Improving knowledge of the subgenus Agrodiaetus ( Lepidoptera : Lycaenidae : Polyommatus ) in Eastern Europe : Overview of the Romanian fauna

The butterfl y subgenus Agrodiaetus of the genus Polyommatus (Lepidoptera: Lycaenidae) is distributed in the western and central Palaearctic and represents a taxonomically challenging group due to its rapid diversifi cation coupled, in many cases, with very limited availability of morphological diagnostic characters. In this study we provide a detailed overview of this subgenus in the Romanian fauna, a country where scattered, poorly documented records suggest the presence of three species: Polyommatus (Agrodiaetus) damon, P. (A.) admetus and P. (A.) ripartii. By analyzing material from museum collections and published records, and combining them with new faunistic fi ndings and DNA data, we show that only P. admetus is currently undoubtedly present in the Romanian fauna (northern Dobrogea), where it reaches its north-eastern range limit in the Balkans. Historical records and DNA data suggest that the occurrence of P. admetus in northern Dobrogea is not likely to refl ect a recent range expansion caused by climate or other environmental changes. Several historical records of this species that suggest its much wider distribution in Romania represent confusion with P. damon and Phengaris alcon. Polyommatus damon, reported mainly from western Romania, lacks records after 1938, while P. ripartii is only known from a single male labelled as originating from the Danube Delta, and requiring confi rmation. There is a great need for directed studies to clarify the status of P. damon and P. ripartii, which are scarce and declining north of the Balkans and may represent taxa of conservation concern.

We sequenced 655 bp of COI for 39 Polyommatus (Agrodiaetus) specimens, seven of which originated from Romania, as well as a 164 bp COI fragment for one Romanian P. admetus (Table 2).For 38 specimens, total genomic DNA was extracted using Chelex 100 resin, 100-200 mesh, sodium form (Biorad), under the following protocol: one leg was removed and introduced into 100 μl of Chelex 10% and 5 μl of Proteinase K (20 mg/ml) were added.The samples were incubated overnight at 55°C and were subsequently incubated at 100°C for 15 min.Samples were then centrifuged for 10 s at 3000 rpm.A 655-bp fragment near the 5´ end of COI was amplifi ed by polymerase chain reaction using the primers LepF1b (a slightly modifi ed version of LepF1) and LepR1 (Table 3).Double-stranded DNA was amplifi ed in 25-μl volume reactions containing: 14.4 μl autoclaved Milli-Q water, 5 μl 5 × buffer, 2 μl 25 mM MgCl 2 , 0.5 μl 10 mM dNTPs, 0.5 μl of each primer (10 μM), 0.1 μl Taq DNA Polymerase (Promega, 5U/μl) and 2 μl of extracted DNA.The typical thermal cycling profi le followed this protocol: fi rst denaturation at 92°C for 60 s, followed by fi ve cycles of 92°C for 15 s, 48°C for 45 s and 62°C for 150 s, and then by 35 cycles of 92°C for 15 s, 52°C for 45 s and 62°C for 150 s and a fi nal extension at 62°C for 420 s.PCR products were purifi ed and sequenced by Macrogen Inc.
Two of the sequences obtained in this study were recovered from specimens that were over 30 years old and were analyzed at the Biodiversity Institute of Ontario, Canada.In one case (sample RVcoll14U820), a full DNA barcode (658-bp) was obtained combining amplicons obtained using the primers [LepF1 + MLepR2] + [MLepF1 + LepR1].In the other case (sample RVcoll14U819), a 164-bp amplicon was obtained using the primer set C_micro-LepF1_t1 + C_TypeR1 (Table 3).For these two samples, standard DNA extraction procedures were used (Ivanova et al., 2006), and DNA sequencing was performed on an ABI 3730xL capillary sequencer (Applied Biosystems).
The new sequences over 200 bp obtained in this study have been submitted to GenBank (see Table 2 for accession numbers) and all sequences are also publicly available in the dataset DS-ADMRIP (dx.doi.org/10.5883/DS-ADMRIP)from the Barcode of Life Data Systems (http://www.boldsystems.org/).

