan before 1950 (plant records at website: http://taif.tfri.gov.tw/taif_en/), and there are no records of Ch. pandava prior to 1976 in Taiwan (Hsu, 1987). Before recent anthropogenically induced outbreaks, Taiwanese populations must have been founded from neighboring regions, considering also that Taiwan is a relatively young island, formed c. 9 Ma (Sibuet and Hsu, 2004). Hsu (1987) has pointed out that the source of the Taiwanese population is likely to have been either mainland China or the Philippines archipelago. According to the haplotype joining network, Taiwanese populations are more closely related to populations from mainland China than those in the Philippine archipelagos which lack a direct connection in the network. Moreover, populations in Taiwan could indeed represent a relatively old colonization. Multiple lines of evidence suggest that southeastern Taiwan constituted a Pleistocene refuge (see examples in Cheng et al., 2005 and Lee et al., 2006). The native hostplant, Cy. taitungensis has high genetic variance indicating a large population during interglacial stages (Huang et al., 2001), and thus sufficient resources for local survival of Ch. pandava over this time. Although Taiwanese populations show only one to four-step mutations from other regions (Figure 3), the significant genetic structure among the three subspecies indicates a long period of isolation because none of Taiwanese 28.

(29) endemic haplotypes (A-D, and F) was found in other native regions of the Cycad Blue.. Origins of the introduced populations in other geographical regions Many native species have been threatened or even extinguished when introduced species successfully establish populations in their native habitats. Therefore, an understanding of the origin, biology, and ecology of alien species could help to focus conservation efforts for native species. So far, Chilades pandava has already been introduced to many parts of the Old World as far apart as Korea (Takeuchi, 2006), Japan (Mitsuhashi, 1992; Takegami, 2001; Hirai, 2009), Hong Kong in 1978 (Bascombe et al., 1999), Pacific islands including Guam in 2005 (Calonje, 2007; Moore, 2008), the neighboring island of Rota in 1996 (Moore, 2008), Saipan in 1996 (Schreiner and Nafus, 1997; Moore et al., 2005) and in the western Indian Ocean in Réunion since 2000 (Martiré and Rochat, 2008; Guillermet, 2009), Mauritius since 2000 (Williams, 2006; Williams, 2007), Madagascar since 2006 (DCL, pers. comm.), and in Miami (SHY, pers. comm.). In some places, the native Cycas plants, Cy. micronesica (Guam) and Cy. thouarsii (Madagascar), are threatened by the Cycad Blue (Table 8). The molecular data could provide enough information to quickly ascertain the origin of the haplotypes of different subspecies or populations, and thus provide a phytosanitary monitoring tool. For example, the most likely origin of the Guam populations is from the Philippine archipelago and that of the Korean populations, from Taiwan because the haplotypes found were shared with these regions (Figure 2 or Table 5). Tracing the source of Cycad Blue populations is more important 29.

(30) in Japan because haplotypes characteristic of two subspecies were detected, Ch. p. peripatria found in Okinawa (No. 13) and Ch. p. pandava found in Honshū (No. 14-15). Besides, the increasing frequency and wide, rapid range expansion of Ch. pandava in Japan (Mitsuhashi, 1992; Hirai, 2009) increases the urgency to protect the last native region of Cy. revoluta, in the Ryukyu Islands of southwestern Japan (Wang et al., 1996). Nevertheless, COII sequence data was unable to trace the origin of Madagascan populations because haplotype O was found in many parts of the present range of Ch. p. pandava. Therefore, to improve the identification of the origin of introduced populations, more native populations should be surveyed and more sensitive genetic methods such as microsatellites (e. g. Habel et al., 2008) should be developed to more finely discriminate the origins of Ch. pandava outbreaks.. The role of the CMR on genetic structure of Ch. pandava The Central Mountain Range (CMR) of Taiwan at over 3000 meters elevation provides a primary north-south barrier considered to divide native populations of many species (examples in Peng, 2006; Wang et al., 2007). Such a dominant biographic barrier that clearly structures populations of many species is exceptional for such a small island. The maximal elevation recorded for Ch. pandava so far is about 700m (Hsu, 1989). As expected, CMR also serves as an effective barrier to divide the eastern and western populations of this Cycad Blue. The scattered nature of larval hostplants is commonly a significant factor in the population structure of tropical insect herbivores (Ehrlich, 1984). However, haplotype distribution and population dynamics of this butterfly are 30.

