présenté Département de biologie UNIVERSITÉ LAVAL JUIN 1997

DANY PIGEON

ANALYSE

GENETIQUE

DE LA SÉGRÉGATION SPATIALE DES LARVES DE DEUX POPULATIONS SYMPATRIQUES D'ÉPERLANS ARC-EN-CIEL (OSMERUS

MORDAX) DANS L'ESTUAIRE MOYEN DU SAINT-LAURENT.

Mémoire présenté

à la Faculté des Btudes supérieures de l'université Laval

pour l'obtention

du grade de maître ès sciences (M.Sc.)

Département de biologie

FACULTÉ DES SCIENCES ET DE &NIE UNIVERSITÉ LAVAL

JUIN 1997

O Dany Pigeon, 1997

National Library Bibliothèque nationale du Canada

Acquisitions and Acquisitions et

Bibliographie Services services bibliographiques

395 Wellington Street 395, nie Wellington Ottawa ON K I A ON4 Ottawa ON K I A O N 4

Canada Canada

The author has granted

a

non- exclusive

licence

allowing

the

National Libfary of

Canada to

reproduce, Ioan, distribute or sell copies of

this

thesis in microfom, paper or electronic formats.

The author

retains

ownership

of

the copyright in this thesis. Neither

the

thesis nor substantial

extracts

fkom it may be printed or otherwise

reproduced without the author's permission.

L'auteur a accordé une licence non exclusive

permettant à

la

Bibliothèque nationale

du

Canada de reproduire, prêter, distribuer ou

vendre

des

copies de

cette thèse

sous la fome de microfiche/film, de

reproduction

sur

papier ou

sur format

électronique.

L'auteur conserve

la propriété du droit

d'auteur

qui protège

cette

thèse.

Ni la thèse ni

des

extraits substantiels de celle-ci ne doivent

être

imprimés ou autrement reproduits sans son autorisation,

Canada

Le partage des ressources trophiques est le principal facteur suggéré pour le maintien dans le temps des populations sympatriques. Le but de cette recherche est de documenter l'importance du partage des ressources au niveau larvaire chez deux populations d'éperlans arc-en-ciel récemment décrites dans I'estuaire moyen du Saint-Laurent. Les résultats de cette recherche ont démontré qu'il n'existe aucun partage spatial des ressources dans ce milieu entre les populations larvaires. En

-

effet, une dominance d'une des populations et une grande homogénéité dans la distribution des individus appartenant à ces deux groupes furent observées pour I'ensem ble de l'estuaire moyen. Ce résultat peut s'expliquer de deux façons: soit qu'il n t a aucun partage des ressources au niveau larvaire chez l'éperlan du St-Laurent, ou encore qu'il a un partage des ressources, mais que celui-ci se fait à un autre stade de vie. Neamoins, ces résultats démontrent l'importance que les facteurs historiques peuvent avoir sur la structure

des

populations. L'importance d'une gestion séparée pour ces deux populations afin d'optimiser leur conservation est également lustrée par les résultats de cette étude.

Je veux tout d'abord exprimer ma reconnaissance envers mon directeur de recherche, le Dr. Louis Bematchez pour sa grande disponibilité, son aide précieuse, sa patience, son dynamisme et la confiance qu'il m'a témoignée au cours de deux dernières années pour mener à bien ce projet de maîtrise ainsi que pour l'année précédant ce projet ou il m'a initié et fait prendre goût à la recherche.

Je tiens également à remercier mon CO-directeur, le Dr. Julian Dodson, pour le matériel mis à ma diposition ainsi que les idées et conseils judicieux qu'il m'a prodigués tout au long de l'élaboration de ce projet.

Mes remerciements s'adressent également à Pascal Sirois avec qui j'ai effectué les campagnes d'échantillonnages et avec lequel j'ai partagé la moitié de chaque larve analysée

.

Merci également à Normand Bertrand qui nous a enseigné les rudiments de la capture de larves, à Martin Lévesque et Daniel St-Pierre, les deux capitaines qui nous ont permis de ne pas faire naufrage durant ['échantillonnage et aussi au GIROQ pour les faciltés reliées à l'échantillonnage.

Merci à tous mes camarades du département de biologie et du laboratoire pour leur support et leur aide, spécialement Sylvain Martin pour m'avoir appris le travail en laboratoire.

Merci également à Lise Boule et Kate O'Malley pour la correction des nombreuses fautes d'orthographe.

Finalement, il me faut mentionner que le chapitre formant le corps de ce mémoire a ete soumis au Journal Canadien des Sciences Aquatiques et Halieutiques pour publication ultérieure.

Les Conseil de Recherche en Sciences Naturelle et en Génie du Canada (CRSNG) m'a supporté financièrement au cour de

ma

_maîtrise.

TABLE DES MATiERES

Pages

AVANT-PROPOS . . . TABLE DES MATIERES . . . LISTE DES TABLEAUX . . . LISTE DES FIGURES . . . INTRODUCTION GENERALE . . .

Concept de partage des ressources entre populations . . . . . . Aspects non-traités du partage des ressources

. . . Biologie de l'espèce à l'étude: l'éperlan arc-en-ciel

. . . Structure génetique de I'eperlan

. . . Théorie du "memberhagrant"

Utilité de IfADN rnitochondrial pour différencier les populations . . . . . . Objedf

CHAPITRE 1: A mtDNA analysis of spatio-temporal distribution of two genetically distinct sympatric larval populations of rainbow smelt (Osmerus mordaw) in the middle estuary of the St .

. . . Lawrence river, Québec, Canada

-Résumé . . . -Abstract . . . . . .

-

Introduction

. . .

-

Materiat and methods

. . . Sample site

. . . Sample collection

Genetic analysis . . . . . . Data analysis

-

Results . . . . . . Sequences analysis

. . . Lawal abundance and distribution

iii

vii

CHAPITRE I (suite)

Discussion

...

.

...

Sequences analysis

Analysis of srnelt larvae

...

Similarity of larval density with previous study

...

Resource partitioning

...

Larval smelt populations and the memberlvagrant hypothesis

.

Historical factors

...

Relevance for management and conservation

. . . ...

CONCLUSION GENERALE

...

REFERENCES BIBLIOGRAPHIQUES

Pages TABLEAU 1. Total nurnber of larvae caught. density, samples sizes for

genetic analysis, relative frequency and absolute density of the two genetic groups of smelt larvae for the

St

Lawrence

middle estuary. . . 37 TABLEAU 2. Hierachical chi-square analysis of mtDNA groups A and 6

frequency distribution among stations and groups of

stations of the St. Lawrence middle estuary. . . . 40 TABLEAU 3. Total number of larvae per 100 m3 belonging each mtDNA

group A and €3, their ratio, south shore contribution and variance estimates. abundance in number of

Iawae

per 100m3 belonging south and north shore populations and their ratio in the St. Lawrence rniddle estuary for each . . .

periods of sampling 41

TABLEAU 4. Relative contribution and variance estimates of the south shore smelt population to the larval samples around Ile

d'Orléans and from the St. Lawrence middle estuary. . . 42

Pages FIGURE 1: Collection sites around Orléans island and approximate

locations of six different water mass salinity that were sampled in the two channels of the middle estuary of St.

