The reference samples of Danubian or Atlantic origin (Table 1; Figs. 2, 3) were mostly identified by single genetic clusters, with only marginal signatures of trans-basin genetic introgression: Drava River samples (DRA_1, DRA_2 and DRA_3, Table 1) were predominated by SA mtDNA haplotypes, with the associated nuclear genetic cluster qSA reaching Q-value frequencies from 0.92 to 0.99 at single localities (Table 1; Figs. 2, 3). Inn and Ramsach River samples (INN_1, RAM_1) were dominated by NA mtDNA haplotypes and the associated nuclear genetic cluster qNA (0.96 ≤ qNA ≤ 0.99) (Table 1; Figs. 2, 3). Sava River samples (OBR_1 and RAD_1) harbored SV mtDNA genetic variants, with the associated nuclear genetic cluster qSV reaching 0.99 (Table 1; Figs. 2, 3). The Upper Rhine River sample (ATL_1) was fixed for the RH mtDNA clade, thus referring the corresponding nuclear genetic cluster as to qRH1 (qRH1 = 0.99) (Table 1; Figs. 2, 3). Finally, hatchery grayling sampled from Italian fish farms (HAT_1 and HAT_2) were all fixed for SC mtDNA haplotypes, thus referring the according genetic cluster as to qSC (qSC = 0.99) (Table 1; Figs. 2, 3).
Distribution patterns of nuclear genetic clusters within northern Adriatic samples were more complex; AD mtDNA haplotypes were predominately linked with two distinct, but closely related, nuclear genetic clusters, referred to as qAD1 and qAD2 (Figs. 2, 3). Cluster qAD1 was widely distributed within the Adige drainage basin and was found at all ADI_ samples, at PAS_1, as well as at ISA_2, which is located at the confluence of Isarco and Adige Rivers. Population-level admixture proportions of qAD1 ranged from 0.69 at ADI_1 to 0.95 at ADI_3 (Table 1). A second Adriatic genetic cluster, qAD2, was found in several Adriatic samples but not in the Adige River, with Q-values ranging from 0.19 at TAG_1 to 0.97 at SES_1. Noteworthy, sample sets were never entirely fixed for qAD1 or qAD2 genetic clusters, and multiple signatures of trans-basin Danubian and/or Atlantic grayling gene pools, either of purebred exotic or of hybrid origin, were detected within all samples from Adriatic water courses. These exotic genetic signatures resembled those found in reference samples, specifically in qSA, qSV, qNA, qRH1, and qSC (Table 1; Figs. 2, 3). However, a sixth exotic genetic cluster not represented by our reference dataset was detected within the northern Adriatic samples. This cluster was found in three purebred specimens, one found at RIE_1 and two at BRE_1 samples, as well as in several hybrid fish of the Adige River samples. This genetic cluster was closely linked to qRH1 in STRUCTURE-based distance tree (Fig. 3B), was associated with RH mtDNA genetic variants, that is, haplotype At28 in the case of purebred specimens, and was thus referred to as qRH2. Moreover, some samples (e.g., ADI_1 and TAG_1) showed exceptionally high levels of non-AD genetic signatures, while at samples BRE_1 (predominance of qNA) and ADD_1 (fixed for qRH1), no native qAD genetic profiles were detected at all. Finally, samples collected from rivers belonging to the eastern Adige basin (RIE_1, RIE_2, AUR_1 and ISA_1) showed no qAD genetic signatures but were predominated by varying frequencies of qSA membership, spanning from 0.46 at RIE_1 to 0.91 at AUR_1 (Table 1; Figs. 2, 3).
STRUCTURE-based analysis of genetic substructure gave similar results when comparing assignment tests run under either the independent or the correlated allele frequency options (Fig. S2): no genetic substructure was revealed within clusters qAD1, qSC, and qSV. In contrast, substantial genetic substructure was found within clusters qAD2, qRH1 as well as qSA. In detail, within qAD2, fish from Sesia (SES_1) were separated from Lower Adda (ADD_2) as well as from Livenza (LIV_1) samples, while within qRH1, Upper Adda samples (ADD_1) were discriminated from Rhine (ATL_1) River samples. Finally, Drava River fish (DRA) were clearly separated from eastern Adige grayling populations (RIE_1, RIE_2, AUR_1 and ISA_1) (Fig. S2).