Morphological and Molecular Revision of the Genus Ozirhincus (Diptera: Cecidomyiidae)—Long-Snouted Seed-Feeding Gall Midges on Asteraceae

The Palaearctic gall-midge genus Ozirhincus is unique among the Cecidomyiidae for its morphology and biology. Unlike most other phytophagous gall midges, species in this genus do not induce galls but develop inside achenes of Asteraceae plants. The heads of adults are characterized by an unusually elongate proboscis, the function of which is unclear. Despite a lot of attention from taxonomists in the 19th and early 20th century, a proper revision of the genus has been hindered by complex host associations, the loss of most relevant type material, and the lack of a thorough comparative study of all life stages. The present revision integrated morphological, molecular, and life-history data to clearly define species boundaries within Ozirhincus, and delimit host-plant ranges for each of them. A phylogenetic analysis based on the mitochondrial COI and 16S genes confirmed the validity of four distinct species but did not resolve the relationships among them. All species are oligophages, and some may occur together on the same host plant. Species with wider host-plant ranges have wider European and circum-Mediterranean distribution ranges, whereas species with narrower host ranges are limited to Europe and the Russian Far East. As part of the present work, O. hungaricus is reinstated from synonymy, O. tanaceti is synonymized under O. longicollis, neotypes are designated for O. longicollis and O. millefolii, and a lectotype is designated for O. anthemidis.


Introduction
Gall midges (Cecidomyiidae) constitute one of the largest families of Diptera, with more than 6200 described species and many that are undescribed or unknown [1]. Most of these species are plant feeders in the larval stage, and induce galls, the complexity and morphological diversity of which is rivalled only by gall wasps (Hymenoptera: Cynipidae) [2], [3]. Despite the ubiquity of gall midges and their fascinating biology, the family is considered to be taxonomically difficult because of the relative scarcity of useful characters, the small size of the midges, and the fact that their study requires special techniques. Knowledge about gall midge systematics is lacking, and many genera are in need of taxonomic revision or even placement to tribe [1]. Such revisions ideally combine morphological, molecular, and life-history data, as this approach produces the most informative and reliable taxonomy, and may offer insight into the evolution of the studied groups.
Ozirhincus Rondani is a small cecidomyiid genus that is unique for its morphology and biology. Adults in this genus differ from other cecidomyiids in having an unusually elongate proboscis and flattened head, and for developing in achenes of Asteraceae rather than inducing galls, as do most other members of the tribe Lasiopterini [4], [1]. Members of this genus are known only from host plants of the tribe Anthemideae and are restricted to the Old World, with the exception of O. millefolii, which was introduced into North America in colonial times [5], [4]. Based on its morphology and biology, the genus belongs to the subtribe Lasiopterina [6], and is apparently closely related to Lasioptera, one of the largest and most loosely defined genera of Cecidomyiidae [4]. The unique morphological characters exhibited by Ozirhincus species, and their specialized life history, strongly suggest that the genus constitutes a monophyletic group.
The genus has drawn the attention of many researchers since it was first described by Rondani in 1840 [7], as attested by unusual nomenclatorial vicissitudes (reviewed by Verrall, [8]), but despite this attention, its taxonomy remained confused until now. A major difficulty stemmed from the fact that Ozirhincus species appeared to have a wide range of host plants, and that some of these hosts support more than one species of gall midge at the same time and place [9][10][11][12]. An added complexity was introduced by Möhn's revision of Ozirhincus [11], [12], which was based solely on larvae that were dissected from dried herbarium material. In that revision, Möhn synonymized 5 of the previously known species and described 7 new ones, basing all taxonomic decisions on problematic morphological characters of the mostly uninformative larvae. Recognizing that larval characters show high intraspecific variability and low interspecific differences, and based on collecting efforts throughout Europe, Skuhravá [13] reversed most of Möhn's decisions, leaving Ozirhincus with 5 valid species, though without clear characters to distinguish between them. The fact that most historical types of species described in the 19 th century are considered lost, led to further complications for clarifying the taxonomy of this genus.
The objective of the present study was to settle the taxonomy of Ozirhincus through extensive sampling from known and potential host plants and the combination of morphological and molecular data. Our specific goals were: 1) to identify reliable morphological characters for distinguishing between the species in the genus based on a thorough study of adults, pupae, and larvae, 2) to clarify the complex host associations in the genus, and 3) to produce a phylogeny for the genus based on genetic markers.

Collecting and rearing of insects
Gall midges in the genus Ozirhincus do not cause the formation of galls, therefore rearing them from their host plants requires collection of normal-looking inflorescences of potential host plants towards the end of flowering, and keeping them in rearing cages in the laboratory until adult emergence. In this study we screened 27 potential host plants (Table 1) in Germany, Israel, and the UK, based on host records from the literature and the fact that Ozirhincus is known only from plants of the tribe Anthemideae [1], [11,12]. Ozirhincus millefolii was also collected in the USA, where it is widespread on its main host plant, Achillea millefolium, which had been introduced from Europe, probably in colonial times [4]. In the following species descriptions, only those names of plants that were confirmed as hosts in the present study are given. Many other plant species have been mentioned in the literature but were not confirmed here. Some of them were sampled in the present work but did not yield gall midges; for those that were not sampled in the present study, further sampling and rearing will be needed to confirm their host status because host records were made based on larvae alone [11], [12], and it is impossible to say to which Ozirhincus species those larvae belonged.