Lawrence River. . . 44 FIGURE 2: A) Examples of a diagnostic RFLP analysis on a 2.4kb

segment of mtDNA with enzyme Apa f for smelt larvae of genetic groups A and B. 6) Examples of a diagnostic RFLP analysis on a 288 bp segment of mtDNA with enzyme Dde I

for smelt larme of genetic groups A and B. . . . 46 FIGURE 3: 5'-3' Iight-stand nucleotide sequences for the two smelt

mtDNA groups compared for the three coding genes

ATPase, ND-5 and ND-6. . . 48 FlG URE 4: Frequency distribution and abundance of larvae belonging

the

two

mtDNA groups A and B for Ile d'Orleans and St.

Lawrence middle estuary for the two channels in May, June

andJuly . . . 57

INTRODUCTION GÉNÉRALE

Concept de partage des ressources entre populations

Selon le concept biologique, l'espèce est un groupe de populations s'échangeant des gènes dans les conditions naturelles, ou qui peuvent potentiellement le faire, qui sont reproductivement isolées des autres groupes semblables et qui occupent une niche spécifique dans la nature (Mayr 1942,1963,1982). Ce concept suppose donc un flux génique (existant ou potentiel) entre les différentes populations. De cette façon. lorsque de telles populations se retrouvent en sympatrie, il est intéressant d'identifier et d'évaluer les forces sélectives permettant la réduction du flux génique homogénisateur, la différenciation et la persistance de ces deux populations dans le temps, tant au niveau écologique qu'évolutif. Jusqu'a maintenant, plusieurs recherches ont été effectuées afin de déterminer les causes écologiques et adaptatives permettant la coexistence en sympatrie. De nombreux auteurs expliquent cette coexistence par un partage différentiel des ressources écologiques (Schmitt and Coyer 1982; Skulason et Smith 1995; Schluter et McPhail 1992). Ce partage est le plus souvent occasionne par la compétition entre les populations et fréquemment illustré par le déplacement de caractères reliés à l'utilisation différentielle des ressources trophiques entre ces populations afin de diminuer la compétition intraspécifique (Schluter et McPhail 1992; Robinson et Wilson 1994;

Skulason et Smith 1995). Beaucoup de théories récentes assument en effet, que le facteur primaire responsable de l'organisation des communautés naturelles, particulièrement chez les animaux, est la compétition (Werner et Hall 1977; Dunham 1979). De plus, les polymorphismes trophiques, engendres par le déplacement de caractères, ne sont pas restreints seulement aux différences morphologiques mais peuvent également se manifester aux niveaux du mode de vie et du comportement associés avec les différences alimentaires et d'habitats (Skulason et Smith 1995).

Parmi les populations sympatriques de poissons d'eau douce, partage des ressources avec déplacement de caracteres et

plusieurs cas de polymorphismes trophiques sont documentés. Notamment, au sein de la plupart des populations sympatriques des régions tempérées et nordiques documentées, les dïfférentiations écologiques se font au niveau de la niche uti-lisée. Effectivement, chez ces espèces, on remarque que les populations sympatriques sont souvent spécialisées morphologiquement pour l'utilisation du benthos d'une part et du plancton d'autre part, réduisant ainsi la compétition intraspécifique. Par exemple, on retrouve de telles spécialisations entre les populations sympatriques d'épinoches à trois épines Gasterosteus aculeatus (Ridgway et McPhail 1983; Bentzen et McPhail 1984;

McPhail 1 984, 1 992; Schuter and McPhail 1 992). de grands corégones Coregonus clupeaformis (Fenderson 1964; Lindsey et al. 1970; Bernatchez et al. 1996) et d'ombles chevalier Salvelinus alpinus (Skulason et al. 1993).

Aspects non-traités du partage des ressources

L'étude du partage des ressources entre différentes populations sympatriques de poissons est fréquente en eau douce. Par contre, de telles études sont rares pour les espèces marines ou anadromes, car ces milieux comportent peu de populations sympatriques connues. En réalite, la majorité des travaux portant sur le partage des ressources et le d6placernent de caractères chez les espèces marines sympatriques sont interspecifiques. Ils n'impliquent donc jamais deux populations sympatriques ayant des capacités d'hybridation. Effectivement, en milieu marin, le partage des ressources est principalement étudié en fonction de l'effet qu'il occasionne sur la compétition entre espéces et particulièrement entre espèces du genre Embiotoca (Hixon ¹ 980; Schmitt et Holbrook 1 986; Schmitt et Coyer ¹ 982). Chez ces

espèces,

le partage des ressources s'exprime surtout par une repartition différentielle de

l'espace et de la nourriture que par des modifications morphologiques évidentes (Hixon 1 980; Schmitt et Coyer 1 982; Schmitt et Holbrook 1986).

De plus, les différentes études de partage des ressources mettent en évidence les spécialisations adaptatives entre les individus adultes des populations. En réalité, aucune Btude connue n'a été recensée jusqu'ici mettant en cause

les

autres stades de vie. Ce manque de recherches est surprenant étant donne que les jeunes stades de vie sont considérés trés importants dans la détermination de la structure populationnelle pour une espèce, particulierementt en milieu marin (Sinclair 1988).

Biologie de l'espèce à l'étude: I'éperlan arc-en-ciel

La famille des Osmeridés de l'ordre des Salmoniformes comporte six genres à travers le monde comprenant dix ou onze espèces dont deux sont présentes sur la côte est de l'Amérique du nord, le capelan (Mallotus ViIIosus) et l'éperlan arc-en-ciel (Osmerus mordar) (Scott et Scott 1988). L'éperlan a

une

distribution qui va de I'estuaire du Lake Melvile (Labrador), au nord jusqu'au New Jersey au sud (Scott et Crossrnan 1973). A u Québec, il est une des principales composantes de la communauté I'ichthyoplanctonique de l'estuaire du Saint-Laurent (Able 1 978; Ouellet et Dodson 19856; Dodson et al. 1989).

L'éperlan est un poisson de petite taille, habituellement une vingtaine de centimètres de longueur, de forme allongée, comprimé latéralement, sans procès axillaire et avec des dents bien d6velopp6es, ce qui le différencie du caplan (Scott et Scott 1988). 11 est carnivore et considéré comme étant trés vorace. En outre, il présente divers modes de vie, il peut aussi bien &re anadrome que dulçaquicole (Nellbring 1989).

Au Canada, l'éperlan d'eau douce se retrouve principalement dans les lacs des Provinces Maritimes et dans les Grands Lacs où il a été introduit artificiellement suite à l'ensemencement de cette espèce dans le Crystal Lake, MI. (Scott et Scott 1988). En eau douce, il peut de plus, adopter différentes formes appellées naines et normales que Iton retrouve tant en allopatrie qu'en sympatrie (Delisle et Veilleux 1969; Nellbring 1989). Lorsqu'il se retrouve en sympatrie, l'éperlan dulçaquicole de l'est du Canada se différencie en deux formes génétiquement distinctes et spécialisées pour l'utilisation de niches alimentaires spécifiques (Delisle et Veilleux 1969; Taylor et Bentzen 1993a, 1993b). Par exemple, les éperlans sympatriques du lac Utopia, NB, Cam se distinguent en deux populations morphologiquement, genetiquement et reproductivement distinctes. Une des populations est caractérisée par des individus de taille normale qui rnaturent à plus de 20 centimètres de longueur standard, avec 31-33 branchiténies et qui frayent directement dans le lac.

