Identity and phylogenetic status of two sibling gall midge species (Diptera: Cecidomyiidae: Contarinia) on the perennial herb Vincetoxicum hirundinaria
Article first published online: 17 OCT 2002
Volume 27, Issue 4, pages 519–528, October 2002
How to Cite
Widenfalk, O., Gyllenstrand, N., Sylvén, E. and Solbreck, C. (2002), Identity and phylogenetic status of two sibling gall midge species (Diptera: Cecidomyiidae: Contarinia) on the perennial herb Vincetoxicum hirundinaria. Systematic Entomology, 27: 519–528. doi: 10.1046/j.1365-3113.2002.00193.x
- Issue published online: 17 OCT 2002
- Article first published online: 17 OCT 2002
- Accepted 28 March 2002
AbstractContarinia comprises one of the largest genera among gall forming midges of Cecidomyiidae, where identification and species relationships are uncertain. Using data on phenological development, morphometric relationships and mitochondrial DNA, the status of two isomorphic species, C. vincetoxici and C. asclepiadis, which attack the perennial herb Vincetoxicum hirundinaria, were investigated. Data show that they are two distinct species. In rearing experiments, the two gall midges were shown to have different times of adult emergence. Small differences in wing morphology were revealed that separate the two species from each other, as well as from C. loti, the type species of Contarinia. Sequence differences in the mitochondrial cytochrome b gene corroborate the specific status of C. vincetoxici and C. asclepiadis. Furthermore, a phylogenetic analysis, also including three other Contarinia species, showed that the two gall midges on V. hirundinaria are not even the most closely related species, suggesting two separate evolutionary colonizations of the host plant.
Contarinia Rondani comprises one of the largest genera of gall forming midges of Cecidomyiidae, with more than 150 described species from the Palaearctic (Skuhravá, 1986) and sixty from the Nearctic Region (Gagné, 1989). Species are known to develop in malformed flower buds of various dicotyledons, grass spikelets, galls on trees and various leaf and inflorescence galls on both mono- and dicotyledons (Harris, 1966). A few species also develop in fruits (Harris, 1966). The genus, however, is not monophyletic and it serves as a catch-all category for supertribe Cecidomyiidi, including species with long and tapered ovipositors and bifilar male flagellomeres. Some natural species groups have been identified, e.g. the Contarinia sorghicola group attacking grass spikelets, and a segregation of these into separate genera has been discussed (Gagné, 1973, 1989). However, studies on the genus are still few and phylogenetic relationships remain uncertain, both at higher levels and within species groups (Harris, 1966; Gagné, 1973).
Most Contarinia species are of very similar appearance, and many species have not been separated morphologically. As a result, descriptions are often based solely on host plant and gall type, which is often spectacular and distinct. Most gall midge species are known to be highly specific both with regard to host plant species and gall structure (Roskam, 1985; Gagné, 1989; Sylvén & Lövgren, 1995), yet there are examples of more relaxed host plant relationships (Jensen, 1946; Barnes, 1950, 1953; Stokes, 1953; Harris, 1966; Roskam, 1985). The swede gall midge C. nasturti Kieffer can form galls on both leaves and flowers on several Brassica species (Barnes, 1950; Stokes, 1953), and the pea midge, C. pisi Winnertz, is known to induce galls in leaf buds and flowers, as well as fruits, on its host plant Pisum sativum (Harris, 1966). However, examples of Contarinia species galling plants from more than one family are few (Jensen, 1946; Gagné, 1986).
There are problems in distinguishing different Contarinia species on morphological grounds, as well as risks of mistaking one species for two or more if only gall characteristics are used. Here we elucidate the status of two presumably different and poorly described Contarinia species that produce galls in different parts on the same host plant, but which are essentially isomorphic.
