A genomic perspective on brown trout migration

(1)

uef.fi

PUBLICATIONS OF

THE UNIVERSITY OF EASTERN FINLAND Dissertations in Forestry and Natural Sciences

Dissertations in Forestry and Natural Sciences

DISSERTATIONS | ALEXANDRE LEMOPOULOS | A GENOMIC PERSPECTIVE ON BROWN TROUT MIGRATION | N

ALEXANDRE LEMOPOULOS

A GENOMIC PERSPECTIVE ON BROWN TROUT MIGRATION

PUBLICATIONS OF

THE UNIVERSITY OF EASTERN FINLAND

Brown trout (Salmo trutta) is extremely diverse in terms of migratory behaviour, ranging from entirely resident to strictly

migratory populations. The reasons underlying this variation have conservation implications. This thesis highlights the genetic underpinning of the decision to migrate. High-

throughput sequencing is used to investigate both the genomic causes and consequences

of the migratory behaviour and provides novel information for the management of

endangered migratory populations.

ALEXANDRE LEMOPOULOS

(2)

(3)

A GENOMIC PERSPECTIVE ON BROWN

TROUT MIGRATION

(4)

2

31181993_UEF_Vaitoskirja_NO_348_Alexandre_Lemopoulos_LUMET_inside_19_07_31.indd 2 01/08/2019 8.26

(5)

Alexandre Lemopoulos

A GENOMIC PERSPECTIVE ON BROWN TROUT MIGRATION

Publications of the University of Eastern Finland Dissertations in Forestry and Natural Sciences

No 348

University of Eastern Finland Joensuu

2019

Academic dissertation

To be presented by permission of the Faculty of Science and Forestry for public examination in the Auditorium N100 in the Natura Building at the University of

Eastern Finland, Joensuu, on August, 16, 2019, at 12 o’clock noon

(6)

4

Grano Oy Jyväskylä, 2019 Editor: Raine Kortet

Distribution: University of Eastern Finland / Sales of publications www.uef.fi/kirjasto

ISBN: 978-952-61-3158-0 (nid.) ISBN: 978-952-61-3159-7 (PDF)

ISSNL: 1798-5668 ISSN: 1798-5668 ISSN: 1798-5676 (PDF)

(7)

Author’s address: Lemopoulos Alexandre University of Eastern Finland

Depart. of Environmental and Biological Sciences P.O. Box 111

80101 JOENSUU, FINLAND email: alexandre.lemopoulos@uef.fi Supervisors: Professor Anssi Vainikka, Ph.D.

University of Eastern Finland

Depart. of Environmental and Biological Sciences P.O. Box 111, 80101 JOENSUU, FINLAND email: anssi.vainikka@uef.fi

Doc. Silva Uusi-Heikkilä, Ph.D.

University of Jyväskylä

Depart. of Environmental and Biological Sciences P.O. Box 35, 40014 Jyväskylä, Finland

email: silva.k.uusi-heikkila@jyu.fi Doc. Pekka Hyvärinen, Ph.D.

Natural Resources Institute Finland

Manamansalontie 90 C 88300 Paltamo, Finland email: pekka.hyvarinen@luke.fi

Professor Robert Arlinghaus, Ph.D.

Leibniz-Institute of Freshwater Ecology and Inland Fisheries

& Humboldt-Universität zu Berlin Müggelseedamm 310, 12587 Berlin email: arlinghaus@igb-berlin.de

Reviewers: Professor Jouni Aspi, Ph.D.

University of Oulu Ecology and Genetics P.O. Box 3000

FI-90014 University of Oulu, Oulu, Finland email: jouni.aspi@oulu.fi

Supervisory Research Geneticist Krista Nichols, Ph.D.

National Oceanic and Atmospheric Administration (NOAA), National Marine Fisheries Service, Northwest Fisheries

(8)

6

Science Center, Seattle, USA email: krista.nichols@noaa.gov Opponent: Professor Claus Wedekind, Ph.D.

Department of Ecology and Evolution, Biophore, University of Lausanne,

1015 Lausanne, Switzerland email: claus.wedekind@unil.ch

(9)

“So comes snow after fire, and even dragons have their endings”

J.R.R. TOLKIEN

(10)

8

(11)

Lemopoulos, Alexandre

A genomic perspective on brown trout migration Joensuu: University of Eastern Finland, 2019 Publications of the University of Eastern Finland Dissertations in Forestry and Natural Sciences 2019; 348 ISBN: 978-952-61-3158-0 (print)

ISSNL: 1798-5668 ISSN: 1798-5668

ISBN: 978-952-61-3159-7 (PDF) ISSN: 1798-5668 (PDF)

ABSTRACT

Migrations are fascinating behaviours that are almost ubiquitous within the animal kingdom. Salmonids in particular perform vast migrations between their natal streams and larger water bodies such as seas, lakes or rivers. Within salmonids, some species also display partial migrations so that some individuals will migrate while others will stay in the rivers for their entire lifespan. These different life-history strategies can be observed within populations, where migratory and resident individuals live in sympatry, but more typically allopatric populations are either migratory or resident depending on the prevailing ecological conditions.

Brown trout Salmo trutta is an especially interesting partially migratory species because of its large diversification and its polytypic morphs. The causes of the migratory life-history dichotomy have been extensively studied from ecological, physiological and environmental perspectives. As such, it has been shown that brown trout is sensitive to a variety of parameters such as water temperature and flow, density of individuals or food availability. In addition, compared to resident fish, migratory individuals often display differences in metabolism, affecting, for instance, growth, brain development and/or adaptation to different salinity.

The genomic basis for migration decisions has been established in some salm- onid species. Rainbow / steelhead trout pair of Onchorynchuss mykiss has been exten- sively studied and multiple candidate genes, as well as specific genomic regions have been proposed to explain the dichotomy in their life-history. Candidate genes for the propensity to migrate have also been proposed for the Sockeye salmon Onchoryn- chuss nerka. In contrast, brown trout had received very little attention from the ge- nomic perspective. It was thus compelling to understand whether the migratory behaviour displayed by brown trout also has a genetic cause or whether it is entirely dependent on individual plasticity.

In this thesis, I used restriction site associated DNA sequencing to investigate the propensity to migrate in brown trout. First, I studied wild populations from dif-

(12)

10

ferent Finnish-Russian watersheds and assessed the population differentiation between populations displaying putatively different migratory behaviour. My co-authors and I revealed the population structure of brown trout in the Oulujärvi watershed, where migratory and resident populations show partial isolation from each other. In the Koutajoki watershed, we observed unequal gene flow between confirm- edly migratory and putatively resident populations. This pointed to behavioural differences among these populations. Finally, we directly compared migratory and resident individuals by using wild populations and a common garden experiment.

These studies yielded candidate markers for the migratory behaviour in brown trout.

Interestingly, these markers mapped against salmon genes involved in functionally important pathways for migration (e.g. brain development, osmoregulation, growth and immunity). Moreover, these gene families had previously been identified in other migratory salmonids supporting an idea of the ancestral evolution and multi- genic nature of migration tendency in all salmonids.