(31) clearly influenced by the CMR (Table 4; Table 7). This barrier may also be reflected in differences in emergence times in eastern and western populations: Ch. pandava in Yilan (Figure 2, No. 6) emerges in September or later whilst populations in Hualian (No. 7) emerge in March (LWW, pers. obs.).. Indirect effects of cycad cultivation on native Cycas species Introduced plants have not only the potential to enlarge the distribution of native insects, but also to increase their biomass (Tallamy, 2004) and thus herbivore pressure on native plants. In Taiwan, the additional Cycad Blue food resource (Cy. revoluta) appears indeed to have augmented the population size of Ch. pandava, as apparent in increased levels of plant attack in southern monitoring sites (Lan, 1999; Wu et al., unpublished data). Although the monitoring and genetic data show that western Taiwanese populations seldom disperse to the eastern side, the extra food resource provided by the introduced Sago Palm that are planted abundantly in eastern Taiwan may still greatly increase the overall population size of the Cycad Blue, threatening the survival of the rare native Cy. taitungensis. Adding greatly to this threat, another harmful pest, the scale insect Aulacaspis yasumatsui, has been introduced to Taiwan with horticultural Sago Palm since 2000 (Germain and Hodges, 2007). This scale continuously sucks nutrients from the leaves, stem and primary root until the host dies (Weissling et al., 1999). It is reported that the cycad scale causes high cycad mortality in Guam (Moore et al., 2005) and in Florida (Howard et al., 1999). While the monitoring of native Cy. taitungensis initially showed no significant mortality from heavy attacks 31.

(32) by Ch. pandava (only 4 of 162 cycads observed died between 2000 to 2004, Wu et al. unpublished data), the increased level of herbivory over many years combined with the presence of the new cycad pest A. yasumatsui may well jeopardise the continued survival of populations of this endemic and already endangered cycad. This is supported by data on Cy. taitungensis in March 2009 (LWW, pers. obs. 2009): 23 of 158 cycads being monitored had died, while the other cycads were in poor condition under the combined attack.. 32.

(33) Part II Molecular phylogenetics of Chilades butterflies. II-1. Introduction Phytophagous insects and plants make up over half of macroscopic organisms, and the relations between phytophagous insects and their hostplants have been widely discussed. 70-80% of phytophagous insects are considered to be monophagous (feeding on plants within a single genus) or oligophagous (feeding on several plants belonging to different genera within one plant family). This phenomenon provides model relationships for studying herbivore-plant interactions (Chapman, 1982, Bernays and Chapman, 1994). Some of the best examples are butterflies, which at the generic level frequently feed on related hostplants (Ehrlich and Raven, 1964; Jermy, 1984; Fiedler, 1996; Boggs et al., 2003; Wahlberg et al., 2009). Major evolutionary associations between phytophagous insects and their hostplants have generally been conceived as scenarios of secondary host shifting, replacing the earlier paradigm of cospeciation based on co-phylogenetic approaches (Janz and Nylin, 1998; Janz et al., 2006). When related insects shift to use another hostplant, it is often based on shared plant secondary chemistry (Carter and Feeny, 1999; Wahlberg, 2001). For example, many Appias pierids, seek hostplants with mustard oil glucosinolates (Wheat et al., 2007). This has allowed their ancestors to colonize both Malpighiales and Brassicales, orders distantly related to each other (Braby and Trueman, 2006). Reconstructing phylogenetic relationships of a selected group is an 33.

(34) important step to investigate the mechanism of host shifting to a novel hostplant. Such investigations may sometimes even cast light on plant taxonomy (e.g. the close relationship of the Euphorbiaceae genera Omphalea and Endospermum: Lees and Smith, 1991; Wurdack and Davis, 2009). Most butterfly species have diversified on angiosperms (Janz and Nylin, 1998) on which they appear to have started their adaptive radiation back into the late Cretaceous (Wahlberg et al., 2009), and only a small number of butterflies are known to feed on gymnosperms (Ehrlich and Raven, 1964; Beccaloni et al., 2008). These butterflies listed in Table 9 include: Neophasia species (Pieridae) which feed on Pinaceae in North America to Mexico, the hesperiid Metardaris cotsinga on Pinus radiata in Bolivia and some amathusiine nymphalids, such as Faunis aerope, Taenaris onolaus, and T. butleri, feeding on Cycadaceae in Asia, and one species of normally monocot-feeding Opsiphanes which feeds on Cycas circinalis in Venezuela. Some Callophrys butterflies (Lycaenidae) utilize Pinaceae or Cupressaceae in North America. Other lycaenids utilize the family Cycadaceae such as Theclinesthes onycha (Lycaenidae) in Australia, Eumaeus species in Central America, and some members of Chilades in Asia. Considering the phylogenetic position of gymnosperm-feeding butterflies recognized so far (Janz and Nylin, 1998; Braby et al., 2006; Wahlberg et al., 2009), most of them appear to represent single host colonizations. from. their. distant. angiosperm-feeding. relatives.. Gymnosperm feeding occurs as recent events in butterflies, whereas it is probably. an. ancient. association 34. in. certain. groups. such. as.