Inflorescences were collected in the field and transferred to the laboratory in large plastic bags. They were then either kept in sealed bags, or placed as bouquets in water in ventilated rearing cages until adult gall midges emerged from them or until they wilted without producing gall midges. Plant species that did not yield gall midges were usually re-sampled several times during their flowering season, sometimes in two or three consecutive years and at several localities, to confirm that they are not hosts. Some inflorescences were dissected under a stereomicroscope to obtain larvae and pupae for morphological study.
To establish the monophyly of Ozirhincus, we chose outgroups representing four genera from the two subtribes of Lasiopterini, the tribe to which Ozirhincus belongs. The subtribe Lasiopterina is represented by Lasioptera, a large, cosmopolitan genus that is closely related to Ozirhincus based on morphological characters, whereas the subtribe Baldratiina, which is

Plant
Asteraceae tribe Confirmed as a host?

Tanacetum vulgare
Anthemideae + restricted to host plants of the Chenopodiaceae is represented by the genera Baldratia, Careopalpis, and Stefaniola. All methods described for Ozirhincus apply also to the outgroups.

Molecular methods
Genomic DNA was extracted from whole adult or immature midges using mostly individual silica-membrane columns from the Blood and Tissue kit by Qiagen (Hilden, Germany). DNA extracts are available from the ZFMK Biobank, Bonn (DNA voucher IDs are given in Table 2). For PCR amplifications, we used the Qiagen Multiplex PCR kit, following the manufacturer's specifications and based on 2-2.5μl undiluted DNA template in 20μl total reaction volumes. The GeneAmp PCR System 2700 (Applied Biosystems, Foster City, CA, USA) was used to amplify two mitochondrial genes. A 658 bp fragment from the 5' part of the mitochondrial cytochrome oxidase subunit I (COI) gene (the so called 'DNA barcoding gene') was amplified using the primers LCO1490 (F) and HCO2198 (R) [14]. Failed reactions were repeated with the primer combination LCO1490 (F) and C1-N-2191 (R, aka 'Nancy') [15]. A 532-533 bp fragment from the 3' end of the 16S ribosomal RNA gene was amplified using the primers 16S-ar-JJ and 16S-1472-JJ [16]. For both genes, the touchdown PCR protocol started at an annealing temperature of 55°C, decreasing to 40°C for the remaining 25 cycles. Following enzymatic clean-up (Exo/SAP), double-stranded sequencing was conducted on an automated ABI 3730XL sequencer (Applied Biosystems) at the Macrogen facility, Amsterdam, NL. Sequences were assembled, inspected and aligned using Geneious vers. R7 (Biomatters, Auckland, New Zealand). The COI + 16S datasets were concatenated in BioEdit [17], resulting in a combined alignment of 1191 bp. Sequences are deposited in GenBank (http://www.ncbi.nlm.nih.gov/) and accession numbers are provided in Table 2.
Phylogenetic analysis MODELTEST ver. 3.7 [18] consistently identified the GTR+G model of nucleotide substitution [19] as the best-fit model for the COI data. For 16S, heterogeneous models were suggested by different algorithms (hierarchical likelihood ratio test, Akaike information criterion, Bayesian information criterion) and conditions (Bayesian information criterion), but most converged on assuming a number of substitution types of 6, so that the same model was applied as for COI. Bayesian analyses were conducted using MrBayes vers. 3.2.0 [20]. Specific parameters for the GTR+G model were equated by MrBayes. Parameters were unlinked between the 3rd versus 1st plus 2nd codon positions and for 16S. We ran two independent replicates (4 chains each) for 70 million generations per analysis. Every 1000th tree was sampled. Negative log-likelihood score stabilization was determined in a separate visualization and the trees for the first 19.000 generations were discarded as burn-in accordingly. We thus retained 139,962 trees, which were used for building a 50%-majority rule consensus tree with posterior probabilities. We reconstructed the evolution of proboscis length as an unordered multistate character onto our phylogenetic tree using the maximum parsimony approach, as implemented in MES-QUITE vers. 3.03 [21].

Taxonomy
Larvae, exuviae, and adults of the gall midges were preserved in 70% ethanol for morphological study, and were later mounted on permanent microscope slides in euparal according to the method outlined by Gagné [22]. Specimens are mounted on slides individually unless otherwise noted. Relevant historical material from the Rübsaamen, and Möhn collections was examined, and its condition evaluated. Material in these collections had originally been mounted on temporary microscope slides in glycerin, or stored in 70% ethanol, and many of the specimens are either lost or have deteriorated to the point that they cannot be used for taxonomic study. Some of that material that was found to be in reasonable condition was remounted on permanent microscope slides in euparal for the purpose of the present study and to ensure its integrity and long-term preservation.
Illustrations of morphological structures were made with the aid of a drawing tube or a Leica DFC295 camera mounted on a Leica DM1000 LED compound microscope. Pupae were studied under a scanning electron microscope. Some adults were also pinned to preserve the color pattern created by the thick covering of scales particularly on the abdomen. Proboscis length was measured from the suture at the base of the labrum to the tip of the labella, and expressed in relation to the length of the antennal scape. Ovipositor length was measured from the base of the eighth abdominal segment to the apex of the cercus, and expressed in relation to the length of segment 8. Some adults and immature stages were preserved in 96% ethanol for molecular study. We compared newly collected material to types and other relevant material deposited in the Staatliches Museum für Naturkunde, Stuttgart, Germany (SMNS), Museum für Naturkunde, Berlin, Germany (ZMHB), the Naturhistorisches Museum, Vienna, Austria (NHMW), the Natural History Museum in London, UK (BMNH), and the private collections of Marcela Skuhravá (Prague) and Eddy Dijktra (Delft). Terminology for adult morphology follows McAlpine et al. [23], and terminology for immature morphology follows Gagné [4]. The specimens examined in this work are deposited in the National Collection of Insects, Zoological Museum, Tel Aviv University (TAUI) unless otherwise indicated.