Par contre, les individus de l'autre population sont nains, ils maturent à moins de 15 centimètres de longueur standard, ont 34-36 branchiténies et frayent dans de petits ruisseaux 3-5 semaines plus tard que la fome normale (Taylor et Bentzen 19936).

L'éperlan anadrome habite les zones côtières, les mers et les estuaires. II remonte généralement les cours d'eau (rivières et ruisseaux) au printemps peu après la débâcle pour frayer (Scott et Scott 1988; Bematchez et Giroux 1991). Dans l'estuaire moyen du Saint-Laurent, l'éperlan remonte les ruisseaux et ies rivières de la rive-sud en mai pour se reproduire. Une à deux semaines après la frai, les oeufs éclosent et les larves sont rapidement expulsées dans I'estuaire où, elles sont transportées vers leur zone d'alevinage située dans la partie nord de I'estuaire moyen (Ouellet et Dodson 1985a, 19856). Les larves sont par la suite retenues dans cette zone de haute productivité planctonique par des migrations verticales actives en combinaison avec la circulation cyclonique du système (Ouellet et Dodson

1 985a; Laprise et Dodson 1 989a, 1 9896). L'analyse des caractères biologiques de l'éperlan anadrome du Québec indique l'existence de trois groupes géographiques, ceux de la Baie des Chaleurs, de la rive sud du fleuve Saint-Laurent et du fjord du Saguenay (Fréchet et Dodson 1983a, 19836). Ces caractères suggèrent également la présence d'un quatrième groupe situe sur la côte nord du bas estuaire du Saint- Laurent (Fréchet et Dodson 1 983a).

Structure génétique de I'éperlan

La structure génétique de I'éperlan arc-en-ciel du nord-est de l'Amérique du Nord a été étudiée depuis quelques

années

avec l'aide de l'ADN rnitochondrial (Baby et al.

1991 ; Taylor et Bentzen 1993a, 1993b; Bematchez et al. 1995; Bernatchez 1995;

Taylor

et Dodson 1994; Bernatchez et Martin 1996; Bematchez 1996). Jusqu'à maintenant plusieurs dizaines de génotypes mitochondriaux ont été identifiés, ces différents génotypes se regroupent en deux différents groupes monophyfétiques nommés A et B et supportés à 98% par une analyse cladistique (Bematchez et Martin 1996). La distribution géographique différentielle des éperlans appartenant a ces deux groupes impliquant une discontinuité phylogénétique, Bernatchez ( 1 996) démontra que ces groupes originaient de deux refuges glaciaires distincts fomés lors du Pléistocène. Brièvement, un de ces groupes, le groupe A, s'est présumément différencié dans le refuge nommé Acadien situé au niveau des Grands Bancs, près de Terre-Neuve et l'autre, le groupe B, dans le refuge Atlantique situé le long de la plaine côtière atlantique incluant le banc de Georges (Bernatchez 1996). Avec le retrait graduel des glaciers il y a 18 000 ans, les éperlans appartenant à ces deux groupes génétiques ont recolonisé le continent suivant deux routes géographiques distinctes (Bematchez 1996).

Des analyses génétiques récentes (RFLP) utilisant l'ADN mitochondrial (Baby et al.

1 991 ; Bematchez et al. 1995; Bernatchez et Martin 1996) ont confirme l'existence des quatres groupes géographiques déterminés antérieurement par Fréchet et Dodson (1 983a. 1983b). Ces analyses génétiques ont également donné

des

résultats très inattendus sur la structure populationnelle de l'éperlan dans l'estuaire du Saint-Laurent. En effet, ces études ont identifié deux populations génétiquement distinctes qui coexistent en sympatrie au niveau de I'estuaire moyen de ce fleuve (Bematchez et al. 1995; Bernatchez et Martin 1996). Ces deux populations sont génétiquement plus différenciées entre elles qu'entre n'importe qu'elles autres populations anadromes, même géographiquement très éloignées. Ce résultat indique que la différenciation génétique n'est pas toujours en corrélation avec les distances géographiques et que les événements historiques sont très importants dans l'élaboration de cette différenciation. En effet, Bernatchez (1996) expliqua I'existence de ces deux populations par un contact secondaire dans I'estuaire moyen des deux races glaciaires décrites précédemment qui auraient. avec le temps, développées des barrières biologiques pour limiter le flux génique entre elles. car chacune des deux populations est dominée par des individus appartenant à des lignées phylogénétiques distinctes.

Théorie du llmember/tagrant"

Même si l'origine des deux populations de l'estuaire moyen est bien déterminée. leur coexistence n'est pas encore comprise. En effet, cette persistance n'est pas en accord avec les considérations théoriques. Selon la théorie du memberkagrant de Sinclair (1988), le nombre de populations

chez

les organismes marins est fonction du nombre de structures oc6anographiques ou physiques qui permettent la survie

des

jeunes

stades

de développement. CintBgrité des populations est donc assurée

par la discontinuité géographique de ces structures. par la capacité des jeunes stades de vie de s'y maintenir et par la capacité des adultes de retourner aux sites

de

frai associés à chaque structure (Bernatchez 1990). Une espèce peut alors être riche ou pauvre en populations suivant le nombre de sites de rétention larvaire (Sinclair et Iles 1988). L'estuaire moyen est une structure océanographique dans laquelle les eaux sont bien mélangées en amont et partiellement stratifiées en aval (El Sabh 1988). Les larves d'éperlans étant asociées avec la partie bien mélangé en amont qui constitue un site de rétention unique (Dosdon et al. 1989), I'hypothése de memberhagrant prédit alors la persistance que d'une seule population dans ce rniiieu,

II est possible que la persistance

de

ces deux populations dans la même zone de rétention larvaire suggère que les larves d'éperlans peuvent utiliser la structure de façon différentielle selon la population. Puisque l'on sait que les larves se maintiennent activement dans le système. elles ont peut-être développé des mécanismes comportementaux distincts leur permettant d'utiliser de façon différentielle la zone d'alevinage par une distribution spatiale différente afin que la coexistence devienne possible. De plus, comme aucune différence morphologique n'a encore été démontrée par une étude entre les larves des deux populations, on peut supposer que le partage des ressources responsables du maintien à long terme de ces populations dans le système soit de nature comportementale. Une étude portant sur une hypothèse de distribution spatio-temporelle difFérente entre les larves des deux populations est donc pertinente, car selon Skulason et Smith (1 995) les différences de comportements dans le partage des ressources précèdent souvent les autres differences telles les spécialisations morphologiques.

UtiM de l'ADN mitochondnal pour dfl6rencier les populations.

