1. Introduction
Given the unprecedented levels of biodiversity losses currently observed
(Barnosky et al., 2011; Pievani, 2014, uncovering cryptic diversity and
providing testable species hypotheses is an urgent task for taxonomists,
especially for the hyperdiverse group of the insects (Hallmann et al.,
2017; Sánchez-Bayo & Wyckhuys, 2019; Seibold et al., 2019).
Traditionally, species were described by examining variation in
morphological traits (Padial, Miralles, De la Riva, & Vences, 2010).
They were delimited in a way to minimize within-species variation and to
maximise between-species variation in sets of variable characters.
However, morphological taxonomy is often challenged by a lack of
variation between taxa, or conversely by sexual or generational
polymorphisms within species. Both of which will lead to an absence of a
“morphological” gap between species and may result in substantial
levels of cryptic diversity (Karanovic, Djurakic, & Eberhard, 2016). To
complement morphology, DNA barcoding was introduced as a reliable, fast,
and cheap identification method (Brunner, Fleming, & Frey, 2002;
Hebert, Cywinska, Ball, & DeWaard, 2003), and has since been
extensively used not only for specimen identification but also for
species delimitation. For insects, the 5’-region of the cytochrome
oxidase subunit I (COI) gene has quickly become the DNA barcode gold
standard due to the fact that, for many species, it demonstrated only
very limited intra-species variation (i.e. generally below 3%) yet
distinct differentiation between species (e.g., Brunner et al., 2002;
Hebert et al., 2003; Meyer & Paulay, 2005). In combination with
morphological identification, COI-barcoding was shown to be a powerful
tool for species delimitation in bees (Pauly, Noël, Sonet, Notton, &
Boevé, 2019; Praz, Müller, & Genoud, 2019; Schmidt, Schmid-Egger,
Morinière, Haszprunar, & Hebert, 2015).
There are, nevertheless, numerous examples where COI-barcoding leads to
an erroneous signal. A number of possible reasons for such problematic
barcoding results have recently emerged. For example, a growing body of
literature is reporting that mitochondrial inheritance is more
complicated than initially thought, with rare cases of paternal leakage,
heteroplasmy or recombination (Ladoukakis & Zouros, 2017; White, Wolff,
Pierson, & Gemmell, 2008). Furthermore, mitochondrial genomes can be
subject to evolutionary forces acting solely at the organelle level
[e.g. mitochondrial introgression, Wolbachia infection or
sex-biased asymmetries; (Toews & Brelsford, 2012)]. Although these
events are generally considered rare (but see Klopfstein, Kropf, &
Baur, 2016; Neumeyer, Baur, Guex, & Praz, 2014; Nichols, Jordan, Jamie,
Cant, & Hoffman, 2012), they can considerably skew phylogenies or
biodiversity estimates (Andriollo, Naciri, & Ruedi, 2015; Hinojosa et
al., 2019; Mutanen et al., 2016). Consequently, species delimitation
should rely on multiple sources of information (Carstens, Pelletier,
Reid, & Satler, 2013) and for molecular markers, species delimitation
should use genes of both mitochondrial and nuclear origin (Dupuis, Roe,
& Sperling, 2012).
In contrast to the uncommon suitability of COI as a species marker, the
quest for similarly well-suited, universal and robust nuclear markers
was so far unsuccessful. Several types of nuclear markers have been
explored, but current candidates are all associated with serious
drawbacks. For instance, single-copy nuclear genes (i.e. elongation
factor 1 alpha [EF-1a] or 28S) or multicopy ribosomal DNA markers
(i.e. internal transcribed spacer [ITS]) were explored (Leneveu,
Chichvarkhin, & Wahlberg, 2009; Martinet et al., 2018; Soltani, Bénon,
Alvarez, & Praz, 2017; Williams, Lelej, & Thaochan, 2019). However,
the usefulness of these nuclear markers is often limited by the lack of
phylogenetic resolution (Dellicour & Flot, 2018), and in insects, by
within-genome variation of the multi-copy ribosomal genes (e.g. ITS),
which is a major impediment to the sequencing workflow. For increased
resolution, some studies have used population genetic markers such as
microsatellites. Although microsatellites provide ample resolution for
species delimitation (McKendrick et al., 2017), one major limitation is
that loci are clade-specific and therefore require clade specific
primers that have to be designed base on available genomic information.
Alternative approaches that are slightly more universal and provide
equal or higher resolution include genomic-reduction techniques such as
RAD- or ddRAD sequencing (Lemopoulos et al., 2019). These methods are
very powerful and can provide high-resolution, intraspecific information
on population dynamics. However, as for microsatellites, RAD-seq is
hampered by cost, workload and/or amount of high-quality DNA required.
More importantly, datasets obtained from different studies and/or taxa
are hardly joinable due to the lack of repeatability. This last
limitation of RAD-sequencing is a severe drawback for species
delimitation, since taxonomic work essentially builds upon previous
hypotheses, with new data continuously complementing earlier datasets.
Ideally, molecular species-delimitation should be based on: (i) nuclear
and mitochondrial markers, to reflect gene flow of both nuclear and
mitochondrial genes; (ii) genomic scale for nuclear markers to cover
numerous independent loci; (iii) sufficiently variable to capture
recently diverged species; (iv) repeatable, so that datasets can be
complemented once more material is available; (v) universal to the
extent that datasets can complement each other. In 2012, ultraconserved
elements (UCEs) were introduced as a quick and essentially universal way
to obtain “thousands of genetic markers spanning multiple evolutionary
timescales” (Faircloth et al., 2012). UCEs appear to fulfil many of the
above mentioned requirements for nuclear markers. However, whether they
harbor enough variation to capture divergence among recently diverged
species remains an open question, since by definition they are highly
conserved.
In this study, we examine the use of UCEs for species delimitation in
Central European bees. We include examples of both putative
mitochondrial introgression and of multiple “barcodes” per species,
investigating how UCEs can overcome the main drawbacks for species
delimitation using DNA barcoding developed above. We focused on the
following European species complexes: Andrena
amieti/allosa/bicolor/montana ; Andrena barbareae/cineraria; A.
dorsata/propinqua; A. carantonica/trimmerana/rosae; Lasioglossum
alpigenum/bavaricum/cupromicans; Nomada goodeniana/succincta .
Mitochondrial introgressions have been suggested for four of these cases
(Schmidt et al., 2015; see details below); low-divergence were suggested
for the controversial A. carantonica /trimmeranacomplex (Schmidt et al. 2015); while deep within-species divergences not
associated with morphological differentiation have been documented theAndrena amieti/allosa/bicolor/montana clade (Praz et al. 2019).
Most of these cases are also controversial with respect to morphological
delimitations, so that current evidence based on the combined
characteristics of morphology and COI-based DNA barcodes does not enable
definite conclusions on the status of these species.