Analyses of DNA sequences
Prior to phylogenetic analysis, we used TCS 1.21 (Clement et al., 2000) to remove duplicated COI haplotypes of P. admetus present in the 114 sequence alignment.The short COI sequence (164 bp) from sample RVcoll14U819 was not subjected to this fi ltering.The 42 COI sequences available for P. admetus were collapsed to 23 unique haplotypes that, in addition to phylogenetic analysis, were also used to construct a maximum parsimony haplotype network using TCS 1.21, with a 95% connection limit.The network presented one loop, which was broken according to frequency and geographic criteria (Excoffi er & Langaney, 1989).The fi ve sequences of P. yeranyani, formerly attributed to P. admetus, have also been collapsed to four unique haplotypes.
Following the removal of duplicated haplotypes, the alignment used for phylogenetic analysis included 94 COI and COII concatenated sequences, and was 2174-bp long.Beneš et al., 2002;Tshikolovets, 2011;Przybyłowicz et al., 2014;Kudrna et al., 2015).P. damon even occurs very locally as far north as the Luga district in the Leningrad region of Russia (Matov & Ivanov, 1999).Recent studies (e.g., Vila et al., 2010;Dincă et al., 2013) have improved knowledge of taxon composition, distribution and genetic structure for several Agrodiaetus species in Europe, but the subgenus remains insuffi ciently documented in the Balkans, despite this being apparently one of the richest European regions for Agrodiaetus taxa (Vila et al., 2010;Kudrna et al., 2015;Vishnevskaya et al., 2016).
Species of Agrodiaetus are poorly known in Romania where scattered, poorly documented records suggest the presence of as many as three species: P. admetus, P. damon and P. ripartii (Székely, 2008;Rákosy, 2013;Rákosy & Török, 2013).In the red list of Romanian butterfl ies (Rákosy, 2002), P. admetus and P. ripartii were listed as data deficient, while P. damon was placed as critically endangered.
In fact, a recent study (Rákosy & Török, 2013) noted that P. ripartii and P. admetus are only represented in museum collections by a single and four specimens from Romania respectively, none collected after 1980.Improving knowledge of Agrodiaetus in Romania would provide data of potential conservation value because its component taxa are very local north of the Balkans, the region for which southern Romania represents the northern limit.
In this study, we combine historical data from museum collections with new faunistic fi ndings and DNA sequences to provide an overview regarding Agrodiaetus in Romania.

Faunistic data
Historical data for P. damon, P. ripartii and P. admetus were gathered by screening all published records and by checking material in collections stored at the Museum of Natural History "Grigore Antipa" Bucharest (MGAB), the Brukenthal National Museum Sibiu (MBSR), the Hungarian Natural History Museum Budapest (HNHM), the Municipal Museum Mediaș (MMM), and the Székely National Museum Sfântu Gheorghe (MOSG).
Field work directed towards the rediscovery of P. admetus in south-eastern Romania (Dobrogea) was done by some of the authors between the end of June and the end of July 2006, 2007, 2014and 2015 (Table 1).

DNA sequencing
The presence in museum collections of two Romanian specimens requiring confi rmation (one P. admetus and one P. ripartii), prompted us to analyse DNA data of several specimens of the above mentioned species.To place the genetic data obtained from Romanian specimens in a broader context, we used sequence records for 114 Polyommatus specimens including 74 mined from GenBank which included sequences for P. ripartii and P. admetus that overlap at least 400 base pairs (bp) of the mitochondrial DNA fragment sequenced in this study.More precisely, we used a part of the cytochrome c oxidase subunit I (COI) and subunit II (COII) dataset from Dincă et al. (2013) and references therein: we selected only sequences representative for each unique haplotype of P. ripartii and therefore retained 72 of the 112 sequences in the original dataset.To this we added two COI sequences of P. admetus from Lukhtanov et al. (2015), as well as 39 COI sequences of P. admetus and one of P. ripartii obtained in this study (Table 2).(1900) and Salay (1910) Only literature data Doubtful record.Caradja (1895) used data from L.
Cosmovici, but mentioned that the specimen is not present in coll.Kemmingers Phylogenetic relationships were inferred using Bayesian inference (BI) through the CIPRES Science Gateway (Miller et al., 2010).Both BI analyses and the estimation of node ages were run in BEAST 1.8.0 (Drummond & Rambaut, 2007) with the data set partitioned by gene.Substitution models for each partition were chosen according to the Akaike's information criterion (AIC) val-ues obtained in JMODELTEST 2.1.3(Darriba et al., 2012), and were GTR + I + G for COI and GTR + G for COII.Base frequencies were estimated, six gamma rate categories were selected and a randomly generated initial tree was used.As previous studies have shown that P. damon is the most basal species of Agrodiaetus (Kandul et al., 2004;Wiemers et al., 2009;Talavera et al.,  2013), we enforced the monophyly of the P. dolus + P. admetus clade so that P. damon is recovered as sister to the rest.Rough estimates of node ages were obtained by applying two molecular clocks with: 1.5% uncorrected pairwise distance per million years estimated for various invertebrates (Quek et al., 2004), and 2.3% estimated for the entire mitochondrial genome of several arthropods (Brower, 1994).A strict clock and a normal prior distribution was used, centred on the mean between the two substitution rates, and the standard deviation was tuned so that the 95% confi dence interval of the posterior density coincided with the 1.5% and 2.3% rates, respectively.
Parameters were estimated using two independent runs of 20 million generations each, and convergence was checked using the program TRACER 1.6.