CADN mitochondrial (ADNmt) est un petit génome d'environ 16 kpb de structure ghetique simple sans séquence d'ADN répétitif, élement tramposable, pseudogene et intron (Bernatchez 1990). 11 est largement utilisé dans les études de structure populationnelle car cette molécule possède

des

qualités intéressantes pour l'étude évolutive des populations (Bernatchez t99O). Pratiquement, c'est une molécule facilement isolable, qui est presente en plusieurs copies (plus de mille copies en moyenne) dans chacune des cellu[es d'un organisme et qui évolue beaucoup plus rapidement que I'ADN nucléaire simple copie (Meyer 1994). Le grand nombre de copie dlADNmt par cellule est tr& intéressant lorsque l'étude populationnelle se fait au niveau de très petits organismes comme les larves, car il offre proportionnellement plus de gabarits pour les travaux moléculaires, comme le PCA, que I'ADN nucléaire. Par exemple, l'amplification d'un segment d'ADN par PCR nécessite théoriquement la pr6sence que d'une copie du segment que Iton veut amplifier, mais pratiquement les amplifications ne sont possibles qu'avec un nombre minimal de gabarits variant selon la taille du segment a amplifier. car le PCR est une technique basée sur la probabilité de rencontre des différentes amorces avec le brin mère. L'utilisation de IIADNmt offre donc un avatage certain lorsque le nombre de cellules de qualitées est limité comme dans le cas des larves qui subissent un stress physique important lors de l'échantillonnage et qui doivent être conservées longtemps avant leur utilisation à des fins moleculaires.

De plus, I'évolution rapide de cette molécule est aussi une qualité attrayante parce qu'elle permet généralement de bonnes differentiations populationnelles au niveau intraspécifique comme c'est le cas chez de nombreuses espèces de poissons notamment le corégone (Coregonus clupeafonnis). la truite brune (Salmo trutta) et

evidemment l'éperlan arc-en-ciel (Osmeros mordax)(Bernatchez et Dodson 1 990, 1 991 ; Bernatchez et Osinov 1 995; Taylor et Bentzen 1 993a, 19936). En effet, les differentes populations de ces espèces ont et6 identifiees clairement soit par des analyses de séquences de I'ADNmt. soit par etudes de

RFLP

(restriction fragment length polymorphisms). Cette demiére méthode consiste a la digestion de I'ADNmt avec des endonucléases de restriction reconnaissant des séquences spécifiques de quatre six nucléotides produisant chacune un certain nombre de fragments de différentes longueurs facilement visualisés suivant leurs migrations selon la taille sur gel d'agarose. C'est l'étude de ces différents fragments qui permet d'identifier et de classifier les individus appartenant aux différentes populations génétiques d'une espèce donnée.

Objectif

L'hypothèse générale de ce projet est de vérifier que le maintien des deux populations d'éperlans arc-en-ciel (Osmerus mordaw) dans l'estuaire moyen du Saint-Laurent est dû à un partage différentiel des ressources écologiques. Comme on sait que le stade larvaire est déterminant dans la structure populationnelle d'une espèce et que le potentiel de partage des ressources peut varier avec le développement. il est de mise pour une &ude sur le partage des ressources de commencer par le stade lamire. Dans cette etude, nous allons identifier les larves appartenant aux deux populations avec des analyses de PCR-RFLP sur I'ADNmt de chaque

larve

afin de tester 11hypotht3se d'une distinction dans leur distribution spatio- temporelle. Cette distinction dans la distribution des larves des deux populations est un mecanisme potentiel de partage des ressources pouvant promouvoir

leur

coexistence dans une même zone de rétention larvaire (voir Chapitre 1 ).

CHAPITRE 1

A mtDNA analysis of spatio-temporal distribution of two genetically distinct sympatric larval populations of rainbow smelt ( O s m e m mordax) in the middle estuary of the

St. Lawrence river, Québec, Canada.

Résumé

Plusieurs auteurs suggèrent que le partage des ressources écologiques est le principal facteur permettant le maintien de populations en sympatrie. Des travaux récents ont démontré l'existence de deux populations sympatriques d'éperlans arc- en-ciel anadromes (Osmenrs mordax) génétiquement distinct mais ayant le potentiel de s'hybrider dans l'estuaire moyen du St-Laurent. La persistance de populations en sympatrie dans ce milieu est contraire aux prédictions théoriques de l'hypothèse du

"member/vagrantu. Dans cette étude, nous avons effectué une analyse de PCR- RFLP sur I'ADNrnt de 922 larves d'éperlans représentant 33 sites d'échantillonnage dans le but de tester l'hypothèse voulant que ces deux populations larvaires soit spatio-temporellement ségrégées et que cette ségrégation explique en partie leur persistance dans le milieu. Les résultats de ces analyses démontrent une homogénéité spatiale dans la distribution des larves appartenant aux deux populations. Conséquemment, ces résultats ne supportent pas I'hypothèse de départ. On peut expliquer ces résultats de deux façons différentes. Premièrement, le partage des ressources est peu important au niveau larvaire pour l'éperlan du St- Laurent. Alternativement, le partage des ressources est significatif, mais

ne

_{se fait}

pas spatialement. Ces resultats démontrent de plus que les effets des événements historiques sur la structure des populations peuvent, dans certaines circonstances, être plus importants que les événements écologiques contemporains.

Abstract

Many authors have suggested that ecological isolation through resource partitioning is one of the more important factor for explaining the persistence of genetically distinct yet closely related sympatric populations. Recent molecular studies demonstrated the existence of two genetically distinct sympatric populations of rainbow smelt (Osmerus mordax) in the middle estuary of the St. Lawrence River.

The persistence of these coexisting populations in sympatry also is in conflict with current theoretical concepts predicting population richness. In the present study, we performed mtDNA PCR-RFLP analysis of 922 individual lawae from 33 sarnpiing stations in order to test the hypothesis that the larvae belonging to the two sympatric smelt populations of the St. Lawrence rniddle estuary are spatially segregated and that such segregation may promote the persistence of the populations. Results clearly revealed spatial homogeneity in the relative distribution of larvae içsued from the two populations. Consequently, they did not support our working hypothesis that larvae belonging to the two sympatric smelt populations in the St. Lawrence middle estuary are spatially segregated. Two alternative explanations rnay account for the lack of spatial partitioning obsewed here. Cornpetition may not be important enough to promote resources partitioning at the larval stage. Alternatively, resources partitioning occurs but not spatially. This study also demonstrated that the effect of historical events on the genetic population structure may have been as important as contemporary ecological settings in deterrnining genetic population structure in smelt.

Introduction

Many north temperate fish species are characterized by the occurrence of sympatric ecotypes that remain genetically distinct despite their potential to hybridize (summarized in Bematchez et

al.

1996). Because of the close relationships and incomplete reproductive isolation, these sympatric ecotypes have been studied to better understand the early stages of speciation. The understanding of the role of selective forces favoring persistence of sympatric foms despite the potential to hybridize has been of particular interest in such studies. Many authors have suggested that ecological isolation through resource partitioning is one of the more important isolating factors that can explain the persistence of genetically distinct yet closely related sympatric populations (Schluter and McPhail 1992, 1993; Hindar 1994; Robinson and Wilson 1994; Skulason and Smith 1995). Ecological resource partitioning can contribute to decrease intraspecific cornpetition between sympatric populations (Skulason and Smith 1995) and thus favour the maintenance of coexisting populations by disruptive or frequency dependent selection (Schiuter and McPhail 1992).

