“And if you gaze long into an abyss, the abyss gazes also into thee”: four morphs of Arctic charr adapting to a depth-gradient in Lake Tinnsjøen

Background The origin of species is a central topic in biology aiming at understanding mechanisms, level and rate of diversification. Ecological speciation is an important driver in adaptive radiation during post-glacial intra-lacustrine niche diversification in fishes. The Arctic charr Salvelinus alpinus L. species complex in the Northern hemisphere freshwater systems display huge morphological and life history divergence in lakes with one or several morphs present, thus offering a unique opportunity to address ongoing speciation mechanisms. We studied Arctic charr in Lake Tinnsjøen by fishing in four nominal lake habitats (pelagial, littoral, shallow-moderate profundal, and deep-profundal habitats) down to 350 meters depth. Research topics addressed were; (1) to illuminate Holarctic phylogeography and lineages colonizing Lake Tinnsjøen, (2) to estimate reproductive isolation of morphs or fish using unbiased methods, and (3) to document eco-morphological and life history trait divergence. Also, we compared Lake Tinnsjøen with four Norwegian outgroup populations of Arctic charr. Results Four field-assigned morphs were identified in Lake Tinnsjøen; the planktivore morph in all habitats except deep-profundal, the dwarf morph in shallow-moderate profundal, the piscivore morph in shallow-moderate profundal (less in littoral and deep-profundal), and an undescribed new morph – the abyssal morph in the deep-profundal only. The morphs displayed extensive life history variation based on age and size patterns. A moderate to high concordance was observed between field-assigned morphs and four unbiased genetic clusters obtained from microsatellite variation. MtDNA suggested the occurrence of two minor endemic clades in Lake Tinnsjøen likely originating from one widespread colonizing clade in the Holarctic. All morphs were genetically differentiated at microsatellites (FST: 0.12-0.20; with some ongoing gene flow among morphs, and for most mtDNA comparisons (FST: 0.04-0.38). Analyses of Norwegian outgroup lakes implied colonization from a river system below Lake Tinnsjøen. Conclusion Our findings suggest post-glacial adaptive radiation of one colonizing mtDNA lineage with divergent niche specialization along a depth-temperature-productivity-pressure gradient. Concordance between reproductive isolation and the realized habitat of the morphs imply that ecological speciation may be the mechanism of divergence. Particularly novel is the extensive morph diversification with depth into the often unexplored deep-water profundal habitat, suggesting we may have systematically underestimated biodiversity present in lakes.

morphs in the lake, and different body proportion, weighting up to 4-6 kg [67]. Thus,when 190 summarizing available information, a set of four morphs were suggested in Lake Tinnsjøen. 191 As no progress occurred considering scientific studies on the small white fish from the 192 bottom of the lake from the ROV team and associated researchers, we decided to perform a 193 fish survey ourselves that was conducted in the lake in 2013 to document the occurrence of 194 morphs. We set up three main research topics with regard to the Lake Tinnsjøen Arctic charr 195 diversity; (1) to illuminate the phylogeography and ancestral lineages colonizing Lake 196 Tinnsjøen (mtDNA-CytB sequences), (2) to estimate reproductive isolation of field assigned 197 morphs or fish assessed using unbiased methods (microsatellites), and (3) to document eco-198 morphological and life history trait divergence (body-shape, proportional catch in habitat, 199 age). To accomplish these tasks we collected fish in different habitats in the pelagial, littoral, 200 shallow-moderate profundal and in the deep-profundal. In the field, we classified fish to 201 morphs from exterior phenotype, while in laboratory we assessed morphological (body shape) 202 and genetic divergence using mtDNA and nDNA markers. We further performed a Holarctic 203 phylogeography with online genetic sequences to evaluate lineages colonizing Lake 204 Tinnsjøen. The strength of association of field-assigned morphs and genetically identified 205 morphs using microsatellites (i.e. genetic clusters) was tested. We compared mtDNA and 206 nDNA in Lake Tinnsjøen with a set of four Norwegian outgroup lakes. Using a putative 207 ancestor below in the same drainage, we compared body shape to the Lake Tinnsjøen morphs.

211
Material used for different analyses 212 The material used for various analyses is summarized in Additional file: Table S1.