Phylogenetic analysis
The complete molecular dataset of COI and 16S consisted of 1191 bp. Unaligned fragment lengths for 16S were 532-533 bp, i.e. included only a single-base indel. The COI fragment consisted of 658 positions and contained no internal gaps. The phylogenetic analysis yielded four strongly supported and genetically quite distant clades, clearly representing four valid species in the genus Ozirhincus (Fig 1), which can be distinguished from each other based on unique combinations of larval, pupal, and adult morphological characters. However, the analysis could not resolve the phylogenetic relationships among the four species. The ancestral states analysis did not yield conclusive results regarding the history of proboscis length in the genus (see character states in Fig 1). It is likely that the short proboscis in O. anthemidis represents the ancestral state, similar to the situation in Lasioptera, and that the typical elongate proboscis evolved only once in the history of the genus, but other scenarios cannot be ruled out.
Mapping the identity of host-plant genera on the phylogenetic tree (colors in Fig 1) suggests that O. millefolii is restricted to the genus Achillea, whereas each of the remaining species is capable of developing in host plants of various genera. It also indicates that certain host plants support more than one Ozirhincus species at the same time and place: Tripleurospermum inodorum is used by O. longicollis and O. hungaricus, while some Anthemis spp. are hosts to both O. longicollis and O. anthemidis. In all of these scenarios it is possible to differentiate among the gall-midge species based at least on their adult and pupal morphology.

Taxonomy
Ozirhincus Rondani 1840 Type species: Ozirhincus longicollis Rondani 1840: 16; by monotypy Ozirhincus is a small genus in the tribe Lasiopterini, apparently closely related to Lasioptera. Like other Lasiopterini, it is characterized by a thick covering of scales (Fig 2A-2C), an irregular number of antennal flagellomeres that are gynecoid in the male, a very short R 4+5 that Phylogenetic tree of Ozirhincus Rondani based on Bayesian analysis of partial sequence of the cytochrome oxidase subunit I (COI) and ribosomal RNA16S mitochondrial genes. Support values are shown next to nodes, above branches. Character states representing proboscis length as suggested by the ancestral states analysis are shown below branches (in square brackets). Letters and numbers following species name refer to collecting localities and dates (details in Table 2). Colors correspond to host-plant genera. reaches C proximal to mid-length of wing, longitudinally divided mediobasal lobes in the male that sheath the aedeagus almost to its apex, a protractible ovipositor with a lateral group of hooked setae on segment 8 of the female abdomen, and variously modified setae on the cercal segment. In the closely related Lasioptera, it was argued that these modified setae in the female function in collecting and carrying conidia [24], [25], but their function in Ozirhincus is unclear because species in this genus do not appear to be associated with fungal symbionts. Adults in this genus have unusually elongate mouthparts that form a short to very long proboscis, composed mostly of the strongly setose labrum and labella, and a considerable elongation of the frontoclypeal membrane (Fig 3A-3E). The head in most species is flattened, with the occiput encroaching the eye area. Palps are four-segmented. Tarsal claws are toothed on all legs. The body is covered by black and white scales that form dorsal transverse stripes on the abdomen. Larvae have a bifid spatula in the third instar, with a reduced number of 3-4 lateral asetose papillae on each side, and usually 2-3 setose papillae on each side of the terminal segment. Pupae have pointed antennal bases ('antennal horns') that terminate in one or two tips. All species develop in achenes of Asteraceae belonging to the tribe Anthemideae ( Fig 2D). Pupation takes place inside the achene and the species complete at least two generations a year.

Descriptions
In the following descriptions, the names of host plants are given first because the identity of the host plant is one of the most important characters that may aid in species identification. Only those plant species that were confirmed as hosts in the present study are listed.
Ozirhincus anthemidis (Rübsaamen 1916 Adult. Head ( Fig 3A): Eye facets round; more spaciously arranged on vertex than laterally; eye bridge 3-4-facets long. Antenna (Fig 5A-5C): scape wide trapezoidal; pedicel globose; number of flagellomeres 11-12 in both sexes, rarely 10, number occasionally differs between antennae of same individual (n = 102♀, 104♂); flagellomeres globular to almost quadrate in female, more cylindrical in male (Fig 5B and 5C); first two flagellomeres usually partially to entirely fused, apical flagellomere often longer, evidently composed of 2-3 entirely or partially fused units; adjacent flagellomeres sometimes fused in mid antenna ( Fig 5C); each flagellomere with two whorls of appressed circumfila and two rows of strong setae originating from prominent sockets, one row proximal to and one row between circumfila ( Fig 5A); entire flagellomere surface other than neck covered by microtrichia. Palpus 4-segmented; segment 1 only slightly longer than wide, segments 3-4 about same length, 3 times longer than wide, with several strong setae and otherwise setulose. Frontoclypeal membrane with several strong setae on each side. Labrum about 4 times as long as width at base, parallel sided on basal two thirds, tapering from apical third towards setulose apex, with a few strong setae dorsally. Labella (Fig 6A) about 2.5 times as long as wide, somewhat concave medially, tapered apically, with several strong setae and densely setose along medio-apical surface.
