OPEN A DNA barcode-based survey of wild urban bees in the Loire Valley, France

OPEN A DNA barcode-based survey of wild urban bees in the Loire Valley, France

The current decline of wild bees puts important ecosystem services such as pollination at risk. Both inventory and monitoring programs are needed to understand the causes of wild bee decline. Effective insect monitoring relies on both mass-trapping methods coupled with rapid and accurate identifications. Identifying wild bees using only morphology can be challenging, in particular, specimens from mass-trapped samples which are often in poor condition. We generated DNA barcodes for two thousand nine hundred thirty-one specimens representing one hundred fifty-seven species (one hundred fifty-six named and one unnamed species) and twenty-eight genera. Automated cluster delineation reveals one hundred seventy-two Barcodes Index Numbers. A total of thirty-six species (twenty-two point nine three percent) were found in highly urbanized areas. The majority of specimens, representing ninety-six point one seven percent of the species barcoded form reciprocally exclusive groups, allowing their unambiguous identification. This includes several closely related species notoriously difficult to identify. A total of one hundred thirty-seven species (eighty-seven point two six percent) show a "one-to-one" match between a named species and the BIN assignment. Fourteen species (eight point nine two percent) show deep conspecific lineages with no apparent morphological differentiation. Only two species pairs shared the same BIN making their identification with DNA barcodes alone uncertain. Therefore, our DNA barcoding reference library allows reliable identification by non-experts for the vast majority of wild bee species in the Loire Valley.

Long-term monitoring programs have documented a sharp decline of insects. The loss of insect pollinators is particularly worrying because of its potential negative ecological and economic consequences. Land use change has been shown to be a major factor involved in the loss of worldwide pollinator populations. Indeed, intensive agriculture has led to the loss of ecological niches for a number of pollinator species, to which are added the adverse effects of pesticide uses. Through the expansion of impervious surfaces, urbanization is also associated with pollinator decline, although some urban green areas such as residential and community gardens, if properly managed, can constitute important refuges for wild bees. This factor and the higher temperatures associated with rapid global warming are accelerating the decline of pollinators worldwide. This scenario has fostered the idea of considering urban areas as potential refuges for pollinators. Consequently, a growing number of studies on urban ecology have emerged describing population dynamics of wild bees in urban areas due to their pivotal importance. In addition, citizen science has been successfully shown to be an efficient way of monitoring urban bees. However, the accurate identification at species level for several bee genera requires advanced taxonomic knowledge, which is limited to a few experts or even not available in many countries. This taxonomic impediment is slowly being overcome through the use of traditional DNA barcoding or more recently developed high-throughput DNA barcoding. Indeed, cox one (cytochrome c oxidase subunit one gene) barcodes have been shown to distinguish between bees difficult to identify due to minor morphological differences or even cryptic bee species.

The accuracy of DNA-based identifications depends on the completeness of DNA barcoding reference libraries. However, only a few DNA barcode reference libraries have been developed for wild bee national faunas, including Ireland; Germany, Canada, and Chile as well as some regional faunas. In addition, identification accuracy depends also on the complete characterization of intraspecific variability.

France has a rich wild bee fauna with over nine hundred fifty-five species recorded, for the whole country but relatively few DNA barcodes of French bees have been published. Our study is the first major contribution to establish a DNA barcoding reference library for French wild bees.

We focused our sampling in central France, a region where over one hundred eighty species of bees have been recorded (unpublished data, Christian Cocquempot personal communication, twenty twenty). Here we present two thousand nine hundred thirty-one barcodes of one hundred fifty-seven wild bee species collected at twenty-nine urban and peri-urban sites in three major cities along the Loire Valley.

Results

Taxon sampling and DNA sequencing success. A total of three thousand five hundred thirty-two bee specimens were collected over the two years of survey; three thousand fifty-seven bees were collected with pan-traps and four hundred seventy-five were collected along transects in twenty-nine sites located in three French cities (Tours, Orléans and Blois).

Of the three thousand five hundred thirty-two specimens collected two thousand nine hundred thirty-one were successfully barcoded (eighty-two point nine eight percent). The overall success rate varied depending on the sequencing method, Sanger versus Single-Molecule Real-Time sequencing and the primers used. Out of the two hundred fifty-two samples analyzed by Sanger sequencing, we obtained a higher barcoding success rate using newly designed primers (one hundred thirty-three sequences out of one hundred eighty-eight specimens, i.e. seventy point seven four percent) than with the traditional Folmer primers (fifteen sequences out of sixty-four specimens, i.e. twenty-three point four three percent). We obtained full DNA barcodes for thirteen species whereas six species were represented only by short sequences (greater than or equal to three hundred base pairs and less than five hundred base pairs). Twenty-seven out of one hundred forty-eight sequences were less than five hundred base pairs long. Of the three thousand three hundred fifty specimens processed using PacBio Sequel platform (Pacific Biosciences, Menlo Park, California, U.S.A.) by Single-Molecule Real-Time sequencing, two thousand seven hundred eighty-three specimens (eighty-three point zero seven percent) yielded a barcode sequence greater than three hundred base pairs. Six out of two thousand seven hundred eighty-three were less than five hundred base pairs. No barcodes could be obtained for four hundred ten additional samples and one hundred fifty-seven samples (four point six nine percent) appeared to be cross contaminated. Sequencing failure was not homogeneous across families, and a few species presented greater amplification problems, especially among the genera Andrena, Hylaeus and Dasypoda.