(35) Araucaria-feeding beetles (Farrell et al., 1998; Sequeira and Farrell, 2001) and Agathiphagidae moths (Dumbleton, 1952). In two cases mentioned above (Neophasia and Eumaeus), gymnosperm-feeding characterizes the genus, suggesting a relatively old association (as in the case study of Callioratis and several apparently related genera of cycad-feeding geometrids: Staude, 2001; Bayliss et al., 2009). In five cases however (Faunis, Taenaris, Callophrys, Theclinesthes and Chilades) the genus also feeds on angiosperms, suggesting a more recent host shift. Species of Neophasia are only known to feed on gymnosperms so host shifts are hard to trace without comprehensive phylogenetic analysis of the groups to which they belong, whilst Eumaeus childrenae is also reported from the toxic monocot family Amaryllidaceae and perhaps more doubtfully, E. atala on Manihot (Ehrlich and Raven, 1964; Contreras-Medina et al., 2003). However, the fact that different species of Chilades feed on both gymnosperms and angiosperms makes the genus an interesting candidate for examining the pattern of hostshifts and asking how such a dramatic shift was adaptively possible. In this study, I examine not only the monophyly of Chilades butterflies but also their host associations mapped onto the phylogenetic framework in order to determine origins of Cycas-feeding and frequency of host shifting in the genus Chilades. So far, no morphological or molecular phylogenies are available for Chilades. Only hypothetical genus-relationships have been proposed (Bálint & Johnson, 1997). Therefore, the main aim of this study is to test the relationships among Chilades butterflies and their relatives based on the genes mitochondrial cytochrome oxidase I and II (COI+COII) and nuclear elongation factor 1 35.

(36) alpha (Ef-1α). Information on insect-hostplant associations and geographic distribution were also used to answer how many shifting events happen for Chilades butterflies to use Cycas-plants.. II-2. Materials and methods Taxon sampling I sampled 82 specimens, representing 13 of 20 Chilades species and 32 related species (Table 1 and 2). Chilades butterflies were sampled representatives from three species groups. One member, C. yunnanensis, not classified to above species groups was also sampled in this study. Related genera in Polyommatus section sampled here included Polyommatus, Albulina, Agriades, Plebejus, and Aricia. In addition, the neotropical genera Echinargus and Hemiargus that were considered to be related to the Polyommatus genus-group (Bálint and Johnson, 1997) were also selected. Moreover, several sections within Polyommatini and Niphanda fusca (Niphandini) were sampled to find out the most related genera to Polyommatus genus-group and to test the monophyly of Chilades.. Molecular methods The procedures of DNA extraction, PCR, and DNA sequencing reaction were described in Part I. The primers used for amplifying the mitochondrial COI+COII, and nuclear Ef1α genes have been described (Caterino and Sperling, 1999; Kandul et al., 2004), and the novel primers were listed in Table 10. 36.

(37) Sequence treatments Molecular sequences of COI+COII and Ef1α genes were checked and assembled into contiguous arrays using Sequencher 4.8 (GeneCode, Boston, USA). Coding regions were determined against to the published sequences of Drosophila yakuba (Clary & Wolstenholme, 1985) and Bombyx mori (Yukuhiro et al., 2002). Primer regions and tRNA-Leu gene were cropped, and the undetermined first codon of COI and the complete or uncompleted stop codons of COI and COII found in Lepidoptera (Wu et al., 2010a) were also removed in the analyses to avoid length variation. The data sets were aligned according to amino sequence similarity by ClustalW in MEGA 3.1 (Kumar et al., 2004) with the default settings (gap opening penalty=10, gap extension penalty=0.2). Missing data and ambiguities were designated by “N”.. Phylogenetic analyses A range of phylogenetic methods was used to infer the phylogeny. Maximum parsimony (MP) were performed in PAUP*10b (Swofford, 2000). Bayesian inference (BI) was carried out using MrBayes v. 3.1 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003), and Maximum Likelihood (ML) was performed in PHYML (Guindon & Gascuel, 2003). To evaluate appropriate outgroup and ingroup combinations for Chilades butterflies, in the first runs, I used two specimens of Niphanda fusca as outgroup and the 80 remaining taxa (44 species) as the ingroup for reconstructing relationships based on MP and BI methods (I refer to 37.