Polyommatus (Agrodiaetus) damon
Ten specimens (six males and four females), collected between 1896 and 1938, were found in museum collections (Fig. 1a-c, Table 1).The most recent Romanian record was a male collected at Nădrag (Timiș county, Banat region), on June 21, 1938 and stored in the collection of the HNHM (Fig. 1b, Table 1).A series of studies mentioned the occurrence of this species in several areas (mainly from western Romania), but without precise locality data and/or numbers of specimens (Table 1).These studies include records from as early as 1853 (Caradja, 1895) to 1917 (Czekelius, 1917), although the limit of the latest dates is hard to establish due to lack of exact data in the publications.A specimen of P. damon from Retezat Mountains, present in the L. Diószeghy collection at the HNHM, lacked a collection date (Table 1), but it was not likely collected later than 1940 (probably several years earlier) (Căpușe & Kovács, 1987).
The distribution of P. damon in Romania based on the available data of acceptable precision, indicates that the species was relatively widespread in the western and south-western parts of the country, but probably very local (Table 1, Fig. 2).

Polyommatus (Agrodiaetus) ripartii
Our survey of museum collections identifi ed a single male specimen of P. ripartii from Romania stored in the collection F. König at the MMM and also reported by Rákosy & Török (2013).This specimen was labelled as having been collected in the Danube Delta, on the 30th of June 1980 (Table 1, Figs 2, 3).In the same collection, there is also a specimen of P. admetus bearing identical collection data to that for P. ripartii (Table 1).

DNA results for P. ripartii
The Bayesian analysis did not recover P. ripartii as monophyletic (Fig. 4).This result is concordant with previous studies (e.g., Vila et al., 2010;Dincă et al., 2013;Vishnevskaya et al., 2016) and highlights the need for further study.However, there was good support for the three main European lineages of P. ripartii reported by previ-  ous studies (Dincă et al., 2013;Vishnevskaya et al., 2016).
The single specimen known from Romania (Danube Delta) (Fig. 3), which has been successfully DNA barcoded, was placed within the Eurasian lineage of P. ripartii (Fig. 4).

Polyommatus (Agrodiaetus) admetus
The fi rst published records of this species in Romania date from 1865 and refer exclusively to northern Dobrogea (south-eastern Romania): two males from Ciucurova (Mann, 1866) and one male from Babadag (Fiebig, 1927) (Table 1, Fig. 5).In addition, our survey of material in museum collections located three more recently collected (1980) specimens from northern Dobrogea (Horia) (Fig. 6a), and one male specimen from the Danube Delta (also 1980), all stored in the coll.F. König at the MMM (Table 1, Fig. 5).These specimens have also been reported by Rákosy & Török (2013).
A male specimen of P. admetus, labelled as originating from the Bucura Lake in the Retezat Mountains was found in the collections of the HNHM.A comment by A. Schmidt (former curator of the Lepidoptera collections at the HNHM, see Bálint & Katona, 2014), attached to the pin of the specimen, expressed doubts regarding the origin of the specimen (Table 1).Two female specimens, stored in the coll.A. Ostrogovich at the MGAB, were published as P. admetus by Popescu-Gorj (1964) (Table 1), but our reexamination revealed that they actually represent females of Phengaris alcon (Denis & Schiffermüller, 1775) (Fig. 7).A specimen present in the coll.L. Diószeghy from the MOSG, was published as P. admetus by Căpușe & Kovács (1987), but we found that it actually represents a female of P. damon (Fig. 1c).
Our research in northern Dobrogea revealed the presence of P. admetus at several sites in the Măcin Mountains and also established its persistence in Babadag forest (Table 1, Figs 5, 6b-d).The species was local, but relatively abundant in the Măcin Mountains, while it was apparently much rarer in the southern part of Babadag forest (Table 1).At all sites, the adults were fl ying in forest clearings and along forest and shrub borders always in xeric areas with tall vegetation and little to no grazing pressure (Fig. 8a-c).