214
Study area, fish sampling and field-assigned morphs 215 Lake Tinnsjøen (60 38 15.6 North, 11 07 15.2 East) is a long (35 km), large (51.38 km 2 ) and 216 deep (max depth of 460 m, 190 m mean depth) oligotrophic lake in southeastern Norway (Fig.   217 1a, b) [68]. High mountain sides surround the lake descending steeply into the lake resulting 218 in a relative small littoral area compared to an extensive pelagic volume and a large profundal 219 area. In the southern and northern end of the lake, larger littoral areas exist with shallow 220 depths. The littoral zone is exposed to the elements such as wind and waves. The shoreline is 221 monotonous with few bays and only one small island. The littoral zone is composed mostly of 222 bedrock, large boulders, smaller rocks as well as sand in less exposed areas and in the deeper Tinnsjøen likely offers a divergent temperature profile (as well as light, pressure and 235 productivity) among habitats, depths and niches, along pelagic and littoral-benthic depth-236 gradients from surface to 460 m. 237 We collected Arctic charr from Lake Tinnsjøen during 2013 and from four additional 238 outgroup populations (see below) North, West, East and South of Lake Tinnsjøen in 2013-239 2015 (Fig. 1a). Fish were caught in four lake habitats (can be viewed as crude nominal niches 240 for individuals and morphs) in Lake Tinnsjøen using equipment described below. At this 241 stage, we do not reveal the exact sampling sites until the taxonomic status of the new abyssal 242 morph has been described and conservation biology authorities in Norway have considered 243 the situation with regard to its conservation value. Particularly relevant here is the population 244 size and uniqueness of the new discovered morph, and what conservation status it merits. As 245 the lake have steep mountain sides entering the lake, it is hard to place equipment precisely at 246 predetermined positions. Thus, habitat and depth ranges fished were grouped to be able to 247 compare catch among four nominal lake habitats. Some fish equipment overlapped to some 248 degree with catches across habitas. However, depth measurements were taken at catch 249 location and noted in the field log, which allowed for later reallocation of the catch effort. As 250 such, the lake habitats defined are quite crude categorizations of habitat and depth ranges, but 251 generally fits with limnologic description of lake niches. The four lake habitats (nominal 252 niches) sampled (and defined by us) in Lake Tinnsjøen in 2013 were; (i) the pelagial (gillnets 253 at <20 m depth, in areas with depths of >30 m, and >50 meters from the shore), (ii) the littoral 254 (gillnets from shore <20 m depth), (iii) the shallow-moderate profundal (gillnets, traps and 255 hook and line from shore at >20 m and <150 m depth), and (iv) the deep-profundal (traps 256 at >150 m depth, > 100 m from the shore). 257 Sampling was conducted with gill-nets, baited anchored longlines, and traps. Initially, 258 we aimed at fishing with a standardized effort x equipment in all niches, but due to the experimental nature of fishing at depths >150 m, and the low fish density, it was difficult to 260 obtain a sufficient sample size. Thus, we intensified the effort in the different habitats with the 261 catch methods that worked best. As such, the material obtained may not be fully 262 representative of fish populations at all depths and habitats, but represents an opportunistic 263 sampling strategy under quite challenging fishing conditions. We used different monofilament 264 series coupled in gangs when fishing with gillnets. In the pelagial, we used a 12-panel being minute, ie. < 10 individuals in few locations). The lake only hold these four fish species.

283
Minnow was introduced into Lake Tinnsjøen recently (likely in the timeframe 1960-1970`s). 284 Fish were killed using an overdose of benzocain and transported dead on ice to the 285 field laboratory at Lake Tinnsjøen. In the field, all the fish were subjectively assigned into 286 four nominal morphs based on exterior morphology, being; (i) planktivore, (ii) dwarf, (iii) 287 piscivore, and (iv) abyssal (see representative individuals in Fig. 1c). ). Each fish was 288 classified as one of the four morphs despite variation within morphs and subsequent 289 uncertainties This field assignment of morphs was labelled as field-assigned morphs (FA-290 morphs). Length and weight were recorded, with sex and maturity stage, and age from otoliths 291 in the laboratory. A DNA sample was taken in the field and stored on 96% EtOH for use in 292 analyses (see description below).