(38) this as the “general” dataset). In the MP method, the initial MP trees were performed with heuristic searches, starting trees determined by 10,000 random taxon addition, tree bisection-reconnection (TBR) branch swapping algorithm, gaps treated as missing data. Multiple character states in the same taxon were treated as substitution codons, and all characters treated as equally weighted. A strict consensus tree was computed in cases where multiple equally parsimonious trees were obtained. The consistency index (CI) and the retention index (RI) were also calculated in PAUP*10b. The topologies of MP trees were evaluated by both bootstrap (Felsenstein, 1985) and Bremer support (Bremer, 1988, 1994). Bootstrap analyses were performed with heuristic searches, 100 random taxon addition, TBR algorithm, and only parsimony-informative sites were used to increase variability. The bootstrap replicates were set to 1000. Bremer support (BS) was carried out using the program TreeRot. v3 (Sorenson and Franzosa, 2007) in conjunction with PAUP*, and the number of heuristic searches with random-taxon addition for each branch was set to 100. In BI method, the software Modeltest v. 3.06 (Posada and Crandall, 1998) was used to determine the best fit model. Model selection was determined according to the hierarchical likelihood ratio test (hLRT). The best fit model of GTR+Г+I [General Time Reversible (GTR), invariable site (I), and gamma distribution (Г)] were used. Six chains (five heated and one cold) were run for one million generations and sampled trees every 100 generations. The log-likelihood scores were plotted against generation time to determine when stationarity was reached. This process became stable before the first 2500-3000 trees, and therefore the first 3000 trees were treated as “burn-in” and the remaining 38.

(39) trees were used for representing the posterior probability. In BI, only one specimen of Niphanda fusca (Voucher ID: L1-2048) was set as outgroup due to the single-outgroup setting of Bayesian analysis. Eliot (1973) in his classification admitted his failure to subdivide the huge tribe Polyommatini, but pointed out the similarity of the male genitalia of the Euchrysops and Polyommatus sections. In addition, according to the first run results (Figure 4), the analyses (MP and BI methods) identified that the Everes section was the closest group to Eliot’s Polyommatus section than to other Polyommatini sections. Therefore, Everes argiades (Voucher ID: L1-270) and Euchrysops cnejus were chosen as functional outgroups, the remaining 57 terminals of the Polyommatus-section as ingroup (here referred to as the “specific” dataset) and all other taxa other than outgroups for the specific dataset were excluded from the analyses. MP was performed using the same settings as the first run. In addition, Partitioned Bremer support (PBS) was calculated for evaluating support and congruence among partitions. In BI, the best fit model for the specific dataset was set as the GTR+Г+I model based on Modeltest results, and subsequently the following settings used those of the first run. The first 3001 trees were treated as “burn-in” and the remaining trees were used to calculate posterior probabilities. ML similarly used the model of GTR+Г+I according to hLRT on this dataset, and the values of Г and I were set according to Modeltest suggestions (Г=0.66 and I=0.55). Tree topology search was set to SPR (Subtree Pruning and Regrafting). The initial tree was used default option (BIONJ), and nucleotide frequencies were based on “empirical” option, and the categories of discrete gamma model were set to four. Bootstrap analyses 39.

(40) based on 1000 pseudoreplicates were carried out to evaluate the support for all branching relationships. For reporting the phylogenies, bootstrap support values ranging from 50-70% are referred to as “weakly supported”, those from 71-90% as “moderately well supported”, and values higher than 90% are as “strongly supported” (in accord with Silva-Brandão et al., 2008). Bayesian posterior probability values from 0.5-0.75 are referred to as “weakly supported”, those from 0.76-0.95 as “moderately well supported”, and those over 0.95 as “strongly supported”. Conflict between the three genes is detected by PBS values.. Ecological character mapping of hostplant associations Evolutionary relationships between the butterflies and their hostplant families were investigated. Hostplant records were collected from several sources (Table 2). The optimized algorithms were based on parsimony, and character states were treated as unordered and equal weight by using Mesquite 2 (Maddison and Maddison, 2009). The characters were coded as binary (0 and 1) according to use or non-use of hostplant families, and missing data was coded as “?”. To detect whether there was phylogenetic signal in hostplant use, the methodology (modified PTP test) proposed by Wahlberg (2001) was used. I performed 300 random reshufflings of character states by using the program Mesquite 2. A significant phylogenetic signal is observed when P value is less than 0.05 (Faith and Cranston, 1991). In this study, the probability that the observed parsimony steps for that character do not differ from random is (n+1)/300, where n is the number of replications as short as or shorter than the 40.