DNA results for P. admetus
Forty-two of the 43 DNA sequences from P. admetus analysed, representing well the range of the species (especially in Europe) (Kudrna et al., 2015) (Fig. 9a), were collapsed to 23 unique haplotypes.The six Romanian specimens originating from the recently discovered populations represented two haplotypes (h1 and h5) that were found exclusively in northern Dobrogea (Table 2, Fig. 9a, b).The single Romanian specimen of P. admetus collected in 1980 in the Danube Delta had only a short 164 bp COI sequence that was not used for haplotype inference.However, for the overlapping gene region, this sequence was identical to h3 from north-eastern Greece, southern Bulgaria and Hungary and to h7 from western Bulgaria, but differed from the other two Romanian haplotypes (h1 and h5) by three mutations.
The Bayesian analysis recovered P. admetus as monophyletic with good support (Fig. 4).It is worth noting that, in previous studies (e.g.Vila et al., 2010;Dincă et al., 2013), this species was not recovered as monophyletic due to the presence of a lineage distributed in eastern Turkey and southern Caucasus.However, Vishnevskaya et al. (2016) proposed that this lineage represents a distinct species (P.yeranyani), also recovered here as monophyletic with good support.Within P. admetus, haplotypes h1 and h5 from the Romanian specimens, sampled in this study in Fig. 2. Distribution of P. damon and P. ripartii in Romania.For P. damon, no record is available after 1938, while P. ripartii has only one record (requiring confi rmation) from 1980 (Danube Delta).Further details are available in Table 1.Records of P. damon with vague locality data were not included on the map.The division of the Romanian territory illustrates the main historical regions of the country.
northern Dobrogea (Măcin Mountains and Babadag forest) clustered together with good support and were part of a well-supported clade that included four more haplotypes (h2, h10, h11 and h15) from the central-eastern Balkans and western Turkey (Figs 4,9a).The short COI sequence of the Romanian specimen from the Danube Delta (RV-coll14U819) clustered (albeit with relatively poor support) with haplotypes found in Albania (h14), north-eastern Greece, southern Bulgaria and Hungary (h3), as well as western Bulgaria (h7) (Figs 4,9a).