Two Contarinia species have been described from Vincetoxicum hirundinaria (Medicus) (Apocynaceae, Asclepiadeae), a perennial herb which is widely distributed in Eurasia with its northwestern distribution limit in eastern Denmark, southeastern Sweden and southwestern Finland (Fig. 1) (Donadille, 1965; Hultén, 1971; Hultén & Fries, 1986). Contarinia vincetoxici Kieffer, which induces galls in the flower buds, has only been described from gall characteristics and larval morphology (Kieffer, 1909). It has been recorded from Germany, former Czechoslovakia, Austria, Denmark, Sweden and Finland (Forsius, 1932; Henriksen, 1944; Sylvén, 1983; Skuhravá, 1986). Contarinia asclepiadis (Giraud), which galls the fruit wall, is described on the basis of gall characteristics and briefly on adult morphology (Giraud, 1863). It has been recorded from the Netherlands, Germany, former Czechoslovakia, Austria and Hungary (Skuhravá, 1986). A preliminary comparison of adult specimens revealed no species-specific differences between C. vincetoxici and C. asclepiadis. Nevertheless, differences in larval feeding habits and adult oviposition behaviour suggested that the two forms of galls represent two different species. Separate species status was further strengthened by differences in the geographical distributions of the two types of galls in Sweden: the flower gall midge is present almost everywhere the host plant occurs, but the fruit midge is very rare and has thus far been found in only two localities.
Using a combination of ecological, morphological and genetic methods, our objective was to determine whether the two forms of galls represent two separate gall midge species, and if so how they relate to each other. Differences between the two forms in oviposition behaviour, larval feeding habits, phenology and geographical distribution are presented, along with a morphometric analysis of wing characters that includes the type species of Contarinia. Finally, including two additional Contarinia species, a phylogenetic analysis using mitochondrial DNA sequences was performed to address the questions of species differences and sister-species status.
Materials and methods
The geographical distribution of the two forms of galls was determined during extensive field surveys across the distribution area of the host plant in Sweden, Finland and Denmark during 1990–2000. During the summer of 1998, seasonal development of gall midges in the field was studied at Bladåker, province of Uppland (Fig. 1, site I), where both types of galls are present. Stands of V. hirundinaria were visited weekly throughout the flowering and fruiting seasons, and all new galls were marked and counted on each visit.
The timing of adult emergence was studied under controlled conditions. Galls were collected in the field and stored in plastic bags at room temperature until larvae emerged. The larvae were then placed in plastic jars (400 ml, thirty larvae in each), with a mixture of sterile peat moss and sand, and kept outdoors (July–August) until the larvae entered the soil. They were then transferred to climate rooms at 0 °C where they were stored for 5 months. After this artificial hibernation, larvae were transferred to a constant temperature of 15 °C and cultures were checked daily for emerging adult midges.
Morphometric analyses were performed on C. vincetoxici, C. asclepiadis and C. loti De Geer, the type species of Contarinia, which causes flower galls on Lotus corniculatus Linnaeus (Fabaceae) (Table 1). The material studied is deposited in the gall midge slide collection in the Department of Entomology, Swedish Museum of Natural History, Stockholm, as voucher specimens.
|Gall midge||Sex||n||Geographical origin|
From specimens stored in 70–80% alcohol, the right wing was removed, dried and mounted between two cover glasses after Sylvén & Antipa Neufeld (1991). Examinations were made under a Zeiss microscope equipped with phase contrast optics and a camera lucida. Drawings were produced at a magnification of 73× and measurements of drawings, taken to the nearest 0.5 mm, were made as indicated in Fig. 2: RW = distance from arculus to distal point of R5, Q = distance from arculus to a point at right angle to the distal point of Cu2; BW = wing breadth at right angle to the midpoint of RW.
The measurement values, divided by 73 and expressed in µm with one decimal place, were tested after Sylvén & Lindberg (1998). The allometric equation (log y = b log x+ loga) is assumed to be applicable, whereby b = the slope and log a = the intercept. Below, log RW or log BW correspond to log x, and log Q to log y. Log Q and log BW values, respectively, were converted to a standard wing length of 1600 µm for females and 1250 µm for males (males of all species are smaller) using the slope b; obtained for each species and sex (for log Q see Fig. 3A,D, log BW not shown). The converted values, Qc and BWc, were expressed as percentages of the standardized RW. The same procedure was used for the lines representing 95% confidence limits of slopes giving the tolerance intervals in Fig. 3C,F. For additional information about methods used, see Sylvén & Lindberg (1998).