Species identification and BIN assignment. Our integrative analyses combining morphology and DNA barcodes, identified a total of one hundred fifty-seven species out of which one hundred fifty-six are described valid species belonging to six families (Andrenidae, Apidae, Colletidae, Halictidae, Megachilidae, Melittidae) and twenty-eight genera. In addition, we found one specimen (MB00149) of an unnamed Andrena species whose BIN was new to BOLD. Seventy species (forty-four point six percent) found in our study had not been reported before within the available regional species inventories. For instance, our barcode of Andrena avara liturata (Warncke) represent a first record for the studied area, and a new BIN in BOLD (Barcode of Life Data).

Of the one hundred fifty-seven species one hundred fifty-six were assigned to one hundred seventy-two Barcodes Index Numbers, sixteen (nine point three zero percent) of them were new to BOLD. Only one of the one hundred fifty-seven species barcoded, did not have a BIN assigned, since its barcode was shorter than five hundred base pairs.

Out of the one hundred fifty-seven species barcoded from our studied area one hundred thirty-seven (eighty-seven point two six percent) had already been barcoded from Germany.

A total of thirty-six species (twenty-two point nine three percent) were found in highly urbanized areas.

Halictidae accounted for most of the specimens, representing just over sixty percent of the total specimens collected and fourteen point three percent of genera. The most diverse families in terms of number species were Halictidae and Andrenidae, accounting for thirty percent and twenty-four percent of species, respectively. Only five species (three point one eight percent) of Melittidae were observed.

The majority of species, ninety-six point one seven percent, had their own unique BIN or were assigned to several BINs that formed single clades allowing unambiguous identification based on DNA barcodes; forty-eight species were represented by a single record. The number of specimens sequenced per BIN ranged from one to three hundred sixty, for Lasioglossum morio, averaging seventeen specimens per BIN. Sixty-one BINs were represented by a single individual, singletons.

Multiple BINS were associated with fourteen species, with up to three BINS for some, including Andrena helvola, Anthophora plumipes, Lasioglossum laticeps, and L. villosulum.

Two species pairs share the same BIN: Andrena carantonica and A. trimmerana, and Halictus simplex and H. langobardicus. The identification of those four species based on DNA barcode data is therefore uncertain. One species barcoded in our study, Lasioglossum mediterraneum, was new to BOLD but the sequence was too short to have a BIN assigned. In addition, one unidentified individual in the genus Andrena was assigned to a new BIN to BOLD. This new BIN remains without species identification pending collection and analysis of more individuals.

Barcode gap and haplotype diversity. The average genetic distance within species and genera in the dataset were zero point two three percent and fifteen point seven zero percent respectively with a maximum intraspecific distance of nine point eight one percent for Andrena lagopus. Intraspecific barcode divergence averaged zero point two three percent whereas distance to the nearest-neighbour species averaged nine point zero eight percent. When we considered only species with a number of records n greater than or equal to three, intraspecific divergence exceeded two percent in three species; Panurgus calcaratus, Andrena bicolor, and Andrena helvola.

The mean intraspecific distance distribution ranged from zero to three point seven eight percent and overlapped slightly with the distance to the nearest neighbour distribution, three point six four to nineteen point five six percent. However, nearest-neighbour distances were on average more than tenfold higher than maximum intraspecific distances. Lasioglossum villosulum maximum intraspecific distance values were higher than their nearest-neighbour distance. Conversely, K two P distances to the nearest neighbour less than two percent were obtained for two species pairs: Andrena trimmerana/A. carantonica and Halictus langobardicus/H. simplex. Maximum intraspecific divergence was not correlated with the number of individuals sampled per species, R equals zero point one seven four, P equals zero point one one three.

Our data set contained seven hundred eighty-seven unique haplotypes with an average of five haplotypes per species and a maximum of seventy haplotypes for Lasioglossum malachurum. We found variable patterns of haplotype diversity among the species. The number of haplotypes per species ranged from one to seventy, and the values of haplotype diversity from zero to zero point nine five. Andrena flavipes had the highest haplotype diversity value, followed by Lasioglossum sabulosum, Apis mellifera, and Lasioglossum malachurum. When we considered species with more than ten individuals, the number of samples collected per species was highly correlated with the number of haplotypes found, R equals zero point eight nine eight, P equals one point one one four times ten to the negative thirteen, but not with the values of haplotype diversity, R equals zero point two two eight, P equals zero point one eight one.