Polyommatus (Agrodiaetus) damon
All specimens of this species are old with no reliable record from Romania after 1938.It is thus not surprising that its national extinction has been suspected (Székely, 2008;Rákosy, 2013).
This species formerly occurred in the western part of Romania (western Oltenia, Banat, Crișana, Transylvania and Maramureș) (Fig. 2).However, this general distribution is based on very few records (only ten specimens found in collections), many with vague data (e.g., Abafi -Aigner, 1911, reported the species from several large areas without precise information, such as Maramureș, Bihorului Mountains, or the valleys of the Criș rivers).For this reason, the distribution map in Fig. 2 includes only data that could be assigned to a locality with acceptable precision.The single record of P. damon from Moldavia (Comănești) was viewed as doubtful by Caradja (1895), who mentioned that the specimen was lacking from its original collection (Table 1).The labelling of the specimen from Saschiz (Mureș county, eastern Transylvania) (Table 1) may also be incorrect in our opinion; although not fi gured as doubtful in Fig. 2, this locality is quite isolated from reliable records and, while virtually impossible to verify, mislabelling cannot be discounted.
Although the species has been mentioned as frequent in Transylvania and Banat (e.g., Abafi -Aigner et al., 1896), the very low number of specimens in collections and the paucity of records with acceptable precision suggest that P. damon was probably rare in Romania even in the second half of the 19 th and the beginning of the 20 th century.Some recent publications also reported this species from south-eastern Romania (Dobrogea) (Székely, 2008;Rákosy, 2013), but our results do not support this.The core of its distribution in Romania seems to have been the Banat region (Fig. 2, Table 1).It is interesting that, although the most recent data known from Romania originate from Banat (June 1938, Nădrag, coll.Sztankov at the HNHM) (Table 1), records suddenly stopped after this date, although Banat has been relatively intensively studied.For example, F. König studied the Lepidoptera fauna of Banat for a very long time , but never reported this species.
Polyommatus damon also appears to be very localized in countries neighbouring Romania.In Hungary it has always been restricted to a few sites in the north of the country (Bálint et al., 2006) and in recent years only one highly endangered population survives near Budapest (Bálint, 2015).In Serbia, several records originate from the east and south of the country, but most are prior to 1988 (Jakšić, 1988;Miljević & Popović, 2014).One recent record (year 2012) originates from Suva Planina and requires confi rmation (Jakšić, 2014;Popović & Đurić, 2014), while a population of the species has been found in 2016 in southwestern Serbia, close to Sjenica city (Miljević & Popović, 2014).In Bulgaria, P. damon has only been reported from two localities in the south-west (Pirin and Rila Mountains), but the records are more than 50 years old.Moreover, since specimens are apparently lacking from collections, its occurrence requires confi rmation in that country (Abadjiev, 2001).In Ukraine the species is declining and is currently known only from one restricted area in the west (Lviv district) (Nekrutenko & Tshikolovets, 2005;Geryak & Kanarsky, 2006).In this context, the historical distribution of P. damon in (mainly western) Romania (Fig. 2) represents an expected link between the populations in western Ukraine, northern Hungary and Serbia.
In Central Europe, P. damon is also very localized and declining.It was, for example, reported as extinct from Poland (where it occurred locally in the south-east) (Buszko & Masłowski, 2008), as near extinct (critically endangered) in the Czech Republic (Beneš et al., 2002;Šlancarová et al., 2012) and as also declining in Germany (Ebert & Rennwald, 1991;Nässig et al., 2004).In fact, according to Van Swaay et al. (2010), P. damon is near threatened in Europe.
The typical habitat for P. damon consists of xeric grasslands with an abundance of its larval food plant, the sainfoin (Onobrychis spp.).In Central Europe, grazing is suspected as the main cause for the decline of this species (Dolek, 1994;Dolek & Geyer, 2002;Kudrna, 1998).A study on P. damon and P. thersites (Cantener, 1835) performed in the Czech Republic (Šlancarová et al., 2012) found that P. damon is likely to be very sensitive to summer grazing and mowing since females lay their eggs high on the food plants (on sainfoin bracts).The same study also suggested that the occurrence of P. damon may be affected by metapopulation processes, so that only large areas of suitable habitat are suffi cient to maintain the species.
The causes for the apparent decline of P. damon in Romania are uncertain, its butterfl y fauna still needs in depth studies.In fact, recent research has revealed new species for the country or confi rmed older records, even in areas that were considered as well investigated (e.g., Dincă & Vila, 2008;Dincă et al., 2008Dincă et al., , 2009Dincă et al., , 2010Dincă et al., , 2011;;Rákosy & Craioveanu, 2016).On the other hand, it is also possible that the decline, or even extinction, of P. damon refl ects its high sensitivity to habitat alteration (Šlancarová et al., 2012).For example, P. damon is suspected to have been introduced in the Benelux due to such agricultural practices and became extinct when sainfoin crops decreased (Bink, 2013).The availability of the larval food plant in western Romania may have also been higher in the 19 th century, when the sainfoin was grown as fodder and as a bee plant, creating habitat for this species.At the end of the 19 th and beginning of the 20 th century, Onobrychis was also  2. For each unique haplotype of P. admetus the number of specimens and country of occurrence are indicated in parentheses.Haplotype codes correspond to those in Table 2.
frequently used in the Carpathian Basin to cover barren hillsides and keep the ground after heavy works of railway construction.This practice promoted it as one of the most important elements dominating meadows for hay-making (Babai et al., 2014).
Until further research is done, P. damon should be treated in Romania as a data defi cient, but potentially extinct species.