Thorax: Dark grey, covered by white scales and setae. Anepimeron with group of 15-20 setae; other pleura without setae. Legs: dorsal part densely covered by black scales other than a patch of white scales on basal half of first tarsomere (Fig 2A-2C); ventral part densely covered by white scales. Tarsal claws (Fig 6C-6E) evenly curved, with thin tooth, strongly curved close to base; empodia longer than bend in claw; pulvilli about 0.3 times as long as claw. Wing: hyaline, with sparse delicate hairs on entire surface and long hairs along posterior margin; length 1.09-1.76 mm in males (n = 65), 1.09-1.66 mm in females (n = 58); R 4+5 joins C around midlength of wing, densely covered by mixed black and white scales; C with break after meeting point with R 4+5 , densely covered by black scales almost to wing apex, except for patch of white scales at meeting point with R 4+5 (Fig 2A-2C); M straight, CuA unforked. Stem of halter light orange, without scales; knob densely covered by black scales on basal third, white scales covering remainder. Female abdomen (Fig 7A and 7B): Dorsum with dense covering of black and white scales: each tergite with wide transverse stripe of black scales on most of surface, and a narrow strip of white scales along posterior margin. Pleuron and venter with white scales. Tergites 1-6 rectangular, with anterior pair of trichoid sensilla, posterior row of strong setae, and otherwise evenly covered by scales; tergite 7 much smaller than preceding, pigmentation evanescent laterally at midlength, with anterior pair of trichoid sensilla and posterior row of setae; tergite 8 divided longitudinally into two elongate sclerites (Fig 7C), each with wide anterior area with pointed dorsal extension and wide posterior area with pointed ventral extension connected by long, narrow band; each sclerite with trichoid sensillum on anterior part of narrow band, and a group of strong, posterior setae on widened posterior area. Eighth tergite 1.65-2.77 times as long as seventh tergite (n = 56). Sternites 2-7 rectangular, with pair of closely approximated trichoid sensilla, posterior row of setae, and several setae laterally and medially, more numerous on more posterior segments; sternite 8 not apparent. Ovipositor long, protrosible, 3.57-7.47 times as long as eighth tergite (n = 57), with pigmented lateral sclerite along segment 9 and lateral group of strong, arched setae originating from prominent sockets, pointed mostly ventrally. Cercal segment (Fig 7B) with dorsolateral sclerotized plate more strongly pigmented along posterior area than elsewhere, laterally with 10-20 short, strong and straight setae; posterior pigmented area with small dorsal projection and bearing 3-4 very long, hook-like, blunt setae. Apical lamella cylindrical, evenly setulose, with numerous strong setae mostly concentrated on dorsal and apical areas. Hypoproct setulose.
Male abdomen (Fig 8A): Color pattern and scale covering as in female. Tergite 1 rectangular, with posterior row of strong setae, and evenly scattered scales; tergites 2-6 similar but larger and with anterior pair of trichoid sensilla; tergite 7 weakly sclerotized posteriorly, with fewer posterior setae, not forming a row; tergite 8 narrow, band-like, without setae other than anterior trichoid sensilla. Sternites 2-7 rectangular, with pair of closely approximated trichoid sensilla and 1-2 posterior rows of strong setae; posterior sternites with several strong setae medially, and otherwise evenly covered by scales; sternite 8 less pigmented and more setose than preceding but without trichoid sensilla. Terminalia (Fig 8B-8E): Gonocoxite cylindrical, about same width throughout length, with mediobasal lobe divided into prominent, globose, densely setose dorsal lobe, and elongate, ventral lobe tapering towards apex, and further subdivided into two longitudinal ridges on apical two thirds, sheathing aedeagus almost to apex ( Fig  8D). Gonostylus widest at about mid length, tapering to wide comb-like tooth, with numerous setae, setulose on basal half dorsally and basal two thirds ventrally, remaining part with shallow ridges (Fig 8E). Aedeagus slightly longer than sheathing mediobasal lobes, wide and blunt apically. Hypoproct entire, blunt apically, or with slight, shallow notch, setose and setulose. Cerci separated by a deep notch, setose and setulose.
Larva (third instar) (Fig 9A-9C). Light to dark yellowish-orange. Cylindrical-ovate. Integument covered by rounded verrucae. Antennae 1.5-2.0 times as long as wide. Cephalic apodeme considerably longer than head capsule ( Fig 9A). Spatula (Fig 9A and 9H) long shafted and bidentate; shape of teeth and distance between them highly variable; when teeth farther apart, sometimes with minute additional projection between them. Sternal papillae without setae; pleural and dorsal papillae with long setae. On each side of spatula 3-4 asetose lateral papillae grouped together, and one asetose ventral papilla somewhat farther away (Fig 9B). Terminal abdominal segment with 2-3 setose papillae on each side ( Fig 9C). All specimens obtained in the present study had 5 lateral papillae on each side; number of terminal papillae varied among individuals from different host plants: those from Chrysanthemum spp. and Anthemis rascheyana had 2 papillae on each side, those from A. tinctoria had 3, and those from A. bornmuelleri had either 2 or 3.
Pupa (Fig 4A and 4B). Light to vivid orange. Antennal bases enlarged, forming straight bidentate horns, tapered and pointed ventroapically. Vertex with long and thin cephalic seta on each side, situated on conspicuous bulge. Face without apparent papillae. Prothoracic spiracle long and slender. Abdominal segments covered by acute spicules.