DNA extraction and amplification
Contarinia vincetoxici was collected from three populations (Fig. 1; sites II, III and IV) and C. asclepiadis from two populations (Fig. 1; sites I and VII). From each population, three individuals were used for genetic analysis. Three additional Contarinia species, all with different host plants and gall types, were included in a phylogenetic analysis. One individual each of C. loti (Fig. 1; site II), C. lysimachiae Rübsaamen, a bud galler on Lysimachia vulgaris Linnaeus (Fig. 1; site V), and C. tritici Kirby, which attacks wheat spikelets (Fig. 1; site II), were analysed. Adult specimens preserved in 70% alcohol were used for DNA extraction, except in the case of C. lysimachiae where only larvae were available. DNA was extracted using a high salt extraction protocol according to Paxton et al. (1996), except that whole insects were used for extraction and the pellet was resuspended in 50 µl of ddH20. A partial region of the cytochrome b gene was PCR-amplified using universal primers CB1 and CB2 (Crozier & Crozier, 1992). PCR reactions contained 1 × PCR buffer (MBI Fermentas, Vilnius, Lithuania), 2.0 units of Taq DNA polymerase (MBI Fermentas), 200 µm of each dNTP (MBI Fermentas), 1.5 mm MgCl (MBI Fermentas), 400 nm of each primer and 4 µl of template DNA in a total volume of 50 µl. Amplification conditions were as follows: 94 °C for 3 min; 35 cycles of 94 °C for 30 s, 45 °C for 30 s and 1 min at 72 °C. These cycles were followed by a prolonged elongation step of 10 min at 72 °C. PCR products were purified using the Wizard PCR Prep kit (Promega, Madison, Wisconsin, U.S.A) according to manufacturer's instructions. Products were either directly or cycle-sequenced in both directions. Sequencing primers were those used for primary amplifications. Direct sequencing of the purified double-stranded PCR product was performed using the T7 Sequencing kit (United States Biochemical Corporation, Cleveland, Ohio, U.S.A). For direct sequencing, the single-stranded DNA was annealed to a primer, then labelled with α-33P-ATP while being extended. Finally, fragments were subjected to dideoxy chain termination (Sanger et al., 1977). Fragments were separated in 6% denaturing polyacrylamide gels and visualized using autoradiography. Cycle-sequencing was performed using the Thermo Sequenase kit (Amersham Pharmacia Biotech, Little Chalfont, U.K.). Cycle-sequenced products were run on an automated DNA electrophoresis system (Model 4200; Li-Cor, Lincoln, Nebraska, U.S.A).
Sequences were base-called and edited using Base ImagIRTM. 4.0 (Li-Cor). Multiple alignments of sequences were performed using the clustal W (Thompson et al., 1994). Phylogenetic analyses were performed using paup* (Swofford, 2000). To determine the shortest phylogenetic tree, maximum likelihood (ML) and maximum parsimony (MP) analyses were performed. The cytochrome b sequence of Anopheles gambiae Giles (L20934) (Beard et al., 1993) was used as an outgroup because this was the phylogenetically closest sequence to Contarinia available in GenBank. The MP analysis was conducted with ten random sequence additions and TBR branch swapping. To estimate the relative levels of support for nodes on the MP tree, 1000 bootstrapped replicates of the original dataset under the heuristic option were performed. In addition, support indices (Bremer, 1988) were calculated. For the ML analysis, the best fit model was estimated using likelihood ratio tests following Huelsenbeck & Crandall (1997), resulting in a model with unequal base frequencies, rate heterogeneity following a gamma distribution (eight rate categories) and a substitution type (general time reversible) with a rate matrix consisting of six parameters. These estimates were used in ML searches using ten random replicate additions with TBR branch swapping. Heuristic bootstrap analysis using NJ starting trees and NNI branch swapping were performed on 100 replicates. Additional sequence analysis was obtained using the DnaSP 3.14 program (Rozas & Rozas, 1999) and the mega 2.0 program (Kumar et al., 2001).
Biology, distribution and seasonal development
The two kinds of Contarinia galls on V. hirundinaria are different in shape, as well as in number of gall midge larvae they host. There are also differences in larval feeding behaviour. The flower gall is induced when the female C. vincetoxici oviposits in the youngest flower buds, by inserting the ovipositor between the petals. After 7–10 days the flower bud develops into a gall as the first-instar larvae start feeding. The galled flower bud remains closed as the flower base grows and becomes swollen. The petals usually become reddish (Fig. 4), but the coloration varies from deep purple to pale green. On average, flower galls contain fifteen larvae, but occasionally there may be up to fifty larvae per gall. The larvae feed on the liquid that fills the gall interior.