Resource partitioning generally occurs in sympatric freshwater fish populations by trophic-based adaptations (Fenderson 1964; McPhail 1983; Taylor and Bentzen 1993a. 19936; McPhail 1994; Snorrason et al. 1994; Bematchez et al. 1996). These resource polymorphisrns are rnanifested in a number of ways, including morphological, behavioral and life history characters (Skiilason and Smith 1995). For example, differentiation in dwarf and normal specialized morphotypes are frequently documented, as in the whitefish Coregonus clupeaformis (Fenderson 1964;

Kirkpatrick and Selander 1979; Bematchez and Dodson 1990) and in the rainbow

smelt Osmerus

mordax

(Nellbring 1989; Taylor and Bentten 1993a, 1993b). Another documented type of trophic specialization involves limnetic and benthic foraging behaviors coevolved with distinct morphological features, as for the threespine stickleback Gasterosteus aculeatus (McPhail 1983; Benizen and McPhail 1984;

Ridgway and McPhail 1984; Schluter 1993). Skulason and Smith (1995) suggested that a relaxation of interspecific competition and the availability of open niches are essential to promote resource polymorphisms. Such conditions are likely prevalent in north temperate freshwater ecosystems (Schluter and McPhail 1993) and have been identified by many authors as an explanation for the rapid adaptive radiation within many north temperate fish species complexes such as lake whitefish Coregonus clupeaformis, arctic charr Salvelinus alpinus, brown trout Salmo trutta, and stickleback Gasterosteus aculeatus (Fergusson and Taggart 1991 ; Schluter and McPhail 1992; Robinson and Wilson 1994; Skulason and Smith 1995; Bematchez et al. 1996, Pigeon et al. 1997). Despite intensive studies of sympatric populations found in freshwater, no study has yet documented resource partitioning between sympatric fish populations of a same species in estuarine or marine ecosystems.

Resource partitioning studies in manne ecosystems have only involved sympatric and reproductively isolated fishes (McEachran and Martin 1977; Hixon 1980; Schmitt and Coyer 1982; Schmitt and Holbrook 1986) but never sympatric populations with the possibility of interbreeding. Furtherrnore, ail studies related to resource partitioning among sympatric fish populations have generally focused on juvenile and adult life-history stages, but have not considered the lawal stage despite the presumed importance of this stage in contributing to population richness and abundance (Sinclair 1988).

The rainbow smelt (Osmerus mordax Mitchill) is an osmerid fish distributed in north eastem America from New-Jersey to Labrador (Scott and Crossman 1973; Scott and

Scott 1988). The species

can

be found in salinity ranging from fresh water to 30 and in water temperatures from below O°C to more than 20°C (Nellbring 1989). It also exhibits a broad life history diversity consisting of anadromous and landlocked populations as well as freshwater dwarf and normal ecotypes (Nellbring 1989; Taylor and Bentzen 1993a) respectively specialized for planktivorous. andlor macrophagous andor piscivorous modes of life (Taylor and Bentzen 1993a. 19936).

These authors demonstrated that sympatric freshwater dwarf and normal ecotypes re present two genetically distinct populations.

Anadromous populations of this species exhibit a similar Iife cycle: adult smelt ascend strearns and rivers in early spring and spawn in freshwater. Hatching occurs 1-2 weeks later and larvae from different spawning sites are usually transported within short periods into estuarine, Fjord or marine coastal waters where they are retained (Ouellet and Dodson 1985a ,1985b; Laprise and Dodson 1989a, 19896) until metamorphosis. In the St. Lawrence estuary, the larvae are transported from spawning sites to a nursery zone situated in the maximum turbidity zone where they are retained by active vertical migration (Dodson et al. 1989; Laprise and Dodson 1989a, 1989b). In Quebec waters, previous studies of variation in biological characters and parasite occurrence revealed the presence of three phenotypically significant geographical groups: the south shore St. Lawrence middle estuary, the Chaleur bay, and the Saguenay Fjord (Fréchet et al. ^{1983a. 1983b).} Recent molecular studies corroborated these results and demonstrated the existence of two genetically distinct syrnpatric populations in the middie estuary of the St. Lawrence river (Bernatchez et al. 1995; Bernatchez and Martin 1996). MtDNA analysis indicated that these populations were associated with two phylogenetically distinct lineages separated by approximately 0.8% average pairwise sequence divergence (Baby et al. 1991 ; Taylor and Bentzen 1993a; Bematchez et al. 1995; Bernatchez et

Martin 1996). SmeQ ascending tnbutanes along the south shore of the estuary to spawn are dominated by phylogenetic group A (80.g0h), whereas those caught in the north channel between Beauport and the Saguenay Fjord are dorninated by the phylogenetic group B (87.1 %) (Bematchez et al. 1995; Bematchez and Martin 1996).

Hereafier, these two populations were respectively named "south shore population"

and "north shore populationn. Restricted gene flow between the two populations was further illustrated by an Fst estimate of 0.1 78 (Bematchez and Martin 1996), a value greater than those commonly observed among local populations of safmonids from geographically separate localities (reviewed in Hindar 1994). Bernatchez (1997) demonstrated that the allopatric ongin of the two phylogenetic groups A and B was associated with distinct glacial refugia in the fast glaciation, and that the middle estuary represents a zone of secondary contact between these races which have devefoped biological bamers to gene flow between them.

The persistence of these two populations does not appear to involve the same mechanisms that have been obsewed in the lacustrine environment. No evidence of dwarf and normal ecotypes have been documented and the geographical mixing of the adults belonging to the two populations has been demonstrated outside the spawning locations (Bematchez et al. 1995). These observations suggest that the maintenance of the sympatric populations may be determined by differential resource partitioning at a younger life-history stage, namely at the larval stage. This persistence also apparently contradicts theoretical predictions on the ecological causes of population structure in the manne environment. According to the mernber- vagrant hypothesis (Sinclair 1988; Sinclair and lles 1988, 1989) population richness and structure are a function of the number and locations of geographic settings pennitting the retention of eariy life-history stages. The well-mixed upstream part of the St. Lawrence middle estuary associated with the smelt nursery area represents a

unique oceanographic system for larval retention (Dodson et al. 1 989).

Consequently, only one population should persist in the systern according to the member-vagrant model. The persistence of the two populations in one oceanographic entity suggests that the larvae might use this structure differently. It has been demonstrated that larval smelt have developed complex behavioral mechanisms to stay in the favorable nursery zone (Laprise and Dodson 1989a, 19896; Dauvin and Dodson 1990). Therefore, it is plausible that these behavioral mechanisms evolved differently for each population resulting in distinct spatio- temporal distributions pemitting coexistence in the same estuarine system. The study of larval dispersion influenced by distinct behavioral mechanisms is particularly relevant as behavioral differences have often been considered to precede segregation in morphological and life history characters for resource partitioning.

(Skulason and Smith 1995).

In this study, we perforrned a mtDNA PCR-RFLP analysis in order to test the hypothesis that lanrae belonging to the two sympatric smelt populations of the St.

Lawrence middle estuary are spatially segregated and that such segregation may prornote the persistence of two populations in the system. As such, this study represents the first characterization of resource partitioning at the larval stage among sympatric fish populations and the first ecological assessrnent of sympatric fish populations in an estuarine environment.

