Salix transect of Europe: additional leaf beetle (Chrysomelidae) records and insights from chrysomelid DNA barcoding

Abstract Occurrence patterns of chrysomelid beetles (Coleoptera: Chrysomelidae), associated with willow (Salix spp.) at 42 sites across Europe, have previously been described. The sites form a transect from Greece (lat. 38.8 °N) to arctic Norway (lat. 69.7 °N). This paper reports additional records and the results of DNA sequencing in certain genera. Examination of further collections from the transect has added 13 species in the genera Aphthona, Chrysomela, Cryptocephalus, Epitrix, Galerucella (2 spp.), Gonioctena, Phyllotreta (2 spp.), Pachybrachis (3 spp.) and Syneta. We also report the sequencing of the DNA regions cytochrome oxidase 1 (CO1) and cytochrome B (cytB) for a number of samples in the genera Plagiodera, Chrysomela, Gonioctena, Phratora, Galerucella and Crepidodera. The cytB sequences are the first available for some of these taxa. The DNA barcoding largely confirmed previous identifications but allowed a small number of re-assignments between related species. Most notably, however, it was evident that the southernmost material (Greece and Bulgaria) of specimens, previously treated as Crepidodera aurata sens. lat., belonged to a distinctive molecular cluster. Morphological re-examination revealed these to be C. nigricoxis Allard, 1878. This is an example of how morphotaxonomy and DNA barcoding can work iteratively to refine identification. Our sequences for C. nigricoxis appear to be the first available for this taxon. Finally, there is little geographic structure evident, even in widely dispersed species.


Introduction
Since early pleas were made for the routine incorporation of a molecular component to taxonomy ("DNA barcoding") (Hebert et al. 2003a, Hebert et al. 2003b, Tautz et al. 2003, a large amount of literature has accrued and a very large number of sequences backed by voucher specimens have been deposited in standard databases. It is now well established that, in many animal groups, sequencing mitochondrial cytochrome c oxidase subunit 1 (COI) provides a straightforward way of gaining taxonomic insight. Early concerns about molecular methods being somehow antagonistic to morphological taxonomy have given way to acceptance that molecular and morphological taxonomy are complementary, reciprocally illuminating and iterative processes.
As part of a study of lowland willow communities sampled from south to north across Europe, we have previously investigated the occurrence and abundance patterns of chrysomelid beetles (Coleoptera: Chrysomelidae) associated with Salix species (Canty et al. 2016). In this study, large numbers of individual beetles were processed and it was impossible with available resources to perform large numbers of genitalia dissections. For this reason, a broad morphospecies concept was used, identifying to species largely using external morphology. We have now been able to test some of these morphospecies assignments using DNA barcoding. This paper reports the new insights that this offers. We also take the opportunity to report additional chrysomelid records from the transect following examination of additional collections.

Collecting methods
Chrysomelid beetles were collected from willows (Salix spp.) by the authors ER and DP at all sites, as previously described (Canty et al. 2016). Details of the sites and the method of their selection have been given in previous papers (Cronk et al. 2015;Canty et al. 2016

Specimen examination and analysis
Morphological procedures followed those used in Canty et al. (2016). A selected subset of specimens was chosen for sequencing (Table 2). These included specimens deemed to be potentially problematic in the original identifications and samples from widespread and variable species. DNA was extracted from material preserved in 90% ethanol. Sequences of mitochondrial cytochrome oxidase subunit 1 (COI) and cytochrome B (cytB) were obtained following protocols for DNA extraction, polymerase chain reaction (PCR) and sequencing described in Percy et al. (2018) with additional primers used for COI (LCO1490 and HCO2198; Folmer et al. 1994). As numerous COI sequences are available on GenBank, we were able to align our own sequences with previously published ones (Table  3). Aligned sequences were analysed using neighbour-joining (NJ) with uncorrected (p) distances in PAUP* (Swofford 2003). Bootstrap support was obtained using 1000 replicates. Sequences generated as a result of this study are all deposited in GenBank (accession numbers MN629748 -MN629886) ( Table 2).   Table 3.

Original species ID
GenBank sequences included in the phylogenetic analysis. The sample in bold under Phratora polaris was downloaded from GenBank as P. tibialis.