The galled fruit caused by C. asclepiadis is easily recognized by being deformed and often twisted with an undulating surface (Fig. 5A). Later, it develops yellow to brownish patches indicating the feeding activities of gall midge larvae on the inner wall tissue. Females oviposit through the fruit wall, depositing the eggs on the inner wall of the fruit where the developing larvae feed (Fig. 5B). Oviposition seems to be more common in small, immature fruits but mature fruits may also be attacked. Fruit gall midges are often seen ovipositing in previously attacked fruits, which explains why the fruit galls host so many larvae of different stages. Repeated ovipositions have not been observed in the flower gall midge, where the number and homogenous size of larvae within a gall indicate that they result from a single oviposition.
Even though there is considerable overlap in the seasonal occurrence of flowers and fruits of V. hirundinaria, the occurrence of the two types of galls are well separated in time. Flower galls usually occur from mid-June until the end of July, whereas the fruit galls usually start to appear in the first week of August and can be found until the first week of September (Fig. 6A). Rearing under controlled conditions shows that both species have only a single breeding generation per year (a fraction of every larval cohort, however, has a prolonged diapause for two or more years). Furthermore, the emergence times for the adult gall midges are well separated in time (Fig. 6B). Contarinia asclepiadis emerges on average 33 days later than C. vincetoxici if kept at a constant temperature of 15 °C (after hibernation for 5 months at 0 °C).
Both gall midges are heavily attacked by parasitoids, but they do not share any parasitoid species. Two species, Omphale salicis Haliday (Eulophidae) and Synopeas acuminatus Kieffer (Platygasteridae), are common on C. vincetoxici, and one Inostemma species (Platygasteridae) attacks C. asclepiadis.
In the Nordic countries, the flower gall is found almost everywhere throughout the distribution range of the host plant (Fig. 1). By contrast, the fruit gall midge was only found in two geographically widely separated places in Scandinavia, namely Bladåker, province of Upland, and Em, province of Småland (Table 1; Fig. 1). Further surveys may reveal additional places with fruit galls, but it is clear that the fruit gall midge is much less prevalent than the flower gall midge. Where it occurs, however, fruit gall densities are usually high.
A detailed morphometric study of the wing shape clearly distinguishes the two gall midges from each other as well as from C. loti. In female C. asclepiadis, the distance Q (Fig. 2) is significantly shorter in relation to both wing length (Fig. 3A) and wing breadth (Fig. 3B) than either C. vincetoxici or C. loti, whereas C. vincetoxici and C. loti are significantly different only with regard to the latter ratio. These differences are further supported by the converted figures (Fig. 3C) where C. asclepiadis shows lower Qc values than C. vincetoxici or C. loti, whereas the latter midge shows much lower BWc values than either C. vincetoxici or C. asclepiadis. Fewer data are available for males, but for C. asclepiadis and C. vincetoxici they show similar trends as those for the females (Fig. 3D–F). Males of C. loti (Fig. 3E), in contrast to females (Fig. 3B), show no tendency to have higher log Q-values than C. vincetoxici in relation to log BW. As in females, male C. asclepiadis seems to have lower Qc values than C. vincetoxici or C. loti (Fig. 3F). However, in C. loti neither the Qc values nor the BWc values deviate significantly from the corresponding values for C. vincetoxici.
Three hundred and ninety-eight base pairs of partial cytochrome b were sequenced and aligned from all five Contarinia species analysed. The region corresponds to bases 10854–11251 of the Anopheles gambiae mitochondrial genome (Beard et al., 1993). Sequences are available in GenBank, accession nos. AY017525-35. No indels or stop codons were found in the sequences when aligned with the A. gambiae sequence using the insect mitochondrial genetic code. The total number of polymorphic sites was 163. The third codon position showed most variation, 60%, whereas first and second codon positions were less variable, with 26.5 and 13.5% of positions varying. This pattern is typical for genes under functional constraints. The sequences were rich in A + T content (78%), as is common in insects. The A + T content in these sequences can be compared to 74% in A. gambiae for the corresponding region, and 81% for the complete cytochrome b gene in the honey bee, Apis mellifera Linnaeus (Crozier & Crozier, 1992).