Material and methods

Sample site

The St. Lawrence middle estuary is approximately 180 km long, receives an annual mean discharge of 10 000 m3 s-1 from the river and comprise three main channels, north, middle and south (D'Anglejan and Smith 1973; Laprise and Dodson 19894 1993). The estuary is characterized by a well-rnixed zone in the upstream section and, a partially-stratified zone downstream (El Sabh 1988). The salinity varies from Opsu (pratical salinity units) upstream of Ile diOrléans to over 30psu downstream of Ile aux Coudres (Fig 1). The estuary is highly energetic; tides Vary between 3 to 5 m in height and current speed may reach 250 cm s-1 (D'Anglejan and Smith 1973).

Due to Coriolis force, tides and the bottom topography of the estuary. important transverse currents are f o n e d upstream and downstream the central islands (Ouellet and Trump 1979; Laprise and Dodson 1989b). Thus, this part of the St- Lawrence estuary between Ile d'Orléans and lie-auxCoudres is characterized by a cyclonic circulation which. in addition to the presence of a nuIl zone (Dodson et al.

1989), contributes to the retention of suspended particulate matter inflowing from the river and salt marshes along the estuary foming a maximum turbidity zone (Laprise and Dodson 1 989 6) (Fig. 1 ).

Sample collection

Larvae were sampled in the north and south channels over three consecutive months (Table 1). Pelagic trawling was carried out at five stations in May and six in June and July between Cap-Brulé and St-Siméon in the north channel and between

St-Jean-Port-Joli and Cacouna in the south channel

(Fig.

1). The stations were identified by the salinity of surface water masses ranging from O to 25psu in each channel. Lanrae were also sarnpled Wice in

May

at four stations around Ile d'Orléans (Table 1. Fig. 1). Sampling consisted of step-oblique tows from the bottom to the surface using a 1 meter plankton net equipped wlh a flowmeter and ftted with a 0.500mm mesh net for the sampling in May and around Ile d'Orléans. Sarnpling in June and July was made in the same rnanner but with a 50 cm, 0,500 mm mesh plankton net fitted to a Tucker trawl. The volume of water filtered was calculated from the distance towed and diameter of the net. Samples were presewed in 95%

ethanol until larvae were sorted for genetic analysis.

Genetic analysis

Total DNA was extracted differently according to larval size. For small size larvae (ranging from 40 to 150 pg and 5 to 13 mm in length), total DNA was extracted following a rapid DNA extraction protocol from ethanol preserved larvae. Briefly, each larvae was soaked twice in distilled water for 15 min in order to remove ethanol. Water was discarded and each larvae was put in 50 pl of extraction buffer (Tris-HCI 10 mM pH 8.0, KCI 50 mM, Tween 20 0.50h, 250 m g h l Proteinase K).

Tubes were incubated for 3 to 4 hours at 65°C and then heated at 95°C for 15 min.

Digestion product was centrifuged for 10 min. For bigger larvae, total DNA was purified from the entire specimen following the procedures described in Bematchez et al. (1 992).

All larvae were aliocated to mtDNA group A or 6, defined previously by Baby et al.

(1 991) and Bematchez and Martin (1 W6), by RFLP analysis performed with restriction endonucleases generating diagnostic sites on PCR-arnplified segments.

We initiated the analysis with the bigger larvae. We performed PCR amplifications on a 2.4 kb segment of the mtDNA molecule encompassing the ND5 and ND6 subunits of the

NADH

dehydrogenase using the prirners developed by Cronin et al.

(1 993) which we named ND56R and ND56F. Each PCR reaction was composed of 1-5 pL of the total DNA extract (approximately 500ng), 5 pl of 10X buffer (500 mM KCI, 1 OOmM Tris-HCI (pH9.0), 1% Triton X-100 and 25

m M

MgCfz), 4 pl of dNTP mix (2.5 mM each of dATP, dCTP, dGTP and dïTP in sterile water), 2 pl of a 20 mM solution of the two prirners, 2-4 units of Taq polymerase and sterile water for a final volume of 50 pL per reaction. DNA amplification was performed in a programmable themal cycler (Perkin-Elmer model 480) using the following profile: a preliminary denaturation at 9S°C for 2 min, followed by 40 cycles involving strand denaturation at 94OC for 1 min, annealing at 4S°C for 1 min, and primer extension at 72°C for 2 min 30 sec, followed by a final elongation at 72OC for 10 min. The segments were subsequently digested with the hexameric enzyme Apa 1 used as recommended by the supplier (Phamacia). This enzyme generated diagnostic fragment patterns for the mtDNA groups A and 6 (Fig. 2a). MtDNA fragments were electrophoretically separated on 1.2% or 1.4% agarose gels at 90 V run for five hours. Restriction fragments were revealed by ethidium bromide staining. A Hind III digest and EcoRI-Hind I I double digest of lambda DNA was used as size standards in each gel (Fig.2a). Reliability of this method was insured by the analysis of 20 adult srnelt previously characterized with RFLP analysis of the entire mtDNA molecule and known to belong to either mtDNA group A or B.

The weak amount of DNA extracted from the small-size larvae in May was problematic for the amplification of the 2.4 kb segment. Development of primers for arnplifying a shorter diagnostic segment was therefore necessary to reduce the stringency of the PCR reactions. To do so, sequencing was performed on three

potentially variable mitochondrial genes, the NADH subunits 5 (ND 5), 6 (ND 6) and the ATPase subunit 6, in order to determine primer sets flanking diagnostic restriction sites between the two mtDNA groups (Fig. 3). PCR amplifications were performed on ND-5 and ND-6 subunl genes as described above and on the ATPase gene with the procedures described in Giuffra et al. (1994).

Amplified DNA was purified with the Quiagen DNA purification kit used as recommended by the supplier. Double-stranded DNA sequencing was performed by applying the dideoxy chain-ending technique using the Sequenase kit (Version 2.0, US Bio-chernical) as described in Bematchez et al. (1992) with the following modifications. Sequence reactions were firstly performed with primer ND56R and ND56F developed by Cronin et al. (1 993) for ND-5 and ND-6 genes and H9208 and L8558 developed by Guiffra et ai. (1 994) for ATPase subunit 6. Internai primers deveioped in the present work were used for subsequent sequence reactions until the identification of diagnostic restriction sites between the two groups (Fig. 3).

Foilowing this procedure, a 288 bp mtDNA segment localized between primers ND56R250 and ND56R450V in the ND-6 gene was chosen to characterize the smallest larvae (Fig. 24. Diagnostic fragment patterns were generated by Dde I on this segment. Procedures to genetically characterize larvae with this segment were done as described for the 2.4 kb segment with the following modifications. Firstly, primers ND56F and ND56R were replaced by the ND56R250 and ND56R450V primers (Fig. 3). Secondly, the elongation step was reduced to a 45 sec step. Finally, Phi-X 174 DNA Hae III digest was used as size standards in each gel (Fig. 26). As previously described, the reliability of the method was insured by the anaiysis of adult srnelt.

Data analysis

The density of larvae at each station (number of larvae per 100 m3 of water filtered) was calculated at each station for each month and averaged for each month. These estirnates allowed to quantitatively evaluate the number of larvae belonging to each genetic group, and therefore. each population for the different stations or periods of sampling.

Heterogeneity in the distribution of larval populations among stations, groups of stations and sampling periods was first evaluated by the analysis of frequency distribution of the two genetic groups for each station or group of stations (in cases where no distinction was detected among stations) using chi-square randomization tests (Roff and Bentzen, 1989) with 1000 randomizations perfoned by the MONTE program of the REAP software package (McElroy et al. 1992). Significant heterogeneity was accepted when the probability p was <50/1000.

