Geographical distribution of the cryptic species Agrodiaetus alcestis alcestis, A. alcestis karacetinae and A. demavendi (Lepidoptera: Lycaenidae) revealed by cytogenetic analysis

Agrodiaetus alcestis (Zerny, 1932) and A. demavendi (Pfeiffer, 1938) belong to the “brown” complex of the genus Agrodiaetus Hübner, 1822. This complex includes several cryptic species which are extremely uniform in wing colouration and genitalia structure, but have distinct chromosome numbers. In this paper we analyse karyotypes of A. alcestis karacetinae Lukhtanov et Dantchenko, 2002 and A. demavendi in populations from Iran. We demonstrate that A. alcestis karacetinae and A. demavendi are sympatric in the provinces Esfahan, Lorestan, Hamadan, Kurdestan, Kermanshah, and Markazi. The haploid chromosome number of A. alcestis karacetinae is found to be n=19 in all the populations studied. The karyotype of A. demavendi is not stable. The lowest chromosome numbers n=63-67 is observed in the south of the revealed distribution range (provinces Esfahan and Lorestan). The highest chromosome numbers (n=73-74) is found in Northwestern Iran in provinces Kurdestan and Zanjan. We also confi rm that A. alcestis sensu lato appears as a polyphyletic taxon on the Bayesian phylogenetic tree inferred from the mitochondrial COI barcodes and should be most likely divided in two different species: A. alcestis sensu stricto and A. karacetinae. The new data on occurrence of A. admetus and A. ripartii in Iran are discussed.

De Lesse (1960aLesse ( , 1960b)), who fi rst studied this complex karyologically, showed that species description, species determination and study of species distribution ranges are impossible without karyotype investigation.De Lesse (1960b) mapped distribution of several "brown" species from north and northwest Iran and Turkey.He ascertained that A. alcestis and A. demavendi had variable chromosome numbers (n=19-22 and n=67-74 correspondingly).Further studies (Larsen, 1975;Lukhtanov et al., 1998) showed that populations of A. alcestis can be divided in two groups with different chromosome numbers: western group with n=20-21 (populations of Lebanon and Turkey, except for SE Turkey) and oriental group with n=19 (Iranian populations, SE Turkey).Wiemers (2003;Wiemers et. al., 2009) established that A. alcestis karacetinae Lukhtanov et Dantchenko, 2002 with n=19 and A. alcestis alcestis with n=20-21 have similar nuclear ITS2 sequences but different and most likely independently evolved COI haplotypes indicating possible specifi c distinctness of these two taxa.
In this study we analyzed karyotypes of A. alcestis karacetinae and A. demavendi from different localities of Western and Central Iran in order to reveal the southernmost and the easternmost limits of distribution ranges of these species.We also tested the Wiemers's hypothesis (Wiemers, 2003;Wiemers et al., 2009) about the polyphyly of A. alcestis sensu lato by using molecular phylogenetic methods.

Insects
Population samples of different taxa of the genus Agrodiaetus were collected by V. Lukhtanov, A. Dantchenko and N. Shapoval in Iran in the period of 2002-2009.In most cases GPS localities data were fi xed (Table 1).
When collecting in the fi eld, we used a protocol that allowed us to obtain molecular and chromosomal information from the same individual specimen (Bulatova et al., 2009).Fresh (not worn) adult males were used to investigate the karyotypes.After capturing a butterfl y in the fi eld, it was placed in a glassine envelope for 1-2 hours to keep it alive until we processed it.Testes were removed from the abdomen and placed into a small 0.5 ml vial with a freshly prepared fi xative (ethanol and glacial acetic acid, 3:1).Then each wing was carefully removed from the body using two sets of forceps: (i) a coarse or "fl attened" set to hold the body and (ii) a much fi ner set to pinch off the wings.The wingless body was placed into a plastic, 2 ml vial with pure 100% ethanol.Each vial with ethanol has already been numbered.This ID number was also used to label a vial with the fi xative and a glassine envelope in which the wings are preserved.Thus, each specimen was individually fi xed.After the fi xation we had three components collected for each butterfl y, each of which was identifi ed by a common ID number: (a) a vial containing the butterfl y testes (for karyotype analysis), (b) a vial containing the butterfl y wingless body (for DNA analysis) and (c)

Chromosome preparation and karyotyping
Testes were stored in the fi xative for 1-12 months at +4°C.Then the gonads were stained in 2% acetic orcein for 30-60 days at +18-20°C.Different stages of male meiosis were examined by using a light microscope Jenaval, Carl Zeiss and photographed by Nikon Coolpix 4500.We have used an original two-phase method of chromosome analysis (Lukhtanov, Dantchenko, 2002a;Lukhtanov et al., 2006Lukhtanov et al., , 2008)).