Polyommatus (Agrodiaetus) ripartii
The only specimen of this species known from Romania was collected in 1980 and, according to its label, originates from the Letea forest in the Danube Delta (Figs 2, 3, Table 1), a sylvo-steppe area with sand dunes.To place this record into a broader context, we consider the status of P. ripartii in the countries neighbouring Romania.From Hungary it is known from a single specimen collected near Budapest (Bálint, 1996).The label on this specimen does not mention the year, but it cannot be later than 1917, the year when the collection owner (E.Ulbrich) died.The species is also known from Serbia, some sites being less than 60 km from the south-western Romanian border, but almost 600 km from the Danube Delta (Miljević & Popović, 2014).In Bulgaria, P. ripartii occurs mainly in the west and south-west of the country, with the nearest records to the southern Romanian border being less than 80 km away, but almost 400 km from the Danube Delta (Abadjiev, 2001).In Ukraine, the species is known exclusively from Crimea (Kolev & De Prins, 1995;Nekrutenko & Tshikolovets, 2005), over 350 km from the Danube Delta.The easternmost records of P. ripartii in the Balkans originate from the European part of Turkey, over 400 km south from the Danube Delta (Hesselbarth et al., 1995;Kudrna et al., 2015).The distribution of P. ripartii in the neighbouring countries suggests that the species may be more likely to occur in south-western Romania than in the Danube Delta, but its presence in south-eastern Romania (Dobrogea) cannot be discounted, especially since the area also hosts popula-tions of P. admetus.The habitat of P. ripartii ranges from xerothermic sylvo-steppe-like environments to shrubs and dry mountain slopes (the latter especially in the southern parts of its range), usually in limestone areas with its larval food plant, Onobrychis spp.Polyommatus ripartii is generally not known to occur near sea level in southern Europe, although towards its northern range limit it can fl y at low elevations (e. g.Poland, southern Ural and Crimea) (Anikin, 1993;Kolev & De Prins 1995;Przybyłowicz, 2014).At least some of these habitat types occur in southern Romania.However, the Danube Delta is a surprising location for its occurrence due to both the very low elevation (0.52 m above mean Black Sea level) (Gâștescu, 2009) and the absence of Onobrychis species (Oprea, 2005;Ciocârlan, 2011;Doroftei et al., 2011).
It is interesting to note that F. König, who collected P. admetus in northern Dobrogea (Babadag forest area) without publishing the records, provided apparently contradictory information in private communications with colleagues, by mentioning both P. admetus and P. ripartii from the forests around Babadag (Rákosy & Török, 2013).In F. König's collection stored at the MMM, besides the P. ripartii from the Danube Delta, there is one specimen of P. admetus with the same labelling as P. ripartii.Besides these, there are three specimens of P. admetus from Babadag forest (Horia) captured by F. König.All fi ve specimens were collected in 1980, although the labels for two specimens from the Danu be Delta mention M. Brătășeanu as collector (Table 1).If the labelling is correct, it is unclear why F. König did not mention the presence of at least one Agrodiaetus species in the Danube Delta, not only in Babadag forest.
The COI sequence of P. ripartii from the Danube Delta belongs to the widespread Eurasian lineage of this species (Fig. 4).This eliminates the possible mislabelling involving specimens belonging to lineages that are unlikely to occur in the Danube Delta, such as the West European one or that of taxon P. ripartii paralcestis (Fig. 4).However, the fact that no species of Onobrychis is known from the  1.The division of the Romanian territory illustrates the main historical regions of the country.The right side of the fi gure represents an enlarged portion of northern Dobrogea, from where all reliable records of P. admetus originate.
Danube Delta (see above) raises doubts about the occurrence of P. ripartii (or any other species of Agrodiaetus) in this area.Therefore, a potential mislabelling cannot be fully excluded: the specimens from the Danube Delta could originate from northern Dobrogea, or even from outside the Romanian borders.Polyommatus admetus has already been confi rmed from northern Dobrogea (see below), and that area offers more suitable habitats for P. ripartii compared to the Danube Delta, including the presence of Onobrychis spp.(Oprea, 2005).
Therefore, both the presence of P. ripartii in Romania and the occurrence of a population in the Danube Delta require confi rmation.If proven, the occurrence of P. ripartii in the Danube Delta would represent the northermost locality for this species in the Balkans and one of the most northerly in Europe, reducing the distribution gap between the populations from the Balkans, Crimea (Nekrutenko & Tshikolovets, 2005) and the isolated ones from south-eastern Poland (Przybyłovicz, 2000).Such a population would also be of conservation concern since P. ripartii has been listed as near threatened in the European Union by Van Swaay et al. (2010).