The results of this thesis are important from a conservation and management perspective, as the previously assumed large phenotypic plasticity in brown trout migration appears to have genetic components. Adequate management, where genomic differences between populations and differentially migrating individuals are considered, is needed in order to maintain regional genetic diversity and both resident and migratory brown trout ecotypes.

(13)

(14)

12

(15)

ACKNOWLEDGEMENTS

Four years ago, I was on Bolivar beach, Athens, holding an iced coffee on my hand, asking myself whether doing my PhD thesis in Finland was a smart decision. At the time, the idea of moving in Finland, in a town I had never heard off, leaving friends and family behind was daunting. Yet, after the 72 hours (thanks for that) given me to decide where I would spend the next four years of my life, I decided to go on an adventure, in Joensuu. Four years after, here I am, wrapping up the last section of my PhD thesis…I am grateful to many people without whom I would not be here today, writing this paragraph.

First and foremost, Anssi. You gave me the chance to prove myself as a scientist back in 2015. You were helpful and supportive when I needed it the most, and you were never afraid to give me freedom and responsibilities in my work. We never had any strong disagreement and were always working with respect and positive atti- tude, which made all the problems we faced feel surmountable. So, thank you for everything and especially for taking the lousy fisherman I am out and trying to teach me tips and tricks.

I would also like to thank Silva for her precious help and communicative good mood. Beside work-related discussions and advice, you helped me tremendously in moving to Turku even though things were not always easy, and that meant a lot to me. Finally, I’d like to thank my third supervisor, Pekka Hyvärinen. Coming to Pal- tamo was always fun and I will only have fond memories of the station (And also thanks to all the staff there!). Finally, I would like to acknowledge Prof. Arlinghaus, even though we did not have the chance to collaborate, he was instrumental in initi- ating the project.

I want to say a special thank you to the whole UEF Aquatic ecology group. You made the whole PhD experience enjoyable and made me quickly feel comfortable in UEF. From nice colleagues, many of you became friends. Aurora, Jenni and Nico, thanks for the justice league, the moral support, and also for NoWPaS (I still haven’t recovered the lost sleep-hours).

Even though I enjoyed Joensuu and its lovely climate, moving to Turku was a big step during the PhD. I would like to thank Craig Primmer’s group for welcoming me there and to the other people in Turku that helped me settling. Anti you taught me so much and were the closest thing to a fourth supervisor to me. I really enjoyed working with you even though you were always adding extra-work to what I thought was a finished thing. So, thank you for always pushing me to do better, but also for all the nice “quick chats” we had.

When I moved to Turku, I expected to find another Finnish city with its saunas and grillis. Little did I know that its University was an extension of France! Antoine, Barbara, Christie, Michaël, Robin, Sophie, Vérane and Yann (let’s extend to French- speaking) it was great sharing (very) long lunches and having passionate discussions

(16)

14

during hour-long coffee breaks with all of you. I will surely miss debates and songs, as well as Sunday’s brunches.

Leaving for Finland was a tough decision, but I am fortunate enough to have solid friendships. I never doubted that I would not lose contact with neither my old-school friends, neither with the ssg guys. I am grateful that even though you did not always understand my choices, you were always there for me without any question asked.

Special thanks to all those who were brave enough to visit me during my time in Finland – and to Bisco who helped with figures (I still like PowerPoint), posters and tried to make me less pessimistic!

My family also deserves a special thanks. I am definitely not the easiest person to keep in touch with and I freely admit that I am terrible at answering my phone (συγνόμη μαμά). Despite all that, you all were a rock upon which I could rely at all times and you always supported me through my crazy plans and ideas. Thank you for everything. Maybe you can admit πατεράκο, that finally, it arrived at the sea.

I wonder where I would now be if Margaux had not had the will to encourage me and follow me in Finland. Margaux you made all this possible and supported me during my whole PhD. You were there through the joy and struggle and pushed me into becoming a better version of myself. Thank you for helping me achieving my dream of becoming a biologist. I will always be grateful.

On a lighter note, I would like to thank many different entities that accompanied me through my entire PhD and helped me staying relatively sane: Paizo, Roti, Wotc, Yahoo fantasy, reddit, St Croix, Solary, Netflix, Hemingway, Sneak Attack, Peter Lahti, Kindle, Studio Reynier Gomez, NBA league pass, Alexandre Dumas, Louis Armstrong, and many more, thank you!

I also owe to acknowledge the Academy of Finland, which funded the whole project for four years. I also thank the two pre-examiners, Prof. Aspi and Dr. Nichols that agreed to review this thesis, as well as my opponent Prof. Wedekind.

Lastly, Sophia, you will probably never read this because it will be written by your boring dad, but you gave me the drive and motivation to finish my PhD. You also played your part by making me stay awake, thinking about how I should for- mulate many different times that yes, there is a genetic component in brown trout migration.

Turku, 10th April 2019 Alexandre Lemopoulos

(17)

LIST OF ABBREVIATIONS

DAPC Discriminant analysis of principal components DDRAD Double digest restriction site associated DNA DNA Deoxyribonucleic acid

FST Fixation index

GWAS Genome wide association study LFMM Latent factor mixed model

HYRAD Hybridisation restriction site associated DNA PCA Principal component analysis

PCR Polymerase chain reaction

RADseq Restriction site associated DNA sequencing SNP Single nucleotide polymorphism

UPGMA Unweighted pair group method with arithmetic mean

(18)

16

(19)

LIST OF ORIGINAL PUBLICATIONS

This thesis is based on data presented in the following articles, referred to by the Roman Numerals I-IV.

I Lemopoulos A*, Prokkola J*, Uusi-Heikkilä S, Huusko A, Hyvärinen P, Koljo- nen ML, Koskiniemi J, Vainikka A. (2019) Comparing RADseq and microsatellites for estimating genetic diversity and relatedness — Implications for brown trout conservation. Ecology and Evolution. 9: 2106–2120.

II Lemopoulos A, Uusi-Heikkilä S, Vasemägi A, Huusko A, Kokko H, Vainikka A. (2018) Genome-wide divergence patterns support fine-scaled genetic struc- turing associated with migration tendency in brown trout. Canadian Journal of Fisheries and Aquatic Sciences. 75: 1680-1692.

III Lemopoulos A, Uusi-Heikkilä S, Huusko A, Vasemägi A, Vainikka A. (2018) Comparison of migratory and resident populations of brown trout reveals candidate genes for migration tendency. Genome Biology and Evolution. 10:

1493–1503.

IV Lemopoulos A, Uusi-Heikkilä S, Hyvärinen P, Alioravainen N, Prokkola J, Elvidge C, Vasemägi A*, Vainikka A*. (In press) Association mapping follow- ing a common-garden migration experiment reveals candidate genes for mi- gration tendency in brown trout. G3: Genes, Genomes, Genetics.

* Equal contribution

The publications are reproduced at the end of the thesis with permission from their copyright holder

(20)

18

AUTHOR’S CONTRIBUTION

I) The author, along with co-authors contributed to designing the study. The author was responsible for extracting the DNA, preparing the RADseq librar- ies and performing the analyses on the RADseq dataset. The author helped writing and editing the final manuscript.

