Genetic diversity and species delimitation
The data from COI and ddRAD revealed deep genetic divergences withinH. bicuspis that might correspond to multiple speciation events. In COI, H. bicuspis was split into several lineages (= putative species), with GMYC suggesting seven and ABGD four lineages based on threshold values between 0.8 - 2.1 % uncorrected p -distance. Based on the distance distribution, we decided to work with the following five lineages: ICOI (east to south-east of Iceland; including station 1159 which was separated by GMYC), IICOI (east of Iceland, close to the shelf; ICOI and IICOI were combined in ABGD), IIICOI (north of Iceland), IVCOI (south and south-west of Iceland; split into two syntopic lineages by GMYC), and VCOI (west and south-west of Iceland) (Figure 1, Table 2). None of these five lineages were syntopically recorded, and most lineages were clearly geographically separated. Only lineages IVCOI and VCOI occurred sympatrically along the Reykjanes Ridge. Lineage VCOI was genetically the most divergent lineage with uncorrected p -distances of 4.9 - 7.3 % to the other lineages.
In the nuclear ddRAD data five lineages could also be delimited, however, the assignment of populations north of the GIF Ridge differed in some important aspects (Figure 3 Table 3). To better discern these lineages north of the GIF Ridge, separate ipyrad analyses were performed for lineages I-III, including PCA, Structure, and coancestry analyses. IRAD includes only south-eastern populations along the GIF Ridge towards the Faeroe Islands. Stations 1159 and 1172, which were also assigned to ICOI, clustered with the near-shelf station 1194 into lineage IIRAD. The latter station had formed IICOI together with station 1219, however, station 1219 clustered in ddRAD with the northern stations (corresponding to IIICOI) into IIIRAD. As a consequence, individuals from several populations were assigned to different lineages depending on the studied markers. This was most notable at station 1219, whose individuals were assigned to IICOI and IIIRAD, respectively. However, Structure analyses assigned a relatively large fraction of ~34 % of their nuclear genome to IRAD(nearly as much as to IIIRAD), but less than 3 % to IIRAD (Figure 3), suggesting hybridization among all three lineages. In the fineRADstructure analyses of all lineages this was not as evident, however, in the separate analyses of lineages IRAD-IIIRAD, individuals from station 1219 exhibited the highest shared coancestry with IIIRADfollowed by IRAD. This putative hybridization might explain why IRAD and IIIRAD were not differentiated in the PCA that included all individuals, but only in the analyses focusing on lineages I-III (Figure 3). South of the GIF, lineages IVRAD and VRAD were delimited identically to IVCOI and VCOI, respectively, without any indication of hybridization or gene flow among these two lineages or with any of the other lineages (Figure 3).
Only two fully matured males were observed (stations 880 and 873; corresponding to lineages ICOI and IRAD). Both featured the wider second antennular segment typical for H. b. bicuspis , as described by Wolff (1962). Pleopod 1 of one specimen had the two distinct corners of H. b. bicuspis,while the other had a more widely rounded corner, indicative of H. b. tepidus (Figure 2). The latter specimen thus features a mix of characters of both subspecies. All other males were juvenile or preparatory, thus pleopod morphologies were non-informative. Their antennula were mostly resembling H. b. bicuspis . Our results indicate a developmental explanation for the different male types described by Wolff (1962).