Distribution. Europe and circum-Mediterranean. Material examined. The type series of O. anthemidis includes males, females, and pupal exuviae that have been kept in ethanol vials since their collection. This material was mounted on permanent microscope slides in euparal for the purpose of the present study, and used for the designation of the following types: LECTOTYPE: ♀, Germany, Werlau, 1895, EH Rübsaamen, reared from Anthemis tinctoria, (115e). The lectotype is mounted on a permanent microscope slide in euparal, is in fair condition, and deposited in ZMHB. PARALECTOTYPES: 1♂, 2 exuviae, Germany, Werlau, 1895, EH Rübsaamen (115e) (ZMHB) (same data as lectotype; both exuviae on same slide); 2♀, 2♂, 4 exuviae, Germany, Oberheimbach, 17 Biology. The phenology of O. anthemidis varies within its distribution range and in relation to its host plants. In dry Mediterranean regions (Middle East and North Africa), the main activity period is in spring, whereas in continental Europe adults are active also in summer. Adults of the overwintered generation emerge in Europe from mid-March to late April and those of the subsequent generation between August and October [11]. In Israel, adults associated with Chrysanthemum coronarium emerge in March from infested achenes in the ground from the previous year, and can be seen hovering over and standing on developing inflorescences, where females lay their eggs into individual flowers. Larvae develop and soon pupate inside the achenes, preventing them from developing seeds. Adults of the subsequent generation emerge already in April, in contrast to the much slower development of larvae in Europe. Adults reared in May may represent either late-developing individuals of the second generation or those of a third generation. Infested achenes are slightly inflated and enlarged compared to normal achenes, and the larvae or pupae fill them completely making it hard to dissect them from the achenes without damaging them. Females of the last generation lay eggs in the flowers, and the hatching larvae develop to third instars but do not pupate. Instead, they enter a diapause inside the achenes that drop to the ground when the plants have dried up, and will emerge as adults the following year. The life history of populations associated with Anthemis spp. in Israel is essentially similar except that the main activity period is in late rather than early spring. Diapause takes place in the summer through winter months in Mediterranean regions, or only during the autumn and winter in continental Europe. Remarks. Rübsaamen [9] repeated Loew's [26] rearings from Tripleurospermum inodorum and Anthemis arvensis, as well as from A. tinctoria, but noticed that he was getting two types of adults from the latter two hosts: some with a rather short proboscis and some with the typical long proboscis mentioned by Loew for his Clinorhyncha chrysanthemi (currently Ozirhincus longicollis). Rübsaamen wondered how Loew had not noticed the two different populations, and speculated that he did not manage to rear the short-snouted individuals for some reason. He correctly attributed the long-snouted adults to Clinorhyncha chrysanthemi, noting that they had 10 antennal flagellomeres as opposed to 11-12 in the short-snouted specimens, and described the latter as a new species-Clinorhyncha anthemidis (currently Ozirhincus anthemidis).
Ozirhincus anthemidis is easily distinguishable from all other species in the genus by its short proboscis, short fourth palpal segment, and 11-12 antennal flagellomeres, as opposed to 8-9 in O. millefolii and 10 in O. longicollis. Ozirhincus hungaricus, the only other species in the genus with a regular number of 11-12 antennal flagellomeres, has a much longer proboscis and is found on different host plants.
Ozirhincus anthemidis is a very common species throughout Europe and the Mediterranean region. It was previously known only from Anthemis hosts [12], [13], but in the present study we found that it is also very common on Chrysanthemum coronarium and C. segetum (and it was the only species we reared from those plants). Host plants that were reported for O. anthemidis by Möhn [12] but not verified in the present study are Anthemis austriaca, A. cretica (as A. montana), A. kotschyana, and A. cotula. Anthemis arvensis is listed here as a verified host plant based on the information in Rübsaamen's original description of O. anthemidis.
Ozirhincus dalmaticus and O. kabylensis were described by Möhn [12] as belonging to the 'anthemidis group' based on small differences in the larval spatula and chaetotaxy, but were later synonymized under O. anthemidis by Skuhravá [13], with whom we concur. Adult. Head (Fig 3B): Occiput extending forward at expense of eye area in frontal view; eye bridge about 4 facets long. Antenna: number of flagellomeres 11-12 in both sexes (n = 100♀, 89♂); flagellomeres short-cylindrical to almost quadrate in female, more elongate in male; first two flagellomeres usually partially to entirely fused, apical flagellomere slightly tapered, sometimes appear as if 'budding' to form the beginning of a small additional segment. Palpus: fourth segment 1.3-2.0 times longer than third. Frontoclypeal membrane greatly extended, with many strong setae on each side. Labrum long, triangular: at least 4 times as long as width at base, tapering and strongly setose on apical quarter. Labella more than 10 times as long as wide, somewhat concave medially, tapered apically, with several strong setae towards apex and densely setose along entire medial surface (Fig 6B). Female abdomen: Elongate sclerites of tergite 8 ( Fig 7D) each with wide anterior area, usually with rounded rather than pointed dorsal and ventral extensions and wide posterior area connected by long, narrow band; each sclerite with trichoid sensillum on anterior third of narrow band, and 8-10 posterior setae on widened posterior area. Eighth tergite 1.64-3.34 times as long as seventh tergite (n = 46). Ovipositor 4.13-7.77 times as long as eighth tergite (n = 46).

Ozirhincus hungaricus
Larva. (third instar). Integument covered by rounded verrucae. Antennae 2-3 times as long as wide. Cephalic apodeme considerably longer than head capsule. Spatula (Fig 9E and 9I) long-shafted and bidentate; teeth widely separated by concave to straight gap, sometimes with minute additional projection between them. Sternal papillae without setae; pleural and dorsal papillae with long setae; on each side of spatula 4 asetose lateral papillae grouped together, and one asetose ventral papilla somewhat farther away (Fig 9E). Eighth abdominal segment with 3 setose papillae on each side (Fig 9D).
Pupa (Fig 4C and 4D). Antennal bases enlarged, bulging, with tiny tapered tip at center of bulge.