II) The author, along with co-authors contributed to designing the study. The author was responsible for extracting the DNA, preparing the RADseq librar- ies and performing all the analyses. The author also wrote the manuscript with help from the co-authors and was the corresponding author.

III) The author, along with co-authors contributed to designing the study. The author was responsible for extracting the DNA, preparing the RADseq librar- ies and performing all the analyses. The author also wrote the manuscript with help from the co-authors and was the corresponding author.

IV) The author, along with co-authors contributed to designing the study. The author extracted the DNA and performed the analysis on the genetic dataset.

The author wrote the final manuscript along with co-authors and was the corresponding author.

(21)

1 INTRODUCTION

Animal migrations have fascinated generations of scientists. Vast movements of un- gulate hordes, seasonal flight of birds across the globe or salmon homing runs are among the most famous displays of such behaviour. These phenomena exist in almost all branches of animal kingdom as they have evolved in several taxa. As such, crustacean, mammals, insects, birds and fish all present some kind of seasonal movements, identified as migrations (Alerstam et al., 2003; Dingle, 2006). The motivation behind these migrations differ between species. Moving to wintering areas, seasonal grazing or search for food (Dingle & Drake, 2007; Dingle, 2014; Jones et al., 2015) are different examples underlying these movement patterns. The evolution of migration is not phylogenetically constrained and has evolved independently across organisms to maximise fitness in ecologically heterogeneous environments (Gross et al., 1988;

Alerstam et al., 2003; Dingle, 2006, 2014).

Migrations are thought to be induced by a combination of proximate, i.e. environmental factors, changes in physiology, morphology and/or behaviour of species (Dingle, 2006; Dingle & Drake, 2007). Genetic components underlying these processes have been discussed in several studies and are nowadays being more and more investigated at the species level (see Liedvogel et al. 2011 for a review). Salm- onid fishes, i.e. salmonids, are among the most famous and economically important migrating animals (Harris & Milner, 2004; Dodson et al., 2013). The origin of their migratory behaviour is a long-standing question (McDowall 1997; McDowall 2002).

Indeed, even though salmon, trout and char derived from a common ancestor, they represent distinct evolutionary lineages, and all perform feeding migrations from rivers to larger water bodies. A recent study (Alexandrou et al., 2013) presented, through phylogenetic reconstruction that such behaviour probably evolved several times within the family (parallel evolution) rather than occurred from a single migratory ancestral organism.

Migratory salmonids typically start from streams and migrate to either saltwa- ter environment (anadromous migration) or to freshwater environment in lakes (adfluvial migration) or larger section of rivers (potamodromous migration). Such a life- history involves a complex adaptation to new environments. Prior to migration, when they are two or three-year-old, some juvenile salmonids, named parr, undergo a major morphological and physiological transformation called smoltification (Folmar & Dickhoff, 1980; McCormick, 2009; McCormick et al., 2013). During this transformation, parrs actively prepare for the upcoming migration. The most appar- ent transformation resides in the changing of body coloration, as migratory fish will adopt a silver colour, best used for camouflage in larger water bodies. Migratory fish will also change shape, becoming slenderer and their snout will get more pointed (Folmar & Dickhoff, 1980; Hard et al., 1999). Physiological changes also occur with fish growing muscle and changing metabolism and tolerance to salinity (e.g. Hoar

(24)

22

1976; Zaugg 1982; Nielsen et al. 1999). After smoltification, the transformed parrs are called smolts and are ready to migrate. After migrating and spending a period at sea or lake, varying from less than one year to more than eight (L’Abee-lund et al., 1989), fish perform a spawning migration, called homing, and move back to their natal rivers to reproduce. Iteroparous salmonids return the feeding areas after spawning, and whether this migration reenacts all aspects of smoltification is not known.

Some salmonids break the standard rule and adopt a partial migration life-cycle (Chapman et al., 2012a, 2012b; Dodson et al., 2013). Individuals of these species can follow either a migratory, or a resident life-history. Resident individuals never undergo smoltification and thus do not migrate, staying in their home waters for their entire lifespan. Populations can range from strictly resident to entirely migratory, and ecotypes (migratory or resident) can thus live in sympatry or allopatry. The mechanisms driving the different life-histories have been extensively studied and the life-history dichotomy is believed to arise from several causes as a combination of environmental factors and genetic interactions are known to play major roles in the decision to migrate (for reviews see Kendall et al. 2015; Ferguson et al. 2017).

The environment can influence the decision to migrate in different phases of on- togeny, and as soon as during the embryonic stage (Jonsson & Jonsson, 2014). Indeed, environment directly impacts the fish condition, which is key for the migration decision (Forseth et al., 1999). The growth and the size of the fish have been found to influence migration decisions in, for instance, brook charr Salvelinus fontinalis (Theriault & Dodson, 2003; Thériault et al., 2007) and Oncorhynchus mykiss (Hecht et al., 2015). However, studies have found contradictory results, with both fast and slow growth seemingly promoting the migratory life-history in brown trout (Bohlin et al., 1996; Olsson et al., 2006). Food availability in contrast is known to significantly pro- mote migration, as brown trout parrs will smoltify more often when food levels are low (Olsson et al., 2006; Wysujack et al., 2009; Jones et al., 2015).

Moreover, high fish density is also enhancing fish migration, as competition for resources increases (Jonsson & Jonsson, 1993; Morita et al., 2000). Abiotic factors such as low water flow (Mills et al., 2012; Berejikian et al., 2013) or temperature are also driving individuals to migrate. Specifically, temperature seems to be one of the most important environmental drivers. For instance, both high temperatures and/or a sud- den increase can trigger smoltification and thus migration in rainbow trout Oncho- rynchus mykiss (Sloat & Reeves, 2014; Kendall et al., 2015).

In addition to the different environmental factors, genetic factors are also deter- minant for the life-history strategy dichotomy. Translocation experiments showed that the migratory behaviour was heritable in brown trout Salmo trutta (Skrochowska, 1969; Näslund, 1993). Since then, many other studies confirmed these findings by showing similar parental-offspring life histories in different salmonids such as brook charr Salvelinus alpinus (Thériault et al., 2007) or Onchorynchus mykiss (Hecht et al., 2012; Hu et al., 2014). However, genetic differentiation between sympatric ecotypes of brown trout has not been observed in various populations (e.g.

(25)

Hindar et al. 1991; Charles et al. 2005) challenging the claim of an existing genetic influence over the migratory behaviour for this particular species.