(41) observed pattern.. II-3. Results Characteristics of the dataset The total alignment data set contained 3437 base pairs (bp), of which 1530bp was from mitochondrial COI, 682bp from mitochondrial COII, and 1225bp from nuclear EF-1α. No stop codon was found throughout the whole dataset, and only COII showed a 6-bp indel, which was caused by a two-codon insertion in Chilades trochylus (Voucher ID: L1-6100 and L1-2443) and a one-codon insertion in Plebejus argyrognomon (Voucher ID: L1-993) in the same region. There were 945 parsimony-informative sites found in the general dataset and 705 parsimonious sites in the specific dataset.. Phylogenetic relationships General dataset The MP method over the whole combined dataset resulted in 16 most equally parsimonious trees (tree length=4950 steps, CI=0.338, RI=0.715). The strict consensus tree showed that all the members within the tribe Polyommatini were strongly supported (Figure 4), whereas the relationships between Polyommatini sections were still unclear. The Everes section is the closest to Polyommatus section among the sampled sections, and all the Chilades butterflies are grouped into the Polyommatus section. In the BI method, the dataset reached stationarity 41.

(42) before generation 300,000. I discarded the first 3,001 trees and computed the consensus tree from the remaining 14,000 trees. The phylogenetic tree has a similar topology with MP method. The monophyly of the tribe Polyommatini was well supported, and the relationships between the Polyommatini sections represent weak to moderate support. The Everes section also represent the closest to Polyommatus section as in the MP method, and Chilades butterflies represent two separated groups in the good support of the Polyommatus section (Figure 5).. Specific dataset Parsimony searches resulted in 14 most parsimonious trees (tree length=2896 steps, CI=0.454, RI=0.751). The strict consensus tree shows that the Polyommatus section is well supported and contains four major groups (clade A-D, Figure 6). Clade A, including the genus Echinargus and Hemiargus, representing the Neotropical Polyommatus section shows both high bootstrap value and Bremer support (Table 11). Clade C which includes most of the Polyommatus genus-group proposed by Bálint and Johnson (1997, 1999) including Aricia, Plebejus, Albulina, and Polyommatus has moderate bootstrap and Bremer supports. The genus Chilades is not included in the clade C, and this genus is divided into another two clades instead. Clade B which includes most species of genus Chilades, has moderate bootstrap and Bremer supports, while clade D, including C. putli, C.minuscula, C. trochylus, and C. yunnanensis, has good bootstrap and Bremer supports. Clade D is more closely related to Clade C than to Clade B. The three Chilades species-groups are mixed between clade B and D 42.

(43) and the genus is therefore polyphyletic. Group B contains all species-groups, with the galba-species group and lajus-species group in its basal part, and the cleotas-species group forming a subordinate clade of clade B. Clade D also belongs to the lajus-species group. The hitherto uncategorized Chilades species, C. yunnanensis, also belongs to the clade D, three of which (C. putli, C. minuscula and C. trochylus) were formerly included in Freyeria (Bascombe et al., 1999; Parsons, 1999; Ackery et al., 1995). Bayesian analyses for the specific dataset also reached stationarity before generation 300,000. Again the first 3,001 trees were discarded and computed the consensus tree from the remaining 14,000 trees. The topology is similar to the parsimony consensus tree and four major clades are revealed. This time however, the major Chilades group (clade B) is more closely related to clade A, and the minor Chilades group (clade D) is more closely related to clade C. The log-likelihood from the ML method was -19380.86, and the resulting bootstrap values are labeled on Figure 6. ML shows a similar topology to the BI method. As before, four major clades with good support are revealed.. Evolution of hostplant associations Tracing hostplant uses in the Chilades strict-consensus tree, the ancestral states of the tribe Polyommatini and the section Polyommatus are the family Fabaceae (Figure 8). The hostplant uses in Chilades butterflies (Clade B) are divided into two components: the basal lineages feed on rosid plants (including Fabaceae and Rutaceae) as other Polyommatus-section butterflies, whereas the tip clade feeds on 43.