293
The four additional outgroup populations of Arctic charr were situated to the North 294 (River Leirfossvassdraget; anadromous sea-running), West (Lake Vatnevatnet), East (Lake 295 Femund) and South (Lake Tyrivatnet), of Lake Tinnsjøen (Fig. 1a). The three latter Arctic 296 charr populations were stationary in freshwater. The sampling equipment, effort and 297 placement varied among lakes comprising gill nets with at least 16.5. 19.5, 22.5. 29.0 mm 298 (knot to knot) and/or modified Jensen series or Nordic multi-mesh panels set in littoral, 299 pelagic, and profundal areas. In the laboratory, these four populations were analysed as 300 described above for Lake Tinnsjøen. A DNA sample was also stored on 96% EtOH for 301 analyses. These four populations were used as selected outgroups in microsatellite analyses, 302 in mtDNA based phylogenetic analyses, and partly in the morphological analyses. Arctic 303 charr in Lake Tyrivatn was inferred as a putative "ancestral state" founder that could have 304 colonized Lake Tinnsjøen, and was thus used for comparative purposes in microsatellite-, 305 mtDNA-and morphometric analyses (Fig. 1a,b). This was anticipated as the lake is situated 306 far below Lake Tinnsjøen in the same watersystem (see supporting argumentation of the most 307 likely colonization routes in the discussion below). Ideally, we would use the real founding 308 population into Lake Tinnsjøen, but this is not known. interpreted mostly based on both forward and reverse readings (but in a few cases, only one 330 sequence direction was readable). A set of 115 Norwegian sequences were retrieved where 331 sample size range 21-22 for the four Lake Tinnsjøen FA-morphs and a sample size of 5-9 for 332 the four Norwegian outgroup lakes (Additional file 1: Table S1).

333
For larger scale comparison of phylogeny, highly similar sequences were retrieved 334 using BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi) (Fig. 2a, haplotypes had good statistical support (89%) from the remaining haplotypes and were 357 selected for further resolution, covering a large geographical range. The purpose with this clade selection was to have an in-depth look at the putative radiation and geographical 359 distribution of the closest genetic relatives to the Lake Tinnsjøen morphs.

360
In addition, for Lake Tinnsjøen, a network was built using TCS v1.21 [80] visualizing 361 only the mtDNA sequences found inside the lake to reveal putative formation of subclades 362 after initial founding colonization.

363
The three 1-mutational step clades (clade 1-1-, 1-2, and 1-3) revealed in TCA in Lake    Based on K-clusters results from the STRUCTURE analysis, we assigned different 423 genetic populations or morphs in Lake Tinnsjøen (GA-morphs). We further contrasted Lake

424
Tinnsjøen with the four outgroup lakes. Assignment analyses was based on K-clusters of 425 individuals with q-values of >0.7 to its own cluster, evaluated as belonging to this population.

426
Individuals with q-values <0.7 were interpreted as being hybrids of unsure population origin.

427
As an alternative way to test genetic differentiation, we first conducted a principal . In addition, we used the 432 same approach for testing differentiation, but now along PC1-3, for four FA-morphs in Lake 433 Tinnsjøen as described above, by only sub-setting Lake Tinnsjøen from the five lake dataset.

435
Eco-morphological and life history trait divergence in the Lake Tinnsjøen charr morphs 436 In Lake Tinnsjøen, association between habitat occurrence with FA-morphs or GA-morphs  See also previous section on fish sampling regarding overall issues related to statistical testing.

441
A discriminant analysis in JMP 11.2 [99] was used to test for association between GA-442 morphs and FA-morphs.

443
A geometric morphometric analysis using landmarks to reveal body shape was 444 conducted using; Lake Tinnsjøen only, and secondly Lake Tinnsjøen and Lake Tyrivatn in the 445 river drainage to the south of Lake Tinnsjøen. In the latter analysis, the idea was to evaluate 446 the phenotype of the putative ancestral founder that could have colonized Lake Tinnsjøen, and 447 how the Arctic charr in Lake Tyrivatn was morphologically assigned to the FA-morphs in 448 Lake Tinnsjøen. with their left side fronting the camera. All fish which had inflated swim bladders were 453 carefully punctuated so that inflation did not affect body shape (subjective correction).