Over the recent years, with the popularisation of high-throughput DNA sequencing techniques, genetic and genomic studies focusing on salmonids migrations have brought new evidence to the genetic origin influencing the migration life-history strategies. In particular, different species of Pacific salmon have been intensively studied. In particular, Oncorhynchus mykiss species complex has received the most attention. Many candidate genes for the propensity to migrate have been proposed (Nichols et al., 2008; Hale et al., 2013; Hecht et al., 2013). The functions of these genes associate with different biological pathways such as circadian rhythms (Arostegui et al., 2019), osmoregulation processes, brain function (McKinney et al., 2015) and development or immunity state (Sutherland et al., 2014). While many of these genes seem to have functional importance, one major region in chromosome 5 (Omy5) in- fluences the behaviour at the species level (Nichols et al., 2008; Pearse et al., 2014;

Leitwein et al., 2017). This “supergene”, maintained through sex-specific dominance, thus establishes the genetic basis for rainbow trout migration. However, it also displays geographical frequency association (Pearse et al., unpublished), making population effects still remarkable (Hale et al., 2013). Another partially migrating species, Sockeye salmon Onchorynchus nerka has also been studied at the genomic level, and candidate genes have been proposed to establish the genetic basis of the migratory behaviour of O. nerka (Nichols et al., 2016; Veale & Russello, 2017)

Brown trout (Salmo trutta) was first described by Linnaeus in 1758. It is part of the Salmonidae family (Crête-Lafrenière et al., 2012) and is one of the most widespread and studied salmonids. Its natural distribution spans across Europe but brown trout has nowadays been imported in Africa, Asia, Australia and in the Americas (MacCrimmon & Marshall, 1968). Brown trout is a very diversified species, and an extensive number of lineages exist. Indeed, within the native range of brown trout Bernatchez (2001) identified five major genetic lineages: Atlantic, Danubian, Mar- moratus, Mediterranean and Adriatic. As brown trout presents a rich evolutionary history, it also displays a large variability. Populations, but also different sympatric ecotypes, can greatly vary in size, morphology or colour but also in feeding behaviour or migration propensity. Therefore, brown trout is considered as a polytypic, diverse and plastic species.

Brown trout display different migration-related morphs, or ecotypes. Salmo trutta fario is the resident trout morph (Fig.1). The mature size often ranges from 25 to 50 cm and a brown colour with red spots is typical of this morph. It usually feeds on insects, but bigger individuals can show piscivorous behaviour. Salmo trutta lacustris or Salmo trutta trutta are the migratory morphs of brown trout (Fig. 1). While lacustris

(26)

24

migrates to larger freshwater bodies (adfluvial), trutta migrates to the sea (anadro- mous).

Figure 1. Brown trout ecotypes. On the left a migratory trout and a resident on the right. Pictures by A. Vainikka

Brown trout populations often show remarkable genetic structures within watersheds (e.g. Carlsson and Nilsson 2000; Griffiths et al. 2009; Swatdipong et al. 2010).

Brown trout often show large genetic variation among populations and presents rapid evolutionary rates of trait change and adaptation to environmental challenges (e.g. Sanz et al., 2000; Griffiths et al., 2009; Stelkens et al., 2012); hence making it a compelling organism model for studying adaptation and population genetics. How- ever, studying signatures of selection among populations and different life-history types has only seldom been done (Leitwein et al., 2016; Wollebaek & Heggenes, 2018).

In contrast to other salmonids, brown trout migrations had not been extensively studied at the genomic level prior to my work. As a result, little was known about the genomic causes and consequences of migratory behaviour in brown trout. It was therefore necessary to first examine if migratory and resident populations differ genetically in systematic ways.

Brown trout is of major socioeconomic value for recreational fisheries in western countries. In Finland, but also across Europe, it is intensively farmed for stocking and to a minor extent for food. However, brown trout faces several threats from anthro- pogenic factors (Syrjänen et al., 2017). Migratory trout is confronted to threats such as dam building, overfishing (ICES, 1994, 2006; Harris & Milner, 2004) or water pol- lution (e.g. Paris et al. 2015; Marques et al. 2019). In addition, the general warming of water is also a major threat to native populations (Hari et al., 2006; Poulet et al., 2011;

Réalis-doyelle et al., 2016). In Finland, most southern brown trout populations are endangered (with no regards to migratory or resident populations), mainly due to human activity. To address this and respond to the large demand from recreational fishers, stocking of hatchery trout is regularly performed in Finland. Therefore, subsequent hybridisation with exogenous and stocks that may have accumulated unintended domestication effects is yet another threat to which native populations are exposed. Such introgression cannot only harm native populations through, for instance, the loss of local adaptations, but it could potentially also significantly alter

(27)

the life-histories of such populations. Therefore, an extensive knowledge of the ori- gins of migratory vs resident life-histories is essential for the planning of sustainable stocking practices. Such knowledge could help providing guidelines for the conservation of this iconic migratory species.

1.1 AIMS OF THE THESIS

The aim of my PhD-thesis was to investigate the genetic components in the propensity to migrate in brown trout. Understanding the origin and evolution of key life- history traits is compelling for any evolutionary biologist, let alone for ichthyologists.

As such, this project was relevant for anyone interested in fundamental processes of behavioural ecology, evolution, genomic research and, in general, fish biology.

In chapter I, I wanted to understand how microsatellite genotyping and RADse- quencing differ in terms of performance, and which method would be more suited for studying signatures of selection. In this comparison, we used Oulujärvi populations as example material while simultaneously providing the key baseline population genetic information for the experimental work in the overall research project. In chapter II, I wanted to assess whether genetic structure existed between main stems of rivers and their tributaries, and if this structure could reflect potential differences in the populations’ life-history strategies. In chapter III, the aim was to specifically compare migratory and resident trout to reveal genes potentially linked with the migratory life-history dichotomy. Finally, in chapter IV, I aimed to strengthen and ex- pand the previous findings on brown trout migratory behaviour by performing an association study using a multi-year common garden experiment.

The applied aim of my thesis was to provide genetic information for the management of the studied populations. Understanding the genetic background for different brown trout migratory ecotypes is also important for any conservation plan consid- ering small remaining resident populations. As only a few strains of migratory hatchery brown trout are used across Finland to enhance local populations, it was also important to address the potential life-history consequences of this stocking practise.

Understanding the implication of genetics into migratory life-history dichotomy would allow a better management of resident and migratory trout stocks. Is each fish stock unique in its propensity to migrate, and is partial migration explained by individual-level genetic differences? Could we potentially use the last remaining wild resident stocks to replenish migratory populations or recover the fitness of hatchery- reared migratory stocks suffering from unintended domestication? These were the general questions needing answers.

(28)

26

(29)

2 MATERIALS AND METHODS

2.1 STUDY AREAS

2.1.1 Koutajoki watershed

The upper River Koutajoki watershed is located in north-eastern Finland (Fig. 2). It consists of three river basins, the Kuusinki, Kitka and Oulanka. The Kitka basin is split by a migration barrier, the Jyrävä falls, separating the main stem in two (Kitka

“above falls” and Kitka “below falls”). These basins are consisting of main stems to which several tributaries discharge into. Apart from the Jyrävä falls total barrier and the Oulanka river’s Kiutaköngäs waterfalls partial barrier, all the three rivers are in- terconnected and free-flowing. The main stems eventually flow into the Oulanka river that discharges into Russian Lake Panozero (Paanajärvi in Finnish), and from it, to Lake Pyazero (Pääjärvi in Finnish) through the river Olanga. Finally, the lake Pyazero eventually empties into the White Sea, through the (dammed) river Kouta- joki.