Polyommatus (Agrodiaetus) admetus
The populations found in this study from the Măcin Mountains and Babadag forest, combined with historical records from Babadag (the most recent from 1980) (Mann, 1866;Fiebig, 1927, Rákosy & Török, 2013), confi rm the presence and extend the known range of P. admetus in north ern Dobrogea (Table 1, Figs 5, 6).The data also sug gest that this species is local, but probably more widespread in the Babadag forest and especially in the Măcin Mountains, where suitable habitats are present (Fig. 8a, b) and where the species is relatively common.The single record from the Danube Delta (Table 1, Fig. 5) (Rákosy & Török, 2013), while plausible, requires confi rmation (see below).
Our survey of museum collections revealed that none of the records for P. admetus from outside northern Dobrogea are reliable (Table 1).In fact, all supposed records from Transylvania and Banat represent misidentifi cations of P. damon (Căpușe & Kovács, 1987) (Fig. 1c) and P. alcon (Popescu-Gorj, 1964) (Fig. 7).These erroneous records resulted in overestimation of the real distribution of P. ad- metus in Romania (e.g., Tshikolovets, 2011;Kudrna et al., 2015;Eckweiler & Bozano, 2016).It is likely that published distribution maps indicating that P. admetus occurs in south-western and/or north-western Romania took over the erroneous literature data mentioned above.
In countries neighbouring Romania, P. admetus is local in the northern part of Hungary (Bálint et al., 2006), and a little further to the north it was reported as local and endangered in the Pannonian part of Slovakia (Kulfan & Kulfan, 1992).In Serbia P. admetus occurs mostly in the southern and south-eastern regions (Miljević & Popović, 2014), while in Bulgaria it is relatively widespread in the central, eastern and southern parts (Abadjiev, 2001), and it is also present in nearby areas including the European part of Turkey (Hesselbarth et al., 1995;Kudrna et al., 2015).In Ukraine the presence of P. admetus requires confi rmation (Nekrutenko & Tshikolovets, 2005) because there are only two records: one specimen collected in 1892 from the Lviv region (western Ukraine) (Garbowski, 1892) and another found in 1904 in the Odessa region (southern Ukraine) (Shugurov, 1906).In this context, P. admetus from Dobrogea represents the north-eastern range limit for this species in the Balkans, and the nearest confi rmed populations occur in north-eastern Bulgaria, circa 200 km to the southeast (Nevsha village, N. Shtinkov personal comment to V. Dincă, 2015).
The six analysed specimens of P. admetus from northern Dobrogea represent two unique haplotypes (h1 and h5), that were differentiated by at least one (h1) and two mutations (h5) from the other haplotypes found in the Balkans (Fig. 9b).The Romanian haplotypes also differed by four (h1) and fi ve (h5) mutations from the Hungarian specimen that was analysed.The genetic data, combined with the historical faunistical data available from Romania, suggests that the presence of P. admetus in Romania does not refl ect a recent range expansion potentially related to climate or other environmental change.The uniqueness of these haplotypes also likely refl ects the geographical isolation of the populations of P. admetus from northern Dobrogea and highlights their conservation value.Interestingly, the COI sequence of the single Romanian specimen labelled as from the Danube Delta, although only 164 bp-long, was placed into a different clade compared to the specimens  (Popescu-Gorj, 1964).The specimen on the left was captured at River Zlaști, while the one on the right was collected at Sovata.For further information, see Table 1.Scale bar: 10 mm.Photos by V. Dincă.from northern Dobrogea (Fig. 4) and, despite geographical proximity (Fig. 9a), differed by three mutations from the haplotypes from northern Dobrogea.This result can have three explanations: (1) DNA analysis of additional specimens would reveal the presence of the Danube Delta haplotype in northern Dobrogea as well; (2) the presence of P. admetus in the Danube Delta refl ects a colonization event independent from that involving northern Dobrogea; (3) the specimen from the Danube Delta is mislabelled and actually originates from outside Romanian territory, or at least from outside Dobrogea.Assuming that the fi rst two hypotheses are not correct, a recent colonization of the Danube Delta by P. admetus from northern Dobrogea is unlikely due to the genetic differences found between the two regions.Moreover, as in the case of P. ripartii, the lack of Onobrychis spp.from the Danube Delta raises further doubts regarding the presence of P. admetus in the area so further studies are needed to clarify this situation.
Although historical records of P. admetus from outside northern Dobrogea have been shown to be unreliable by our study, it is possible that the species occurs in other parts of the country, especially in the south-west and west, where potentially suitable habitats are present.Indeed, the nearest populations from Hungary are circa 150 km away from the western Romanian border, while the ones from eastern Serbia and western Bulgaria are less than 70 km away from the south-western Romanian border.It is therefore possible that directed research would reveal populations of P. admetus outside northern Dobrogea.
The populations of P. admetus recently discovered in northern Dobrogea are within protected areas, namely the Măcin Mountains National Park (ROSCI0123) and the Babadag-Codru forest (ROSPA0091).Based on our observations (years 2014-2016), the impact of grazing is low in Măcin Mountains and moderate in Babadag forest so it probably does not severely affect the species and its food plant, Onobrychis spp.However, in Babadag forest, where P. admetus seems rarer than in Măcin Mountains (Table 1), the areas with forest clearings and meadows where the species has been found, appear to be decreasing because of overgrowth by shrubs and trees, including Pinus plantations (Fig. 8c).Without a suitable management plan aiming to maintain the sylvo-steppe character of the area, the species may severely decline and even disappear from its currently known location.