Sequence analysis and phylogeny inference
For molecular phylogenetical analysis we used COI barcodes (658-bp 5' segments of mitochondrial cytochrome oxidase subunit I) from 2 specimens of A. alcestis alcestis, 4 specimens of A. alcestis karacetinae and 30 other representatives of the A. alcestis-A.dolus clade.This fragment was selected as it was available from Genbank for almost all taxa of the "brown" complex, and its effectiveness for solving species-level taxonomical problems in butterfl ies was previously demonstrated (Wiemers, 2003;Hebert et al., 2004;Lukhtanov et al., 2009).
Neighbour-joining (NJ) analysis was performed using Kimura's two-parameter Comp.Cytogenet., 2010 4(1) Comparative Cytogenetics model of base substitution as implemented in MEGA4 (Tamura et al., 2007).All positions containing missing data were eliminated only in pairwise sequence comparisons (Pairwise deletion option).Maximum parsimony (MP) analysis was performed using a heuristic search as implemented in MEGA4 (Tamura et al., 2007).A heuristic search was carried out using the close-neighbour-interchange algorithm with search level 3 (Nei, Kumar, 2000) in which the initial trees were obtained with the random addition of sequences (10 replicates).We used nonparametric bootstrap values (Felsenstein, 1985) to estimate branch support on the recovered tree.The bootstrap consensus trees were inferred from 1000 replicates by MEGA4 software for both NJ and MP analyses.
Bayesian analyses were performed using the program MrBayes 3.1.2(Huelsenbeck, Ronquist, 2001;Ronquist, Huelsenbeck, 2003).A GTR substitution model with gammadistributed rate variation across sites and a proportion of invariable sites was specifi ed before running the program for 5,000,000 generations with default settings.The fi rst 1250 trees (out of 5000) were discarded as a burn-in prior to computing a consensus phylogeny and posterior probabilities.

Karyotypes
A. alcestis karacetinae (Fig. 1, a) The haploid chromosome number n=19 was found in MI and MII cells of twenty one studied individuals.In MI cells, all bivalents formed a gradient size row.The karyotype contained no exceptionally large or small bivalents.
A. demavendi (Fig. 1, b) In most cases the chromosome numbers were only approximately established.They are similar in several examined populations (Table 1).The karyotype contains 2 large and 2 medium-sized bivalents.All other bivalents are relatively small and form a gradient series in MI.
A. admetus (Fig. 1, c) The haploid chromosome number n=77 was found in MI cells of two studied individuals.In the specimen E456 the number n=80 was found.In MI cells, the karyotype contains one large and three medium-sized bivalents.All other bivalents are relatively small and form a gradient series in MI cells.
A. ripartii Freyer, 1830 (Fig. 1, d) The haploid chromosome number n=ca.89 was found in MI cell of the single studied specimen.The count was done with approximation due to the overlapping of some chromosomes.In MI cells, the karyotype contains one large and one medium-sized bivalents.All other bivalents are relatively small and form a gradient series in MI cells.

Phylogenetic analysis of molecular data
We have analyzed 43 (including outgroup) COI barcode sequences.The fi nal data set alignment included 690 sites, 106 sites were variable, and 71 sites were parsimonyinformative.
The average nucleotide frequencies were 0.329 (A), 0.367 (T), 0.155 (C), and 0.148 (G).The test of the homogeneity of substitution patterns between sequences did not reject the null hypothesis that the sequences have evolved with the The Bayesian and NJ phylogenetic analyses support monophyly of A. alcestis karacetinae with n=19, however statistical support for this clade was relatively low.On the Bayesian tree A. alcestis karacetinae appeared as a taxon closely related to A. dantchenkoi Lukhtanov et Wiemers, 2003, not to A. alcestis alcestis as expected (Fig. 4). A. alcestis alcestis with n=21, 21 did not appear as monophyletic group on the MP and Bayesian trees; it appeared as monophyletic group only on the NJ tree but with low bootstrap support.The phylogenetic relationships between A. alcestis karacetinae and A. alcestis alcestis were not resolved on the NJ and MP trees.At the same time, on the Bayesian tree A. alcestis sensu lato (A.alcestis karacetinae + A. alcestis alcestis) appeared as a clearly polyphyletic taxon (Fig. 4).