Large-growing, heavily river-fished, iconic brown trout in this system have been monitored and studied for decades (Huusko et al., 2017). Several radio-tagging studies have shown that Lake Pyazero is the main feeding area for migratory trout origi- nating from Kuusinki, Kitka “below the falls” and Oulanka rivers (Huusko et al., 1999; Saraniemi et al., 2008). In contrast, due to the natural migration barrier created by the Jyrävä falls, migratory trout from above the falls migrate upstream to lake Kitka. The origin of the brown trout populations is unsure, but it is thought that the

“above the fall” trout originate from an ancient Baltic Sea population (Heikkinen &

Kurimo, 1977; Koutaniemi, 1999)

The region is partially part of the Oulanka national park in Finland, and the Paanajärvi national park in Russia. Therefore, human impact in the near-river environment is minimal, and rivers and environment are mainly untouched with the ex- ception of intensive recreational river fisheries that take place both in Finland and in Russia. The estimated economic value of the recreational brown trout fisheries in Kuusamo area is about 1-2 million euros (Kuosku et al., 2014).

2.1.2 Oulujärvi and Kainuu fisheries research station

Lake Oulujärvi is found in northern central Finland (Fig. 2). The watershed consists of one large lake (928 km²) and several rivers flowing into it. The watershed itself is connected to the Bothnian sea by heavily dammed former salmon river Oulujoki.

Both the Hyrynsalmi and Sotkamo watercourses flowing to Lake Oulujärvi are heavily dammed without fishways, and their original migratory brown trout and salmon stocks have died out. The small undammed but often stretched rivers are inhabited

(30)

28

by different brown trout populations, supposedly resident and often original. In contrast, lake Oulujärvi is stocked with migratory trout, coming often from private fish farms and representing non-native stocks. The original brown trout stock collected from Rivers Varisjoki and Kongasjoki is maintained by Taivalkoski hatchery of Nat- ural Resources Institute Finland and partially by Kainuu Fisheries Research station (www.kfrs.fi) in Paltamo. The contemporary hatchery strain was founded in October 2000 by crossing two previous hatchery stocks coming from two neighbouring rivers (Varisjoki and Kongasjoki). One of these stocks was originally founded in the 1960- 1970’s, while the other was founded in the 1970-1980’s.

Kainuu Fisheries Research station is a research station focusing on aquatic ecology. It is part of the Natural Resources Institute Finland (LUKE) and is found in Pal- tamo north from Lake Oulujärvi by the River Varisjoki. It offers different types of facilities, permitting researcher to use various strains of fish as well as performing small and large-scale experiments.

Figure 2. Two watershed locations in Finland. On the top right, the Koutajoki watershed where 6 populations were sampled: 3 residents in tributaries (T) and 3 migratory in main stems (MS.

On the bottom right, the Oulujoki watershed where three resident populations (R) and one migratory populations (M) were sampled.

(31)

2.2 MOLECULAR TOOLS

2.2.1 Extracting the DNA

In any genetic analysis, the first step is to extract DNA from samples. These can be hairs, saliva, blood or any tissue containing DNA. In ichthyology, the most common tissues for the extraction of DNA are small pieces of fin (usually caudal, pectoral or adipose) or dried scales. All the extractions performed during my PhD work were done using commercial kits from either Omega Bio-Tek (E.Z.N.A.® Tissue DNA Kit) or Macherey and Nagel (NucleoSpin® Tissue).

The protocols were similar and were performed according to manufacturers’ in- structions. The extraction consisted of one overnight digestion phase using a protein- ase. For pigmented fin tissues, an additional step was added at this point to remove the disturbing effect of melanin. I added a solution of 1:24 chloroform and centri- fuged the samples for 5 minutes. I then removed the pigments from the solution by selecting only the upper phase of the solution and ligated the DNA to micro-columns.

Afterwards one or two washing steps were performed using washing buffers, before the final elution of the DNA into a warmed (70°C) elution buffer. DNA was then stored at -20°C for future use.

Obtaining good quality DNA is not always trivial, and individual assessment must be performed after the extractions. The quality checks were performed with a fluorometric measurement using the Qubit 2.0. This allowed us to estimate the con- centration of the DNA in our samples, and therefore proceed to the next steps, be it for subsequent microsatellite sequencing, or library preparation for RAD sequencing (see below).

2.2.2 Microsatellites (I)

Microsatellites are short tandem repeats (from 1-10bp) found randomly in the DNA.

They present high mutation rates and are usually highly polymorphic. Because these markers represent assumedly neutral variation, they have been used in many studies to investigate, for instance, genetic structure of populations (Moore et al., 2014), kin- ship analyses (Carlsson, 2007) or heterozygosity levels (Carlsson et al., 1999). These markers have been extremely popular for over three decades and the microsatellite body of work is extensive (e.g. Carlsson and Nilsson 2000; Pearse et al. 2009;

Swatdipong et al. 2010). However, analyses based on microsatellites also have vari- ous limitations such as the presence of null alleles, insufficient polymorphic loci or the need for previous labour (i.e. developing a reliable panel) for them to be efficient (e.g. Zhang and Hewitt 2003; Putman and Carbone 2014). Moreover, as they only convey information about neutral evolution, obtaining information about functional traits is generally impossible when using microsatellites.

(32)

30

2.2.3 Restriction site associated DNA sequencing-RADseq (II-IV)

Restriction site-associated DNA sequencing (RADseq) was first introduced by Baird et al. (2008). Over the last decade, this method broke out and revolutionized the fields of ecological, evolutionary and conservation genomics (Andrews et al., 2016). The method underlying RADseq is fairly simple (Fig.3). Restriction enzymes cut the DNA strands at different restriction sites (depending on the enzymes, the most common being 6’ or 8’ cutters). Millions of short fragments (~100bp) are then obtained, and adaptors as well as individual barcodes are ligated to the fragment ends. These fragments are then amplified through polymerase chain reaction (PCR) and purified (and size-selected when necessary). Afterwards, the fragments are sequenced, and millions of individually barcoded reads are obtained. These are then processed and compared for similarity, and single nucleotide polymorphisms (SNPs) can be detected from these comparisons, i.e. called for further analyses through bioinformatics pipelines (see below).

The RADseq method has seen an astonishing expansion, and many protocols have been developed with subtle variations from the original method. As such, ez- RAD (Toonen et al., 2013), ddRAD (Peterson et al., 2012) or hyRAD (Suchan et al., 2016) are only some examples of RAD-derived techniques used in different studies.

They all present different advantages and limitations (see Andrews et al. 2016), and the study questions along with the available resources should indicate which method is optimal before performing any study. In our project, we used both ddRAD and the original RADseq protocols. The main criteria included local expertise, budget limitations, schedule restrictions but also factors such as the desired number of markers, the number of individual samples available or the previous literature (Puritz et al., 2014).