471
To evaluate how concordant body shape was to FA-morphs in Lake Tinnsjøen, we 472 used a discriminant analysis in in JMP 11.2 [99] with linear, common covariance using 473 residuals from the five PC-axes in MorphoJ. Similarly, we tested morphological resemblance 474 in body shape of the FA-morphs with their putative ancestral founder from Lake Tyrivatn.

475
Assignment percentages to each of the categories were recorded for both analyses.

476
A subset of the fish was used for determining age based on otoliths, immersed in 95% 477 EtOH, read using a microscope [103]. An unfortunate challenge was encountered as the 478 Arctic charr heads had been stored in unbuffered formalin which partly prevented age reading 479 in some fish due to unbuffered formalin eating up parts of the otoliths. However, for age 480 determined fish, we were confident in their age. Some of the morphs had few individuals 481 analyzed for age. Also, it was difficult to determine maturity stage in some fish. This situation 482 prevented a thorough life history analysis. Thus, we present age and body weight distributions 483 revealing the youngest sexually mature male and female (also for body weight distributions).

498
Phylogeography and the ancestral lineages colonizing Lake Tinnsjøen based on mtDNA 499 A set of 13 haplotypes (h1-13) were found in the combined dataset of Lake Tinnsjøen and the 500 four Norwegian outgroup lakes ( Table 3). The 13 haplotype sequences obtained in our study 501 are deposited on GenBank (accession number x-y). Here, 12 of the 13 haplotypes were only 502 found in Lake Tinnsjøen (which lacked h2). The four outgroup lakes all had haplotype h1, 503 which also occurred in all of the four FA-morphs, while only one outgroup lake, Lake 504 Vatnevatnet, had an additional haplotype h2. 505 From the samples in the larger scale phylogeography (Fig. 2a), a total of 75 new 506 haplotypes were retrieved from Blast, comprising 88 haplotypes including the 13 Norwegian 507 haplotypes (Additional file: Table S2). Comparing these 75 haplotypes to the ones found in 508 Norway revealed that only h1 (in several lakes) and h13 (in one lake) were found outside 509 Lake Tinnsjøen and the four Norwegian outgroups. Lake Tinnsjøen harbored a set of 10 510 endemic haplotypes (h3-h12).

511
The major branch in Figure 2b (light purple) including the Lake Tinnsjøen haplotypes 512 were used for drawing a minimum spanning network, not considering frequencies of 513 haplotypes. This major clade with 21 haplotypes had good statistical support (89%), covering 514 a large geographical range (Fig. 2b,c). Within the light purple clade, a total of 6 haplotypes or 515 sub-clades was supported with good statistical bootstrap values between 77-93%.

516
In figure 2b the phylogeny of the 13 haplotypes in Lake Tinnsjøen revealed moderate 517 to high bootstrap support for clustering of three "clades"; clade I (h1, h2, h10) with bootstrap 518 support of 88%, clade II (h5-h9. h11, h12) with bootstrap support of 93%, and clade III (h3, 519 h4) with bootstrap support of 85%. Here, clade I consisted of more haplotypes (i.e. h13-18, 520 h21, h32, h33) that were found outside Lake Tinnsjøen and the four Norwegian outgroup 521 lakes. One haplotype link, h5-h13, had unresolved cluster groupings, where it was interpreted 522 that h5, being one mutational step away from h1, belonged to clade II rather than to clade I, 523 and that h13 belonged to clade I. The tree topology in figure 2b support these evaluations.

524
The minimum spanning network drawn using only the 13 haplotypes in Lake 525 Tinnsjøen revealed variable frequency and their internal phylogenetic relationship (Fig. 2d). 526 Using the FA-morphs within Lake Tinnsjøen as units, the number of haplotypes 527 ranged from 4 in the piscivore morph to 6 in the dwarf and planktivore morph (Table 3).
The genetic diversity (Table 3) of FA-morphs ranged from a low haplotype diversity 535 of 0.476 (planktivore morph) to a high 0.743 (dwarf morph) in Lake Tinnsjøen, and from 0-536 0.222 (highest in Lake Vatnevatnet) in outgroup lakes. In Lake Tinnsjøen combined, the 537 haplotype diversity was 0.711. Similarly for nucleotide diversity, a low value was observed 538 for the abyssal morph (0.00078) and a high value for the dwarf morph (0.00128), while the 539 four outgroup lakes varied 0-0.00026 (highest in Lake Vatnevatnet). In Lake Tinnsjøen 540 combined, nucleotide diversity was 0.00124.