The two gall midges on V. hirundinaria show nucleotide differences and sequence distances that confirm their specific status (Table 2). In C. asclepiadis, six specimens revealed four haplotypes, and four haplotypes were found in nine C. vincetoxici. This results in a haplotype diversity of 0.8 ± 0.17 in C. asclepiadis and 0.69 ± 0.15 in C. vincetoxici, which indicates substantial mitochondrial variation in Swedish populations of both species. The number of intraspecific nucleotide differences within the two species ranges from none to eleven for C. asclepiadis and from none to five for C. vincetoxici, whereas the difference between them ranges from forty-eight to fifty-three (Table 2). Of 132 amino acids coded for by the sequences there were none to four amino acid substitutions within C. asclepiadis, none to two within C. vincetoxici and from thirteen to sixteen between the two species. The nucleotide diversity (π) in C. asclepiadis and C. vincetoxici sequences was 0.017 (SD 0.003) and 0.005 (SD 0.001), respectively, and the estimate of the proportion of interpopulational diversity (Nst) was 0.965 ± 0.017. None of the pair-wise sequence comparisons failed Tajima's relative rate test (Tajima, 1993), which implies equal mutation rates (µ) in C. asclepiadis and C. vincetoxici. The transition/transversion ratio (R) of 1.12 (outgroup pruned from analysis) indicates a bias towards transitions as compared to random expectation of 0.5.
|1. C. asclepiadis (I)||–||0.011||0.016||0.028||0.189||0.183||0.183||0.178||0.135||0.054||0.236||0.694|
|2. C. asclepiadis (VII)||4||–||0.016||0.022||0.191||0.186||0.185||0.181||0.131||0.050||0.218||0.691|
|3. C. asclepiadis (VII)||6||6||–||0.031||0.204||0.199||0.198||0.193||0.129||0.061||0.242||0.668|
|4. C. asclepiadis (VII)||10||8||11||–||0.201||0.195||0.196||0.191||0.138||0.051||0.238||0.695|
|5. C. vincetoxici (II)||50||51||53||52||–||0.008||0.013||0.010||0.259||0.214||0.260||0.851|
|6. C. vincetoxici (II)||49||50||52||51||3||–||0.005||0.003||0.253||0.208||0.254||0.829|
|7. C. vincetoxici (IV)||49||50||52||51||5||3||–||0.003||0.254||0.209||0.255||0.809|
|8. C. vincetoxici (IV)||48||49||51||50||4||1||1||–||0.248||0.203||0.249||0.814|
|9. C. tritici (II)||36||35||35||35||59||58||58||57||–||0.122||0.269||0.821|
|10. C. lysimachiae (V)||18||17||20||17||55||54||54||53||35||–||0.201||0.730|
|11. C. loti (II)||57||54||58||56||60||59||59||58||59||53||–||1.161|
|12. Anopheles gambiae||105||105||103||104||115||114||113||113||112||108||126||–|
The phylogenetic analysis shows a clear separation of C. asclepiadis and C. vincetoxici, and furthermore indicates that these two species are not sister species (Fig. 7A,B). However, the topologies of the two trees are somewhat different. The parsimony tree (Fig. 7A) suggests clustering of C. asclepiadis with C. lysimachiae and C. vincetoxici with C. lotii, whereas the position of C. tritici is unresolved. In an analysis using only first and second codon positions, the basic topology is the same, with C. asclepiadis and C. lysimachiae forming one cluster and C. vincetoxici and C. loti forming another cluster. In the likelihood tree (Fig. 7B), C. vincetoxici clusters with C. lotii, in agreement with the parsimony tree, but one of the C. asclepiadis haplotypes is unresolved, and C. lysimachiae and C. tritici forms a group together with C. vincetoxici and C. loti.
By employing a combination of ecological, morphometric and molecular techniques, we show that the two isomorphic Contarinia species on V. hirundinaria are two separate species. Field observations on phenology in combination with rearing experiments show that each gall maker has only one larval generation per summer, and that the timings of adult emergence and gall initiation are widely separated in the two described species (Fig. 6A,B). Observed differences in oviposition behaviour, larval feeding habits, geographical distribution patterns, as well as in parasitoid guilds, further illustrate their ecological separation.