RADseq is a powerful method for working on non-model organisms such as brown trout. It is particularly interesting as no previous labour, such as primer opti- mization, is needed. Therefore, working on any given organism is possible with only small adjustments to the original protocol. In addition, RADseq opens a realm of study possibilities. Notwithstanding traditional genomic analyses (population structure, heterozygosity levels…), RADseq also allows to investigate markers for signature of selection or for genotype-phenotype association. As such, genome scans can be executed on RADseq derived data while understanding the limitations of the low coverage of the sequence data. As such, functional genetics underlying ecologically important traits can be quickly explored through this method.

(33)

Figure 3. RAD sequencing methods.

2.3 BIOINFORMATICS METHODS

2.3.1 RAD-Reads processing

After sequencing, RAD-reads must be cleaned from sequencing errors and the sequences quality checked. As each library contains up to 96 individuals (but often less, depending on the desired coverage), the reads must first be demultiplexed. Each read possesses a barcode (4-6 bp) that is assigned to a unique individual in each library.

The demultiplexing step consists of assigning each read - based on barcodes - to its individual. Afterwards, the sequencing quality is checked (mainly if some bases are ambiguous). Once demultiplexing is performed and the quality assessed, cut-sites and barcodes sequences are trimmed from the reads and one single file (fastq), containing all the reads for a given individual, is obtained.

2.3.1.1 De novo assembly

One of the incentives of RAD sequencing is that it allows working on non-model organisms. Therefore, no previous knowledge, such as available genome or primers, is needed before undergoing sequencing and processing the obtained sequences. As millions of small reads are obtained for each individual, it is necessary to align the

(34)

32

reads corresponding to similar locations among individuals. Since brown trout ge- nome has not been published yet, a de novo catalog needs to be built in each study.

The catalog is created using a high number of individuals from the sampling design and represents a database against which the reads can be aligned. Reads from each individual are then mapped according to sequence similarity against the catalog. Af- terwards, based on the catalog, SNPs can be detected, i.e. called according to various parameters such as a given number of individuals or populations where the locus has to be present. The main difficulty when building the catalog is avoiding any type of artefactual reconstruction. The catalog needs to be fine-tuned according to each dataset respectively. Many phases of testing need to be performed to find suitable parameters for each study (Shafer et al., 2017; Leary et al., 2018). For catalog building, the number of mismatches allowed between reads and between stacks are important, as well as the minimum coverage needed for each read to be processed. Moreover, key metrics such as minor allele frequencies (Roesti et al., 2012), maximum individual heterozygosities (Hohenlohe et al., 2011) or population and individual percentages of shared loci are important for optimal SNP calling. The most commonly used program nowadays is Stacks (Catchen et al., 2013). The guidelines into a suitable usage of the program are well documented, and correct usage is strongly advised to avoid any potential bias (Mastretta-Yanes et al., 2015; Paris et al., 2017; Rochette & Catchen, 2017). Stacks has been used in hundreds of studies with a diversity of organisms (e.g.

Keller et al. 2013; Benestan et al. 2015; Recknagel et al. 2015) and, while having re- ceived some criticism (Puritz et al., 2014), currently remains the “go-to” option for any RAD sequencing study.

2.3.1.2 Using a reference genome

When a genome (or genome from a closely related organism) is available, there is no need for creating a de novo catalog. The genome can instead be used to map the reads and call the SNPs. Atlantic salmon (Salmo salar) is the sister species of brown trout and is therefore its closest related species with an annotated genome (Crête- Lafrenière et al., 2012). Its genome has been recently published (Lien et al., 2016) and is a useful resource for many salmonid researchers. In addition to its availability, it is also well annotated and allows inferring the functions of genes that contain or are closely located to the identified SNPs. In order to align the reads, mapping software need to be used before SNP calling through Stacks (Catchen et al., 2013) or another software such as Angsd (Korneliussen et al., 2014). The most commonly used map- ping software is bowtie2 (Langmead & Salzberg, 2013). For each individual, bowtie2 maps the reads to an index, created from a chosen reference genome (here Atlantic salmon). Reads can then be aligned and SNPs called in the same manner as in the de novo analysis.

(35)

Using a reference genome can be very advantageous compared to the de novo ap- proach. Using a genome reference prevents erroneous or artefactual catalog recon- struction than can occur when using a de novo approach (e.g. by unsuitable number of mismatches allowed between reads, see Puritz et al., 2014; Mastretta-Yanes et al., 2015; Leary et al., 2018). However, other concerns can arise. As Atlantic salmon has diverged from brown trout a long time ago, it is expected that only a portion of the reads can be aligned to its genome. In addition, retaining only uniquely aligned reads can be even more challenging.

Salmonids underwent a series of genome duplications and present residual tetra- ploidy in their genome (Johnson et al., 1987; Sutherland et al., 2016). As such, parts of the genome can be tetraploid and the reads can therefore be aligned in multiple locations of the genome. Therefore, we needed to make sure to only retain uniquely aligned reads by keeping only the highest mapping scores for every single read, before performing the SNP calling.

2.4. Population Genomics (I-II)

We evaluated the population structure present in two different watersheds (Oulujärvi I, Koutajoki II). The genetic diverges between sampling sites were tested with different methods based on 16 microsatellites (I) and thousands of SNPs obtained through ddRADseq (4876 and 3972 respectively). Discriminant principal component analyses (DAPC, Jombart et al. 2010) were used as explorative method to uncover basic population structuring (I-II). In addition, individuals were reassigned to populations using principal components (II). Population structures were also unrav- elled using Bayesian clustering methods (I-II). To this end, we used the Evanno method (Evanno et al., 2005), to find the theoretical optimal number of clusters (K) present in the data and then used the STRUCTURE software (Pritchard et al., 2000;

Falush et al., 2007) to reveal the population structure. We also reconstructed a phy- logeographic tree using an unweighted pair group method with arithmetic mean (UPGMA) clustering algorithm (II). This allowed to decipher the relationships between the rivers and assess the genetic proximity between each population. Individ- ual relatedness and family structures were also assessed, and effective population sizes estimated based on linkage disequilibrium estimation. These measures were obtained using COLONY (Jones & Wang, 2010) and Nestimator (Do et al., 2014).

In chapter I, we used three wild brown trout populations along with one hatchery strain to understand how different markers performed. We compared the two datasets using two different approaches, the first with more individuals genotyped for the microsatellite than the SNP panel (120 vs 75 individuals) and the second with the same number of individuals for the two datasets (75 individuals for both). In chapter II, we used 16-30 individuals per sampling site for a total of 11 locations separated in tributaries and main stems.

(36)

34

2.5 Association analyses (III-IV)

2.5.1 Genome scans and Environmental analysis (III)

Local adaptation and signature of selection can be detected through genome scans and environmental association analyses (e.g. Berg et al. 2015; Babin et al. 2017). The aim of these analyses is to detect interesting markers that are significantly different between studied populations and which might associate with given environmental variables or phenotypes. These markers are often deviating from a given assumption (i.e. neutrality) and are thus called outliers. The deviation from neutrality implies a non-neutral inheritance of the markers and can then represent local adaptation or signatures of selection (Kirk & Freeland, 2011; Ahrens et al., 2018). Thus, outliers can be significantly associated with the environment (e.g. Berg et al. 2015; Babin et al.