541
Pairwise distance FST values based mtDNA Cytochrome B among the four morphs in 542 Lake Tinnsjøen ranged from a low 0.042 (planktivore vs piscivore morphs) to a high 0.38 543 (planktivore vs dwarf) (Additional file: Table S4). All other FST comparisons than the 544 planktivore versus the piscivore morphs were significantly different. and Li`s D; p-value > 0.10).

554
Reproductive isolation of field assigned morphs or fish assessed using unbiased methods 555 The hierarchical STRUCTURE analysis suggested K=8 genetic clusters with the four morphs 556 in Lake Tinnsjøen and the four Norwegian outgroup lakes occurred as distinct clusters (Fig.   557 5ad, Additional file: Table S9, hierarchical STRUCTURE plot in Additional file: Fig. S1).

574
Using the same principal components on microsatellite as above, but only contrasting 575 the four FA-morphs in Lake Tinnsjøen, revealed that four out of the six comparisons were 576 significantly different for PC1 (q=2.57, alpha =0.05), and five of six were significantly 577 different for PC2 (Fig. 5c). For PC1, the piscivore morph was not different from the abyssal 578 morph, and the planktivore morph was not different from the dwarf. Along PC2, the dwarf 579 morph was not different from the abyssal morph, while for PC3, the piscivore and abyssal 580 morph did not differ significantly.

591
In the contingency analysis of habitat-specific catch by the four revealed GA-morphs 592 the association was also significant (N=344, Df=12, R 2 (U)=0.4283, Likelihood ratio test; 593  2 =302.55 and P<0.0001), although less than 20% of cells in the tests had expected count <5

594
(suggesting x 2 to be suspect)(Additional file: Table S6). The same general pattern emerged as 595 previously described for FA-morphs above in the FA-morphs by habitat-specific catch 596 contingency analysis.

604
Similarly, back assignment using body shape of the FA-morphs and the putative 605 ancestor from Lake Tyrivatn showed that Lake Tyrivatn had highest assignment to plantivore 606 morph (18.8%), a lower assignment to dwarf (6.3%) and piscivore (3.1), while no fish from 607 Lake Tyrivatn was assigned to the abyssal morph in Lake Tinnsjøen (Additional file: Table   608 S7).

609
For comparative purposes, the FA-morphs and GA-morphs were visually contrasted 610 with regard to age and weight distribution, suggesting large difference among morphs (Fig.   611 4bc). It seems that the planktivore morph has the lowest age span followed by a roughly equal 612 life span in the dwarf and abyssal morph. The piscivore morph has the longest life span. There 613 were large differences in weight distribution, where the piscivore attained the largest size 614 followed by the planktivore morph. The dwarf and abyssal morphs were minute in 615 comparison, but the dwarf morph attained a larger body size than the abyssal morph. The 616 comparison of FA-morphs and GA-morphs broadly gave the same picture in age and weight.

619
In our study we have found empirical support for evaluating the three main research questions 620 addressed. First, we find it reasonable to postulate that members of one Holarctically 621 widespread mtDNA lineage colonized Lake Tinnsjøen, likely suggesting one single common 622 ancestor that later diversified into the observed four sympatric morphs. Further, the number of 623 endemic haplotypes, and signatures of population demographic expansion, in one of the Lake 624 Tinnsjøen clades (clade II), support a mechanism of intralacustrine diversification. Secondly, 625 we found that the four field assigned morphs were genetically divergent at microsatellite loci 626 (and partly mtDNA), thus suggesting reproductive isolation among morphs (although with 627 some degree of gene flow). Further, there were a close association between field assigned 628 morphs and unbiased genetic analyses (microsatellites) revealing four distinct genetic clusters 629 in the lake, supporting morph differentiation. Thirdly, we evaluate that the four morphs were  Tinnsjøen was glaciated and we thus assume that it could not have been accessible for fish 662 immigration prior to that period -setting a crude frame for colonization to <9 700 ybp. We 663 further infer that the fish colonization have proceeded from the southeast through the River 664 Skienselva, or alternatively through any existing non-identified pro-glacial lakes situated 665 southeast of Lake Tinnsjøen. This is also logic given the elevation level of the landscape 666 surrounding Lake Tinnsjøen, where colonization along the suggested direction is most likely 667 as the alternative routes imply crossing mountains and elevated slopes. Furthermore, the estimated ice-flow directions ( Fig. 1b; [105],) support that the Arctic charr colonized Lake 669 Tinnsjøen along the River Skienselva from the coastline and upwards. As the Arctic charr can 670 be anadromous and live short periods in the sea [19], and as the Skagerak area at certain times 671 during deglaciation was carrying a brackish water upper layer [107,108], it seems reasonable 672 to infer that the Arctic charr came from the south and colonized lake Tinnsjøen from the coast.