DISCUSSION
We found that in Esfahan, Lorestan, Hamadan, Kurdestan, Kermanshah, and   Markazi provinces A. alcestis karacetinae and A. demavendi were sympatric in their distribution (Fig. 5).In all these localities imago of both species fl ow together: syntopically and synchronously.The stable chromosome number n=19 was found in all the studied populations of A. alcestis karacetinae.This chromosome number was also established in other populations from NW Iran (de Lesse, 1960b) and SE Turkey (Lukhtanov et al., 1998;Lukhtanov, Dantchenko, 2002a, 2002b), whereas A. alcestis alcestis from other parts of Turkey and from Lebanon had n=20 or n=21 (de Lesse, 1960b;Larsen, 1975).In populations of A. demavendi chromosome numbers were not stable: there was a tendency towards increasing the chromosome numbers from n=64-67 in the south of revealed distributional area (Esfahan and Lorestan provinces) to n=73-74 in the north (Kurdestan) (Fig. 5-6).Thus, despite the morphological similarity, A. alcestis sensu lato and A. demavendi can be easily distinguished by their karyotypes.Fereydun-Shahr (province Esfahan) was the southernmost locality where A. alcestis karacetinae and A. demavendi were discovered by us.This locality seems to be close to the southernmost limit of entire distribution ranges of these species, as no representatives Comp. Cytogenet., 2010 4(1) Comparative Cytogenetics of the "brown" complex are known from more southern regions (Nazari, 2003).At the same time, this locality seems to be close to the easternmost limit of distribution range of A. alcestis sensu lato (Fig. 6).
In Zanjan (West part), Azerbaijan-e-Gharbi and Ardebil provinces we found specimens with chromosome numbers n=77, n=80 (Fig. 1, c, Table 1).These chromosome numbers as well as the structure of entire karyotype are similar to those known in A. admetus, another representative of the "brown" complex.A. admetus is known from Balkan Peninsula, Turkey, Armenia, and Azerbaijan (Kandul et al., 2007).Although A. admetus was pre-viously mentioned for Iran (see : Carbonell, 2001: 106), this record was not confi rmed by chromosomal or molecular data.Thus, our fi nding seems to represent the fi rst confi rmed evidence for the presence of A. admetus in Iran.
In Azerbaijan-e-Sharqi province we found a specimen with chromosome number n=ca.89 (Fig. 1, d, Table 1).This chromosome number as well as the karyotype structure is similar to those known in A. ripartii (Freyer, 1830) (Lukhtanov, Dantchenko, 2002a, b). A. ripartii was not previously mentioned for Iran, except for A. ripartii eriwanensis Forster, 1960 (Nazari, 2003).However, the latter record was not confi rmed by chromosomal or molecular data.It should be also noted that the taxon A. eriwanensis is not closely related to A. ripartii (see: Figs 2-4).Thus, our fi nding seems to represent the fi rst evidence for the occurrence of A. ripartii in Iran.
Our karyological studies confi rm the conclusion of de Lesse (1960aLesse ( , 1960b)), that species determination within the "brown" complex is impossible without investigation of karyotypes.Preliminary species determinations made by us in the fi eld, were proved to be incorrect in many cases.We found two sorts of errors: (a) specimens recognized as A. alcestis turned out to be A. demavendi or vice versa, (b) specimen recognized as A. alcestis or A. demavendi turned out to be another species (A. admetus, A. ripartii).Wiemers (2003) proposed a hypothesis about non-conspecifi city of A. alcestis karacetinae and A. alcestis alcestis.We tested this hypothesis by analysing the Wiemers' original COI sequences as well as other samples from GenBank representing additional target and outgroup taxa.The analysis of more representative data set generally confi rmed this hypothesis.It is demonstrated that A. alcestis sensu lato Comp.Cytogenet., 2010 4(1)    Comparative Cytogenetics same pattern of substitution.The disparity index indicated no larger differences in base composition biases than expected based on evolutionary divergence between the sequences and by chance alone.The NJ and MP bootstrap consensus trees are shown on the Fig.2and Fig.3correspondingly.Branches corresponding to partitions reproduced in less than 50% bootstrap replicates are collapsed.The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test is shown above the branches.The 50% majority rule consensus tree was recovered from the trees sampled during Bayesian analyses and is shown on the Fig.4.The posterior probability is shown above every branch on the Bayesian tree.

Fig. 2 .
Fig. 2. Bootsrtap consensus NJ tree of the "brown" Agrodiaetus complex inferred from COI barcodes.Bootstrap values >50% are shown above the branches.Haploid chromosome number of A. alcestis are shown after name of a taxon.

Fig. 3 .
Fig. 3. Bootstrap consensus MP tree of the "brown" Agrodiaetus complex inferred from COI barcodes.Bootstrap values >50% are shown above the branches.Haploid chromosome number of A. alcestis are shown after name of a taxon.

Fig. 4 .
Fig. 4. Consensus Bayesian tree of the "brown" Agrodiaetus complex inferred from COI barcodes.Posterior probability values >50% are shown above the branches.Haploid chromosome number of A. alcestis are shown after name of a taxon.

Fig. 5 .
Fig. 5. Distribution map of A. alcestis karacetinae (white circle) and A. demavendi (white square) in Iran with their haploid chromosome numbers (original data).

Table 1 .
List of the studied Agrodiaetus samples with their haploid numbers (n) and locality data.
Comp.Cytogenet., 2010 4(1)Comparative Cytogenetics a glassine envelope containing the wings.The set specimens of the donor butterfl ies (the butterfl y wingless bodies in ethanol and wings in glassine envelopes) are kept in the department of Karyosystematics, Zoological Institute of Russian Academy of Science.