2017; Pritchard et al. 2018) or even phenotypes (e.g. Colosimo et al. 2005; Keller et al.

2013; Veale and Russello 2017) and the outliers are often referred to as “candidate”

loci or genes. Identifying candidate markers through genome scans and environment association analysis is a powerful method, but the detected outliers can often arise from false positives and/or negatives, and also from confounded characteristics, rather than the environment itself (Ahrens et al., 2018). To address this issue, it is often recommended to perform multiple scans and to cross-validate the results.

In chapter III, we studied two different watersheds with brown trout populations from main stems (migratory) and tributaries (resident). We hypothesized that the differences between these populations would associate with the phenotypic differences of the ecotypes. To reveal markers associated to main stem environment (and thus to migratory brown trout) and to tributaries (and thus to resident brown trout), we performed four different analyses. Two of them were genome scans and two of them environmental association analyses. Bayescan (Foll & Gaggiotti, 2008) identifies markers with FST values departing from assumed neutrality, while PCAdapt (Luu et al., 2016) detects outlier markers in a multivariate analysis context (PCA). Latent Factor Mixed Modelling (LFMM; Frichot and François 2015)) is an environmental association analysis that identifies outliers using a Bayesian framework. Finally, BayeScEnv (Villemereuil & Gaggiotti, 2015) is based on FST and environmental variables to identify outliers. Correction for multiple testing (False discovery rate, FDR) was applied for all tests: genome scans outliers were identified under a 0.05 threshold and environmental analyses under a 0.1 threshold. To reduce potential Type I and II errors, we only considered markers identified by at least three of the four methods.

In chapter III, we used the results from the two previous chapters to design our study. Since population life-histories were assessed in the previous studies, we took the four populations from study I (three resident vs one migratory) and six populations from study II (three resident vs three migratory) and compared each watershed separately to uncover potential candidate markers for migratory vs resident life history strategies. In contrast to the other studies, we mapped the previously sequenced

(37)

RAD reads to the Atlantic salmon (Salmo salar) genome. We kept uniquely aligned reads with the best alignment scores and obtained two datasets of 5519 and 5670 SNPs, respectively.

2.5.2 Genome-wide association studies (IV)

Genome-wide association studies (GWAS) aim to understand the links between genotypes and phenotypes. They were initially developed for medical studies (Ku et al., 2010), as case-control studies took place and the genetic origin of many disease was discovered through an impressive body of work (e.g. Rivas et al. 2011; Chung et al.

2013; Demontis et al. 2019). Such studies revolutionized our understanding of complex diseases and the way we could potentially cure them and represented a major advance in the genomics field.

Molecular ecology studies recently caught up on the association study trend. In- deed, with the increased availability of next generation sequencing, studying the origin of key life-history traits has become possible. Understanding of the mechanisms underlying important traits is extremely interesting from an evolutionary standpoint, but it could also improve the management and conservation of species of interest. Among the breakthrough studies that emerged from ecological GWAS, some focused on economically important salmonids, such as Atlantic salmon Salmo salar (Johnston et al., 2014; Barson et al., 2015; Gutierrez et al., 2015) and Pacific Salmon Onchorynchus mykiss (Hale et al., 2013; Hecht et al., 2013; Prince et al., 2017).

Indeed, the Atlantic salmon age at maturity trait, investigated by Barson et al. (2015) is extremely important in an aquaculture context. Moreover, the migration propensity (Hale et al., 2013; Hecht et al., 2013) and timing of Pacific salmon migration (Prince et al., 2017) are also key aspects in the life-history of the valued Pacific salm- ons.

Linking genotypes to phenotypes can be particularly challenging. Studied traits are often confounded with population structure, and signals of association can therefore be buried among markers differing between populations through neutral pro- cess, such as genetic drift. It is therefore important to account and remove population stratification (to remove neutral variation) when performing association studies.

Therefore, GWAS ideally need to have large sample size to increase statistical power and several independent populations which segregate independently from the studied traits. In addition, GWAS needs an extensive coverage of the genome, thus a large number of markers, in order to identify potentially interesting markers. The imper- fect coverage itself creates a challenge because of the increased chances of type II error. Correcting for multiple testing inherent in testing thousands of individual markers is an important, even though very conservative, step of GWAS to avoid type I error.

(38)

36

Another characteristic of GWAS is their sensitivity to the used pipeline (Torkamaneh et al., 2016) and the effect of even slight genotyping differences on the association results (Hong et al., 2010). To address these issues, we used two pipelines and association methods to identify putative candidate genes for migration propensity in brown trout. One pipeline consisted of SNPs called through Stacks (Catchen et al., 2013) and the GWAS performed using a mixed linear model in GAPIT (Lipka et al., 2012). The other pipeline consisted of SNPs and GWAS called and performed in Angsd (Skotte et al., 2012; Korneliussen et al., 2014). In both cases, population strat- ification was corrected using the first principal component as a covariate. Candidate markers were identified as the overlapping markers in the top 0.1% of each of the two pipelines.

Here, we performed a GWAS to identify markers associated with migration distance on a common garden experiment. Brown trout from two distinct populations (see Chapters I and II) were raised in common conditions and placed in eight circular concrete channels equipped with four antennas at equal intervals. The fish were individually tagged and their movements within the streams were recorded over a two-year period. A linear mixed model was fitted to explain the total distance swam by each individual to account for various environmental effects (sex, tank, year). The residuals from this model were kept to represent intrinsic differences among individuals. These residuals were averaged over years and used as proxies of migratory behaviour in two different GWAS. Based on the 116 sequenced individuals (58 from each population, selected to represent the most extreme phenotypes) we used two different bioinformatic pipelines and obtained 24330 SNPs (Angsd) and 18264 SNPs (Stacks) that were mapped against the Atlantic salmon genome to retrieve potential candidate markers for the migration propensity.

(39)

3 RESULTS AND DISCUSSIONS

3.1 DIFFERENCES IN MARKER EFFICIENCY AND EVALUATION OF GENOMIC DIFFERENCES BETWEEN OULUJÄRVI POPU- LATIONS (I)

Genetic diversity and effective population sizes were roughly similar between all the comparisons, with the hatchery strain presenting higher level of heterozygosity and allelic richness and all populations displaying very small effective population sizes (Table 1). While family structures were similar between datasets, the estimations using the SNP dataset were more trustworthy, as shown by higher assignment proba- bilities. Individual heterozygosities showed some correlation between datasets, yet the individual variance was much higher when using the microsatellite panel.

Table 1. Genetic diversity and Ne in all studied populations across the three datasets. Ex- pected heterozygosity (He) and effective population size (Ne) shown as measured from microsatellites in 120 individuals (A) and RADseq (B) and microsatellites (C) on the same 75 individuals. LD-based estimates of Ne are shown with 95% non-parametric jackknifed confi- dence intervals.

The 4 populations were genetically different from each other, presenting moderate FST values that were similar across datasets, even though slightly higher when using the microsatellite data of 75 individuals. DAPC showed similar patterns when comparing the microsatellite complete dataset (120 individuals) to the SNP dataset.