674
Holarctic phylogenetic patterns using mtDNA CytB in Arctic charr and Salvelinus spp. 675 We have screened a moderate number of each of the four morphs in Lake Tinnsjøen and only

697
For the five Norwegian lakes combined, we observed a set of 13 haplotypes (h1-h13).

698
Haplotype h1 was found in all the five Norwegian lakes, as well as in some other Holarctic 699 lakes (Sweden, Finland, Russia, and Canada). Haplotype h13 was found in Lake Tinnsjøen 700 and one other lake in Sweden. Haplotype h2 was only found in one Norwegian lake (Lake 701 Vatnevatnet). A set of 10 haplotypes (h3-h12) were found to be endemic in Lake Tinnsjøen.

702
In the phylogenetic analyses using the full dataset of the 88 mtDNA CytB sequences, a 703 moderate to strong statistical support for branching events were observed (Fig. 3). These 704 major branches were found in different parts of the Holarctic and reflects a deeper taxonomic 705 partitioning than only containing Arctic charr (Additional file: Table S2a). However, our main 706 purpose of this large scale comparison of CytB sequences (using different described charr 707 (Salvelinus spp.) taxa) was to visualize, and polarize, the closest relatives in the major branch 708 that also contained the Arctic charr in Lake Tinnsjøen. With that in mind, we focused on the 709 major light purple lineage in Figure 3 (named #1 in Fig. 3a,b). This lineage included a set of 710 21 haplotypes widely distributed (Fig. 3c) in Transbaikalia -Kamchatka -Bering Sea (h17, 711 h18, h21, h32, h33), Quebec -Taimyr -Chucotka - Fennoscandia (h1, h2, h13, h14, h15, h16), 712 and in Lake Tinnsjøen (h3-h13). Based on these results, it seems plausible to evaluate that the 713 ancestors of this major lineage colonized a large geographic area throughout the Holarctic, 714 where some ancestral individuals also colonized Lake Tinnsjøen. Here, the founders of Lake 715 Tinnsjøen could potentiall have carried the h1 haplotype (clade I), subsequently giving rise to 716 clade II (h5, h7, h8, h9, h11, h12) and clade III (h3, h4) (Fig. 3d).
In our study, considering the major lineage 1 (light purple; named #1 in in Fig. 3a,b) 718 with 21 haplotypes (h1-h18, h21, h32, h33), it is interesting to see what species designations 719 are given to haplotypes as these should be the closest relatives to morphs in Lake Tinnsjøen.

730
In general, there were few similar CytB sequences in GenBank, and specifically with 731 regard to the Fennoscandian area (as revealed in Fig. 3d). However, we contrast our findings Siberia revealed that the Atlantic group was distributed all the way to Taimyr (see also [117]).

743
Further, Alekseyev et al. [118] found no strong support for a separation of Atlantic and 744 Siberian haplotypes into two distinctive groups when analyzing Arctic charr in Siberia using 745 the mtDNA control region. Thus, these studies in general seem to support our crude inference 746 of a widespread Atlantic mtDNA lineage (Fig. 3a,b). However, we seem to lack the Siberian, these two studies, it seems that the piscivore morph is less genetically distinct than the three 847 other morphs in that lake. In comparison, the FST values among the four sympatric morph in 848 Lake Tinnsjøen were 0.042-0.382. Here, the only comparison that was non-significant was 849 between the planktivore and the piscivore morph. Thus, five out of the conducted six morph 859 not yet seen outside the lake. However, that could also reflect limited geographical coverage 860 nearby, or far from, Lake Tinnsjøen. Thus, one should be cautious interpreting these results.