Taxonomic insights from molecular barcoding
We used DNA sequencing to test and, if necessary, refine our morphospecies assignments made previously (Canty et al. 2016). Generally, the barcoding results confirmed the morphospecies assignments and provide well-supported species clusters (Figs 1, 2). However, the Chrysomelidae barcoding analysis revealed that some specimens were incorrectly assigned in Canty et al. (2016) (Table 2; Fig. 2). These were all due to using broad morphospecies concepts for Phratora vitellinae (Linnaeus, 1758) and Crepidodera aurata Marsham, 1802. In Phratora, three specimens assigned to Phratora vitellinae clustered in the barcoding data with sequences identified on GenBank as P. polaris Schneider, 1886; and one specimen assigned to Phratora vitellinae clustered with GenBank sequences of P. vulgatissima (Linnaeus, 1758). In Crepidodera, two specimens assigned to Crepidodera aurata clustered with GenBank sequences, plus our own sequences, for C. fulvicornis Fabricius, 1792.
In addition, we noted that certain specimens assigned to Crepidodera aurata formed a distinct molecular cluster, distinct from our own C. aurata sequences and from all others downloaded from GenBank. These specimens were the southernmost specimens of our C. aurata from sites 3 and 4 (Greece) and site 7 (Bulgaria). This prompted a morphological reexamination of these samples, including dissections of genitalia and these specimens were identified with C. nigricoxis Allard, 1878 ( Fig. 3; Table 2). The two species are very similar in external morphology and variable (Fig. 3). Nevertheless, the molecular data clearly separates them (Figs 1, 2). Our sequences for C. nigricoxis appear to be the first to be made available for this taxon. Gavrilović and Ćurčić (2013) note that C. nigricoxis is found on Salix alba L. Although we did not distinguish willow species at the point of collection, Salix alba was present at all the sites where we recorded C. nigricoxis (Cronk et al. 2015). DNA analysis (NJ tree) using COI and cytB sequences generated in this study. Node support shown only for nodes ≥ 90% bootstrap support. DNA barcoding analysis using COI sequences generated in this study and from GenBank. Sequences from this study show the site number and those obtained from GenBank are indicated by a black circle (GenBank accessions given in Table 3). Node support shown only for nodes > 90% bootstrap support. Maximum intraspecific divergences are shown (for our transect samples only), estimated using uncorrected (p) distances (see methods).

Phylogeographic patterns
There is little phylogeographic structure evident from the sequence data, even for widely dispersed taxa along the transect. Fig. 2 (COI data) is suggestive of a split in Crepidodera fulvicornis between northern samples (Finland: 31,35,39) in one clade and southern samples (Hungary: 16,Poland: 23,Latvia: 27) in the other (e.g. a zoogeographic boundary around Estonia or the Gulf of Finland), but one sample from Finland (site 33) that only sequenced for cytB ( Fig. 1) clusters with the southern clade. The absence of clear phylogeographic patterns in the chrysomelids is similar to our findings for curculionids (Canty et al. in review), but differs from those found in a hemipteran taxon (the nettle psyllid; Psylloidea, Hemiptera) sampled along the transect in which population structure suggests distinct regional clades (Wonglersak et al. 2017). . Some of these are not generally associated with willows and are probably accidental by-catch (e.g. Galerucella nymphaeae and Galerucella sagittariae). These additional records do not materially change the basic data or conclusions of Canty et al. (2016), but bring the total number of species to 47 (not 34).

Discussion
The barcoding, described here, provides a good example of the value of iterative molecular and morphological processes in taxonomy. In this case, a broad morphospecies concept allowed determination of those species that have the greatest geographic and morphological variation. These could then be targeted for barcoding to determine patterns of molecular variation. In the case of Crepidodera aurata sens. lat., this led to the distinguishing of two divergent molecular clusters. This in turn led to a re-appraisal of the morphology and to the refinement of the concept of C. aurata and the recognition of C. nigricoxis as its apparent replacement (at least in our sampling) in southern Europe (Greece and Balkans). This very small example thus serves to emphasise that morphological and molecular taxonomy, taken together and applied iteratively, are powerful adjuncts.