Distribution. Widespread in Europe to Siberia. Remarks. Möhn [12] described O. hungaricus, O. parvus, and O. hispanicus as belonging to the 'tanaceti group' based on characters of the larvae alone, but these species were later synonymized under O. tanaceti by Skuhravá [13]. The larvae of all species in the 'tanaceti group' have 4 lateral papillae on each side of the spatula and 3 terminal papillae on each side of the anus. Because the holotype larva of Ozirhincus hungaricus survived in a recognizable condition while those of O. parvus and O. hispanicus did not, we reinstate O. hungaricus as the name of the common species associated with Tanacetum vulgare in Europe, and synonymize the other two species under it. Tanacetum corymbosum is listed here as a confirmed host because the holotype of O. hungaricus had been found in that plant [12].
In the present study we found that, in addition to T. vulgare, O. hungaricus also develops in Tripleurospermum inodorum, which it sometimes shares with O. longicollis. When the two species occur together on that host plant, they can be distinguished from each other easily by: the number of antennal flagellomeres (10 in O. longicollis vs. 11-12 in O. hungaricus); the characteristic constrictions of the male flagellomeres in O. longicollis, where they are girdled by the circumfila loops (Figs 3E, 5D and 5E); the clear difference in the shape of the pupal antennal horns (Fig 4E vs. Fig 4C); and by the following combination of characters in the third-instar larva: the teeth of the spatula in O. longicollis are usually closer together than in O. hungaricus (compare Fig 9F and 9E); O. longicollis has 3 lateral papillae on each side of the spatula and 2 terminal papillae on each side of the anus, as opposed to 4 lateral papillae and 3 terminal papillae on each side in O. hungaricus. Adult. Head (Fig 3D and 3E): Occiput extending forward at expense of eye area in frontal view; eye bridge about 4 facets long. Antenna: number of flagellomeres 10 in both sexes (n = 57♀, 36♂); flagellomeres short-cylindrical to almost quadrate in female, more elongate in male ( Fig 5E); first two flagellomeres usually partially to entirely fused, apical flagellomere slightly tapered, sometimes appears as if 'budding' to form the beginning of a small additional segment; male flagellomeres characteristically constricted in mid-section by circumfila loops (Figs 3E and 5D). Palpus: fourth segment usually 1.3-1.6 times longer than third. Frontoclypeal membrane greatly extended, with many strong setae on each side. Labrum long-triangular: at least 4 times as long as width at base, tapering and strongly setose on apical quarter. Labella more than 10 times as long as wide, somewhat concave medially, tapered apically, with several strong setae towards apex and densely setose along entire medial surface.
Female abdomen: Elongate sclerites of tergite 8 (Fig 7E), each with wide anterior area with pointed dorsal and ventral extensions and wide posterior area connected by long, narrow band; each sclerite with trichoid sensillum on anterior part of narrow band, and 4-6 posterior setae on widened posterior area. Eighth tergite 1.53-2.51 times as long as seventh tergite (n = 22). Ovipositor 3.90-7.24 times as long as eighth tergite (n = 12).
Larva. (third instar). Integument covered by rounded verrucae. Antennae 2-3 times as long as wide. Cephalic apodeme considerably longer than head capsule. Spatula (Fig 9F) long shafted and bidentate; teeth separated by narrow gap. On each side of spatula 3 asetose lateral papillae grouped together, and one asetose ventral papilla somewhat farther away. Eighth abdominal segment with 2 setose papillae on each side.
Distribution. Europe, Israel. Probably circum-Mediterranean. Material examined. The type of Ozirhincus longicollis Rondani is considered lost, based on an exhaustive study of Rondani's collection by Gagné and Solinas [27]. Given that this is the type species of Ozirhincus, we hereby designate a neotype for it in order to clarify the application of the name O. longicollis Rondani and the generic concept of Ozirhincus as a whole. Although the host plant from which O. longicollis was described in Italy is unknown, the morphological description given by Rondani is distinctive, and the species has been reported since from various locations and host plants throughout Europe [9][10][11], [26]. Based on this information, we designate the neotype from Germany, from the first host plant with which this species has been associated [26]. Biology. The life history of this species is similar to that of O. anthemidis. In the present study it was most abundant on Tripleurospermum inodorum and Anthemis bornmuelleri, from which adults emerged at the same time as those of O. hungaricus and O. anthemidis, respectively. The association of O. longicollis with the several Tanacetum spp. that are mentioned above is based on examination of material reared by HF Barnes [10] and housed at the BMNH, material received from the private collection of E. Dijkstra, and on Kieffer's original description of O. tanaceti.
Remarks. Rondani [7] described Ozirhincus longicollis without a host plant association, but mentioned that the proboscis is long, tapered, and held perpendicular to the head. He named the genus after this character, and the species after the long neck. Loew [26] later described Clinorhyncha chrysanthemi, also noting that the proboscis is very long and is bent under the head against the thorax, but he did not give a reason for its separation from O. longicollis. The two species were later synonymized by Möhn [11]. Two other species synonymized by Möhn [11] under O. longicollis in the same work are Clinorrhyncha leucanthemi and C. crassipes. The types of these species are lost, but information in subsequent publications [26], [9], [10], as well as our own findings support Möhn's synonymy. Winnertz [28] described C. crassipes without an associated host but mentioned that it had a long proboscis and 10 antennal flagellomeres in the male. This description places it clearly in O. longicollis. Kieffer [29] described C. leucanthemi from Leucanthemum vulgare (as Chrysanthemum leucanthemum) without any further information, but detailed information was given in Barnes et al. [10], who reared it from the same host plant. Barnes et al. show the long proboscis and long fourth segment of the palp, and state that the species has 10 antennal flagellomeres. Again, this combination of characters fits only with O. longicollis.