861
In summary, genetic divergence (using different markers) among sympatric Arctic 862 charr morphs in lakes throughout the Holarctic varies widely, and we expect them to do so 863 given their different evolutionary histories, genetic load and evolvability, biotic and abiotic 864 environmental conditions, and ecological opportunities to radiate. Indeed, we can see systems 865 with one to four morphs in different lakes. However, few studies have addressed nuclear and 866 mtDNA markers at the same time. The best studied system so far is Lake Thingvallavatn in Iceland, where four morphs reside -where one of the morphs (i.e. piscivore morph) could be 868 recently evolved as a separate genetic cluster or being induced due to growth threshold 869 dynamics. In comparison, in Lake Tinnsjøen we have described four morphs that are different 870 with regard to microsatellites and mostly with regard to mtDNA haplotypes. The Arctic charr 871 morphs in Lake Tinnsjøen seem to be differentially distributed along a depth-temperature-872 productivity-pressure gradient. The evolutionary branching in their phylogeny, and the high 873 number of endemic haplotypes in Lake Tinnsjøen, with signatures of demographic expansion, 874 could support an intra-lacustrine origin of these morphs. However, the evolutionary scenarios 875 remain to be tested in detail using a set of higher resolution markers. Although the Arctic 876 charr species complex has been studied for a long time, researchers still need to address the 877 important mechanisms underlying origin, presence and temporal persistence of sympatric 878 morphs. Thus, a multi-method based eco-evo-devo approach with ecological, morphological 879 and life history studies [146], and state of the art genomics as performed in Lake  would apply for a system with three niches and morphs -evolving a morph adapted to the 907 profundal. Based on the number of sequence of morphs from monomorphic to four morph 908 systems, it seems that there is a predictable temporal pattern in evolutionary branching 909 associated with niche radiation. Here, the littoral (or pelagial) may be the first niche to be 910 filled -then the pelagial (or littoral) -then the profundal, with a piscivore morph originating 911 putatively due to growth threshold dynamics from one of the units, or evolving independently.

912
Adding upon this complexity, moving away from an assumption of only three discrete 913 niches in given a lake, one can imagine that there could be gradients of expected predictable 914 fitness along environmental variation such as e,g, the depth-temperature-productivity-pressure 915 gradient in Lake Tinnsjøen. Indeed, a study on polymorphic European whitefish (Coregonus 916 lavaretus) in the Swizz Alpine Lake Neuchâtel suggest adaptive diversification and build up of reproductive isolation along ecological gradients when assessing morphs spawning at 918 different time and place [147]. Ohlberger et al. [148] used an adaptive-dynamics model, 919 calibrated with empirical data, finding support for an evolutionary diversification of the two 920 German Lake Stechelin Coregonus sp. morphs likely driven by selection for physiologically 921 depth-related optimal temperatures. In the 1.6 km deep Lake Baikal, Russia, one of the oldest 922 freshwater lakes on earth, adaptive radiations have occurred in several taxa such as e.g. 923 reflected by the depth-gradient and the environmental niche radiation of the freshwater 924 sculpins (Cottidae, Abyssocottidae and Comephoridae) [149]. Also, speciation along depth 925 gradients in the ocean are strongly suggested [150]. A study by Chavarie et al. [151] tested a 926 multi-trait depth gradient diversification of morphs in Lake trout (Salvelinus namaycush) in 927 Bear Lake in Canada, but did not find a strong association in differentiation with depth (but, 928 partly association with genetic structure), suggesting that a highly variable nature of 929 ecological opportunities existed for divergent selection and phenotypic plasticity. In 930 comparison with these studies, it seems reasonable to infer that there is a depth-temperature-  Thus, revealing mechanisms in speciation trajectories in the Arctic charr complex is indeed a 943 challenging task.