Using morphometric characters not previously used on Contarinia, small but significant differences in wing morphology were revealed. These characters separate the two gall midges on V. hirundinaria from each other, as well as from the generic type species C. loti. The values for females in particular provide strong evidence for C. asclepiadis being morphologically different from C. vincetoxici and C. loti (Fig. 3A–F). The values for females (but not those for males) also indicate morphological differences between C. loti and C. vincetoxici (Fig. 3A–C). This shows that the specific biologies are accompanied by differences, albeit small, in morphology.
The analysis of mtDNA confirms the specific status of C. asclepiadis and C. vincetoxici. Genetic distances between them (Table 2) are comparable with distance values obtained from cytochrome b sequences of other insect species such as bumble bees (Koulianos & Schmid-Hempel, 2000), and even though there is a high genetic diversity within each of the two gallers on Vincetoxicum, most of the genetic variation (97%) was found between them. Species differences are further expressed in the phylogenetic trees, where C. asclepiadis and C. vincetoxici cluster in two separate groups (Fig. 7A,B). In the likelihood tree (Fig. 7B), however, one of the C. asclepiadis haplotypes is unresolved, which reflects the high genetic diversity within this species. The higher genetic diversity observed in the much less prevalent C. asclepiadis as compared to C. vincetoxici may be explained by population subdivision, selection or a bottle neck effect, as the two species have equal mutation rates.
Our findings are in line with the generally high specialization that seems to prevail among many species of gall midges with regard to host plant species and gall structure (Roskam, 1985; Gagné, 1989). Some Contarinia species that have been shown to produce more than one type of gall (Barnes, 1950; Stokes, 1953) are aberrant in that they utilize short-lived host plants, whereas gall midges in general seem to utilize perennials and trees (Roskam, 1985). This suggests that the unpredictable nature of short-lived host plants might be a driving force toward more varied feeding habits.
The present phylogenetic analysis not only demonstrates the specific status of C. vincetoxici and C. asclepiadis, but also suggests that the two species are not sister species (Fig. 7A,B). Although there are topological differences between the two phylogenetic trees, they are consistent in the paraphyletic origin of C. asclepiadis and C. vincetoxici, with, for example, C. loti always closer to C. vincetoxici than C. asclepiadis (Fig. 7A,B). This in turn indicates that the two gall midges have adopted V. hirundinaria as host plant on two separate occasions. This finding differs from patterns observed in, for example, many cynipid gall wasps (Liljeblad & Ronquist, 1998) and in some gall midge genera (Roskam, 1985), where gallers on the same host plant species or on host plants of the same family often are of monophyletic origin. Whether or not the evolutionary pattern of the gall midges on V. hirundinaria is common among Contarinia gall midges in general, we cannot tell. It is likely, however, that host plant shifts are more common in gallers with less complex gall structures and where larvae feed on, for example, nectarlike substances in flowers, than among species whose galls involve more intricate manipulations of the plant tissue (Roskam, 1985; F. Ronquist, personal communication).
We thank Göran Nordlander, Fredrik Ronquist and three anonymous referees for valuable comments on earlier drafts of this paper. We also thank Cia Olsson for laboratory assistance, Gunnel Sellerholm for mounting specimens examined, and for drawing Fig. 2. This study was supported by grants to C.S. from The Swedish Natural Science Research Council.