Finally, O. tanaceti, described by Kieffer [30] from Tanacetum vulgare, is also synonymized here under O. longicollis. As in the above-mentioned cases, the type series of O. tanaceti is considered lost, hence our decision is based on Kieffer's original description and our own findings. In that description, Kieffer emphasized the fact that all 14 specimens he examined had 10 antennal flagellomeres, a character that fits only O. longicollis, based on our examination of 189 specimens from T. vulgare from the UK, Germany, Austria, Czech Republic, and the Netherlands. A single female we reared from T. vulgare, which had 10 antennal flagellomeres, was found to be O. longicollis based on DNA sequencing, corroborating our conclusion that Kieffer's specimens, upon which he based the description of O. tanaceti, actually constituted a series of O. longicollis individuals. This finding also confirms our conclusion that flagellomere number is a reliable character in Ozirhincus.
In the present study we found that O. longicollis often occurs together with O. hungaricus on Tripleurospermum inodorum. In this scenario, one can tell the two species apart by the differ- Characters as in O. anthemidis except for the following: Host plants. Achillea millefolium, Achillea ptarmica. Adult. Head (Fig 3C): Occiput extending forward at expense of eye area in frontal view; eye bridge 2-facets long. Palpus: segment four usually 1.2-1.7 times longer than segment three. Frontoclypeal membrane greatly extended, with many strong setae on each side. Labrum long triangular: at least 4 times as long as width at base, tapering and strongly setose on apical quarter. Labella at least 4 times as long as wide, somewhat concave medially, tapered apically, with several strong setae towards apex and densely setose along entire medial surface. Antenna ( Fig  5F): number of flagellomeres 8, occasionally 9 in both sexes (n = 49♀, 32♂); number occasionally differs between antennae of same individual; flagellomeres cylindrical, barrel-shaped, about 1.3 times as long as wide in female, 1.7 times as long as wide in male; apical flagellomere slightly tapered.
Thorax: Brownish-orange, covered by black and white scales creating three longitudinal black stripes along dorsum separated by thinner white stripes; pleura with white scales. Anepimeron with a group of 5-18 setae. Empodia about as long as bend in claw. Wing length: 1.06-1.44 mm in males (n = 17), 1.14-1.44 mm in females (n = 21). Halter white.
Female abdomen: General color brownish-orange covered by a mixture of reddish-brown scales speckled with black scales and narrow transverse strip of white scales along posterior margin. Sclerites of tergite 8 ( Fig 7F) strongly pigmented, with wide anterior area narrowing gradually towards posterior, with 3-7 setae on widened posterior section. Eighth tergite 1.46-2.16 times as long as seventh tergite (n = 21). Ovipositor 5.05-7.04 times as long as eighth tergite (n = 18). Dorso-lateral plate on cercal segment with 10-14 strong, straight setae.
Larva (third instar). Antennae about twice as long as wide. Spatula (Fig 9G) bidentate, with narrow gap between teeth; shaft pigmentation evanescent posteriorly. On each side of spatula 4 asetose lateral papillae, one of which farther laterally than others. Eighth abdominal segment with 2 setose papillae on each side.
Pupa (Fig 4F). Antennal bases form short bi-dentate horns; lateral lobe slightly longer than median lobe; both tapered, pointed ventroapically. Horns widely separated on vertex by horizontal ridge.
Distribution. Widespread in Europe to Siberia. Introduced into and presently widespread in North America.
Material examined. The type series of Ozirhincus millefolii (Wachtl) could not be found in the Natural History Museum in Vienna (NHMW), where it was supposedly deposited (Peter Sehnal, pers. comm.) and is considered lost. We therefore designate a Neotype for it as follows: NEOTYPE: ♀, Germany, NRW, Ägedienberg, 11.viii.2011, N. Dorchin, reared from Achillea millefolium. The neotype is mounted on a permanent microscope slide in euparal and deposited in TAUI.
OTHER Biology. The life history of this species is similar to that of O. longicollis. Its main host plant is Achillea millefolium, from which it can be reared in great numbers in July-August in Europe and the USA.
Remarks. This species was described in detail by Wachtl [31] from Austria, with comparison to Clinorrhyncha chrysanthemi, but not to Ozirhincus longicollis, which was placed in a different genus at the time. It was probably introduced into North America with the seeds of its main host plant, Achillea millefolium, during colonial times [4], and was first recognized there by Felt [32], who described it as Clinorhyncha filicis, and again in 1908 as C. kanerensis [33]. These two species were later synonymized under O. millefolii by Gagné [5].
Ozirhincus millefolii is usually smaller than other Ozirhincus spp., the scale covering on its thorax and abdomen is largely reddish-brown rather than strictly black-and-white, and it is easily distinguishable from the remaining three species by having only 8-9 antennal flagellomeres, and by the position of the antennal horns in the pupa, which are widely splayed. Möhn [12] listed Santolina chamaecyparissus as a host plant based on larvae he found in dried herbarium material from France but this plant genus was not verified as a host in the present study.
Ozirhincus trichatus Möhn was described from Achillea fragrantissima based on three larvae, which were extracted from herbarium material originally collected in Syria, near Damascus [12]. Achillea fragrantissima is found along the eastern Mediterranean region, from Syria, through Jordan and Israel to Egypt, as well as in the Arabian Peninsula. It is a common and widespread plant in Israel, and we have sampled it many times at different localities and dates in an effort to rear Ozirhincus from it, without success. We therefore suspect that the identification of the host plant might have been erroneous. Because morphological characters of the larvae are generally unreliable for distinguishing among Ozirhincus species, it is currently impossible to verify that O. trichatus is a valid species rather than a synonym of O. millefolii, which has been recorded by Möhn [12] from numerous Achillea species. In view of these facts, we make O. trichatus a nomen dubium until adults are reared from A. fragrantissima that can be identified to species with certainty.