However, when using microsatellite genotypes of 75 individuals, differences arose in the patterns, as the DAPC did not discriminate the Tuhkajoki and Oulujärvi populations. In contrast, the Bayesian analysis yielded 4 different clusters (K=4) in all situations with very low level of admixture between populations.

A B C

He Ne He Ne He Ne

Pohjajoki 0.53 14.8 (6.9-34.7) 0.09 13.0 (6.3-28.3) 0.48 9.7(2.8-73) Tuhkajoki 0.61 22.0 (12.4-46.3) 0.1 2.10 (2.9-27.4) 0.58 10.1 (4.4-28.8) Vaarainjoki 0.59 32.8 (11.5 - ∞) 0.11 24.5 (12.3-123.8) 0.59 28.8(10.1-∞) Hatchery stock 0.66 53.0(30.8-133) 0.14 96.5 (60.8-149.0) 0.66 57.1(31.2-316.8)

(40)

38

Overall, the study yielded three main results. Firstly, a well-established microsatellite panel can perform as well as several thousands of SNPs in estimating population differences. Second, multivariate analyses perform better with SNPs than microsatellites when using a small sample size. Finally, in order to estimate individual heterozygosities, thousands of SNPs are necessary to obtain reliable results.

While RADseq is a more modern and potentially more powerful method, we here showed that microsatellites are often enough for estimating population structure, genetic differentiation and effective population size. For conservation plans, these results are valuable. Management of populations does not always require expensive genotyping such as RAD sequencing, and resources could thus be better invested by genotyping individuals with microsatellites. However, when management requires

Figure 4. Two analyses of population structure of four brown trout populations. On the left a principal component analysis, and on the right a Bayesian reconstruction through Structure software; using a) 120 individ-

uals and 16 microsatellites b) 75 individuals and 4876 SNPs

and c) 75 individuals and 16 microsatellites.

(41)

measures such as individual level of heterozygosities, then RAD sequencing is a much more reliable source of information. For studies and conservation plans based on metrics such as heterozygosity fitness correlation, RAD sequencing is therefore more adequate. Moreover, in the case of small, vulnerable population (such as the studied brown trout populations), sampling large number of individuals is not always feasible, and RAD sequencing is a better alternative than microsatellites with small samples sizes. However, in the scope of my thesis, RAD sequencing was also a better alternative due to the nature of the obtained markers. Indeed, in contrast to microsatellite that only represent neutral variation, some SNPs can potentially bear functional adaptation. Therefore, different ecologically relevant analyses can be performed for SNP datasets. For instance, genome scans and genome-wide analyses were key methods in my thesis, thus leading us to choose RAD sequencing rather than microsatellites for studies II-IV.

This study also yielded some interesting biological information in addition to the technical details. The 4 studied populations corresponded to three wild and one hatchery population. The hatchery population is migratory whereas the wild populations display a tentative resident life-history strategy. Both Bayesian and multivariate results showed four different clusters corresponding to totally genetically isolated populations, despite the lack of physical barriers between the sampling sites (except for the River Tuhkajoki that is beyond dams). These genetic differences also showed that no- or very limited introgression occurs from hatchery to wild populations (just one pure hatchery fish was found among Vaarainjoki fish). Also, since the life-history differences were linked with high population divergence, we were able to compare these populations in a genome scan context (see chapter III).

3.2 POPULATION GENOMICS OF THE BROWN TROUT IN KOUTA- JOKI WATERSHED (II)

After filtering, we obtained 3976 SNPs that we further used to investigate the population genomics of brown trout collected in 11 different sampling sites. These sites corresponded to 7 tributaries and 4 main stems. Interestingly, the sampling sites also corresponded to putative ecotypes, with tributaries supporting mainly resident trout while main stems were inhabited by migratory trout (identified through radio-tagging studies, see Saraniemi et al. 2008; Huusko et al. 2017)

Fish from tributaries showed lower genetic diversities compared to main stem fish and, in general, also lower effective population sizes. FST-values were moderate and tributaries individuals were not found closer to their discharging main-stem individuals compared to others. Fish from one main stem (Kitkajoki-above falls) showed a slightly bigger differentiation from the others.

(42)

40

The multivariate and structure analyses showed a pattern where tributaries individuals represented more isolated and unique genotypes whereas main stems individuals presented more admixture and were more difficult to separate from each other (Fig. 5).

Figure 5. Discriminant principal component analysis of 3972 SNPs. Four tributary populations stand out from the others. (KUU = Kuusinkijoki; Raa = Raatepuro; KIT = Kitkajoki below falls;

Pes= Pesospuro; OUL = Oulankajoki; Mer = Merenoja; Maa =Maaninkajoki; Ast = Astumajoki;

Khy = Kitkajoki hatchery; Kil = Kirintöjöki- Lohijoki)

Local topology seemingly played a role in the differentiation between these populations. From the four main stems, the Kitkajoki-above individuals showed higher differentiation from the others, reflecting that the Jyrävä falls act as an impassable barrier between populations. In addition, tributaries that were sampled close to the mouth of the rivers (e.g. Raatepuuro, Merenoja) displayed a considerable level of admixture with the main stem they discharged in. However, tributaries sampled higher (e.g. Maaninkajoki, Juumajoki, Pesospuuro) showed clear isolation. This indicates that within a dendritic system such as the Koutajoki one, geography and topology have, unsurprisingly, an important impact on genetic differentiation between the populations (and as such formation of distinct populations).

Interestingly, the genetic variation between these rivers also indicated different life-history strategies related to migration, suggesting that the migratory behaviour can have an impact on the overall genomic pattern. As tributaries are apparently inhabited by mainly resident trout, the genomic pattern displayed here largely show

A genomic perspective on brown trout migration

uef.fi

Dissertations in Forestry and Natural Sciences

ALEXANDRE LEMOPOULOS

A GENOMIC PERSPECTIVE ON BROWN TROUT MIGRATION

ALEXANDRE LEMOPOULOS

A GENOMIC PERSPECTIVE ON BROWN

TROUT MIGRATION

Alexandre Lemopoulos

A GENOMIC PERSPECTIVE ON BROWN TROUT MIGRATION

ABSTRACT

ACKNOWLEDGEMENTS

LIST OF ABBREVIATIONS

LIST OF ORIGINAL PUBLICATIONS

AUTHOR’S CONTRIBUTION

CONTENTS

1 INTRODUCTION

1.1 AIMS OF THE THESIS

2 MATERIALS AND METHODS

2.1 STUDY AREAS

2.2 MOLECULAR TOOLS

2.3 BIOINFORMATICS METHODS

2.5.2 Genome-wide association studies (IV)

3 RESULTS AND DISCUSSIONS

3.1 DIFFERENCES IN MARKER EFFICIENCY AND EVALUATION OF GENOMIC DIFFERENCES BETWEEN OULUJÄRVI POPU- LATIONS (I)

3.2 POPULATION GENOMICS OF THE BROWN TROUT IN KOUTA- JOKI WATERSHED (II)