944
A novel finding in our study was the appearance of the deep-profundal abyssal morph 945 with its distinctive phenotypic features, apparently being adaptations to the cold, dark and 946 low-productive high-pressure environment in deeper parts of the oligotrophic Lake Tinnsjøen.

947
Our finding of the four morphs could reflect ongoing divergence along a depth-temperature-948 productivity-pressure gradient from surface to deep profundal environments. This imply large 949 differences in yearly cumulative temperature sum at different depths and productivity, likely 950 strongly affecting life history evolution. In shallow Fennoscandian lakes, the littoral seem to 951 have the highest biotic production, followed by the pelagial and profundal [152]. In the 1.6 952 km deep Lake Baikal, oligochaetes was found from the surface down to maximum depth, occupation could be even stronger than previously anticipated, selecting traits that have not been seen in other morphs from other lakes. In Lake Tinnsjøen, the small eyes (an apparent 968 reduction of size and potential function?) in the abyssal morph bear apparent similarities with 969 eye-reduction seen in cave fishes [e.g. 155]. This seems somehow logical given that cave 970 environments often can be described as nutrient-poor, cold, and harboring few present species.

971
In speculation, there might be a temporal cascade effect in the adaptation process to a 972 given niche where e.g. body size, growth rate, sexual maturation, coloration, secondary sexual 973 traits, physiology and morphology are highly contingent upon the ecological opportunity, 974 constraining environmental conditions, and the evolutionary optimal solutions in any given 975 niche. Here, e.g. morphology and physiology may reflect specialization to niches, growth rate 976 and size and age at sexual maturity may reflect food conditions and predation regimes, while 977 body size, coloration and secondary sexual traits may reflect the optimal visual conditions, 978 affecting mate choice behavior and thus sexual selection. Here, evolution would likely result 979 in optimal solutions to obtain the highest overall life-time fitness in a given niche, and also 980 due to relative fitness rewards in yet other niches in the lake due to overall e.g. frequency or 981 density dependent fitness of the present morphs. Further, in this adaptive process e.g. major 982 histocompatability complex (MHC) genes, which are related to e.g. kin recognition, parasite 983 and disease resistance, as well as niche occupation, may be important as previously shown to 984 differ among morphs and lakes in Arctic charr [156][157][158][159]. A study by Baillie et al. [160] 985 surveying microsatellite and a MHC gene in Lake Trout in Lake Superior revealed that 986 variation was partitioned more by water stratum than by ecomorph with a stronger association  [163][164][165][166][167]). However, the process of ecological speciation is complex and remains to be 1002 tested awaiting ecological niche studies and using higher resolution genetic markers under an 1003 evolutionary scenario framework comparing simulated and empirical data. As a crucial and 1004 fundamental basis in ecological theory, we would also here, in our newly discovered Lake

1005
Tinnsjøen system, expect a niche-specific fitness trade-off in adaptations to evolve so that no 1006 one phenotype will be optimal in all the available lake niches. Thus, the saying "Jack of all 1007 trades, master of none, but oftentimes better than master of one" might nicely reflect the early 1008 postglacial stages of the ongoing evolutionary dynamics in adaptive radiation of Arctic charr. below the species level. Whether or not Lake Tinnsjøen represents a true sympatric speciation 1025 process remains to be tested using a combined set of genetic markers to contrast evolutionary 1026 scenarios. Lake Tinnsjøen offers a rare research window into an ongoing speciation process -1027 evidently revealing an important part of the worldwide evolutionary legacy of Arctic charr. 1028 We suggest that the Norwegian management authorities merit Lake Tinnsjøen special 1029 biodiversity protection as it is one of the most divergent Arctic charr systems seen worldwide. Ethics approval and consent to participate 1035 Not applicable.

1037
Fishing license 1038 Fish were sampled after initial consent from local authorities at Tinn County Administration 1039 giving us permission to fish in Lake Tinnsjøen after consent was approved also by the local   Further, as this morph is not found elsewhere in the world, it merits the highest conservation 1056 status possible. The Lake Tinnsjøen represents a unique window into speciation for scientists.

1058
Competing interests 1059 The authors declare no conflict of interest.    Fig. 6). Summarize statistics for genetic variation in the morphs and 1772 lakes is also given.

1980
Values in bold denote a significant test with regard to population expansion events.