Discussion
Species delimitation and host-plant ranges in Ozirhincus have been unclear for over a century, mainly due to complex host associations and the fact that the only revision of the genus [11], [12] was based on the mostly uninformative larvae. As noted by Barnes et al. [10], resolving the taxonomic issues in Ozirhincus necessitated a thorough comparative study of specimens from a wide array of host plants. In the present work we combined such a study with molecular methods in order to infer the phylogeny of the genus and provide, for the first time, discrete characters for recognizing the species within it. We found that Ozirhincus includes four clearly defined species that can be distinguished from each other based on morphological characters of adults and pupae. Our molecular data supported the morphological findings and corroborated our taxonomic conclusions. It is noteworthy that the most useful morphological character we found is the number of antennal flagellomeres, which is generally considered an unreliable character in the supertribe Lasiopteridi (to which Ozirhincus belongs) [22]. While in other Lasiopteridi flagellomere numbers may vary among individuals within the same species (e.g., [6], [34][35][36][37][38]), we showed that in Ozirhincus flagellomere numbers are consistent within a species and are diagnostic.
The most conspicuous morphological character of Ozirhincus is the elongate proboscis, although we do not know what function it serves. While it is assumed that the short-lived adults of most cecidomyiids do not feed [4], some are known to consume nectar or pollen from flowers, and may effect pollination (e.g., [39], [40]). This is in contrast to cecidomyiid species that effect pollination by using flowers as oviposition sites rather than a food source for the adults (e.g., [41][42][43]). In the case of Ozirhincus, because the females spend some time on the inflorescences when ovipositing into flowers, they end up carrying some pollen grains on their body, and possibly contribute towards pollination, but we do not know if they actually feed on the flowers by inserting their proboscis into them. If the adults do feed on nectar or pollen, one might expect that the length of their proboscis will correlate with the length of the flowers of their hosts, but this is not the case. Ozirhincus anthemidis, whose main hosts are Chrysanthemum species, has the shortest proboscis in the genus, whereas O. millefolii and O. longicollis have much longer proboscises, yet they are associated with the shorter flowers of Achillea and Anthemis species. Determining if and how Ozirhincus adults use the flowers of their host plants as a food source requires more study.
Ozirhincus, which evolved for larval development in achenes of Anthemideae host plants, can be regarded as an offshoot of the closely related genus Lasioptera. Möhn [12] argued that a short proboscis, as seen in O. anthemidis, is the ancestral state in the genus, because it represents a transitional state between this genus and Lasioptera. The molecular analysis conducted in the present study, while clearly supporting the validity of four species in Ozirhincus, was unable to resolve the phylogenetic relationships among them, and ancestral states reconstruction does not offer insight into the evolution of proboscis size. Nevertheless, a position of the short-snouted O. anthemidis at the base of the tree would be the most parsimonious inference, with the very long snout of O. longicollis representing a derived state. Morphologically, Ozirhincus is much closer to Lasioptera than to other Lasiopterini, as has also been confirmed by our molecular data. However, it is not clear if Ozirhincus evolved from Asteraceae-feeding species within the biologically diverse Lasioptera (currently with 130 species) because most of the relevant Lasioptera species are associated with different Asteraceae tribes. Of the six Lasioptera species that develop in Anthemideae, five are associated with Artemisia, which is not used by Ozirhincus [1]. The sixth species, L. francoisi Kieffer, was recorded from Achillea millefolium but we did not rear it in the present study despite repeated collections of this host plant. Further large-scale sampling of Lasioptera species was beyond the scope of the present work. The fact that Ozirhincus is limited to host plants of the tribe Anthemideae is of interest given that other cecidomyiid genera, such as the large genus Rhopalomyia Rübsaamen, also appear to prefer this tribe [1]. However, within the Anthemideae, no Ozirhincus species has ever been reared from Artemisia, a genus that hosts more than 170 Rhopalomyia species [44], [1]. One of the main goals achieved in the present study, was to clarify the complex pattern of host use in Ozirhincus, which had been a major cause for confusion and uncertainty in the taxonomy of this genus. With the exception of O. millefolii, each Ozirhincus species is now known to use several host-plant species in more than one genus, and each host-plant genus other than Achillea is known to support more than one Ozirhincus species. This means that a single sample of some plants can yield two Ozirhincus species simultaneously, but in such cases the key and descriptions we provided here make it easy to recognize the species.
This work shows that Ozirhincus species are oligophagous, similar to some species of Lasioptera (e.g., L. buhri Möhn, L. carophila Löw), but in stark contrast to the mostly monophagous species in the subtribe Baldratiina (e.g., Baldratia, Careopalpis, and Stefaniola) ([6] and Dorchin, unpubl. data). The extent of oligophagy within Ozirhincus varies, with species currently known from between 1-4 genera and 3-12 species of host plants. In those species that exhibit the wider host ranges (O. longicollis and O. anthemidis), we did not find any evidence for host-race formation, as individuals from different host plants were intermixed in our phylogenetic tree. It is noteworthy that the two species with the smaller number of host species are limited to Europe and the Russian Far East, whereas those with a larger number of host species extend all the way to the southern Mediterranean region. Not surprisingly, this observation suggests that having a greater number of host species contributes toward a wider distribution range.
Gagné, Junichi Yukawa, and Chris Borkent provided helpful comments on an earlier version of this manuscript. This study was supported by the Israeli Taxonomy Initiative.