The relative contribution of the two breeding populations to larval samples was estimated for each station and group of stations for each period of sampling. This was calculated using the method of Lane et al. (1990) by first considering the frequency of the two genetic groups A and B in the two reproductive stocks characterized previously as baseline populations (Bernatchez et al. 1995). Final estimates were however calculated by using the distribution of the two mtDNA groups in May larvae as the baseline data for the north shore population. This substitution generated more plausible estimates (see Results section for explanations). Lane's et al. (1990) formula to estirnate the contribution of the south shore population and the asymptotic variance formula are:

where x represents the contribution of the south shore population, is the observed frequency for one genetic group (A in this case) in larvae samples, p l is the frequency observed for this genetic group in the south shore reproductive population, pz is the observed frequency for the genetic group A in the other base line population, V(pi) = pi(1

-

^pi)/2Ni, where Ni is the sample size from which pi was estimated (north shore estimates were obtained by subtraction).

Results

Sequences analysis

Sequence analysis was perfomed until finding suitable primers flanking diagnostic restriction sites. This resulted in a 5'-3' sequence effort of 3042 bp of the light strand of mtDNA (Fig. 3). The following sequences were obtained for one representative of each mtDNA group A and 6: 1) 618 bp from the 68th nucleotide of the ATPase subunit 6, to 10 bp in the adjacent cytochrorne oxidase subunit 3 (COlll) gene, 2) the cornplete ND-5 gene (1833 bp)

and

32 bp of the adjacent Leucine t-RNA gene, and 3) the complete ND-6 gene (522 bp) and 28 bp of the adjacent Glucine t-RNA gene.

Only nine mutations were observed in these sequences and their spatial distribution was heterogeneous. Thus, the ATPase subunit 6 gene segment represented the least variable region, with no substitutions detected, compared to six in the ND-5 gene and three in the ND-6 gene discriminating the two mtDNA groups (Fig. 3). The overall sequence divergence estimate between the groups was 0.30%. Seventy- eight percent (719) of these substitutions

were

transitions, three implicated purine- purine differences (al1 in the ND-5 gene) and four implicated pyrimidine-pyrimidine differences (one in N D 4 gene and three in ND-6 gene, Fig.3). The two trançversions occuned in the ND5 gene for a final transition:transversion ratio of 7:2. Among these nine mutations, seven were silent, eight implying substitutions at the third base of codons and one at the second nucleotide. Two of them involved amino acid replacement (Fig. 3). In addition, among the nine mutations, five represented diagnostic restriction sites between the two groups (Fig. 3).

Larval abundance and distribution

A total of O to 2797 larvae were captured at each of the 42 stations of the middle estuary and around Ile d'Orléans in May, June and July 1994 (Table 1, Fig. 4a-û).

Low densities of larvae were observed at the downstream sites of the middle estuary. Thus, at the 18 stations having a salinity between 15 and 25psu and for the three sampling months, 17 had a larval density inferior to two per 100rn3 of water (Table 1). Larval density was much higher in the upstream section of the middle estuary (10psu and less), with the maximum reached at pointe Dauphine in May where a density of 396.8 larvae per lOOrn3 was observed.

Depending on abundance, 2 to 59 larvae per station were genetically characterized (al1 larvae when ne40 and random sub-sampling of 40 to 59 for the others), for a

total of 922 larvae from 33 stations. No smelt larvae were caught at the other nine stations. Globally, the frequency distribution of the larvae belonging to mtDNA groups A and B did not support the hypothesis of differential spatial segregation between the two larval populations. Thus, the distribution of larvae characterized by either mtDNA group A or 6 was very homogeneous among stations (Table 1).

Larvae characterized by mtDNA group B dominated each station at a frequency varying between 76 to 100% (Table 1, Fig. 4a-d) and no difference in the frequency distribution of the two mtDNA groups was detected among individual samples for each month (Table 2). Heterogeneity analysis was then perfomed each month among stations of the north and south channels individually and no significant difference was observed except at Montmorency R. and St-Anne R. on May 31 (Table 2). Samples of each channel were then pooled for each month to compare the distribution of rntDNA groups between channels (May 31 observations excluded) and no heterogeneity was detected (Table 2). Finally, samples were pooled by month in order to test for temporal heterogeneity in the frequency distribution of the two groups. These tests indicated distinct distributions of the two mtDNA groups between May and June and May and July (Table 2). The difierences were explained by an increase in the number of larvae belonging to group A in time. This increase is illustrated by the ratio of absolute numbers of larvae characterized by mtDNA groups A and B for a given month (Table 3).

The distribution of mtDNA groups A and

B

among samples allowed the estimation of the relative contribution of the south shore breeding population to the number of larvae in each station and for each month (Table 4). lnitially calculated using the baseline populations of Bematchez et al. (1 995), contribution estimates of the south shore population were negative for May. Such nonsensical results were due to the frequency of mtDNA group A (6%) observed in May, which was lower than the

frequency reported by Bematchez et al. (1995) for the north shore breeding population (13%). This difference may be due to yhe stochastic effects of sampling because of the relatively low number (n=101) of fish analyzed by Bematchez and al.

(1995) and/or partial rnixing of the two aduit populations in the north channel (Bematchez et al. 1995). As the sample size for May 24-26 sampling was greater (n=305) and always avoided negative contributions, we felt that the proportion of larvae belonging to A and 8 groups in this month was more representative of the north shore baseline population for estimating relative contributions. This implies that the laivae belonging to the south shore population were practically absent from the system in May.

Congruent with the results based on the frequency distribution of mtDNA groups A and B, the contribution of the south shore population to larval samples was very weak and homogeneous among stations for a given month (Table 4). Seventeen of the 33 contribution estimates were nuIl and the average contribution of the 16 others was 10.0%. A temporal increase in the frequency of the mtDNA group A was

also

reflected by the estimate of the south shore contribution to lanral samples (pooled) for each month. Thus, the estimated contribution for this population increased frorn 0% in May, to 5% in June and to 8.7% in July (Table 3).

The total larval density belonging to the south shore population in the systern was calculated each month (Table 3) by rnultiplying the contribution estimate of the south shore population by the larval density for each month (Table 3). The total abundance estimates of laivae belonging to the north shore population were estimated by the subtraction between the total and south shore larval density (Table 3). No detectable contribution of the south shore population was observed in May. For the two other months, the larval southmorth shore population density ratios were approximately

1 :20 in June and 1:10 in July (Table 3). Therefore, the total lamal density indicated an increase of the larvae belonging to the south shore population in tirne.

Discussion

Sequence analysis

Sequence analysis revealed heterogeneity among genes in the mutational distribution between the two smelt mtDNA groups. This suggests differential mutation rates among genes, which corroborate previous studies (Thomas and Beckenbach 1989; Chapman et al. 1994; Giuffra et al. 1994; Meyer 1994; Zardoya et al. 1995). The divergence estimates between the two mtDNA groups were ,0.16%

(0/638) for ATPase subunit 6, 0.33% (6/1830) for ND-5 gene and 0.57% (3/522) for ND-6 gene.