- 1950) The identity of the swede midge, with notes on its biology. Annals of Applied Biology, 37, 241–248. (
- 1953) The biological approach to the species problem in gall midges (Dipt., Cecidomyiidae). Annales Entomologici Fennici, 19, 2–24. (
- 1993) The mitochondrial genome of the mosquito Anopheles gambiae: DNA sequence, genome organization, and comparisons with mitochondrial sequences of other insects. Insect Molecular Biology, 2, 103–124. , & (
- 1988) The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution, 42, 795–803. (
- 1992) The cytochrome b and ATPase genes of honeybee mitochondrial DNA. Molecular Biology and Evolution, 9, 474–482. & (
- 1965) Contribution a l'étude du Vincetoxicum officinalis Moench, et de ses ‘variétés’ dans le sud-est Français. Naturalia Monspeliensia Serié Botanique, 16, 17–42. (
- 1932) Cecidiologische Beiträge VII. Memoranda Societatis pro Fauna et Flora Fennica, 8, 256–258. (
- 1973) A generic synopsis of the Nearctic Cecidomyiidi (Diptera: Cecidomyiidae: Cecidomyiinae). Annals of the Entomological Society of America, 66, 857–889. (
- 1986) Revision of Prodiplosis (Diptera: Cecidomyiidae) with descripions of three new species. Annals of the Entomological Society of America, 79, 235–245. (
- 1989) The Plant Feeding Gall Midges of North America. Cornell University Press, London. (
- 1863) Verhandlungen der Kaiserlich-Königlichen Zoologisch-Botanischen Gesellschaft in Wien, 13, 1303. (
- 1966) Gall midge genera of economical importance (Diptera: Cecidomyiidae) Part I: introduction and subfamily Cecidomyiinae; supertribe Cecidomyiidi. Transactions of the Royal Entomological Society of London, 118, 313–358. (
- 1944) Fortegnelse over de danske galler (zoocecidier). Spolia Zoologica Musei Haunensis, 6, 257–258. (
- 1997) Phylogeny estimation and hypothesis testing using maximum likelihood. Annual Review of Ecology and Systematics, 28, 437–466. & (
- 1971) Atlas Över Växternas Utbredning I Norden. 2nd edn. Generalstabens Litografiska Anstalts Förlag, Stockholm. (
- 1986) Atlas of North European Vascular Plants North of the Tropic of Cancer. II. Taxonomic Index to the Maps 997–1936. Koeltz Scientific Books, Königstein. & (
- 1946) The identity and host plants of blossom midge in Hawaii (Diptera: Cecidomyiidae: Contarinia). Proceedings of the Hawaiian Entomological Society, 12, 525–534. (
- 1909) Contributions à la connaissance des insectes gallicoles. Bulletin de la Société d'Historie Naturelle de Metz, 26, 1–35. (
- 2000) Phylogenetic relationships among bumble bees (Bombus, Latreille) inferred from mitochondrial cytochrome b and cytochrome oxidase I sequences. Molecular Phylogenetics and Evolution, 14, 335–341. & (
- 2001) MEGA: Molecular Evolutionary Genetics Analysis, Version 2.0.Arizona State University, Tempe, Arizona. , , & (
- 1998) A phylogenetic analysis of higher-level gall wasp relationships (Hymenoptera: Cynipidae). Systematic Entomology, 23, 229–252. & (
- 1996) Mating structure and nestmate relatedness in a communal bee, Andrena jacobi (Hymenoptera, Andrenidae), using microsatellites. Molecular Ecology, 5, 511–519. , , , & (
- 1985) Evolutionary patterns in gall midge–host plant associations (Diptera, Cecidomyiidae). Tijdschrift voor Entomologie, 3, 193–213. (
- 1999) DnaSP Version 3.0: an integrated program for molecular population genetics and molecular evolution analysis. Bioinformatics, 15, 174–175. & (
- 1977) DNA sequencing with chain-terminating inhibitors. Proceedings of the National Academy of Sciences USA, 74, 5463–5467. , & (
- 1986) Family Cecidomyiidae. Catalogue of Palaearctic Diptera, Vol. 4. Sciaridae – Anisopodidae (ed. by A. Soós and L. Papp), pp. 72–297. Elsevier, Amsterdam. (
- 1953) Biological investigations into the validity of Contarinia species living on Cruciferae, with special reference to the swede midge, Contarinia nasturtii (Kieffer). Annals of Applied Biology, 40, 726–741. (
- 2000) paup*: Phylogenetic Analysis Using Parsimony, Version 4.0 Beta. Sinauer, Sunderland, Massachusetts. (
- 1983) Gallmyggor påÖlands alvar. Entomologisk Tidskrift, 104, 136–150. (
- 1991) A preparation method for gall midges. Entomologisk Tidskrift, 112, 153–155. & (
- 1998) Morphometric analyses of Palaearctic Dasineura and Jaapiella species (Diptera: Cecidomyiidae, Oligotrophini) developing in flower heads of Asteraceae, with description of four new species. Entomologica scandinavica, 29, 241–265. & (
- 1995) Dasineura ingeris sp.n. (Diptera: Cecidomyiidae) on Salix viminalis in Sweden, including comparisons with some other Dasineura species on Salix. Systematic Entomology, 20, 59–71. & (
- 1993) Simple methods for testing molecular clock hypothesis. Genetics, 135, 599–607. (
- 1994) Clustal W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Research, 22, 4673–4680. , & (