CONCLUSIONS
By critically examining literature and museum collection data and by adding new data, this study provides an overview of Agrodiaetus in the Romanian fauna.Three species belonging to this subgenus have been reported from Romania: P. damon, P. ripartii and P. admetus.
We showed that the most recent record of P. damon from Romania dates back to 1938 and that urgent research is needed to determine whether the species is now extinct.
The single specimen of P. ripartii known from Romania was collected in 1980 and originates from the Danube Delta.DNA sequencing has shown that this specimen belongs to the widespread Eurasian lineage of P. ripartii.However, the lack of Onobrychis spp.(the larval food plant of this butterfl y) in the Danube Delta suggests that the presence of P. ripartii in this area is unlikely and hence the single available record requires confi rmation.Recent fi ndings prove the presence of P. admetus in Romania (northern Dobrogea) by confi rming historical records from Babadag forest and, for the fi rst time, identifying populations in Măcin Mountains.All records of P. admetus from outside Dobrogea (Transylvania, Banat) were found to represent either unreliable labelling or confusion with P. damon and Phengaris alcon.The Romanian populations from northern Dobrogea represent haplotypes that are distinct from all other material analysed and likely do not represent recent colonizations due to climate or other environmental change.The single specimen from the Danube Delta is, unexpectedly, genetically differentiated from the relatively near populations in northern Dobrogea.This result and also the lack of Onobrychis spp. in the Danube Delta, indicate the need to confi rm the presence of P. admetus in this region.
Given the current data, P. admetus certainly occurs in the Romanian fauna, while P. damon may be extinct, and P. ripartii requires confi rmation.

Fig. 4 .
Fig. 4. Bayesian ultrametric tree based on cytochrome c oxidase subunit I (COI) and subunit II (COII) sequences.Bayesian posterior probabilities higher than 0.7 are displayed above recovered nodes.Node bars represent 95% highest posterior density for age estimations.The three main lineages of P. ripartii occurring in Europe as defi ned by Dincă et al. (2013) are indicated by grey vertical bars, and the clade of P. admetus highlighted in grey.Romanian specimens of P. ripartii and P. admetus are in bold green and blue, respectively.With the exception of sample RVcoll14U819, specimens are named using COI GenBank accession numbers.Sample codes, as well as COII GenBank accession numbers, are available in Table2.For each unique haplotype of P. admetus the number of specimens and country of occurrence are indicated in parentheses.Haplotype codes correspond to those in Table2.

Fig. 5 .
Fig. 5. Distribution of P. admetus in Romania.Further details are available in Table1.The division of the Romanian territory illustrates the main historical regions of the country.The right side of the fi gure represents an enlarged portion of northern Dobrogea, from where all reliable records of P. admetus originate.

Fig. 7 .
Fig. 7. Female specimens of Phengaris alcon (dorsal and ventral view for each) stored in the coll.A. Ostrogovich from the MGAB, erroneously attributed to Polyommatus admetus in previous studies(Popescu-Gorj, 1964).The specimen on the left was captured at River Zlaști, while the one on the right was collected at Sovata.For further information, see Table1.Scale bar: 10 mm.Photos by V. Dincă.

Fig. 9 .
Fig. 9. a -geographic distribution of COI haplotypes of P. admetus.The white square corresponds to a 164 bp COI sequence of P. admetus from the Danube Delta (Romania) that, for the overlapping COI region, differed by three mutations from the Romanian haplotypes from northern Dobrogea (h1 and h5).b -maximum parsimony haplotype network based on 42 COI sequences of P. admetus.The circles are scaled to represent the relative frequency of each haplotype in the data set.Each branch represents one point mutational step, and the black dots represent unsampled haplotypes.The Romanian haplotypes are coloured in light blue.

Table 1 .
Synthesis of Polyommatus damon, P. ripartii and P. admetus records from Romania, based on literature surveys and museum collection data, as well as original fi ndings of this study.Abbreviations of museum names follow those indicated in the material and methods section of the study.GPS coordinates are approximate and have been provided where relatively precise locality data were available.

Table 2 .
Specimens used for the DNA analyses.The sample IDs of specimens sequenced in this study are in bold.The sequence of sample RVcoll14U819, which was not submitted to GenBank because it was shorter than 200 bp, has been assigned the BOLD process ID corresponding to its COI sequence.

Table 3 .
Primers used in this study.