However, an unexpected result was the overall low divergence (0.29%) between the two mtDNA groups obtained by sequencing compared to RFLP analysis performed over the entire mtDNA molecule (approxirnately 0.8%) (Baby et al. 1991 ; Taylor and Bentzen 1993a; Bematchez and Martin 1996). It has generally been reported that RFLP analysis detects less variation than sequencing because it is an indirect method of variation detection (Carr and Marshall 1991 ; Bematchez and Danzman 1993; Brown et al 1993; Meyer 1994). The almost three-fold difference in divergence estimates between the two methods indicates that the regions chosen for sequencing were less variable than averaged over the entire molecule. This was also unexpected since the chosen mtDNA regions were reported to be highly variable at the intraspecific level in other fishes, narnely in brown trout, Salmo trutta (Giuffra et al. 1994; Bematchez and Osinov 1995) and lake whitefish, Coregonus

clupeaforrnis (Pigeon et al. 1997). Aitogether, these results indicated that mutational differences between the two groups were concentrated in regions that were not sequenced, such as the D-Loop. The D-Loop region was however not selected a priori since it has not been reported as particulariy polyrnorphic in fish (Meyer et al.

1990; Bematchez et al. 1992). These results also suggested that it may not be possible to generally classify mtDNA genes as highly variable or not, since the extent of polymorphism in specific genes is apparently species-dependent as stated above. A striking example of this is the study of the cytochrome b gene in two closely related osmerids; rainbow smelt (Osmerus mordax) and capelin (Maîlotus villosus) (Taylor and Dodson 1994; Dodson et al., submitted). These authors reported high polymorphism in capelin cytochrome b sequence but cornplete monomorphism among smelt known a priori to belong to the distinct mtDNA groups A and B. As mutation rate in mitochondria protein coding genes is rnainly affected by selective functional constraints on the gene product (Meyer 1994), these results corroborate previous observations that different mtDNA genes may be subject to differential selective constraints even among closely related species (Bematchez et al. 1992).

Another argument in favour of differential selective constraints among mtDNA genes for different species is in the nature of nucleotide substitutions. It has been generally observed that transitions greatly outnumber transversions by a 10:l-32:l ratio, in both protein-coding genes and noncoding regions at the intraspecific level, in birds and mammals (Moritz et al. 1987; Beckenbach et al. 1990), as well as in coding regions of fishes (Carr and Marshall 1991). The 7:2 ratio observed in the present study contrasts with these obsenrations and those of Giuffra et al. (1 994) on brown trout for which a 17:O ratio was observed in cytochrome b and ATPase subunit 6 protein-coding genes.

Analysis of smelt lawae

This study revealed four relevant results to the issue of the sympatric persistence of fish populations and to the biology of rainbow smelt in the estuary: the overall similarity of larval density distribution with previous studies, the lack of spatial segregation over time between the two larval populations, the increase of the south shore larvae relative to the north shore larvae with time and finally, the overall greater abundance of the north shore larvae population.

Sirnilarity of larval density with previous study

Our results indicated the upstream displacement of larvae during the season, particulariy notable in the north channel, as maximum densities moved from water masses between 5 and IOpsu in May to waters of less than Spsu in July. These results are identical to those obtained in detailed studies of smelt larval ecology in the St. Lawrence Estuary. The low density of lawae in the downstrearn section of the middle estuary comparative to the upstream section was previously described by Ouellet and Dodson (1985) and Laprise and Dodson (1989a, 1989b). Larval dispersal seemed also to follow the same pattern described previously (Ouellet and Dodson 1985, Laprise and Dodson 1989a). Briefly, early after hatching the larvae are transported into the estuarine nursery zone where they are retained by active vertical displacements in combination with the circulation system. The fact that older larvae use the tidal currents more efficiently by increasing the amplitude of their vertical migration explains the increase in density farther upstrearn in July (Laprise and Dodson 19896). These results indicate that the adaptive mechanisrns of vertical displacements exhibited by smelt larvae for their retention in a highly productive nursery area (Dauvin and Dodson 1990, Vincent et al 1996) are temporally stable.

Resource partitioning

Results revealed spatial homogeneity in the relative distribution of the larvae issued from the south and north shore populations. Consequently, they did not support our working hypothesis that the larvae belonging to the two sympatric srnelt populations in the St. Lawrence middle estuary are spatially segregated, and that such segregation may promote the persistence of two populations in the system. Two alternative explanations rnay account for these observations.

First, these results may suggest that resource partitioning is not important at the larval stage to promote persistence of sympatric populations still capable of interbreeding. As this work represented a first study of resource partitioning at the larval stage in sympatric fish populations. it cannot be discussed in the light of eariier findings, but rather in the light of theoretical considerations. It has generally been assumed that a major force driving resource partitioning is competition for trophic resources (Schluter 1993; Robinson and Wilson 1994; Skulason and Smith 1995).

Consequently, the persistence of the two larval populations without apparent segregation may indicate that there is not sufficient competition to drive resource partitioning in the system at the larval stage. The maximum turbidity zone of St.

Lawrence middle estuary where the maximum density of smelt larvae is found, represents a productive food web with high densities of both micro and macrozooplankton (Dodson et al. 1989; Dauvin and Dodson 1990; Frenette et al.

1995; Vincent et al. 1996) which constitutes a nursery for larval smelt. The densities of potential food for larvae, including copepods and Neomysis americana are respectively 10 and 100 times greater within the turbidity zone than at its downstream Iimit (Dodson et al. 1989; Dauvin and Dodson 1990). This abundance is presumably due to the common hydrodynamic control of suspended sediment and

some pefagic and Dodson

components of the estuarine ecosystem (Dodson el l989a, 19896; Dauvin and Dodson 1990). The

al. 1989; Laprise high content of alimentary resources in this zone is also illustrated by a growth rate of smelt larvae of 0.33 mm d-1 within the turbidity zone as compared to growai rates varying from 0.1 0 to 0.24 mm d-1 downstream of the turbidity zone (Laprise and Dodson 19896).

The major potential cornpetitive species for rainbow smelt resources in the middle estuary are the lawae of tomcod (Microgadus tomcod), capelin (Mallotus villosus) and hemng (Clupea harengus) (Courtois et al. 1982). The ecological and physical separation of capelin and herring from each other and from tomcod and smelt by various physical factors and species-specific reproductive strategies are well defined for the estuary (Courtois et al. 1982). However, tomcod and srnelt appear to occupy the sarne area (salinity &PSU) and feed on the same copepod resources as young larvae in June (Laprise and Dodson 1989b). Longitudinal segregation occurs later in summer (in July) by species-specific use of the vertical pattern of current, with smelt migrating upstream and tomcod downstream to exploit more manne food resources (Laprise and Dodson 1989b, 1990). In June, tomcod use the net upstream residual currents in deeper waters to remain in salinity less than Spsu, while smelt larvae concentrate in surface waters during flood tides and disperse almost homogeneously in the water column during ebbs to remain in the same zone (Laprise and Dodson 1989a, 1989b 1990).

The abundance of trophic resources and the few potential interspecific cornpetitors for these resources renders plausible the persistence of coexisting populations without spatial resource partitioning at the larval stage. In fact, in this system, not only do two populations of the same species coexist, but a different species is also present at the same time in the same environment. A spatial segregation between