(44) Cycadaceae. In Freyeria butterflies (Clade D), although the hostplant use of the species C. yunnanensis is still unknown, the ancestral hostplant use is also the family Fabaceae. The modified PTP test used to detect phylogenetic signal was significant (P=0.003), implying that the distribution of Cycas-feeding trait among taxa is phylogenetically associated.. II-4. Discussion Phylogenetic relationships of Chilades and Freyeria butterflies The molecular data support the monophyly of Eliot’s Polyommatus section and Chilades butterflies as currently conceived are clearly polyphyletic. Bálint and Johon (1997, 1999) placed the genera Luthrodes, Edales, Lachildes, and Freyeria as synonyms of the genus Chilades. The results support the treatment of first three genera, but the genus Freyeria should be treated as an independent group because both clades are well supported (Figure 6, clade B and D). The genus Freyeria here was formally resurrected, including members of F. trochylus and F. minuscula, F. putli, and F. yunnanensis, comb. nov. Although phylogenetic relationships among the basal lineages of Polyommatus section are still unresolved, Freyeria (clade D) is more closely related to other members of the Polyommatus genus-group (clade C) than to the genus Chilades (clade B) (Figure 6). Moreover, three Chilades species-groups exhibit a mixed pattern, even when the members of Freyeria as conceived here are excluded. The cleotas species-group represent the terminal group in clade 44.

(45) B (Figure 6), however, this species-group and the galba species-group are the two members of clade B. If these species-groups are removed from clade B, the remaining taxa are rendered paraphyletic. C. yunnanensis is rare and only one female specimen is known in the Natural History Museum (BMNH). It has been recollected in the western Yunnan, where it inhabits mountainous areas. Several new observations were made during the course of this study. Firstly, the molecular data show that the Everes section are more closely related to the Polyommatus section, with good support (Figure 4), even though Eliot (1973) pointed out that the Euchrysops section have similar morphological characters in the aedeagus to the Polyommatus section. Moreover, a hypothetical relationship proposed by Bálint and Johnson (1997) was that the Nabokovia genus-group is sister to the Polyommatus genus-group. The molecular data provide no support on this point, whereas the Nabokovia genus-group is evident as one of the basal lineages of the large section Polyommatus based on the combined molecular dataset (Figure 6). Moreover, the topology indicates that Chilades butterflies are a group that originated out of Africa (similar cases are represented by Kodandaramaiah and Wahlberg, 2007 and Aduse-Poku et al., 2009). African Chilades represent basal lineages of clade B, Central Asian Chilades intermediate lineages, and Asian species tip lineages. It is interesting to note that Freyeria represent the reverse pattern to Chilades butterflies, the Asian species represent in the basal part and African species at the tip clade.. 45.

(46) Relationships among the Polyommatini sections Although the main focus of this study was to examining the monophyly of Chilades butterflies, the relationships among the genera of the Polyommatus section and the sections among the tribe Polyommatini are worth discussing. Eliot’s great work (1973) built a basal classification of Lycaenidae, and many authors have followed his fundamental classification (Scoble, 1986; Bascombe et al., 1999; Braby, 2000). Although Eliot (1973) admits himself a failure to find good characters for subdividing the very large tribe Polyommatini into a few major natural groups, the molecular data still strongly support the monophyly of the tribe Polyommatini (Figure 4). Bálint and Johnson pioneered the task of reclassifying Eliot’s Polyommatus section, based on male and female genitalic characters, with a focus on the aedeagus and female terminalia. They provided valuable information about Neotropical blues using good taxon sampling within this section of lycaenids. However, their results indicated that the generic grouping still needs further clarification because the sampled species in the genera Aricia and Polyommatus seems not to compose monophyletic groups (Figures 6 and 7). More sampling efforts are needed to clarify the relationship in this large group.. Evolution of hostplant associations of Chilades butterflies Chilades mindora, C. cleotas, and C. pandava, the three species that are known to be specialized on Cycas plants form a monophyletic group within the clade of Chilades sensu stricto (Figure 8), suggesting the colonization of cycads occurred just once in the evolutionary history of these lycaenids. The phylogeny indicates that the cycad-association was 46.